CN114519923B - Intelligent diagnosis and early warning method and system for power plant - Google Patents
Intelligent diagnosis and early warning method and system for power plant Download PDFInfo
- Publication number
- CN114519923B CN114519923B CN202111408475.0A CN202111408475A CN114519923B CN 114519923 B CN114519923 B CN 114519923B CN 202111408475 A CN202111408475 A CN 202111408475A CN 114519923 B CN114519923 B CN 114519923B
- Authority
- CN
- China
- Prior art keywords
- data
- power plant
- early warning
- model
- diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000003745 diagnosis Methods 0.000 title claims abstract description 50
- 238000013499 data model Methods 0.000 claims abstract description 22
- 239000011159 matrix material Substances 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 14
- 238000011156 evaluation Methods 0.000 claims description 30
- 238000001514 detection method Methods 0.000 claims description 27
- 238000012549 training Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000007621 cluster analysis Methods 0.000 claims description 4
- 238000013461 design Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 230000007787 long-term memory Effects 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000009795 derivation Methods 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 229910052739 hydrogen Inorganic materials 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 238000012800 visualization Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 14
- 230000000007 visual effect Effects 0.000 abstract description 9
- 230000003993 interaction Effects 0.000 abstract description 8
- 238000006731 degradation reaction Methods 0.000 abstract description 7
- 230000015556 catabolic process Effects 0.000 abstract description 6
- 238000004422 calculation algorithm Methods 0.000 abstract description 3
- 230000035784 germination Effects 0.000 abstract description 2
- 238000012423 maintenance Methods 0.000 description 12
- 230000015654 memory Effects 0.000 description 12
- 238000007405 data analysis Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 3
- 238000013178 mathematical model Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
- -1 wind Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an intelligent diagnosis early warning method and system for a power plant, which belong to the field of state monitoring and fault diagnosis. The data access module realizes the reading operation with the power plant database; the diagnosis and early warning analysis module constructs a data model according to historical data, then based on model output and real-time data, utilizes a robust non-negative matrix factorization algorithm to diagnose early failure germination, combines the operation performance of a system, determines a failure threshold value, and realizes the prediction of the complete failure time of the system based on Gamma degradation modeling; the visual interaction module is used for displaying the current monitoring state of the system and the future fault trend.
Description
Technical Field
The invention belongs to the field of state monitoring and fault diagnosis, and relates to an intelligent diagnosis early warning method and system for a power plant.
Background
The equipment of the thermal power plant can be basically divided into large professional series of machines, furnaces, electricity and instruments 4, mainly including boilers, steam turbines, generators, transformers, pumps and fans with large and small sizes, electric switches, pipelines and pressure vessels of steam, water, wind, smoke, oil and the like, and the corresponding fault types mainly include mechanical faults and electrical faults. The system can be divided into a series of complex multi-coupling thermodynamic control loops of coordination control, AGC control, primary air control, water supply control, overheat air temperature control and the like, and corresponding faults comprise actuator faults, sensor faults, control performance degradation and the like. At present, most thermal power plants still adopt a strategy of scheduled maintenance and fault maintenance, so that economic resource waste of the power plants is caused, equipment maintenance is insufficient or excessive, and health management of the equipment is not facilitated. Economic losses due to equipment failure are also increasingly appreciated and focused by power generation enterprises. In this case, the operation and maintenance mode of performing state monitoring before the equipment fails, that is, pre-judging the possible failure in advance, must draw attention from the power plant.
Conventional state monitoring and fault diagnosis methods are generally classified into three types, namely, analytical model-based methods, knowledge-based methods, and data analysis-based methods. In 1971, beard doctor of the american college of bureau of technology first proposed analytic redundancy technology and replaced hardware redundancy technology with it, which laid the theoretical foundation for fault diagnosis. In 1976, a review article of fault diagnosis was published by Willsky on Automatica, which was considered the earliest article on fault diagnosis. In 1978, himmelblau written a first academic book related to the content of fault diagnosis, and then a method for fault diagnosis by using an analytical model was widely used. The method based on the analytic model utilizes prior information of system performance parameters contained in the dynamic mathematical model, namely, the unchanged analytic relation between the input and the output of the system, compares the actual output value of the system with the expected output value to obtain residual errors, and utilizes the residual errors to detect and identify the states and faults of the system. The fault diagnosis method based on the analytical model is generally suitable for a simple system or process with relatively less input and output and state numbers, and can obtain a mathematical model with higher precision under the condition, and can detect the type of the fault more accurately. However, for a complex control system, characteristics such as nonlinearity, coupling, time variability and the like often exist among variables, and an effective and accurate mathematical model is difficult to build. In addition, the actual control system is influenced by uncertain factors such as noise, external interference and the like, so that the method is invalid, and the application of the method is limited. Knowledge-based methods are classified into fault tree analysis, expert systems, symbolic directed graphs, neural networks, and the like. The Fault tree analysis (Fault TREE ANALYSIS) is from top to bottom and layer by layer analysis from the system terminal Fault, so as to find out the most fundamental Fault factors, and the relation between the events and the system Fault is expressed by a logic graphic form. Such methods are applicable to systems with a great deal of knowledge. However, the current knowledge processing method still has certain difficulty, such as bottleneck problem of knowledge acquisition, limitation of knowledge reasoning, poor self-learning ability and self-adaptation ability, and the like.
Based on the data analysis method, the implicit information in the data is mined through various data processing and analysis methods such as statistical analysis, cluster analysis, spectrum analysis, wavelet analysis and the like. The main advantage of the data analysis method is that effective information is extracted by utilizing a large amount of related data of the control system through a mathematical method, and useful statistical data and calculation reasoning information are provided for on-site staff to improve the monitoring performance of the system. The data analysis-based method projects the main features into a low-dimensional space through strict statistical analysis, and can greatly simplify and improve the diagnosis process. Therefore, the fault diagnosis method based on data analysis has great theoretical significance and practical application value. Along with the rapid development of information technology, particularly the large-scale construction of a DCS system, an SIS system and various information systems, the data volume accumulated by a power plant is promoted to be rapidly increased, and a foundation is laid for the large data analysis and research work of key equipment of power generation enterprises. Through a special and efficient real-time data mining technology, the new generation of equipment state on-line monitoring and fault early warning system can help a user to realize intelligent management of equipment states, further give full play to professional efficiency of an equipment manager, change fault post-treatment into pre-prevention, grasp overall dynamic change in equipment operation in real time, greatly improve operation safety level and efficiency of equipment in the life cycle of each equipment, reduce unplanned shutdown and accidents caused by equipment reasons, reduce equipment operation and maintenance cost and create more benefits for power generation enterprises.
Traditional power plant equipment overhaul has the condition of insufficient maintenance or excessive maintenance, and the realization of a state monitoring method based on analytical models and knowledge requires a great deal of expertise and field experience.
Disclosure of Invention
The invention aims to overcome the defects of insufficient maintenance or excessive maintenance and a state detection method in the prior art in the overhaul of power plant equipment, and provides an intelligent diagnosis and early warning method and system for a power plant.
In order to achieve the above purpose, the invention is realized by adopting the following technical scheme:
An intelligent diagnosis and early warning method for a power plant comprises the following steps:
step 1) acquiring real-time power plant SIS data, and preprocessing the acquired power plant SIS data;
Step 2) constructing a data model by utilizing data in a power plant SIS historical database and a DCS database;
Step 3) performing fault detection and fault diagnosis based on the preprocessed real-time power plant SIS data and data model;
And 4) constructing an evaluation system of the power plant operation system, calculating an evaluation benchmark and a failure threshold of the evaluation system, establishing an early warning model of the equipment, and calculating to obtain the failure time of the equipment.
Preferably, in step 1), the synchronous access mode and the asynchronous access mode are respectively adopted, power plant SIS data operated on site are collected, and the collected power plant SIS data are stored in a real-time database.
Preferably, in step 1), preprocessing power plant SIS data by using a cluster analysis method;
the preprocessing comprises filtering and bad value elimination.
Preferably, in step 2), different data modeling methods are adopted for different components and systems to construct a data model;
for a single-input single-output component, a data model is constructed by adopting a piecewise linearization method.
Preferably, the method for constructing the data model by adopting piecewise linearization is specifically as follows:
wherein f is an output signal of the component, U is an input signal of the component, [ U i,Ui+1 ] represents an i-th segment characteristic interval, and k 1、k2、k3 is the slope of input signal connecting lines at different stages in the model respectively; b 1、b2、b3 are constants of input signal connecting lines at different stages in the model respectively; Representing the estimated values of the parameters of the model, Is an output estimate of the model; j is the error deviation of the estimated output from the actual output.
Preferably, in step 3), the fault detection is specifically:
Assuming that the training set data under normal conditions is a matrix X e R n×m, the following decomposition is performed: x=wh+s,
Wherein W εR n×k is the base matrix, H εR k×m is the representation matrix, S εR n×m is the residual matrix; the design optimization function is as follows:
s.t.W≥0，H≥0
the update rule of W, H, S is understood by derivation and correlation as follows:
Finally, the fault detection statistics are as follows:
S2＝diag(S(X)S(x))
Wherein S (X) represents a residual matrix obtained by using the training set, and S (X) represents a residual matrix obtained by using the test set;
The fault judging method comprises the following steps: and calculating a control limit S m according to the kernel density estimation method by using the statistic obtained by calculation of the training set, then calculating the statistic by using the test set, and if the statistic exceeds the control limit, indicating that a fault occurs.
An intelligent diagnosis and early warning system for a power plant comprises
The data acquisition unit is used for acquiring real-time power plant operation data and preprocessing the acquired power plant operation data;
the model building unit interacts with the data acquisition unit, divides the components and the subsystems, and builds a data model by using the components and the subsystems;
The fault detection and diagnosis unit is interacted with the model building unit and is used for carrying out fault detection and fault diagnosis based on the output data of the data model and real-time power plant SIS data;
The early warning evaluation unit is interacted with the fault detection diagnosis unit, an early warning system is built, an evaluation system of the early warning system is built, an evaluation standard and a failure threshold value of the evaluation system are calculated, an early warning model is built, and the failure time is calculated.
Preferably, a visualization unit is also included for monitoring the operating status of the power plant.
Preferably, the control unit comprises three control gates;
The control gates are respectively a forgetting gate, an input gate and an output gate;
the forgetting gate is used for screening and reserving important parts in long-term memory;
The input gate is used for selecting and updating the current short-term information;
And outputting the information of the first two of the gate summaries and outputting.
Compared with the prior art, the invention has the following beneficial effects:
The invention discloses an intelligent diagnosis method applied to a power plant, which comprises a three-layer structure, the system comprises a data access module, a diagnosis and early warning analysis module and a visual interaction module. The data access module realizes the reading operation with the power plant database; the diagnosis and early warning analysis module constructs a data model according to historical data, then based on model output and real-time data, utilizes a robust non-negative matrix factorization algorithm to diagnose early failure germination, combines the operation performance of a system, determines a failure threshold value, and realizes the prediction of the complete failure time of the system based on Gamma degradation modeling; the visual interaction module is used for displaying the current monitoring state of the system and the future fault trend. Based on the angle of data driving, the invention utilizes abundant data resources stored in the power plant, realizes fault detection of the power plant system and equipment based on big data analysis technology, evaluates the performance of the power plant system and equipment, predicts the future running condition of the system or equipment based on the fault detection, predicts the failure time of the system or equipment in the future, and therefore, provides reasonable maintenance and overhaul suggestions for field personnel, and greatly improves the working efficiency.
The invention also discloses an intelligent diagnosis early warning system of the power plant, which is characterized in that the intelligent diagnosis early warning system of the power plant is based on abundant historical and real-time data in a power plant database by utilizing a data driving method, and state information of degradation and faults of an evaluation system and equipment is obtained through mining analysis of an intelligent algorithm, and early warning can be carried out on future performance degradation. The data access module is designed to read the operation data of each device and system stored in the power plant database. The diagnosis and early warning analysis module completes modeling, fault detection and fault early warning functions according to the data, and is specifically characterized in that a data model of a power plant system and equipment is built; detecting a failure of the system and the device using the model and the real-time data; and establishing a performance evaluation system of the system and the equipment based on the data analysis method, determining an evaluation benchmark and an failure threshold of the system and the equipment in the system, and finally establishing an early warning model, and predicting the failure time of the system and the equipment according to the evaluation system. The visual interaction module is an interaction interface for presenting the information such as the system and equipment information of the power plant and the detection, evaluation, early warning and the like of the visual interaction module to the user.
Drawings
FIG. 1 is a flow chart of the intelligent diagnosis and early warning method of the power plant;
FIG. 2 is a functional implementation logic of the intelligent diagnosis and early warning method of the power plant of the invention;
FIG. 3 is a block diagram of piecewise linear modeling in an embodiment of the present invention;
FIG. 4 is a LSTM basic unit structure diagram of the intelligent diagnosis and early warning method of the power plant;
FIG. 5 is a visual representation of the intelligent diagnostic and warning method of the power plant of the present invention.
Detailed Description
The invention is described in further detail below with reference to the attached drawing figures:
Example 1
An intelligent diagnosis and early warning method for a power plant comprises the following steps:
step 1) acquiring real-time power plant SIS data, and preprocessing the acquired power plant SIS data;
Step 2) constructing a data model by utilizing data in a power plant SIS historical database and a DCS database;
Step 3) performing fault detection and fault diagnosis based on the preprocessed real-time power plant SIS data and data model;
And 4) constructing an evaluation system of the power plant operation system, calculating an evaluation benchmark and a failure threshold of the evaluation system, establishing an early warning model of the equipment, and calculating to obtain the failure time of the equipment.
Example 2
Taking a control loop of a certain thermal power generating unit as an example, as shown in fig. 1, the system structure diagram of the invention is provided, and the system is composed of a data access module, a diagnosis and early warning analysis module and a visual interaction module. The system establishes a data model of an object (component equipment and a subsystem) by a data driving method through reading historical and real-time data stored in a power plant SIS database, a DCS database and the like, outputs and actually data according to the model, and detects the occurrence of faults by a big data analysis method; in addition, by making a performance evaluation standard, predicting the performance degradation process of the object, thereby realizing the early warning of faults; finally, the monitoring and operation of the field personnel are realized through the design of the visual client.
Specifically, in the data access module, based on 0PC industrial standard, the field data acquisition technology is researched according to the synchronous access mode and the asynchronous access mode respectively, so that an OPC client is developed, and the acquired field data is dumped to a real-time database; and secondly, based on a cluster analysis method, smoothing of random errors and forensic detection of error of the field data are realized, so that filtering and bad value elimination of the field data are realized.
The diagnosis and early warning analysis module mainly comprises functions of modeling, detection, evaluation, early warning and the like.
First, different data modeling methods are adopted for different components and systems.
Modeling by adopting a piecewise linearization method aiming at a simple single-input single-output component;
Defining f to represent the output signal of the component, U being the input signal of the component, [ U i,Ui+1 ] to represent the i-th segment characteristic interval, and constructing the following piecewise linear model by using a least squares method:
assuming the number of samples is N, the following model performance index is defined:
wherein f is an output signal of the component, U is an input signal of the component, [ U i,Ui+1 ] represents an i-th segment characteristic interval, and k 1、k2、k3 is the slope of input signal connecting lines at different stages in the model respectively; b 1、b2、b3 are constants of input signal connecting lines at different stages in the model respectively; Representing the estimated values of the parameters of the model, Is an output estimate of the model; j is the error deviation of the estimated output from the actual output.
And when the performance index of the model is not higher than the set threshold, approving the constructed model, otherwise, shortening the characteristic interval and re-modeling.
Training a model by adopting a long-short-term memory network (LSTM) aiming at a complex multiple-input multiple-output system; it comprises three control gates: forget door, input door, output door; the forgetting gate is used for screening and reserving important parts in long-term memory; the input gate is used for selecting and updating the current short-term information; finally, outputting information of the first two summarized by the gate and outputting the information; when the input data is x, the memory unit is c, and the predicted output is h, firstly calculating the candidate memory unit value at the current moment:
Wherein W xc、Whc is the weight value output by the LSTM unit at the last moment and the input data respectively; x t is an input data value at time t, and h t-1 is a predicted output value at time t-1;
the value of the input gate is then calculated:
it＝σ(Wxixt+Whiht-1+Wcict-1+bi)
wherein W xi、Whi、Wci is the weight value of input data, predicted output data and memory unit in the input gate, b i is bias, sigma is common sigmoid function, and the value range is (0, 1);
calculating the value of the forgetting gate:
ft＝σ(Wxfxt+Whfht-1+Wcfct-1+bf)
wherein W xf、Whf、Wcf is the weight value of the input data, the predicted output data and the memory unit in the forgetting gate, and b f is the bias;
calculating the current memory cell state:
calculating the value of the output gate:
ot＝σ(Wxoxt+Whoht-1+Wcoct-1+bo)
Wherein W xo、Who、Wco is the weight value of the input data, the predicted output data and the memory unit in the output gate, and b o is the bias;
finally, calculating the output of the LSTM:
ht＝ot tanh Ct
After the piecewise linear model and the LSTM model are trained, standard output information of the object can be obtained by inputting control signals of the system or the component.
Secondly, the method for realizing the fault detection function is as follows:
Assuming that training set data under normal conditions (namely, real-time signals are used as model inputs, and obtained standard object output signals) are matrix X epsilon R n×m, decomposing the training set data as follows:
X＝WH+S
Wherein W εR n×k is the base matrix, H εR k×m is the representation matrix, S εR n×m is the residual matrix; the design optimization function is as follows:
s.t.W≥0，H≥0
the update rule of W, H, S is understood by derivation and correlation as follows:
Finally, the fault detection statistics are as follows:
S2＝diag(S(X)S(x))
wherein S (X) represents a residual matrix obtained by using the training set, and S (X) represents a residual matrix obtained by using the test set; the fault judging method comprises the following steps: and calculating a control limit S m according to the kernel density estimation method by using the statistic obtained by calculation of the training set, then calculating the statistic by using the test set, and if the statistic exceeds the control limit, indicating that a fault occurs.
Next, the Hurst index was used to evaluate the performance of the system:
for a time series Y containing N data, the average value thereof is calculated
Calculating the dispersion accumulation sequence Y 'of Y'
Dividing the dispersion accumulation sequence Y' into W non-overlapping equal-length intervals with a window length L, (W=N/L, an integer); for each interval, a first order linear fit is performed on the L data points contained therein by using a least square method:
let L data points be (t (1), y (1)), (t (2), y (2)), …, (t (k), y (k)), …, (t (L), y (L)) in this order, there are
The fit over the j-th interval is y j＝ajt+bj (j=1, 2, 3., W);
calculating the sum of the mean square deviations of the j-th interval after filtering the trend:
Calculating a DFA fluctuation function F (L):
Taking different window lengths L, multiple groups (L, F (L)) can be obtained, and F (L) and L satisfy the power law relation:
F(L)＝a×Lα
In the double logarithmic coordinates (ln (L), ln (F (L)) the data points are fitted by the least squares method, there are
lnF(L)＝αln(L)+lna
Wherein the slope α of the straight line portion, i.e. the Hurst index;
If α=0.5, it indicates that the system performance is good; whereas the more it deviates from 0.5, the worse the performance is explained; therefore, based on a large amount of historical data, the Hurst indexes under different conditions are calculated statistically, an optimal evaluation standard is screened out, and an index threshold value l of system performance failure is determined.
Finally, the method for early warning the system failure is as follows:
The system degradation model is built as follows:
Dm(t)＝D(t)+ε(t)
Wherein D m (t) is the Hurst performance index measurement value of the system at the measurement time t, D (t) is the actual Hurst performance index value of the system at the measurement time t, epsilon (t) to N (0, sigma 2)
D (t) obeys Gamma distribution, and the probability density function is as follows:
wherein,
The estimated value of each parameter is obtained by calculation as follows:
a＝E(Δω(t))/β
σ2＝0.5(E(Δω(t)2)-E(Δω(t))2-βE(Δω(t)))
Then, the predicted system life distribution is as follows:
And obtaining the system or component failure or failure time under the appointed failure threshold according to the formula.
The visual interaction module presents diagnosis and early warning information of the power plant through a B/S or C/S release mode, including but not limited to modeling results of components and systems, real-time state monitoring, performance evaluation, fault early warning and the like. And the analysis and maintenance of on-site operators are facilitated by the presentation of graph curves and other forms.
In FIG. 4, the input data is x, the memory cell is c, the predicted output is h, the first σ is the input gate, the second σ is the forgetting gate, the third σ is the output gate, the "+" is the current memory cell, and the tan h on the left is the candidate memory cellFirst, the data stream x t,ht-1,ct-1 is obtained through the input gate:
it＝σ(Wxixt+Whiht-1+Wcict-1+bi)
Obtained through a forgetting door:
ft＝σ(Wxfxt+Whfht-1+Wcfct-1+bf)
Finally, the input gate output, c t-1, the forget gate output and the candidate unit are multiplied and converged after "+" to obtain the current memory unit state:
the data stream x t,ht-1,ct-1 is obtained through the output gate:
ot＝σ(Wxoxt+Whoht-1+Wcoct-1+bo)
finally, the output of the current memory unit and the output gate is subjected to X-calculation to obtain the output of LSTM:
ht＝ot tanh Ct
Example 3
An intelligent diagnosis and early warning system for a power plant comprises
The data acquisition unit is used for acquiring real-time power plant operation data and preprocessing the acquired power plant operation data;
the model building unit interacts with the data acquisition unit, divides the components and the subsystems, and builds a data model by using the components and the subsystems;
The fault detection and diagnosis unit is interacted with the model building unit and is used for carrying out fault detection and fault diagnosis based on the output data of the data model and real-time power plant SIS data;
The early warning evaluation unit is interacted with the fault detection diagnosis unit, an early warning system is built, an evaluation system of the early warning system is built, an evaluation standard and a failure threshold value of the evaluation system are calculated, an early warning model is built, and the failure time is calculated.
And the visualization unit is used for monitoring the running state of the power plant.
The control unit comprises three control gates;
The control gates are respectively a forgetting gate, an input gate and an output gate; the forgetting gate is used for screening and reserving important parts in long-term memory; the input gate is used for selecting and updating the current short-term information; and outputting the information of the first two of the gate summaries and outputting.
In summary, the invention uses abundant data resources stored in the power plant based on the data driving angle, realizes fault detection of the power plant system and equipment based on big data analysis technology, evaluates the performance of the power plant system and equipment, predicts the future running condition of the system or equipment based on the fault detection, predicts the failure time of the system or equipment when the system or equipment breaks down, and thereby provides reasonable maintenance and overhaul suggestions for field personnel, and greatly improves the working efficiency.
The above is only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited by this, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (6)
1. The intelligent diagnosis and early warning method for the power plant is characterized by comprising the following steps of:
step 1) acquiring real-time power plant SIS data, and preprocessing the acquired power plant SIS data;
Step 2) constructing a data model by utilizing data in a power plant SIS historical database and a DCS database; wherein, different data modeling methods are adopted for different components and systems to construct a data model;
for a single-input single-output component, constructing a data model by adopting a piecewise linearization method;
The method for constructing the data model by adopting piecewise linearization is specifically as follows:
wherein, As an output signal of the component(s),Is the input signal to the component and,Represent the firstThe section characteristic interval, k 1、k2、k3, is the slope of the input signal connecting line at different stages in the model respectively; b 1、b2、b3 are constants of input signal connecting lines at different stages in the model respectively; Representing the estimated values of the parameters of the model, Is an output estimate of the model; j is the error deviation of the estimated output and the actual output;
Step 3) performing fault detection and fault diagnosis based on the preprocessed real-time power plant SIS data and data model; the fault detection specifically comprises the following steps:
Assuming that the training set data under normal conditions is a matrix It was decomposed as follows:，
wherein, Is a base matrix of the type that,Is a representation of a matrix of the matrix,Is a residual matrix; the design optimization function is as follows:
the update rule of W, H, S is understood by derivation and correlation as follows:
Finally, the fault detection statistics are as follows:
wherein, Representing the residual matrix obtained using the training set,Representing a residual matrix obtained using the test set;
The fault judging method comprises the following steps: the statistic obtained by training set calculation is used for calculating the control limit according to the kernel density estimation method Then calculating statistics by using the test set, and if the statistics exceed a control limit, indicating that a fault occurs;
And 4) constructing an evaluation system of the power plant operation system, calculating an evaluation benchmark and a failure threshold of the evaluation system, establishing an early warning model of the equipment, and calculating to obtain the failure time of the equipment.
2. The intelligent diagnosis and early warning method for a power plant according to claim 1, wherein in step 1), a synchronous access mode and an asynchronous access mode are respectively adopted, power plant SIS data of field operation are collected, and the collected power plant SIS data are stored in a real-time database.
3. The intelligent diagnosis and early warning method for a power plant according to claim 1, wherein in the step 1), the power plant SIS data is preprocessed by a cluster analysis method;
the preprocessing comprises filtering and bad value elimination.
4. A power plant intelligent diagnosis and early warning system for implementing the power plant intelligent diagnosis and early warning method according to any one of claims 1 to 3, characterized by comprising
The data acquisition unit is used for acquiring real-time power plant operation data and preprocessing the acquired power plant operation data;
the model building unit interacts with the data acquisition unit, divides the components and the subsystems, and builds a data model by using the components and the subsystems;
The fault detection and diagnosis unit is interacted with the model building unit and is used for carrying out fault detection and fault diagnosis based on the output data of the data model and real-time power plant SIS data;
The early warning evaluation unit is interacted with the fault detection diagnosis unit, an early warning system is built, an evaluation system of the early warning system is built, an evaluation standard and a failure threshold value of the evaluation system are calculated, an early warning model is built, and the failure time is calculated.
5. The intelligent diagnostic and warning system of claim 4, further comprising a visualization unit for monitoring the operational status of the power plant.
6. The intelligent diagnostic and warning system of a power plant of claim 4, comprising a control unit comprising three control gates;
The control gates are respectively a forgetting gate, an input gate and an output gate;
the forgetting gate is used for screening and reserving important parts in long-term memory;
The input gate is used for selecting and updating the current short-term information;
And outputting the information of the first two of the gate summaries and outputting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111408475.0A CN114519923B (en) | 2021-11-24 | 2021-11-24 | Intelligent diagnosis and early warning method and system for power plant |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111408475.0A CN114519923B (en) | 2021-11-24 | 2021-11-24 | Intelligent diagnosis and early warning method and system for power plant |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114519923A CN114519923A (en) | 2022-05-20 |
| CN114519923B true CN114519923B (en) | 2024-09-24 |
Family
ID=81596248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111408475.0A Active CN114519923B (en) | 2021-11-24 | 2021-11-24 | Intelligent diagnosis and early warning method and system for power plant |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114519923B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115081926B (en) * | 2022-07-14 | 2022-11-11 | 石家庄良村热电有限公司 | Operation safety early warning method and system suitable for intelligent power plant |
| CN115639470B (en) * | 2022-09-23 | 2024-01-30 | 贵州北盘江电力股份有限公司光照分公司 | Generator monitoring method and system based on data trend analysis |
| CN116340427B (en) * | 2023-04-25 | 2023-10-13 | 北京工业大学 | Method for environmental protection data early warning system |
| CN117519067B (en) * | 2023-10-20 | 2024-08-20 | 东北大学 | Multi-frame control performance evaluation method in continuous rolling process |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108460207A (en) * | 2018-02-28 | 2018-08-28 | 上海华电电力发展有限公司 | A kind of fault early warning method of the generating set based on operation data model |
| CN109657836A (en) * | 2018-11-19 | 2019-04-19 | 东莞理工学院 | A kind of Power System Analysis method for early warning based on big data |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112257988A (en) * | 2020-09-29 | 2021-01-22 | 中广核工程有限公司 | Complex accident feature identification and risk early warning system and method for nuclear power plant |
| CN113189941A (en) * | 2021-03-16 | 2021-07-30 | 珠海市钰海电力有限公司 | Intelligent fault early warning method and system for power plant power generation equipment |
| CN113065580B (en) * | 2021-03-17 | 2024-04-16 | 国能大渡河大数据服务有限公司 | Power plant equipment management method and system based on multi-information fusion |
-
2021
- 2021-11-24 CN CN202111408475.0A patent/CN114519923B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108460207A (en) * | 2018-02-28 | 2018-08-28 | 上海华电电力发展有限公司 | A kind of fault early warning method of the generating set based on operation data model |
| CN109657836A (en) * | 2018-11-19 | 2019-04-19 | 东莞理工学院 | A kind of Power System Analysis method for early warning based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114519923A (en) | 2022-05-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114519923B (en) | Intelligent diagnosis and early warning method and system for power plant | |
| Udo et al. | Data-driven predictive maintenance of wind turbine based on SCADA data | |
| Song et al. | Wind turbine health state monitoring based on a Bayesian data-driven approach | |
| CN100470416C (en) | Power plant thermal equipment intelligent state diagnosing and analyzing system | |
| CN117930815B (en) | Wind turbine generator remote fault diagnosis method and system based on cloud platform | |
| Su et al. | Study on an intelligent inference engine in early-warning system of dam health | |
| CN117713221B (en) | Micro-inversion photovoltaic grid-connected optimization system | |
| CN112834211A (en) | Fault early warning method for transmission system of wind turbine generator | |
| CN116050665B (en) | Heat supply equipment fault prediction method | |
| CN113036913B (en) | Method and device for monitoring state of comprehensive energy equipment | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| CN117273440B (en) | Engineering construction Internet of things monitoring and managing system and method based on deep learning | |
| CN117408162B (en) | Power grid fault control method based on digital twin | |
| CN117808166A (en) | Chemical industry safety automation detection monitoring system of clouding PLC | |
| CN111931849A (en) | Hydroelectric generating set operation data trend early warning method | |
| CN117689373A (en) | Maintenance decision support method for energy router of flexible direct-current traction power supply system | |
| CN118035814A (en) | State fault early warning method and system based on digital twin field data analysis equipment | |
| CN117989074A (en) | Intelligent monitoring method for offshore wind turbine based on sensing calculation coordination | |
| Wang | Key techniques in intelligent predictive maintenance (IPdM)–a framework of intelligent faults diagnosis and prognosis system (IFDaPS) | |
| CN118096131B (en) | Operation and maintenance inspection method based on electric power scene model | |
| CN116821610B (en) | Method for optimizing wind power generation efficiency by utilizing big data | |
| KR20220089853A (en) | Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology | |
| CN115600695B (en) | Fault diagnosis method for metering equipment | |
| CN117196575A (en) | Ground equipment fault prediction and health management system universal architecture and use method thereof | |
| CN114320773B (en) | Wind turbine generator system fault early warning method based on power curve analysis and neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |