CN117648543A - Self-evolving substation equipment learning method - Google Patents
Self-evolving substation equipment learning method Download PDFInfo
- Publication number
- CN117648543A CN117648543A CN202410127488.8A CN202410127488A CN117648543A CN 117648543 A CN117648543 A CN 117648543A CN 202410127488 A CN202410127488 A CN 202410127488A CN 117648543 A CN117648543 A CN 117648543A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- decision tree
- substation
- accuracy
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000003066 decision tree Methods 0.000 claims abstract description 69
- 238000012300 Sequence Analysis Methods 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 22
- 230000008859 change Effects 0.000 claims abstract description 20
- 238000013138 pruning Methods 0.000 claims abstract description 13
- 238000000586 desensitisation Methods 0.000 claims abstract description 10
- 238000005516 engineering process Methods 0.000 claims abstract description 9
- 238000012423 maintenance Methods 0.000 claims description 53
- 230000007613 environmental effect Effects 0.000 claims description 34
- 230000008569 process Effects 0.000 claims description 16
- 230000001932 seasonal effect Effects 0.000 claims description 16
- 238000012544 monitoring process Methods 0.000 claims description 11
- 238000013480 data collection Methods 0.000 claims description 8
- 238000011156 evaluation Methods 0.000 claims description 8
- 230000006872 improvement Effects 0.000 claims description 8
- 230000007246 mechanism Effects 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 230000000737 periodic effect Effects 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000007774 longterm Effects 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 4
- 101100391182 Dictyostelium discoideum forI gene Proteins 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000007547 defect Effects 0.000 claims description 3
- 230000001788 irregular Effects 0.000 claims description 3
- 238000013450 outlier detection Methods 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 3
- 238000012731 temporal analysis Methods 0.000 claims description 3
- 238000000700 time series analysis Methods 0.000 claims description 3
- 230000032683 aging Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000015556 catabolic process Effects 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010921 in-depth analysis Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Human Resources & Organizations (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Medical Informatics (AREA)
- Bioethics (AREA)
- Strategic Management (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Water Supply & Treatment (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Public Health (AREA)
- Computer Hardware Design (AREA)
- Databases & Information Systems (AREA)
- Computer Security & Cryptography (AREA)
- Probability & Statistics with Applications (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a self-evolving substation equipment learning method, which comprises the steps of S1, collecting data from a substation by using a plurality of sensors, and establishing a substation data set; s2, processing the collected data set; s3, collecting data in a fixed time interval; s4, data desensitization is carried out by adopting a differential privacy technology; s5, performing time sequence analysis on the equipment operation parameters and the external parameters by adopting an autoregressive integrated moving average model; s6, performing fault prediction by using a decision tree model; s7, collecting the latest operation data from the transformer substation sensor at fixed time intervals; s8, carrying out a time sequence analysis method by the continuation model, and updating key features periodically; s9, adjusting a splitting criterion and a pruning strategy of the decision tree; s10, calculating the accuracy, recall rate and F1 score of the decision tree model; s11, identifying the change trend of the model performance; s12, fine tuning of the model is performed. The invention improves the accuracy and timeliness of fault prediction.
Description
Technical Field
The invention relates to the technical field of information, in particular to a self-evolving substation equipment learning method.
Background
In the operation of a power system, a transformer substation is used as a key node for electric energy transmission, and the stability and safety of the whole power grid are directly influenced by the health condition of equipment. Conventional substation equipment monitoring and fault prediction methods rely on periodic physical inspection and empirically driven predictive models, which have significant limitations in facing complex, dynamically changing operating environments.
First, conventional methods often fail to fully utilize the large amount of data generated by the substation. With the development of sensor technology and data acquisition systems, substation equipment can generate a large amount of data about the running state in real time, including parameters such as temperature, pressure, current and the like. However, conventional methods fail to integrate and analyze such data effectively, thereby missing the opportunity to use such data for in-depth analysis and prediction. Second, conventional fault prediction methods often lack adaptivity. These methods are typically designed based on a fixed set of rules or models that are difficult to adapt to widely varying actual operating conditions. Due to the complexity of the substation environment, such as seasonal changes, weather effects, and equipment aging, these fixed models often fail to accurately predict equipment failure. Furthermore, the prior art has a disadvantage in terms of privacy protection. In collecting and analyzing sensitive operational data, how to ensure data security and privacy protection is an important challenge. Conventional methods often lack effective privacy protection measures in processing sensitive data, which may lead to risks of data disclosure or abuse. Therefore, how to provide a self-evolving substation equipment learning method is a problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The invention aims to provide a self-evolving substation equipment learning method, equipment faults can be predicted more accurately by comprehensively utilizing operation data collected in real time, and the accuracy and timeliness of fault prediction are improved by combining time sequence analysis and a machine learning model.
According to the embodiment of the invention, the self-evolving substation equipment learning method comprises the following steps of:
s1, collecting temperature from a transformer substation by using a plurality of sensorsPressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->And build up a substation data set；
S2, collecting data setsPerforming outlier processing by using the Z-Score, and adopting Min-Max standardization to process data in a data set;
s3, processing the high-speed data stream by adopting a time window method, and collecting data in a fixed time interval;
s4, adding random noise in the data set, and performing data desensitization by adopting a differential privacy technology;
s5, performing time sequence analysis on equipment operation parameters and external parameters by adopting an autoregressive integrated moving average model, and extracting trend, seasonal and random components;
S6, performing fault prediction by using a decision tree model, and predicting whether the equipment is likely to be in fault or not based on the characteristics extracted from the time sequence analysis;
s7, configuring a high-frequency acquisition module, and collecting the latest operation data from the transformer substation sensor at fixed time intervals;
s8, transmitting the latest operation data to a central data warehouse in real time, marking the latest operation data with a time stamp, continuing the model to perform a time sequence analysis method, and updating key features regularly;
s9, adopting an online learning algorithm increment version of the decision tree to enable the model to learn and self-adjust based on the latest data, and adjusting the splitting criterion and pruning strategy of the decision tree;
s10, regularly calculating the accuracy, recall rate and F1 score of the decision tree model;
s11, comparing the current performance index with preset reference or historical performance data, and identifying the change trend of the model performance;
and S12, periodically collecting user feedback including accuracy and practicability evaluation of model prediction, and carrying out fine adjustment of the model by combining the user feedback and performance monitoring results, modifying a decision threshold or introducing new data features.
Optionally, the S2 specifically includes:
s21, regarding transformer substation data sets In relation to temperature->Pressure->Current->Record of operations->Historical maintenance dataExternal environmental information->And weather conditions->Data of->Calculate each data +.>Mean->And standard deviation->For each data point->Calculate Z-Score value:
；
s22, for each dataDefining a specific threshold for identifying and eliminating outliers of corresponding data dimensionsFor some data->When->The data point is regarded as an outlier;
s23, regarding the data points identified as abnormalTaking an average of the data points from which the anomaly was removed or replaced with neighboring data points of the dimension;
s24, repeating S21-S23 until the substation data setMiddle temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->No more abnormal values exceeding the respective threshold value exist for the data of (a), the processed data set is defined as +.>；
S25 forIs +.>Calculate its minimum +.>Maximum->；
S26, applying Min-Max standardization to each data point：
；
Wherein,is the data dimension +.>Normalized data points in the middle, the normalized data set after processing is defined as。
Optionally, the step S4 specifically includes:
S41, data setIs +.>Definition of +.>Pressure->Current->Operation recordHistorical maintenance data->External environmental information->And weather conditions->Privacy budget for each data type of (2)>And calculates the corresponding data type +.>Sensitivity reflecting maximum range of variation of substation operation data in differential privacy processing>；
S42, each data point：
；
Wherein,represents->Data points after desensitization in dimensions;
s43, finally generating a temperaturePressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Desensitization dataset->。
Optionally, the step S5 specifically includesAn autoregressive integrated moving average model for analyzing and predicting the operating state of the substation equipment is constructed, and trend, seasonal and random components are extracted from the time series data:
；
wherein,represents time t>Predictive value of type data->、And->Constant term, autoregressive coefficient and moving average coefficient of model respectively, < >>、And->The long-term trend, periodic variation of seasonal components, and irregular fluctuation of random components, which are trend components of the disclosure data, are coordinated.
Optionally, the step S6 specifically includes:
s61, constructing a decision tree model based on key features extracted by time sequence analysis, wherein the decision tree model uses data pointsAs input->Representative dataset +.>Trend, seasonal and random components extracted in the time series analysis:
；
wherein,parameters representing the decision tree, including split point and threshold value +.>Is based on data set->For evaluating whether a specific condition is met;
s62, each internal node of the decision tree represents a test condition:
；
wherein,、and->For determining the current temperature->Pressure->And current->Whether the set corresponding thresholds t, p and i are exceeded;
s63, each branch of the decision tree represents the result of the test condition and points to a further node or a final leaf node if、And->If a certain item exceeds a threshold value, the path points to a fault class, otherwise, the path points to a normal class;
s64, each leaf node of the decision tree represents a fault or a normal state based on the data pointsAnd a predefined fault criterion, meaning that the substation is at risk of a fault according to the current analysis if the path of the decision tree points to the fault.
Optionally, the step S8 specifically includes:
S81, defining data acquisition frequency f, and in each acquisition periodIn, for all collection about temperature +.>Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Data points>Adding a timestamp mark;
s82, data points with time stampsThe updated data set is formed by outlier detection, min-Max standardization and differential privacy technology processing>；
S83, useUpdating a comprehensive time series model which comprehensively considers all data points and time stamps thereof:
；
s84, according to the decision tree modelUpdating the structure and the condition correspondingly;
s85, continuously updating a time sequence analysis model of key operation parameters of the substation equipment according to the latest data collected by the high-frequency acquisition module;
s86, for each running parameter, the autoregressive integrated moving average model periodically uses the newly acquired dataTo update the autoregressive coefficients in the model parameters +.>Moving average coefficient->And model constant term->：
。
Optionally, the step S9 specifically includes:
s91, applying an incremental learning version of the decision tree model to process a new data set in real timeThe latest operation parameter data of the transformer substation are collected from the real-time data collection module;
Wherein, the time series characteristic: time-dependent features such as trends, seasonal patterns are extracted from the temperature, pressure and current data. Maintenance and operation features: key features such as maintenance frequency, operational anomalies are extracted from historical repair and operational records. Environmental characteristics: the effect of weather and external environment on the performance of the device is considered.
S92, the decision tree model is based on the new data setDynamically updating, wherein the updating rule is based on a time sequence analysis result:
；
；
wherein,is data point +.>Middle temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->The amount of change at successive time points, +.>Is a preset update threshold;
s93, the update operation includes splitting and pruning of the nodes for adapting to the new data pattern.
Optionally, the step S10 specifically includes:
s101, periodically using the slave data setAnd evaluating the performance of the model with the updated model;
s102, calculating accuracy and data variationAnd (3) associating:
；
wherein each data pointRepresentative includes temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Is a different data type of (a);
S103, the recall rate is related to updating of the decision tree model:
；
the recall reflects the capability of the model to identify the actual faults, and changes of various operation parameters are considered;
s104, F1 score combines accuracy and recall, and is closely related to a model updating mechanism:
；
wherein each ofThe calculation is as follows:
；
s105, according to the accuracy, recall and F1 score results, adjusting the splitting criterion and pruning strategy of the decision tree and updating the parameters of the time sequence model, and carrying out real-time fault prediction of the transformer substation.
Optionally, the step S11 specifically includes:
s111, periodically evaluating and comparing indexes of the accuracy rate, the recall rate and the F1 score in the current model with preset reference or historical performance data;
s112, in the change trend, a mathematical formula is used for quantifying the change of the performance index and the performance changeExpressed as the difference between the current index value and the historical index value:
；
wherein,accuracy, recall and F1 score;
s113, identifying the improvement or reduction of the model performance and the field possibly needing further adjustment by analyzing the accuracy, recall and F1 score change trend;
s114, if the recall rate is reduced, indicating that the decision tree model needs to be further adjusted or the time sequence analysis needs to be updated to reflect the latest trend and mode;
If the accuracy rate is reduced, indicating that the overall performance of the model is reduced in distinguishing between the fault and non-fault states of the substation, properly processing the abnormal values by checking the data collection and integration steps, properly selecting and standardizing the characteristics, and re-evaluating the initial model construction process, including whether the training data set needs to be adjusted, the model parameters need to be changed or different algorithms need to be tried;
if F1 score is reduced, the defect of the model on the recognition capability of the fault state is indicated, and the model can not only recognize the fault but also correctly distinguish the non-fault condition by optimizing the splitting criterion and pruning strategy of the decision tree model.
Optionally, the step 12 specifically includes:
s121, user feedback from substation operators and maintenance teams is collected regularly, and accuracy and practicability evaluation of model prediction are focused on;
s122, combining user feedback with performance monitoring results of accuracy, recall and F1 score to comprehensively evaluate the performance and practicability of the model;
s123, based on the collected feedback and performance data, performing fine adjustment for modifying decision threshold or introducing new data features on the model, wherein a specific adjustment formula is expressed as an adjusted performance index Coefficients associated with user feedback +.>Is the product of:
；
wherein,is based on the userFeedback quantization score, representing improvement or decrease of model prediction accuracy or practicality, ++>Is an adjustment coefficient for controlling the degree of influence of user feedback on the model performance adjustment;
the adjustment process includes modifying splitting criteria or time series model parameters of the decision tree model;
s124, the effect of model adjustment is evaluated by recalculating the accuracy, recall and F1 score.
The beneficial effects of the invention are as follows:
the invention can more accurately predict the equipment faults by comprehensively utilizing the operation data collected in real time, and improves the accuracy and timeliness of fault prediction by combining time sequence analysis and a machine learning model. Meanwhile, the self-evolution learning mechanism enables the model to learn and self-adjust based on the latest data, and the model can adapt to operation changes such as ageing, seasonal changes, weather influences and the like of substation equipment, so that the prediction performance of the model is kept.
According to the invention, the accuracy, recall rate and F1 score of the model are regularly calculated, and the performance indexes are compared with the historical data, so that the degradation of the performance of the model can be timely identified and solved, the prediction model is ensured to always maintain an optimal state, and meanwhile, the model is regularly collected and adjusted by combining the feedback, so that the model is ensured to be not only theoretically effective, but also has high practicability and reliability in actual operation. This enhances the consistency between the model and the actual operation.
The invention can reduce the downtime caused by equipment faults, and enables a maintenance team to plan and execute maintenance tasks more effectively by predicting and identifying potential problems in time, thereby improving the overall operation and maintenance efficiency.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
fig. 1 is a flowchart of a self-evolving substation equipment learning method according to the present invention;
fig. 2 is a flowchart of initial model construction in a self-evolving substation equipment learning method provided by the invention;
fig. 3 is a flowchart of a performance evaluation model in a self-evolving substation equipment learning method provided by the invention.
Description of the embodiments
The invention will now be described in further detail with reference to the accompanying drawings. The drawings are simplified schematic representations which merely illustrate the basic structure of the invention and therefore show only the structures which are relevant to the invention.
Referring to fig. 1-3, a self-evolving substation equipment learning method includes the steps of:
s1, collecting temperature from a transformer substation by using a plurality of sensors Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->And build up a substation data set；
S2, collecting data setsPerforming outlier processing by using the Z-Score, and adopting Min-Max standardization to process data in a data set;
in this embodiment, S2 specifically includes:
s21, regarding transformer substation data setsIn relation to temperature->Pressure->Current->Record of operations->Historical maintenance dataExternal environmental information->And weather conditions->Data of->Calculate each data +.>Mean->And standard deviation->For each data point->Calculation of Z-Score value:
；
s22, for each dataDefining a specific threshold for identifying and eliminating outliers of corresponding data dimensionsFor some data->When->The data point is regarded as an outlier;
s23, regarding the data points identified as abnormalTaking an average of the data points from which the anomaly was removed or replaced with neighboring data points of the dimension;
s24, repeating S21-S23 until the substation data setMiddle temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->No more abnormal values exceeding the respective threshold value exist for the data of (a), the processed data set is defined as +. >The consistency and reliability of the data are ensured;
the multi-dimensional data, including the key operation parameter temperature, relied on by the substation equipment learning method can be ensured through the steps S21-S24Pressure->Current->And other important information operation records->Historical maintenance dataExternal environmental information->And weather conditions->Proper outlier processing is carried out before analysis and modeling, so that the quality of data and the accuracy of a model are improved.
S25 forIs +.>Calculate its minimum +.>Maximum->；
S26, applying Min-Max standardization to each data point：
；
Wherein,is the data dimension +.>Normalized data points in the middle, the normalized data set after processing is defined as，For training of self-evolving learning models.
According to the method, through the step S25-the step S26, the multi-dimensional data collected by the transformer substation is ensured to keep the data quality before the subsequent self-evolution learning model training, and meanwhile, the data privacy is effectively protected.
S3, processing the high-speed data stream by adopting a time window method, and collecting data in a fixed time interval;
s4, adding random noise in the data set, and performing data desensitization by adopting a differential privacy technology;
In this embodiment, S4 specifically includes:
s41, data setIs +.>Definition of +.>Pressure->Current->Operation recordHistorical maintenance data->External environmental information->And weather conditions->Privacy budget for each data type of (2)>And calculates the corresponding data type +.>Sensitivity reflecting maximum range of variation of substation operation data in differential privacy processing>；
S42, each data point：
；
Wherein,represents->Data points after desensitization in dimensions;
s43, finally generating a temperaturePressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Desensitization dataset->The method and the device keep the integral statistical property of the data while protecting the privacy of the data, and are suitable for training of a self-evolution learning model.
The present embodiment considersThe characteristics and the importance of each data dimension ensure that effective learning and prediction are performed, and meanwhile, possible privacy leakage is prevented, so that the method is particularly suitable for transformer substation data containing sensitive information, the privacy protection of the key data of the transformer substation in the self-evolution learning process is ensured, meanwhile, the effectiveness and the reliability of the data are maintained, and a solid data base is provided for the efficient operation and fault prediction of transformer substation equipment.
S5, performing time sequence analysis on equipment operation parameters and external parameters by adopting an autoregressive integrated moving average model, and extracting trend, seasonal and random components;
in the present embodiment, S5 specifically includes, forAn autoregressive integrated moving average model for analyzing and predicting the operating state of the substation equipment is constructed, and trend, seasonal and random components are extracted from the time series data:
；
wherein,represents time t>Predictive value of type data->、And->Constant term, autoregressive coefficient and moving average coefficient of model respectively, < >>、And->The long-term trend, periodic variation of seasonal components, and irregular fluctuation of random components, which cooperate as trend components of the revealed data, reflect contributions of different data types, such as environmental information and historical maintenance records, to model predictions.
ModelThe self-evolution learning model of the transformer substation can process and analyze different types of data, such as physical measured values of temperature, pressure, current and the like, operation records, historical maintenance data and other types of data, simultaneously can help predict and understand behaviors of transformer substation equipment under different conditions, such as temperature rise or current fluctuation can indicate impending equipment faults, and maintenance records and environmental factors, such as weather changes, can influence the performance and maintenance requirements of the equipment, so that the self-evolution learning model can accurately predict future states according to the historical and current data, deep data driving insight is provided for the transformer substation, preventive measures or optimized maintenance plans can be adopted in advance by the transformer substation, unexpected shutdown and fault risks are reduced, and the reliability and efficiency of the whole system are improved.
S6, performing fault prediction by using a decision tree model, and predicting whether the equipment is likely to be in fault or not based on the characteristics extracted from the time sequence analysis;
in this embodiment, S6 specifically includes:
s61, constructing a decision tree model based on key features extracted by time sequence analysis, wherein the decision tree model uses data pointsAs input->Representative dataset +.>Trend, seasonal and random components extracted in the time series analysis:
；
wherein,parameters representing the decision tree, including split point and threshold value +.>Is based on data set->For evaluating whether a specific condition is met;
s62, each internal node of the decision tree represents a test condition:
；
wherein,、and->For determining the current temperature->Pressure->And current->Whether the set corresponding thresholds t, p and i are exceeded;
s63, each branch of the decision tree represents the result of the test condition and points to a further node or a final leaf node if、And->Something exceedingA threshold value, the path points to the fault class, otherwise, the path points to the normal class;
s64, each leaf node of the decision tree represents a fault or a normal state based on the data pointsAnd a predefined fault criterion, meaning that the substation is at risk of a fault according to the current analysis if the path of the decision tree points to the fault.
According to the method, new operation data and user feedback are continuously received, and the model periodically updates the structure and the splitting criterion of the decision tree to adapt to the change of the operation environment of the transformer substation, so that the accuracy and the efficiency of fault prediction are improved.
S7, configuring a high-frequency acquisition module, and collecting the latest operation data from the transformer substation sensor at fixed time intervals;
s8, transmitting the latest operation data to a central data warehouse in real time, marking the latest operation data with a time stamp, continuing the model to perform a time sequence analysis method, and updating key features regularly;
in this embodiment, S8 specifically includes:
s81, defining data acquisition frequency f, and in each acquisition periodIn, for all collection about temperature +.>Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Data points>Adding a timestamp mark;
s82, data points with time stampsThe updated data set is formed by outlier detection, min-Max standardization and differential privacy technology processing>；
S83, useUpdating a comprehensive time series model which comprehensively considers all data points and time stamps thereof:
；
S84, according to the decision tree modelCorrespondingly updating the structure and the condition of the device to better reflect the latest fault prediction and the device state;
according to the method, the model can be self-adjusted and optimized according to the latest operation data and time sequence characteristics of the transformer substation through the steps S81-S84, the self-evolution learning model of the transformer substation can be used for learning and predicting by effectively utilizing the latest time sequence coherent data, and the accuracy and efficiency of monitoring and fault prediction of transformer substation equipment are improved. By associating each data point with its corresponding timestamp, the model is able to more accurately track and analyze changes in device status over time, thereby more effectively predicting potential faults and guiding maintenance decisions.
S85, continuously updating a time sequence analysis model of key operation parameters of the substation equipment according to the latest data collected by the high-frequency acquisition module;
s86, for each running parameter, the autoregressive integrated moving average model periodically uses the newly acquired dataTo update the autoregressive coefficients in the model parameters +.>Moving average coefficient->And model constant term->：
。
According to the method, the continuous updating mechanism of the step S85 and the step S86 can adapt to the dynamic changes of the environment and the equipment performance of the transformer substation, real-time data can be processed and analyzed, the prediction mechanism of the transformer substation can be dynamically adjusted according to the latest data trend, the transformer substation can quickly respond to the changes of the equipment state, the maintenance decision is optimized, the running efficiency and the reliability of the whole power grid are finally improved, faults can be prevented better by continuously updating key features, the downtime is reduced, and therefore the running efficiency and the service life of the equipment are improved.
S9, adopting an online learning algorithm increment version of the decision tree to enable the model to learn and self-adjust based on the latest data, and adjusting the splitting criterion and pruning strategy of the decision tree;
in this embodiment, S9 specifically includes:
s91, applying an incremental learning version of the decision tree model to process a new data set in real timeThe latest operation parameter data of the transformer substation are collected from the real-time data collection module;
s92, the decision tree model is based on the new data setDynamically updating, wherein the updating rule is based on a time sequence analysis result:
；/>
；
wherein,is data point +.>Middle temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->The amount of change at successive time points, +.>Is presetUpdating a threshold value;
s93, the update operation includes splitting and pruning of nodes to accommodate new data patterns, for example, if the amount of change in temperature exceeds a threshold, which may mean that the device temperature is abnormally rising, the decision tree needs to be adjusted to better predict potential failures.
Each update of the decision tree also considers the latest key features of the time series model, ensures that the decision tree structure and conditions reflect the latest running state of the device, and in the update process, the model follows a differential privacy technology to ensure the privacy of data processing and model update.
The self-evolution learning model of the transformer substation can adapt to data changes in real time, effectively predicts and deals with equipment faults, for example, a decision tree node updated according to current and temperature data can accurately predict the possibility of faults, and a decision tree node updated according to weather conditions and external environment information can accurately predict equipment fault risks caused by extreme weather.
Through continuous updating and optimizing of the model, the transformer substation can respond to the change of the state of the equipment in time, and equipment faults and unexpected shutdown are reduced.
S10, regularly calculating the accuracy, recall rate and F1 score of the decision tree model;
in this embodiment, S10 specifically includes:
s101, periodically using the slave data setAnd evaluating the performance of the model with the updated model;
s102, calculating accuracy and data variationAnd (3) associating:
；
wherein each data pointRepresentative includes temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Is a different data type of (a);
s103, the recall rate is related to updating of the decision tree model:
；
the recall reflects the capability of the model to identify the actual faults, and changes of various operation parameters are considered;
S104, F1 score combines accuracy and recall, and is closely related to a model updating mechanism:
；
wherein each ofThe calculation is as follows:
；
s105, according to the accuracy, recall and F1 score results, adjusting the splitting criterion and pruning strategy of the decision tree and updating the parameters of the time sequence model, and carrying out real-time fault prediction of the transformer substation.
The method ensures that the self-evolution learning model of the transformer substation can accurately reflect the latest equipment state and running conditions, and the transformer substation can timely discover and correct the defects in the model and keep the prediction capability and the actual running environment synchronous by periodically calculating and monitoring the key performance indexes.
S11, comparing the current performance index with preset reference or historical performance data, and identifying the change trend of the model performance;
in this embodiment, S11 specifically includes:
s111, periodically evaluating and comparing indexes of the accuracy rate, the recall rate and the F1 score in the current model with preset reference or historical performance data;
s112, in the change trend, a mathematical formula is used for quantifying the change of the performance index and the performance changeExpressed as the difference between the current index value and the historical index value:
；
wherein,accuracy, recall and F1 score;
S113, identifying the improvement or reduction of the model performance and the field possibly needing further adjustment by analyzing the accuracy, recall and F1 score change trend;
s114, if the recall rate is reduced, indicating that the decision tree model needs to be further adjusted or the time sequence analysis needs to be updated to reflect the latest trend and mode;
when the accuracy drops significantly, this indicates an overall performance degradation of the model in distinguishing between positive and negative classes (i.e., fault and non-fault conditions). In this case, the following needs to be reviewed:
data preprocessing and feature selection: checking the data collection and integration steps, ensuring that the data is accurate, the outliers are processed correctly, and the features are properly selected and standardized.
Training process of model: the model build process is reinitialized, including whether the training dataset needs to be adjusted, the model parameters changed, or a different algorithm tried.
A significant drop in F1 score indicates that the model has a problem in the balance of accuracy and recall. This typically involves the ability of the model to identify a minority class (e.g., a fault condition). In this case, a re-inspection is required:
constructing and optimizing a decision tree model: in particular, the splitting criteria and pruning strategy of the decision tree model, so as to ensure that the model not only can identify faults, but also can correctly distinguish non-fault conditions.
Unbalanced data processing: considering whether the distribution of the classes in the dataset is balanced, resampling is taken or class-specific weights are used as necessary to adjust the model's attention to the different classes.
The transformer substation can continuously optimize the prediction model, keep the consistency with the state of actual equipment, and the continuous performance evaluation and optimization strategy based on the data is beneficial to maintaining the long-term stable operation of the transformer substation.
And S12, periodically collecting user feedback including accuracy and practicability evaluation of model prediction, and carrying out fine adjustment of the model by combining the user feedback and performance monitoring results, modifying a decision threshold or introducing new data features.
In this embodiment, step 12 specifically includes:
s121, user feedback from substation operators and maintenance teams is collected regularly, and accuracy and practicability evaluation of model prediction are focused on;
s122, combining user feedback with performance monitoring results of accuracy, recall and F1 score to comprehensively evaluate the performance and practicability of the model;
s123, modifying the decision threshold for the model based on the collected feedback and performance dataOr introducing new fine adjustment of data characteristics, the specific adjustment formula is expressed as an adjusted performance index Coefficients associated with user feedback +.>Is the product of:
；
wherein,based on the quantitative score of the user feedback, representing the improvement or decrease of the model prediction accuracy or practicality, +.>Is an adjustment coefficient for controlling the degree of influence of user feedback on the model performance adjustment;
the adjustment process includes modifying splitting criteria or time series model parameters of the decision tree model;
s124, the effect of model adjustment is evaluated by recalculating the accuracy, the recall rate and the F1 score, so that the transformer substation can flexibly adjust the model thereof to adapt to the continuously-changing running conditions and the user demands, and better performance and usability are achieved in practical application.
According to the method and the system, the prediction model of the transformer substation can be ensured to continuously adapt to the actual operation requirement through the model adjustment strategy based on the user feedback and the performance data, and a dynamic and interactive mode is provided for the transformer substation to optimize the prediction capability of the transformer substation.
Examples:
in summer 2023 (6 months to 8 months), a self-evolving substation equipment learning method is adopted to improve the efficiency and accuracy of equipment monitoring in a certain large-scale substation located in the eastern China. The main challenges faced by this substation are equipment aging and environmental factor changes (often facing weather conditions with large high temperature and humidity changes), which lead to an increase in equipment failure rate and an increase in prediction difficulty.
To address these issues, workers implement the following strategies:
the staff installs the key parameter such as high accuracy sensor real-time supervision temperature, pressure, electric current to data collection and integration have been carried out in three months:
| month of month | Data volume | Data type |
| 6 months of | 0.91TB | Temperature, pressure, current, operational records, historical maintenance data, environmental information, weather conditions |
| 7 months of | 1.63TB | Temperature, pressure, current, operational records, historical maintenance data, environmental information, weather conditions |
| 8 months of | 0.92TB | Temperature, pressure, current, operational records, historical maintenance data, environmental information, weather conditions |
Table 1 data collection and integration
As can be seen from table 1 above, over three months, approximately 2.46TB of data was collected, including equipment operational data, maintenance history, and weather-related environmental data. The time series data of temperature and current were analyzed using an autoregressive integrated moving average model, and some periodic wave patterns were found. Through the decision tree model, workers can predict potential equipment faults according to historical fault data and real-time parameters.
The staff is provided with an automatic data acquisition system, the latest operation data is collected every one hour, and the model is continuously updated through an online learning algorithm. Throughout the summer, the model underwent multiple iterations and optimizations to accommodate environmental changes and fluctuations in device status, while weekly evaluations were made of model accuracy, recall, and F1 score:
TABLE 2 Performance monitoring and optimization before and after use
| Performance index | 6 months old (reference) | 7 middle ten days of month (after improvement) | 8 months bottom (end) |
| Accuracy rate (ACC) | 85.2% | 95.12% | 98.34% |
| Recall (Recall) | 88.33% | 993.32% | 95.89% |
| F1 Score of | 86.16% | 94.65% | 95.92% |
As can be seen from the above Table 2, in the middle of 7 months, the accuracy rate is improved to 95.12%, the recall rate is 93.32%, and the F1 score reaches 94.65%. The accuracy is improved to 98.34% at the end of 8 months, the recall rate is 95.89%, the F1 score reaches 95.92%, and by comparing the performance indexes with the reference data at the beginning of 6 months, the performance of the model is obviously improved by workers.
Meanwhile, staff periodically collect feedback from substation operators, and particularly, the model prediction accuracy and the practicability are evaluated. At the end of 7 months, a series of valuable advice and observations were collected through a detailed user feedback meeting. Based on these feedback, the staff adjusts some parameters in the decision tree model and adds new data features such as grid load changes over a specific period of time.
In three months of implementation, the model successfully predicts the failure of one transformer on a high-temperature and high-humidity weekend (on the day of 7 months), and the potential failure of the cooling system of the primary transformer at the beginning of 8 months. This prediction enables the maintenance team to take action in advance, avoiding long downtime and potential loss. Compared with 6 months, the equipment failure rate of 8 months is reduced by about 71%, and the effect of model optimization is obviously demonstrated. This reduction not only means less maintenance costs, but also means a significant improvement in the stability and reliability of the overall operation. And meanwhile, according to the model after the user feedback adjustment, higher robustness is shown when data under extreme weather conditions are processed. In a thunderstorm weather of 8 months, the model successfully predicts and avoids multiple equipment failures caused by voltage fluctuations.
By implementing the self-evolving substation equipment learning method, the substation achieves remarkable results in high-efficiency fault prediction, equipment maintenance and environmental change coping. The integration of the user feedback enables the model to be more fit with actual operation requirements, and the practicality of the model and the satisfaction degree of the operation and maintenance team are improved. In general, the method has obvious advantages in the aspects of improving the prediction accuracy, reducing the failure rate, reducing the maintenance cost, improving the operation and maintenance efficiency and the like, and provides powerful technical support for the development of the intelligent power grid.
The invention can more accurately predict the equipment faults by comprehensively utilizing the operation data collected in real time, and improves the accuracy and timeliness of fault prediction by combining time sequence analysis and a machine learning model. Meanwhile, the self-evolution learning mechanism enables the model to learn and self-adjust based on the latest data, and the model can adapt to operation changes such as ageing, seasonal changes, weather influences and the like of substation equipment, so that the prediction performance of the model is kept.
According to the invention, the accuracy, recall rate and F1 score of the model are regularly calculated, and the performance indexes are compared with the historical data, so that the degradation of the performance of the model can be timely identified and solved, the prediction model is ensured to always maintain an optimal state, and meanwhile, the model is regularly collected and adjusted by combining the feedback, so that the model is ensured to be not only theoretically effective, but also has high practicability and reliability in actual operation. This enhances the consistency between the model and the actual operation.
The invention can reduce the downtime caused by equipment faults, and enables a maintenance team to plan and execute maintenance tasks more effectively by predicting and identifying potential problems in time, thereby improving the overall operation and maintenance efficiency.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.
Claims (10)
1. The self-evolving substation equipment learning method is characterized by comprising the following steps of:
s1, makingCollecting temperature from a substation with multiple sensorsPressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->And establishes a substation data set +.>；
S2, collecting data setsPerforming outlier processing by using the Z-Score, and adopting Min-Max standardization to process data in a data set;
s3, processing the high-speed data stream by adopting a time window method, and collecting data in a fixed time interval;
s4, adding random noise in the data set, and performing data desensitization by adopting a differential privacy technology;
S5, performing time sequence analysis on equipment operation parameters and external parameters by adopting an autoregressive integrated moving average model, and extracting trend, seasonal and random components;
s6, performing fault prediction by using a decision tree model, and predicting whether the equipment is likely to be in fault or not based on the characteristics extracted from the time sequence analysis;
s7, configuring a high-frequency acquisition module, and collecting the latest operation data from the transformer substation sensor at fixed time intervals;
s8, transmitting the latest operation data to a central data warehouse in real time, marking the latest operation data with a time stamp, continuing the model to perform a time sequence analysis method, and updating key features regularly;
s9, adopting an online learning algorithm increment version of the decision tree to enable the model to learn and self-adjust based on the latest data, and adjusting the splitting criterion and pruning strategy of the decision tree;
s10, regularly calculating the accuracy, recall rate and F1 score of the decision tree model;
s11, comparing the current performance index with preset reference or historical performance data, and identifying the change trend of the model performance;
and S12, periodically collecting user feedback including accuracy and practicability evaluation of model prediction, and carrying out fine adjustment of the model by combining the user feedback and performance monitoring results, modifying a decision threshold or introducing new data features.
2. The self-evolving substation equipment learning method according to claim 1, wherein S2 specifically comprises:
s21, regarding transformer substation data setsIn relation to temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Data of->Calculate each data +.>Mean->And standard deviation->For each data point->Calculate Z-Score value:
；
s22, for each dataDefining a specific threshold for identifying and eliminating outliers of corresponding data dimensionsFor some data->When->The data point is regarded as an outlier;
s23, regarding the data points identified as abnormalTaking an average of the data points from which the anomaly was removed or replaced with neighboring data points of the dimension;
s24, repeating S21-S23 until the substation data setMiddle temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->No more abnormal values exceeding the respective threshold value exist for the data of (a), the processed data set is defined as +.>；
S25 forIs +.>Calculate its minimum +.>Maximum->；
S26, applying Min-Max standardization to each data point ：
；
Wherein,is the data dimension +.>Normalized data points in (1), the normalized data set after processing is defined as +.>。
3. The self-evolving substation equipment learning method according to claim 2, wherein S4 specifically comprises:
s41, data setIs +.>Definition of +.>Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Privacy budget for each data type of (2)>And calculates the corresponding data type +.>Sensitivity reflecting maximum range of variation of substation operation data in differential privacy processing>；
S42, each data point：
；
Wherein,represents->Data points after desensitization in dimensions;
s43, finally generating a temperaturePressure->Current->Record of operations->Historical maintenance data->External environment informationAnd weather conditions->Desensitization dataset->。
4. A self-evolving substation equipment learning method according to claim 3, characterized in that S5 comprises in particular forAn autoregressive integrated moving average model for analyzing and predicting the operating state of the substation equipment is constructed, and trend, seasonal and random components are extracted from the time series data:
；
Wherein,represents time t>Predictive value of type data->、And->Constant term, autoregressive coefficient and moving average coefficient of model respectively, < >>、And->The long-term trend, periodic variation of seasonal components, and irregular fluctuation of random components, which are trend components of the disclosure data, are coordinated.
5. The self-evolving substation equipment learning method according to claim 4, wherein S6 specifically comprises:
s61, constructing a decision tree model based on key features extracted by time sequence analysis, wherein the decision tree model uses data pointsAs input->Representative dataset +.>Trend, seasonal and random components extracted in the time series analysis:
；
wherein,parameters representing the decision tree, including split point and threshold value +.>Is based on data set->For evaluating whether a specific condition is met;
s62, each internal node of the decision tree represents a test condition:
；
wherein,、and->For determining the current temperature->Pressure->And current->Whether the set corresponding thresholds t, p and i are exceeded;
s63, each branch of the decision tree represents the result of the test condition and points to a further node or a final leaf node if 、And->If a certain item exceeds a threshold value, the path points to a fault class, otherwise, the path points to a normal class;
s64, each leaf node of the decision tree represents a fault or a normal state based on the data pointsAnd a predefined fault criterion, meaning that the substation is at risk of a fault according to the current analysis if the path of the decision tree points to the fault.
6. The self-evolving substation equipment learning method according to claim 5, wherein S8 specifically comprises:
s81, defining data acquisition frequency f, and in each acquisition periodIn, for all collection about temperature +.>Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Data points>Adding a timestamp mark;
s82, data points with time stampsThe updated data set is formed by outlier detection, min-Max standardization and differential privacy technology processing>；
S83, useUpdating a comprehensive time series model which comprehensively considers all data points and time stamps thereof:
；
s84, according to the decision tree modelUpdating the structure and the condition correspondingly;
s85, continuously updating a time sequence analysis model of key operation parameters of the substation equipment according to the latest data collected by the high-frequency acquisition module;
S86, for each running parameter, the autoregressive integrated moving average model periodically uses the newly acquired dataTo update the autoregressive coefficients in the model parameters +.>Moving average coefficient->And model constant term->：
。
7. The self-evolving substation equipment learning method according to claim 1, wherein S9 specifically comprises:
s91, applying an incremental learning version of the decision tree model to process a new data set in real timeThe latest operation parameter data of the transformer substation are collected from the real-time data collection module;
s92, the decision tree model is based on the new data setDynamically updating, wherein the updating rule is based on a time sequence analysis result:
；
；
wherein,is data point +.>Middle temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->The amount of change at successive time points, +.>Is a preset update threshold;
s93, the update operation includes splitting and pruning of the nodes for adapting to the new data pattern.
8. The self-evolving substation equipment learning method according to claim 7, wherein S10 specifically comprises:
s101, periodically using the slave data setAnd evaluating the performance of the model with the updated model;
S102, calculating accuracy and data variationAnd (3) associating:
；
wherein each data pointRepresentative includes temperature->Pressure->Current->Record of operations->Historical maintenance data->External environmental information->And weather conditions->Is a different data type of (a);
s103, the recall rate is related to updating of the decision tree model:
；
the recall reflects the capability of the model to identify the actual faults, and changes of various operation parameters are considered;
s104, F1 score combines accuracy and recall, and is closely related to a model updating mechanism:
；
wherein each ofThe calculation is as follows:
；
s105, according to the accuracy, recall and F1 score results, adjusting the splitting criterion and pruning strategy of the decision tree and updating the parameters of the time sequence model, and carrying out real-time fault prediction of the transformer substation.
9. The self-evolving substation equipment learning method according to claim 8, wherein S11 specifically comprises:
s111, periodically evaluating and comparing indexes of the accuracy rate, the recall rate and the F1 score in the current model with preset reference or historical performance data;
s112, in the change trend, a mathematical formula is used for quantifying the change of the performance index and the performance changeExpressed as the difference between the current index value and the historical index value:
；
Wherein,accuracy, recall and F1 score;
s113, identifying the improvement or reduction of the model performance and the field possibly needing further adjustment by analyzing the accuracy, recall and F1 score change trend;
s114, if the recall rate is reduced, indicating that the decision tree model needs to be further adjusted or the time sequence analysis needs to be updated to reflect the latest trend and mode;
if the accuracy rate is reduced, indicating that the overall performance of the model is reduced in distinguishing between the fault and non-fault states of the substation, properly processing the abnormal values by checking the data collection and integration steps, properly selecting and standardizing the characteristics, and re-evaluating the initial model construction process, including whether the training data set needs to be adjusted, the model parameters need to be changed or different algorithms need to be tried;
if F1 score is reduced, the defect of the model on the recognition capability of the fault state is indicated, and the model can not only recognize the fault but also correctly distinguish the non-fault condition by optimizing the splitting criterion and pruning strategy of the decision tree model.
10. The self-evolving substation equipment learning method according to claim 9, wherein the step 12 specifically includes:
s121, user feedback from substation operators and maintenance teams is collected regularly, and accuracy and practicability evaluation of model prediction are focused on;
S122, combining user feedback with performance monitoring results of accuracy, recall and F1 score to comprehensively evaluate the performance and practicability of the model;
s123, based on the collected feedback and performance data, performing fine adjustment for modifying decision threshold or introducing new data features on the model, wherein a specific adjustment formula is expressed as an adjusted performance indexAssociated with user feedbackCoefficient of->Is the product of:
；
wherein,based on the quantitative score of the user feedback, representing the improvement or decrease of the model prediction accuracy or practicality, +.>Is an adjustment coefficient for controlling the degree of influence of user feedback on the model performance adjustment;
the adjustment process includes modifying splitting criteria or time series model parameters of the decision tree model;
s124, the effect of model adjustment is evaluated by recalculating the accuracy, recall and F1 score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410127488.8A CN117648543A (en) | 2024-01-30 | 2024-01-30 | Self-evolving substation equipment learning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410127488.8A CN117648543A (en) | 2024-01-30 | 2024-01-30 | Self-evolving substation equipment learning method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117648543A true CN117648543A (en) | 2024-03-05 |
Family
ID=90046483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410127488.8A Pending CN117648543A (en) | 2024-01-30 | 2024-01-30 | Self-evolving substation equipment learning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117648543A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117841028A (en) * | 2024-03-08 | 2024-04-09 | 安徽国智数据技术有限公司 | Comprehensive pipe gallery inspection robot based on artificial intelligence |
| CN118157326A (en) * | 2024-05-09 | 2024-06-07 | 国能大渡河检修安装有限公司 | Method and device for adjusting running state of power station transformer, medium and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114817985A (en) * | 2022-04-22 | 2022-07-29 | 广东电网有限责任公司 | Privacy protection method, device, equipment and storage medium for electricity consumption data |
| CN116595584A (en) * | 2023-05-19 | 2023-08-15 | 西安体育学院 | Physical medicine data fusion privacy protection method based on cloud and fog architecture longitudinal federal learning |
| CN117216702A (en) * | 2023-09-19 | 2023-12-12 | 国网北京市电力公司 | Power transformation equipment parameter processing method and device and electronic equipment |
| CN117349782A (en) * | 2023-12-06 | 2024-01-05 | 湖南嘉创信息科技发展有限公司 | Intelligent data early warning decision tree analysis method and system |
| CN117390944A (en) * | 2023-11-28 | 2024-01-12 | 江苏阿尔发云信息技术有限公司 | Substation operation condition simulation system |
-
2024
- 2024-01-30 CN CN202410127488.8A patent/CN117648543A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114817985A (en) * | 2022-04-22 | 2022-07-29 | 广东电网有限责任公司 | Privacy protection method, device, equipment and storage medium for electricity consumption data |
| CN116595584A (en) * | 2023-05-19 | 2023-08-15 | 西安体育学院 | Physical medicine data fusion privacy protection method based on cloud and fog architecture longitudinal federal learning |
| CN117216702A (en) * | 2023-09-19 | 2023-12-12 | 国网北京市电力公司 | Power transformation equipment parameter processing method and device and electronic equipment |
| CN117390944A (en) * | 2023-11-28 | 2024-01-12 | 江苏阿尔发云信息技术有限公司 | Substation operation condition simulation system |
| CN117349782A (en) * | 2023-12-06 | 2024-01-05 | 湖南嘉创信息科技发展有限公司 | Intelligent data early warning decision tree analysis method and system |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117841028A (en) * | 2024-03-08 | 2024-04-09 | 安徽国智数据技术有限公司 | Comprehensive pipe gallery inspection robot based on artificial intelligence |
| CN117841028B (en) * | 2024-03-08 | 2024-05-24 | 安徽国智数据技术有限公司 | Comprehensive pipe gallery inspection robot based on artificial intelligence |
| CN118157326A (en) * | 2024-05-09 | 2024-06-07 | 国能大渡河检修安装有限公司 | Method and device for adjusting running state of power station transformer, medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117630758B (en) | Method and system for monitoring health state of power station transformer | |
| CN107862463B (en) | Method and control system for scheduling load of power plant | |
| US7409303B2 (en) | Identifying energy drivers in an energy management system | |
| CN117648543A (en) | Self-evolving substation equipment learning method | |
| CN117454093A (en) | Intelligent power module management method and system | |
| CN117670378B (en) | Food safety monitoring method and system based on big data | |
| CN117930815A (en) | Wind turbine generator remote fault diagnosis method and system based on cloud platform | |
| WO2020197533A1 (en) | Surrogate of a simulation engine for power system model calibration | |
| CN117878925B (en) | Power transmission data control method and system of smart power grid | |
| CN117391459B (en) | Electric power operation risk early warning method and system based on deep learning | |
| CN118432000B (en) | High-voltage isolation protection method and system | |
| CN117176560A (en) | Monitoring equipment supervision system and method based on Internet of things | |
| CN117408162A (en) | Power grid fault control method based on digital twin | |
| Ghasemi et al. | Equipment failure rate in electric power distribution networks: An overview of concepts, estimation, and modeling methods | |
| CN118017502A (en) | Digital twinning-based power distribution calculation power prediction method, system and medium | |
| CN117933447A (en) | Fault prediction method and system based on data analysis | |
| CN118395358B (en) | Intelligent anti-misoperation topology analysis method for transformer substation | |
| KR20220089853A (en) | Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology | |
| CN117494009A (en) | Electrical equipment state evaluation method based on insulating material pyrolysis analysis and cloud platform | |
| CN118449261A (en) | Intelligent detection management method and system for load of public transformer area | |
| CN117977536A (en) | Smart power grid load identification method | |
| CN117009741A (en) | Abnormal analysis method and system based on generated energy | |
| CN115936663A (en) | Maintenance method and device for power system | |
| Righetto et al. | Predictive maintenance 4.0: Concept, architecture and electrical power systems applications | |
| Bagchi et al. | Data analytics and stochastic modeling in a semiconductor fab |
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 |