Electric power data consistency evaluation method and system based on deep learning

Abstract

The invention provides an electric power data consistency evaluation method and system based on deep learning, and the method comprises the steps: carrying out the scene analysis based on the operation state information of an electric power system in a current operation state, and determining a target electric power scene type in the current operation state; matching in a pre-established power model library based on the target power scene category to obtain a target power model; inputting power data of the power system in the current operation state into the target power model to obtain a preliminary consistency evaluation result output by the target power model; and determining a final consistency evaluation result of the power data based on correlation and difference analysis between the initial consistency evaluation results under different target power scene categories. According to the method, the problem of evaluation deviation caused by insufficient scene adaptability is avoided, and a reliable basis is provided for stable operation and optimization decision making of a power system.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and system for evaluating the consistency of power data based on deep learning. Background Technology

[0002] In the operation and management of power systems, the consistency assessment of power data is a crucial step in ensuring the safe, stable, and efficient operation of the system. Existing power data consistency assessment methods typically rely on a single power model and fixed assessment rules to perform a comprehensive evaluation of the power data. This method involves simple verification and comparison of the power data to determine whether it conforms to a preset standard format and range, thereby arriving at a data consistency conclusion. However, this assessment method has significant drawbacks. Because power systems operate in complex and diverse scenarios, encompassing various conditions such as normal operation, fault handling, and load adjustment, a single model and fixed rules are insufficient to adapt to the characteristics and changes of power data under different scenarios. In complex scenarios, this method cannot accurately identify consistency issues under different business logics and physical constraints, leading to biased assessment results and failing to provide a reliable basis for power system optimization decisions. Summary of the Invention

[0003] This invention provides a deep learning-based method and system for power data consistency assessment, aiming to avoid assessment bias caused by insufficient scenario adaptability and provide a reliable basis for the stable operation and optimization decision-making of power systems.

[0004] In a first aspect, the present invention provides a deep learning-based method for evaluating the consistency of power data, comprising: Based on the operating status information of the power system under the current operating state, scenario analysis is performed to determine the target power scenario category under the current operating state; The target power model is obtained by matching the target power scenario category with a pre-established power model library. The power data of the power system under the current operating state is input into the target power model to obtain the preliminary consistency assessment result output by the target power model; the preliminary consistency assessment result characterizes whether the power data meet the physical laws under the target power scenario category; Based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories, the final consistency assessment results of power data are determined.

[0005] In a second aspect, the present invention also provides a deep learning-based power data consistency assessment system, applied to the deep learning-based power data consistency assessment method described in the first aspect; the deep learning-based power data consistency assessment system includes: The scenario analysis module is used to perform scenario analysis based on the operating status information of the power system under the current operating state, and to determine the target power scenario category under the current operating state; The model matching module is used to match the target power scenario category in a pre-established power model library to obtain the target power model; The consistency prediction module is used to input the power data of the power system under the current operating state into the target power model to obtain the preliminary consistency evaluation result output by the target power model; the preliminary consistency evaluation result characterizes whether the power data meet the physical laws under the target power scenario category; The consistency assessment module is used to determine the final consistency assessment result of power data based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories.

[0006] Thirdly, the present invention also provides an electronic device, comprising: a memory for storing computer software programs; and a processor for reading and executing the computer software programs, thereby implementing the deep learning-based power data consistency assessment method as described above.

[0007] Fourthly, the present invention also provides a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements the deep learning-based power data consistency assessment method described above.

[0008] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the deep learning-based power data consistency assessment method as described above.

[0009] The deep learning-based power data consistency assessment method provided in this invention classifies power scenarios into different categories based on operational status information and selects corresponding power models for evaluation. The power models for different scenario categories reflect the unique physical laws and operational rules of that scenario, ensuring that power data can be evaluated in accordance with actual conditions under various operating conditions. Furthermore, after obtaining preliminary results from the evaluation by power scenario category, a comprehensive judgment is made, fully considering the correlation between data consistency across different power scenario categories. Therefore, it can more accurately identify consistency issues of power data in different scenarios, effectively avoiding evaluation bias caused by insufficient scenario adaptability, thus providing a reliable basis for stable operation and optimization decisions of the power system. Attached Figure Description

[0010] Figure 1This is a flowchart illustrating the power data consistency assessment method based on deep learning provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the power data consistency assessment system based on deep learning provided in an embodiment of the present invention; Figure 3 An embodiment diagram of the electronic device provided in this invention; Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0013] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0014] See Figure 1 , Figure 1 This is a flowchart illustrating the deep learning-based power data consistency assessment method provided by the present invention. In this embodiment, the execution entity of the deep learning-based power data consistency assessment method is the power management system. Therefore, the deep learning-based power data consistency assessment method includes: Step 10: Perform scenario analysis based on the operating status information of the power system under the current operating state to determine the target power scenario category under the current operating state.

[0015] Optionally, the power management system can collect various operating status information of the power system in real time under the current operating state, including but not limited to electrical parameters such as voltage, current, active power, reactive power, and frequency, as well as the operating status of equipment (such as the opening and closing status of switches, transformer taps, etc.).

[0016] Furthermore, the power management system performs scenario analysis based on the operating status information under the current operating state to determine the target power scenario category under the current operating state, specifically as in steps 101 to 105. Optionally, the target power scenario category in this embodiment of the invention includes a normal operation scenario category, a fault scenario category, and a load adjustment scenario category. The fault scenario category indicates that the operating status information matches a certain fault feature, and the load adjustment scenario category indicates that the load has undergone a significant change.

[0017] Step 20: Match the target power scenario category with a pre-established power model library to obtain the target power model.

[0018] Furthermore, a power model library is pre-established in the power management system. This library contains power models for different target power scenarios. For example, the steady-state power model is used to simulate and analyze various electrical characteristics of the power system under stable operating conditions; the short-circuit analysis model is specifically used to analyze the changes in current, voltage, and other parameters when a short circuit or other fault occurs in the power system; and the power flow calculation model is used to calculate the voltage and power distribution of each node in the power system when the load changes.

[0019] Therefore, once the target power scenario category is determined, the power management system searches for the power model corresponding to the target power scenario category in the power model library and selects it as the target power model under that target power scenario category. Optionally, in this embodiment of the invention, a steady-state power model corresponding to the normal operation scenario category, a short-circuit analysis model corresponding to the fault scenario category, and a power flow calculation model corresponding to the load adjustment scenario category are pre-defined in the power model library.

[0020] Step 30: Input the power data of the power system under the current operating state into the target power model to obtain the preliminary consistency assessment results output by the target power model. The preliminary consistency assessment results characterize whether the power data meet the physical laws under the target power scenario category.

[0021] Furthermore, the power management system collects various power data from the power system under its current operating state, such as real-time data on voltage, current, and power, as well as static data such as equipment parameters and network topology, and converts and organizes them according to the input requirements of the target power model. Then, the organized power data is input into the selected target power model.

[0022] Furthermore, the target power model calculates and analyzes the input power data according to its internal logic algorithm to determine whether the data satisfy the physical laws under the target power scenario category. If they satisfy the laws, the output power data shows a preliminary assessment result of consistency; if they do not satisfy the laws, the output power data shows a preliminary assessment result of inconsistency, as detailed in steps 301 to 304.

[0023] Step 40: Based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories, determine the final consistency assessment results of the power data.

[0024] Furthermore, the power management system analyzes the correlations between preliminary assessment results under different target power scenario categories. For example, the changing trends of certain power data may be correlated under normal operation and load adjustment scenarios; data differences between fault scenarios and normal operation scenarios may reflect the severity and scope of the fault. Further, the power management system identifies the specific causes and manifestations of data inconsistencies by comparing the differences in preliminary assessment results under different target power scenario categories. Further still, the power management system comprehensively considers these correlations and differences, combining the actual operation of the power system with historical data, to conduct a comprehensive assessment of the consistency of the power data, determining the final consistency assessment result, as detailed in steps 401 to 406.

[0025] In one embodiment, within a certain time period, the preliminary consistency assessment result under the normal operation scenario category was "consistent", and the preliminary consistency assessment result under the load adjustment scenario category was also "consistent". Furthermore, through analysis, it was found that during the load adjustment process, the trend of power data change was consistent with the expected change during normal operation, and there was a reasonable correlation between the two. Therefore, the final consistency assessment result of the power data was determined to be "consistent".

[0026] This invention categorizes power scenarios based on operational status information and selects corresponding power models for evaluation. The power models for each scenario reflect the unique physical laws and operational rules of that scenario, ensuring that power data is evaluated in accordance with actual conditions under different operating circumstances. After obtaining preliminary results from the evaluation by power scenario category, a comprehensive judgment is made, fully considering the correlation between data consistency across different power scenario categories. This allows for more accurate identification of consistency issues in power data across different scenarios, effectively avoiding evaluation biases caused by insufficient scenario adaptability, and thus providing a reliable basis for stable operation and optimization decisions of the power system.

[0027] In one embodiment, steps 101 to 105 are described as follows: Step 101: Identify abnormal devices based on the device operating parameters of each operating device.

[0028] Optionally, real-time data collection of equipment operating parameters, such as voltage, current, temperature, and speed, is performed on each device in the power system. For each device, the power management system establishes a fluctuation range for its normal operating parameters, which is derived from statistical analysis of the device's design parameters and historical operating data. When the operating parameters of a device exceed the normal fluctuation range, the device is identified as abnormal. Specifically, this embodiment of the invention employs the confidence interval method from statistics. It is assumed that the historical data of a certain parameter of the device conforms to a normal distribution. ,in, The mean, For standard deviation, take 3. The principle is that when the current operating parameters of the equipment exceed the range... If this occurs, the device is determined to be faulty.

[0029] In one embodiment, taking a transformer in a substation as an example, the power management system collects oil temperature data of the transformer over a long period of time. Statistical analysis shows that the historical oil temperature data conforms to a normal distribution. (Unit: °C). At a certain moment, the transformer oil temperature was measured at 68 °C. 68 > 50 + 3 * 5, which exceeds the normal fluctuation range. Therefore, the power management system identified the transformer as an abnormal device.

[0030] Step 102: Determine the anomaly index based on the equipment operating parameters of each abnormal device and its historical operating parameters, and determine the equipment anomaly impact value based on the importance coefficient of each abnormal device in the power system.

[0031] Furthermore, for each anomalous device, its anomalous index is first calculated. The anomalous index is calculated using a combination of Dynamic Time Warping (DTW) and cosine similarity. Let the current operating parameter sequence of the anomalous device be... The historical normal operation parameter sequence is as follows First, find using the DTW algorithm. and Find the optimal alignment path, and then calculate the cosine similarity between the two aligned sequences. The anomaly index E can be expressed as E=1- .

[0032] Furthermore, in determining the importance coefficient of each abnormal device in the power system, this embodiment of the invention employs the Analytic Hierarchy Process (AHP). A judgment matrix is ​​constructed, and pairwise comparisons are made from multiple dimensions such as the device's location in the power grid, its impact on power supply reliability, and its capacity to obtain the importance coefficient I of each device. The formula for calculating the abnormal impact value V of the device is V=E*I.

[0033] Continuing with the example of the transformer previously identified as abnormal, its current oil temperature parameter sequence is as follows: =[68, 67, 66] (unit: ℃, data collection time interval is 10 minutes), selected historical normal oil temperature sequence =[52, 53, 54]. After finding the optimal alignment path using the DTW algorithm, the cosine similarity is calculated. =0.7, then the anomaly index of the transformer is E=1-0.7=0.3. Using the analytic hierarchy process (AHP), the importance coefficient of the transformer in the power system is determined to be I=0.8, so the equipment anomaly impact value of the transformer is V=0.3*0.8=0.24.

[0034] Step 103: Determine the intensity of the topology change based on the change in connectivity of each node in the power grid topology before and after the topology change, and the initial connectivity value of each node before the change.

[0035] Furthermore, the power management system records the power grid topology, and for each node, its connectivity is measured by the number of branches directly connected to that node.

[0036] In one embodiment, the node before the topology change The initial connectivity value is After the change, it becomes Then the node Change in connectivity =| - | Intensity of topological changes The information entropy method is used to calculate the probability of changes in connectivity at each node. ( (total number of nodes), then according to the information entropy formula The intensity of the topological change is obtained.

[0037] In one embodiment, a power grid contains 5 nodes, and the initial connectivity value of node 1 before the topology change is... =3, after the change =2, then =|2-3|=1; Similarly, calculate the change in connectivity of other nodes. =0, =2, =1, =0. Then , , , , The intensity of topological changes can be obtained from the information entropy formula. .

[0038] Step 104: Determine the load fluctuation trend value based on the load change rate within the preset time interval, combined with the maximum load value, minimum load value, and average load value within the preset time interval.

[0039] Furthermore, the power management system calculates the load change rate within a preset time interval. In one embodiment, the load at the beginning of the preset time interval is... The load at the end time is Then the load change rate Combined with the maximum load during that time interval Minimum value and average In this embodiment of the invention, the Support Vector Regression (SVR) algorithm is used to predict the fluctuation trend of the load. , , , The load fluctuation trend value is input as a feature vector into the trained SVR model and outputs the feature vector. The SVR model is trained using historical load data and employs a radial basis function (RBF) kernel, with the optimization objective being to minimize structural risk.

[0040] For example, in a city's power grid, the preset time interval is the past hour, and the load at the start time... =1000kW, load at the end of the period =1200kW, then the load change rate =0.2. Maximum load value within this interval. =1300kW, minimum value =900kW, average value =1100kW. The feature vector [0.2, 1300, 900, 1100] is input into the trained SVR model, and the output load fluctuation trend value is... =0.6 (The larger this value is, the more obvious the upward trend of the load).

[0041] Step 105: Based on the equipment anomaly impact value, topology change intensity, and load fluctuation trend value, perform scenario analysis to determine the target power scenario category under the current operating state.

[0042] Furthermore, the power management system performs scenario analysis based on the equipment anomaly impact value, the intensity of topology changes, and the load fluctuation trend value to determine the target power scenario category under the current operating state, as detailed in steps 1051 to 1059.

[0043] This invention classifies different power scenario categories based on the operating status information under the operating state, providing a data foundation for subsequent processing. Therefore, power models corresponding to different power scenario categories can be selected for evaluation. Power models under different power scenario categories can reflect the unique physical laws and business rules of the scenario, so that power data can be evaluated in accordance with the actual situation under different operating conditions.

[0044] In one embodiment, steps 1051 to 1059 are described as follows: Step 1051: Based on the impact value of equipment anomalies and the intensity of topology changes, perform correlation analysis to determine the first degree of correlation between topology changes and equipment anomalies. Based on the impact value of equipment anomalies and the trend value of load fluctuations, perform correlation analysis to determine the second degree of correlation between load fluctuations and equipment anomalies.

[0045] Optionally, the power management system employs canonical correlation analysis (CCA) to calculate the correlation between equipment anomaly impact values ​​and topology change intensity, and between equipment anomaly impact values ​​and load fluctuation trend values. CCA finds the linear combination of two sets of variables that maximizes the correlation coefficient between them; this maximum correlation coefficient is the canonical correlation coefficient between the two sets of variables, and is used as the correlation degree.

[0046] For example, the vector formed by the impact values ​​of equipment anomalies is The vector formed by the intensity of topological changes is The maximum correlation coefficient obtained through CCA is the first correlation between topology change and equipment anomaly. Similarly, let the vector formed by the load fluctuation trend values ​​be... The maximum correlation coefficient, or second correlation degree, between the equipment anomaly impact value and the load fluctuation trend value is calculated using CCA. .

[0047] In one embodiment, within a certain time period, the abnormal impact values ​​of five devices were monitored as [0.2, 0.3, 0.1, 0.4, 0.25], and the corresponding topology change intensity values ​​were recorded as [0.8, 1.2, 0.5, 1.5, 1.0], and the load fluctuation trend values ​​as [0.3, 0.6, 0.2, 0.7, 0.4]. The device abnormal impact value vector X = [0.2, 0.3, 0.1, 0.4, 0.25] and the topology change intensity vector Y = [0.8, 1.2, 0.5, 1.5, 1.0] were used to perform CCA calculations to obtain the first correlation degree. =0.7; Perform CCA calculation on the equipment anomaly impact value vector X and the load fluctuation trend value vector Z=[0.3, 0.6, 0.2, 0.7, 0.4] to obtain the second correlation degree. =0.5.

[0048] For situations where the impact value of equipment malfunction exceeds the preset malfunction impact threshold: Step 1052: If the first correlation degree is greater than the first preset correlation threshold, then the target power scenario category is determined to be the fault scenario category.

[0049] Furthermore, when the impact value of the equipment malfunction is determined to be greater than the preset malfunction impact threshold... At that time, check the first correlation between topology changes and device anomalies. .like Greater than the first preset association threshold This indicates a strong correlation between equipment malfunction and topology changes. In this case, the equipment failure likely caused the topology change, so the power management system determines the target power scenario category as the fault scenario category.

[0050] Step 1053: If the first correlation degree is less than or equal to the first preset correlation threshold, and the load fluctuation trend value is greater than the preset load fluctuation threshold, then the target power scenario category is determined to be the fault scenario category.

[0051] Furthermore, when the impact value of equipment malfunction exceeds the preset malfunction impact threshold... But the first degree of relevance Less than or equal to the first preset association threshold At that time, the power management system continues to assess the load fluctuation trend. .like Greater than the preset load fluctuation threshold This indicates that although the equipment malfunction is not closely related to changes in the topology, it is potentially related to load fluctuations. Large load fluctuations may put pressure on the equipment and thus cause malfunctions. This situation is also likely to be classified as a fault scenario.

[0052] Step 1054: If the first correlation degree is less than or equal to the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is less than or equal to the second preset correlation threshold, then the target power scenario category is determined to be the load adjustment scenario category.

[0053] Furthermore, when the impact value of equipment malfunction exceeds the preset malfunction impact threshold... First degree of relevance Less than or equal to the first preset association threshold Load fluctuation trend value Less than or equal to the preset load fluctuation threshold And the second degree of correlation Less than or equal to the second preset association threshold This indicates that the correlation between equipment malfunctions and topology changes or load fluctuations is not strong. In this case, the equipment malfunctions are more likely caused by fluctuations generated during normal load adjustment, so the power management system determines the target power scenario category as the load adjustment scenario category.

[0054] Step 1055: If the first correlation degree is less than or equal to the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is greater than the second preset correlation threshold, then the target power scenario category is determined to be the fault scenario category.

[0055] Furthermore, when the impact value of equipment malfunction exceeds the preset malfunction impact threshold... First degree of relevance Less than or equal to the first preset association threshold Load fluctuation trend value Less than or equal to the preset load fluctuation threshold But the second degree of correlation Greater than the second preset association threshold When this occurs, it indicates a strong correlation between equipment malfunction and load fluctuation. Even if the load fluctuation does not exceed the preset threshold, this strong correlation suggests a potential fault. Therefore, the power management system determines the target power scenario category as the fault scenario category.

[0056] For cases where the impact value of equipment anomalies is less than or equal to the preset anomaly impact threshold: Step 1056: If the first correlation degree is less than or equal to the first preset correlation threshold, then the target power scenario category is determined to be the normal operation scenario category.

[0057] Furthermore, when the power management system determines that the abnormal impact value of the equipment is less than or equal to the preset abnormal impact threshold... At that time, check the first correlation between topology changes and device anomalies. .like Less than or equal to the first preset association threshold This indicates that the equipment is operating normally and is not significantly related to changes in the topology. The power management system determines the target power scenario category as the normal operation scenario category.

[0058] Step 1057: If the first correlation degree is greater than the first preset correlation threshold and the load fluctuation trend value is greater than the preset load fluctuation threshold, then the target power scenario category is determined to be the load adjustment scenario category.

[0059] Furthermore, when the equipment malfunction impact value is less than or equal to the preset malfunction impact threshold... But the first degree of relevance Greater than the first preset association threshold And load fluctuation trend value Greater than the preset load fluctuation threshold This indicates that although the impact of equipment malfunctions is relatively small, there is a strong correlation between topology changes and load fluctuations. This situation is more consistent with the topology changes caused during load adjustment. Therefore, the power management system determines the target power scenario category as the load adjustment scenario category.

[0060] Step 1058: If the first correlation degree is greater than the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is greater than the second preset correlation threshold, then the target power scenario category is determined to be the load adjustment scenario category.

[0061] Furthermore, when the equipment malfunction impact value is less than or equal to the preset malfunction impact threshold... First degree of relevance Greater than the first preset association threshold Load fluctuation trend value Less than or equal to the preset load fluctuation threshold And the second degree of correlation Greater than the second preset association threshold When the topology changes are strong, it indicates that the equipment is operating normally. However, this correlation is more likely to be a normal phenomenon during the load adjustment process. Therefore, the power management system determines the target power scenario category as the load adjustment scenario category.

[0062] Step 1059: If the first correlation degree is greater than the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is less than or equal to the second preset correlation threshold, then the target power scenario category is determined to be the normal operation scenario category.

[0063] Furthermore, when the equipment malfunction impact value is less than or equal to the preset malfunction impact threshold... First degree of relevance Greater than the first preset association threshold Load fluctuation trend value Less than or equal to the preset load fluctuation threshold And the second degree of correlation Less than or equal to the second preset association threshold When the topology changes are highly correlated with equipment anomalies, they are not closely related to load fluctuations. This situation still falls under the normal operation scenario, so the target power scenario category is determined to be the normal operation scenario category.

[0064] When the impact value of equipment anomalies is large, this invention accurately identifies fault scenarios caused by equipment failure leading to topology changes or related to load fluctuations by combining topology changes and load fluctuation correlation. When the impact value of equipment anomalies is small, it distinguishes normal operation scenarios and load adjustment scenarios caused by load adjustments leading to topology changes or related to load fluctuations through multi-dimensional correlation. Therefore, it can more accurately capture the inherent connections and change characteristics of operating states, improving the accuracy and reliability of power scenario classification.

[0065] In one embodiment, steps 301 to 304 are described as follows: Step 301: In the target power model, the power data is split according to different electrical components and lines to obtain power data vectors. The power data vectors include sub-vectors related to nodes and loops. The power data includes voltage data, current data, power data, and frequency data from various data sources.

[0066] Optionally, the power management system acquires various types of power data under the current operating status of the power system, including voltage, current, power, and frequency data from different data sources. In the target power model, the power data is broken down according to electrical components (such as transformers, generators, and loads) and lines within the power system. Node-related power data is further integrated into node-related sub-vectors, such as node voltage and injected current; circuit-related power data is integrated into circuit-related sub-vectors, such as circuit current and voltage drop. This ultimately forms a complete power data vector, which comprehensively and systematically contains the operating information of all parts of the power system.

[0067] In one embodiment, a power system comprising 3 nodes and 2 branches, as well as 1 generator, 1 transformer, and 1 load, is monitored. The power data collected by the system includes: the voltage of node 1 is... =10kV, injection current =50A; Voltage at node 2 =9.8kV, injected current =-30A (the negative sign indicates that the current flows out of the node); Voltage at node 3 =10.2kV, injected current =20A; Current in branch 1 =40A, voltage drop =0.5kV; Current in branch 2 =10A, voltage drop =0.3kV; Generator output power =800kW, frequency =50Hz. In the target power model, node-related data is split into node-related sub-vectors. The data related to the loop (assuming the loop is formed by branches) is split into loop-related sub-vectors. Finally, the power data vector is obtained. .

[0068] Step 302: Construct the node-branch association matrix and the loop-branch association matrix based on the power system topology. For the matrix elements in the node-branch association matrix, if a node is connected to a branch and current flows out of the node, it is 1; if current flows in, it is -1; and if they are not connected, it is 0. For the matrix elements in the loop-branch association matrix, if a loop contains a branch and the direction is the same, it is 1; if the direction is opposite, it is -1; and if it does not contain a branch, it is 0.

[0069] Furthermore, the power management system constructs a node-branch association matrix based on the power system's topology. Correlation matrix of loop branches .

[0070] For the node-branch association matrix Its rows correspond to nodes, and its columns correspond to branches. If a node and branch road Connected, and current flows from the node. Flowing out to the branch road Then matrix elements If the current flows from the branch Inflow node ,but =-1; if node and branch road If not connected, then =0.

[0071] For the correlation matrix of loop branches The rows correspond to loops, and the columns correspond to branches. If a loop... Includes branch roads And branch roads Current direction and loop If the reference directions are consistent, then the matrix elements =1; if the direction is opposite, then =-1; if the loop Excluding branches ,but =0.

[0072] Continuing with the example of the aforementioned power system, its topology is as follows: Node 1 is connected to Node 2 via branch 1, and Node 2 is connected to Node 3 via branch 2. Therefore, a node-branch association matrix needs to be constructed. : .

[0073] In this diagram, the first row represents node 1, which is connected to branch 1 and from which current flows. =1, not connected to branch 2, so =0; The second line represents node 2, where current flows in from branch 1 and out from branch 2, so =-1, =1; The third line represents node 3, where current flows in from branch 2, so =-1, not connected to branch 1, so =0.

[0074] Assuming that branch 1 and branch 2 form a loop, with the reference direction being clockwise, construct the loop branch association matrix. : .

[0075] Because the current direction in both branch 1 and branch 2 is consistent with the loop reference direction, therefore =1, =1.

[0076] Step 303: Determine the node current vector based on the node branch correlation matrix and the power data vector, and determine the loop voltage vector based on the loop branch correlation matrix and the power data vector.

[0077] Furthermore, the power management system utilizes the node-branch association matrix. and power data vector The node current vector is determined by matrix multiplication. The calculation formula is: ,in, It is a sub-vector related to current in the power data vector.

[0078] Similarly, using the loop branch correlation matrix and power data vector The loop voltage vector is determined by matrix multiplication. The calculation formula is: ,in, It is a voltage-related subvector within the power data vector.

[0079] Continuing in the example above, the subvectors related to current in the power data vector =[50, 40, 10], voltage-related subvectors =[0.5, 0.3].

[0080] Calculate the node current vector : .

[0081] Calculate the loop voltage vector : .

[0082] Step 304: Perform physical law mapping based on node current vector and / or loop voltage vector to obtain physical law mapping results, and perform consistency prediction based on physical law mapping results and scenario constraint adaptation conditions to output preliminary consistency evaluation results.

[0083] Furthermore, the power management system performs physical law mapping based on the node current vector and / or loop voltage vector to obtain the physical law mapping result, as specifically in steps 3041 to 3043.

[0084] Furthermore, the power management system performs consistency prediction based on the physical law mapping results and scenario constraint adaptation conditions, and outputs preliminary consistency assessment results, as detailed in steps 3044 to 3047.

[0085] The embodiments of this invention systematically decompose power data, construct an association matrix, determine key vectors, and analyze them in conjunction with physical laws and scenario constraints, forming a complete and rigorous preliminary consistency assessment process for power data. Therefore, it can more comprehensively and accurately uncover potential inconsistencies in power data, effectively improving the accuracy and reliability of power data consistency assessment.

[0086] In one embodiment, steps 3041 to 3043 are described as follows: Step 3041: If the target power scenario category is the normal operation scenario category, then the node power vector is obtained by mapping the physical laws based on the first node voltage amplitude and first node voltage phase angle of each node, the second node voltage amplitude and second node voltage phase angle of each node's adjacent nodes, branch conductance, branch susceptance, and power-related elements in the node current vector.

[0087] Optionally, when the target power scenario category is a normal operation scenario category, the power management system performs mapping based on the steady-state operation physical laws of the power system. For each node in the power system, the first node voltage amplitude of the node... Phase angle with the first node voltage and the voltage amplitude of the second node of its adjacent node. Phase angle of the second node voltage At the same time, the power management system obtains the conductance of each branch. and susceptance Combining the power-related elements in the node current vector, the node power is calculated using the following formula: .

[0088] .

[0089] in, and They are nodes Active power and reactive power, For nodes The set of adjacent nodes, The power-related term in the node current vector There are 10 elements. Therefore, by combining the active and reactive power of all nodes into a node power vector, the physical law mapping is completed.

[0090] Step 3042: If the target power scenario category is a fault scenario category, then the loop voltage vector is converted according to the short circuit type and short circuit location of the short circuit point to obtain the short circuit current vector.

[0091] Furthermore, when the target power scenario category is a fault scenario category, the power management system transforms the loop voltage vector based on the short-circuit type (e.g., three-phase short circuit, two-phase short circuit, single-phase ground fault, etc.) and short-circuit location. Taking a three-phase short circuit as an example, positive-sequence, negative-sequence, and zero-sequence network analysis methods are used to establish equations to solve for the short-circuit current based on the short-circuit boundary conditions. Let the node number of the short-circuit point be... For a three-phase short circuit, the positive-sequence, negative-sequence, and zero-sequence voltages at the short-circuit point are all 0. The short-circuit current can be calculated using the following equation: .

[0092] in, , , These represent the positive-sequence, negative-sequence, and zero-sequence currents at the short-circuit point, respectively. , , For the network's order self-impedance, , , The sequence combination impedances from the power source to the short-circuit point are as follows: Given the voltage at the short-circuit point before the short circuit, the calculated sequence currents are combined into a short-circuit current vector.

[0093] Step 3043: If the target power scenario category is a load adjustment scenario category, then the node current vector and loop voltage vector are used as initial boundary conditions for power flow iteration calculation. During the iteration calculation, the node current vector and loop voltage vector are combined with intermediate results to obtain the voltage distribution vector and power distribution vector of each node.

[0094] Furthermore, when the target power scenario category is a load adjustment scenario category, the power management system uses the node current vector and loop voltage vector as initial boundary conditions and employs the Newton-Raphson method for iterative power flow calculation. In each iteration, based on the node power balance equation and loop voltage equation, the node current vector and loop voltage vector are combined with intermediate calculation results to continuously update the voltage magnitude and phase angle of each node until the convergence condition is met, thus obtaining the voltage distribution vector and power distribution vector of each node.

[0095] The node power balance equation is: .

[0096] .

[0097] in, and They are nodes and The conductivity and susceptance between them , The power-related first term in the loop voltage vector One element, The power-related term in the node current vector Each node has an element. By iteratively solving the above equations, the node voltage and node power are updated. After multiple iterations, when the power deviation and voltage deviation meet the set convergence conditions, the voltage distribution vector and power distribution vector of each node are obtained.

[0098] The embodiments of the present invention perform targeted processing according to the characteristics of different scenarios, which can accurately mine the inherent physical relationships of power data under different operating states, improve the accuracy and effectiveness of mapping the physical laws of power data, and provide a reliable basis for evaluating the consistency of power data in different scenarios.

[0099] In one embodiment, steps 3044 to 3047 are described as follows: Step 3044: Based on the scenario constraint adaptation conditions, perform constraint judgment on the physical law mapping results to determine the first number of elements in the node current vector and / or loop voltage vector that do not meet the scenario constraint adaptation conditions.

[0100] Optionally, the power management system checks the node current vector and loop voltage vector in the physical law mapping results one by one according to the scenario constraint adaptation conditions of different target power scenario categories. Among them, each target power scenario category has its specific constraints. For example, in the normal operation scenario category, the node voltage amplitude must be within a certain percentage range of the rated value, and the current magnitude must also conform to the normal load conditions; in the fault scenario category, the short-circuit current must meet the calculation result range of a specific fault type; in the load adjustment scenario category, the changes in power, voltage and other parameters after load change must conform to certain rules.

[0101] Therefore, the power management system compares the elements of the node current vector and the loop voltage vector in the physical law mapping result with the corresponding scenario constraint adaptation conditions, counts the number of elements that do not meet the conditions, and records them as the first quantity. .

[0102] In one embodiment, if the current scenario is determined to be a normal operation scenario, the current vector of a certain node is: =[50, -10, 150], loop voltage vector is =[0.5, 0.8, 1.2]. Under normal operating conditions, the scenario constraints stipulate that the node current amplitude cannot exceed the rated current of 100A, and the loop voltage drop cannot exceed 1kV. In the node current vector, 150>100, meaning one element does not meet the conditions; in the loop voltage vector, 1.2>1, 0.8>1, meaning two elements do not meet the conditions. Therefore, the first quantity... =1+2=3.

[0103] Step 3045: Project the node current vector and / or loop voltage vector into a multidimensional space to obtain the projected data vector. When traversing each path based on the topology, verify the projected data vector to determine the second number of elements in the node current vector and / or loop voltage vector that do not conform to Hough's current law and Hough's voltage law.

[0104] Furthermore, the power management system projects the node current vectors and loop voltage vectors into a multi-dimensional space, constructing a spatial coordinate system with parameters such as current and voltage as dimensions, thus obtaining the projected data vector. Further, based on the power system's topology, the power management system traverses each path in the topology. During this traversal, the projected data vector is validated. According to Kirchhoff's Current Law (KCL) and Voltage Law (KVL), it checks whether the algebraic sum of the currents at each node is zero, and whether the algebraic sum of the voltage drops in each loop is zero. The number of elements in the projected data vector that do not satisfy the two laws is counted and denoted as the second quantity. .

[0105] Continuing with the example of the aforementioned power system, the node current vector =[50, -10, 150], loop voltage vector is =[0.5, 0.8, 1.2]. Project this onto a three-dimensional space with node current and loop voltage as dimensions (assuming three dimensions: node 1 current, node 2 current, and loop 1 voltage). Through topological traversal, it is found that in a certain loop, the sum of voltage drops calculated from the loop voltage vector is 0.5 + 0.8 + 1.2 = 2.5 ≠ 0, which does not satisfy KVL; at a certain node, the sum of related currents in the node current vector is 50 - 10 + 150 = 190 ≠ 0, which does not satisfy KCL. Statistically, there are two elements that do not satisfy KCL and KVL, namely the second quantity. =2.

[0106] Step 3046: Based on historical power data under the current operating state, perform dynamic trend prediction on the node current vector and / or loop voltage vector, and determine the third number of elements in the node current vector and / or loop voltage vector whose trend prediction deviation value is greater than a preset deviation threshold.

[0107] Furthermore, the power management system utilizes historical power data under current operating conditions and employs algorithms such as ARIMA (Autoregressive Integral Moving Average) in time series analysis to dynamically predict the trends of elements in the node current vector and loop voltage vector. For each element, the deviation between its predicted and actual values ​​is calculated to obtain the trend prediction deviation value. A preset deviation threshold is set; when the trend prediction deviation value exceeds this threshold, the element is considered to not conform to the trend change pattern. Therefore, the number of elements with trend prediction deviation values ​​exceeding the preset deviation threshold is counted and denoted as the third number. .

[0108] Continuing with the above embodiment, hourly data on the current of a certain node and the voltage of a certain loop were collected over the past week. The current vector of the current node was then analyzed using the ARIMA model. =[50, -10, 150], loop voltage vector is The elements in the range [0.5, 0.8, 1.2] are used for prediction. The preset deviation threshold is 10% of the actual value. Calculations show that the predicted value of the third element 150 in the node current vector is 100, with a deviation of 50% > 10%; the predicted value of the second element 0.8 in the loop voltage vector is 0.6, with a deviation of 33.3% > 10%. Therefore, the third quantity... =2.

[0109] Step 3047: Perform consistency prediction based on the sum of the first quantity, the second quantity, and the third quantity, and output the preliminary consistency assessment results.

[0110] Furthermore, the power management system calculates the first quantity. Second quantity and the third quantity sum Set a consistency threshold. ,when > At that time, the output power data showed inconsistent preliminary consistency assessment results; when At that time, the output power data has a consistent preliminary consistency assessment result.

[0111] Continuing with the above embodiments, in the previous examples, =3, =2, =2, then =3+2+2=7. Set a consistency threshold. =5, because 7>5, so the power management system outputs the preliminary consistency assessment result that there is inconsistency in the power data.

[0112] The embodiments of this invention conduct in-depth analysis of power data from multiple dimensions and construct a comprehensive power data consistency assessment system. Therefore, it can accurately assess the consistency of power data and effectively capture potential problems in terms of data adaptation to scenarios, adherence to physical laws, and trend changes, thereby improving the accuracy and reliability of power data consistency assessment.

[0113] In one embodiment, steps 401 to 406 are described as follows: Step 401: Construct a scenario result correlation matrix based on the preliminary consistency assessment results under different target power scenario categories. Each matrix element in the scenario result correlation matrix represents the preliminary consistency assessment result of each power data indicator under each target power scenario category, with a value of 1 for consistency and 0 for inconsistency.

[0114] Optionally, after obtaining the preliminary consistency assessment results for different target power scenario categories (normal operation scenario category, fault scenario category, and load adjustment scenario category), the power management system constructs a scenario result correlation matrix for each power data indicator (such as voltage, current, power, frequency, etc.). The rows of the matrix correspond to different target power scenario categories, and the columns correspond to each power data indicator. If the preliminary consistency assessment result for a power data indicator under a certain target power scenario category is "consistent," then the element at the corresponding position in the matrix takes a value of 1; if it is "inconsistent," then the value is 0. In this way, the preliminary assessment results of all power data indicators under different scenarios are presented intuitively in matrix form.

[0115] Continuing with the monitoring of a regional power grid using a power management system, the power data indicators involved include four metrics: node 1 voltage, node 2 current, line 1 power, and system frequency. Under the normal operation scenario, the preliminary consistency assessment results are: node 1 voltage is consistent, node 2 current is consistent, line 1 power is inconsistent, and system frequency is consistent. Under the fault scenario, the preliminary consistency assessment results are: node 1 voltage is inconsistent, node 2 current is inconsistent, line 1 power is inconsistent, and system frequency is consistent. Under the load adjustment scenario, the preliminary consistency assessment results are: node 1 voltage is consistent, node 2 current is consistent, line 1 power is consistent, and system frequency is consistent. The constructed scenario result correlation matrix... for: .

[0116] The first row corresponds to the normal operation scenario category, the second row corresponds to the fault scenario category, and the third row corresponds to the load adjustment scenario category; the first column corresponds to the node 1 voltage, the second column corresponds to the node 2 current, the third column corresponds to the line 1 power, and the fourth column corresponds to the system frequency.

[0117] Step 402: For each power data indicator, calculate the indicator difference degree of the preliminary consistency assessment results under different target power scenario categories, and determine the power data indicators with indicator difference degrees greater than the preset difference threshold as key difference indicators.

[0118] Furthermore, for each power data indicator, the power management system calculates the degree of difference in the indicator of its preliminary consistency assessment results under different target power scenario categories.

[0119] Optionally, in this embodiment of the invention, the calculation of the index difference degree adopts the information entropy method. Let a certain power data index be... The preliminary consistency assessment results under each target power scenario category constitute a vector. (in (Values ​​can be 0 or 1), first calculate the probability of each value. , Then the degree of difference of this indicator The calculation formula is Set a preset difference threshold. When the difference of a certain power data indicator > When this occurs, the indicator is identified as a key difference indicator.

[0120] Continuing with the voltage index of node 1 in the correlation matrix of the above scenario results, its corresponding vector is: =[1, 0, 1], n=3, the number of 0s is 1, and the number of 1s is 2. Then... , Calculate the degree of difference according to the formula. 0.918. Set the preset difference threshold. =0.8, because Since the value is greater than 0.8, the voltage index at node 1 is the key difference indicator. Similarly, the vector corresponding to the current index at node 2 is calculated. Differences in the range [1, 0, 1] >0.8, the current index at node 2 is the key difference indicator; the power index at line 1 corresponds to the vector. =[0, 0, 1], 0.918 > 0.8, the power index of Line 1 is the key difference indicator; the corresponding vector for the system frequency index. =[1, 1, 1], =0<0.8, the system frequency index is not a key difference index.

[0121] Step 403: For key difference indicators, determine the weighting factor for each target power scenario category based on the correlation pattern between the preliminary consistency assessment results under different target power scenario categories.

[0122] Furthermore, for key difference indicators, the power management system analyzes the correlation patterns between their preliminary consistency assessment results under different target power scenario categories. Principal component analysis (PCA) is used to construct a data matrix from the preliminary consistency assessment results of key difference indicators under different target power scenario categories. The contribution rate of each principal component is calculated using PCA, and the weighting factor for each target power scenario category is determined based on the principal component contribution rate. Let the weighting factors for the normal operation scenario category, fault scenario category, and load adjustment scenario category be as follows: , , PCA calculation makes =1, and the weight allocation can reflect the correlation characteristics of the evaluation results of key difference indicators in different scenarios to the greatest extent.

[0123] Continuing with the three key difference indicators identified above—node 1 voltage, node 2 current, and line 1 power—we construct a data matrix based on their preliminary consistency assessment results across different target power scenario categories: .

[0124] For matrix Principal component analysis was performed, and the contribution rates of the first principal component (0.6), the second principal component (0.3), and the third principal component (0.1) were calculated. Based on these principal component contribution rates, the weighting factors for the normal operation scenario categories were determined. =0.6, the weight factor for the fault scenario category =0.3, the weighting factor for the load adjustment scenario category. =0.1.

[0125] Step 404: Based on the weight factors of each target power scenario category, fuse the preliminary consistency evaluation results under each target power scenario category to obtain the fused evaluation value of each power data indicator.

[0126] Furthermore, the power management system merges the preliminary consistency assessment results for each target power scenario category based on the weighting factors of each category. For each power data indicator, the merged assessment value... The calculation formula is ,in, , , These represent the preliminary consistency assessment results (values ​​of 0 or 1) for the indicator under the normal operation scenario, fault scenario, and load adjustment scenario categories, respectively. The fused assessment value for each power data indicator is then calculated using this formula, comprehensively reflecting the consistency of the indicator across different scenarios.

[0127] Continuing with the voltage parameters for node 1, =1, =0, =1, =0.6, =0.3, =0.1, calculate the fusion evaluation value according to the formula. =0.6*1+0.3*0+0.1*1=0.7. Similarly, calculate the fused evaluation value of the current index at node 2. =0.6*1+0.3*0+0.1*1=0.7; The combined evaluation value of the power index of Line 1. =0.6*0+0.3*0+0.1*1=0.1; Since the system frequency index is not a key difference indicator, the average value of its evaluation results in each scenario is directly taken as the fused evaluation value. = (1+1+1) / 3 = 1.

[0128] Step 405: If the fused evaluation value of each power data indicator is greater than or equal to the preset evaluation threshold, then the final consistency evaluation result is determined to be consistent.

[0129] Furthermore, the power management system sets a preset evaluation threshold. When the merged evaluation value of each power data indicator is greater than or equal to the preset evaluation threshold When the time is right, it indicates that all power data indicators show a high degree of consistency after comprehensively considering the evaluation results under different target power scenario categories. At this time, the final consistency evaluation result is determined to be "consistent".

[0130] Continuing with the above embodiments, a preset evaluation threshold is set. =0.6, the fused evaluation value of the voltage index at node 1 =0.7>0.6, the fused evaluation value of the node 2 current index =0.7>0.6, the combined evaluation value of the power index of line 1 =0.1<0.6, the fused evaluation value of the system frequency index =1>0.6, because the power index of line 1 after fusion is less than the preset evaluation threshold, so the condition is not met at present.

[0131] Assuming that the combined evaluation value of all indicators is greater than or equal to 0.6, such as =0.7, =0.7, =0.6, If the value is 0.8, the power management system determines the final consistency assessment result as "consistent".

[0132] Step 406: If the fusion evaluation value of a preset number of power data indicators is less than the preset evaluation threshold, then the final consistency evaluation result is determined to be inconsistent.

[0133] Furthermore, the power management system sets preset quantities. When the combined evaluation value of a preset number of power data indicators is less than a preset evaluation threshold, If this occurs, it indicates that there are many inconsistencies in the comprehensive evaluation of power data indicators. In this case, the final consistency evaluation result is determined to be "inconsistent".

[0134] Continuing with the above embodiments, a preset quantity is used. =1, the combined evaluation value of line 1 power index =0.1<0.6, which satisfies the condition that there is one power data indicator whose fused evaluation value is less than the preset evaluation threshold. Therefore, the power management system determines the final consistency evaluation result as "inconsistent".

[0135] This invention, through constructing a scenario result correlation matrix, determining key difference indicators, setting weight factors, integrating evaluation results, and making judgments, forms a scientific and rigorous final consistency evaluation system for power data. Therefore, it can comprehensively explore the inherent connections and differences of power data in different scenarios, effectively improving the accuracy and reliability of power data consistency evaluation.

[0136] Furthermore, the deep learning-based power data consistency assessment system provided by the present invention will be described below. The deep learning-based power data consistency assessment system described below can be referred to in correspondence with the deep learning-based power data consistency assessment method described above.

[0137] Optional, refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of the deep learning-based power data consistency assessment system provided by the present invention. The deep learning-based power data consistency assessment system includes: The scenario analysis module 210 is used to perform scenario analysis based on the operating status information of the power system under the current operating state, and to determine the target power scenario category under the current operating state. The model matching module 220 is used to match the target power scenario category in a pre-established power model library to obtain the target power model; The consistency prediction module 230 is used to input the power data of the power system under the current operating state into the target power model and obtain the preliminary consistency evaluation result output by the target power model; the preliminary consistency evaluation result characterizes whether the power data meet the physical laws under the target power scenario category; The consistency assessment module 240 is used to determine the final consistency assessment result of power data based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories.

[0138] This invention categorizes power scenarios based on operational status information and selects corresponding power models for evaluation. The power models for each scenario reflect the unique physical laws and operational rules of that scenario, ensuring that power data is evaluated in accordance with actual conditions under different operating circumstances. After obtaining preliminary results from the evaluation by power scenario category, a comprehensive judgment is made, fully considering the correlation between data consistency across different power scenario categories. This allows for more accurate identification of consistency issues in power data across different scenarios, effectively avoiding evaluation biases caused by insufficient scenario adaptability, and thus providing a reliable basis for stable operation and optimization decisions of the power system.

[0139] Please see Figure 3 , Figure 3 An embodiment diagram of an electronic device provided in accordance with the present invention. For example... Figure 3 As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, it performs the following steps: Based on the operating status information of the power system under the current operating state, scenario analysis is performed to determine the target power scenario category under the current operating state; The target power model is obtained by matching the target power scenario category with a pre-established power model library. The power data of the power system under the current operating state is input into the target power model to obtain the preliminary consistency assessment results output by the target power model; the preliminary consistency assessment results characterize whether the power data meet the physical laws under the target power scenario category; Based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories, the final consistency assessment results of power data are determined.

[0140] Please see Figure 4 , Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with an embodiment of the present invention is shown. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400 on which a computer program 311 is stored. When the computer program 311 is executed by a processor, it performs the following steps: Based on the operating status information of the power system under the current operating state, scenario analysis is performed to determine the target power scenario category under the current operating state; The target power model is obtained by matching the target power scenario category with a pre-established power model library. The power data of the power system under the current operating state is input into the target power model to obtain the preliminary consistency assessment results output by the target power model; the preliminary consistency assessment results characterize whether the power data meet the physical laws under the target power scenario category; Based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories, the final consistency assessment results of power data are determined.

[0141] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the deep learning-based power data consistency assessment method provided by the above methods, the method comprising: Based on the operating status information of the power system under the current operating state, scenario analysis is performed to determine the target power scenario category under the current operating state; The target power model is obtained by matching the target power scenario category with a pre-established power model library. The power data of the power system under the current operating state is input into the target power model to obtain the preliminary consistency assessment results output by the target power model; the preliminary consistency assessment results characterize whether the power data meet the physical laws under the target power scenario category; Based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories, the final consistency assessment results of power data are determined.

[0142] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deep learning-based method for evaluating the consistency of power data, characterized in that, include: Based on the operating status information of the power system under the current operating state, scenario analysis is performed to determine the target power scenario category under the current operating state; The target power model is obtained by matching the target power scenario category with a pre-established power model library. The power data of the power system under the current operating state is input into the target power model to obtain the preliminary consistency assessment result output by the target power model; the preliminary consistency assessment result characterizes whether the power data meet the physical laws under the target power scenario category; Based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories, the final consistency assessment results of power data are determined.

2. The power data consistency assessment method based on deep learning according to claim 1, characterized in that, The correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories determines the final consistency assessment result of the power data, including: A scenario result association matrix is ​​constructed based on the preliminary consistency assessment results under different target power scenario categories. Each matrix element in the scenario result association matrix represents the preliminary consistency assessment result of each power data indicator under each target power scenario category, with a value of 1 for consistency and 0 for inconsistency. For each power data indicator, calculate the indicator difference degree of the preliminary consistency assessment results under different target power scenario categories, and determine the power data indicators with the indicator difference degree greater than the preset difference threshold as key difference indicators. For the key difference indicators, a weighting factor for each target power scenario category is determined based on the correlation pattern between the preliminary consistency assessment results under different target power scenario categories. Based on the weighting factors of each target power scenario category, the preliminary consistency assessment results under each target power scenario category are fused to obtain the fused assessment value of each power data indicator; If the fused evaluation value of each power data indicator is greater than or equal to the preset evaluation threshold, then the final consistency evaluation result is determined to be consistent. If the fused evaluation value of a preset number of power data indicators is less than a preset evaluation threshold, then the final consistency evaluation result is determined to be inconsistent.

3. The power data consistency assessment method based on deep learning according to claim 1, characterized in that, The specific steps for processing the power data under the current operating state based on the target power model and outputting the preliminary consistency assessment result of the power data under the current operating state include: In the target power model, power data is split according to different electrical components and lines to obtain power data vectors; the power data vectors include sub-vectors related to nodes and loops; the power data includes voltage data, current data, power data and frequency data from various data sources; Based on the topology of the power system, a node-branch association matrix and a loop-branch association matrix are constructed. For the matrix elements in the node-branch association matrix, if a node is connected to a branch and current flows out of the node, it is 1; if current flows in, it is -1; and if they are not connected, it is 0. For the matrix elements in the loop-branch association matrix, if a loop contains a branch and the direction is the same, it is 1; if the direction is opposite, it is -1; and if it does not contain a branch, it is 0. The node current vector is determined based on the node branch correlation matrix and the power data vector, and the loop voltage vector is determined based on the loop branch correlation matrix and the power data vector. Physical law mapping is performed based on the node current vector and / or the loop voltage vector to obtain the physical law mapping result. Consistency prediction is then performed based on the physical law mapping result and the scenario constraint adaptation conditions to output the preliminary consistency evaluation result.

4. The power data consistency assessment method based on deep learning according to claim 3, characterized in that, The consistency prediction based on the physical law mapping result and scene constraint adaptation conditions, and the output of the preliminary consistency evaluation result, include: Based on the scenario constraint adaptation conditions, the physical law mapping results are constrained and determined to determine the first number of elements in the node current vector and / or the loop voltage vector that do not conform to the scenario constraint adaptation conditions. The node current vector and / or the loop voltage vector are projected into a multidimensional space to obtain a projected data vector. When traversing each path based on the topology, the projected data vector is verified to determine the second number of elements in the node current vector and / or the loop voltage vector that do not conform to Hough's current law and Hough's voltage law. Based on historical power data under the current operating state, dynamic trend prediction is performed on the node current vector and / or the loop voltage vector to determine the third number of elements in the node current vector and / or the loop voltage vector whose trend prediction deviation value is greater than a preset deviation threshold. Consistency prediction is performed based on the sum of the first quantity, the second quantity, and the third quantity, and the preliminary consistency assessment result is output.

5. The power data consistency assessment method based on deep learning according to claim 3, characterized in that, The target power scenario categories include normal operation scenario categories, fault scenario categories, and load adjustment scenario categories; The target power models corresponding to the normal operation scenario category, the fault scenario category, and the load adjustment scenario category are respectively the steady-state power model, the short-circuit analysis model, and the power flow calculation model. The physical law mapping based on the node current vector and / or the loop voltage vector to obtain the physical law mapping result includes: If the target power scenario category is a normal operation scenario category, then the node power vector is obtained by mapping the physical laws based on the first node voltage amplitude and first node voltage phase angle of each node, the second node voltage amplitude and second node voltage phase angle of each node's adjacent nodes, branch conductance, branch susceptance, and power-related elements in the node current vector. If the target power scenario category is a fault scenario category, then the loop voltage vector is converted according to the short circuit type and short circuit location of the short circuit point to obtain the short circuit current vector; If the target power scenario category is a load adjustment scenario category, then the node current vector and the loop voltage vector are used as initial boundary conditions for power flow iteration calculation. During the iteration calculation, the node current vector and the loop voltage vector are combined with intermediate results to obtain the voltage distribution vector and power distribution vector of each node.

6. The power data consistency assessment method based on deep learning according to any one of claims 1 to 5, characterized in that, The operational status information includes equipment operating parameters, changes in power grid topology, and load fluctuations. The scenario analysis based on the operating status information of the power system under the current operating state, to determine the target power scenario category under the current operating state, includes: Identify abnormal devices based on the operating parameters of each operating device; An anomaly index is determined based on the equipment operating parameters of each abnormal device in conjunction with its historical operating parameters, and the impact value of the equipment anomaly is determined based on the importance coefficient of each abnormal device in the power system. The intensity of the topology change is determined based on the change in connectivity of each node in the power grid topology before and after the topology change, as well as the initial connectivity value of each node before the change. The load fluctuation trend value is determined based on the load change rate within a preset time interval, combined with the maximum load value, minimum load value, and average load value within the preset time interval. Based on the abnormal impact value of the equipment, the intensity of the topology change, and the load fluctuation trend value, scenario analysis is performed to determine the target power scenario category under the current operating state.

7. The power data consistency assessment method based on deep learning according to claim 6, characterized in that, The scenario analysis, based on the abnormal impact value of the equipment, the intensity of the topology change, and the load fluctuation trend value, determines the target power scenario category under the current operating state, including: A correlation analysis is performed based on the impact value of the equipment anomaly and the intensity of the topology change to determine the first degree of correlation between the topology change and the equipment anomaly. A correlation analysis is then performed based on the impact value of the equipment anomaly and the load fluctuation trend value to determine the second degree of correlation between the load fluctuation and the equipment anomaly. For cases where the abnormal impact value of the device exceeds the preset abnormal impact threshold: If the first correlation degree is greater than the first preset correlation threshold, then the target power scenario category is determined to be a fault scenario category; If the first correlation degree is less than or equal to the first preset correlation threshold, and the load fluctuation trend value is greater than the preset load fluctuation threshold, then the target power scenario category is determined to be a fault scenario category. If the first correlation degree is less than or equal to the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is less than or equal to the second preset correlation threshold, then the target power scenario category is determined to be the load adjustment scenario category. If the first correlation degree is less than or equal to the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is greater than the second preset correlation threshold, then the target power scenario category is determined to be a fault scenario category. For cases where the abnormal impact value of the device is less than or equal to the preset abnormal impact threshold: If the first correlation degree is less than or equal to the first preset correlation threshold, then the target power scenario category is determined to be a normal operation scenario category; If the first correlation degree is greater than the first preset correlation threshold, and the load fluctuation trend value is greater than the preset load fluctuation threshold, then the target power scenario category is determined to be the load adjustment scenario category. If the first correlation degree is greater than the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is greater than the second preset correlation threshold, then the target power scenario category is determined to be the load adjustment scenario category. If the first correlation degree is greater than the first preset correlation threshold, and the load fluctuation trend value is less than or equal to the preset load fluctuation threshold, and the second correlation degree is less than or equal to the second preset correlation threshold, then the target power scenario category is determined to be the normal operation scenario category.

8. A power data consistency assessment system based on deep learning, characterized in that, The method is applied to the deep learning-based power data consistency assessment method as described in any one of claims 1 to 7; the deep learning-based power data consistency assessment system comprises: The scenario analysis module is used to perform scenario analysis based on the operating status information of the power system under the current operating state, and to determine the target power scenario category under the current operating state; The model matching module is used to match the target power scenario category in a pre-established power model library to obtain the target power model; The consistency prediction module is used to input the power data of the power system under the current operating state into the target power model to obtain the preliminary consistency evaluation result output by the target power model; the preliminary consistency evaluation result characterizes whether the power data meet the physical laws under the target power scenario category; The consistency assessment module is used to determine the final consistency assessment result of power data based on the correlation and difference analysis between the preliminary consistency assessment results under different target power scenario categories.

9. An electronic device, comprising: Memory, used to store computer software programs; A processor for reading and executing the computer software program, characterized in that, when the processor executes the computer software program, it implements the deep learning-based power data consistency assessment method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, wherein a computer software program is stored therein, characterized in that, When the computer software program is executed by the processor, it implements the deep learning-based power data consistency assessment method as described in any one of claims 1 to 7.

Images