CN117408162A - Power grid fault control method based on digital twin - Google Patents
Power grid fault control method based on digital twin Download PDFInfo
- Publication number
- CN117408162A CN117408162A CN202311514043.7A CN202311514043A CN117408162A CN 117408162 A CN117408162 A CN 117408162A CN 202311514043 A CN202311514043 A CN 202311514043A CN 117408162 A CN117408162 A CN 117408162A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- power grid
- digital twin
- fault
- 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 description 47
- 238000012544 monitoring process Methods 0.000 claims abstract description 17
- 238000012502 risk assessment Methods 0.000 claims abstract description 12
- 238000011217 control strategy Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 46
- 238000010801 machine learning Methods 0.000 claims description 45
- 238000013179 statistical model Methods 0.000 claims description 40
- 238000012423 maintenance Methods 0.000 claims description 30
- 230000008859 change Effects 0.000 claims description 20
- 238000012360 testing method Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 14
- 238000001514 detection method Methods 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 230000004044 response Effects 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 claims description 7
- 230000006399 behavior Effects 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000004088 simulation Methods 0.000 claims description 4
- 238000007619 statistical method Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 description 5
- 238000003745 diagnosis Methods 0.000 description 5
- 230000000737 periodic effect Effects 0.000 description 5
- 230000003449 preventive effect Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000007774 longterm Effects 0.000 description 3
- 238000013468 resource allocation Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000004580 weight loss Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000001483 mobilizing 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
- 238000011084 recovery Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
- G06F17/13—Differential equations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid distribution networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Geometry (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Hardware Design (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Water Supply & Treatment (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Computational Linguistics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
Abstract
By collecting real-time power grid operation data and equipment information, a digital twin model is constructed, and the model can accurately reflect the state and the operation condition of an actual power grid. And the sensor and the monitoring equipment are used for collecting parameters such as voltage, current, frequency and the like of each node of the power grid in real time, and feeding the data back to the digital twin model. In the digital twin model, potential fault hidden dangers such as equipment overload and unstable voltage and the like existing in the power grid are identified through analyzing real-time data. And carrying out risk assessment on the detected fault hidden danger, determining the influence degree of the fault hidden danger on the safety and stability of the power grid, and classifying the fault hidden danger into high, medium and low risk grades. And (5) formulating a corresponding control strategy according to the risk level of the hidden trouble. After the control measures are implemented, the state of the power grid is continuously monitored, and the real-time feedback of the control effect to the digital twin model is ensured.
Description
Technical Field
The invention relates to the field of data identification processing, in particular to a digital twinning-based power grid fault control method and a power grid.
Background
With the rapid increase of the power consumption load of the urban power distribution network and the increasing increase of the power reliability requirement, the load scale is gradually increased, and the topology structure of the power distribution network is also more and more complex; in addition, with the application of distributed energy access and distributed power supply, the operation complexity of the power distribution network is greatly improved. For the distribution network with a complex topological structure, if a grid fault occurs, the result is serious, even an electrical accident can be caused, and huge loss is brought to social life. The occurrence of power failure accidents in the power distribution network can never be avoided, the power failure accidents caused by weather factors, and the faults caused by trees lead to power failure of users and influence the reliability index of the power distribution network. However, many power outages are caused by recurrent defects or by gradual failure of the equipment, which may have occurred within weeks to months before the power outage caused by the final failure. As such, conventional power distribution network fault diagnosis techniques for handling faults have failed to accommodate the need for safe, economical, reliable operation of active power distribution networks.
To meet the above requirements, research on intelligent fault diagnosis and prediction technology of the power distribution network is urgently needed. The equipment in the distribution network may have a fault tendency due to some external factors under long-term operation, or some equipment may have some defects due to cumulative effects after being subjected to a plurality of disturbances or faults, although the equipment can still normally operate. Therefore, a failure prediction is required for these potential safety hazards. The fault prediction can be used as a starting point according to the running condition of the current equipment, and predict whether the future trend of faults occurs or not through the existing running environment conditions of the equipment, historical data and the like.
The fault diagnosis difficulty is increased due to factors such as power electronization of electric equipment in the active power distribution network, access of a large number of distributed power supplies, unidirectional to bidirectional conversion of tide and the like. Traditional methods based on current equipment operation conditions and historical data prediction are not suitable for active power distribution networks. A new method for analyzing faults of a power distribution network is needed to solve the problem of fault analysis of an active power distribution network.
Disclosure of Invention
The invention aims to solve the problem that a two-dimensional image acquired by an RGB camera in a dim light environment lacks detection precision, and provides a digital twin-based power grid fault control method and a power grid.
The invention adopts the following technical means for solving the technical problems:
the invention provides a digital twinning-based power grid fault control method, which comprises the following steps:
step S1: dividing a power grid into a plurality of units, installing a sensor in each unit, acquiring electrical data of each unit, performing calibration, denoising and standardization processing on the acquired electrical data, wherein the electrical data comprises physical data and historical change data, the physical data comprises voltage, current, power and frequency data, and the historical change data comprises historical performance and change trend of the power grid in each period of a statistical period;
step S2: establishing a physical model based on the physical data of the power grid, analyzing the historical change data of the power grid by using a statistical method to establish a statistical model, extracting the characteristics of the electrical data by a machine learning algorithm to establish a machine learning model, and fusing the physical model, the statistical model and the machine learning model to form a comprehensive digital twin model;
step S3: inputting the real-time data into a digital twin model, and identifying potential fault hidden danger in the power grid by analyzing the real-time data;
step S4: performing risk assessment on the detected fault hidden danger, determining the influence degree of the fault hidden danger on the safety and stability of the power grid, and classifying the fault hidden danger into high, medium and low risk grades; according to the risk level of the hidden trouble, a corresponding control strategy is formulated; for the hidden danger of high risk, immediate treatment measures are adopted; and for medium and low risk hidden dangers, selecting corresponding accurate monitoring and early warning schemes.
Further, in the step S2, the method for establishing the physical model is as follows:
defining model variables, specifically including voltage, current, power and frequency, establishing a basic equation according to the model variables, establishing a physical model through a differential equation, adjusting the physical model by utilizing collected electrical data to enable the output of the model to be consistent with actual data, solving the differential equation by using a numerical method, simulating specific behaviors of a power grid in a period of time, and verifying the established physical model by comparing a simulation result with the actual data.
Further, in the step S2, the method for establishing the statistical model is as follows:
according to the physical data, establishing statistical models respectively corresponding to the physical data;
obtaining physical data corresponding to various fault types, and determining parameter state data corresponding to the physical data according to a statistical model of an operation parameter, wherein the parameter state data records the normal state or abnormal state of various physical data;
for each fault type, a statistical model of the operating parameter status data corresponding thereto is established.
Further, in the step S2, the method for establishing the machine learning model is as follows:
performing feature classification, combination and reconstruction on the physical data based on the historical change data to obtain a general data set, dividing the general data set into a training set and a testing set, and training a machine learning model through the training set;
the machine learning model is subjected to repeated iterative training, error calculation is carried out by using a multi-weight loss function in each iterative training, a counter propagation algorithm is used, the weight and bias of the model are updated according to gradient information of the loss function, repeated iterative training is carried out until a preset stopping condition is reached, and finally the machine learning model with all parameter values determined is obtained.
Further, in the step S2, the method for establishing the digital twin model is as follows:
establishing a joint probability model, integrating a physical model, a statistical model and a machine learning model in the same framework, using a Bayesian network to represent the dependency relationship among the models, forming a joint probability graph model, creating a joint data set containing the output of the models and the actual power grid state, dividing the joint data set into a joint training set and a joint testing machine, using the joint training set to perform joint training on the whole joint probability model, using the joint testing set to perform verification and evaluation on the whole joint probability model, so as to establish a digital twin model, embedding the digital twin model into the actual power grid, receiving power grid feedback data in real time, and performing iterative updating on the model.
Further, the specific method for receiving the power grid feedback data in real time and iteratively updating the model comprises the following steps: and comparing the actual data with the data output by the digital twin model, judging whether the difference exceeds the level of random variation, wherein the difference comparison method is root mean square error or average absolute error, if the difference exceeds the level of random variation, judging that the digital twin model cannot be identified, and iterating the digital twin model.
Further, in the step S3, the data collected in real time is calibrated, the data is denoised and filtered by using a signal processing technology, interference and noise are eliminated, key features are extracted from the real-time data by using a machine learning technology, the extracted features are used as input, each component in the digital twin model including a physical model, a statistical model and a machine learning model is input, abnormal detection is performed by using the model, a mode inconsistent with normal operation is identified, whether the real-time data is normal operation data is judged, and if not, a specific fault type is detected.
Further, risk assessment is carried out on the detected hidden trouble, the influence degree of the hidden trouble on the safety and stability of the power grid is determined, and the hidden trouble is classified into high, medium and low risk grades; wherein, the hidden danger which can cause major faults, accidents of the power grid or affect large-area users is given a high risk level; for hidden dangers that cause problems but do not have a significant impact, a medium risk level is given; the hidden danger that the influence on the power grid is small and effective control can be achieved within a specified time is given to low risk level;
for the hidden danger with high risk, emergency measures are taken, an emergency response plan is started, emergency resources are mobilized, and faults are rapidly processed;
for the hidden danger with high risk, equipment maintenance and overhaul are carried out, and the change of the hidden danger in the wind is monitored in real time;
for the low risk hidden danger, regular maintenance is carried out, and the low risk hidden danger is monitored in real time.
The invention provides a digital twin-based power grid fault control method and a power grid, which have the following beneficial effects: the digital twin model can provide real-time monitoring and prediction of the state of the power grid through real-time data analysis. Therefore, operation and maintenance personnel can know the running condition of the power grid more timely, predict potential faults and take corresponding measures. The digital twin model can identify potential fault hazards in the power grid, including high, medium and low risk classifications. This helps to take preventive measures to improve the reliability and stability of the grid before the problem progresses to an emergency phase. Through risk assessment of fault hidden danger, different risk grades can be reasonably divided, and corresponding control strategies are formulated. This helps to optimize resource allocation, placing emphasis on high risk hidden trouble, improving the efficiency of fault handling. By accurate monitoring and prediction of the digital twin model, emergency maintenance costs due to sudden failures can be reduced. Meanwhile, periodic maintenance and preventive maintenance can be reasonably arranged, and unnecessary shutdown and maintenance are avoided. Through real-time monitoring and accurate risk assessment, the stable running state of the power grid can be better maintained. And the potential problems are identified in time, and a control strategy is adopted, so that the risk of instability of the power grid is reduced. The digital twin model is established and applied based on analysis of a large amount of real-time data and historical data, so that a power grid manager can be helped to make more scientific and data-driven decisions, and the accuracy of the decisions is improved. Through implementing digital twin model, can more effectively deal with the potential safety hazard in the electric wire netting, take measures in time in order to prevent accident occurrence, promote the holistic security of electric wire netting. The automated nature and real-time monitoring function of the digital twin model reduces the likelihood of human error. The operation and maintenance personnel can make decisions more quickly and accurately, and the fault risk caused by manual operation is reduced.
Drawings
Fig. 1 is a schematic flow chart of an embodiment of a digital twin-based power grid fault control method according to the present invention.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present invention, as the achievement, functional features, and advantages of the present invention are further described with reference to the embodiments, with reference to the accompanying drawings.
The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a digital twin-based power grid fault control method according to an embodiment of the present invention includes:
step S1: dividing a power grid into a plurality of units, installing a sensor in each unit, acquiring electrical data of each unit, performing calibration, denoising and standardization processing on the acquired electrical data, wherein the electrical data comprises physical data and historical change data, the physical data comprises voltage, current, power and frequency data, and the historical change data comprises historical performance and change trend of the power grid in each period of a statistical period;
step S2: establishing a physical model based on the physical data of the power grid, analyzing the historical change data of the power grid by using a statistical method to establish a statistical model, extracting the characteristics of the electrical data by a machine learning algorithm to establish a machine learning model, and fusing the physical model, the statistical model and the machine learning model to form a comprehensive digital twin model;
step S3: inputting the real-time data into a digital twin model, and identifying potential fault hidden danger in the power grid by analyzing the real-time data;
step S4: performing risk assessment on the detected fault hidden danger, determining the influence degree of the fault hidden danger on the safety and stability of the power grid, and classifying the fault hidden danger into high, medium and low risk grades; according to the risk level of the hidden trouble, a corresponding control strategy is formulated; for the hidden danger of high risk, immediate treatment measures are adopted; and for medium and low risk hidden dangers, selecting corresponding accurate monitoring and early warning schemes.
The digital twin model can provide real-time monitoring and prediction of the state of the power grid through real-time data analysis. Therefore, operation and maintenance personnel can know the running condition of the power grid more timely, predict potential faults and take corresponding measures. The digital twin model can identify potential fault hazards in the power grid, including high, medium and low risk classifications. This helps to take preventive measures to improve the reliability and stability of the grid before the problem progresses to an emergency phase. Through risk assessment of fault hidden danger, different risk grades can be reasonably divided, and corresponding control strategies are formulated. This helps to optimize resource allocation, placing emphasis on high risk hidden trouble, improving the efficiency of fault handling. By accurate monitoring and prediction of the digital twin model, emergency maintenance costs due to sudden failures can be reduced. Meanwhile, periodic maintenance and preventive maintenance can be reasonably arranged, and unnecessary shutdown and maintenance are avoided. Through real-time monitoring and accurate risk assessment, the stable running state of the power grid can be better maintained. And the potential problems are identified in time, and a control strategy is adopted, so that the risk of instability of the power grid is reduced. The digital twin model is established and applied based on analysis of a large amount of real-time data and historical data, so that a power grid manager can be helped to make more scientific and data-driven decisions, and the accuracy of the decisions is improved. Through implementing digital twin model, can more effectively deal with the potential safety hazard in the electric wire netting, take measures in time in order to prevent accident occurrence, promote the holistic security of electric wire netting. The automated nature and real-time monitoring function of the digital twin model reduces the likelihood of human error. The operation and maintenance personnel can make decisions more quickly and accurately, and the fault risk caused by manual operation is reduced.
In this embodiment, in the step S2, the method for establishing the physical model is as follows:
defining model variables, specifically including voltage, current, power and frequency, establishing a basic equation according to the model variables, establishing a physical model through a differential equation, adjusting the physical model by utilizing collected electrical data to enable the output of the model to be consistent with actual data, solving the differential equation by using a numerical method, simulating specific behaviors of a power grid in a period of time, and verifying the established physical model by comparing a simulation result with the actual data.
And determining the overall structure and the topological relation of the power grid, wherein the overall structure and the topological relation specifically comprise the connection and interaction relation of power stations, substations, transmission lines and distribution network components.
Modeling various devices in the power grid, including generators, transformers, lines, and switching devices; based on the physical equation of the power grid, the mathematical description of the power grid is established, and a model comprising a generator, a motor and a frequency modulator is established. And integrating real-time data of the actual power grid into the model, and carrying out parameter adjustment on the model by utilizing the actual operation data to ensure that the model reflects the actual behavior of the power grid. The accuracy of the model is verified using historical data or simulated test data. The model prediction results can be compared with actual operation data for evaluation.
By establishing a model based on a physical equation, the dynamic behavior of the power grid can be more accurately simulated, including the changes of parameters such as voltage, current, power, frequency and the like. The physical model building process involves considering the physical characteristics of various elements in the power grid, which is helpful for a deep understanding of the operation mechanism and dynamic response of the power grid. The physical model is adjusted by using the collected electrical data, so that the model can be better adapted to the characteristics of an actual power grid, and the accuracy of the model is improved. The physical model can capture characteristics of various equipment faults and grid faults. The digital twin model may simulate these fault scenarios with the physical model for fault detection, localization and diagnosis. By comparing with the actually collected data, the digital twin model can adjust and calibrate the parameters of the physical model, so that the digital twin model is better suitable for the characteristics of an actual power grid.
Further, in the step S2, the method for establishing the statistical model is as follows:
according to the physical data, establishing statistical models respectively corresponding to the physical data;
obtaining physical data corresponding to various fault types, and determining parameter state data corresponding to the physical data according to a statistical model of an operation parameter, wherein the parameter state data records the normal state or abnormal state of various physical data;
for each fault type, a statistical model of the operating parameter status data corresponding thereto is established.
According to the physical data, a statistical method can be adopted to establish a statistical model corresponding to the physical data. This helps the digital twin model understand and describe the distribution and variation of the physical data.
For various fault types, physical data related to these fault types is collected. Specifically including data during normal operation and when a fault occurs, for subsequent establishment of a statistical model associated with the fault.
The physical data is mapped to parameter state data using a statistical model of the operating parameters. These parameter status data record the normal status or abnormal status of various physical data. By analyzing the physical data, it is possible to identify which parameter states indicate that the power grid is operating normally and which indicate that the power grid has a fault.
For each fault type, a statistical model of the operating parameter status data corresponding to the fault type is established. This may include using a supervised learning approach, such as a classification algorithm, to distinguish between normal states and particular fault states. The statistical model may be learned by training a dataset and applied to fault detection in actual operation.
The established statistical model is validated and a separate test dataset is used to evaluate the performance of the model. According to the verification result, parameters or algorithms of the model need to be adjusted to improve the accuracy and generalization capability of the model.
The physical model and the statistical model are integrated into a digital twin model. The comprehensive model can fully utilize the physical principle and the statistical rule, and improves the comprehensive understanding and predicting capability of the power grid state.
Statistical models built for various fault types help to more accurately locate where the fault occurred. By analyzing the statistical model of the operation parameters, the abnormal part in the power grid can be rapidly identified. The establishment of the statistical model considers probability distribution of normal operation and abnormal operation, so that the false alarm rate can be reduced, and false alarm to normal operation is reduced. The application of the statistical model can improve the generalization capability through training and verification, so that the statistical model is more adaptive and predictive in actual operation. Through more accurate fault detection and positioning, more effective maintenance planning can be realized, and the cost of power grid maintenance is reduced. Timely fault discovery and accurate positioning can reduce downtime and maintenance costs. Through more comprehensive fault detection and more accurate positioning, measures can be taken in time to solve the problem, and the reliability and stability of the power grid are improved.
Further, in the step S2, the method for establishing the machine learning model is as follows:
performing feature classification, combination and reconstruction on the physical data based on the historical change data to obtain a general data set, dividing the general data set into a training set and a testing set, and training a machine learning model through the training set;
the machine learning model is subjected to repeated iterative training, error calculation is carried out by using a multi-weight loss function in each iterative training, a counter propagation algorithm is used, the weight and bias of the model are updated according to gradient information of the loss function, repeated iterative training is carried out until a preset stopping condition is reached, and finally the machine learning model with all parameter values determined is obtained.
The physical data is processed, including sorting, merging and reconstruction, to obtain a generic data set having representative characteristics. The method specifically comprises the preprocessing steps of extracting features and reducing dimensions of different physical data. The generic data set is divided into a training set and a testing set. The training set is used to train the machine learning model and the test set is used to evaluate the performance of the model on the new data.
The machine learning model is continually optimized using the training set through a training process of multiple iterations. In each iterative training, the model performs error calculation according to the weight loss function, and the weight and bias of the model are updated by using a back propagation algorithm.
Multiple weight loss functions are used in each iterative training. This may include various loss functions, such as mean square error (Mean Squared Error, MSE), etc., to better reflect the performance of the model on the training set.
The model weights and biases are updated using a back-propagation algorithm to better fit the model to the training data by calculating gradient information for the loss function to minimize the loss function.
The updating of the weights and the offsets is repeated through a plurality of iterative training until a predetermined stop condition is satisfied.
In the training process, a machine learning model with all the parameter values determined is finally obtained.
Machine learning models can provide data driven grid simulation and prediction by learning patterns and correlations in historical data. Digital twin models may benefit from such models to more accurately reflect the dynamic behavior of the actual grid. Through the training process of multiple iterations, the self-adaptability and generalization capability of the machine learning model can be improved, so that the machine learning model can adapt to different working conditions and changes. Digital twin models can benefit from it to better simulate the complexity of an actual grid. The training process of the machine learning model can generally be completed in a relatively short time, so that the digital twin model can be built more quickly. In addition, the online learning and updating function of the model can enable the digital twin model to continuously adapt to the change of the power grid. The machine learning model learns the normal and abnormal modes of the power grid during training, and can be used for fault detection and abnormal mode identification. The digital twin model can take advantage of these capabilities for more accurate grid monitoring and fault prediction.
In this embodiment, in the step S2, the method for establishing the digital twin model is as follows:
establishing a joint probability model, integrating a physical model, a statistical model and a machine learning model in the same framework, using a Bayesian network to represent the dependency relationship among the models, forming a joint probability graph model, creating a joint data set containing the output of the models and the actual power grid state, dividing the joint data set into a joint training set and a joint testing machine, using the joint training set to perform joint training on the whole joint probability model, using the joint testing set to perform verification and evaluation on the whole joint probability model, so as to establish a digital twin model, embedding the digital twin model into the actual power grid, receiving power grid feedback data in real time, and performing iterative updating on the model.
When a joint probability model is established, first, nodes in the Bayesian network are determined, and each node corresponds to a model or a specific variable. Including physical model nodes, statistical model nodes, machine learning model nodes, and nodes representing actual grid conditions.
And establishing directed edges between nodes to represent the dependency relationship between the models. And determining the connection mode between the nodes according to the mutual influence and the dependency relationship between different models.
Defining a probability distribution for each node, which can be determined using a physical equation for the physical model node; for statistical and machine learning model nodes, the probability distribution may be determined using training data.
And performing joint training on the physical model, the statistical model and the machine learning model, and creating a joint data set containing the output of each model and the actual power grid state, wherein the joint data set is divided into a joint training set and a joint testing machine. The combined training data set comprises physical model output, statistical model output, machine learning model output and actual power grid state; will be
The entire model is jointly trained using a probability map model defined in a bayesian network. The joint test set is divided into sections that can be independently validated for each model to evaluate the performance of each model.
The entire joint probability model is validated using the joint test set. The joint probability model is embedded into the digital twin model to form a component of the digital twin model.
And embedding a digital twin model into an actual power grid, and receiving power grid feedback data in real time, wherein the power grid feedback data comprises physical sensor data and model output.
Based on the data received in real time, the digital twin model is iteratively updated. And updating the weight and other relevant parameters of the machine learning model according to the actual power grid data.
In this embodiment, the specific method for receiving the feedback data of the power grid in real time and iteratively updating the model includes: and comparing the actual data with the data output by the digital twin model, judging whether the difference exceeds the level of random variation, wherein the difference comparison method is Root Mean Square Error (RMSE) or Mean Absolute Error (MAE), if the difference exceeds the level of random variation, judging that the digital twin model cannot be recognized, and iterating the digital twin model.
In this embodiment, in the step S3, the data collected in real time is calibrated, the data is denoised and filtered by using a signal processing technology, interference and noise are eliminated, the key features are extracted from the real-time data by using a machine learning technology, the extracted features are input into each component in the digital twin model, including the physical model, the statistical model and the machine learning model, the model is used for anomaly detection, a mode inconsistent with normal operation is identified, whether the real-time data is normal operation data is judged, and if not, a specific fault type is detected.
Key features are extracted from the preprocessed real-time data using machine learning techniques. Algorithms such as Principal Component Analysis (PCA), wavelet transformation, time-frequency analysis may be used as long as the key features can be extracted.
The extracted key features are combined into feature vectors which are used as the input of the subsequent model. It is ensured that the feature vector can sufficiently reflect the characteristics of the real-time data.
The feature vector is input to a physical model, a statistical model section, and a machine learning model section of the digital twin model, respectively.
Outputs are obtained from the various components of the digital twin model. Abnormality detection is performed using these outputs. And comparing the model output with the actual data, identifying an abnormal mode in the model output, and judging whether the real-time data is consistent with normal operation.
If an abnormality is detected, a fault diagnosis is performed using information in the model output. This may involve further analysis of the anomaly pattern to determine a specific fault type.
In the embodiment, risk assessment is carried out on the detected fault hidden danger, the influence degree of the fault hidden danger on the safety and stability of the power grid is determined, and the fault hidden danger is classified into high, medium and low risk grades; wherein, the hidden danger which can cause major faults, accidents of the power grid or affect large-area users is given a high risk level; for hidden dangers that cause problems but do not have a significant impact, a medium risk level is given; the hidden danger that the influence on the power grid is small and effective control can be achieved within a specified time is given to low risk level;
for the hidden danger with high risk, emergency measures are taken, an emergency response plan is started, emergency resources are mobilized, and faults are rapidly processed;
for the hidden danger with high risk, equipment maintenance and overhaul are carried out, and the change of the hidden danger in the wind is monitored in real time;
for the low risk hidden danger, regular maintenance is carried out, and the low risk hidden danger is monitored in real time.
And utilizing a digital twin model and a fault detection mechanism to detect fault hidden dangers. And performing risk assessment on the detected fault hidden danger, and considering possible influence and consequences thereof. The evaluated metrics may include area of impact, number of users, grid safety stability, etc. The hidden trouble is classified into high, medium and low risk grades. High risk means that a major grid failure may result, medium risk means that problems may occur but no major impact is caused, low risk means that the impact is small and that effective control can be obtained within a specified time. For fault hazards classified as high risk, emergency measures are initiated. Including initiating emergency response plans, mobilizing emergency resources, and quickly handling faults to minimize possible impact. Aiming at the hidden danger of high risk, equipment maintenance and overhaul are carried out, and the equipment is ensured to run in the optimal state. This helps reduce the likelihood of future similar failures. And carrying out real-time key monitoring on hidden trouble risks classified as stroke risks. Through the digital twin model and real-time data, the change of the risk potential of stroke and whether the risk is evolved into higher risk are monitored. And aiming at the fault hidden dangers classified as low risk, carrying out periodic maintenance and monitoring the state of the low risk hidden dangers in real time. Periodic maintenance helps ensure long-term stability of the device.
By classifying the hidden trouble and timely responding to emergency, the power grid can more effectively cope with possible risks and faults. This helps to improve the overall safety of the grid and reduce the probability of major faults and accidents. The emergency measures and equipment maintenance on the high-risk hidden trouble can reduce the influence of accidents on users and reduce the power failure time and service interruption. This helps to improve user satisfaction and reduce service interruption. Through reasonable evaluation and classification of risks, resources can be allocated more pertinently, and high-risk hidden dangers are processed preferentially. This helps to reduce the cost of emergency maintenance and emergency response, and improves the efficiency of resource utilization. Real-time monitoring and periodic maintenance of low and medium risk potential hazards are helpful for finding potential problems in advance and taking preventive measures. This may reduce possible future failures and reduce maintenance costs. By setting up an emergency response plan in advance, the system can be started up quickly when a fault occurs and mobilize necessary resources for processing. This helps to shorten the fault recovery time and improve the accuracy of the emergency response. Through the use of digital twin models and real-time data, more accurate decisions can be made based on actual conditions. This helps the grid manager to better understand the grid conditions, optimize the resource allocation and make a reasonable maintenance plan. Regular maintenance and monitoring, and timely treatment of high risk hidden danger, and are helpful for improving the long-term stability of the power grid.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. The power grid fault control method based on digital twinning is characterized by comprising the following steps of:
step S1: dividing a power grid into a plurality of units, installing a sensor in each unit, acquiring electrical data of each unit, performing calibration, denoising and standardization processing on the acquired electrical data, wherein the electrical data comprises physical data and historical change data, the physical data comprises voltage, current, power and frequency data, and the historical change data comprises historical performance and change trend of the power grid in each period of a statistical period;
step S2: establishing a physical model based on the physical data of the power grid, analyzing the historical change data of the power grid by using a statistical method to establish a statistical model, extracting the characteristics of the electrical data by a machine learning algorithm to establish a machine learning model, and fusing the physical model, the statistical model and the machine learning model to form a comprehensive digital twin model;
step S3: inputting the real-time data into a digital twin model, and identifying potential fault hidden danger in the power grid by analyzing the real-time data;
step S4: performing risk assessment on the detected fault hidden danger, determining the influence degree of the fault hidden danger on the safety and stability of the power grid, and classifying the fault hidden danger into high, medium and low risk grades; according to the risk level of the hidden trouble, a corresponding control strategy is formulated; for the hidden danger of high risk, immediate treatment measures are adopted; and for medium and low risk hidden dangers, selecting corresponding accurate monitoring and early warning schemes.
2. The digital twin-based power grid fault control method according to claim 1, wherein in the step S2, the physical model building method is as follows:
defining model variables, specifically including voltage, current, power and frequency, establishing a basic equation according to the model variables, establishing a physical model through a differential equation, adjusting the physical model by utilizing collected electrical data to enable the output of the model to be consistent with actual data, solving the differential equation by using a numerical method, simulating specific behaviors of a power grid in a period of time, and verifying the established physical model by comparing a simulation result with the actual data.
3. The digital twin-based power grid fault control method according to claim 1, wherein in the step S2, the statistical model building method is as follows:
according to the physical data, establishing statistical models respectively corresponding to the physical data;
obtaining physical data corresponding to various fault types, and determining parameter state data corresponding to the physical data according to a statistical model of an operation parameter, wherein the parameter state data records the normal state or abnormal state of various physical data;
for each fault type, a statistical model of the operating parameter status data corresponding thereto is established.
4. The digital twin based grid fault control method according to claim 1, wherein in the step S2, the method for creating the machine learning model is as follows:
performing feature classification, combination and reconstruction on the physical data based on the historical change data to obtain a general data set, dividing the general data set into a training set and a testing set, and training a machine learning model through the training set;
the machine learning model is subjected to repeated iterative training, error calculation is carried out by using a multi-weight loss function in each iterative training, a counter propagation algorithm is used, the weight and bias of the model are updated according to gradient information of the loss function, repeated iterative training is carried out until a preset stopping condition is reached, and finally the machine learning model with all parameter values determined is obtained.
5. The method for controlling a power grid fault based on digital twinning according to claim 1, wherein in the step S2, the method for creating the digital twinning model is as follows:
establishing a joint probability model, integrating a physical model, a statistical model and a machine learning model in the same framework, using a Bayesian network to represent the dependency relationship among the models, forming a joint probability graph model, creating a joint data set containing the output of the models and the actual power grid state, dividing the joint data set into a joint training set and a joint testing machine, using the joint training set to perform joint training on the whole joint probability model, using the joint testing set to perform verification and evaluation on the whole joint probability model, so as to establish a digital twin model, embedding the digital twin model into the actual power grid, receiving power grid feedback data in real time, and performing iterative updating on the model.
6. The digital twin-based power grid fault control method according to claim 1, wherein the specific method for receiving power grid feedback data in real time and iteratively updating the model is as follows: and comparing the actual data with the data output by the digital twin model, judging whether the difference exceeds the level of random variation, wherein the difference comparison method is root mean square error or average absolute error, if the difference exceeds the level of random variation, judging that the digital twin model cannot be identified, and iterating the digital twin model.
7. The grid fault control method based on digital twin according to claim 1, wherein in the step S3, the data collected in real time is calibrated, the data is denoised and filtered by using a signal processing technology, interference and noise are eliminated, key features are extracted from the real-time data by using a machine learning technology, the extracted features are taken as input, each component in the digital twin model including a physical model, a statistical model and a machine learning model is input, anomaly detection is performed by using the model, anomaly data inconsistent with normal data is identified, whether the real-time data is normal operation data is judged, and if not, a specific fault type is detected.
8. The digital twin-based power grid fault control method according to claim 1, wherein the detected fault hidden danger is subjected to risk assessment, the degree of influence on the safety and stability of the power grid is determined, and the degree of influence is classified into high, medium and low risk grades; wherein, the hidden danger which can cause major faults, accidents of the power grid or affect large-area users is given a high risk level; for hidden dangers that cause problems but do not have a significant impact, a medium risk level is given; the hidden danger that the influence on the power grid is small and effective control can be achieved within a specified time is given to low risk level;
for the hidden danger with high risk, emergency measures are taken, an emergency response plan is started, emergency resources are mobilized, and faults are rapidly processed;
for the hidden danger with high risk, equipment maintenance and overhaul are carried out, and the change of the hidden danger in the wind is monitored in real time;
for the low risk hidden danger, regular maintenance is carried out, and the low risk hidden danger is monitored in real time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311514043.7A CN117408162A (en) | 2023-11-14 | 2023-11-14 | Power grid fault control method based on digital twin |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311514043.7A CN117408162A (en) | 2023-11-14 | 2023-11-14 | Power grid fault control method based on digital twin |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117408162A true CN117408162A (en) | 2024-01-16 |
Family
ID=89497989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311514043.7A Pending CN117408162A (en) | 2023-11-14 | 2023-11-14 | Power grid fault control method based on digital twin |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117408162A (en) |
-
2023
- 2023-11-14 CN CN202311514043.7A patent/CN117408162A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104595170B (en) | A kind of air compressor machine monitoring and diagnosis system and method for self-adaptive kernel gauss hybrid models | |
| CN109977624A (en) | Photovoltaic plant soft fault monitoring method based on deep neural network | |
| CN107203816B (en) | Fault maintenance method and system for secondary equipment of power system | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| Mathew et al. | Regression kernel for prognostics with support vector machines | |
| CN113888353A (en) | Energy efficiency diagnosis method, system and medium for distributed photovoltaic power generation equipment | |
| Zhou | A fuzzy petri-net approach for fault analysis considering factor influences | |
| CN116308304A (en) | New energy intelligent operation and maintenance method and system based on meta learning concept drift detection | |
| CN117394529A (en) | SCADA-based auxiliary decision method and system for main distribution network loop-closing reverse power supply control conditions | |
| CN110412417B (en) | Micro-grid data fault diagnosis method based on intelligent power monitoring instrument | |
| Li et al. | Deep learning based covert attack identification for industrial control systems | |
| CN117273440A (en) | Engineering construction Internet of things monitoring and managing system and method based on deep learning | |
| CN114666117A (en) | Network security situation measuring and predicting method for power internet | |
| CN117494009A (en) | Electrical equipment state evaluation method based on insulating material pyrolysis analysis and cloud platform | |
| KR20220089853A (en) | Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology | |
| CN117031201A (en) | Multi-scene topology anomaly identification method and system for power distribution network | |
| JP7062505B2 (en) | Equipment management support system | |
| CN117408162A (en) | Power grid fault control method based on digital twin | |
| CN115936663A (en) | Maintenance method and device for power system | |
| US20180087489A1 (en) | Method for windmill farm monitoring | |
| CN115146727A (en) | Intelligent power utilization system fault detection method and system | |
| CN114414938B (en) | Dynamic response method and system for power distribution network faults | |
| CN117113157B (en) | Platform district power consumption fault detection system based on artificial intelligence | |
| CN117878925B (en) | Power transmission data control method and system of smart power grid | |
| JP7257455B2 (en) | Method and monitoring system for monitoring controller decision logic performance |
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 |