Self-adaptive scene perception and dynamic optimization method and system for line loss of transformer area

Abstract

The invention discloses a transformer area line loss adaptive scene perception and dynamic optimization method and system, and relates to the technical field of power system line loss analysis and governance, and the method comprises the steps: collecting voltage and current data of a transformer area distribution transformer, carrying out the time sequence decomposition of the data, obtaining a trend term and a period term, calculating a scene discrimination index, and carrying out the operation scene division; constructing a line loss estimation equation according to the operation scene and the scene discrimination index, dividing power distribution branches and calculating a theoretical line loss value; when the deviation between the theoretical line loss value and the actually measured line loss value exceeds a threshold value, calculating a line loss contribution degree and determining an abnormal line loss branch; calculating the ratio of the compensation cost to the line loss improvement value for the abnormal line loss branch as a target function, and solving a compensation strategy through a gradient descent method; and executing the compensation strategy and dynamically adjusting the constraint condition until a preset line loss rate requirement is met. According to the invention, self-adaptive perception of the operation scene of the transformer area and dynamic optimization compensation of the line loss can be realized, and the line loss treatment effect is improved.

Description

Technical Field

[0001] This invention relates to the field of power system line loss analysis and management technology, and in particular to a method and system for adaptive scenario perception and dynamic optimization of transformer substation line loss. Background Technology

[0002] Current line loss analysis methods for transformer substations are mainly based on static load models, which cannot adapt to the actual situation of dynamic load changes. Existing technologies typically use fixed time windows for data collection and analysis, making it difficult to accurately identify line loss characteristics under different operating scenarios, resulting in significant deviations between line loss assessment results and actual conditions.

[0003] Methods for managing line losses in distribution networks generally suffer from problems such as delayed response and simplistic compensation strategies. Traditional methods often rely on fixed thresholds for anomaly detection, which is insufficient to handle complex and variable load characteristics. Furthermore, the formulation of compensation strategies lacks in-depth analysis of the contribution of line losses, easily leading to a waste of compensation resources.

[0004] Existing adaptive compensation methods primarily focus on the compensation effect of a single branch, failing to adequately consider the coupling relationship between branches. Furthermore, the optimization objectives of the compensation strategies are overly simplistic, failing to achieve a balance between economy and reliability. The lack of a dynamic adjustment mechanism during compensation execution makes it difficult to guarantee the continuity and stability of the compensation effect. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0006] The technical solution of this invention is: a method for adaptive scene perception and dynamic optimization of transformer area line loss, comprising the following steps: S1. Collect voltage and current data of the distribution transformers in the distribution area; S2. Perform time-series decomposition on voltage and current data to obtain trend and periodic terms. Calculate scenario discrimination indicators based on the fluctuation amplitude of the trend and the duration of the periodic terms. Adaptively divide the operating status of the transformer area based on the scenario discrimination indicators to obtain the operating scenario. S3. Construct line loss estimation equations based on operating scenarios and scenario discrimination indicators. Divide the distribution network of the transformer area into multiple distribution branches based on the line loss estimation equations and calculate the theoretical line loss value of each distribution branch. S4. When the deviation between the theoretical line loss value and the measured line loss value exceeds the preset deviation threshold, calculate the line loss contribution of each distribution branch according to the line loss estimation equation, and determine the branch with the largest line loss contribution as the abnormal line loss branch. S5. Calculate the ratio of compensation cost to line loss improvement value for abnormal line loss branches, use the ratio as the objective function, adjust the constraints of the objective function according to the contribution of line loss, and solve the compensation strategy using the gradient descent method. S6. Execute the compensation strategy and dynamically adjust the constraints until the line loss rate after compensation is less than the preset line loss rate threshold, output the optimal compensation strategy, convert the optimal compensation strategy into control commands and send them to the execution device.

[0007] In step S2, the voltage and current data are decomposed into time series to obtain trend and periodic terms. A scenario discrimination index is calculated based on the fluctuation amplitude of the trend term and the duration of the periodic term. Based on this scenario discrimination index, the operating status of the transformer substation is adaptively divided to obtain operating scenarios, including: S2.1. Perform empirical mode decomposition on the voltage and current data according to the time series to obtain multiple time-series intrinsic mode components; S2.2. Calculate the energy distribution value of each time series intrinsic mode component, construct a time series reconstruction matrix based on the energy distribution value, and reconstruct each time series intrinsic mode component based on the time series reconstruction matrix to obtain the trend term and periodic term; S2.3. Extract the fluctuation amplitude of the trend item and the duration of the period item, calculate the cross-correlation coefficient between the fluctuation amplitude and the duration, determine the weight coefficient based on the cross-correlation coefficient, and weight the fluctuation amplitude and duration according to the weight coefficient to obtain the scene discrimination index. S2.4. Calculate the data density distribution of the scene discrimination index, determine multiple cluster centers based on the data density distribution, and calculate the distance from the operating status of the transformer area to each cluster center; S2.5. Update the cluster center position based on the distance and reclassify the affiliation of the station area operation status until the cluster center position converges to obtain the operation scenario.

[0008] In step S3, a line loss estimation equation is constructed based on the operating scenario and scenario discrimination index. The distribution network of the transformer substation is then divided into multiple distribution branches based on this equation, and the theoretical line loss value for each distribution branch is calculated, including: S3.1. Extract power flow distribution features based on the operating scenario, construct a dynamic weight matrix based on the scenario discrimination index, and combine the power flow distribution features with the dynamic weight matrix to construct a line loss estimation equation; S3.2. Calculate the voltage-current correlation degree of the distribution branches in the distribution network of the transformer substation according to the line loss estimation equation, determine the boundary points of the distribution branches according to the rate of change of the voltage-current correlation degree, and divide the distribution network of the transformer substation into multiple distribution branches according to the boundary points of the distribution branches; S3.3. Apply the line loss estimation equation to multiple distribution branches, calculate the conductor loss and joint loss of each distribution branch, and add the conductor loss and joint loss to obtain the theoretical line loss value of each distribution branch.

[0009] In step S4, when the deviation between the theoretical line loss value and the measured line loss value exceeds a preset deviation threshold, the line loss contribution of each distribution branch is calculated according to the line loss estimation equation, and the branch with the largest line loss contribution is identified as the abnormal line loss branch, including: S4.1. Construct a time-series deviation sequence by combining the theoretical line loss value and the measured line loss value in chronological order. Normalize the time-series deviation sequence to obtain a deviation feature vector. Determine whether the deviation exceeds a preset deviation threshold based on the time-series distribution characteristics of the deviation feature vector. S4.2. Extract voltage and current coefficients from the line loss estimation equation, construct a parameter matrix based on the voltage and current coefficients, multiply the parameter matrix with the deviation eigenvector to obtain the line loss contribution of each distribution branch, and determine the distribution branch with the largest line loss contribution as the abnormal line loss branch.

[0010] In step S5, the ratio of compensation cost to line loss improvement value is calculated for abnormal line loss branches. This ratio is used as the objective function. The constraints of the objective function are adjusted according to the line loss contribution. The gradient descent method is used to solve the compensation strategy, including: S5.1. Decompose the active and reactive components according to the load change curve of the abnormal line loss branch, calculate the compensation capacity demand based on the active and reactive components, and add the investment cost and operation and maintenance cost corresponding to the compensation capacity demand to obtain the compensation cost. S5.2. Construct a branch impedance matrix based on the compensation capacity requirement and the line loss contribution. Calculate the difference in power loss before and after compensation based on the branch impedance matrix to obtain the line loss improvement value. Construct an objective function using the ratio of the compensation cost to the line loss improvement value. S5.3. Generate the constraint benchmark value of the objective function based on the line loss contribution, and multiply the change value of the line loss contribution by the constraint benchmark value to obtain the constraint adjustment amount; S5.4. Update the constraints of the objective function using the constraint adjustment amount, and use the gradient descent method to iteratively calculate the updated objective function to obtain the compensation strategy.

[0011] In step S5.2, a branch impedance matrix is ​​constructed based on the compensation capacity requirement and the line loss contribution. The line loss improvement value is obtained by calculating the power loss difference before and after compensation based on the branch impedance matrix. An objective function is constructed using the ratio of the compensation cost to the line loss improvement value, including: S5.2.1 Divide the compensation capacity demand into compensation response intervals according to the load fluctuation characteristics, construct an adaptive update mechanism for compensation weights based on the line loss contribution, dynamically adjust the impedance compensation coefficients of each compensation response interval using the adaptive update mechanism for compensation weights, calculate the impedance interconnection relationship between adjacent compensation response intervals based on the impedance compensation coefficients, and combine the impedance interconnection relationship to construct a branch impedance matrix. S5.2.2. Substitute the branch voltage and current data collected before compensation into the branch impedance matrix to calculate the power loss before compensation, and substitute the branch voltage and current data collected after compensation into the branch impedance matrix to calculate the power loss after compensation. Use the difference between the power loss before and after compensation as the line loss improvement value. S5.2.3. Calculate the dynamic compensation benefit index of the branch based on the impedance compensation coefficient, generate the priority compensation coefficient according to the dynamic compensation benefit index, and multiply the priority compensation coefficient by the ratio of compensation cost and line loss improvement value to obtain the objective function.

[0012] In step S6, the compensation strategy is executed and the constraints are dynamically adjusted until the line loss rate after compensation is less than the preset line loss rate threshold. The optimal compensation strategy is then output, converted into control commands, and sent to the execution device, including: S6.1. Parse the compensation strategy into a combination of compensation capacity switching information and switch status information, generate a device action timing table based on the compensation capacity switching information and switch status combination information, and execute the device action timing table; S6.2. Collect the branch voltage and current parameters after execution, and calculate the compensated line loss rate based on the branch voltage and current parameters; S6.3. Calculate the deviation between the compensated line loss rate and the preset line loss rate threshold, calculate the constraint adjustment amount based on the deviation value, and update the constraint conditions based on the constraint adjustment amount; S6.4. Regenerate and execute the compensation strategy according to the updated constraints. Repeat the above steps until the line loss rate after compensation is less than the preset line loss rate threshold to obtain the optimal compensation strategy. Convert the optimal compensation strategy into a control command and send it to the execution device.

[0013] The transformer substation line loss adaptive scene perception and dynamic optimization system provided in this embodiment of the invention includes: The data acquisition unit is used to collect voltage and current data from the distribution transformers in the transformer substation area. The scene recognition unit is used to perform time-series decomposition of voltage data and current data to obtain trend items and periodic items, calculate scene discrimination index based on the fluctuation amplitude of the trend item and the duration of the periodic item, and adaptively divide the operating status of the transformer area based on the scene discrimination index to obtain the operating scene. The branch division unit is used to construct line loss estimation equations based on the operating scenario and scenario discrimination index, divide the distribution network of the transformer area into multiple distribution branches according to the line loss estimation equations, and calculate the theoretical line loss value of each distribution branch. The anomaly determination unit is used to calculate the line loss contribution of each distribution branch according to the line loss estimation equation when the deviation between the theoretical line loss value and the measured line loss value exceeds the preset deviation threshold, and to determine the branch with the largest line loss contribution as the abnormal line loss branch. The compensation calculation unit is used to calculate the ratio of compensation cost to line loss improvement value for abnormal line loss branches, use the ratio as the objective function, adjust the constraints of the objective function according to the line loss contribution, and solve the compensation strategy using the gradient descent method. The compensation execution unit is used to execute the compensation strategy and dynamically adjust the constraints until the line loss rate after compensation is less than the preset line loss rate threshold, output the optimal compensation strategy, convert the optimal compensation strategy into control commands and send them to the execution device.

[0014] This invention accurately identifies the load characteristics of different operating scenarios in distribution areas by performing time-series decomposition of voltage and current data and calculating scenario discrimination indicators, thus improving the adaptability of line loss analysis. Based on the operating scenarios, a line loss estimation equation is constructed, enabling precise division of distribution branches and improving the accuracy of line loss calculation. Abnormal line loss branches are determined according to their line loss contribution, avoiding the limitations of traditional fixed threshold judgment methods. By using the ratio of compensation cost to line loss improvement value as the objective function and dynamically adjusting constraints based on line loss contribution, the economic efficiency of the compensation strategy is ensured, continuous optimization of the compensation effect is achieved, and the line loss rate is stably controlled within a preset threshold range. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart of the adaptive scene perception and dynamic optimization method for line loss in transformer areas provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the adaptive scene perception and dynamic optimization system for transformer area line loss provided in an embodiment of the present invention. Detailed Implementation

[0017] like Figure 1 As shown, this invention provides a method for adaptive scene perception and dynamic optimization of transformer area line loss, including the following steps: S1. Collect voltage and current data of the distribution transformers in the distribution area; S2. Perform time-series decomposition on voltage and current data to obtain trend and periodic terms. Calculate scenario discrimination indicators based on the fluctuation amplitude of the trend and the duration of the periodic terms. Adaptively divide the operating status of the transformer area based on the scenario discrimination indicators to obtain the operating scenario. S3. Construct line loss estimation equations based on operating scenarios and scenario discrimination indicators. Divide the distribution network of the transformer area into multiple distribution branches based on the line loss estimation equations and calculate the theoretical line loss value of each distribution branch. S4. When the deviation between the theoretical line loss value and the measured line loss value exceeds the preset deviation threshold, calculate the line loss contribution of each distribution branch according to the line loss estimation equation, and determine the branch with the largest line loss contribution as the abnormal line loss branch. S5. Calculate the ratio of compensation cost to line loss improvement value for abnormal line loss branches, use the ratio as the objective function, adjust the constraints of the objective function according to the contribution of line loss, and solve the compensation strategy using the gradient descent method. S6. Execute the compensation strategy and dynamically adjust the constraints until the line loss rate after compensation is less than the preset line loss rate threshold, output the optimal compensation strategy, convert the optimal compensation strategy into control commands and send them to the execution device.

[0018] In step S2, the voltage and current data are decomposed into time series to obtain trend and periodic terms. A scenario discrimination index is calculated based on the fluctuation amplitude of the trend term and the duration of the periodic term. Based on this scenario discrimination index, the operating status of the transformer substation is adaptively divided to obtain operating scenarios, including: S2.1. Perform empirical mode decomposition on the voltage and current data according to the time series to obtain multiple time-series intrinsic mode components; S2.2. Calculate the energy distribution value of each time series intrinsic mode component, construct a time series reconstruction matrix based on the energy distribution value, and reconstruct each time series intrinsic mode component based on the time series reconstruction matrix to obtain the trend term and periodic term; S2.3. Extract the fluctuation amplitude of the trend item and the duration of the period item, calculate the cross-correlation coefficient between the fluctuation amplitude and the duration, determine the weight coefficient based on the cross-correlation coefficient, and weight the fluctuation amplitude and duration according to the weight coefficient to obtain the scene discrimination index. S2.4. Calculate the data density distribution of the scene discrimination index, determine multiple cluster centers based on the data density distribution, and calculate the distance from the operating status of the transformer area to each cluster center; S2.5. Update the cluster center position based on the distance and reclassify the affiliation of the station area operation status until the cluster center position converges to obtain the operation scenario.

[0019] First, the voltage and current data are subjected to empirical mode decomposition (EMD) according to the time series to obtain multiple time-series intrinsic mode components (EMCs). Empirical mode decomposition is an adaptive time-frequency analysis method for nonlinear and non-stationary signals, extracting intrinsic mode functions from the signal through an iterative screening process. Specifically, for the collected voltage and current data of the transformer area, upper and lower envelopes containing all local maxima and minima are constructed respectively. The mean of the upper and lower envelopes is calculated, and the original signal is subtracted from the mean to obtain the first candidate component. The above process is repeated until the candidate component meets the conditions of the intrinsic mode function, thus becoming a time-series EMC. For example, if the voltage data of the transformer area for a certain time period is 220V, 219V, 221V, 223V, and 220V, multiple EMCs with different frequencies can be obtained after empirical mode decomposition, reflecting the fluctuation characteristics of voltage at different scales.

[0020] Next, the energy distribution values ​​of each time-series intrinsic mode component are calculated. A time-series reconstruction matrix is ​​constructed based on these values. The time-series reconstruction matrix is ​​then used to reconstruct each time-series intrinsic mode component, yielding trend and periodic terms. The energy distribution value is calculated by dividing the sum of the squares of each intrinsic mode component by the sum of the squares of the original signal, reflecting the proportion of each component in the original signal. The intrinsic mode components are sorted from largest to smallest energy distribution. Low-frequency components with a larger energy proportion are selected for reconstruction as the trend term, while high-frequency components with a smaller energy proportion are selected for reconstruction as the periodic term. Among the several intrinsic mode components obtained from the decomposition of the transformer voltage data, the sum of the energy of the top three components accounts for 85% of the total energy. The signals reconstructed from these three components are used as the trend term, and the signals reconstructed from the remaining components are used as the periodic term.

[0021] Subsequently, the fluctuation amplitude of the trend term and the duration of the periodic term are extracted. The cross-correlation coefficient between fluctuation amplitude and duration is calculated, and weighting coefficients are determined based on the cross-correlation coefficient. The fluctuation amplitude and duration are then weighted and combined according to the weighting coefficients to obtain the scenario discrimination index. The fluctuation amplitude refers to the difference between the maximum and minimum values ​​of the trend term within a unit of time, reflecting the drastic degree of load change in the distribution area. The duration refers to the length of time during which a similar fluctuation pattern is maintained continuously in the periodic term, reflecting the stability of the electricity consumption behavior in the distribution area. The cross-correlation coefficient is obtained by calculating the correlation between the fluctuation amplitude sequence and the duration sequence at different time delays. The correlation coefficient corresponding to the time delay point with the strongest correlation is used to determine the weights. For example, if the fluctuation amplitude of the trend term in a certain distribution area is 15V, the duration of the periodic term is 3 hours, and the cross-correlation coefficient is 0.72, then the weighting coefficient for the fluctuation amplitude is determined to be 0.6, the weighting coefficient for the duration is 0.4, and the weighted combination yields a scenario discrimination index of 10.2.

[0022] Next, the data density distribution of the scene discrimination index is calculated. Based on the data density distribution, multiple cluster centers are determined, and the distance from the operating status of the transformer substation to each cluster center is calculated. The data density distribution is obtained using the kernel density estimation method. This method places a kernel function at each data point location, and the superposition of all kernel functions forms a density curve. The local maxima of the density curve are selected as the initial cluster centers, representing high-density areas in the dataset. For each operating status of the transformer substation, the Euclidean distance between its scene discrimination index and the scene discrimination index of each cluster center is calculated. The smaller the distance, the closer the operating status is to the corresponding cluster center. For example, for the operating status of the transformer substation with a discrimination index of 10.2, its distances to the three cluster centers are calculated to be 2.3, 5.1, and 8.7, respectively, with the closest cluster center being the first one.

[0023] Finally, the cluster center positions are updated based on distance, and the affiliation of the station's operating status is redefined until the cluster center positions converge to obtain the operating scenarios. The method for updating the cluster centers is to calculate the average of the scenario discrimination index for all operating states belonging to the same category. After each cluster center update, the distance from all operating states to the new cluster center is recalculated, and the affiliation is adjusted according to the minimum distance principle. This iterative process continues until the change in cluster center positions is less than a preset threshold or the maximum number of iterations is reached. For example, through multiple iterations, the initial cluster centers are adjusted from 7.8, 12.5, and 18.2 to 8.3, 13.1, and 17.6, forming stable operating scenario division results, corresponding to low-load stable scenarios, medium-load fluctuating scenarios, and high-load drastic change scenarios, respectively.

[0024] This invention achieves adaptive classification of transformer substation operating states by performing time-series decomposition on voltage and current data, extracting the fluctuation amplitude of the trend term and the duration of the period term, and calculating scenario discrimination indicators. This method can accurately identify different operating scenarios for transformer substations, improving the accuracy of line loss analysis. Through adaptive scenario classification, corresponding optimization measures can be taken for different operating scenarios, reducing line losses in transformer substations and improving power grid operating efficiency. This method fully utilizes the time-series characteristics of voltage and current data, overcoming the limitations of traditional fixed-threshold classification methods, and realizing dynamic perception and intelligent classification of transformer substation operating scenarios.

[0025] In step S3, a line loss estimation equation is constructed based on the operating scenario and scenario discrimination index. The distribution network of the transformer substation is then divided into multiple distribution branches based on this equation, and the theoretical line loss value for each distribution branch is calculated, including: S3.1. Extract power flow distribution features based on the operating scenario, construct a dynamic weight matrix based on the scenario discrimination index, and combine the power flow distribution features with the dynamic weight matrix to construct a line loss estimation equation; S3.2. Calculate the voltage-current correlation degree of the distribution branches in the distribution network of the transformer substation according to the line loss estimation equation, determine the boundary points of the distribution branches according to the rate of change of the voltage-current correlation degree, and divide the distribution network of the transformer substation into multiple distribution branches according to the boundary points of the distribution branches; S3.3. Apply the line loss estimation equation to multiple distribution branches, calculate the conductor loss and joint loss of each distribution branch, and add the conductor loss and joint loss to obtain the theoretical line loss value of each distribution branch.

[0026] Power flow distribution characteristics refer to the distribution of power flow in a distribution network under specific operating scenarios, including active power distribution, reactive power distribution, power factor distribution, and load density distribution. The extraction method utilizes historical operating data of the distribution transformers in the distribution area, establishing statistical feature extraction techniques for each operating scenario. For low-load stable scenarios, the focus is primarily on load density distribution characteristics; for medium-load fluctuating scenarios, the focus is on extracting power factor distribution characteristics; for high-load drastic change scenarios, both active and reactive power distribution characteristics are considered. In a medium-load fluctuating scenario in a certain distribution area, the extracted power factor distribution characteristics range from 0.85 to 0.92, showing a slight decreasing trend with increasing branch distance.

[0027] The scenario discrimination index, as a comprehensive indicator characterizing the operating status of a transformer substation, is used to construct a dynamic weight matrix. The dynamic weight matrix is ​​a weight coefficient matrix that dynamically adjusts with the scenario discrimination index, used to balance the importance of various power flow distribution characteristics in line loss estimation. The scenario discrimination index is mapped to a preset weight interval, generating corresponding weight coefficients. The weight interval is determined through regression analysis of historical operating data of the transformer substation; a low scenario discrimination index corresponds to a high load density distribution weight, and a high scenario discrimination index corresponds to a high power distribution weight. For a transformer substation with a scenario discrimination index of 10.2, the constructed dynamic weight matrix has a load density distribution weight of 0.45, a power factor distribution weight of 0.35, an active power distribution weight of 0.12, and a reactive power distribution weight of 0.08.

[0028] The basic structure for line loss estimation is formed by weighting the power flow distribution characteristics according to the corresponding weight coefficients in the dynamic weight matrix. The line loss estimation equation is constructed using multiple regression analysis, with power flow distribution characteristics as independent variables and historical line loss rates as dependent variables, combined with parameter optimization using the dynamic weight matrix. The general form of the line loss estimation equation is a weighted linear combination of the line loss rate and each power flow distribution characteristic, plus a nonlinear correction term related to the operating scenario. In practical applications, the line loss estimation equation also needs to be appropriately adjusted according to the topological characteristics of the distribution network in the transformer substation. For transformer substations with medium load fluctuation scenarios, the coefficient for the power factor distribution characteristic in the line loss estimation equation is 0.35, the coefficient for the load density distribution characteristic is 0.45, and the nonlinear correction term is 0.03, together forming a complete line loss estimation structure.

[0029] When calculating the voltage-current correlation of distribution branches in a transformer substation network based on the line loss estimation equation, the equation needs to be applied to each node in the network to calculate the ratio of voltage change rate to current change rate between adjacent nodes. The voltage-current correlation is an indicator that measures the correlation between voltage and current changes between adjacent nodes in a distribution network; a higher correlation indicates stronger electrical coupling between nodes. The calculation method is to take the ratio of the voltage difference to the current difference between adjacent nodes and multiply it by the corresponding weighting coefficient in the line loss estimation equation. In a transformer substation network, the voltages of two adjacent nodes are 220V and 218V, and the currents are 45A and 43A, respectively. The corresponding weighting coefficient in the line loss estimation equation is 0.35, and the calculated voltage-current correlation is 0.35 × (220 - 218) / (45 - 43) = 0.35.

[0030] Determining the boundary points of distribution branches based on the rate of change of voltage-current correlation is a crucial step in identifying natural segmentation points in a distribution network. The rate of change of voltage-current correlation refers to the degree of change in voltage-current correlation calculated node by node along the distribution line. Nodes with a large rate of change often indicate significant changes in load characteristics or line parameters, making them suitable as boundary points of distribution branches. The determination method involves calculating the rate of change of voltage-current correlation for each node, marking nodes with rates exceeding a preset threshold as candidate boundary points, and then using cluster analysis to filter out the true boundary points from the candidate points. The preset threshold is typically set to 1.5 to 2 times the average rate of change of voltage-current correlation. In a certain distribution network, the voltage-current correlation between node 1 and node 2 is 0.35, and the voltage-current correlation between node 2 and node 3 is 0.52, with a rate of change of 0.17, exceeding the preset threshold of 0.15. Therefore, node 2 is determined as the boundary point of the distribution branch.

[0031] When dividing a transformer substation's distribution network into multiple distribution branches based on their boundary points, it's crucial to ensure the division conforms to the physical characteristics and operational rules of the power system. Starting from the transformer outgoing line, proceed segment by segment along the distribution line. Upon encountering a boundary point, mark the current path as a distribution branch, and then begin a new path from that boundary point. The division process must consider the network's topology, handling complex structures such as branches and loops. After division, each distribution branch should include a clearly defined start and end point, as well as all nodes and connections along its path. For complex transformer substations, iterative optimization of the division results may be necessary to ensure its rationality and practicality. For example, a transformer substation's distribution network, after division, forms three distribution branches: branch A (containing nodes 1 and 2), branch B (containing nodes 2, 3, and 4), and branch C (containing nodes 2, 5, and 6).

[0032] When applying the line loss estimation equation to multiple distribution branches and calculating the conductor loss of each branch, factors such as conductor type, length, cross-sectional area, and current carrying capacity must be considered. Conductor loss is mainly generated by line impedance and current, and is calculated by summing the square of the branch current and the conductor resistance. Conductor resistance is related to conductor material, cross-sectional area, and temperature, and is usually obtained using empirical formulas or table lookup methods. In practical applications, the influence of conductor temperature changes with load must also be considered, and a temperature correction factor must be used for adjustment. For example, a distribution branch with a length of 200m and a cross-sectional area of ​​70mm²... 2 The aluminum wire has a resistivity of 0.0328 Ω·mm² / m and a current of 45 A. The calculated wire resistance is 0.0328 × 200 / 70 = 0.094 Ω, and the wire loss is 0.094 × 45 A. 2 =190.35W.

[0033] Joint loss is an easily overlooked but significantly impactful part of line loss calculations, primarily caused by the contact resistance at the connection points. Joint loss is related to factors such as the number of joints, joint type, degree of aging, and current carrying capacity. It is calculated by summing the product of the contact resistance of each joint and the square of the current flowing through it. Contact resistance is usually estimated using empirical formulas or obtained through regression analysis based on historical maintenance data. In line loss estimation, joint loss typically accounts for 5% to 15% of total line loss, increasing with the degree of joint aging. For newly installed distribution branches, the proportion of joint loss is lower; for distribution branches that have been in operation for more than 5 years, the proportion may exceed 10%. For example, a distribution branch with 3 connection joints, an average contact resistance of 0.002Ω, and a current of 45A, calculates a joint loss of 3 × 0.002 × 45² = 12.15W.

[0034] The final step in line loss calculation is to add the conductor loss and joint loss to obtain the theoretical line loss value for each distribution branch. The theoretical line loss value is the expected line loss level calculated under ideal conditions based on the line loss estimation equation and branch parameters, serving as a benchmark for subsequent abnormal line loss detection. During the calculation, it is important to ensure unit consistency; power loss is typically converted to kW or MW, and the power loss and line loss rate are calculated. The line loss rate is the ratio of the theoretical line loss value to the input power of the distribution branch, expressed as a percentage. After comprehensively considering the branch characteristics, a correction factor can be added to improve the accuracy of the theoretical line loss value. For example, in a distribution branch with a conductor loss of 190.35W and a joint loss of 12.15W, the theoretical line loss is 190.35 + 12.15 = 202.5W. With an input power of 15kW, the line loss rate is 202.5 / 15000 = 1.35%.

[0035] This invention can accurately identify the boundary points of distribution branches, enabling reasonable division of the distribution network; it can calculate conductor loss and joint loss separately, providing precise location for subsequent line loss management; it adapts to the operating conditions of transformer substations with different load characteristics and power flow distributions, exhibiting strong versatility and adaptability; by introducing a dynamic weight matrix, the line loss calculation becomes more closely aligned with actual operating conditions, reducing estimation errors; it achieves refined analysis of line losses from overall to local perspectives, providing a scientific basis for transformer substation line loss management, ultimately achieving the goal of reducing transformer substation line loss rate and improving power supply efficiency.

[0036] In step S4, when the deviation between the theoretical line loss value and the measured line loss value exceeds a preset deviation threshold, the line loss contribution of each distribution branch is calculated according to the line loss estimation equation, and the branch with the largest line loss contribution is identified as the abnormal line loss branch, including: S4.1. Construct a time-series deviation sequence by combining the theoretical line loss value and the measured line loss value in chronological order. Normalize the time-series deviation sequence to obtain a deviation feature vector. Determine whether the deviation exceeds a preset deviation threshold based on the time-series distribution characteristics of the deviation feature vector. S4.2. Extract voltage and current coefficients from the line loss estimation equation, construct a parameter matrix based on the voltage and current coefficients, multiply the parameter matrix with the deviation eigenvector to obtain the line loss contribution of each distribution branch, and determine the distribution branch with the largest line loss contribution as the abnormal line loss branch.

[0037] A time-series deviation sequence is constructed by arranging the theoretical and measured line loss values ​​in chronological order. Power data of the total outgoing circuits of the transformer substation are collected using smart meters or dedicated monitoring devices, and the measured line loss value is obtained after deducting user electricity consumption. For example, taking a certain transformer substation as an example, the theoretical and measured line loss values ​​for each day of the week are collected. The theoretical line loss values ​​are 28.01, 27.85, 28.22, 27.98, 28.56, 29.01, and 28.45 kWh, respectively, and the measured line loss values ​​are 28.35, 28.02, 28.45, 31.25, 32.78, 33.12, and 32.95 kWh, respectively. The deviations between the two are calculated to be 0.34, 0.17, 0.23, 3.27, 4.22, 4.11, and 4.50 kWh, respectively. Arranging these values ​​in chronological order constitutes the time-series deviation sequence.

[0038] The time-series deviation sequence is normalized to obtain the deviation eigenvector. The purpose of normalization is to eliminate the influence of dimensions and make the deviation values ​​comparable. The normalization method is to divide each deviation value in the time-series deviation sequence by the average of the total theoretical line loss values ​​for that transformer area, obtaining the relative deviation value. For the above deviation sequence, the average total theoretical line loss value for the transformer area is 28.30 kWh, and the normalized deviation eigenvectors are 0.012, 0.006, 0.008, 0.116, 0.149, 0.145, and 0.159. Normalization eliminates the influence of differences in transformer area size, making the line loss deviations of different transformer areas comparable.

[0039] The determination of whether a preset deviation threshold is exceeded is based on the temporal distribution characteristics of the deviation feature vector. The temporal distribution characteristics refer to the changing patterns of the deviation feature vector over time, including mean, variance, and trend. The preset deviation threshold is determined based on historical operating data of the transformer area and professional experience, and is typically set between 0.05 and 0.10. In the example above, the mean of the deviation feature vector is 0.085, the variance is 0.005, the maximum value is 0.159, and the minimum value is 0.006. The deviation values ​​for the first three days are all less than the preset threshold of 0.05, while the deviation values ​​for the following four days are all greater than the preset threshold, showing a continuous upward trend. Therefore, it is determined that the line loss deviation exceeds the preset deviation threshold, indicating an abnormal line loss situation.

[0040] Voltage and current coefficients are extracted from the line loss estimation equation, which contains multiple variables and their coefficients, including voltage and current. The voltage coefficient reflects the impact of voltage changes on line loss, and the current coefficient reflects the impact of current changes on line loss. Parameter separation is performed on the line loss estimation equation to obtain the coefficients of voltage-related and current-related terms. For the three distribution branches in the example above, the voltage coefficient for the first branch is 0.015, and the current coefficient is 0.035; the voltage coefficient for the second branch is 0.012, and the current coefficient is 0.028; and the voltage coefficient for the third branch is 0.013, and the current coefficient is 0.032.

[0041] A parameter matrix is ​​constructed based on voltage and current coefficients. This parameter matrix is ​​a two-dimensional array with two rows, each corresponding to a voltage coefficient, and two columns, representing the number of distribution branches. The voltage and current coefficients of each distribution branch are arranged sequentially to form a matrix. The line loss contribution of each distribution branch is obtained by multiplying the parameter matrix by the deviation eigenvector. This contribution represents the proportion of each distribution branch's contribution to the total line loss deviation. The calculation method involves performing a dot product operation between each row of the parameter matrix and the deviation eigenvector to obtain the corresponding line loss contribution of the distribution branch. To perform the dot product operation, the deviation eigenvector needs to be expanded to a vector with the same number of columns as the parameter matrix. This expansion is achieved by calculating the mean and variance of the deviation eigenvector, which are then used as the two elements of the expanded vector.

[0042] The distribution branch with the largest line loss contribution is identified as the abnormal line loss branch. The line loss contribution of each distribution branch is compared, and the branch with the largest line loss contribution is selected as the abnormal line loss branch. The largest line loss contribution means that the branch has the greatest impact on the total line loss deviation, and there is a high probability that there are abnormal line loss problems, such as aging wires, loose joints, or illegal connections.

[0043] The preset deviation threshold and line loss contribution calculation method can be adjusted according to the specific conditions of the transformer substation to improve the accuracy of abnormal line loss branch location. For larger or more complex transformer substations, the distribution branch division can be further refined to improve the accuracy of anomaly location. Furthermore, historical data can be combined with trend analysis to identify the development patterns of abnormal line losses, enabling early warning and handling of potential problems.

[0044] This invention can promptly detect abnormal line losses in the distribution network of a transformer substation, preventing energy waste; it achieves line loss analysis from the overall transformer substation to specific branches, improving the accuracy of anomaly location; it uses normalization processing and parameter matrix technology to eliminate the impact of scale differences between different transformer substations; through quantitative analysis of line loss contribution, it accurately identifies abnormal line loss branches, providing precise guidance for subsequent maintenance; it reduces the workload of manual inspections, lowers operation and maintenance costs, and improves power grid operating efficiency; it provides technical support for the refined management of line losses in transformer substations, contributing to the goal of intelligent operation and maintenance and energy conservation in the distribution network.

[0045] In step S5, the ratio of compensation cost to line loss improvement value is calculated for abnormal line loss branches. This ratio is used as the objective function. The constraints of the objective function are adjusted according to the line loss contribution. The gradient descent method is used to solve the compensation strategy, including: S5.1. Decompose the active and reactive components according to the load change curve of the abnormal line loss branch, calculate the compensation capacity demand based on the active and reactive components, and add the investment cost and operation and maintenance cost corresponding to the compensation capacity demand to obtain the compensation cost. S5.2. Construct a branch impedance matrix based on the compensation capacity requirement and the line loss contribution. Calculate the difference in power loss before and after compensation based on the branch impedance matrix to obtain the line loss improvement value. Construct an objective function using the ratio of the compensation cost to the line loss improvement value. S5.3. Generate the constraint benchmark value of the objective function based on the line loss contribution, and multiply the change value of the line loss contribution by the constraint benchmark value to obtain the constraint adjustment amount; S5.4. Update the constraints of the objective function using the constraint adjustment amount, and use the gradient descent method to iteratively calculate the updated objective function to obtain the compensation strategy.

[0046] A load variation curve refers to the data set showing the load change over time for a branch with abnormal line loss within a certain period. Data is typically collected in hourly increments over 24 hours or 7 days. The active component represents the actual effective power consumed, while the reactive component represents the power required to establish the electromagnetic field. The decomposition method involves calculating the power factor from the collected voltage and current data to obtain the active and reactive power. In practical applications, the Fourier transform method can be used to decompose the load variation curve into fundamental and harmonic components for more accurate acquisition of the active and reactive components. For a branch with abnormal line loss, the average active power of the 24-hour load curve is 85kW, the average reactive power is 45kVar, the power factor is 0.883, and the load fluctuation range is 70kW to 110kW.

[0047] Calculating the compensation capacity requirement based on active and reactive components is a crucial step in determining the specifications of the required compensation equipment. The compensation capacity requirement refers to the size of the reactive power compensation equipment needed to improve line losses, measured in kVar. The calculation method involves determining the required reactive power compensation capacity based on the magnitude and fluctuation characteristics of the reactive components, combined with the target power factor. The target power factor is typically set between 0.95 and 0.98. The compensation capacity calculation considers load fluctuations, generally taking 80% to 90% of the reactive component as the basic compensation capacity, and adding a 10% to 20% margin based on load fluctuations. For the aforementioned abnormal line loss branch, if the target power factor is 0.95, the calculated basic compensation capacity is 35 kVar. Considering a 15% load fluctuation margin, the final compensation capacity requirement is determined to be 40 kVar.

[0048] Adding the investment costs and operation and maintenance costs corresponding to the compensation capacity demand to obtain the compensation cost is a crucial step in economic evaluation. Investment costs include the purchase price, installation cost, and auxiliary equipment cost of the compensation equipment. These costs are directly proportional to the compensation capacity, but there is a scale effect; the investment cost per unit capacity decreases as capacity increases. Operation and maintenance costs include equipment maintenance, repairs, replacement of parts, and energy losses. These costs are typically calculated annually and are related to the equipment's lifespan and the operating environment. When calculating the compensation cost, the operation and maintenance costs must be discounted to their present value over the equipment's lifespan and added to the investment costs to obtain the total compensation cost. For a compensation capacity requirement of 40kVar, the unit price of the equipment is 50 yuan / kVar, the installation cost is 20% of the equipment cost, the auxiliary equipment cost is 15% of the equipment cost, the annual operation and maintenance cost is 5% of the investment cost, the equipment service life is 10 years, and the discount rate is 5%. The calculated investment cost is 40 × 50 × (1 + 0.2 + 0.15) = 5400 yuan, and the present value of the operation and maintenance cost is 5400 × 0.05 × [1 - (1 + 0.05)]. -10 ] / 0.05=4169 yuan, and the total compensation cost is 5400+4169=9569 yuan.

[0049] Constructing a branch impedance matrix based on compensation capacity requirements and line loss contribution is a prerequisite for calculating line loss improvement values. The branch impedance matrix describes the electrical characteristics of a distribution branch, including the branch's resistance, reactance, and the coupling relationships between them. The construction method involves calculating the basic impedance value based on the branch's physical parameters (such as wire diameter, length, and material), and then weighting the impedance value according to the line loss contribution; the higher the line loss contribution, the larger the adjustment coefficient. In practical applications, the branch impedance matrix also needs to consider the influence of temperature, frequency, and load level on the impedance. For an abnormal line loss branch with a line loss contribution of 0.75, a resistance baseline of 0.32 Ω / km, a reactance baseline of 0.35 Ω / km, and a branch length of 1.2 km, the calculated branch impedance is (0.32 + j0.35) × 1.2 × (1 + 0.75 × 0.2) = (0.46 + j0.50) Ω, where 0.2 is the adjustment coefficient.

[0050] The line loss improvement value, calculated based on the difference in power loss before and after compensation using the branch impedance matrix, is an important indicator for evaluating the compensation effect. Power loss refers to the power consumed due to line impedance during power transmission, and is proportional to the square of the current and the impedance. The calculation method uses the branch impedance matrix and the load current to calculate the power loss before and after compensation, respectively; the difference between the two is the line loss improvement value. The load current after compensation is calculated using the power triangle relationship, that is, while keeping the active power constant, the total apparent power and current decrease due to the reduction in reactive power. For the above-mentioned abnormal line loss branch, the power factor before compensation is 0.883, the current is 120A, and the branch impedance is (0.46+j0.50)Ω, so the calculated power loss before compensation is 120²×0.46=6624W; after compensation, the power factor increases to 0.95, and the current decreases to 112A, so the calculated power loss after compensation is 112A. 2 ×0.46=5773W, the line loss improvement value is 6624-5773=851W. If calculated based on 8760 hours of operation per year, the annual line loss improvement is 851×8760=7454.76kWh. Calculated at an electricity price of 0.6 yuan / kWh, the annual economic benefit is 7454.76×0.6=4472.86 yuan.

[0051] Constructing an objective function based on the ratio of compensation cost to line loss improvement is the core of optimizing compensation strategies. The objective function is a mathematical expression guiding the optimization of the compensation strategy; this ratio reflects the compensation input required for a unit of line loss improvement, and a smaller ratio indicates a higher return on investment. The construction method involves dividing the compensation cost by the annualized equivalent of the line loss improvement value to obtain the return on investment index. In practical applications, the objective function can also incorporate weighting coefficients to consider factors such as equipment lifespan, maintenance difficulty, and reliability. For the example above, the compensation cost is 9569 yuan, and the annual line loss improvement economic benefit is 4472.86 yuan. The calculated return on investment is 9569 / 4472.86 = 2.14, meaning that every 2.14 yuan invested yields 1 yuan of annual line loss improvement benefit. Considering a 10-year equipment lifespan, the overall return on investment is 0.214, which is less than 1, indicating that the investment is economically feasible.

[0052] Generating constraint benchmark values ​​for the objective function based on line loss contribution is a crucial step in ensuring the rationality of the compensation strategy. Constraint benchmark values ​​refer to the baseline standards for each constraint condition during the objective function optimization process, used to limit the feasible range of the compensation strategy. The generation method involves setting different levels of constraint benchmark values ​​based on the line loss contribution; the higher the line loss contribution, the more lenient the constraints, prioritizing the resolution of abnormal line losses with high contributions. Constraint benchmark values ​​typically include maximum investment limit, minimum rate of return, longest payback period, and minimum power factor. For an abnormal line loss branch with a line loss contribution of 0.75, the maximum investment limit constraint benchmark value can be set to 120% of the standard value (12,000 yuan), the minimum rate of return constraint benchmark value to 80% of the standard value (8%), the longest payback period constraint benchmark value to 120% of the standard value (3 years), and the minimum power factor constraint benchmark value to 0.92.

[0053] Multiplying the change in line loss contribution by the constraint baseline value to obtain the constraint adjustment amount is a key step in dynamically optimizing constraints. The change in line loss contribution refers to the amount of change in the line loss contribution relative to its initial value during the compensation process. The constraint adjustment amount represents the magnitude of adjustment to the constraints based on the change in line loss contribution, used to dynamically adjust constraints during optimization. The calculation method involves multiplying the change in line loss contribution by a preset sensitivity coefficient and the constraint baseline value to obtain the specific adjustment amount. The sensitivity coefficient is typically set between 0.5 and 2, reflecting the degree to which the constraints are sensitive to changes in line loss contribution. In a certain iteration, the line loss contribution decreased from 0.75 to 0.68, a change of -0.07. The sensitivity coefficient was 1.5, and the maximum investment limit constraint baseline value was 12,000 yuan. The calculated constraint adjustment amount was -0.07 × 1.5 × 12,000 = -1,260 yuan, meaning the maximum investment limit was adjusted to 12,000 - 1,260 = 10,740 yuan.

[0054] Updating the constraints of the objective function using constraint adjustment amounts is a measure to ensure the adaptive nature of the optimization process. The updated constraints are new constraints formed by adding constraint adjustment amounts to the original constraints, and are used to guide the next round of optimization calculations. The update method involves adding or multiplying each constraint baseline value with its corresponding constraint adjustment amount to obtain the new constraint value. In practical applications, constraint updates also need to consider upper and lower boundaries to ensure that the updated constraints are within a reasonable range.

[0055] For the example above, the updated maximum investment limit constraint is 10,740 yuan, the minimum rate of return constraint is 8% + (-0.07) × 1.5 × 8% = 8% × (1 - 0.07 × 1.5) = 7.16%, the longest payback period constraint is 3 × (1 + 0.07 × 1.5) = 3.315 years, and the minimum power factor constraint is 0.92 + (-0.07) × 1.5 × 0.06 = 0.914, where 0.06 is the power factor adjustment coefficient.

[0056] The final step in the optimization process is to iteratively calculate the updated objective function using gradient descent to obtain the compensation strategy. Gradient descent is a numerical optimization algorithm that iteratively searches for the optimal solution along the negative gradient direction of the objective function. The iterative calculation steps include calculating the current objective function value, calculating the gradient of the objective function with respect to each parameter, updating the parameters along the negative gradient direction, and checking the convergence condition. The gradient calculation can use the numerical difference method. The parameter update step size is typically set between 0.01 and 0.1. The convergence condition is that the change in the objective function value between two consecutive iterations is less than a preset threshold (e.g., 0.001) or the maximum number of iterations (e.g., 100 iterations) is reached. In the example above, after iterative calculation using gradient descent, the optimal compensation capacity is 38 kVar, corresponding to an investment cost of 8800 yuan, an annual benefit of 4320 yuan for line loss improvement, and an input-output ratio of 2.04, which meets the updated constraints.

[0057] This invention ensures the rationality of investment returns, making compensation strategies more aligned with the actual conditions of specific branches. Through the decomposition and analysis of active and reactive components, it accurately calculates the compensation capacity requirements, avoiding over- or under-compensation, achieving precise management of abnormal line losses, and significantly reducing the line loss rate of distribution areas. It supports a strategy combining fixed and dynamic compensation to adapt to load fluctuation characteristics, providing a scientific basis for optimizing line losses in distribution areas, and improving the operating efficiency and power quality of the distribution network.

[0058] In step S5.2, a branch impedance matrix is ​​constructed based on the compensation capacity requirement and the line loss contribution. The line loss improvement value is obtained by calculating the power loss difference before and after compensation based on the branch impedance matrix. An objective function is constructed using the ratio of the compensation cost to the line loss improvement value, including: S5.2.1 Divide the compensation capacity demand into compensation response intervals according to the load fluctuation characteristics, construct an adaptive update mechanism for compensation weights based on the line loss contribution, dynamically adjust the impedance compensation coefficients of each compensation response interval using the adaptive update mechanism for compensation weights, calculate the impedance interconnection relationship between adjacent compensation response intervals based on the impedance compensation coefficients, and combine the impedance interconnection relationship to construct a branch impedance matrix. S5.2.2. Substitute the branch voltage and current data collected before compensation into the branch impedance matrix to calculate the power loss before compensation, and substitute the branch voltage and current data collected after compensation into the branch impedance matrix to calculate the power loss after compensation. Use the difference between the power loss before and after compensation as the line loss improvement value. S5.2.3. Calculate the dynamic compensation benefit index of the branch based on the impedance compensation coefficient, generate the priority compensation coefficient according to the dynamic compensation benefit index, and multiply the priority compensation coefficient by the ratio of compensation cost and line loss improvement value to obtain the objective function.

[0059] The compensation capacity demand is divided into compensation response intervals based on load fluctuation characteristics. Load fluctuation characteristics refer to the regularity of electricity load changes within a transformer area over time, which can be obtained by analyzing the 24-hour load curve. For a branch with abnormal line loss in a certain transformer area, load data for one week is collected, and analysis shows that this branch has a "double-peak" characteristic, with peak electricity consumption from 7-9 am and 6-10 pm. Based on this characteristic, the compensation capacity demand is divided into three compensation response intervals: peak interval, off-peak interval, and off-peak interval. For the calculated total compensation capacity demand, 60% of the compensation capacity is allocated in the peak interval, 30% in the off-peak interval, and 10% in the off-peak interval.

[0060] An adaptive update mechanism for compensation weights is constructed based on line loss contribution. Line loss contribution is an indicator that measures the degree of influence of a branch on the total line loss; the higher the value, the greater the contribution of that branch to the line loss. The adaptive update mechanism for compensation weights is a method for dynamically adjusting the allocation of compensation resources, which can automatically adjust the compensation weights of each compensation response interval according to changes in line loss contribution. When line loss contribution increases, the compensation weight of the peak interval increases, and the increase is proportional to the rate of change of line loss contribution; when line loss contribution decreases, the compensation weight of the peak interval decreases. The compensation weights of the off-peak and valley intervals are adjusted accordingly based on changes in the compensation weight of the peak interval, ensuring that the sum of the compensation weights of the three intervals is 1.

[0061] An adaptive update mechanism for compensation weights is used to dynamically adjust the impedance compensation coefficient for each compensation response interval. The impedance compensation coefficient refers to the degree of adjustment of the branch impedance by the compensation equipment; the larger the value, the more significant the compensation effect. The dynamic adjustment method combines the compensation weight with the branch impedance characteristics to calculate the impedance compensation coefficient for each compensation response interval. For peak intervals, the impedance compensation coefficient equals the compensation weight multiplied by the ratio of branch reactance to resistance; for off-peak and valley intervals, the calculation method is similar but slightly different. This differentiated calculation method takes into account the differences in compensation effect under different load levels.

[0062] The impedance interconnection between adjacent compensated response intervals is calculated based on the impedance compensation coefficient. This impedance interconnection describes the electrical coupling characteristics between different compensated response intervals and is a crucial parameter for constructing the branch impedance matrix. The calculation method involves multiplying the impedance compensation coefficients of two adjacent compensated response intervals, and then multiplying by an interconnection factor to obtain the impedance interconnection. The magnitude of the interconnection factor is related to the correlation between the load changes of the two compensated response intervals; the higher the correlation, the larger the interconnection factor. In this way, the impedance interconnection between three pairs of compensated response intervals is calculated.

[0063] The branch impedance matrix is ​​constructed by combining impedance interconnections. This symmetric matrix has diagonal elements representing the impedance characteristics of each compensation response interval, and off-diagonal elements representing the impedance interconnections between different compensation response intervals. The impedance compensation coefficients of each compensation response interval are used as diagonal elements, and the calculated impedance interconnections are used as the corresponding off-diagonal elements. This matrix comprehensively describes the impedance characteristics of abnormal line loss branches under different load levels and their internal interactions.

[0064] The power loss before compensation is calculated by substituting the branch voltage and current data collected before compensation into the branch impedance matrix. The power loss is calculated by multiplying the current vector by the branch impedance matrix, and then multiplying by the transpose of the current vector. For a branch with abnormal line loss in a certain transformer area, 24 hours of voltage and current data are collected before compensation to form a current vector. The current vector is then substituting into the constructed branch impedance matrix to calculate the power loss before compensation.

[0065] The compensated branch voltage and current data are substituted into the branch impedance matrix to calculate the power loss after compensation. The compensated current data is obtained through simulation calculation, taking into account the influence of the compensation equipment on the current magnitude and phase. The calculation method is the same as before compensation, and the difference in power loss before and after compensation is used as the line loss improvement value, which is then converted into an annual line loss improvement value.

[0066] The dynamic compensation benefit index of the branch is calculated based on the impedance compensation coefficient. This dynamic compensation benefit index is a comprehensive indicator that measures the compensation effect, considering the economic and technical feasibility of compensation under different load levels. The calculation method involves multiplying the impedance compensation coefficient of each compensation response interval by the duration of that interval and then summing the results. The calculated dynamic compensation benefit index reflects the comprehensive effect of the compensation strategy throughout the 24-hour period.

[0067] A priority compensation coefficient is generated based on the dynamic compensation benefit index. This coefficient determines the compensation priority when multiple abnormal line loss branches require simultaneous compensation. It is generated by dividing the dynamic compensation benefit index by a benchmark value. A higher dynamic compensation benefit index results in a larger priority compensation coefficient and a higher compensation priority. This mechanism ensures that limited compensation resources are used where they are most needed, improving the overall compensation effectiveness.

[0068] The objective function is obtained by multiplying the priority compensation coefficient by the ratio of compensation cost to line loss improvement. The objective function is the core of the optimization problem, representing the line loss improvement effect achieved per unit of input cost. The calculation method is to divide the compensation cost by the line loss improvement value to obtain the cost-benefit ratio, and then multiply this by the priority compensation coefficient to obtain the objective function value. The smaller the objective function value, the better the input-output ratio, and the higher the economic efficiency of the compensation strategy.

[0069] This invention divides compensation response intervals based on load fluctuation characteristics, making the compensation strategy more closely aligned with actual load change patterns. It can dynamically adjust the allocation of compensation resources according to changes in line loss contribution, quantifies the compensation effect using the impedance compensation coefficient method, and provides a basis for accurately calculating line loss improvement values. It considers the impedance interconnection relationship between adjacent compensation response intervals, comprehensively reflecting the branch impedance characteristics, and introduces dynamic compensation benefit indicators and priority compensation coefficients to achieve optimal allocation of compensation resources. The overall method is highly adaptable and can automatically adjust parameters according to transformer area characteristics, making it widely applicable.

[0070] In step S6, the compensation strategy is executed and the constraints are dynamically adjusted until the line loss rate after compensation is less than the preset line loss rate threshold. The optimal compensation strategy is then output, converted into control commands, and sent to the execution device, including: S6.1. Parse the compensation strategy into a combination of compensation capacity switching information and switch status information, generate a device action timing table based on the compensation capacity switching information and switch status combination information, and execute the device action timing table; S6.2. Collect the branch voltage and current parameters after execution, and calculate the compensated line loss rate based on the branch voltage and current parameters; S6.3. Calculate the deviation between the compensated line loss rate and the preset line loss rate threshold, calculate the constraint adjustment amount based on the deviation value, and update the constraint conditions based on the constraint adjustment amount; S6.4. Regenerate and execute the compensation strategy according to the updated constraints. Repeat the above steps until the line loss rate after compensation is less than the preset line loss rate threshold to obtain the optimal compensation strategy. Convert the optimal compensation strategy into a control command and send it to the execution device.

[0071] The compensation strategy analysis process involves parsing the calculated compensation strategy into specific compensation capacity switching information and switch status combination information. Compensation capacity switching information refers to the specific capacity value that is activated or deactivated for each compensation device, such as activating 50 kVar of compensation capacity for the reactive power compensation device of an abnormal line loss branch in a certain distribution area. Switch status combination information refers to the switching status of each compensation device, including closed, open, or maintaining the current state. For a certain abnormal line loss branch, the compensation strategy may require activating all compensation capacity during peak load periods, activating part of the compensation capacity during off-peak load periods, and deactivating all compensation capacity during off-peak load periods.

[0072] The process of generating a device action sequence table based on the combination of compensation capacity switching information and switch status information takes into account the operation sequence and time intervals of different devices. The device action sequence table includes the action type, action time, and execution priority for each device. Action types include closing, opening, and capacity regulation. Action time refers to the specific moment when the device performs the action, taking into account load change trends and grid operating conditions. Execution priority determines the execution order when multiple devices need to operate simultaneously. For reactive power compensation devices, opening operations are usually performed first, followed by closing operations, to avoid repeatedly operating the same device within a short period. For voltage regulation devices, small-scale adjustments are performed first, followed by large-scale adjustments, to avoid excessive voltage fluctuations. The device action sequence table generated in this way ensures that each device performs actions in a reasonable order and at reasonable time intervals, guaranteeing the safe and stable operation of the power grid.

[0073] The execution device action sequence table is implemented by sending control commands to each execution device. These control commands include a device identifier, action type, action parameters, and execution time. The device identifier uniquely identifies the device performing the action; the action type specifies the specific action the device needs to perform; the action parameters provide the specific values ​​required to perform the action; and the execution time specifies when the command is executed. The control commands are sent to each execution device via a communication network, and upon receiving the command, the execution device executes the corresponding action according to the specified time.

[0074] The branch voltage and current parameters after the execution of the compensation strategy are collected through the distribution network monitoring equipment in the transformer substation. The monitoring equipment collects electrical parameters such as voltage, current, and power factor of each branch at preset time points after the compensation strategy is completed. The collected voltage parameters include phase voltage, line voltage, and neutral point voltage; the current parameters include phase current and zero-sequence current. The collection frequency is once per minute, continuously for 30 minutes. The collected voltage and current parameters undergo data preprocessing to remove outliers and noise, and the average value is calculated as the branch voltage and current parameters.

[0075] The compensated line loss rate is calculated based on the collected branch voltage and current parameters. The calculation method involves substituting the branch voltage and current parameters into the branch impedance matrix to calculate the power loss, and then dividing the power loss by the input power to obtain the line loss rate. For a three-phase balanced system, the power loss equals the sum of the squares of the three-phase currents multiplied by the branch resistance. For a three-phase unbalanced system, the power loss also needs to consider the effects of zero-sequence current and mutual impedance. Input power refers to the total power input from the upstream power grid to the distribution area.

[0076] The deviation between the compensated line loss rate and the preset line loss rate threshold is calculated using a simple subtraction operation. The deviation equals the compensated line loss rate minus the preset line loss rate threshold. The preset line loss rate threshold is set based on the historical line loss level and management objectives of the transformer area, and is generally 1.2 times the normal line loss rate of the area. A deviation greater than zero indicates that the compensated line loss rate is still higher than the preset threshold, requiring further adjustment of the compensation strategy; a deviation less than or equal to zero indicates that the compensation effect meets the requirements, and the optimal compensation strategy can be output.

[0077] Calculating the constraint adjustment based on the deviation value involves an adaptive adjustment mechanism. The constraint adjustment is proportional to the deviation value; when the deviation value is large, the constraint adjustment is also large to accelerate convergence; when the deviation value is small, the constraint adjustment is also small to improve accuracy. The calculation method involves multiplying the deviation value by an adjustment coefficient to obtain the constraint adjustment. The adjustment coefficient is related to the historical adjustment results, with an initial value set to 0.5, which decreases as the number of iterations increases to avoid oscillations.

[0078] Updating constraints based on constraint adjustments refers to modifying the constraints in the compensation strategy optimization problem according to the calculated constraint adjustments. The constraints mainly include compensation capacity constraints, voltage constraints, and power factor constraints. For compensation capacity constraints, the maximum allowable compensation capacity is increased based on the constraint adjustment; for voltage constraints, the allowable voltage deviation range is appropriately relaxed based on the constraint adjustment; for power factor constraints, the power factor requirement is increased based on the constraint adjustment. When updating constraints, it is necessary to ensure that the new constraints remain within the physical limitations of the equipment and the safe operation range of the power grid.

[0079] Regenerating and executing the compensation strategy based on the updated constraints is an iterative optimization process. In each iteration, the optimization problem is resolved based on the updated constraints to obtain a new compensation strategy. This new strategy is then executed and its effectiveness evaluated. During the iteration process, the compensation strategy and corresponding line loss rate are recorded each time to track optimization progress. To improve efficiency, when the deviation value changes by no more than a preset threshold for three consecutive times, the optimization is considered close to convergence, and the iteration process can be terminated early.

[0080] The iteration process ends when the compensated line loss rate is less than the preset line loss rate threshold, and the optimal compensation strategy is output. The optimal compensation strategy refers to the strategy with the lowest overall cost and easiest implementation among all compensation strategies that meet the line loss rate requirements. The optimal compensation strategy is converted into control instructions through the strategy parsing module. These control instructions include the specific operation content and timing for each execution device. The control instructions are sent to each execution device via the communication network. Upon receiving the instructions, the execution devices either execute them immediately or perform the corresponding operations according to the specified time.

[0081] This invention improves compensation accuracy by adjusting compensation parameters through real-time feedback. It dynamically updates constraints based on the compensation effect and adopts an iterative optimization method, taking into account the operating characteristics and safety constraints of different equipment to generate a reasonable equipment action sequence table, thus ensuring the safe and stable operation of the power grid. The overall method is highly adaptable and can configure parameters according to the characteristics and management objectives of different distribution areas. It effectively reduces the line loss rate of distribution areas, improves power supply quality and economic benefits, and provides strong support for distribution area line loss management.

[0082] like Figure 2 As shown, the adaptive scene perception and dynamic optimization system for transformer area line loss includes: The data acquisition unit is used to collect voltage and current data from the distribution transformers in the transformer substation area. The scene recognition unit is used to perform time-series decomposition of voltage data and current data to obtain trend items and periodic items, calculate scene discrimination index based on the fluctuation amplitude of the trend item and the duration of the periodic item, and adaptively divide the operating status of the transformer area based on the scene discrimination index to obtain the operating scene. The branch division unit is used to construct line loss estimation equations based on the operating scenario and scenario discrimination index, divide the distribution network of the transformer area into multiple distribution branches according to the line loss estimation equations, and calculate the theoretical line loss value of each distribution branch. The anomaly determination unit is used to calculate the line loss contribution of each distribution branch according to the line loss estimation equation when the deviation between the theoretical line loss value and the measured line loss value exceeds the preset deviation threshold, and to determine the branch with the largest line loss contribution as the abnormal line loss branch. The compensation calculation unit is used to calculate the ratio of compensation cost to line loss improvement value for abnormal line loss branches, use the ratio as the objective function, adjust the constraints of the objective function according to the line loss contribution, and solve the compensation strategy using the gradient descent method. The compensation execution unit is used to execute the compensation strategy and dynamically adjust the constraints until the line loss rate after compensation is less than the preset line loss rate threshold, output the optimal compensation strategy, convert the optimal compensation strategy into control commands and send them to the execution device.

[0083] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for adaptive scene perception and dynamic optimization of transformer area line loss, characterized in that, Includes the following steps: S1. Collect voltage and current data of the distribution transformers in the distribution area; S2. Perform time-series decomposition on voltage and current data to obtain trend and periodic terms. Calculate scenario discrimination indicators based on the fluctuation amplitude of the trend and the duration of the periodic terms. Adaptively divide the operating status of the transformer area based on the scenario discrimination indicators to obtain the operating scenario. S3. Construct line loss estimation equations based on operating scenarios and scenario discrimination indicators. Divide the distribution network of the transformer area into multiple distribution branches based on the line loss estimation equations and calculate the theoretical line loss value of each distribution branch. S4. When the deviation between the theoretical line loss value and the measured line loss value exceeds the preset deviation threshold, calculate the line loss contribution of each distribution branch according to the line loss estimation equation, and determine the branch with the largest line loss contribution as the abnormal line loss branch. S5. Calculate the ratio of compensation cost to line loss improvement value for abnormal line loss branches, use the ratio as the objective function, adjust the constraints of the objective function according to the contribution of line loss, and solve the compensation strategy using the gradient descent method. S6. Execute the compensation strategy and dynamically adjust the constraints until the line loss rate after compensation is less than the preset line loss rate threshold, output the optimal compensation strategy, convert the optimal compensation strategy into control commands and send them to the execution device.

2. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 1, characterized in that, In step S2, the voltage and current data are decomposed into time series to obtain trend and periodic terms. A scenario discrimination index is calculated based on the fluctuation amplitude of the trend term and the duration of the periodic term. Based on this scenario discrimination index, the operating status of the transformer substation is adaptively divided to obtain operating scenarios, including: S2.

1. Perform empirical mode decomposition on the voltage and current data according to the time series to obtain multiple time-series intrinsic mode components; S2.

2. Calculate the energy distribution value of each time series intrinsic mode component, construct a time series reconstruction matrix based on the energy distribution value, and reconstruct each time series intrinsic mode component based on the time series reconstruction matrix to obtain the trend term and periodic term; S2.

3. Extract the fluctuation amplitude of the trend item and the duration of the period item, calculate the cross-correlation coefficient between the fluctuation amplitude and the duration, determine the weight coefficient based on the cross-correlation coefficient, and weight the fluctuation amplitude and duration according to the weight coefficient to obtain the scene discrimination index. S2.

4. Calculate the data density distribution of the scene discrimination index, determine multiple cluster centers based on the data density distribution, and calculate the distance from the operating status of the transformer area to each cluster center; S2.

5. Update the cluster center position based on the distance and reclassify the affiliation of the station area operation status until the cluster center position converges to obtain the operation scenario.

3. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 1, characterized in that, In step S3, a line loss estimation equation is constructed based on the operating scenario and scenario discrimination index. The distribution network of the transformer substation is then divided into multiple distribution branches based on this equation, and the theoretical line loss value for each distribution branch is calculated, including: S3.

1. Extract power flow distribution features based on the operating scenario, construct a dynamic weight matrix based on the scenario discrimination index, and combine the power flow distribution features with the dynamic weight matrix to construct a line loss estimation equation; S3.

2. Calculate the voltage-current correlation degree of the distribution branches in the distribution network of the transformer substation according to the line loss estimation equation, determine the boundary points of the distribution branches according to the rate of change of the voltage-current correlation degree, and divide the distribution network of the transformer substation into multiple distribution branches according to the boundary points of the distribution branches; S3.

3. Apply the line loss estimation equation to multiple distribution branches, calculate the conductor loss and joint loss of each distribution branch, and add the conductor loss and joint loss to obtain the theoretical line loss value of each distribution branch.

4. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 1, characterized in that, In step S4, when the deviation between the theoretical line loss value and the measured line loss value exceeds a preset deviation threshold, the line loss contribution of each distribution branch is calculated according to the line loss estimation equation, and the branch with the largest line loss contribution is identified as the abnormal line loss branch, including: S4.

1. Construct a time-series deviation sequence by combining the theoretical line loss value and the measured line loss value in chronological order. Normalize the time-series deviation sequence to obtain a deviation feature vector. Determine whether the deviation exceeds a preset deviation threshold based on the time-series distribution characteristics of the deviation feature vector. S4.

2. Extract voltage and current coefficients from the line loss estimation equation, construct a parameter matrix based on the voltage and current coefficients, multiply the parameter matrix with the deviation eigenvector to obtain the line loss contribution of each distribution branch, and determine the distribution branch with the largest line loss contribution as the abnormal line loss branch.

5. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 1, characterized in that, In step S5, the ratio of compensation cost to line loss improvement value is calculated for abnormal line loss branches. This ratio is used as the objective function. The constraints of the objective function are adjusted according to the line loss contribution. The gradient descent method is used to solve the compensation strategy, including: S5.

1. Decompose the active and reactive components according to the load change curve of the abnormal line loss branch, calculate the compensation capacity demand based on the active and reactive components, and add the investment cost and operation and maintenance cost corresponding to the compensation capacity demand to obtain the compensation cost. S5.

2. Construct a branch impedance matrix based on the compensation capacity requirement and the line loss contribution. Calculate the difference in power loss before and after compensation based on the branch impedance matrix to obtain the line loss improvement value. Construct an objective function using the ratio of the compensation cost to the line loss improvement value. S5.

3. Generate the constraint benchmark value of the objective function based on the line loss contribution, and multiply the change value of the line loss contribution by the constraint benchmark value to obtain the constraint adjustment amount; S5.

4. Update the constraints of the objective function using the constraint adjustment amount, and use the gradient descent method to iteratively calculate the updated objective function to obtain the compensation strategy.

6. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 5, characterized in that, In step S5.2, a branch impedance matrix is ​​constructed based on the compensation capacity requirement and the line loss contribution. The line loss improvement value is obtained by calculating the power loss difference before and after compensation based on the branch impedance matrix. An objective function is constructed using the ratio of the compensation cost to the line loss improvement value, including: S5.2.1 Divide the compensation capacity demand into compensation response intervals according to the load fluctuation characteristics, construct an adaptive update mechanism for compensation weights based on the line loss contribution, dynamically adjust the impedance compensation coefficients of each compensation response interval using the adaptive update mechanism for compensation weights, calculate the impedance interconnection relationship between adjacent compensation response intervals based on the impedance compensation coefficients, and combine the impedance interconnection relationship to construct a branch impedance matrix. S5.2.

2. Substitute the branch voltage and current data collected before compensation into the branch impedance matrix to calculate the power loss before compensation, and substitute the branch voltage and current data collected after compensation into the branch impedance matrix to calculate the power loss after compensation. Use the difference between the power loss before and after compensation as the line loss improvement value. S5.2.

3. Calculate the dynamic compensation benefit index of the branch based on the impedance compensation coefficient, generate the priority compensation coefficient according to the dynamic compensation benefit index, and multiply the priority compensation coefficient by the ratio of compensation cost and line loss improvement value to obtain the objective function.

7. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 1, characterized in that, In step S6, the compensation strategy is executed and the constraints are dynamically adjusted until the line loss rate after compensation is less than the preset line loss rate threshold. The optimal compensation strategy is then output, converted into control commands, and sent to the execution device, including: S6.

1. Parse the compensation strategy into a combination of compensation capacity switching information and switch status information, generate a device action timing table based on the compensation capacity switching information and switch status combination information, and execute the device action timing table; S6.

2. Collect the branch voltage and current parameters after execution, and calculate the compensated line loss rate based on the branch voltage and current parameters; S6.

3. Calculate the deviation between the compensated line loss rate and the preset line loss rate threshold, calculate the constraint adjustment amount based on the deviation value, and update the constraint conditions based on the constraint adjustment amount; S6.

4. Regenerate and execute the compensation strategy according to the updated constraints. Repeat the above steps until the line loss rate after compensation is less than the preset line loss rate threshold to obtain the optimal compensation strategy. Convert the optimal compensation strategy into a control command and send it to the execution device.

8. The adaptive scene perception and dynamic optimization method for transformer area line loss according to claim 7, characterized in that, The control instructions include the specific operation content and timing of each execution device, and the control instructions are sent to each execution device through a communication network.

9. A transformer substation line loss adaptive scene perception and dynamic optimization system, used to implement the transformer substation line loss adaptive scene perception and dynamic optimization method according to any one of claims 1-8, characterized in that, The system includes: The data acquisition unit is used to collect voltage and current data from the distribution transformers in the transformer substation area. The scene recognition unit is used to perform time-series decomposition of voltage data and current data to obtain trend items and periodic items, calculate scene discrimination index based on the fluctuation amplitude of the trend item and the duration of the periodic item, and adaptively divide the operating status of the transformer area based on the scene discrimination index to obtain the operating scene. The branch division unit is used to construct line loss estimation equations based on the operating scenario and scenario discrimination index, divide the distribution network of the transformer area into multiple distribution branches according to the line loss estimation equations, and calculate the theoretical line loss value of each distribution branch. The anomaly determination unit is used to calculate the line loss contribution of each distribution branch according to the line loss estimation equation when the deviation between the theoretical line loss value and the measured line loss value exceeds the preset deviation threshold, and to determine the branch with the largest line loss contribution as the abnormal line loss branch. The compensation calculation unit is used to calculate the ratio of compensation cost to line loss improvement value for abnormal line loss branches, use the ratio as the objective function, adjust the constraints of the objective function according to the line loss contribution, and solve the compensation strategy using the gradient descent method. The compensation execution unit is used to execute the compensation strategy and dynamically adjust the constraints until the line loss rate after compensation is less than the preset line loss rate threshold, output the optimal compensation strategy, convert the optimal compensation strategy into control commands and send them to the execution device.

Images