Regional power grid distributed situation awareness method and system based on federated learning

Abstract

The invention relates to the technical field of power grid situation awareness and fault detection, in particular to a regional power grid distributed situation awareness method and system based on federated learning, and the method comprises the steps: obtaining first-class feature data through data collection and processing, carrying out the training through the first-class feature data, constructing a second-class model, and carrying out the further training. And finally analyzing the fault condition by using the second-class model. The federated learning and edge computing technologies are combined, and data do not need to leave a local edge computing terminal, so that the data privacy of each power supply area is effectively protected, and the security of the system is improved; compared with a traditional centralized model, data does not need to be transmitted to a central server, and the risk of privacy disclosure is avoided; meanwhile, the edge calculation is combined with the central server, and local data processing and model training are completed at the edge calculation terminal, so that the calculation burden of the central server is reduced, the resource utilization rate is optimized, and an efficient solution is provided for power grid fault perception.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid situation awareness and fault detection, and in particular to a regional power grid distributed situation awareness method and system based on federated learning. Background Art

[0002] Traditional centralized systems require all data to be centralized to a central server, which may lead to the leakage of users' sensitive data or unauthorized access, and take a long time to process data and detect faults, which leads to slow response. Delayed detection and response may affect the stability and reliability of the power grid.

[0003] At the same time, centralized data processing also increases the risk of privacy leakage. Centralized fault detection systems require central servers to have strong computing power to process large amounts of data. This will lead to high hardware costs and energy consumption, and may become a system bottleneck.

[0004] This invention patent proposes a regional power grid distributed situational awareness method based on federated learning. The distributed situational awareness method based on edge computing and federated learning provides efficient visualization of fault information through real-time monitoring and fault detection of the power grid, and assists in adjusting the operation mode of the power grid. Power grid fault perception not only expands the ability of power grid status monitoring and fault handling, but also extends the scope of fault perception to assist decision-making, providing comprehensive, fast and accurate technical support for power grid dispatching, and effectively responding to various faults and maintenance events. Summary of the invention

[0005] In view of the above problems existing in the prior art, the present invention is proposed.

[0006] Therefore, the technical problem to be solved by the present invention is: for the perception detection of power grid faults. This technical solution processes data locally and only shares model parameters, thereby effectively protecting the privacy data of users, while distributing computing tasks to various nodes, reducing the computing burden and data transmission requirements of the central server, and improving the efficiency of the system. Federated learning does not transmit original data, but only transmits model parameters, which makes data more secure during transmission and reduces the risk of data tampering or leakage. And faults can be detected and predicted faster.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a regional power grid distributed situation awareness method based on federated learning, including: collecting and processing data to obtain a type of feature data; using the first type of feature data for training; constructing a second type of model for further training; and using the second type of model to analyze fault conditions.

[0008] As a preferred solution of the regional power grid distributed situational awareness method based on federated learning described in the present invention, the data collection and processing includes storing the collected data and performing preliminary processing on the data to standardize the processed data.

[0009] As a preferred solution of the regional power grid distributed situational awareness method based on federated learning described in the present invention, wherein: the training using a type of feature data includes optimizing the feature data, improving training efficiency and forming accurate analysis results.

[0010] As a preferred solution of the regional power grid distributed situational awareness method based on federated learning described in the present invention, the construction of the second type of model includes processing based on running data to provide a model suitable for further training.

[0011] As a preferred solution of the regional power grid distributed situation awareness method based on federated learning described in the present invention, the method of analyzing the fault condition using the two-type model includes judging the operating condition based on the model, and further performing fault analysis and status identification.

[0012] As a preferred solution of the regional power grid distributed situational awareness method based on federated learning described in the present invention, the data collection and processing includes performing a type of processing on a type of characteristic data collected, and using the processed standard data for training.

[0013] As a preferred solution of the regional power grid distributed situation awareness method based on federated learning described in the present invention, wherein: the one type of characteristic data includes main performance characteristic data of the operating status obtained from the monitoring and collection device and the sensor;

[0014] The second type of model includes a model generated after training and optimization based on the primary feature data, which is constructed by integrating operation data and local model information to support real-time status classification and fault detection in a distributed environment.

[0015] Another object of the present invention is to provide a regional power grid distributed situational awareness system based on federated learning, which can achieve efficient data collection, training, aggregation and fault analysis through modular design, improve the accuracy and real-time performance of power grid status perception, effectively protect data privacy, and optimize the collaborative operation efficiency in a distributed environment.

[0016] In order to solve the above technical problems, the present invention provides the following technical solutions: a regional power grid distributed situation awareness system based on federated learning, comprising: a data acquisition and preprocessing module, a local training module, a model aggregation module and a fault state identification module;

[0017] The data collection and preprocessing module collects and processes data to obtain a type of feature data;

[0018] The local training module is trained using a type of feature data;

[0019] The model aggregation module constructs a second type of model for further training;

[0020] The fault status identification module uses two types of models to analyze the fault situation.

[0021] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the regional power grid distributed situation awareness method based on federated learning are implemented as described above.

[0022] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a regional power grid distributed situation awareness method based on federated learning as described above are implemented.

[0023] Beneficial effects of the present invention: This technical solution adopts federated learning technology, which enables data to be processed at the local edge computing terminal without being transmitted to the central server, thereby effectively protecting the data privacy of the power supply area. In contrast, the traditional centralized model requires data to be transmitted to the central server, which not only increases the risk of privacy leakage, but also may cause data security issues. In addition, this solution completes local data processing and training at the edge computing terminal through the combination of edge computing and central servers, significantly reducing the computing pressure of the central server and optimizing resource utilization efficiency. Traditional methods often rely too much on central servers, which can easily lead to resource waste and performance bottlenecks. The present solution achieves efficient distributed computing while protecting privacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0025] Figure 1 An overall flow chart of a regional power grid distributed situation awareness method based on federated learning provided for one embodiment of the present invention.

[0026] Figure 2 A distribution network fault perception model diagram based on federated learning of a regional power grid distributed situation awareness method based on federated learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0028] Example 1, reference Figure 1-Figure 2 , is an embodiment of the present invention, which provides a regional power grid distributed situation awareness method based on federated learning, including:

[0029] S1: Collect and process data to obtain a type of characteristic data.

[0030] It should be noted that if Figure 1 As shown in S1, data collection and processing include storing the collected data and performing preliminary processing on the data to normalize the processed data; wherein data collection and processing include performing a type of processing on a type of characteristic data collected, and using the processed normalized data for training; a type of characteristic data includes main performance characteristic data of the operating status obtained from the monitoring collection device and the sensor.

[0031] Furthermore, S1 belongs to the data acquisition and preprocessing module, which obtains the main performance characteristic data of the operating status from the monitoring and acquisition devices or sensors of each substation in the power grid and stores it in the local historical data sample library. To obtain the operating status data of each substation in real time and improve the training efficiency and convergence speed of the model, the data needs to be standardized and normalized; in the distributed situation awareness system of the power grid, in order to improve the data quality, the collected raw data is first smoothed. The moving average method can be used to calculate the average value of the data in a certain window, and the exponential smoothing method can be used to give a higher weight to the recent data, thereby effectively reducing noise and making the data more stable; this process helps to highlight the core features of the data and facilitates subsequent model training.

[0032] Furthermore, the preprocessed data is subjected to anomaly detection, and outliers are identified using a set threshold or a statistical model-based method. For these abnormal data, the mean filling method can be used to replace the abnormal points with the average value of the adjacent normal data, or the median filling method can be used to reduce the impact of extreme values ​​on the data set. In addition, a regression model can be used to predict reasonable values ​​based on historical data to replace the abnormal points; through such a preprocessing process, it is ensured that the model can use high-quality data during the training process, thereby improving the accuracy of power grid fault detection and the overall stability of the system.

[0033] In the embodiments of the present application, one type of characteristic data is the main performance characteristic data of the operating status obtained from the monitoring and acquisition device and the sensor. These characteristic data can fully reflect the operating status of the power grid equipment. One type of characteristic data is the basic data for realizing model training and operating status analysis. After these data are collected by various sensors, they can reflect the normal or abnormal state characteristics of the power grid operation; for example, when the equipment has an abnormal load rate or excessive voltage fluctuations, these data can be directly used as the basis for fault warning. One type of characteristic data usually includes the operating performance indicators of the power equipment. These indicators can help determine whether the equipment has potential hidden dangers, such as overload operation, abnormal power factor or excessive frequency fluctuations.

[0034] In an optional embodiment, a type of characteristic data can also be realized in other ways, for example, by extracting key performance indicator data through the historical operation logs of various power grid equipment. These log data can be derived from the long-term monitoring records of the equipment, including the equipment's operating hours, maintenance records, and historical trends in power fluctuations. Compared with real-time monitoring data, these historical log data can more comprehensively reflect the long-term operating conditions and potential degradation trends of the equipment. In addition, the degree of matching between dispatch instructions and equipment responses can be obtained through the data interface of the power grid dispatching system, and these data can be used to further explore the changing patterns of the power grid operating status. A type of characteristic data obtained in this way can provide richer background information for power grid status assessment and fault prediction.

[0035] In an optional embodiment, a type of characteristic data can also be implemented in other ways, for example, through the integration of external data sources, such as meteorological data, load forecast data, and user electricity consumption behavior data, as auxiliary features of the operating status. These data can establish an association model through the environment and external conditions related to the operating status of the power grid, and explore the potential influencing factors of the operating characteristics of the power grid; for example, under meteorological conditions, changes in temperature, humidity and wind speed may directly affect the load capacity and operating efficiency of the transmission equipment; load forecast data can verify the dispatching efficiency of the equipment in combination with the actual operating conditions; user electricity consumption behavior data can reveal the demand-side characteristics of the power supply of the power grid; through the combination of these external data with the main performance characteristic data, the operating status of the power grid can be described more comprehensively, providing more dimensional data support for subsequent model training and status evaluation.

[0036] In the embodiment of the present application, one type of processing is preprocessing a type of feature data, that is, standardization and normalization processing, specifically smoothing processing, which is used to improve data quality and model training efficiency. Smoothing processing includes two methods: moving average method and exponential smoothing method. The moving average method can effectively reduce random noise in the data by calculating the average value of data in a certain window, making the data more stable and easy to analyze; the exponential smoothing law gives a higher weight to recent data, making it more suitable for scenarios with large dynamic changes; these processing methods help to highlight the core features of the data and provide high-quality input data for subsequent training;

[0037] Furthermore, when preprocessing a type of feature data, standardization can scale the data values ​​to a uniform scale range, for example, adjusting data of different units or magnitudes to zero mean and unit variance, thereby eliminating dimensional differences between dimensions; normalization improves the convergence speed and stability of the model under different input ranges by scaling the data values ​​to a specified range; in addition, the preprocessing of a type of feature data also includes anomaly detection and repair, using threshold setting or statistical model methods to identify and repair outliers in the data, such as using mean filling or median filling to replace abnormal data points to reduce the impact of extreme values ​​on model training; by preprocessing a type of feature data, not only can the quality and standardization of the data be ensured, but also the training efficiency of the model and the accuracy of the results can be effectively improved; this process provides high-quality input data for the distributed power grid situation awareness system, ensuring the accuracy of subsequent model optimization and fault detection.

[0038] In an optional embodiment, a type of processing can also be achieved in other ways. For example, instead of strictly normalizing the data, the continuous data is discretized into multiple intervals through binning, so that a rule-based analysis method can be adopted in the subsequent model; a fixed number of intervals can be divided according to the actual distribution of the data, and a specific category label is assigned to each interval; this processing method helps to reduce the impact of subtle fluctuations in the data and is more suitable for some application scenarios that need to focus on the overall trend; in addition, in terms of anomaly detection, a simple sliding window detection method can be used to judge anomalies by monitoring the short-term fluctuation range of the data in real time, rather than relying on complex statistical models or regression predictions, but this method has reduced accuracy.

[0039] In an optional embodiment, a type of processing can also be implemented in other ways, for example, directly using a rule-based processing method to compare the collected raw data with a preset reference standard, eliminate obviously abnormal values, and group and organize the data that meets the range. This method can quickly screen out abnormal data points and process them without using complex statistical models; in addition, a simple interpolation algorithm can be used to supplement missing data. It is easy to operate and has high computational efficiency, and is suitable for application in scenarios with small data volumes or high real-time requirements.

[0040] Example 2, reference Figure 1-Figure 2 , is an embodiment of the present invention, which provides a regional power grid distributed situation awareness method based on federated learning, including:

[0041] S2: Use one type of feature data for training.

[0042] It should be noted that if Figure 1 As shown in S2, using a type of feature data for training includes optimizing the feature data, improving training efficiency and forming accurate analysis results.

[0043] Furthermore, the local training module uses the feature data obtained from the data acquisition and preprocessing module to train the local model. Each edge computing terminal independently trains its local model and adopts a fault detection algorithm based on federated learning and edge computing.

[0044] Furthermore, each edge computing terminal initializes a local model first, uses the random weight initialization method, uses a fixed-size window to split the time series data, and each window is used as a training sample. The window size and step size are selected according to the specific task. The cross-validation method is used to divide the data set into K parts, and K-1 parts are used as training sets and 1 part as validation sets in turn, and the cycle is repeated K times; the local model training is applied;

[0045] Combining the above model selection, training process and optimization methods, an efficient and accurate local model for power grid fault perception based on the federated learning local training module is constructed.

[0046] S3: Build a second-category model for further training.

[0047] It should be noted that if Figure 1 As shown in S3, building the second type of model includes processing based on the running data to provide a model suitable for further training; wherein, the second type of model includes a model generated after training and optimization based on the primary feature data, which is constructed by integrating the running data and local model information to support real-time status classification and fault detection in a distributed environment.

[0048] Furthermore, the model aggregation runs on the central server, which is responsible for receiving the encrypted model parameters uploaded by each edge computing terminal and aggregating them. The federated averaging algorithm is used to perform weighted averaging on the feature quantities of the received models to generate a globally shared fault detection model:

[0049]

[0050] Among them, w global represents the global model parameters after aggregation, w i represents the local model parameters of the i-th edge terminal, n i represents the data volume of the ith edge terminal, n total Indicates the total amount of data of all edge terminals.

[0051] Furthermore, parameter aggregation is a key step in federated learning, which aims to integrate the local model parameters uploaded by each edge node into a global model. Specifically, after each edge node completes model training locally, it sends the model parameters to the central server. The server weights these parameters according to the amount of data involved in the training of each node to generate a global model. This process does not require the transmission of original data, but is achieved only by aggregating model parameters, which not only improves model performance but also effectively protects data privacy. Finally, the server sends the updated global model to each edge node for the next round of training. This method improves computing efficiency while ensuring data security and collaborative training effects in a distributed environment.

[0052] S4: Analyze the fault condition using the second type model.

[0053] It should be noted that if Figure 1 As shown in S4 of FIG. 4 , using the second type of model to analyze the fault condition includes judging the operation condition based on the model, and further performing fault analysis and state identification.

[0054] Furthermore, the state set of the power grid equipment model can be defined as A={A1,A2,A3,A4}, where A1,A2,A3,A4 are divided into four levels: normal, caution, abnormal, and severe, namely: A1 all parameter values ​​meet the standards; A2 equipment needs to be repaired, but it does not affect equipment operation; A3 equipment has a tendency to deteriorate to a fault condition; A4 equipment is faulty and needs to be repaired immediately.

[0055] The above is a schematic scheme of a regional power grid distributed situation awareness method based on federated learning in this embodiment. It should be noted that the technical solution of the system of the regional power grid distributed situation awareness method based on federated learning and the technical solution of the regional power grid distributed situation awareness method based on federated learning above belong to the same concept. The details of the technical solution of the regional power grid distributed situation awareness system based on federated learning in this embodiment that are not described in detail can all be referred to the description of the technical solution of the regional power grid distributed situation awareness method based on federated learning above.

[0056] Example 3, reference Figure 1-Figure 2 , which is an embodiment of the present invention, provides a regional power grid distributed situation awareness method based on federated learning. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0057] Using the digital twin computing method of power grid equipment based on federated learning, the method first uses the federated learning model to train and predict the performance data (i.e., current eigenvalues) of the power equipment to obtain the predicted eigenvalues. By further analyzing these predicted eigenvalues, the characteristic state value can be calculated, and finally the overall state value can be obtained, as shown in Table 1 and Table 2:

[0058] Table 1 Health status assessment model

[0059]

[0060] Table 2 Circuit breaker state model

[0061]

[0062]

[0063] A federated learning algorithm is used to perform distributed training on the power grid fault perception model.

[0064] Step 1: Each edge computing terminal configures local training parameters, sets the learning rate to 0.005, the batch size to 64, and the number of training iterations to 1000.

[0065] Step 2: Each edge computing terminal extracts the operating status data of the power grid equipment from the local data storage. The first 75% of the data set is used for training, and the last 25% is used for verification and testing.

[0066] Step 3: The edge computing terminal uses local training data to train the fault perception model. During the model training process, input data and train the local model parameters w i .

[0067] Step 4: After training is completed, each edge computing terminal will set the local model parameter w i The data is encrypted and uploaded to the central server.

[0068] Step 5: The central server receives the encrypted model parameters of each edge computing terminal, and uses the federated averaging algorithm (FedAvg) to aggregate the parameters to generate global model parameters.

[0069] In summary, the innovative federated learning architecture of the present invention ensures that power grid fault data and feature quantities are processed in local edge computing terminals; innovatively designs a distributed power grid equipped with an edge computing terminal model, which is responsible for local data collection, preprocessing and model training; innovatively designs an efficient parameter aggregation mechanism, aggregates model parameters in the central server through a federated averaging algorithm to generate a global model; innovatively proposes to use a global model to perform fault detection and status classification on the real-time operation data of the power grid.

[0070] Embodiment 4 is an embodiment of the present invention, which provides a regional power grid distributed situation awareness system based on federated learning, including: a data acquisition and preprocessing module, a local training module, a model aggregation module and a fault state identification module;

[0071] The data collection and preprocessing module collects and processes data to obtain a type of feature data;

[0072] The local training module is trained using a type of feature data;

[0073] The model aggregation module constructs a second type of model for further training;

[0074] The fault status identification module uses two types of models to analyze the fault situation.

[0075] This embodiment further provides a computing device, which is applicable to a regional power grid distributed situation awareness method based on federated learning, including:

[0076] A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement a regional power grid distributed situation awareness method based on federated learning as proposed in the above embodiment.

[0077] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, a regional power grid distributed situation awareness method based on federated learning as proposed in the above embodiment is implemented.

[0078] The storage medium proposed in this embodiment and the regional power grid distributed situational awareness method based on federated learning proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0079] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0080] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0081] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A regional power grid distributed situation awareness method based on federated learning, characterized by: include: Collect and process data to obtain a type of characteristic data; Use one type of feature data for training; Build a second-class model for further training; The fault condition is analyzed using the two-category model.

2. A regional power grid distributed situation awareness method based on federated learning as claimed in claim 1, characterized in that: The data collection and processing includes storing the collected data and performing preliminary processing on the data to standardize the processed data.

3. A regional power grid distributed situation awareness method based on federated learning as claimed in claim 2, characterized in that: The use of a type of feature data for training includes optimizing the feature data, improving training efficiency and forming accurate analysis results.

4. A regional power grid distributed situation awareness method based on federated learning as claimed in claim 3, characterized in that: The construction of the second type of model includes processing based on the running data to provide a model suitable for further training.

5. A regional power grid distributed situation awareness method based on federated learning as claimed in claim 4, characterized in that: The use of the second type of model to analyze the fault condition includes judging the operating condition based on the model, and further performing fault analysis and state identification.

6. A regional power grid distributed situation awareness method based on federated learning as claimed in claim 5, characterized in that: The data collection and processing includes performing a type of processing on a type of characteristic data collected, and using the processed standard data for training.

7. A regional power grid distributed situation awareness method based on federated learning as claimed in claim 6, characterized in that: The first type of characteristic data includes main performance characteristic data of the operating status obtained from the monitoring and collection device and the sensor; The second type of model includes a model generated after training and optimization based on the primary feature data, which is constructed by integrating operation data and local model information to support real-time status classification and fault detection in a distributed environment.

8. A regional power grid distributed situational awareness system based on federated learning according to any one of claims 1 to 7, characterized in that: include: Data acquisition and preprocessing module, local training module, model aggregation module and fault status identification module; The data collection and preprocessing module collects and processes data to obtain a type of feature data; The local training module is trained using a type of feature data; The model aggregation module constructs a second type of model for further training; The fault status identification module uses two types of models to analyze the fault situation.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a regional power grid distributed situational awareness method based on federated learning as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a regional power grid distributed situational awareness method based on federated learning as described in any one of claims 1 to 7 are implemented.

Images