Power transmission line intelligent management and control method and system based on power distribution data feedback

Abstract

The invention discloses a power transmission line intelligent management and control method and system based on power distribution data feedback, and relates to the technical field of power transmission line management and control, and the method comprises the steps: integrating multi-source data, constructing a historical environment-electrical coupling feature matrix, extracting time sequence features through a long and short term memory network, and capturing a spatial distribution rule through combining a convolutional neural network; and establishing a dynamic line loss prediction model to simulate minimum line loss values in different environment scenes. And detecting an abnormal time period through an isolated forest algorithm, screening key damage factors in combination with Pearson correlation analysis and SHAP value analysis, dynamically generating and executing a loss reduction strategy priority event, and optimizing model parameters by using error back propagation and reinforcement learning. And finally, generating a typical environment scene loss reduction case library based on the optimization model, and simulating strategy effectiveness under an extreme condition through a digital twin technology to form a reusable decision rule set, thereby improving the operation efficiency of a power transmission line and the reliability of power supply.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission line control and management, and specifically relates to an intelligent control method and system for transmission lines based on distribution data feedback. Background Art

[0002] With the continuous expansion of the scale of the power system, the operating environment of transmission lines has become increasingly complex. Traditional control and management methods are difficult to meet the requirements. The intelligent control method and system for transmission lines based on distribution data feedback can accurately grasp the status of transmission lines with the help of real-time feedback of distribution data, timely discover potential faults and risks, optimize the operation and maintenance strategies of transmission lines, improve power supply reliability, reduce operation and maintenance costs, meet the development requirements of smart grids, and ensure stable power supply.

[0003] A Chinese patent application with the publication number CN118551177B discloses a line loss control method, device and electronic equipment for distribution lines. First, based on the operating parameters of the target line, line loss-related analysis can be performed on at least one distribution line to be analyzed to determine whether each distribution line to be analyzed belongs to a high-loss distribution line. Secondly, when there is a distribution line to be analyzed that belongs to a high-loss distribution line among at least one distribution line to be analyzed, for each distribution line to be analyzed that belongs to a high-loss distribution line, determination processing of the loss reduction potential and high-loss causes of the distribution line to be analyzed is performed to obtain the corresponding loss reduction potential parameters and high-loss cause information. Then, based on the corresponding loss reduction potential parameters and high-loss cause information, line loss control operations can be performed. Based on the above content, the problem of relatively low reliability of line loss control in existing technologies can be improved.

[0004] In the above-mentioned existing technology, based on the operating parameters of the target line, line loss-related analysis is performed on at least one distribution line to be analyzed to determine whether each distribution line to be analyzed belongs to a high-loss distribution line. Then, by performing determination processing on the loss reduction potential and high-loss causes of the distribution line to be analyzed, the corresponding loss reduction potential parameters and high-loss cause information are obtained. Finally, based on the corresponding loss reduction potential parameters and high-loss cause information, line loss control operations are performed. However, when the above-mentioned existing technology controls and manages transmission lines, it lacks the analysis of the influence of environmental factors on the loss reduction potential and high-loss causes of the lines, and cannot ensure the accuracy of transmission line fault detection, thus unable to ensure the accuracy of transmission line control and management.

[0005] Therefore, the present invention provides an intelligent control method and system for transmission lines based on distribution data feedback to solve the above-mentioned problems. Summary of the Invention

[0006] To solve the above technical problems, an intelligent control method and system for transmission lines based on distribution data feedback are provided. This technical solution solves the problem in the above-mentioned background technology that when the existing technology controls transmission lines, it lacks the analysis of the impact of environmental factors on the line loss reduction potential and the causes of high line losses, and cannot ensure the accuracy of transmission line fault detection, thus unable to ensure the accuracy of transmission line control.

[0007] To achieve the above objectives, the technical solution adopted in the present invention is as follows: An intelligent control method for transmission lines based on distribution data feedback, characterized by comprising: S1. Obtain the historical operation parameters of each transmission line to be analyzed, and then obtain the historical meteorological data, historical geographical data, and historical humanistic data of the area where each transmission line to be analyzed is located, and perform spatio-temporal alignment with the historical operation parameters of the line to construct a historical environment - electrical coupling feature matrix; S2. Use a long short-term memory network to extract the historical environment time series features in the historical environment - electrical coupling feature matrix, combine a convolutional neural network to capture the spatial distribution law, establish a dynamic line loss prediction model, and obtain the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model; S3. Obtain the historical actual line loss value and compare the difference with the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model. According to the difference comparison result, judge whether to trigger the isolation forest algorithm to detect abnormal periods of environmental parameters, and then screen key loss-causing factors through Pearson correlation analysis; S4. Obtain environmental prediction data and the real-time operation parameters of the line, and combine key loss-causing factors and the dynamic line loss prediction model to dynamically generate loss reduction strategy priority events; S5. Execute the loss reduction strategy priority events, and obtain the actual line loss data after execution. Synchronously perform error backpropagation on the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and use the reinforcement learning algorithm to optimize the dynamic line loss prediction model; S6. Based on the optimized dynamic line loss prediction model, generate a loss reduction case library for typical environmental scenarios, and simulate the effectiveness of strategies under extreme conditions through digital twin technology to form a reusable decision rule set.

[0008] In an optional embodiment, step S1 specifically includes: S1.1. Collect the historical operation parameters of each transmission line to be analyzed through the SCADA system. The historical operation parameters of the line include historical current, historical voltage, historical load rate, and historical fault records; S1.2. Collect historical temperature, historical humidity, historical wind speed, historical rainfall, and historical icing degree from the weather stations in the area where each transmission line to be analyzed is located to obtain historical meteorological data; S1.3. Collect historical terrain slope, historical vegetation coverage rate, historical soil resistivity, and historical altitude from the weather stations in the area where each transmission line to be analyzed is located; S1.4. Collect historical industrial pollution concentration, historical traffic noise decibel value, and historical electromagnetic radiation intensity of wireless base stations from the weather stations in the area where each transmission line to be analyzed is located to obtain historical humanistic data; S1.5. Apply moving average filtering to the historical meteorological data, historical geographical data, and historical line operation parameters to eliminate high-frequency noise, and apply median filtering to the historical humanistic data to suppress mutation interference; S1.6. Use the KNN interpolation method to fill in the missing values in the historical meteorological data, and at the same time use linear regression to fill in the missing values of the geographical data; S1.7. Eliminate abnormal data points beyond the normal range based on the 3σ principle, and project all the data processed in steps 1.4 to 1.6 onto the UTM coordinate system. Taking the center point of the transmission line as the origin, establish a local coordinate grid; S1.8. Based on the GPS timing system, align the historical meteorological data, historical geographical data, historical humanistic data, and historical line operation parameters projected onto the UTM coordinate system according to the minute-level time stamps; S1.9. Use principal component analysis for dimensionality reduction, retain the principal components with a cumulative variance contribution rate greater than or equal to 95%, and then generate a historical environment - electrical coupling feature matrix with a dimension less than or equal to 50.

[0009] In an alternative embodiment, step S2 specifically includes: S2.1. Design a double-layer long short-term memory network, set the input layer dimension to 50, the number of hidden layer units to 128, the activation function to tanh, and the output layer predicts the line loss sequence for the next 24 hours; S2.2. Divide the historical data in the historical environment - electrical coupling feature matrix into training samples according to a 24-hour sliding window, set the step size to 1 hour, and generate N time series segments; S2.3. Use a 3-layer convolutional neural network, set the convolutional kernel size to 3×3, the step size to 1, and the padding method to SAME to extract the spatial distribution features of the historical geographical data in the historical environment - electrical coupling feature matrix; S2.4. Concatenate the N time series segments generated by the double-layer long short-term memory network and the spatial distribution features obtained by the 3-layer convolutional neural network, input them into the fully connected layer, and the output layer generates the dynamic line loss prediction value, thus completing the construction of the dynamic line loss prediction model; S2.5. Obtain the historical environmental sensitivity coefficient matrix based on the N time series segments generated by the double-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network, and use the historical environmental sensitivity coefficient matrix to correct the historical line operation parameters in the historical environment-electrical coupling feature matrix; S2.6. Set the loss function and dynamically adjust the learning rate using the Adam algorithm. Then, input the N time series segments generated by the corrected double-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network into the dynamic line loss prediction model to train the model; S2.7. Simulate and generate the environmental-electrical coupling feature simulation matrix under different environmental scenarios, and input the simulated meteorological data, simulated geographical data, and simulated human data in the environmental-electrical coupling feature simulation matrix into the trained dynamic line loss prediction model to obtain the minimum line loss values under different environmental scenarios output by the trained dynamic line loss prediction model.

[0010] In an alternative embodiment, step S3 specifically includes: Obtain the difference between the historical actual line loss value corresponding to the N time series segments and the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model; If the difference between the historical actual line loss value and the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model is greater than or equal to 0.15 times the minimum line loss value, trigger the isolation forest algorithm to obtain the anomaly score for each data point; Filter the simulated meteorological data, simulated geographical data, and simulated human data in the environmental-electrical coupling feature simulation matrix corresponding to the data points with an anomaly score higher than T threshold = μ(s) + 3σ(s), to obtain the first abnormal data, where μ(s) is the global mean of the minimum line loss values corresponding to all data points, σ(s) is the global standard deviation of the minimum line loss values corresponding to all data points, and T threshold is the abnormal data screening threshold; Based on the first abnormal data and its corresponding minimum line loss value, obtain the first correlation coefficient, and at the same time obtain the first abnormal data with the first correlation coefficient greater than or equal to 0.8; Through the SHAP value analysis model, quantify the contribution degree of each factor in the first abnormal data to the line loss, and mark the factor with the largest contribution degree as the key loss-causing factor; Among them, the calculation formula for triggering the isolation forest algorithm to obtain the anomaly score for each data point is: In the formula, is the anomaly score under the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment, is the historical actual line loss value at the i-th time series segment corresponding to the power transmission circuit under the t-th environmental scenario, is the mean value under the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment, is the standard deviation of the minimum line loss value under the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment.

[0011] In an alternative embodiment, step S4 specifically includes: Obtain meteorological data, geographical data, and human data for the next 24 hours from the meteorological bureau to obtain environmental prediction data, and synchronously obtain the real-time operation parameters of each power transmission line to be analyzed; Input the environmental prediction data and the real-time operation parameters of the line into the dynamic line loss prediction model to obtain the minimum line loss value corresponding to the environmental prediction data and the real-time operation parameters of the line; Compare the environmental prediction data and the real-time operation parameters of the line with the key loss-causing factors to dynamically generate a dynamic loss-causing value; Perform a weighted sum of the minimum line loss value and the dynamic loss-causing value to obtain a reference value for the loss reduction strategy; Based on the reference value for the loss reduction strategy, dynamically generate a priority event for the loss reduction strategy.

[0012] In an alternative embodiment, step S5 specifically includes: After executing the priority event for the loss reduction strategy, obtain the actual line loss value of the power transmission line corresponding to the priority event for the loss reduction strategy during the abnormal period to obtain the actual line loss data after execution; Synchronously obtain the actual environment-electrical coupling characteristic matrix of the power transmission line corresponding to the priority event for the loss reduction strategy; Import the data in the actual environment-electrical coupling characteristic matrix into the dynamic line loss prediction model to obtain the predicted line loss data output by the dynamic line loss prediction model; Compare the predicted line loss value included in the predicted line loss data output by the dynamic line loss prediction model with the actual line loss value to obtain the difference between the two for error backpropagation, and synchronously use the Adam optimizer to update the parameters in the dynamic line loss prediction model to optimize the dynamic line loss prediction model using the reinforcement learning algorithm.

[0013] In an alternative embodiment, step S6 specifically includes: Based on the optimized dynamic line loss prediction model, obtain the minimum line loss values under different environmental scenarios; Perform a typical scenario division on the meteorological data, geographical data, human data, and line operation parameters corresponding to the minimum line loss values under different environmental scenarios, and combine with the priority event for the loss reduction strategy to generate a typical environmental scenario loss reduction case library; Build a 3D line model based on the Unreal Engine, and use physical simulation to inject extreme environmental parameters and extreme line operation parameters into the 3D line model to simulate the effectiveness of strategies under extreme conditions; According to the effectiveness of the strategies under extreme conditions, obtain the effective loss reduction events under extreme conditions, and use the effective loss reduction events under extreme conditions to supplement the typical environmental scenario loss reduction case library to form a reusable decision rule set.

[0014] Furthermore, a transmission line intelligent control system based on distribution data feedback is proposed, which is used to implement the control method as described in any one of the above, including: An acquisition module, which is used to obtain the historical operation parameters of each transmission line to be analyzed and the historical meteorological data, historical geographical data, and historical humanistic data of the area where each transmission line to be analyzed is located, and is used to obtain environmental prediction data and real-time line operation parameters; A data processing module, which is used to perform spatio-temporal alignment on the historical meteorological data, historical geographical data, historical humanistic data, and historical operation parameters of each transmission line to be analyzed, and construct a historical environment-electrical coupling feature matrix, and is used to extract the historical environment time series features in the historical environment-electrical coupling feature matrix by using a long short-term memory network, capture the spatial distribution law by combining a convolutional neural network, and establish a dynamic line loss prediction model; A main control module, which is used to obtain the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model, obtain the historical actual line loss values, compare the differences with the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model, and judge whether to trigger the isolation forest algorithm to detect abnormal environmental parameter periods according to the difference comparison results, and then screen the key loss-causing factors through Pearson correlation analysis, and is used to obtain environmental prediction data and real-time line operation parameters, and combine the key loss-causing factors and the dynamic line loss prediction model to dynamically generate loss reduction strategy priority events, execute the loss reduction strategy priority events, and obtain the actual line loss data after execution, and is used to perform error backpropagation on the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, optimize the dynamic line loss prediction model by using a reinforcement learning algorithm, generate a typical environmental scenario loss reduction case library based on the optimized dynamic line loss prediction model, and simulate the effectiveness of strategies under extreme conditions through digital twin technology to form a reusable decision rule set; A display module, which is used to display the data and information collected by the acquisition module, display the data processing process and results of the data processing module, and display the data analysis process and results of the main control module.

[0015] In an alternative embodiment, the acquisition module includes: The first acquisition unit is used to obtain the historical operation parameters of each transmission line to be analyzed, as well as the historical meteorological data, historical geographical data, and historical humanistic data of the area where each transmission line to be analyzed is located; The second acquisition unit is used to obtain environmental prediction data and real-time operation parameters of the line.

[0016] In an alternative embodiment, the main control module includes: The line loss evaluation unit is used to obtain the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model; The damage-causing factor judgment unit is used to obtain the historical actual line loss value, compare the difference with the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model, and judge whether to trigger the isolation forest algorithm to detect abnormal periods of environmental parameters according to the difference comparison result, and then screen the key damage-causing factors through Pearson correlation analysis; The line loss reduction execution unit is used to obtain environmental prediction data and real-time operation parameters of the line, and combine the key damage-causing factors and the dynamic line loss prediction model to dynamically generate line loss reduction strategy priority events, execute the line loss reduction strategy priority events, and obtain the actual line loss data after execution; The model optimization unit is used to perform error backpropagation on the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and optimize the dynamic line loss prediction model using the reinforcement learning algorithm; The decision rule generation unit is used to generate a typical environmental scenario line loss reduction case library based on the optimized dynamic line loss prediction model, and simulate the effectiveness of the strategy under extreme conditions through digital twin technology to form a reusable decision rule set.

[0017] Compared with the prior art, the beneficial effects of the present invention are: An intelligent control method for transmission lines based on distribution data feedback proposed in this solution, by integrating multi-source data (line operation parameters, meteorology, geography, and human data) and constructing a historical environment - electrical coupling feature matrix, combining LSTM and CNN to construct a dynamic line loss prediction model, realizes the prediction of the minimum line loss value under different environmental scenarios. Detects abnormal periods through the Isolation Forest algorithm, combines Pearson correlation and SHAP values to analyze and screen key loss-causing factors, dynamically generates priority events for loss reduction strategies and executes them, and optimizes model parameters using error backpropagation and reinforcement learning. Solves the technical problems of inaccurate fault detection and lagging strategy optimization caused by traditional methods ignoring environmental factors, has the advantages of improving detection accuracy, enhancing dynamic adaptability, reducing line loss rate, and improving power supply reliability, and simulates extreme conditions through digital twin technology to form a reusable decision rule set, promoting the refined control of smart grids. Brief Description of the Drawings

[0018] Figure 1 Is a flowchart of an intelligent control method for transmission lines based on distribution data feedback proposed by the present invention; Figure 2 Is a flowchart for obtaining the historical environment - electrical coupling feature matrix in the present invention; Figure 3 Is a flowchart for obtaining the minimum line loss value in the present invention; Figure 4 Is a system framework diagram of an intelligent control system for transmission lines based on distribution data feedback proposed by the present invention. Detailed Embodiment

[0019] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations.

[0020] Refer to Figure 1 - Figure 4 As shown, an intelligent control method for transmission lines based on distribution data feedback, characterized by comprising: S1. Obtain the historical operation parameters of each transmission line to be analyzed, and then obtain the historical meteorological data, historical geographical data, and historical human data of the area where each transmission line to be analyzed is located, and perform spatio-temporal alignment with the historical operation parameters of the line to construct a historical environment - electrical coupling feature matrix.

[0021] Further, step S1 specifically includes: S1.1. Collect the historical operation parameters of each transmission line to be analyzed through the SCADA system, and the historical operation parameters of the line include historical current, historical voltage, historical load rate, and historical fault records; S1.2. Collect historical temperature, historical humidity, historical wind speed, historical rainfall and historical ice coverage at each meteorological station in the area where the power transmission line to be analyzed is located to obtain historical meteorological data; S1.3. Collect historical terrain slope, historical vegetation coverage, historical soil resistivity and historical altitude through the meteorological station in the area where each transmission line to be analyzed is located; S1.4. Collect historical industrial pollution concentration, historical traffic noise decibel value and historical wireless base station electromagnetic radiation intensity through the meteorological station in each area where the power transmission line to be analyzed is located, and obtain historical humanities data. The data collected is historical data for the past 5 to 10 years; S1.5. Use sliding average filtering (window size is 5) to eliminate high-frequency noise for historical meteorological data, historical geographical data and historical line operation parameters, and use median filtering to suppress mutation interference for historical cultural data; S1.6. Use KNN interpolation to fill missing values ​​in historical meteorological data (k value is 5), and use linear regression to fill missing values ​​in geographic data; S1.7. Based on the 3σ principle, remove abnormal data points that are beyond the normal range, and project all data processed from step 1.4 to step 1.6 into the UTM coordinate system, taking the center point of the transmission line as the origin, and establish a local coordinate grid (resolution is less than or equal to 10m×10m); S1.8. Based on the GPS timing system, the historical meteorological data, historical geographical data, historical cultural data and historical line operation parameters projected into the UTM coordinate system shall be aligned according to minute-level timestamps, and the maximum time deviation allowed shall be controlled within ±1 minute; S1.9. Use principal component analysis to reduce the dimension, retain the principal components with cumulative variance contribution greater than or equal to 95%, and then generate a historical environment-electrical coupling feature matrix with a dimension less than or equal to 50.

[0022] S2. Use the long short-term memory network to extract the historical environmental time series features in the historical environmental-electrical coupling feature matrix, combine the convolutional neural network to capture the spatial distribution law, establish a dynamic line loss prediction model, and obtain the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model.

[0023] Specifically, based on the 3σ principle, abnormal data points beyond the normal range are eliminated, such as abnormal records with temperature greater than 50°C or humidity greater than 100%.

[0024] Furthermore, step S2 specifically includes: S2.1. Design a two-layer long short-term memory network, set the input layer dimension to 50 (the number of features after PCA dimension reduction), set the number of hidden layer units to 128, set the activation function to tanh, and the output layer predicts the line loss sequence for the next 24 hours; S2.2. Divide the historical data in the historical environment - electrical coupling feature matrix into training samples according to a 24-hour sliding window, with a step size of 1 hour, to generate N time series segments; S2.3. Use a three-layer convolutional neural network, with the convolutional kernel size set to 3 by 3, the step size set to 1, and the padding method set to SAME, to extract the spatial distribution features of the historical geographical data in the historical environment - electrical coupling feature matrix (such as terrain slope gradient, vegetation coverage hotspots); S2.4. Concatenate the N time series segments generated by the double-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network, input them into the fully connected layer (with 256 neurons), and the output layer generates the dynamic line loss prediction value, thus completing the construction of the dynamic line loss prediction model; S2.5. According to the N time series segments generated by the double-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network, obtain the historical environment sensitivity coefficient matrix, and use the historical environment sensitivity coefficient matrix to correct the historical line operation parameters in the historical environment - electrical coupling feature matrix; S2.6. Set the loss function, and use the Adam algorithm to dynamically adjust the learning rate. Then, input the corrected N time series segments generated by the double-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network into the dynamic line loss prediction model to train the model; S2.7. Simulate and generate the environment - electrical coupling feature simulation matrix under different environmental scenarios, and input the simulated meteorological data, simulated geographical data, and simulated human data in the environment - electrical coupling feature simulation matrix into the trained dynamic line loss prediction model to obtain the minimum line loss values under different environmental scenarios output by the trained dynamic line loss prediction model; Among them, the loss function is: In the formula, , is the model validation set, are the model parameters; The mathematical definition of MSE is: In the formula, is the historical actual line loss value corresponding to the i-th time series segment, is the predicted line loss value corresponding to the i-th time series segment output by the model.

[0025] It can be understood that the steps to obtain the historical environment sensitivity coefficient matrix are: S2.5.1. Input data: Time series features : The historical environment time series features extracted by the double-layer LSTM, T is the number of time steps, dLSTM is the output dimension of the LSTM.

[0026] Spatial feature h CNN ∈R H×W×C : Geospatial distribution features extracted by a three - layer CNN, where H×W is the spatial resolution and C is the number of channels.

[0027] S2.5.2. Concatenate the temporal feature h LSTM and the spatial feature h CNN into a comprehensive feature vector: h fusion =Concat(h LSTM， h CNN ), where h fusion ∈R dfusion , d fusion =d LSTM +C·H·W; S2.5.3. Map the fused feature to the environmental sensitivity coefficient matrix through a fully - connected layer: K env =σ(W·h fusion +b); Where: W∈R n env ×d fusion is the weight matrix of the fully - connected layer, b∈R n env is the bias term, n env is the dimension of the environmental sensitivity coefficient matrix (e.g., n env =n env_params ×n elec_params ), and σ(·) is the activation function (e.g., ReLU); S2.5.4. Reconstruct the output K env of the fully - connected layer into the target shape (n env_params , n elec_params ), that is: K env ∈R n env_params ×n elec_params ; Where n env_params is the number of types of environmental data, geographical data, and human data (such as temperature, humidity, wind speed, etc.), and n elec_params is the number of types of line operation parameters (such as current, etc.).

[0028] Example: Assume that the LSTM output dimension d LSTM =128, and the CNN output spatial feature h CNN∈R32 (after global average pooling), the weights of the fully connected layer W ∈ R 6×160 (mapping 160-dimensional features to 6 environmental parameters × 3 line operation parameters).

[0029] Then the environmental sensitivity coefficient matrix is: K env = ReLU(W · [h LSTM ; h CNN ); The output shape is 6 × 3, for example: .

[0030] It can be understood that the matrix element k i，j represents the sensitivity of the i-th environmental data, geographical data, and human data to the j-th line operation parameter. For example, K temp,R = 0.002 means that when the temperature rises by 1°C, the resistance increases by 0.2%.

[0031] It can also be understood that the historical line operation parameters in the historical environment - electrical coupling feature matrix are corrected using the historical environmental sensitivity coefficient matrix. Taking the AC resistance of the transmission line as an example: R env-adjusted = R 0 · (1 + αΔT + βΔH + γΔU wind ) R 0 is the reference resistance (the resistance value under normal temperature and dry conditions); ΔT, ΔH, ΔU wind are the change amounts of temperature, humidity, and wind speed respectively; α, β, γ are the historical environmental sensitivity coefficients (obtained by looking up the historical environmental sensitivity coefficient matrix).

[0032] S3. Obtain the historical actual line loss value and compare the difference with the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model. Then, based on the difference comparison result, determine whether to trigger the isolation forest algorithm to detect abnormal periods of environmental parameters, and then screen key loss-causing factors through Pearson correlation analysis.

[0033] Furthermore, step S3 specifically includes: Obtain the difference between the historical actual line loss value corresponding to N time series segments and the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model; If the difference between the historical actual line loss value and the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model is greater than or equal to 0.15 times the minimum line loss value, then trigger the isolation forest algorithm to obtain the anomaly score of each data point; Screen the anomaly scores higher than Tthreshold The simulated meteorological data, simulated geographical data, and simulated humanistic data in the environmental-electrical coupling feature simulation matrix corresponding to the data points of = μ(s) + 3σ(s) are used to obtain the first abnormal data, where μ(s) is the global mean of the minimum line loss values corresponding to all data points, σ(s) is the global standard deviation of the minimum line loss values corresponding to all data points, and T threshold is the abnormal data screening threshold; Based on the first abnormal data and its corresponding minimum line loss value, a first correlation coefficient is obtained, and at the same time, the first abnormal data with the first correlation coefficient greater than or equal to 0.8 is obtained; Through the SHAP value analysis model, the contribution degree of each factor in the first abnormal data to the line loss is quantified, and the factor with the largest contribution degree is marked as the key loss-causing factor; Among them, when triggering the isolation forest algorithm, the calculation formula for the abnormal score of each data point is: In the formula, is the abnormal score in the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment, is the historical actual line loss value at the i-th time series segment of the transmission circuit corresponding to the t-th environmental scenario, is the mean value in the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment, is the standard deviation of the minimum line loss value in the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment; The calculation process for obtaining the first correlation coefficient based on the first abnormal data and its corresponding minimum line loss value is: S3.1.1. Let the first abnormal data be , and its corresponding minimum line loss value be ; S3.1.2. Through the Pearson correlation coefficient calculation formula , the first correlation coefficient r is obtained, where is the value of the th abnormal data point, is the line loss value corresponding to the th abnormal data point, is the mean value corresponding to the abnormal data point, is the mean value of the line loss values corresponding to the abnormal data point.

[0034] It can be understood that the specific steps for quantifying the contribution degree of each factor in the first abnormal data to the line loss through the SHAP value analysis model and marking the factor with the largest contribution degree as the key loss-causing factor are S3.2.1. Conduct SHAP value analysis on the first abnormal data (such as humidity, pollution level) screened out to quantify its contribution to line loss: , where represents the contribution of each factor in the first abnormal data to line loss (the marginal contribution of this feature to the model output); S3.2.2. If the C of a certain factor j > 0.3·∑C j , then it is marked as a key loss-causing factor.

[0035] S4. Obtain environmental prediction data and real-time line operation parameters, and combine key loss-causing factors and a dynamic line loss prediction model to dynamically generate loss reduction strategy priority events.

[0036] Further, step S4 specifically includes: Obtain meteorological data, geographical data, and human data for the next 24 hours from the meteorological bureau to get environmental prediction data, and synchronously obtain the real-time line operation parameters of each transmission line to be analyzed; Input the environmental prediction data and real-time line operation parameters into the dynamic line loss prediction model to obtain the minimum line loss value corresponding to the environmental prediction data and real-time line operation parameters; Compare the environmental prediction data and real-time line operation parameters with the key loss-causing factors to dynamically generate a dynamic loss-causing value. If the environmental prediction data and real-time line operation parameters meet one of the key loss-causing factors, then the dynamically generated dynamic loss-causing value +1, and normalize the dynamically generated dynamic loss-causing value to obtain a dynamic loss-causing value within the range of 0-1. For example, assume the key loss-causing factors are temperature (K = 1), humidity (K = 2), wind speed (K = 3), and the total number of factors K = 3: Environmental prediction data: T env = 32°C, H env = 75%, W env = 8m / s Real-time operation parameters: T real = 31°C, H real = 82%, W real = 9m / s Trigger judgment: Factor 1 (temperature): T real = 31°C > 30°C → Trigger (I(1) = 1).

[0037] Factor 2 (humidity): H real = 82% > 80% → Trigger (I(2) = 1).

[0038] Factor 3 (wind speed): W env = 8m / s < 10m / s, W real= 9 m / s < 10 m / s → Not triggered (I(3) = 0).

[0039] Dynamic damage value calculation: D raw = 1 + 1 + 0 = 2, Dnorm = 32 ≈ 0.67; Perform weighted summation of the minimum line loss value and the dynamic damage value to obtain a reference value for the loss reduction strategy; Based on the reference value of the loss reduction strategy, dynamically generate loss reduction strategy priority events.

[0040] S5. Execute the loss reduction strategy priority event, obtain the actual line loss data after execution, synchronously perform error backpropagation on the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and optimize the dynamic line loss prediction model using the reinforcement learning algorithm.

[0041] Furthermore, step S5 specifically includes: After executing the loss reduction strategy priority event, obtain the actual line loss value of the transmission line corresponding to the loss reduction strategy priority event during the abnormal period to obtain the actual line loss data after execution; Synchronously obtain the actual environment - electrical coupling characteristic matrix of the transmission line corresponding to the loss reduction strategy priority event; Import the data in the actual environment - electrical coupling characteristic matrix into the dynamic line loss prediction model to obtain the predicted line loss data output by the dynamic line loss prediction model; Compare the predicted line loss value included in the predicted line loss data output by the dynamic line loss prediction model with the actual line loss value to obtain the difference between the two for error backpropagation, and synchronously update the parameters in the dynamic line loss prediction model using the Adam optimizer to realize the optimization of the dynamic line loss prediction model using the reinforcement learning algorithm.

[0042] S6. Based on the optimized dynamic line loss prediction model, generate a loss reduction case library for typical environmental scenarios, and simulate the effectiveness of the strategy under extreme conditions through digital twin technology to form a reusable decision rule set.

[0043] Furthermore, step S6 specifically includes: Based on the optimized dynamic line loss prediction model, obtain the minimum line loss values under different environmental scenarios; Perform typical scenario division on the meteorological data, geographical data, human data, and line operation parameters corresponding to the minimum line loss values under different environmental scenarios, and combine with the loss reduction strategy priority events to generate a loss reduction case library for typical environmental scenarios; Build a 3D model of the line based on the Unreal Engine, and inject extreme environmental parameters and extreme line operation parameters into the 3D model of the line using physical simulation to simulate the effectiveness of the strategy under extreme conditions; According to the effectiveness of strategies under extreme conditions, obtain the loss reduction effective events under extreme conditions, and use the loss reduction effective events under extreme conditions to supplement the loss reduction case library for typical environmental scenarios, form a reusable decision rule set, and use the reusable decision rule set to intelligently control the transmission line.

[0044] Furthermore, a transmission line intelligent control system based on distribution data feedback is proposed to implement the control method as described in any one of the above, including: A collection module, which is used to obtain the historical operation parameters of each transmission line to be analyzed and the historical meteorological data, historical geographical data, and historical humanistic data of the area where each transmission line to be analyzed is located, for obtaining environmental prediction data and real-time operation parameters of the line; A data processing module, which is used to perform spatio-temporal alignment on the historical meteorological data, historical geographical data, historical humanistic data, and historical operation parameters of the area where each transmission line to be analyzed is located, and construct a historical environment-electrical coupling feature matrix, for using a long short-term memory network to extract the historical environment time series features in the historical environment-electrical coupling feature matrix, and combining a convolutional neural network to capture the spatial distribution law, and establish a dynamic line loss prediction model; A main control module, which is used to obtain the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model, for obtaining the historical actual line loss value and comparing the difference with the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model, and according to the difference comparison result, judge whether to trigger the isolation forest algorithm to detect abnormal periods of environmental parameters, and then screen key loss-causing factors through Pearson correlation analysis, for obtaining environmental prediction data and real-time operation parameters of the line, and combining the key loss-causing factors and the dynamic line loss prediction model, dynamically generate loss reduction strategy priority events, execute the loss reduction strategy priority events, and obtain the actual line loss data after execution, for performing error backpropagation on the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and using a reinforcement learning algorithm to optimize the dynamic line loss prediction model, for generating a loss reduction case library for typical environmental scenarios based on the optimized dynamic line loss prediction model, and simulating the effectiveness of strategies under extreme conditions through digital twin technology, and forming a reusable decision rule set; A display module, which is used to display the data and information collected by the collection module, display the data processing process and results of the data processing module, and display the data analysis process and results of the main control module.

[0045] Furthermore, the collection module includes: A first collection unit, which is used to obtain the historical operation parameters of each transmission line to be analyzed and the historical meteorological data, historical geographical data, and historical humanistic data of the area where each transmission line to be analyzed is located; The second acquisition unit is used to obtain environmental prediction data and real-time line operation parameters.

[0046] Furthermore, the main control module includes: The line loss assessment unit is used to obtain the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model; The damage factor judgment unit is used to obtain the historical actual line loss value, compare the difference with the minimum line loss values under different environmental scenarios output by the dynamic line loss prediction model, and judge whether to trigger the isolation forest algorithm to detect abnormal periods of environmental parameters according to the difference comparison result, and then screen the key damage factors through Pearson correlation analysis; The line loss reduction execution unit is used to obtain environmental prediction data and real-time line operation parameters, combine the key damage factors and the dynamic line loss prediction model, dynamically generate line loss reduction strategy priority events, execute the line loss reduction strategy priority events, and obtain the actual line loss data after execution; The model optimization unit is used to perform error backpropagation on the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and optimize the dynamic line loss prediction model using the reinforcement learning algorithm; The decision rule generation unit is used to generate a line loss reduction case library for typical environmental scenarios based on the optimized dynamic line loss prediction model, and simulate the effectiveness of the strategy under extreme conditions through digital twin technology to form a reusable decision rule set.

[0047] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent control of power transmission lines based on power distribution data feedback, characterized in that: include: S1. Obtain the historical operation parameters of each transmission line to be analyzed, then obtain the historical meteorological data, historical geographical data and historical cultural data of the area where each transmission line to be analyzed is located, and align them with the historical operation parameters of the line in time and space to construct a historical environment-electrical coupling feature matrix; S2. Use the long short-term memory network to extract the historical environmental time series features in the historical environmental-electrical coupling feature matrix, combine the convolutional neural network to capture the spatial distribution law, establish a dynamic line loss prediction model, and obtain the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model; S3. Obtain the historical actual line loss value and compare the difference with the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model. Based on the difference comparison result, determine whether to trigger the isolation forest algorithm to detect the abnormal period of environmental parameters, and then screen the key loss factors through Pearson correlation analysis; S4. Obtain environmental prediction data and real-time line operation parameters, and dynamically generate loss reduction strategy priority events by combining key loss-causing factors and dynamic line loss prediction models; S5. Execute the loss reduction strategy priority event, obtain the actual line loss data after the execution, synchronously perform error back propagation between the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and optimize the dynamic line loss prediction model using the reinforcement learning algorithm; S6. Based on the optimized dynamic line loss prediction model, a loss reduction case library for typical environmental scenarios is generated, and the effectiveness of strategies under extreme conditions is simulated through digital twin technology to form a reusable decision rule set.

2. According to claim 1, a method for intelligent control of power transmission lines based on power distribution data feedback is characterized in that: Step S1 specifically includes: S1.

1. The historical operating parameters of each transmission line to be analyzed are collected through the SCADA system, wherein the historical operating parameters of the line include historical current, historical voltage, historical load rate and historical fault records; S1.

2. Collect historical temperature, historical humidity, historical wind speed, historical rainfall and historical ice coverage at each meteorological station in the area where the power transmission line to be analyzed is located to obtain historical meteorological data; S1.

3. Collect historical terrain slope, historical vegetation coverage, historical soil resistivity and historical altitude through the meteorological station in the area where each transmission line to be analyzed is located; S1.

4. Collect historical industrial pollution concentrations, historical traffic noise decibel values, and historical wireless base station electromagnetic radiation intensity through the meteorological station in each area where the power transmission line to be analyzed is located, and obtain historical cultural data; S1.

5. Use sliding average filtering to eliminate high-frequency noise for historical meteorological data, historical geographical data and historical line operation parameters, and use median filtering to suppress mutation interference for historical cultural data; S1.

6. Use KNN interpolation to fill missing values ​​in historical meteorological data, and use linear regression to fill missing values ​​in geographic data; S1.

7. Based on the 3σ principle, remove abnormal data points that are beyond the normal range, and project all data processed from step 1.4 to step 1.6 into the UTM coordinate system, and establish a local coordinate grid with the center point of the transmission line as the origin; S1.

8. Based on the GPS timing system, align the historical meteorological data, historical geographic data, historical cultural data and historical route operation parameters projected into the UTM coordinate system according to minute-level timestamps; S1.

9. Use principal component analysis to reduce the dimension, retain the principal components with cumulative variance contribution greater than or equal to 95%, and then generate a historical environment-electrical coupling feature matrix with a dimension less than or equal to 50.

3. According to claim 1, a method for intelligent control of transmission lines based on power distribution data feedback is characterized in that: Step S2 specifically includes: S2.

1. Design a two-layer long short-term memory network, set the input layer dimension to 50, the number of hidden layer units to 128, the activation function to tanh, and the output layer to predict the line loss sequence for the next 24 hours; S2.

2. Divide the historical data in the historical environment-electrical coupling feature matrix into training samples according to a 24-hour sliding window with a step size of 1 hour to generate N time series segments; S2.

3. A 3-layer convolutional neural network is used with the convolution kernel size set to 3×3, the step size set to 1, and the padding method set to SAME to extract the spatial distribution characteristics of historical geographic data in the historical environment-electrical coupling feature matrix; S2.

4. The N time series segments generated by the two-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network are spliced ​​and input into the fully connected layer. The output layer generates the dynamic line loss prediction value, thereby completing the construction of the dynamic line loss prediction model; S2.

5. Obtain the historical environmental sensitivity coefficient matrix based on the N time series segments generated by the two-layer long short-term memory network and the spatial distribution characteristics obtained by the three-layer convolutional neural network, and use the historical environmental sensitivity coefficient matrix to correct the historical line operation parameters in the historical environment-electrical coupling characteristic matrix; S2.

6. Set the loss function and use the Adam algorithm to dynamically adjust the learning rate. Then, input the N time series segments generated by the modified two-layer long short-term memory network and the spatial distribution features obtained by the three-layer convolutional neural network into the dynamic line loss prediction model to train the model. S2.

7. Simulate and generate the environment-electrical coupling characteristic simulation matrix under different environmental scenarios, and input the simulated meteorological data, simulated geographical data and simulated cultural data in the environment-electrical coupling characteristic simulation matrix into the trained dynamic line loss prediction model to obtain the minimum line loss value under different environmental scenarios output by the trained dynamic line loss prediction model.

4. According to claim 1, a method for intelligent control of transmission lines based on power distribution data feedback is characterized in that: Step S3 specifically includes: Obtain the difference between the historical actual line loss values ​​corresponding to N time series segments and the minimum line loss values ​​under different environmental scenarios output by the dynamic line loss prediction model; If the difference between the historical actual line loss value and the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model is greater than or equal to 0.15 times the minimum line loss value, the isolation forest algorithm is triggered to obtain the anomaly score of each data point; Screening abnormality score higher than T threshold =μ(s)+3σ(s) corresponding to the data point of the environment-electrical coupling characteristic simulation matrix, the simulated meteorological data, simulated geographical data and simulated cultural data are used to obtain the first abnormal data, where μ(s) is the global mean of the minimum line loss value corresponding to all data points, σ(s) is the global standard deviation of the minimum line loss value corresponding to all data points, T threshold Filter thresholds for abnormal data; Based on the first abnormal data and the minimum line loss value corresponding thereto, a first correlation coefficient is obtained, and at the same time, first abnormal data with a first correlation coefficient greater than or equal to 0.8 is obtained; The SHAP value analysis model is used to quantify the contribution of each factor in the first abnormal data to the line loss, and the factor with the largest contribution is marked as the key loss-causing factor; Among them, the calculation formula for triggering the isolation forest algorithm and obtaining the abnormal score of each data point is: In the formula, is the abnormal score of the t-th environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment, is the historical actual line loss value at the i-th time segment corresponding to the transmission circuit under the t-th environmental scenario, is the mean value of the dynamic line loss prediction model output under the t-th environmental scenario corresponding to the i-th time series segment, It is the standard deviation of the minimum line loss value in the tth environmental scenario output by the dynamic line loss prediction model corresponding to the i-th time series segment.

5. The method for intelligent control of power transmission lines based on power distribution data feedback according to claim 1, characterized in that: Step S4 specifically includes: Obtain the meteorological data, geographic data and humanities data for the next 24 hours from the Meteorological Bureau, obtain environmental forecast data, and simultaneously obtain the real-time operating parameters of each transmission line to be analyzed; Inputting environmental prediction data and line real-time operation parameters into a dynamic line loss prediction model to obtain a minimum line loss value corresponding to the environmental prediction data and line real-time operation parameters; Compare environmental prediction data and real-time line operation parameters with key damage factors to dynamically generate dynamic damage values; The minimum line loss value and the dynamic loss value are weighted and summed to obtain a reference value for the loss reduction strategy; Based on the loss reduction strategy reference value, loss reduction strategy priority events are dynamically generated.

6. The method for intelligent control of power transmission lines based on power distribution data feedback according to claim 1, characterized in that: Step S5 specifically includes: After executing the loss reduction strategy priority event, the actual line loss value of the transmission line corresponding to the loss reduction strategy priority event during the abnormal period is obtained to obtain the actual line loss data after execution; Synchronously obtain the actual environment-electrical coupling characteristic matrix of the transmission line corresponding to the priority events of the loss reduction strategy; Importing data in the actual environment-electrical coupling characteristic matrix into the dynamic line loss prediction model to obtain predicted line loss data output by the dynamic line loss prediction model; The predicted line loss value contained in the predicted line loss data output by the dynamic line loss prediction model is compared with the actual line loss value to obtain the difference between the two for error back propagation. The Adam optimizer is simultaneously used to update the parameters in the dynamic line loss prediction model to optimize the dynamic line loss prediction model using the reinforcement learning algorithm.

7. The method for intelligent control of power transmission lines based on power distribution data feedback according to claim 1, characterized in that: Step S6 specifically includes: Based on the optimized dynamic line loss prediction model, the minimum line loss value in different environmental scenarios is obtained; The meteorological data, geographical data, cultural data, and line operation parameters corresponding to the minimum line loss values ​​in different environmental scenarios are divided into typical scenarios, and combined with the priority events of the loss reduction strategy, a typical environmental scenario loss reduction case library is generated; Build a 3D line model based on Unreal Engine, and use physical simulation to inject extreme environmental parameters and extreme line operation parameters into the line 3D model to simulate the effectiveness of strategies under extreme conditions; According to the effectiveness of strategies under extreme conditions, effective loss reduction events under extreme conditions are obtained, and the effective loss reduction events under extreme conditions are used to supplement the loss reduction case library of typical environmental scenarios to form a reusable decision rule set.

8. An intelligent control system for power transmission lines based on power distribution data feedback, used to implement the control method according to any one of claims 1 to 7, characterized in that: include: A collection module, which is used to obtain the historical operation parameters of each transmission line to be analyzed and the historical meteorological data, historical geographical data and historical cultural data of the area where each transmission line to be analyzed is located, and is used to obtain environmental prediction data and real-time operation parameters of the line; A data processing module, which is used to perform spatiotemporal alignment on the historical meteorological data, historical geographical data, historical cultural data and historical operation parameters of each transmission line to be analyzed, and to construct a historical environment-electrical coupling feature matrix, and to extract the historical environment time series features in the historical environment-electrical coupling feature matrix using a long short-term memory network, and to capture the spatial distribution law in combination with a convolutional neural network, so as to establish a dynamic line loss prediction model; A main control module, which is used to obtain the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model, to obtain the historical actual line loss value and perform a difference comparison with the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model, and to determine whether to trigger the isolation forest algorithm to detect the abnormal period of environmental parameters according to the difference comparison result, and then screen the key loss factors through Pearson correlation analysis, to obtain environmental prediction data and real-time line operation parameters, and to dynamically generate loss reduction strategy priority events in combination with the key loss factors and the dynamic line loss prediction model, execute the loss reduction strategy priority events, and obtain the actual line loss data after execution, to perform error back propagation between the actual line loss data and the predicted line loss data output by the dynamic line loss prediction model, and to optimize the dynamic line loss prediction model using a reinforcement learning algorithm, to generate a typical environmental scenario loss reduction case library based on the optimized dynamic line loss prediction model, and to simulate the effectiveness of the strategy under extreme conditions through digital twin technology to form a reusable decision rule set; The display module is used to display the data and information collected by the acquisition module, the data processing process and results of the data processing module, and the data analysis process and results of the main control module.

9. The intelligent control system for power transmission lines based on power distribution data feedback according to claim 8, characterized in that: The acquisition module comprises: A first acquisition unit, the first acquisition unit is used to obtain the historical operation parameters of each transmission line to be analyzed and the historical meteorological data, historical geographical data and historical cultural data of the area where each transmission line to be analyzed is located; The second acquisition unit is used to obtain environmental prediction data and real-time line operation parameters.

10. The intelligent control system for power transmission lines based on power distribution data feedback according to claim 8, characterized in that: The main control module comprises: A line loss evaluation unit, which is used to obtain the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model; A damage factor judgment unit, which is used to obtain the historical actual line loss value and compare the difference with the minimum line loss value under different environmental scenarios output by the dynamic line loss prediction model, and judge whether to trigger the isolation forest algorithm to detect the abnormal period of environmental parameters according to the difference comparison result, and then screen the key damage factors through Pearson correlation analysis; A loss reduction execution unit, which is used to obtain environmental prediction data and real-time line operation parameters, and dynamically generate loss reduction strategy priority events in combination with key loss-causing factors and a dynamic line loss prediction model, execute the loss reduction strategy priority events, and obtain actual line loss data after execution; A model optimization unit, the model optimization unit is used to perform error back propagation between actual line loss data and predicted line loss data output by a dynamic line loss prediction model, and optimize the dynamic line loss prediction model using a reinforcement learning algorithm; A decision rule generation unit is used to generate a loss reduction case library for typical environmental scenarios based on the optimized dynamic line loss prediction model, and to simulate the effectiveness of strategies under extreme conditions through digital twin technology to form a reusable decision rule set.

Images