JP5793299B2 - Process monitoring and diagnosis device - Google Patents

Description

The present invention relates to a process monitoring / diagnosis apparatus having a diagnostic algorithm capable of detecting a state change of a process system and a sign of abnormality such as a sewage treatment process, an industrial wastewater treatment process, a sludge digestion process, a water purification process, a water supply / distribution process, and a chemical process.

  In the operational management of water treatment / water operation processes such as sewage treatment processes, sludge digestion processes, water purification processes, and water supply / distribution processes, petrochemical processes, and semiconductor manufacturing processes, etc. Operation that leads to energy and cost savings is required.

  Here, as examples of the specified target performance, compliance with the effluent water quality standard in the sewage treatment process, securing a predetermined amount of generated energy (methane, hydrogen, etc.) in the sludge digestion process, and water supply and distribution water quality by disinfection and sterilization in the water purification process For example, compliance with standards, achievement of a predetermined target value for the yield of refined products such as petroleum in the petrochemical process, and achievement of a predetermined target value or more for the yield of semiconductor products in the semiconductor manufacturing process. Examples of energy-saving and cost-saving operations include reduction of blower and pump drive power and chemical injection in the sewage treatment process, maximization of generated energy efficiency in the sludge digestion process, and minimization of chemical injection in the water purification process. In the petrochemical process and semiconductor manufacturing process, the yield can be maximized.

  In order to achieve these, the process status related to the target performance is monitored so as not to fall into a state where the predetermined target cannot be achieved, and a state change or abnormal state that impedes the achievement of the predetermined target is quickly detected in advance. It is an important operational management point to take measures. In addition, in order to perform operations that lead to energy savings and cost savings after achieving the predetermined targets, the process state related to the target performance and energy savings and cost savings is always kept in a good state, and the process state changes that are likely to deviate from the good state Need to be detected quickly.

  Multivariate statistical process monitoring (MSPC) using the “multivariate statistical analysis method” developed mainly in the petrochemical process field as a method for diagnosing such process changes and abnormalities A method called “Control” is known (see, for example, Non-Patent Document 1, Patent Document 1, Patent Document 2, and Patent Document 3).

  MSPC is sometimes referred to as a chemometrics method, and a method based on Principal Component Analysis (PCA) is widely used as the most basic and frequently used method in MSPC. . In addition, monitoring methods using principal component regression (PCR), latent variable projection method / partial least square method (PLS), etc., as advanced methods based on PCA Is also used (see Non-Patent Document 1 described above).

These methods usually use the correlation information between many process data from a large number of measurement data to generate a small number of statistical data, and detect a change in the process state using the small number of statistical data generated. Based on the idea of trying. For example, in MSPC using PCA, a highly correlated data set (data subspace) is generated using PCA, and the data in this subspace (similar to the Mahalanobis distance used in the Taguchi method in the field of quality engineering) a statistic called) T ² statistic concepts, data at each time perform a process state monitoring by statistic called Q statistic indicating how much deviation from the subspace.

  When constructing such a condition monitoring / abnormality diagnosis system, first select all of the many measurement variables measured in the target process or some variables necessary for monitoring, and then select the selected measurement. A time series data (stored in a data server or the like) of variables is input offline to construct / identify a monitoring diagnosis model. Then, the time series data of the same measurement variable is input to the constructed monitoring diagnosis model online. Thereafter, for example, in a monitoring method based on PCA, a process state change or abnormality is detected (state change / abnormality detection) according to a predetermined procedure, and a measurement variable that causes the process is estimated (factor separation). After that, usually, the operator who is presented with the detection and factor separation results identifies the true factor of the state change / abnormality and takes a procedure of countermeasures against the situation.

  Such advanced condition monitoring / abnormality diagnosis techniques are pioneered in the petrochemical process field, but are rarely used in the field of water and sewage processes. Process managers and operators who manage the operation and operation of normal processes monitor the time series data of measurement variables used in the above diagnostic system with a trend graph, etc., and monitor process status changes and abnormalities on the trend graph. Yes. In addition to this, there are many cases where several management indexes and performance indexes calculated from measured variables and plant structure data are monitored. For example, in the sewage treatment process field, there is a management index called SRT (sludge residence time) calculated from the sludge concentration and flow rate, which are measurement variables, and the volume of the structure. We are operating. As another example, in the field of water supply processes, the water supply business guidelines established by the Japan Waterworks Association (JWWA) propose to evaluate the performance of water purification and water supply / distribution processes using performance indicators called performance indicators (PI). .

JP-A-8-241121 JP 2004-303007 A JP 2007-65883 A

URL: http://tech.chase-dream.com/spc.html

  Such management indices and performance indices are indices that well express certain characteristics of the plant related to the performance and stable operation of the plant, and are useful information for operators and plant managers. However, the conventional MSPC state monitoring system mainly used in the petrochemical process field is not consistent with plant monitoring based on these management indexes. In addition, because the management index expresses certain characteristics of the plant well, the change in the management index is very deeply related to the change in the state of the plant, but the conventional MSPC considers such a management index. Not.

  In addition, the process state change is not measured directly from the process measurement variable data, but the measured data is subjected to a non-linear operation, the differential value or integral value of the measurement data, or partial information of the measurement data. It is often possible to monitor the process state more appropriately by monitoring periodic information. For example, in the sewage treatment process field, there is a so-called respiration rate meter for monitoring the activity state of microorganisms, but there are few treatment plants where respiration rate meters are installed, and the dissolved oxygen (DO) concentration is an alternative indicator. By monitoring the differential value (rate of change), the active state of the microorganism may be grasped. As another example, water leakage is a problem in the water distribution process, but in order to obtain an approximate value of the amount of water leakage, the amount of water leakage can be calculated using the water distribution data only during the night hours when there is little demand for water purification. You may get an estimate. As another example, there is a daily fluctuation caused by a person's life pattern in clean water and sewage, and it is better to monitor the process for each time zone to catch a change in the state of the process.

  However, in the conventional MSPC, a monitoring system is constructed using all items of measurement variables or all data of selected items.

  It is an object of the present invention to improve the state monitoring performance by MSPC, that is, to detect a sign of a state change or an abnormal state. An object of the present invention is to provide a process monitoring diagnostic apparatus capable of diagnosis.

The process monitoring diagnostic apparatus of the present invention collects time-series data of a plurality of measurement variables including state quantities and manipulated variables of the target process measured at a predetermined cycle by a plurality of process sensors provided in the target process, and holding to keep data collection and storage unit, and using said past time-series data of the stored plurality of measurement variables into the data collection and storage unit, to construct a process monitoring model supplies process model building and supply unit , Process monitoring / diagnosis by detecting the status change and abnormal signs by monitoring the process status using the online data extracted from the data collection / storage unit and the process monitoring model constructed by the process model construction / supply unit and a section, wherein the process model building and supply unit, past time-series data of a plurality of measurement variables stored in the data collection and storage unit A selection variable determination unit that selects all or some of the variables necessary for constructing the process monitoring model, and a plurality of measurement variables stored in the data collection / storage unit, A variable conversion formula determination unit in which a predetermined conversion formula for obtaining a management index useful for operation and a useful index for early detection of process state changes and abnormal signs is set, and selected by the selection variable determination unit The selected variable obtained by removing abnormal data such as outliers from the time series data of the past conversion variable converted by using the selection variable and the expression of the variable conversion formula determination unit, and the normal time series data of the conversion variable against (xi (t) -ai) / the normalized parameter determination unit that determine the parameters ai and bi for normalizing data by bi, the normalization parameter determination One of multivariate analysis means represented by principal component analysis (PCA), principal component regression (PCR), and partial least squares (PLS) for data normalized using the normalization parameters determined by A diagnostic model construction unit that defines an expression for generating at least one or more diagnostic statistics data,
A statistic threshold value setting unit for detecting a change in state with respect to the one or more diagnostic statistical data generated by the diagnostic model construction unit, and the process monitoring / diagnostic unit is the selection variable determination unit From the current data of the data collection / storage unit using the variable selection formula determined by the variable conversion formula determination unit and the variable selection unit that sequentially retrieves the current data corresponding to the selected variable determined in step 1 from the data collection / storage unit A variable conversion unit that performs variable conversion to obtain a current index, and a data normalization that normalizes the variables selected using the normalization parameters determined by the normalization parameter determination unit and the online data of the converted variables And statistical data is generated from the online data normalized by the data normalization unit based on the statistic generation formula defined by the diagnostic model construction unit. Statistic monitoring unit that sets the status to be monitorable, and when the online statistic data generated by the statistic monitoring unit exceeds the threshold determined by the statistic threshold setting unit, it is detected as a process state change or abnormality And a state change detection unit that performs the above operation.

  Xi: i-th selected variable / transform variable, ai: constant representing shift for i-th select variable / transform variable (shift parameter), bi: constant representing scaling for i-th select variable / transform variable (scaling) Parameter).

  In the present invention, the process model construction / supply unit further includes a state change factor contribution amount setting unit that estimates a factor when a state change occurs from the selection variable and the conversion variable, and the process monitoring -When the state change detection unit detects a process state change or abnormality, the diagnosis unit estimates a variable that is a factor by a contribution amount calculation set by the state change factor contribution amount setting unit The structure which further has an item (variable) estimation part may be sufficient.

  In the present invention, the variable conversion equation determination unit includes nonlinear conversion (including: product (multiplication) and quotient (division)), differentiation / difference conversion, integration / integration conversion, decimation conversion of a predetermined period, and interpolation of a predetermined period. At least one conversion formula is included from the poration conversion and the management index / performance index conversion.

  Further, in the present invention, the process measurement variables collected and stored in the data collection / storage unit at a pre-stage of the processing in the selection variable determination unit and the variable selection unit are predetermined in a predetermined time unit T. The process over the period R is shifted so that a process measurement variable is newly generated, and a process for configuring an expanded process measurement variable that is R / T times the number of the original process measurement variables is configured. Also good.

  In the present invention, the process measurement variables collected and stored in the data collection / storing unit in the previous stage of processing in the selected variable determination unit and the variable selection unit are decomposed / re-processed by discrete wavelet transform. By applying a digital filter configured by a configuration algorithm, the original process data may be divided into N pieces, and an expanded process measurement variable N times the number of original process measurement variables may be configured.

Further, the present invention is a process monitoring diagnostic apparatus, the M for each processing unit such as a processing sequence for each and water distribution block each (M: the number of processing units) was constructed, further, the M-number of each monitoring diagnostic apparatus It may be configured in a hierarchical form having an overall process monitoring / diagnostic device by MSPC which receives each statistic calculated from

  In the present invention, the target process is a biological wastewater treatment process such as a sewage treatment process / industrial wastewater process, and sludge residence time (SRT), aerobic tank sludge is used as a conversion formula by the variable conversion formula determination unit. Residence time (A-SRT), hydraulic residence time (HRT), Log (SRT) / water temperature, Log (A-SRT) / water temperature, excess sludge generation, organic matter (COD and / or BOD) load, nitrogen Load, phosphorus load, organic matter (BOD and / or COD) -SS load, water area load, phosphorus load / nitrogen load, organic matter load / nitrogen load, organic matter load / phosphorus load, pH / ORP, DO change rate ( Differential value), air volume change rate (differential value), ammonia concentration change rate, nitric acid concentration change rate, phosphorus concentration change rate, pH change rate, ORP change rate, sludge interface change rate, water temperature change rate, predetermined Rainfall integrated value (integrated value) for the period, measurement data for weekdays / holidays Or having one or more conversion formula.

  Further, in the present invention, the target process is a sludge treatment process, and as a conversion formula by the variable conversion formula determination unit, a concentration tank surplus sludge mixing rate, a concentrate tank solid matter recovery rate, a concentrate tank HRT, a concentrate tank sludge solid matter retention Time, Concentration tank sludge interface change rate, Centrifugal concentrator centrifugal effect, Centrifugal concentrator screw conveyor and bowl rotation speed differential speed, Pressure / normal pressure concentrator gas-solid ratio, Pressure / normal pressure concentrator floss thickness change rate, Pressure / normal pressure concentrator flotation sludge scraping frequency, digester digestibility, digester digested sludge volume, digester digestion days, digester digestion days / digestion temperature, gas generation rate, methane gas composition ratio, CO2 composition ratio, sulfurization Hydrogen composition ratio, digester solid load, digester organic load, digester organic load / nitrogen load, pH / ORP, pH change rate, ORP change rate, digester effluent SS change rate, temperature change rate, gas generation Volume change rate, dehydrator Ramp rate, it may be configured with any one or more of the conversion formula.

  Furthermore, in the present invention, the target process is a water purification / supply / distribution process, and as the conversion formula by the variable conversion formula determination unit, the sludge amount / coagulant injection amount, the chlorine requirement amount / hypochlorous acid injection amount are defined in advance. The amount of water distribution during the night time, the amount of water distribution or water supply per predefined time, or the performance indicator (PI) measured from online measurement data, as seen from the raw water effective utilization rate (%), and the mold odor Delicious water achievement rate (%), Delicious water achievement rate (%) viewed from chlorine odor, Total trihalomethane concentration water quality standard ratio (%), Organic matter (TOC) concentration water quality standard ratio (%), Activated carbon input rate (%), Chemical stockpiling days (days), fuel stockpiling days (days), supply unit price (yen / cubic meter), water supply cost (yen / cubic meter), yield (%), power consumption per cubic meter of water distribution (kWh / cubic meter) Water distribution Energy consumption per cubic meter (MJ / cubic meter), renewable energy utilization rate (%), effective utilization rate of purified water generation soil (%), carbon dioxide (CO2) emissions per cubic meter of water distribution (g · CO2 / cubic meter) It may be configured to have one or more conversion formulas of: groundwater rate (%), pump average operation rate (%), water leakage rate (%), and water leakage amount per water supply (cubic meter / year / case).

  According to the present invention, it is possible to improve the state monitoring performance by the MSPC, that is, to detect a sign of a state change or an abnormal state by linking the information useful for the plant operation manager and the MSPC. In addition, since the index focused on by the operator for monitoring and the MSPC are linked, it is possible to perform state monitoring and abnormality diagnosis that are easier for the operator to understand.

It is a functional block diagram which shows one Embodiment of the process monitoring diagnostic apparatus which concerns on this invention. It is a system block diagram which shows the case where one embodiment same as the above is applied to a sewage treatment process. It is a system block diagram explaining the sludge process to which one embodiment same as the above is applied. It is a system block diagram explaining the purified water and water supply / distribution process to which one embodiment same as the above is applied. It is a figure explaining embodiment which applied the process which considers process delay in this invention. It is a block diagram explaining embodiment which uses discrete wavelet transform together in this invention. It is a block diagram explaining embodiment with the hierarchical structure in this invention. The present invention is a diagram showing a table 1 showing the structure of a variable transformation formula determining section for obtaining an indication of when applied to biological waste water treatment processes. It is a figure showing the same Table 1. FIG. It is a figure showing Table 2 which shows the structure of the variable conversion formula determination part for obtaining the parameter | index at the time of applying this invention to a sludge concentration and digestion process. It is a figure showing the same table 2. FIG. It is a figure showing the same table 2. FIG. It is a figure showing Table 3 which shows the structure of the variable conversion formula determination part for obtaining the parameter | index at the time of applying this invention to a water supply, water purification, and water supply / distribution process. It is a figure showing the same Table 3. FIG.

  Hereinafter, an embodiment of a process monitoring diagnosis apparatus according to the present invention will be described in detail with reference to the drawings.

  1 and 2 show the basic configuration of this embodiment. As an object process, nitrogen and phosphorus removal, which is an example of a biological wastewater treatment process such as a sewage treatment process / industrial wastewater process, is shown. This shows the monitoring system applied to the target advanced sewage treatment process. First, in FIG. 2, a sewage advanced treatment process which is a process to be monitored will be described.

The advanced sewage treatment process 1 is configured by sequentially connecting a first sedimentation tank 101, an anaerobic tank 102, an oxygen-free tank 103, an aerobic tank 104, and a final sedimentation tank 105 in series. The ponds and tanks of the sewage altitude treatment process 1 are provided with pumps and sensors described below as actuators and their operation amount sensors. That is, the first sedimentation basin 101 has an excess sludge drain pump and its extraction flow sensor 111, and the aerobic tank 104 has a blower for supplying oxygen and its supply air flow sensor 112, and the aerobic tank 104 and its upstream stage. A circulation pump and its circulation flow sensor 113 are connected to the circulation path to the anaerobic tank 103, and a return sludge pump and its return flow sensor 114 are further connected to the return path from the final sedimentation tank 105 to the anaerobic tank 102. The pond 105 is provided with an excess sludge extraction pump and an extraction flow rate sensor 115 thereof.

The sewage advanced treatment process 1 is provided with the following process sensors. That is, for the inflow pipeline to the first sedimentation basin 101, the rainfall sensor 121 that measures the rainfall in the surrounding area, the sewage inflow sensor 122 that measures the inflow sewage, and the total nitrogen amount contained in the inflow sewage. An inflow TN sensor 123 for measuring, an inflow TP sensor 124 for measuring the total amount of phosphorus contained in the inflowing sewage, and an inflow UV sensor or an inflow COD sensor 125 for measuring the amount of organic matter contained in the inflowing sewage are provided.

  The anaerobic tank 102 is provided with an anaerobic tank ORP sensor 126 that measures the ORP (oxygen reduction potential) and an anaerobic tank pH sensor 127 that measures pH, and the anaerobic tank 103 measures the ORP. An anaerobic tank ORP sensor 128 and an anoxic tank pH sensor 129 for measuring pH are provided, and the aerobic tank 104 includes a phosphoric acid sensor 1210 for measuring the phosphoric acid concentration and a DO sensor 1211 for measuring the dissolved oxygen concentration. , And an ammonia sensor 1212 for measuring the ammonia concentration.

  In addition, for each of the reaction tanks 102 to 104, an MLSS sensor 1213 that measures the amount of activated sludge in at least one of these tanks (anaerobic tank 102 in the example in the figure) is also provided with each of the reaction tanks 102 to 104. On the other hand, water temperature sensors 1214 for measuring the water temperature in at least one of these tanks (the oxygen-free tank 103 in the example in the figure) are provided.

  In addition, the final sedimentation basin 105 includes an excess sludge SS sensor 1215 that measures the solid concentration of the amount of sludge drawn from here, a discharge SS sensor 1216 that measures the SS concentration of discharged water discharged from here, and a final sedimentation A sludge interface sensor 1217 for measuring the sludge interface level of the pond 105 is provided.

  Furthermore, the discharge pipe from the final sedimentation basin 105 includes a sewage discharge sensor 1218 that measures the amount of discharged sewage, a discharge TN sensor 1219 that measures the total amount of nitrogen contained in the discharged sewage, and a total phosphorus contained in the discharged sewage. A discharge TP sensor 1220 for measuring the amount and a discharge UV sensor or a discharge COD sensor 1221 for measuring the amount of organic matter contained in the discharged sewage are provided.

  The various actuators 111 to 115 described above operate at a predetermined cycle, and the operation amount sensors 111 to 115 and the various process sensors 121 to 1221 represented by the same reference numerals measure at a predetermined cycle.

  1 and 2 includes a process measurement data collection / storage unit 2, a past data (offline data) extraction unit 3, a process monitoring model construction / supply unit 4, and a current data (online data) extraction unit 5. And a process monitoring / diagnostic unit 6 and a user interface unit 7.

  The process measurement data collection / storage unit 2 collects and holds process data obtained from various actuators / operation amount sensors 111 to 115 and various process sensors 121 to 1221 of the sewage advanced treatment process 1 at a predetermined cycle. The past data extraction unit 3 extracts past data (offline data) from various time series data stored in the process measurement data collection / storage unit 2. The process monitoring model construction / supply unit 4 constructs a process monitoring / diagnostic model offline in advance using the offline data extracted by the past data extraction unit 3. The current data extraction unit 5 extracts current data (online data) from various time series data stored in the process measurement data collection / storage unit 2. The process monitoring / diagnostic unit 6 uses the online data extracted by the current data extraction unit 5 and the process monitoring model constructed by the process monitoring model construction / supply unit 4 to monitor the process status, Detect abnormal signs. The user interface unit 7 notifies the plant manager and the operator of information related to the state change or abnormality sign detected by the process monitoring / diagnostic unit 6 and the factor variable candidates.

    As shown in FIG. 1, the process monitoring model construction / supply unit 4 includes a selection variable determination unit 41, a variable conversion formula determination unit 42, a normal data extraction unit 43, a normalization parameter determination unit 44, and a diagnostic model construction. It is preferable that a unit 45 and a statistic threshold value setting unit 46 are provided, and further a state change factor contribution amount formula setting unit 47 is provided.

The selection variable determination unit 41 is a variable necessary for constructing a process monitoring model from information on past time series data of measurement variables extracted from the process measurement data collection / storage unit 2 through the past data (offline data) extraction unit 3. Determine and select. The variable conversion formula determination unit 42 performs appropriate variable conversion on the measurement variable and provides a new information that is easy to understand for the operation management of the operator and the quick detection of the state change or abnormality sign of the process monitoring model. A new conversion variable (index). The normal data extraction unit 43 removes missing values and obvious abnormal values from the measurement variable selected by the selection variable determination unit 41 and the index generated by the variable conversion formula determination unit 42, and is normal. Extract data only. The normalization parameter determination unit 44 is an expression that normalizes various selection / conversion variables by (xi (t) −ai) / bi with respect to the selection variables and conversion variables in the normal state extracted by the normal data extraction unit 43. Shift parameter a i and scaling parameter bi are determined.

  Xi (t): i-th selected / transformed variable, ai: constant representing the shift for the i-th selected / transformed variable (shift parameter), bi: constant representing the scaling for the i-th selected / transformed variable (scaling) Parameter).

The diagnostic model construction unit 45 performs principal component analysis (PCA), principal component regression (PCR) or partial least squares (PCA) on the normalized data defined by the normalization parameter determination unit 44. In order to calculate the loading matrix (load matrix) and the score matrix by applying multivariate analysis means such as PLS (Partial Least Squares), and to calculate the Q statistic and Hotelling's T ² statistic defined using them Set the calculation formula (model). The statistic threshold value setting unit 46 uses the diagnostic model constructed by the diagnostic model construction unit 45 to set a threshold value for determining abnormality / normality for statistical data calculated using past offline data. Set. When the Q statistic or the Hotelling T2 statistic exceeds the threshold of the statistic threshold setting unit 46, the state change factor contribution formula setting unit 47 calculates the contribution of each selected / transformed variable to that statistic. Determine the formula to do.

  The process monitoring / diagnostic unit 6 also includes a variable selection unit 61, a variable conversion unit 62, an outlier removal unit 63, a data normalization unit 64, a statistic monitoring unit 65, as shown in FIG. It is preferable that a state change detection unit 66 and a factor item (variable) estimation unit 67 are further provided.

The variable selection unit 61 extracts the selected variable determined by the selection variable determination unit 41 from the current time series data of the measurement variable extracted from the process measurement data collection / storage unit 2 through the current data (online data) extraction unit 5. . The variable conversion unit 62 performs variable conversion on the current time series data using the variable conversion formula determined by the variable conversion formula determination unit 42, and calculates an index. The outlier removal unit 63 removes missing values and outliers from the current data of the measurement variable selected by the variable selection unit 61 and the current index (conversion variable) generated by the variable conversion unit 62. . The data normalization unit 64 uses the shift parameter ai and the scaling parameter bi determined by the normalization parameter determination unit 44 for the selected variable and conversion variable in the current normal state extracted by the outlier removal unit 63. Perform normalization. The statistic monitoring unit 65 applies these statistics to the current data normalized by the data normalization unit 64 according to the Q statistic determined by the diagnostic model construction unit 45 and the calculation formula of Hotelling's T ² statistic. Calculate When the statistic monitored by the statistic monitoring unit 65 exceeds the threshold defined by the statistic threshold setting unit 46, the state change detection unit 66 detects a process state change or an abnormal sign. The factor item (variable) estimation unit 67, when the state change detection unit 66 detects a change in the Q statistic or the Hotelling T2 statistic, determines the contribution amount of the selection / conversion variable that becomes the change factor. Calculation is performed according to the formula set by the factor contribution formula setting unit 47, and the selection / conversion variable that is the factor is estimated.

  Here, as described above, the conventional status monitoring system using MSPC does not use management indices and performance indices, which are useful information for operators and plant managers, but includes all items of measurement variables or selected items. A monitoring system was constructed using all data.

  Therefore, in the present invention, information useful for the plant operation manager and the MSPC are linked to improve the state monitoring performance by the MSPC, that is, it is possible to detect a sign of a state change or an abnormal state. Further, by linking the index that the operator is paying attention to to the MSPC for monitoring, it is possible to perform state monitoring and abnormality diagnosis that are easier for the operator to understand.

  In order to realize these, the characteristic part of the present invention is that, in the embodiment shown in FIGS. 1 and 2, the process monitoring model construction / supply unit 4 is provided with a variable conversion formula determination unit 42, and the process monitoring The variable conversion unit 62 is provided in the diagnosis unit 6 to obtain the various indexes described above and apply them to the MSPC.

  Next, the operation of the above-described embodiment will be described.

First, in the sewage altitude treatment process 1, process information is measured at predetermined intervals by the operation amount sensors 111 to 115 and the various process sensors 121 to 1221. These pieces of measurement information are stored as time series data by the process measurement data collection / storage unit 2 according to a predetermined format.

In the present invention, when constructing a process monitoring and diagnosis device, first, past process data stored in the process measurement data collection / storage unit 2 over a predetermined period is extracted by the past data extraction unit 3. . The process monitoring model construction / supply unit 4 constructs a process monitoring model by using past process data for a predetermined period extracted by the past data extraction unit 3.

  In the process monitoring model construction / supply unit 4, the selection variable determination unit 41 determines a selection method of measurement variables necessary for constructing the process monitoring model. In the normal sewage treatment process, not only the items measured by the operation amount sensors 111 to 115 and the various process sensors 121 to 1221, but also the feedback control target values, measurement variables related to equipment such as blowers and pumps, or with time There are usually thousands of measurement variables, such as the accumulated amount. In the multivariate statistical process monitoring method, a process monitoring model can be constructed in principle even if all these measurement variables are input, but the selection method is determined so as to select only the necessary variables.

  For example, the target value of feedback control usually has little information because it does not change at a constant value over a long period of time. It is preferable not to select a variable that does not have such information because it may deteriorate the diagnostic performance. Further, since the integrated amount is a variable that increases monotonously, the process monitoring model cannot be correctly constructed if it is selected as input data for constructing the process monitoring model. Also, in the case where multiple pumps and blowers are prepared, if the blower or pump flow data that rarely starts up is input as it is, it is 0 in most time zones, so the monitoring model must be constructed correctly. I can't. Further, in order to detect a change in the state of the processing process or an abnormality sign, almost no data on the device side such as the current value of the device and the piping pressure is necessary.

  Therefore, the selection variable determination unit 41 does not select the above-described variables, but selects only the measurement variables necessary for the purpose. For example, in the case of the process of FIG. 2, the measurement variables measured by the operation amount sensors 111 to 115 and the various process sensors 121 to 1221 are selected.

As described above, the variable conversion formula determination unit 42 is a characteristic part of the present invention, and is a management index useful for the operator for a plurality of measurement variables stored in the process measurement data collection / storage unit 2. And conversion formulas of indicators useful for early detection of process state changes and abnormal signs.

  Examples of conversion formulas are listed below. The variable conversion formulas shown below are summarized in Table 1 shown in FIGS. 8A and 8B.

  As an example of the operation management index, the sludge residence time (SRT), the aerobic tank sludge residence time (A-SRT), or the hydraulic residence time (HRT) in Table 1 will be described. These can be calculated using some of the variables measured by the manipulated variable sensors 111-115 and the various process sensors 121-1221. These indicators are always managed in the operation of the sewage treatment process, and process managers and operators operate with reference to these indicators (Sewerage Maintenance Guidelines Part 2003, Japan Sewage Association, reference). Therefore, when this index is used as input information for process monitoring by MSPC, if any change occurs in this index, it will be detected as a change in the statistic of MSPC described later. In addition, as shown in Table 1, Log (SRT) / water temperature and Log (A-SRT) / water temperature are usually managed so as to have a linear relationship (same sewerage maintenance guidelines, second part 2003 version, Japanese sewerage system) Therefore, these indicators can also be used for the same purpose.

  In addition, the amount of excess sludge generated is an index managed by the operator or process manager. As shown in Table 1, sludge conversion coefficients a, b, and c determined by investigating and determining the inflow solubility BOD, the inflow SS, and the MLSS concentration in the reaction tank, respectively, as shown in Table 1. Can be calculated. This index can also be used as an important operation management index.

  In addition, organic substances, nitrogen, phosphorus, etc. are usually measured by various water concentration sensors related to them, but from the viewpoint of processing, it is not a concentration, but a method that manages the load by multiplying the concentration by the treatment amount. Is often appropriate. Therefore, it is preferable to generate indexes such as inflow organic matter load amount, inflow nitrogen load amount, inflow phosphorus load amount, or discharge organic matter load amount, discharge nitrogen load amount, and discharge phosphorus load amount shown in Table 1.

  In addition, since the operation management is often performed not only by these load amounts themselves but also by the ratio of the load amounts, in this case, for example, the organic matter-SS load, which is an indicator of the treatment characteristics of the organic matter shown in Table 1, Water area load related to sedimentation characteristics of sedimentation basin is also generated as an index. Further, in an advanced treatment process aimed at removing nitrogen and phosphorus, the treatment may be performed efficiently when the ratio of organic matter: nitrogen: phosphorus is maintained at a predetermined ratio that roughly corresponds to the composition ratio of microorganisms. Are known. For this reason, if it produces | generates as an index of these load ratios, ie, phosphorus load / nitrogen load, organic substance load / nitrogen load, organic substance load / phosphorus load, it can be used as a standard of the removal performance of phosphorus or nitrogen.

  Further, the anaerobic tank 102 and the anoxic tank 103 are usually provided with an ORP meter 128 and a pH meter 127 in many cases. These ORP meter 128 and pH meter 127 show a strong correlation when there is no significant change in the total ion concentration. Therefore, if the ORP / pH ratio is managed, changes in the ion concentration other than pH are detected. It becomes easy. For example, the concentration of dissolved oxygen in the anaerobic tank 102 and the anoxic tank 103 may increase during rainy weather and the processing may deteriorate. In such a case, since the change in ORP becomes larger than the change in pH, monitoring this ratio can be used for detection of dissolved oxygen contamination.

  Furthermore, among the measurement variables measured by the operation amount sensors 111 to 115 and the various process sensors 121 to 1221, not only the values of the measurement variables but also the information on the rate of change (difference, differentiation) is useful for process operation management. It may be important. For example, the rate of change of dissolved oxygen (DO) concentration can be considered as a substitute for a microbial respiratory rate meter. Microbial respiration rate monitors the activity state of microorganisms by utilizing the property that dissolved oxygen is reduced when the activity of microorganisms treating sewage is weakened. If the change rate of DO is monitored, the activity state of microorganisms can be monitored. On the other hand, since the aeration air volume is often controlled so as to keep the DO concentration at a constant value in the treatment plant, in such a case, the DO concentration does not change because it is controlled. If the rate of change is monitored, the activity state of the microorganism can be known indirectly. Using the same principle, it is possible to monitor the activity state of specific microorganisms involved in the removal of nitrogen and phosphorus. In other words, in order to see the activity state of nitrifying bacteria that change ammonia into nitric acid, it is only necessary to monitor the rate of change in ammonia concentration. To see the active state of denitrifying bacteria that reduce nitric acid to nitrogen gas, the rate of change in nitric acid concentration Can be monitored. In order to monitor the active state of the phosphorus accumulating microorganism that removes phosphorus, the change rate of the phosphorus concentration in the anaerobic tank and the aerobic tank may be monitored. Furthermore, since the pH and ORP may change abruptly when poisoning is mixed or when processing is inhibited, information on the rate of change of pH and ORP is also useful for process monitoring. In addition, the water temperature may suddenly decrease during rainy weather or when the snowmelt water flows in. However, since the decrease in the water temperature is an impediment to processing, the water temperature change rate is also an important monitoring item.

  On the other hand, in contrast to the information on the change rate, the integrated amount in a predetermined period often affects the process. For example, the integrated amount of rainfall during a predetermined period is highly likely to affect the performance of phosphorus removal. This is due to the fact that dissolved oxygen is brought into an anaerobic tank and an anaerobic tank due to rain, and that acetic acid-based organic matter necessary for phosphorus removal flows out due to rain. Such effects of rain may appear directly during rainy weather, but may appear for a while after raining. In this case, since the accumulated amount of rain and information on how long the clear sky has continued before is often important, the accumulated rainfall over a predetermined period can be an index.

  In addition, since a normal sewage treatment process has a load pattern corresponding to people's life patterns, for example, there may be a case where a load pattern that differs between a holiday such as Sunday and a weekday is displayed. In such a case, if the holiday / weekday data is taken out separately and used as MSPC input, the diagnostic performance may be further improved. For example, extracting Sunday data corresponds to an operation of thinning out (decimating) data on 1st / 7th, and extracting weekday data corresponds to decimation on 6th / 7th. On the other hand, in order to supply data continuously, operations such as zero input for weekday data and zero input for weekday data for Sunday data can be considered. Corresponds to the (poration) operation. Therefore, it is expected that the diagnostic performance of MSPC can be improved by inputting an index obtained by appropriately converting data using decimation and interpolation.

In this way, the variable conversion equations shown in Table 1 for performing various variable conversions, that is, nonlinear conversion (including: product (multiplication) and quotient (division)), differentiation / difference conversion, integration / integration conversion, predetermined cycle The function of the variable conversion formula determination unit 42 determines a variable conversion formula for executing at least one conversion from among decimation conversion, interpolation conversion at a predetermined cycle, and management index / performance index conversion.

  Next, the normal data extraction unit 43 removes missing data and outliers from the offline data extracted by the past data (offline data) extraction unit 3 in order to construct a process monitoring model by MSPC. Extract only useful data for construction. As the processing method of the normal data extraction unit 43, a plurality of methods can be considered. However, points need to be removed from outliers and missing values that are not considered to be actual process values, but data from which the process state deviates from the normal state should not be removed much. That is the point.

  Specifically, for example, when the measurement in a predetermined cycle is a relatively fast cycle such as one minute, a simple operation of extracting and using the median data for every hour can be considered. Usually, the residence time of the sewage treatment process is several hours to several tens of hours. Therefore, many outliers and missing values can be removed even with such a simple operation of performing median treatment in units of one hour. Other methods include, for example, using a method of robust statistics, adopting median as a median value index of data, and using median absolute deviation (MAD) as a data dispersion index, from median to MAD. A method of removing data that is more than a predetermined distance is also conceivable.

  Next, the normalization parameter determination unit 44 determines the values of the shift parameter ai and the scaling parameter bi necessary for normalization.

For example, a robust sample average and a robust sample standard deviation are used. Here, “robust sample average” and “robust sample standard deviation” are to obtain the sample average and sample standard deviation after removing about several percent of data near the maximum and minimum values of the process data in advance. . If this procedure is followed, the shift parameter and the scaling parameter can be determined as follows, after removing some data near the upper and lower limit values in advance.

ai = 1 / N * Σ _{k = 1} ^N xi (k)
bi = Σ _{k = 1} ^N (xi (k) −ai) ² / (N−1)
... (1)
Here, N is the number of cut out data.

  Alternatively, the aforementioned shift parameter may be the median described above, and the scaling parameter may be the median absolute deviation (MAD).

The diagnostic model construction unit 45 defines a calculation formula for statistics necessary for process monitoring. For example, when PCA is used as the multivariate analysis means, data decomposition is first performed as follows.

Formulas for calculating the Q statistic and the Hotelling T ² statistic using the loading matrix P for the data thus decomposed are defined as follows.

Q statistics:
Q (x (t)) = x ^T (t) (I-PP ^T ) x (t) (3)
Hotelling's T ² statistic:

The statistic threshold value setting unit 46 sets the threshold values of the expressions (3) and (4). This threshold setting value is important for detection of state changes and abnormal signs, so its setting method is important, but since it is not directly related to the present invention, its details are not mentioned, and only a typical setting method is set. Indicates. If there is no prior information for past offline data, the statistical confidence limit value for the Q statistic and the statistical confidence limit value for the Hotel ² T2 statistic may be used as the default setting method. Yes (C. Rosen \ Monitoring Wastewater Treatment Systems ", Lic. Thesis, Dept. of Industrial Electrical Engineering and Automation, Lund University, Lund, Sweden (1998))
These can be written as follows:

  In this way, the threshold value of the statistic can be set based on the equations (5) and (6).

Other methods, for example, outside the scope of the robust statistics using aforementioned median ME and the median absolute deviation (MAD) Q statistic and T ² ME ± relative statistic k * MAD (k parameters) upon that the data removal, the threshold maximum value of T ² statistic Q statistic and Hotelling removed or higher alpha% from the maximum value: determining the value of (alpha parameter) as a threshold value. A means for setting the threshold value by such a method is a function of the statistic threshold value setting unit 46.

The state change factor contribution amount setting unit 47 sets the measurement variables and conversion variables determined by the selection variable determination unit 41 and the variable conversion equation determination unit 42 for the statistic defined by the equations (3) and (4). Set the contribution formula. There are a plurality of methods for defining the contribution amount. For example, the contribution amount can be defined as follows.

  Here, n means the nth variable, and t is a variable representing a certain time. Expressions (7) and (8) are the actions of the state change factor contribution amount setting unit 47.

  After the process monitoring model is constructed by the process monitoring model construction / supply unit 4 according to the above procedure, the process monitoring / diagnosis unit 6 supplies the process monitoring model constructed by the process monitoring model construction / supply unit 4, Process monitoring is performed using this process monitoring model.

In the process monitoring / diagnostic unit 6, first, online data at the time point at which diagnosis is desired (hereinafter referred to as “current” or “current”) is obtained from the data collected by the process measurement data collecting / saving unit 2 as current data (online data). Extraction is performed by the extraction unit 5. The process monitoring / diagnostic unit 6 uses the current data extracted by the current data extraction unit 5 to monitor the process state, and detects any change in the state or indication of an abnormality. . The operation of the process monitoring / diagnostic unit 6 will be described in detail below.

  The variable selection unit 61 takes out the current data corresponding to the variable determined to be selected by the selection variable determination unit 41.

  Similarly, the variable conversion unit 62 calculates an index such as a current operation management index from the current data by the variable conversion formula determined by the variable conversion formula determination unit 42.

The outlier removal unit 63 performs processing when the current measurement data and index data selected / calculated by the variable selection unit 61 and the variable conversion unit 62 are outliers. As this process, for example, a simple process such as holding the zero order only when the data at the corresponding time is missing data may be used. Alternatively, a simple median process of about 3 to 7 steps may be performed before the calculation of the variable selection unit 61 and the variable conversion unit 62. Since the process monitoring / diagnostic unit 6 is in a phase of actually monitoring and diagnosing, the outlier processing is not necessarily essential and may be very simple. If a order to be diagnosed as abnormal after the diagnosis even if not the outliers processing.

  Next, in the data normalization unit 64, the variable selection unit 61 and the variable conversion unit 62 select / calculate using the normalization parameter shown in the equation (1) determined by the normalization parameter determination unit 44. Normalize the current measurement data and index data.

Next, in the statistic monitoring unit 65, the measurement data and index data normalized by the data normalization unit 64 are defined by the statistic defined by the diagnostic model construction unit 45, for example, the equations (3) and (4). The current Q statistic and T ² statistic are monitored by substituting the calculated Q statistic and T ² statistic into X (t). Since this statistic changes from time to time, it may be monitored in the form of a time series graph (trend graph).

Next, in the state change detection unit 66, the current Q statistic or T ² statistic exceeds the threshold set by the statistic threshold setting unit 46, for example, the threshold defined by the equations (5) and (6). In this case, it is determined that a change in state has occurred in the process, and the fact is notified to the operator or process manager through the user interface unit 7 shown in FIG. In this case, for example, a rule such as notifying the operator when the number of times exceeding the threshold value continues r times may be inserted to avoid frequent alarms.

  When a state change of the process is detected by the state change detection unit 66, it is preferable that the factor variable (variable) estimation unit 67 estimates the measurement variable or index that is the factor. At this time, the contribution amounts of the measurement variable and the index are calculated based on the contribution amount expressions set by the state change factor contribution amount expression setting unit 46, for example, the expressions (7) and (8). Then, for example, by determining in advance the rules listed below as (a), (b), and (c), a measurement variable or index that is considered to be a cause of the state change is estimated, and this is passed through the user interface unit 7 to the operator. Or notify the process manager.

(a) The state contribution factor variable having the largest contribution amount is used.
(b) Three variables from the largest contribution amount are used as state change factor variables.
(c) The value of the contribution amount exceeds the average contribution amount ± k * standard deviation (k: parameter) of the contribution amount as a state change factor variable.

The user interface unit 7 shown in FIG. 2 not only presents the abnormality detection result and the factor variable estimation result as described above, but as described above, when the statistical data such as the Q statistic and the T ² statistic is used. A series graph (trend graph) may be constantly monitored. In addition, index data such as an operation management index converted by the variable conversion unit 62 may be constantly monitored as a trend graph.

As described above, in the embodiment shown in FIG. 1 and FIG. 2, in the advanced sewage treatment process, changes in operation management indexes (SRT and load ratio) and the like that are constantly managed by the plant manager are represented by MSPC. Can be detected within the framework of statistical process monitoring, and can provide useful diagnostic information for plant managers and operators. In other words, process abnormalities that are key to the operation management of advanced sewage treatment processes can be caused by changing the process state, such as differential values (difference values), integral values (integrated values) of measurement variables, or measurement variables every weekday / holiday. By inputting to the MSPC with an easy-to-understand index, it is possible to detect a sign of a state change or abnormality more quickly and accurately in the MSPC framework.

  Next, an embodiment when the monitoring / diagnosis system shown in FIG. 1 is applied to the sludge treatment system shown in FIG. 3 will be described.

Sludge treatment process is a target process shown in FIG. 3 includes a first sedimentation tank 801, a biological reactor 802, made up of the final sedimentation tank 803, a sewage treatment process the sludge supply source. In addition, in order to treat the sludge generated from these, a sludge treatment process comprising a centrifugal concentrator 804, a pressure / normal pressure concentrator 805, a sludge concentrator 806, a sludge digester 807, and a dehydrator 808. Have

  Further, as a process sensor, an initial sedimentation basin sludge flow rate sensor 811 and an initial sedimentation basin sludge concentration sensor 812 are provided for the initial sedimentation basin 801. The first sedimentation basin sludge flow rate sensor 811 measures the sludge flow rate discharged from the first sedimentation basin 801 to the sludge concentration tank 806. The first sedimentation basin sludge concentration sensor 812 measures the sludge concentration in the first sedimentation basin 801.

  Further, for the final sedimentation basin 803, there are a concentration tank surplus sludge input sensor 813 for measuring the amount input to the sludge concentration tank 806 and a surplus sludge concentration sensor 814 for measuring the surplus sludge concentration in the final sedimentation tank 803. Is provided.

  The sludge concentration tank 806 is provided with a sludge concentration tank separation liquid flow sensor 815 for measuring the flow rate and turbidity of the separation liquid discharged from the sludge concentration tank 806, and a sludge concentration tank separation liquid SS sensor 816. Yes. Concentrated sludge flow for measuring the concentration sludge flow rate (sludge digestion tank input sludge amount) and its concentration (sludge digestion tank input sludge concentration) flowing in this pipe from the sludge concentration tank 806 to the sludge digestion tank 807 A sensor 817 and a concentrated sludge concentration sensor 818 are provided. Further, a sludge concentration tank solid concentration sensor 819 for measuring the solid concentration in the sludge concentration tank 806 and a sludge concentration tank interface level sensor 8110 are provided.

For the centrifugal concentrator 804, the centrifugal concentrator electric sensor 8111 that measures the motor output, the centrifugal concentrator screw conveyor rotational sensor 8112 that measures the rotational speed of the screw conveyor, and the bowl rotational speed A centrifugal concentration tank bowl rotational speed sensor 8113 is provided for measuring the amount of excess sludge charged to the centrifugal concentration tank 804 from the final sedimentation tank 803, and an excess sludge input amount sensor 8114 for measuring the amount of excess sludge charged to the centrifugal concentration tank 804 is provided. Yes,
In addition, for the pressurized / normal pressure concentrating tank 805, a pressure / normal pressure concentrating tank surplus sludge input sensor 8115 for measuring the surplus sludge flow rate introduced from the final sedimentation tank 803, and the pressurized / normal pressure concentrating tank 805 Pressurized / normal pressure water sensor 8116, floss thickness sensor 8117, flotation sludge concentration sensor 8118, and flotation sludge scraping to measure pressurized / normal pressure water volume, floss thickness, flotation sludge concentration, flotation sludge scraping thickness respectively in tank 805 Thickness sensors 8119 are provided respectively.
For the sludge digestion tank 807, a sludge digestion tank digested sludge amount sensor 8120 for measuring the amount of digested sludge discharged from the sludge digestion tank 807 and the sludge concentration in the pipeline to the dehydrator 808 at the subsequent stage, The sludge digestion tank digested sludge concentration sensor 8121 is provided, and the sludge digester input organic substance concentration sensor 8122 is provided in the input pipe line from the previous stage. Further, in this sludge digestion tank 807, digested sludge organic matter concentration sensor 8123 and digestion temperature sensor 8124 for measuring digested sludge organic matter concentration, digestion temperature, digestion gas generation amount, methane concentration, CO2 concentration, hydrogen sulfide concentration, pH, ORP, respectively. A digestion gas generation amount sensor 8125, a methane concentration sensor 8126, a CO2 concentration sensor 8127, a hydrogen sulfide concentration sensor 8128, a pH sensor 8129, and an ORP sensor 8130 are provided. Further, a desorption liquid SS concentration sensor 8131 in the sludge digestion tank 807 is provided in the circulation line to the first sedimentation tank 801.

  The dehydrator 808 is provided with a dehydrator filtration flow rate sensor 8132 that measures the filtration flow rate.

  Next, the operation of the embodiment shown in FIGS. 1 and 3 will be described. In this embodiment, since the target process is different from the above-described embodiment of FIGS. 1 and 2, only the operation of the variable conversion equation determination unit 42 in FIG. 1 is different. Therefore, only the operation of this part will be described.

In the variable conversion formula determination unit 42, a management index useful for an operator from various measurement variables collected and stored by the process measurement data collection / storage unit 2, and an index useful for early detection of a process state change or abnormal sign. Determine the conversion formula. Examples of conversion formulas are listed below. These conversion equations are summarized in Table 2 shown in FIGS. 9A, 9B, and 9C.

  For example, in the sludge concentration tank 806, in addition to the sludge from the first settling tank 801, excess sludge from the final settling tank 803 may be concentrated. In this case, if the excess sludge mixing rate is high, the sedimentation in the sludge concentration tank 806 may be disrupted. Therefore, it is preferable to calculate the excess sludge mixing rate in Table 2 as an index. Further, it is preferable to calculate the concentration rate of solids in the concentration tank as an index for evaluating the treatment efficiency of the sludge concentration tank 806. Furthermore, since the treatment of the sludge concentration tank 806 is managed by the residence time (HRT), calculating the HRT index is useful for operation management of the sludge concentration tank 806. Similarly, as indices useful for operation management of the sludge concentration tank 806, there are a concentrate tank solid load, a sludge solid retention time, and the like. Furthermore, when the sludge concentration deteriorates, it is considered that the sludge interface rapidly rises, so the rate of change of the sludge interface level is an important index for the concentration treatment.

  If such an index is calculated using measurement data according to the calculation formula shown in Table 2, operation management for the sludge concentration tank 806 becomes more appropriate.

  In addition, since excess sludge may be difficult to concentrate, it is not possible to concentrate on machinery using a pressure / normal pressure concentrator 805 such as a centrifugal concentrator 804, a pressurized levitation concentrator, or an atmospheric levitation concentrator, instead of the sludge concentration tank 806. May be concentrated.

  In the centrifugal concentrator 804, if the centrifugal effect is increased, the concentrated sludge concentration and the solids recovery rate are increased. However, since power costs are required for this purpose, the centrifugal effect may be monitored in a balance between processing efficiency and energy saving. Also, since the difference in the rotation speed between the screw conveyor and the bowl is related to the concentrated sludge concentration and the solids recovery rate, it is better to monitor this differential speed.

  Further, in the pressure flotation type concentrator and the normal pressure flotation type concentrator, the gas-solid ratio is managed as a management index for stably floating the sludge. For this reason, the gas-solid ratio is an important management item as a management index of the pressure / normal pressure concentrator 805. In addition, since the sludge is levitated and recovered by scraping the sludge in the pressurized flotation type concentrator and the normal pressure flotation type concentrator, the rate of change in the thickness of the flotation sludge (floss) is also an important index. Moreover, the frequency of flotation sludge scraping is also an important management item for pressurized flotation type concentrators and atmospheric flotation type thickeners.

  Therefore, if such an index is calculated using the measurement data according to the calculation formula shown in Table 2, operation management for the centrifugal concentrator 804 and the pressurized / normal pressure concentrator 805 becomes more appropriate.

In the sludge digestion tank 807, methane gas is recovered as energy by digestion, so the digestibility, digested sludge amount, digestion days, digestion days / digestion temperature, etc. are important management indexes. A more direct index is a gas generation rate, and an index that is a quality index of the generated gas and also a management index is a methane gas composition ratio, a CO2 composition ratio, and a hydrogen sulfide composition ratio. Furthermore, since the load amount is often managed in the same manner as the advanced sewage treatment process, it is preferable to manage the digester solid load, digester organic load, digester nitrogen load, etc. The ratio of organic load and nitrogen load is also an important management item as an index that affects For the purpose of diagnosing abnormalities, it is preferable that the ratio of pH and ORP, pH change rate, and ORP change rate are also input to the MSPC as in the advanced sewage treatment process. Furthermore, since it is known that when the SS of the digester detachment liquid rapidly increases, the SS change rate of the detachment liquid may be input to the MSPC. Moreover, since it is conceivable that the gas generation amount changes suddenly at the time of abnormality, the rate of change of the gas generation amount can also be used as an index.

Finally, in the dehydrator 808, but dewatering sludge, the rate of change of the filtered flow rate for filtration flow rate is important for operational management can be selected as an index.

  The variable conversion formula determination unit 42 determines the operation management index of each sub-process (concentration, digestion, dehydration) and the abnormal sign detection index as described above using the measurement variables of the various process sensors 811 to 8132. .

According to this embodiment, in each sub-process of the sludge treatment process, a change in the operation management index that is managed with care by the plant manager is detected in the statistical process monitoring framework by MSPC. It is possible to present diagnostic information useful for sludge treatment plant managers and operators. In addition, by inputting the process abnormality that is the key in the operation management of the sludge treatment process into the MSPC with an index that makes it easier to catch the process state change such as the differential value (difference value) of the measurement variable, the MSPC framework Among them, signs of state changes and abnormalities can be detected more quickly and accurately.

  Next, the embodiment shown in FIGS. 1 and 4 will be described. In this embodiment, the monitoring / diagnosis system shown in FIG. 1 is applied to the water purification / distribution process shown in FIG.

  In the water purification / supply / distribution process, which is the target process shown in FIG. 4, a landing well 901, a mixing basin 902, a settling basin 903, a filtration basin 904, a clean basin 905, and a water basin 906 are sequentially arranged in series. Further, for the filtration basin 904, a drainage basin 907 and a drainage basin 908 for storing the wastewater generated there are provided.

In addition, a water intake pump 911 for the landing well 901 is provided as an actuator, a mixing tank 902 is provided with a stirrer 912, and a filter 904 is provided with a washing pump 913 for washing with water from the water purification tank 905. It has been. Further, a water supply pump 914 is provided between the water purification tank 905 and the distribution reservoir 906, and a water distribution pump 915 is provided between the distribution reservoir 906 and the distribution area. A return pump 916 is provided between the distribution reservoir 907 and the mixing basin 902, and a concentrator 917 and a dehydrator 918 are provided on the outlet side of the mud basin 908.

Further, as a process sensor, a water intake flow rate sensor 921 is provided in the pipe line to the landing well 901. The landing well 901 includes a chlorine demand sensor 922, a geosmin concentration sensor 923, a 2MIB concentration sensor 924, a trihalomethane concentration sensor 925, And a TOC sensor 926 is provided. The mixing basin 902 is provided with a hypochlorous acid injection amount sensor 927, a flocculant injection amount sensor 928, a pH adjuster injection amount sensor 929, and an activated carbon injection amount sensor 9210. Further, a sludge extraction amount sensor 9211 and a sludge concentration sensor 9212 are provided in the pipeline from the settling basin 903 to the waste mud basin 908. Furthermore, the filtration basin 904 is provided with the residual chlorine concentration sensor 9213, the entrance side of the distribution reservoir 906 is provided with distributed water flow sensor 9214, the water distribution ku is provided with water supply Ryose Nsa 9215.

  Next, the operation of the embodiment shown in FIGS. 1 and 4 will be described. This embodiment is also different from the above-described embodiment of FIGS. 1 and 2 in the target process, and only the operation of the variable conversion equation determination unit 42 in FIG. 1 is different. Therefore, only the operation of this part will be described.

In the variable conversion formula determination unit 42, management indexes useful for the operator and conversion of indexes useful for early detection of process state changes and abnormal signs from various measurement variables collected and stored by the process measurement data collection / storage unit 2 Determine the formula. Examples of this conversion formula are listed. These conversion equations are summarized in Table 3 shown in FIGS. 10A and 10B.

  In the water purification plant, flocculant such as PAC is injected, but sludge is generated by the flocculant. The ratio between the generated sludge amount and the flocculant injection amount is an index for an appropriate injection amount. Similarly, since hypochlorous acid is often injected with respect to the chlorine demand, the ratio between the chlorine demand and the hypochlorous acid injection is also an index.

  On the other hand, water leakage is a big problem in the distribution pipe network in the distribution area. When evaluating the estimated value of this amount of water leakage, the amount of water distribution during night hours when there is not much demand for water may be referred to. In this case, decimating not only the amount of water distributed at each time but also the amount of water distributed in a predetermined night time period may be useful for diagnosing the amount of water leakage. Also, if daytime time zone data is missing, after extracting the nighttime zone data for a specific day by decimation, the data may be interpolated appropriately to become that day's data. Good. That is, for example, when time zone data of 8 hours at night is extracted using the upsampler concept, 8 × 3 = 24 hours of data can be obtained by interpolating with the same data every 3 points. Can be generated. In this way, the amount of water distribution at night time can be used as one index data. Moreover, since the demand for water has daily fluctuations according to people's life patterns, measurement data such as the amount of water distribution every predetermined time may be useful. Similarly, measurement data for each weekday / holiday can be a useful index.

In the water purification / distribution process, the Japan Water Works Association (JWWA) has published a performance indicator (PI), and the water purification / distribution process is often managed by PI. Some of the PIs can also be operational management indicators that can be calculated from sensors that can be measured online (at least in principle). PI is usually defined in units of years, but if the year unit is changed to a predetermined time unit and monitored, it will be possible to monitor online how PI is changing, and useful management information It can be. As PIs that can be monitored online (at least in principle), the raw water effective utilization rate, the delicious water achievement rate seen from the mold odor, the delicious water achievement rate seen from the chlorine odor, the total trihalomethane concentration water quality standard ratio, the organic matter ( TOC) concentration water quality standard ratio, activated carbon input ratio, chemical stockpiles days, fuel Storage days, the supply unit price, YuOsamuritsu, power consumption (per water distribution amount 1 m ^3), energy consumption (per water distribution amount 1 m ^3), renewable energy utilization, effective utilization rate of water purification soil generated, CO2 emissions (per water distribution amount 1m ^3), underground water rate, pump average occupancy rate, leakage rate, water supply number per leakage amount, there is.

  These indices can be calculated from the measurement values obtained by the process sensors 921 to 9215 by the calculation formula shown in Table 3. However, since the amount of electric power and the amount of CO2 generated are not usually measured, they are converted from the flow rate and specifications of equipment such as a pump.

According to this embodiment, in the water purification / supply / distribution process, it is possible to detect changes in the performance index and operation management index of the water purification process within the framework of statistical process monitoring by MSPC. Useful diagnostic information for the operator or operator. In addition, MSPC frames can be entered by inputting process abnormalities, which are key in the operation management of water purification and water supply / distribution processes, into the MSPC with indices that make it easier to capture process state changes, such as differential values (difference values) of measurement variables. It is possible to detect signs of state changes and abnormalities more quickly and accurately in the workpiece.

  Next, for the process measurement variables collected and stored in the data collection / storage unit 2 before the selected variable determination unit 41 and the variable selection unit 61 shown in FIG. FIG. 5 shows an embodiment in which a process measurement variable is newly generated by shifting the time over R, and an expanded process measurement variable that is R / T times the number of original process measurement variables is configured. I will explain.

  This embodiment is characterized in that the processing shown in FIG. 5 is performed immediately after the past data (offline data) extraction unit 3 and immediately after the current data (online data) extraction unit 5. explain.

FIG. 5 shows a set of measurement data stored in the process measurement data collection / storage unit 2 as X, and a data set starting at the time t as X (t). In FIG. 5 , X (t−1) to X (t−7) are described in which the time is shifted from 1 step to 7 steps in a predetermined time unit.

In each of the above-described embodiments, after data is extracted from the measurement data stored in the process measurement data collection / storage unit 2 by the past data (offline data) extraction unit 3 or the current data (online data) extraction unit 5, Processing of the selection variable determination unit 41, the variable conversion formula determination unit 42, the variable selection unit 61, and the variable conversion unit 62 has been performed. This means that treatment with X (t) in FIG. On the other hand, in the present embodiment, an expanded measurement variable: Y (t) = [X (t) X (t−1)... X (t−M)] is used instead of X (t). The process of the selection variable determination unit 41, the variable conversion formula determination unit 42, the variable selection unit 61, and the variable conversion unit 62 is performed. Here, t−M is M = 7, but has no particular meaning, and is determined in advance. Subsequent processing is as in each embodiment.

  In this way, even in a process in which a time delay associated with the residence time of the process is clearly present, the same effects as those of the above-described embodiments can be obtained in consideration of the process time delay. it can.

Next, prior to the selection variable determination unit 41 and the variable selection unit 61 shown in FIG. 1, the process measurement variables collected and stored in the process measurement data collection / storage unit 2 are decomposed by discrete wavelet transform. Embodiment in which original process data is divided into N by applying a digital filter constituted by a reconstruction algorithm, and an expanded process measurement variable N times the number of original process measurement variables is configured Will be described with reference to FIG.

  This embodiment is also characterized in that the processing shown in FIG. 6 is performed immediately after the past data (offline data) extraction unit 3 and immediately after the current data (online data) extraction unit 5. To do.

  In FIG. 6, the set of measurement data stored in the process measurement data collection / storage unit 2 is X (t), and the discrete wavelet transform is applied to this and the data reconstructed and returned to time-series data is X1. (t) to X4 (t).

  In the embodiment described above, the past data (offline data) extraction unit 3 or the current data (online data) extraction unit 5 extracts data from the measurement data stored in the process measurement data collection / storage unit 2 and then selects it. The process of the variable determination part 41, the variable conversion formula determination part 42, the variable selection part 61, and the variable conversion part 62 was implemented. This means that processing is performed using X (t) in FIG.

  On the other hand, in this embodiment, instead of X (t), measurement variables decomposed for each frequency by discrete wavelet transformation: Y (t) = [X1 (t) X2 (t)... Xm ( t)], the processing of the selected variable determination unit 41, the variable conversion formula determination unit 42, the variable selection unit 61, and the variable conversion unit 62 is performed. Here, m = 4 in FIG. 6 is not particularly significant, and is determined in advance. Subsequent processing is as in each of the embodiments described above.

  In this way, even if there are variations in the change speed of multiple process measurement variables or there are many non-stationary fluctuations such as a mixture of fast and slow changes, taking into account the difference in change speed and non-stationarity The same effects as those of the above-described embodiments can be obtained.

  Next, M process monitoring / diagnosis devices are constructed partially (locally) for each processing unit such as each processing sequence or each distribution block, and each of these M monitoring units is further monitored. An embodiment for performing overall (global) process monitoring / diagnosis by MSPC using each statistic calculated from the diagnostic apparatus as an input will be described with reference to FIG.

In FIG. 7, the process monitoring / diagnosis apparatus of series 1 to the process monitoring / diagnosis apparatus of series N are the same as those shown in FIG. 1, but the Q statistics calculated by the monitoring apparatuses of each series It is characterized in that it has an overall process monitoring / diagnostic device that receives the quantity and T ² statistics.

  Also in this embodiment, the operation of characteristic points different from the above-described embodiment will be described below.

In each series of process monitoring / diagnostic apparatuses, processes are monitored by the method according to the above-described embodiment described with reference to FIG. 1, but each Q statistic and T ² statistic are monitored independently. In addition, an overall process monitoring / diagnostic apparatus based on MSPC using these statistics as inputs is operating at a higher level. The operation of the overall process monitoring / diagnostic apparatus is exactly the same as that of the above-described embodiment except that the input is the Q statistic and T ² statistic from each series.

  By having this overall process monitoring / diagnostic device, if any state change is detected by the overall process monitoring / diagnostic device, it is possible to determine in which sequence it was detected using the concept of contribution amount. . If the state change of a specific series, it can be determined that the abnormality is only in that series, but if there are state changes in multiple series at the same time, it is an abnormality that affects multiple series. Can be judged. For example, when the target process is the sewage treatment process shown in FIG. 2, when an operation state of a certain series is more difficult than an operation state of another series, and a state change occurs in the process, the entire process monitoring / diagnosis device The specific amount of the contribution of the detected abnormality is high. On the other hand, when poisonous substances are mixed into the inflowing sewage, the entire series is affected. Therefore, the contribution amount of the abnormality detected by the overall process monitoring / diagnosis apparatus appears in the entire series.

  By configuring in this way, when the process is managed for each processing sequence, managed for each water distribution block, or when the processing process is operated in a network with a connecting pipe or connecting pipe Even so, the same effects as those of the above-described embodiments can be obtained for each processing unit. In addition, by having an overall process monitoring / diagnostic device, it is possible to simultaneously determine whether the state changes for each processing unit or for the entire process, and process monitoring / diagnosis over the entire plant becomes possible. .

  In this way, in process monitoring of water and sewage plants, etc., changes in operational management indices and plant performance indices normally managed by plant managers and operators should be detected within the statistical process monitoring framework by MSPC. It is possible to provide useful diagnostic information for plant managers and operators.

Also, in process management of processes such as water and sewage plants, key process abnormalities can be measured by measuring differential values (difference values) of measurement variables, integral values of measurement variables (integrated values), products and ratios of measurement variables at the same time, and measurement. By inputting into MSPC an index that makes it easier to capture process state changes, such as extracting characteristic data from data, it is possible to quickly and accurately detect signs of state changes and abnormalities within the MSPC framework. And can provide useful awareness tools for plant operations managers.

2 ... Process measurement data collection / storage unit 3 ... Past data (offline data) extraction unit 4 ... Process monitoring model construction / supply unit 5 ... Current data (online data) extraction unit 6 ... Process monitoring / diagnosis unit 41 ... Selection variable determination Unit 42 ... variable conversion formula determination unit 43 ... normal data extraction unit 44 ... normalization parameter determination unit 45 ... diagnostic model construction unit 46 ... statistic threshold value setting unit 47 ... state change factor contribution amount formula setting unit 61 ... variable selection unit 62 ... variable conversion unit 63 ... outlier removal unit 64 ... data normalization unit 65 ... statistic monitoring unit 66 ... state change detection unit 67 ... factor item (variable) estimation unit

Claims (9)

A data collection / storage unit that collects and holds time-series data of a plurality of measurement variables consisting of state quantities and manipulated variables of the target process measured at a predetermined cycle by a plurality of process sensors provided in the target process When,
A process model construction / supply unit that constructs and supplies a process monitoring model using past time-series data of a plurality of measurement variables stored in the data collection / storage unit ;
A process monitoring / diagnostic unit that monitors the state of a process using online data extracted from the data collection / storing unit and a process monitoring model constructed by the process model construction / supply unit, and detects state changes and abnormal signs And
The process model construction / supply section
A selection variable determination unit that selects all or some of the variables necessary to construct the process monitoring model from past time series data of a plurality of measurement variables stored in the data collection / storage unit ;
From a plurality of measurement variables stored in the data collection / storage unit, a predetermined conversion formula for obtaining a management index useful for operation of the target process and an index useful for early detection of process state changes and abnormal signs is provided. A set variable conversion formula determination unit;
The selected variable obtained by removing abnormal data such as outliers from the time series data of the past converted variable converted by using the selected variable selected by the selected variable determining unit and the formula of the variable conversion formula determining unit a normalization parameter determination unit that determine the parameters ai and bi for normalizing data by (xi (t) -ai) / bi to normal time series data of the conversion variables,
For data normalized using the normalization parameters determined by the normalization parameter determination unit, there are many such as principal component analysis (PCA), principal component regression (PCR), and partial least squares (PLS). A diagnostic model construction unit that defines an expression for generating at least one or more diagnostic statistical data using one of the variable analysis means;
A statistic threshold value setting unit for detecting a change in state with respect to the one or more diagnostic statistic data generated by the diagnostic model construction unit;
The process monitoring / diagnosis unit
A variable selection unit that sequentially retrieves the current data corresponding to the selected variable determined by the selected variable determination unit from the data collection and storage unit;
A variable conversion unit that performs variable conversion to obtain a current index from the current data of the data collection and storage unit using the variable conversion formula determined by the variable conversion formula determination unit;
A data normalization unit for normalizing the online data of the variable selected and the converted variable using the normalization parameter determined by the normalization parameter determination unit;
A statistic monitoring unit that generates statistic data from the online data normalized by the data normalization unit based on the statistic generation formula defined by the diagnostic model construction unit and makes it monitorable,
A state change detection unit that detects a change in process state or an abnormality when online statistical data generated by the statistics monitoring unit exceeds a threshold value determined by the statistics threshold value setting unit. Process monitoring and diagnosis device.
Xi: i-th selected variable / transform variable, ai: constant representing shift for i-th select variable / transform variable (shift parameter), bi: constant representing scaling for i-th select variable / transform variable (scaling) Parameter). The process model construction / supply unit further includes a state change factor contribution expression setting unit that estimates a factor when a state change occurs from the selection variable and the conversion variable,
The process monitoring / diagnostic unit, when a state change or an abnormality of a process is detected by the state change detection unit, calculates a variable that is a factor by a contribution amount calculation set by the state change factor contribution amount setting unit. The process monitoring diagnosis apparatus according to claim 1, further comprising a factor item (variable) estimation unit for estimation.   The variable conversion equation determination unit includes nonlinear conversion (including: product (multiplication) and quotient (division)), differential / difference conversion, integration / integration conversion, decimation conversion with a predetermined period, interpolation conversion with a predetermined period, management index The process monitoring diagnosis apparatus according to claim 1, wherein at least one conversion formula is included among the / performance index conversions.   In the preceding stage of the processing in the selected variable determining unit and the variable selecting unit, the process measurement variables collected and stored in the data collection / storage unit are timed over a predetermined period R in a predetermined time unit T. A process for generating a new process measurement variable by shifting and configuring an expanded process measurement variable that is R / T times the number of original process measurement variables is included. 4. The process monitoring diagnostic apparatus according to any one of 3 above.   The process measurement variables collected and stored in the data collection / storing unit are configured by a decomposition / reconstruction algorithm based on discrete wavelet transform in the previous stage of processing in the selected variable determination unit and the variable selection unit. The original process data is divided into N pieces by applying a digital filter, and an expanded process measurement variable N times the number of original process measurement variables is configured. Item 4. The process monitoring diagnostic apparatus according to any one of Items 3 to 3. The process monitoring diagnostic apparatus according to claim 1, M-number for each processing unit such as a processing sequence for each and water distribution block each (M: the number of processing units) was constructed, further, the M-number of each monitoring diagnostic apparatus A hierarchical process monitoring / diagnostic apparatus characterized by having an overall process monitoring / diagnosis apparatus using MSPC that receives each statistic calculated from The target process is a biological wastewater treatment process such as a sewage treatment process / industrial wastewater process,
The conversion formula by the variable conversion formula determination unit includes sludge residence time (SRT), aerobic tank sludge residence time (A-SRT), hydraulic residence time (HRT), Log (SRT) / water temperature, Log (A- SRT) / water temperature, excess sludge generation, organic matter (COD and / or BOD) load, nitrogen load, phosphorus load, organic matter (BOD and / or COD) -SS load, water area load, phosphorus load / nitrogen load , Organic load / nitrogen load, organic load / phosphorus load, pH / ORP, DO change rate (differential value), air flow rate change rate (differential value), ammonia concentration change rate, nitric acid concentration change rate, phosphorus concentration change rate, One or more conversion formulas of pH change rate, ORP change rate, sludge interface change rate, water temperature change rate, rainfall integrated value (integrated value) for a predetermined period, measurement data for weekdays / holidays A process monitoring diagnostic apparatus according to any one of claims 1 to 5, further comprising: The target process is a sludge treatment process,
As a conversion formula by the variable conversion formula determination unit, the concentration rate of the excess concentration of sludge in the concentration tank, the recovery rate of the concentration of solid in the concentration tank, the concentration tank HRT, the concentration time of the concentration tank sludge solids, the rate of change of the concentration tank sludge interface, the centrifugal concentrator centrifugal effect, Centrifugal concentrator screw conveyor and bowl speed differential speed, pressure / normal pressure concentrator gas-solid ratio, pressure / normal pressure concentrator floss thickness change rate, pressure / normal pressure concentrator flotation sludge scraping frequency, digester Digestibility, digester digestion sludge volume, digester digestion days, digester digestion days / digestion temperature, gas generation rate, methane gas composition ratio, CO2 composition ratio, hydrogen sulfide composition ratio, digester solid load, digester organic load, Digestor organic load / nitrogen load, pH / ORP, pH change rate, ORP change rate, digester detachment liquid SS change rate, temperature change rate, gas generation rate change rate, dehydrator filtration flow rate change rate It must have more than one conversion formula Process monitoring diagnostic apparatus according to any one of claims 1 to 5 and. The target process is water purification / distribution process,
As a conversion formula by the variable conversion formula determination unit, sludge amount / flocculating agent injection amount, chlorine demand amount / hypochlorous acid injection amount, water distribution amount in a predefined night time period, water distribution amount in a predetermined time or Performance indicator (PI) measured from the amount of water supply or online measurement data, raw water effective utilization rate (%), delicious water achievement rate seen from mold odor (%), delicious water achievement rate seen from chlorine odor (%), Total Trihalomethane Concentration Water Quality Standard Ratio (%), Organic Substance (TOC) Concentration Water Quality Standard Ratio (%), Activated Carbon Input Rate (%), Chemical Stockpiling Days (Day), Fuel Stockpiling Days (Day), Supply Unit Price ( Yen / cubic meter), water supply cost (yen / cubic meter), yield (%), water consumption per cubic meter (kWh / cubic meter), energy consumption per cubic meter (MJ / cubic meter), renewable energy Lugi utilization rate (%), effective utilization rate of purified water generation soil (%), carbon dioxide (CO2) emissions per cubic meter of water distribution (g · CO2 / cubic meter), groundwater rate (%), pump average operation rate (% ), Water leakage rate (%), water leakage amount per number of water supply (cubic meter / year / case), or any one or more of conversion formulas. Process monitoring diagnostic device.

Images