Parameter estimation-based intelligent diagnosis method for abnormal line loss of power distribution network
By using a method based on parameter estimation and convolutional neural networks, abnormal line losses in distribution networks can be quickly and accurately identified, solving the problems of low efficiency and misjudgment in existing line loss diagnosis technologies, and achieving efficient and accurate line loss management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-03-27
AI Technical Summary
Existing methods for diagnosing line losses in distribution networks suffer from low timeliness, reliance on manual judgment, and inaccurate calculations of theoretical line loss rates, leading to misjudgments of anomalies and making it difficult to effectively diagnose abnormal line loss problems.
A parameter estimation-based approach is adopted, which involves preprocessing data collected from the distribution network, estimating line impedance parameters using the steepest descent method, and combining this with a convolutional neural network for anomaly diagnosis. A multi-objective optimization model is then constructed to improve the accuracy and efficiency of parameter estimation.
It enables rapid and accurate identification of line loss anomalies, reduces false positives and false negatives, improves the efficiency and accuracy of line loss anomaly diagnosis, and enhances the accuracy of equipment parameters and the speed of optimization solution for unconstrained models.
Smart Images
Figure CN121741333A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of power distribution network management, and particularly relates to an intelligent diagnosis method for abnormal line loss of a power distribution network based on parameter estimation. BACKGROUND
[0002] A power distribution network is the last link of power transmission to users, and its line loss rate is a key indicator for measuring the planning level, operation efficiency and economic benefit of the power grid. However, abnormal line loss problems caused by line aging, insulation damage, connection point oxidation, electricity stealing and metering equipment failure and the like have long existed in the operation of the power distribution network, causing huge power loss and economic loss to power grid enterprises, so that effective methods need to be taken to diagnose the line loss abnormal problems and take corresponding loss reduction measures to improve the operation safety and economy of the power distribution network.
[0003] At present, the main method for line loss management is statistical analysis and comparison, which screens high-loss or negative-loss lines that may have line loss abnormalities by monthly statistics of line loss rate, comparison of theoretical line loss rate with actual line loss rate and the like, analyzes the abnormal reasons, and determines the abnormal reasons and takes loss reduction measures according to the analysis results.
[0004] The traditional line loss diagnosis method has problems of low timeliness and dependence on manual judgment, and in the process of calculating the theoretical line loss rate, it depends on the theoretical parameters of the equipment and ignores the influence of temperature, humidity, corrosion degree, equipment aging condition and the like on the equipment impedance, resulting in distortion of the theoretical line loss rate and easy misjudgment of abnormalities. SUMMARY
[0005] In view of the above problems, the purpose of the present application is to provide an intelligent diagnosis method for abnormal line loss of a power distribution network based on parameter estimation.
[0006] The specific technical scheme for realizing the purpose of the present application is as follows:
[0007] An intelligent diagnosis method for abnormal line loss of a power distribution network based on parameter estimation, comprising the following steps:
[0008] Step 1, collecting power distribution network data and performing preprocessing;
[0009] Step 2, based on the power distribution network data, based on the topological relationship of the power distribution network, taking power calculation and voltage calculation comprehensive optimization as the target, and using the steepest descent method to estimate the line impedance parameters of the power distribution network;
[0010] Step 3, calculating the theoretical line loss according to the parameter estimation result;
[0011] Step 4, using a pre-trained convolutional neural network to diagnose the abnormal reasons of the sample with line loss abnormality.
[0012] Compared with the prior art, the application has the beneficial effects that:
[0013] (1) The application provides an abnormal line loss intelligent diagnosis method for a power distribution network based on parameter estimation, which uses real-time measurement data, corrects model impedance parameters through a parameter estimation method, calculates a theoretical line loss rate, compares an actual line loss rate to identify line loss abnormalities, and uses a convolutional neural network to diagnose the causes of line loss abnormalities.
[0014] (2) The parameter estimation method provided by the application takes power calculation and voltage calculation as targets, constructs a multi-objective optimization model, avoids the situation that a single method makes parameter solving fall into local optimization, improves the accuracy of equipment parameters, reduces the calculation error of theoretical line loss, reduces the misjudgment and omission of line loss abnormalities, and selects the steepest descent method to solve the comprehensive optimal result of parameter estimation, thereby improving the optimization solving speed of the unconstrained model.
[0015] (3) In the abnormal diagnosis, the application uses a convolutional neural network to construct a line loss abnormality diagnosis model, the completed model can quickly analyze a large amount of sample data, and constantly updates the model parameters, thereby improving the efficiency and accuracy of line loss abnormality diagnosis.
[0016] The application will be further described below in combination with specific embodiments. BRIEF DESCRIPTION OF DRAWINGS
[0017] Figure 1 FIG. 1 is a flowchart of the abnormal line loss intelligent diagnosis method for the power distribution network based on parameter estimation of the application.
[0018] Figure 2 FIG. 2 is a power distribution network topology diagram in the embodiment of the application.
[0019] Figure 3 FIG. 3 is a flowchart of the power distribution network line impedance parameter estimation using the steepest descent method in the application.
[0020] Figure 4 FIG. 4 is a convolutional neural network structure diagram in the embodiment of the application.
[0021] Figure 5 FIG. 5 is an IEEE33 node model diagram in the embodiment of the application. DETAILED DESCRIPTION
[0022] EMBODIMENT
[0023] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. The described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.
[0024] As shown in the present application and claims, unless the context clearly indicates otherwise, the words "one", "an", "a", and / or "the" do not mean to specify a single number, but can also include a plurality. Generally, the terms "comprising" and "including" only indicate the inclusion of the steps and elements explicitly identified, and these steps and elements do not constitute an exclusive list, and the method or device can also include other steps or elements.
[0025] Unless specifically stated otherwise, the relative arrangement of components and steps, numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application. At the same time, it should be understood that the sizes of the various parts shown in the drawings are not drawn in accordance with the actual proportional relationship. The technology, methods and devices known to those skilled in the relevant art can not be discussed in detail, but in appropriate cases, the technology, methods and devices should be considered as part of the authorized description. In all examples shown and discussed here, any specific value should be interpreted as merely exemplary, and not as a limitation. Therefore, other examples of exemplary embodiments can have different values. It should be noted that similar reference numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.
[0026] In conjunction with Figure 1 An intelligent diagnosis method for abnormal line loss of a power distribution network based on parameter estimation includes the following steps:
[0027] Step 1, collecting power distribution network data and performing preprocessing;
[0028] The power distribution network data includes power distribution network topology data, device parameter data and power distribution network operation data, and the operation data is collected by a micro-synchronous phasor measurement unit (PMU) at a cycle of 15 minutes to collect node voltage, current and power
[0029] The preprocessing of the data includes smoothing processing and standardization of the data;
[0030] The smoothing processing uses a Savitzky-Golay smoothing filter to filter the original operation data collected to eliminate noise interference;
[0031] Standardization is to normalize the data according to the parameter distribution, eliminate the influence of dimension on neural network training:
[0032]
[0033] In the formula, x' is the standardized data, μ and σ are the mean and standard deviation of the original data respectively;
[0034] Step 2, according to the distribution network data, based on the distribution network topology, the optimal power calculation and voltage calculation are integrated as the target, and the steepest descent method is used to estimate the distribution network line impedance parameters:
[0035] First, the voltage-based parameter estimation is carried out:
[0036] Combined with Figure 2 , for the general case of radial low-voltage area topology structure, point O represents the first end of the area transformer, points L1, L2, L3... are user load nodes, and points A, B... are intermediate intersection nodes;
[0037] For any branch, take adjacent nodes L2 and B as an example, the node voltage relationship is
[0038]
[0039] In the formula, , represents the corresponding branch impedance; , are the voltages at nodes B and L2 respectively; is the current flowing from node B to L2, where , are the measured data.
[0040] The voltage at node B can be expressed as:
[0041]
[0042] Continue to calculate the voltage of the upstream node along the line, and finally the voltage calculation value at the first end of the area can be obtained The equation with line impedance parameters as variables can be expressed as:
[0043]
[0044] In the formula, is the first end voltage calculated from the end node L2; is the impedance parameter value of each branch between the end node L2 and the first end node O; n is the number of intermediate branches;
[0045] Starting from different end nodes, a first end voltage can be calculated, and the calculated value is compared with the actual measured value of the voltage at the first end of the transformer The error of voltage estimation is represented by the square of the difference between the estimated value and the measured value in time series T:
[0046]
[0047] Then, the power-based parameter estimation is performed:
[0048] For any branch in the transformer area, the current flowing through it is equal to the sum of the currents on all adjacent downstream lines, which can be represented as:
[0049]
[0050] In the formula, j is the line number; h is the number of branches downstream of line j;
[0051] The transformer power is divided into active power P and reactive power Q. The power value measured at the end node is added to the power loss caused by line impedance to obtain the estimated value of the transformer power:
[0052]
[0053] In the formula, m and n are the number of end nodes and the total number of line branches, respectively; 、 and represent the current flowing through line j and the corresponding impedance parameter
[0054] The error of power estimation is represented by the square of the difference between the estimated value and the measured value in time series T:
[0055]
[0056] In the formula, 、 represents the power estimation value at time t, 、 represents the power measured value at time t.
[0057] A multi-objective optimization model is constructed by comprehensively optimizing the power estimation and voltage estimation. The power estimation and voltage estimation are used as targets to solve the comprehensive optimal result, avoiding the impedance parameter estimation value from falling into a local optimal solution due to the pursuit of minimum voltage error or power error. The objective function of the multi-objective optimization model is:
[0058]
[0059]
[0060]
[0061] wherein, The weighting coefficient for the voltage term. This represents the error value in the voltage calculation. This represents the error value in the power estimation. This represents the starting-end voltage value calculated based on the measurement data of the end node Lk of the distribution network topology at time t. This represents the voltage value measured at time t, and m represents the number of terminal nodes. , This represents the estimated active and reactive power values at time t. , This represents the measured values of active and reactive power at time t.
[0062] Considering that this model falls under the category of unconstrained optimization with a large number of parameters, the steepest descent method (STDM) is chosen as the iterative method to ensure faster iteration. Figure 3 As shown;
[0063] When using the steepest descent method to iterate k times, from... Departure, to make The direction in which the function value decreases the most within the neighborhood is taken as the search direction. That is, the direction of the negative gradient:
[0064]
[0065] Iterating along this direction yields the following new variable values:
[0066]
[0067] in, Let the step size be denoted as . The step size for each iteration must satisfy the following condition:
[0068]
[0069] The convergence condition for model iteration is:
[0070]
[0071] Where ε is the convergence accuracy of the iteration, and after satisfying the convergence condition, the minimum value of the objective function is obtained. This is the optimal solution for the impedance parameters.
[0072] Step 3: Calculate the theoretical line loss based on the parameter estimation results;
[0073] Step 4: Use a pre-trained convolutional neural network to diagnose the causes of abnormalities in samples with line loss anomalies.
[0074] Theoretical line loss calculation is performed by using the estimated impedance parameters, and when the measured line loss rate is greater than or less than the set threshold value, and |measured line loss rate-theoretical line loss rate|≥a, a is the set threshold value, it is determined that the distribution network may have a line loss problem;
[0075] In this embodiment, the impedance parameters obtained by parameter estimation are used for theoretical line loss calculation, and when the measured line loss rate is greater than or less than the set threshold value, and |measured line loss rate-theoretical line loss rate|≥3%, it is determined that the distribution network may have a line loss abnormal problem. The common line loss abnormal reason classification is shown in the following table.
[0076] Table 1
[0077]
[0078] For samples that may have line loss problems, a convolutional neural network with an abnormal reason label sample set data is used to extract the feature relationship between the input data, determine the abnormal reason classification, and realize intelligent diagnosis of line loss.
[0079] In combination Figure 4 , the convolutional neural network includes an input layer, a hidden layer, and an output layer;
[0080] The hidden layer includes two convolutional layers, two pooling layers, and one fully connected layer.
[0081] The convolutional layer mapping formula is:
[0082]
[0083] In the formula, represents the jth value of the lth layer, represents the value of the l-1th layer, represents a weight matrix, represents a bias matrix, is the convolution kernel range, and f(·) represents the convolution function.
[0084] The maximum layer pooling is used to reduce the dimension of the high-dimensional matrix obtained by the convolutional layer operation, reduce the parameter and data dimension, and the calculation formula is:
[0085]
[0086] In the formula, represents the data value of each point in the pooling region, and max(·) is the maximum value function.
[0087] The fully connected layer calculates the results in the form of probability by using the features extracted from the previous layers, and finally outputs the class with the maximum probability. The formula of the fully connected layer is:
[0088]
[0089] where h(x) represents the output of the fully connected layer, and f(·) represents a nonlinear activation function of the fully connected layer:
[0090]
[0091] where, is the input of the fully connected layer.
[0092] In the training process, the convolutional neural network adopts a cross-entropy loss function as the loss function of the model, which is used to measure the difference between the model diagnosis result and the actual abnormal reason, and the model parameters are adjusted by minimizing the cross-entropy loss, so that the model can better approximate the true abnormal reason, and the accuracy of the line loss abnormal diagnosis is improved:
[0093]
[0094] where LOSS is the cross-entropy loss, and n is the total number of samples; is the actual abnormal label of sample i; is the model diagnosis result of sample i.
[0095] During the use of the model, new line loss abnormal sample data can be continuously supplemented to continuously train the model, and the accuracy and calculation efficiency of the model for line loss abnormal diagnosis can be further improved.
[0096] The effectiveness of the parameter estimation method is verified by using an IEEE33 node model, as shown in FIG. 1, which includes 33 nodes and 32 branches. The simulation measures the voltage, current, active power and reactive power data of each node every 15 minutes in a day. Considering the error existing in actual measurement, a random error within 0.5% is added accordingly. Figure 5
[0097] The initial impedance parameter value of the line is set to 1, the weight coefficient is set to 50, and the algorithm convergence precision ε is set to 10-10. The calculation is completed for 181 iterations, the calculation time is 6.32s, the average impedance error is 0.00091, and as the sample data increases, the error gradually decreases and tends to approach the actual impedance of the line.
[0098] Line loss abnormal diagnosis:
[0099] 10000 groups of sample data are randomly selected for model training, and the sample data includes topological relationship, device parameter, node voltage, current, active power, reactive power, measured line loss rate, and abnormal problem label. The sample data is preprocessed, the impedance estimation value is calculated by using the parameter estimation method, and the theoretical line loss rate is calculated. 80% is randomly selected as the training sample set, and 20% is randomly selected as the test sample set.
[0100] The convolutional neural network algorithm proposed in the present application is compared with other common intelligent algorithms such as random forest, LightGBM, and support vector machine, and the results of line loss anomaly diagnosis are compared.
[0101] The accuracy (ACC), precision (PRE), and recall (REC) are used as evaluation indexes of the model. The calculation formulas are as follows
[0102]
[0103]
[0104]
[0105] In the formula, N is the total number of abnormal reason types; i is the abnormal reason label; represents the number of samples diagnosed as i and actually as i, represents the number of samples not diagnosed as i and actually not as i, represents the number of samples diagnosed as i but actually not as i, represents the number of samples not diagnosed as i but actually as i. The value range of each index is [0, 1], and the closer to 1, the better the model diagnosis effect.
[0106] The performance of different algorithms is compared, and the convolutional neural network has better performance in accuracy, precision, and recall, can accurately diagnose the line loss abnormal reason, and improves the line loss management level.
[0107] Table 2 Comparison of performance of different algorithms
[0108] Intelligent algorithm Accuracy Precision Recall Convolutional neural network 96.45% 94.47% 95.28% Random forest 83.39% 84.22% 80.69% LightGBM 87.78% 90.21% 90.05% Support vector machine 90.53% 89.81% 88.24%
[0109] The present scheme also provides a power distribution network abnormal line loss intelligent diagnosis system based on parameter estimation, comprising the following modules:
[0110] A data acquisition module is used to acquire power distribution network data and perform preprocessing;
[0111] An impedance estimation module is used to estimate the line impedance parameters of the power distribution network based on the power distribution network topology relationship, with the power calculation and voltage calculation comprehensive optimization as the target, and the steepest descent method is used for parameter estimation.
[0112] An abnormal diagnosis module is used to calculate the theoretical line loss according to the parameter estimation result, and use the pre-trained convolutional neural network to diagnose the abnormal reason of the sample with line loss anomaly.
[0113] The scheme also provides a computer device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:
[0114] Step 1, collecting power distribution network data and preprocessing;
[0115] Step 2, according to the power distribution network data, based on the power distribution network topology relationship, taking the comprehensive optimization of power calculation and voltage calculation as the target, using the steepest descent method to estimate the line impedance parameters of the power distribution network;
[0116] Step 3, performing theoretical line loss calculation according to the parameter estimation result;
[0117] Step 4, using the pre-trained convolutional neural network to diagnose the abnormal reason of the sample with line loss anomaly.
[0118] The scheme also provides a computer storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the following steps:
[0119] Step 1, collecting power distribution network data and preprocessing;
[0120] Step 2, according to the power distribution network data, based on the power distribution network topology relationship, taking the comprehensive optimization of power calculation and voltage calculation as the target, using the steepest descent method to estimate the line impedance parameters of the power distribution network;
[0121] Step 3, performing theoretical line loss calculation according to the parameter estimation result;
[0122] Step 4, using the pre-trained convolutional neural network to diagnose the abnormal reason of the sample with line loss anomaly.
[0123] The above-mentioned embodiments only express one embodiment of the present application, and the description is more specific and detailed, but it cannot be understood as limiting the scope of the patent. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.
Claims
1. A method for intelligent diagnosis of abnormal line losses in distribution networks based on parameter estimation, characterized in that, Includes the following steps: Step 1: Collect data from the power distribution network and perform preprocessing. Step 2: Based on the distribution network data and the distribution network topology, with the goal of optimizing both power and voltage estimation, the steepest descent method is used to estimate the impedance parameters of the distribution network lines. Step 3: Calculate the theoretical line loss based on the parameter estimation results; Step 4: Use a pre-trained convolutional neural network to diagnose the causes of abnormalities in samples with line loss anomalies.
2. The intelligent diagnosis method for abnormal line losses in distribution networks based on parameter estimation according to claim 1, characterized in that, The distribution network data includes distribution network topology data, equipment parameter data, and distribution network operation data; The data preprocessing includes data smoothing and standardization.
3. The intelligent diagnosis method for abnormal line losses in distribution networks based on parameter estimation according to claim 1, characterized in that, The estimation of distribution network line impedance parameters in step 2 is specifically as follows: A multi-objective optimization model is constructed by combining power estimation and voltage estimation. ； ； ； in, The weighting coefficient for the voltage term. This represents the error value in the voltage calculation. This represents the error value in the power estimation. This represents the starting-end voltage value calculated based on the measurement data of the end node Lk of the distribution network topology at time t. This represents the voltage value measured at time t, and m represents the number of terminal nodes. , This represents the estimated active and reactive power values at time t. , This represents the measured values of active and reactive power at time t.
4. The intelligent diagnosis method for abnormal line losses in distribution networks based on parameter estimation according to claim 3, characterized in that, The aforementioned estimation of the impedance parameters of the distribution network lines specifically includes: The steepest descent method is selected as the iterative method. When using the steepest descent method to iterate k times, from... Departure, to make The direction in which the function value decreases the most within the neighborhood is taken as the search direction. That is, the direction of the negative gradient: ； Iterating along this direction yields the following new variable values: ； in, Let the step size be denoted as . The step size for each iteration must satisfy the following condition: ； The convergence condition for model iteration is: ； Where ε is the convergence accuracy of the iteration, and after satisfying the convergence condition, the minimum value of the objective function is obtained. This is the optimal solution for the impedance parameters.
5. The intelligent diagnosis method for abnormal line losses in distribution networks based on parameter estimation according to claim 1, characterized in that, The abnormal line loss diagnosis in step 4 specifically includes: Theoretical line loss is calculated using the estimated impedance parameters. When the measured line loss rate exceeds or falls below a set threshold, and when |measured line loss rate - theoretical line loss rate| ≥ a, where a is the set threshold. It was determined that there might be line loss issues in the distribution network; For samples that may have line loss problems, a convolutional neural network is trained using sample set data with abnormal cause labels to extract the feature relationships between input data, determine the abnormal cause classification, and realize intelligent line loss diagnosis.
6. The intelligent diagnosis method for abnormal line losses in distribution networks based on parameter estimation according to claim 5, characterized in that, The convolutional neural network includes an input layer, a hidden layer, and an output layer; The hidden layer includes two convolutional layers, two pooling layers, and one fully connected layer. The mapping formula for the convolutional layer is: ； In the formula, This represents the j-th value in the l-th layer. This represents the value of the (l-1)th layer. Represents the weight matrix, Represents the bias matrix. The range of the convolution kernel is defined, and f(·) represents the convolution function. Max pooling is used to reduce the dimensionality of the high-dimensional matrix obtained from convolutional layer operations, thereby reducing the parameters and data dimensions. The calculation formula is as follows: ； In the formula, This represents the data values at each point within the pooling region, and max(·) is the maximum value function; The fully connected layer calculates the probabilities of the features extracted by the previous layers and finally outputs the category with the highest probability. The formula for the fully connected layer is: ； In the formula, h(x) represents the output of the fully connected layer, and f(·) represents the nonlinear activation function of the fully connected layer: ； In the formula, This is the input for the fully connected layer.
7. The intelligent diagnosis method for abnormal line losses in distribution networks based on parameter estimation according to claim 5, characterized in that, During training, the convolutional neural network uses the cross-entropy loss function as the model's loss function to measure the difference between the model's diagnostic results and the actual causes of anomalies. By minimizing the cross-entropy loss, the model parameters are adjusted to better approximate the true causes of anomalies, thereby improving the accuracy of line loss anomaly diagnosis. ； In the formula, LOSS is the cross-entropy loss, and n is the total number of samples; The actual anomaly label for sample i; The diagnostic results for sample i are from the model.
8. A smart diagnostic system for abnormal line losses in a distribution network based on parameter estimation, characterized in that, Includes the following modules: Data acquisition module: Used to collect data from the power distribution network and perform preprocessing; Impedance estimation module: Based on distribution network data and topology, this module estimates the impedance parameters of distribution network lines using the steepest descent method, with the goal of optimizing a combination of power and voltage estimation. Anomaly Diagnosis Module: Used to calculate theoretical line loss based on parameter estimation results, and to diagnose the cause of anomalies in samples with abnormal line loss using a pre-trained convolutional neural network.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-7.
10. A computer-storable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.