A power distribution network feeder line loss anomaly evaluation method and device

Abstract

The application discloses a power distribution network feeder line loss abnormality evaluation method and device, constructs a power distribution network line loss characteristic index system containing a distributed power supply, uses collected power distribution network line loss data and line loss characteristic index data of a line area for training an LSTM neural network model, obtains an optimal power distribution network line loss prediction model, so that line loss prediction based on the optimal power distribution network line loss prediction model on to-be-predicted line loss characteristic index data is performed, first power distribution network line loss prediction data is obtained and based on the first power distribution network line loss prediction data, a line loss reasonable confidence interval of the power distribution network line loss is generated; based on the line loss reasonable confidence interval and an improved fuzzy C-means algorithm, the first power distribution network line loss prediction data is clustered, a suspected abnormal feeder line is obtained, and the abnormality degree of the power distribution network line loss is evaluated by calculating an abnormality coefficient of the suspected abnormal feeder line. Compared with the prior art, the technical scheme of the application can improve the accuracy of the power distribution network line loss abnormality evaluation by creating an abnormality coefficient to evaluate the line loss abnormality degree.

Description

Technical Field

[0001] This invention relates to the technical field of distribution network line loss prediction, and in particular to a method and apparatus for assessing abnormal line losses in distribution network feeders. Background Technology

[0002] With the introduction of the national "dual-carbon" target, implementing renewable energy substitution actions and building a new power system based on new energy sources are important national strategic deployments. The large-scale integration of distributed power sources such as wind and solar power into the distribution network has significantly impacted grid operating parameters. On the one hand, the integrated distributed power sources affect power flow in the distribution network, directly impacting line losses; on the other hand, the instability of their output power brings a series of problems such as harmonics and voltage fluctuations, affecting the power supply voltage quality of the distribution network. The disturbances brought by distributed power sources interact with existing disturbance factors in the distribution network, making the power quality disturbance mechanism more complex and increasing the difficulty of distribution network loss analysis. Therefore, analyzing the impact of distributed power source integration on distribution network losses in light of the actual development of the distribution network, and accurately diagnosing abnormal line losses, is of great significance for effectively carrying out energy conservation and loss reduction work and improving the economic benefits of power companies.

[0003] For the diagnosis of line loss anomalies in distribution networks, current research includes line loss analysis models based on smart meters, which use clustering algorithms to identify line loss anomalies; and line loss anomaly identification based on k-means clustering algorithms, which can determine whether line loss anomalies exist in a distribution area. Therefore, using data mining techniques for line loss anomaly analysis is of great research significance. Although there are relatively comprehensive studies on the analysis of line loss influencing factors, these studies have not fully considered the impact of distributed generation on line loss characteristics in the context of new power systems. In particular, research on the spatiotemporal distribution characteristics of power systems is still lacking, making it difficult to effectively conduct spatiotemporal analysis of line losses and adapt to the flexible and variable situation of distributed generation in actual distribution networks. Furthermore, current research is mainly limited to theoretical calculations of line losses in the context of traditional power systems, and does not adequately consider the proportion and type of new energy sources, making it difficult to form a reasonable range of line losses with reference value.

[0004] Most existing line loss anomaly diagnoses, both domestically and internationally, analyze anomaly data from a data mining perspective, which improves the efficiency of judgment to some extent. However, the incompleteness of characteristic indicators may affect the accuracy of anomaly line loss diagnosis. Furthermore, many diagnostic schemes only consider traditional distribution networks, neglecting the impact of dynamic changes in distributed generation on line loss characteristics and lacking secondary analysis of high-loss lines. This makes them difficult to apply to new power systems dominated by renewable energy sources, and ultimately hinders accurate diagnosis of abnormal line losses in distribution networks. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and device for assessing abnormal line losses in distribution network feeders, which can improve the accuracy of assessing abnormal line losses in distribution networks.

[0006] To address the aforementioned technical problems, this invention provides a method for assessing abnormal line losses in distribution network feeders, comprising:

[0007] Based on the line loss characteristic index data, a line loss characteristic index system for distribution networks including distributed power sources is constructed. The line loss characteristic index data includes weather factors, equipment grid connection parameters, and distribution network operation data.

[0008] Collect all distribution network line loss data for the selected line area within a preset time period, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data for the line area within the preset time period. Perform data processing on all distribution network line loss data and all line loss characteristic index data to obtain standard distribution network line loss data and standard line loss characteristic index data. Divide the standard distribution network line loss data and the standard line loss characteristic index data into training dataset and test dataset.

[0009] An LSTM neural network model is constructed based on the LSTM algorithm. The LSTM neural network model is trained according to the training dataset and the test dataset to determine the optimal distribution network line loss prediction model.

[0010] Obtain the line loss characteristic index data of different distributed power source types, input the line loss characteristic index data of different distributed power source types into the optimal distribution network line loss prediction model, output and calculate the reasonable confidence interval of the line loss of the distribution network for different distributed power source types based on the first distribution network line loss prediction data corresponding to different distributed power source types.

[0011] The improved fuzzy C-means algorithm is used to perform a first clustering of the predicted line loss data of the first distribution network to obtain a first clustering result. Based on the reasonable confidence interval of the line loss, the first clustering result is then subjected to a second clustering to obtain suspected abnormal feeders. The abnormality coefficient of the suspected abnormal feeders is calculated. Based on the abnormality coefficient, the degree of abnormality of the distribution network line loss of the suspected abnormal feeders is evaluated and obtained.

[0012] In one possible implementation, all distribution network line loss data for the selected line area within a preset time period are collected, and based on the distribution network line loss characteristic index system, all line loss characteristic index data for the line area within the preset time period are collected, specifically including:

[0013] Obtain the distributed power penetration rate of each line area in the distribution network, select the line areas where the distributed power penetration rate is greater than the preset distributed power penetration threshold, and collect the distribution network line loss data of the line areas within one year.

[0014] Based on the aforementioned power distribution network line loss characteristic index system, line loss characteristic index data for the aforementioned line area are collected.

[0015] In one possible implementation, all distribution network line loss data and all line loss characteristic index data are processed separately to obtain standard distribution network line loss data and standardized line loss characteristic index data, specifically including:

[0016] Negative values ​​of distribution network line loss data are removed from all distribution network line loss data. At the same time, distribution network line loss data with a line loss rate greater than 25% are also removed from all distribution network line loss data to obtain the first distribution network line loss data.

[0017] Determine whether there are missing data values ​​in all the first distribution network line loss data or all the line loss characteristic index data. If there are missing data values, determine whether all the first distribution network line loss data or all the line loss characteristic index data conform to a uniform distribution. If yes, obtain the mean of all the first distribution network line loss data or all the line loss characteristic index data and fill the missing data values ​​with the mean. If no, obtain the median of all the first distribution network line loss data or all the line loss characteristic index data and fill the missing data values ​​with the median to obtain standard distribution network line loss data and first line loss characteristic index data.

[0018] The first line loss characteristic index data is normalized to obtain standard line loss characteristic index data.

[0019] In one possible implementation, an LSTM neural network model is constructed based on the LSTM algorithm. The LSTM neural network model is then trained using the training dataset and the test dataset to determine the optimal distribution network line loss prediction model. Specifically, this includes:

[0020] An LSTM neural network model is constructed based on the LSTM algorithm. The training dataset is input into the LSTM neural network model for model training to obtain an initial distribution network line loss prediction model. The test dataset is then input into the initial distribution network line loss prediction model to obtain the distribution network line loss prediction data corresponding to the test dataset.

[0021] The predicted line loss data of the distribution network is compared with the predicted line loss data of the standard distribution network to determine the optimal line loss prediction model of the distribution network.

[0022] In one possible implementation, the line loss characteristic index data includes weather factors, equipment grid connection parameters, and distribution network operation data. The weather factors include average temperature, average wind speed, sunshine duration, sunshine intensity, and rainfall. The equipment grid connection parameters include grid connection capacity, grid connection location, grid connection operation mode, and power supply. The distribution network operation data includes data date, power supply type, power supply radius, line load rate, and power supply quantity.

[0023] In one possible implementation, predicting line loss characteristic index data for different types of distributed power sources is obtained, and the predicting line loss characteristic index data is input into the optimal distribution network line loss prediction model, specifically including:

[0024] Acquire the first predictable line loss characteristic index data for distributed power generation type wind power generation, wherein the weather factors in the first predictable line loss characteristic index data include average temperature and average wind speed.

[0025] Acquire the second predictable line loss characteristic index data for distributed power source type photovoltaic power generation, wherein the weather factors in the second predictable line loss characteristic index data include sunshine duration and sunshine intensity;

[0026] Obtain the third predictable line loss characteristic index data for distributed power generation of hydropower type, wherein the weather factors in the third predictable line loss characteristic index data include average temperature and rainfall.

[0027] The first line loss characteristic index data to be predicted, the second line loss characteristic index data to be predicted, and the third line loss characteristic index data to be predicted are input into the optimal distribution network line loss prediction model.

[0028] In one possible implementation, an anomaly coefficient is calculated for the suspected abnormal feeder, and the degree of distribution network line loss anomaly of the suspected abnormal feeder is assessed and obtained based on the anomaly coefficient, specifically including:

[0029] Using the cluster center of the category where the suspected abnormal feeder loss is located as the baseline feeder, obtain the suspected abnormal feeder loss prediction data and suspected abnormal feeder loss characteristic index data of the suspected abnormal feeder, and at the same time obtain the baseline feeder loss prediction data and baseline feeder loss characteristic index data of the baseline feeder.

[0030] Calculate the difference between the predicted data of the suspected abnormal feeder line loss corresponding to the first preset time and the predicted data of the baseline feeder line loss, and calculate the average difference of all differences;

[0031] Calculate the deviation between the suspected abnormal feeder line loss characteristic index data and the benchmark feeder line loss characteristic index data, and calculate the average deviation value corresponding to each line loss characteristic index based on the number of line loss characteristic indices.

[0032] The abnormality coefficient of the suspected abnormal feeder is calculated and obtained based on the average difference and the average deviation.

[0033] The abnormality coefficient is compared with a preset abnormality coefficient threshold. If the abnormality coefficient is greater than or equal to the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a high degree of abnormality. If the abnormality coefficient is less than the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a low degree of abnormality.

[0034] The present invention also provides a distribution network feeder line loss anomaly assessment device, comprising: a distribution network line loss characteristic index system construction module, a distribution network data processing module, a distribution network line loss prediction model construction module, a line loss reasonable confidence interval calculation module, and a distribution network line loss anomaly degree assessment module;

[0035] The distribution network line loss characteristic index system construction module is used to acquire and construct a distribution network line loss characteristic index system including distributed power sources based on the line loss characteristic index data. The line loss characteristic index data includes weather factors, equipment grid connection parameters and distribution network operation data.

[0036] The distribution network data processing module is used to collect all distribution network line loss data of the selected line area within a preset time, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data of the line area within the preset time, process all distribution network line loss data and all line loss characteristic index data respectively to obtain standard distribution network line loss data and standard line loss characteristic index data, and divide the standard distribution network line loss data and the standard line loss characteristic index data into training dataset and test dataset;

[0037] The distribution network line loss prediction model construction module is used to construct an LSTM neural network model based on the LSTM algorithm, train the LSTM neural network model according to the training dataset and the test dataset, and determine the optimal distribution network line loss prediction model.

[0038] The reasonable confidence interval calculation module for line loss is used to obtain the line loss characteristic index data to be predicted for different types of distributed power sources, input the line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model, and output and calculate the reasonable confidence interval for line loss of distribution network for different types of distributed power sources based on the first distribution network line loss prediction data corresponding to different types of distributed power sources.

[0039] The distribution network line loss anomaly assessment module is used to perform a first clustering of the first distribution network line loss prediction data based on the improved fuzzy C-means algorithm to obtain a first clustering result, and then perform a second clustering of the first clustering result based on the reasonable confidence interval of the line loss to obtain suspected abnormal feeders. The module calculates the anomaly coefficient of the suspected abnormal feeder and assesses and obtains the distribution network line loss anomaly degree of the suspected abnormal feeder based on the anomaly coefficient.

[0040] In one possible implementation, the distribution network data processing module is used to collect all distribution network line loss data for a selected line area within a preset time period, and, based on the distribution network line loss characteristic index system, collect all line loss characteristic index data for the line area within the preset time period, specifically including:

[0041] Obtain the distributed power penetration rate of each line area in the distribution network, select the line areas where the distributed power penetration rate is greater than the preset distributed power penetration threshold, and collect the distribution network line loss data of the line areas within one year.

[0042] Based on the aforementioned power distribution network line loss characteristic index system, line loss characteristic index data for the aforementioned line area are collected.

[0043] In one possible implementation, the distribution network data processing module is used to process all distribution network line loss data and all line loss characteristic index data respectively to obtain standard distribution network line loss data and standard line loss characteristic index data, specifically including:

[0044] Negative values ​​of distribution network line loss data are removed from all distribution network line loss data. At the same time, distribution network line loss data with a line loss rate greater than 25% are also removed from all distribution network line loss data to obtain the first distribution network line loss data.

[0045] Determine whether there are missing data values ​​in all the first distribution network line loss data or all the line loss characteristic index data. If there are missing data values, determine whether all the first distribution network line loss data or all the line loss characteristic index data conform to a uniform distribution. If yes, obtain the mean of all the first distribution network line loss data or all the line loss characteristic index data and fill the missing data values ​​with the mean. If no, obtain the median of all the first distribution network line loss data or all the line loss characteristic index data and fill the missing data values ​​with the median to obtain standard distribution network line loss data and first line loss characteristic index data.

[0046] The first line loss characteristic index data is normalized to obtain standard line loss characteristic index data.

[0047] In one possible implementation, the distribution network line loss prediction model construction module is used to construct an LSTM neural network model based on the LSTM algorithm, train the LSTM neural network model according to the training dataset and the test dataset, and determine the optimal distribution network line loss prediction model, specifically including:

[0048] An LSTM neural network model is constructed based on the LSTM algorithm. The training dataset is input into the LSTM neural network model for model training to obtain an initial distribution network line loss prediction model. The test dataset is then input into the initial distribution network line loss prediction model to obtain the distribution network line loss prediction data corresponding to the test dataset.

[0049] The predicted line loss data of the distribution network is compared with the predicted line loss data of the standard distribution network to determine the optimal line loss prediction model of the distribution network.

[0050] In one possible implementation, the line loss characteristic index data in the distribution network line loss characteristic index system construction module includes weather factors, equipment grid connection parameters, and distribution network operation data. The weather factors include average temperature, average wind speed, sunshine duration, sunshine intensity, and rainfall. The equipment grid connection parameters include grid connection capacity, grid connection location, grid connection operation mode, and grid-connected power consumption. The distribution network operation data includes data date, power supply type, power supply radius, line load rate, and power supply quantity.

[0051] In one possible implementation, the line loss reasonable confidence interval calculation module is used to acquire the line loss characteristic index data to be predicted for different distributed power generation types, and input the line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model, specifically including:

[0052] Acquire the first predictable line loss characteristic index data for distributed power generation type wind power generation, wherein the weather factors in the first predictable line loss characteristic index data include average temperature and average wind speed.

[0053] Acquire the second predictable line loss characteristic index data for distributed power source type photovoltaic power generation, wherein the weather factors in the second predictable line loss characteristic index data include sunshine duration and sunshine intensity;

[0054] Obtain the third predictable line loss characteristic index data for distributed power generation of hydropower type, wherein the weather factors in the third predictable line loss characteristic index data include average temperature and rainfall.

[0055] The first line loss characteristic index data to be predicted, the second line loss characteristic index data to be predicted, and the third line loss characteristic index data to be predicted are input into the optimal distribution network line loss prediction model.

[0056] In one possible implementation, the distribution network line loss anomaly assessment module is used to calculate the anomaly coefficient of the suspected abnormal feeder, and to assess and obtain the distribution network line loss anomaly degree of the suspected abnormal feeder based on the anomaly coefficient, specifically including:

[0057] Using the cluster center of the category where the suspected abnormal feeder loss is located as the baseline feeder, obtain the suspected abnormal feeder loss prediction data and suspected abnormal feeder loss characteristic index data of the suspected abnormal feeder, and at the same time obtain the baseline feeder loss prediction data and baseline feeder loss characteristic index data of the baseline feeder.

[0058] Calculate the difference between the predicted data of the suspected abnormal feeder line loss corresponding to the first preset time and the predicted data of the baseline feeder line loss, and calculate the average difference of all differences;

[0059] Calculate the deviation between the suspected abnormal feeder line loss characteristic index data and the benchmark feeder line loss characteristic index data, and calculate the average deviation value corresponding to each line loss characteristic index based on the number of line loss characteristic indices.

[0060] The abnormality coefficient of the suspected abnormal feeder is calculated and obtained based on the average difference and the average deviation.

[0061] The abnormality coefficient is compared with a preset abnormality coefficient threshold. If the abnormality coefficient is greater than or equal to the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a high degree of abnormality. If the abnormality coefficient is less than the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a low degree of abnormality.

[0062] The present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method for assessing abnormal line losses of distribution network feeders as described in any of the preceding claims.

[0063] The present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the line loss anomaly assessment method for distribution network feeders as described in any of the preceding claims.

[0064] The present invention provides a method and apparatus for assessing abnormal line losses in distribution network feeders, which has the following advantages compared with the prior art:

[0065] By considering line loss characteristic indicators such as weather factors, equipment grid connection parameters, and distribution network operation data, a line loss characteristic indicator system for distribution networks including distributed generation is constructed to analyze the impact of distributed generation on line loss characteristics. The collected distribution network line loss data and line loss characteristic indicator data for the line area are used to train an LSTM neural network model to obtain the optimal distribution network line loss prediction model. This model is then used to predict line losses based on the line loss characteristic indicator data to be predicted. Based on the first distribution network line loss prediction data, a reasonable confidence interval for distribution network line loss is generated. Using the reasonable confidence interval and an improved fuzzy C-means algorithm, the first distribution network line loss prediction data is clustered to identify suspected abnormal feeders. By calculating the abnormality coefficient of the suspected abnormal feeders, the degree of abnormality in distribution network line loss is assessed. Compared with existing technologies, the technical solution of this invention breaks away from the traditional method of using a line loss rate exceeding a certain threshold as the condition for assessing line loss anomalies. It utilizes a Long Short-Term Memory (LSTM) network and an improved fuzzy C-means clustering algorithm to create anomaly coefficients to assess the degree of line loss anomalies, which can improve the accuracy of line loss anomaly assessment in distribution networks. At the same time, it solves the one-sidedness of existing distribution network line loss anomaly diagnosis that does not consider the dynamic changes of distributed power sources. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating an embodiment of a method for assessing abnormal line losses in a power distribution network feeder provided by the present invention.

[0067] Figure 2 This is a schematic diagram of an embodiment of a line loss anomaly assessment device for a power distribution network feeder provided by the present invention. Detailed Implementation

[0068] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] Example 1

[0070] See Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of a method for assessing abnormal line losses in distribution network feeders provided by the present invention. Figure 1 As shown, the method includes steps 101-105, as detailed below:

[0071] Step 101: Obtain and construct a distribution network line loss characteristic index system with distributed power sources based on the line loss characteristic index data, wherein the line loss characteristic index data includes weather factors, equipment grid connection parameters and distribution network operation data.

[0072] In one embodiment, the distributed power source types include wind power generation, photovoltaic power generation, and hydropower generation, and the obtained line loss characteristic index data are the line loss characteristic index data corresponding to different distributed power source types.

[0073] In one embodiment, the line loss characteristic index data includes weather factors, equipment grid connection parameters, and distribution network operation data. The weather factors include average temperature, average wind speed, sunshine duration, sunshine intensity, and rainfall. The equipment grid connection parameters include grid connection capacity, grid connection location, grid connection operation mode, and power supply. The distribution network operation data includes data date, power supply type, power supply radius, line load rate, and power supply.

[0074] Specifically, regarding the selection of weather factors, since the main influencing factor for wind power generation is wind energy density, which depends on air pressure, temperature, and wind speed; photovoltaic power generation is mainly determined by the total solar radiation, the total area of ​​photovoltaic cells, and the photoelectric conversion efficiency. Given a fixed conversion efficiency of the solar cell modules, the power generation of the photovoltaic system is determined by the solar radiation intensity; hydropower generation mainly utilizes the changes in water levels upstream and downstream. Continuous high temperatures or rainfall will have a certain impact on water level changes, thus affecting hydropower generation. Therefore, through correlation analysis, average temperature, average wind speed, sunshine duration, sunshine intensity, and rainfall were determined as weather factors in the characteristic indicators.

[0075] Specifically, regarding the selection of grid connection parameters for equipment, since grid connection of distributed power sources will change the network structure, operation mode and power flow direction of the distribution network, the main indicators are determined by analyzing the correlation of grid connection parameters of distributed power sources, namely grid connection capacity, grid connection location, grid connection operation mode and on-grid power.

[0076] Specifically, for the selection of distribution network operation data, based on big data analysis of the main influencing factors of distribution network line loss, the data date, power supply type, power supply radius, line load rate, and power supply amount are determined as distribution network operation data.

[0077] In this embodiment, due to the uncertainty and randomness of distributed power generation output, the changes in distribution network line losses are affected by the access of distributed power sources. The trend of line loss changes is closely related to the characteristics of distributed power sources. Moreover, the current line loss characteristic indicators of distribution networks containing distributed power sources only consider a few factors and are not comprehensive. Therefore, in this embodiment, a line loss characteristic indicator system for distribution networks containing distributed power sources is established from three dimensions: weather factors, equipment grid connection parameters, and distribution network operation data. This solves the problem of incomplete characteristic indicators for distribution networks containing distributed power sources.

[0078] Step 102: Collect all distribution network line loss data for the selected line area within a preset time period, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data for the line area within the preset time period. Process all distribution network line loss data and all line loss characteristic index data to obtain standard distribution network line loss data and standard line loss characteristic index data, and divide the standard distribution network line loss data and standard line loss characteristic index data into training dataset and test dataset.

[0079] In one embodiment, the distributed generation penetration rate of each line area in the distribution network is obtained. Line areas with a distributed generation penetration rate greater than a preset distributed generation penetration rate threshold are selected, and the corresponding daily distribution network line loss data for each line area over a year is collected. The distributed generation penetration rate is the ratio of distributed generation to power supply of the selected line. Preferably, the preset distributed generation penetration rate threshold is 50%.

[0080] In one embodiment, based on the distribution network line loss characteristic index system, the line loss characteristic index data corresponding to each day of the line area within the year is collected. The weather factors in the line loss characteristic index data can be obtained from the China Meteorological Administration website, and the equipment grid connection parameters and distribution network operation parameters in the line loss characteristic index data can be obtained from the power grid management platform.

[0081] In one embodiment, after obtaining all the distribution network line loss data and all the line loss characteristic index data, data processing is performed on all the distribution network line loss data and all the line loss characteristic index data.

[0082] Specifically, negative values ​​of distribution network line loss data are removed from all distribution network line loss data, and distribution network line loss data with a line loss rate greater than 25% are also removed from all distribution network line loss data to obtain the first distribution network line loss data.

[0083] Specifically, it is further determined whether there are missing data values ​​in all the first distribution network line loss data or all the line loss characteristic index data. If there are missing data values, it is determined whether all the first distribution network line loss data or all the line loss characteristic index data conform to a uniform distribution. If so, the mean of all the first distribution network line loss data or all the line loss characteristic index data is obtained, and the mean is used to fill the missing data values. If not, the median of all the first distribution network line loss data or all the line loss characteristic index data is obtained, and the median is used to fill the missing data values, thus obtaining standard distribution network line loss data and first line loss characteristic index data.

[0084] Specifically, since the dimensions of the various feature indicators in the line loss feature index data are different, it is necessary to normalize the first line loss feature index data, mapping the data to [0, 1] to obtain the standard loss feature index data. The normalization formula is as follows:

[0085]

[0086] In the formula, x max x represents the maximum value of a single-class feature index. min denoted as the minimum value of a single-class feature index data, x is the original single feature index data to be normalized, and x′ is the single-class feature index data after normalization, i.e., a single standard loss feature index data.

[0087] In one embodiment, the standard distribution network line loss data and the standard line loss characteristic index data are divided into a training dataset and a test dataset. Specifically, by setting a division ratio, the standard distribution network line loss data and the standard line loss characteristic index data are divided into a training dataset and a test dataset according to the division ratio; preferably, the division ratio is 8:2; and one set of standard distribution network line loss data and standard line loss characteristic index data in the training dataset and the test dataset are data acquired on the same day.

[0088] Step 103: Construct an LSTM neural network model based on the LSTM algorithm, train the LSTM neural network model according to the training dataset and the test dataset, and determine the optimal distribution network line loss prediction model.

[0089] In one embodiment, an LSTM neural network model is constructed based on the LSTM algorithm. Specifically, the LSTM neural network model is a Long Short-Term Memory neural network model, which includes one input layer, one hidden layer, and one output layer. The number of LSTM neurons in the hidden layer is minimized by using ReLU as the activation function. The Adam algorithm is selected to solve the optimization problem of distribution network line loss data and parameters.

[0090] In one embodiment, weather factor influencing indicators are determined based on the type of distributed power source, namely wind power, photovoltaic power, and hydropower. The weather factors in the input feature indicator system differ for different distributed power source types. When the distributed power source type is wind power, the input is average temperature and average wind speed; when the distributed power source type is photovoltaic power, the input is sunshine duration and sunshine intensity; and when the distributed power source type is hydropower, the input is average temperature and rainfall. For all three types of distributed power source types, the input is the grid connection parameters of the equipment and the characteristic indicators of the distribution network operation data, which are used as inputs to the neural network model.

[0091] In one embodiment, the training dataset is input into the LSTM neural network model for model training to obtain an initial distribution network line loss prediction model, wherein the initial distribution network line loss prediction model is an initial distribution network line loss prediction model including distributed generation; and the test dataset is input into the initial distribution network line loss prediction model to obtain the distribution network line loss prediction data corresponding to the test dataset.

[0092] In one embodiment, the predicted distribution network line loss data is compared with the standard distribution network line loss data to determine the optimal distribution network line loss prediction model. Specifically, the predicted distribution network line loss data corresponding to the test dataset is compared with the standard distribution network line loss data in the test set, that is, the predicted value and the actual value are compared and analyzed to evaluate the prediction result. The mean square error and mean absolute error of the predicted distribution network line loss data corresponding to the test dataset and the standard distribution network line loss data in the test set are calculated. The optimal distribution network line loss prediction model is determined when the mean square error and mean absolute error are minimized. The calculation formulas for the mean absolute error (MAPE) and mean square error (MSE) are as follows:

[0093]

[0094]

[0095] ′

[0096] In the formula, y i It is the i-th distribution network line loss prediction data in the test dataset; y iis the line loss data of the i-th standard distribution network in the test set; n is the number of samples in the test set, i.e., the number of distribution network lines.

[0097] Step 104: Obtain the line loss characteristic index data to be predicted for different distributed power source types, input the line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model, output and calculate the reasonable confidence interval of the distribution network line loss for different distributed power source types based on the first distribution network line loss prediction data corresponding to different distributed power source types.

[0098] In one embodiment, predictable line loss characteristic index data for different types of distributed power sources are acquired. Specifically, first predictable line loss characteristic index data for wind power generation is acquired, wherein the weather factors in the first predictable line loss characteristic index data include average temperature and average wind speed; second predictable line loss characteristic index data for photovoltaic power generation is acquired, wherein the weather factors in the second predictable line loss characteristic index data include sunshine duration and sunshine intensity; and third predictable line loss characteristic index data for hydropower generation is acquired, wherein the weather factors in the third predictable line loss characteristic index data include average temperature and rainfall.

[0099] In one embodiment, the first line loss characteristic index data to be predicted, the second line loss characteristic index data to be predicted, and the third line loss characteristic index data to be predicted are input into the optimal distribution network line loss prediction model, so that the optimal distribution network line loss prediction model outputs the first distribution network line loss prediction data corresponding to different distributed power generation types.

[0100] In one embodiment, based on the first distribution network line loss prediction data corresponding to different distributed power source types, the reasonable confidence interval of the distribution network line loss for different distributed power source types is calculated.

[0101] Specifically, based on the predicted line loss data of the first distribution network corresponding to different distributed power source types, confidence intervals for the predicted line loss data of the first distribution network corresponding to different distributed power source types at confidence levels of 90%, 95%, and 99% are calculated respectively. These confidence intervals are defined by the upper and lower confidence limits of the statistics, respectively, as shown below:

[0102]

[0103] In the formula, σ is the average value of the predicted line loss data for the first distribution network; the C-value is 1.64 at 90% confidence level, 1.96 at 95% confidence level, and 2.58 at 99% confidence level; σ is the standard deviation of the population data; n is the number of predicted line loss data for the first distribution network.

[0104] In one embodiment, a confidence interval with a 95% confidence level is selected as the reasonable confidence interval for the line loss of the distribution network.

[0105] Step 105: Perform a first clustering on the first distribution network line loss prediction data based on the improved fuzzy C-means algorithm to obtain the first clustering result. Based on the reasonable confidence interval of the line loss, perform a second clustering on the first clustering result to obtain suspected abnormal feeders. Calculate the abnormality coefficient of the suspected abnormal feeder. Based on the abnormality coefficient, evaluate and obtain the degree of distribution network line loss abnormality of the suspected abnormal feeder.

[0106] In one embodiment, the fuzzy coefficients and iteration stopping threshold are set in the improved fuzzy C-means clustering algorithm. The first distribution network line loss prediction data obtained in step 104, and the equipment grid connection parameters and distribution network operation data in the corresponding line loss characteristic index data to be predicted, are used as inputs to the improved fuzzy C-means clustering algorithm, so that the improved fuzzy C-means clustering algorithm can cluster the first distribution network line loss prediction data. Preferably, the fuzzy coefficient m is set to 2, and the iteration stopping threshold is set to 0.001.

[0107] In one embodiment, the clustering process is the process of minimizing the objective function. Through repeated iterative calculations, the error value of the objective function is gradually reduced. When the objective function converges, the final clustering result can be obtained. The clustering process is as follows:

[0108] make Given N samples, S sample space dimension, and C number of clusters, the objective function is:

[0109]

[0110] Make:

[0111]

[0112]

[0113] u ij ≥0, 1≤i≤N, 1≤j≤C;

[0114] In the formula, U is the membership matrix; V is a matrix consisting of C cluster centers; u ij Let x be the membership degree of the i-th sample to the j-th class; membership degree represents the degree to which an object x belongs to set Q, and the closer it is to 1, the higher the membership degree. i For the i-th sample; v j Let m be the j-th cluster center; m is the fuzzy coefficient; ||x i -v j || represents the sample point x iTo the cluster center v j Euclidean distance.

[0115] In one embodiment, the optimal number of clusters is selected using an effectiveness function, specifically the Xie-Beni clustering effectiveness function.

[0116] The XB metric can be used to find a balance between intra-class compactness and inter-class separation. The formula is:

[0117]

[0118] In the formula: x i For the i-th sample; v j v is the j-th cluster center; i Let i be the i-th cluster center.

[0119] Among them, intra-class compactness = numerator of formula / number of samples N, the smaller the better; inter-class separation = denominator of formula * number of samples N, the larger the better; the smaller the XB index calculation result, the better the clustering effect.

[0120] In one embodiment, based on the reasonable confidence interval of the line loss, the results of the first clustering are subjected to a second clustering to obtain suspected abnormal feeders.

[0121] Specifically, based on the endpoints of the reasonable confidence interval for line loss, the cluster centers and their respective classes in the first clustering are analyzed; the classes containing cluster centers that are outside the reasonable confidence interval for line loss and less than 0.5 of the endpoints of the reasonable confidence interval for line loss are identified, and secondary clustering is performed on these classes. Through secondary clustering, data containing cluster centers within the reasonable confidence interval for line loss are filtered out, and the classes containing cluster centers outside the reasonable confidence interval for line loss are identified as suspected abnormal feeders.

[0122] In one embodiment, in order to quantitatively analyze the degree of abnormality of feeder line loss, based on the spatiotemporal distribution characteristics of line loss, the cluster center of the category of the suspected abnormal feeder is used as the reference feeder, and considering both time and spatial discrete aspects, the abnormality coefficient of the suspected abnormal feeder is calculated.

[0123] Specifically, taking the cluster center of the category where the suspected abnormal feeder loss is located as the baseline feeder, the suspected abnormal feeder loss prediction data and suspected abnormal feeder loss characteristic index data are obtained, and the baseline feeder loss prediction data and baseline feeder loss characteristic index data are also obtained.

[0124] Specifically, the difference between the predicted line loss data of the suspected abnormal feeder corresponding to the first preset time and the predicted line loss data of the benchmark feeder is calculated, and the average difference of all differences is calculated; the deviation between the characteristic index data of the suspected abnormal feeder and the characteristic index data of the benchmark feeder is calculated, and based on the number of line loss characteristic indicators, the average deviation corresponding to each line loss characteristic indicator is calculated; wherein, the calculation process of the average difference A and the average deviation B is as follows:

[0125]

[0126]

[0127] In the formula, T represents the number of days for the first preset time; X i X represents the predicted line loss data for the i-th suspected abnormal feeder; c Z represents the baseline feeder line loss; S represents the number of line loss characteristic indicators; Z represents the baseline feeder line loss. i Z represents the characteristic index data of the i-th suspected abnormal feeder line loss; c The characteristic index data for the reference feeder.

[0128] Preferably, the first preset time is the same time period of each day within a month, that is, the difference between the predicted data of the suspected abnormal feeder line loss and the predicted data of the baseline feeder line loss corresponding to the same time period of each day within the preset month is calculated.

[0129] Specifically, based on the average difference and the average deviation, the average difference and the average deviation are added together to obtain the abnormality coefficient F of the suspected abnormal feeder, where F = A + B.

[0130] In one embodiment, the abnormality coefficient is compared with a preset abnormality coefficient threshold. If the abnormality coefficient is greater than or equal to the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a high degree of abnormality, that is, the feeder is likely to have abnormal line loss in a certain period of time. If the abnormality coefficient is less than the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a low degree of abnormality, that is, the feeder is unlikely to have abnormal line loss in a certain period of time.

[0131] Example 2

[0132] See Figure 2 , Figure 2 This is a schematic diagram of an embodiment of a line loss anomaly assessment device for distribution network feeders provided by the present invention, as shown below. Figure 2As shown, the device includes a distribution network line loss characteristic index system construction module 201, a distribution network data processing module 202, a distribution network line loss prediction model construction module 203, a line loss reasonable confidence interval calculation module 204, and a distribution network line loss anomaly assessment module 205, as detailed below:

[0133] The distribution network line loss characteristic index system construction module 201 is used to acquire and construct a distribution network line loss characteristic index system containing distributed power sources based on the line loss characteristic index data. The line loss characteristic index data includes weather factors, equipment grid connection parameters, and distribution network operation data.

[0134] The distribution network data processing module 202 is used to collect all distribution network line loss data of the selected line area within a preset time, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data of the line area within the preset time, process all distribution network line loss data and all line loss characteristic index data respectively to obtain standard distribution network line loss data and standard line loss characteristic index data, and divide the standard distribution network line loss data and the standard line loss characteristic index data into training dataset and test dataset.

[0135] The distribution network line loss prediction model construction module 203 is used to construct an LSTM neural network model based on the LSTM algorithm, train the LSTM neural network model according to the training dataset and the test dataset, and determine the optimal distribution network line loss prediction model.

[0136] The line loss reasonable confidence interval calculation module 204 is used to obtain the line loss characteristic index data to be predicted for different distributed power source types, input the line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model, and output and calculate the line loss reasonable confidence interval of the distribution network line loss for different distributed power source types based on the first distribution network line loss prediction data corresponding to different distributed power source types.

[0137] The distribution network line loss anomaly assessment module 205 is used to perform a first clustering of the first distribution network line loss prediction data based on the improved fuzzy C-means algorithm to obtain a first clustering result, and perform a second clustering of the first clustering result based on the reasonable confidence interval of the line loss to obtain suspected abnormal feeders, calculate the anomaly coefficient of the suspected abnormal feeders, and assess and obtain the distribution network line loss anomaly degree of the suspected abnormal feeders based on the anomaly coefficient.

[0138] In one embodiment, the distribution network data processing module 202 is used to collect all distribution network line loss data of the selected line area within a preset time period, and collect all line loss characteristic index data of the line area within the preset time period based on the distribution network line loss characteristic index system. Specifically, this includes: obtaining the distributed power penetration rate of each line area in the distribution network, selecting line areas where the distributed power penetration rate is greater than a preset distributed power penetration rate threshold, collecting distribution network line loss data of the line area within one year; and collecting line loss characteristic index data of the line area based on the distribution network line loss characteristic index system.

[0139] In one embodiment, the distribution network data processing module 202 is used to process all distribution network line loss data and all line loss characteristic index data respectively to obtain standard distribution network line loss data and standard line loss characteristic index data. Specifically, this includes: removing negative values ​​from all distribution network line loss data, and simultaneously removing distribution network line loss data with a line loss rate greater than 25% from all distribution network line loss data to obtain first distribution network line loss data; determining whether there are missing values ​​in all first distribution network line loss data or all line loss characteristic index data; if there are missing values... Then, determine whether all the first distribution network line loss data or all the line loss characteristic index data conform to a uniform distribution. If yes, obtain the mean of all the first distribution network line loss data or all the line loss characteristic index data, and fill the missing data values ​​with the mean. If not, obtain the median of all the first distribution network line loss data or all the line loss characteristic index data, and fill the missing data values ​​with the median, to obtain standard distribution network line loss data and first line loss characteristic index data. Normalize the first line loss characteristic index data to obtain standard line loss characteristic index data.

[0140] In one embodiment, the distribution network line loss prediction model construction module 203 is used to construct an LSTM neural network model based on the LSTM algorithm, train the LSTM neural network model according to the training dataset and the test dataset, and determine the optimal distribution network line loss prediction model. Specifically, it includes: constructing an LSTM neural network model based on the LSTM algorithm; inputting the training dataset into the LSTM neural network model for model training to obtain an initial distribution network line loss prediction model; inputting the test dataset into the initial distribution network line loss prediction model to obtain distribution network line loss prediction data corresponding to the test dataset; and comparing the distribution network line loss prediction data with the standard distribution network line loss data to determine the optimal distribution network line loss prediction model.

[0141] In one embodiment, the line loss characteristic index data in the distribution network line loss characteristic index system construction module 201 includes weather factors, equipment grid connection parameters, and distribution network operation data. The weather factors include average temperature, average wind speed, sunshine duration, sunshine intensity, and rainfall. The equipment grid connection parameters include grid connection capacity, grid connection location, grid connection operation mode, and grid-connected power consumption. The distribution network operation data includes data date, power supply type, power supply radius, line load rate, and power supply quantity.

[0142] In one embodiment, the line loss reasonable confidence interval calculation module 204 is used to acquire line loss characteristic index data to be predicted for different distributed power generation types, and input the line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model. Specifically, this includes: acquiring first line loss characteristic index data to be predicted for distributed power generation of wind power generation, wherein the weather factors in the first line loss characteristic index data to be predicted include average temperature and average wind speed; acquiring second line loss characteristic index data to be predicted for distributed power generation of photovoltaic power generation, wherein the weather factors in the second line loss characteristic index data to be predicted include sunshine duration and sunshine intensity; acquiring third line loss characteristic index data to be predicted for distributed power generation of hydropower generation, wherein the weather factors in the third line loss characteristic index data to be predicted include average temperature and rainfall; and inputting the first line loss characteristic index data to be predicted, the second line loss characteristic index data to be predicted, and the third line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model.

[0143] In one embodiment, the distribution network line loss anomaly assessment module 205 is used to calculate the anomaly coefficient of the suspected abnormal feeder, and assess and obtain the distribution network line loss anomaly degree of the suspected abnormal feeder based on the anomaly coefficient. Specifically, this includes: using the cluster center of the category where the suspected abnormal feeder loss belongs as the reference feeder, obtaining the suspected abnormal feeder line loss prediction data and suspected abnormal feeder line loss characteristic index data of the suspected abnormal feeder; simultaneously obtaining the reference feeder line loss prediction data and reference feeder line loss characteristic index data of the reference feeder; calculating the difference between the suspected abnormal feeder line loss prediction data and the reference feeder line loss prediction data corresponding to a first preset time, and calculating the... The average difference of the difference is calculated; the deviation between the suspected abnormal feeder line loss characteristic index data and the benchmark feeder line loss characteristic index data is calculated, and the average deviation of each line loss characteristic index is calculated based on the number of line loss characteristic indices; the abnormality coefficient of the suspected abnormal feeder is calculated based on the average difference and the average deviation; the abnormality coefficient is compared with a preset abnormality coefficient threshold. If the abnormality coefficient is greater than or equal to the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a high degree of abnormality; if the abnormality coefficient is less than the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a low degree of abnormality.

[0144] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0145] It should be noted that the above-described embodiment of the distribution network feeder line loss anomaly assessment device is merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0146] Based on the above-described embodiments of the distribution network feeder line loss anomaly assessment method, another embodiment of the present invention provides a distribution network feeder line loss anomaly assessment terminal device. The distribution network feeder line loss anomaly assessment terminal device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the distribution network feeder line loss anomaly assessment method of any embodiment of the present invention.

[0147] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the line loss anomaly assessment terminal equipment of the distribution network feeder.

[0148] The terminal equipment for assessing line loss anomalies in the distribution network feeder can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal equipment for assessing line loss anomalies in the distribution network feeder may include, but is not limited to, a processor and a memory.

[0149] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the distribution network feeder line loss anomaly assessment terminal equipment, connecting various parts of the equipment via various interfaces and lines.

[0150] The memory can be used to store the computer programs and / or modules. The processor, by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory, realizes various functions of the distribution network feeder line loss anomaly assessment terminal equipment. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0151] Based on the above embodiments of the method for assessing abnormal line losses of distribution network feeders, another embodiment of the present invention provides a storage medium, the storage medium including a stored computer program, wherein, when the computer program is running, the device where the storage medium is located controls the execution of the method for assessing abnormal line losses of distribution network feeders according to any embodiment of the present invention.

[0152] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0153] In summary, this invention discloses a method and apparatus for assessing line loss anomalies in distribution network feeders. By constructing a line loss characteristic index system for distribution networks including distributed generation sources, and using collected distribution network line loss data and line loss characteristic index data from the line area to train an LSTM neural network model, an optimal distribution network line loss prediction model is obtained. This model is then used to predict line loss based on the line loss characteristic index data to be predicted. Based on the first distribution network line loss prediction data, a reasonable confidence interval for distribution network line loss is generated. Using the reasonable confidence interval and an improved fuzzy C-means algorithm, the first distribution network line loss prediction data is clustered to identify suspected abnormal feeders. By calculating the anomaly coefficient of the suspected abnormal feeders, the degree of distribution network line loss anomaly is assessed. Compared with existing technologies, this invention breaks away from the traditional method of using a line loss rate exceeding a certain threshold as the condition for line loss anomaly assessment. By creating anomaly coefficients to assess the degree of line loss anomaly, the accuracy of distribution network line loss anomaly assessment can be improved.

[0154] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for assessing abnormal line losses in a distribution network feeder, characterized in that, include: Based on the line loss characteristic index data, a line loss characteristic index system for distribution networks including distributed power sources is constructed. The line loss characteristic index data includes weather factors, equipment grid connection parameters, and distribution network operation data. Collect all distribution network line loss data for the selected line area within a preset time period, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data for the line area within the preset time period. Perform data processing on all distribution network line loss data and all line loss characteristic index data to obtain standard distribution network line loss data and standard line loss characteristic index data. Divide the standard distribution network line loss data and the standard line loss characteristic index data into training dataset and test dataset. Specifically, the process of processing all distribution network line loss data and all line loss characteristic index data to obtain standard distribution network line loss data and standardized line loss characteristic index data includes: Negative values ​​of distribution network line loss data are removed from all distribution network line loss data. At the same time, distribution network line loss data with a line loss rate greater than 25% are also removed from all distribution network line loss data to obtain the first distribution network line loss data. Determine whether there are missing data values ​​in the first distribution network line loss data or the data of all line loss characteristic indicators. If there are missing data values, determine whether the first distribution network line loss data or the data of all line loss characteristic indicators conforms to a uniform distribution. If yes, obtain the mean of the first distribution network line loss data or the data of all line loss characteristic indicators and fill the missing data values ​​with the mean. If no, obtain the median of the first distribution network line loss data or the data of all line loss characteristic indicators and fill the missing data values ​​with the median to obtain standard distribution network line loss data and first line loss characteristic indicator data. The first line loss characteristic index data is normalized to obtain standard line loss characteristic index data. An LSTM neural network model is constructed based on the LSTM algorithm. The LSTM neural network model is trained according to the training dataset and the test dataset to determine the optimal distribution network line loss prediction model. Obtain the line loss characteristic index data of different distributed power source types, input the line loss characteristic index data of different distributed power source types into the optimal distribution network line loss prediction model, output and calculate the reasonable confidence interval of the line loss of the distribution network for different distributed power source types based on the first distribution network line loss prediction data corresponding to different distributed power source types. The improved fuzzy C-means algorithm is used to perform a first clustering of the predicted line loss data of the first distribution network to obtain a first clustering result. Based on the reasonable confidence interval of the line loss, the first clustering result is then subjected to a second clustering to obtain suspected abnormal feeders. The abnormality coefficient of the suspected abnormal feeders is calculated. Based on the abnormality coefficient, the degree of abnormality of the distribution network line loss of the suspected abnormal feeders is evaluated and obtained.

2. The method for assessing abnormal line losses in a distribution network feeder as described in claim 1, characterized in that, Collect all distribution network line loss data for the selected line area within a preset time period, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data for the line area within the preset time period, specifically including: Obtain the distributed power penetration rate of each line area in the distribution network, select the line areas where the distributed power penetration rate is greater than the preset distributed power penetration threshold, and collect the distribution network line loss data of the line areas within one year. Based on the aforementioned power distribution network line loss characteristic index system, line loss characteristic index data for the aforementioned line area are collected.

3. The method for assessing abnormal line losses in a distribution network feeder as described in claim 1, characterized in that, An LSTM neural network model is constructed based on the LSTM algorithm. The LSTM neural network model is then trained using the training dataset and the test dataset to determine the optimal distribution network line loss prediction model. This process specifically includes: An LSTM neural network model is constructed based on the LSTM algorithm. The training dataset is input into the LSTM neural network model for model training to obtain an initial distribution network line loss prediction model. The test dataset is then input into the initial distribution network line loss prediction model to obtain the distribution network line loss prediction data corresponding to the test dataset. The predicted line loss data of the distribution network is compared with the predicted line loss data of the standard distribution network to determine the optimal line loss prediction model of the distribution network.

4. The method for assessing abnormal line losses in a distribution network feeder as described in claim 1, characterized in that, The line loss characteristic index data includes weather factors, equipment grid connection parameters, and distribution network operation data. The weather factors include average temperature, average wind speed, sunshine duration, sunshine intensity, and rainfall. The equipment grid connection parameters include grid connection capacity, grid connection location, grid connection operation mode, and power supply. The distribution network operation data includes data date, power supply type, power supply radius, line load rate, and power supply.

5. The method for assessing abnormal line losses in a distribution network feeder as described in claim 4, characterized in that, Acquire predictable line loss characteristic index data for different types of distributed power sources, and input the predictable line loss characteristic index data into the optimal distribution network line loss prediction model, specifically including: Acquire the first predictable line loss characteristic index data for distributed power generation of wind power generation, wherein the weather factors in the first predictable line loss characteristic index data include average temperature and average wind speed. Acquire the second line loss characteristic index data for distributed power source type photovoltaic power generation, wherein the weather factors in the second line loss characteristic index data include sunshine duration and sunshine intensity; Obtain the third predictable line loss characteristic index data for distributed power generation of hydropower type, wherein the weather factors in the third predictable line loss characteristic index data include average temperature and rainfall. The first line loss characteristic index data to be predicted, the second line loss characteristic index data to be predicted, and the third line loss characteristic index data to be predicted are input into the optimal distribution network line loss prediction model.

6. The method for assessing abnormal line losses in a distribution network feeder as described in claim 1, characterized in that, Calculate the anomaly coefficient of the suspected abnormal feeder, and based on the anomaly coefficient, assess and obtain the degree of distribution network line loss anomaly of the suspected abnormal feeder, specifically including: Using the cluster center of the category where the suspected abnormal feeder loss is located as the baseline feeder, obtain the suspected abnormal feeder loss prediction data and suspected abnormal feeder loss characteristic index data of the suspected abnormal feeder, and at the same time obtain the baseline feeder loss prediction data and baseline feeder loss characteristic index data of the baseline feeder. Calculate the difference between the predicted data of the suspected abnormal feeder line loss corresponding to the first preset time and the predicted data of the baseline feeder line loss, and calculate the average difference of all differences; Calculate the deviation between the suspected abnormal feeder line loss characteristic index data and the benchmark feeder line loss characteristic index data, and calculate the average deviation value corresponding to each line loss characteristic index based on the number of line loss characteristic indices. The abnormality coefficient of the suspected abnormal feeder is calculated and obtained based on the average difference and the average deviation. The abnormality coefficient is compared with a preset abnormality coefficient threshold. If the abnormality coefficient is greater than or equal to the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a high degree of abnormality. If the abnormality coefficient is less than the preset abnormality coefficient threshold, the distribution network line loss of the suspected abnormal feeder is considered to be of a low degree of abnormality.

7. A device for assessing abnormal line losses in a distribution network feeder, characterized in that, include: The system includes modules for constructing a characteristic index system for distribution network line losses, data processing for distribution networks, building a prediction model for distribution network line losses, calculating reasonable confidence intervals for line losses, and assessing the degree of anomalies in distribution network line losses. The distribution network line loss characteristic index system construction module is used to acquire and construct a distribution network line loss characteristic index system including distributed power sources based on the line loss characteristic index data. The line loss characteristic index data includes weather factors, equipment grid connection parameters and distribution network operation data. The distribution network data processing module is used to collect all distribution network line loss data of the selected line area within a preset time, and based on the distribution network line loss characteristic index system, collect all line loss characteristic index data of the line area within the preset time, process all distribution network line loss data and all line loss characteristic index data respectively to obtain standard distribution network line loss data and standard line loss characteristic index data, and divide the standard distribution network line loss data and the standard line loss characteristic index data into training dataset and test dataset; Specifically, the process of processing all distribution network line loss data and all line loss characteristic index data to obtain standard distribution network line loss data and standardized line loss characteristic index data includes: Negative values ​​of distribution network line loss data are removed from all distribution network line loss data. At the same time, distribution network line loss data with a line loss rate greater than 25% are also removed from all distribution network line loss data to obtain the first distribution network line loss data. Determine whether there are missing data values ​​in the first distribution network line loss data or the data of all line loss characteristic indicators. If there are missing data values, determine whether the first distribution network line loss data or the data of all line loss characteristic indicators conforms to a uniform distribution. If yes, obtain the mean of the first distribution network line loss data or the data of all line loss characteristic indicators and fill the missing data values ​​with the mean. If no, obtain the median of the first distribution network line loss data or the data of all line loss characteristic indicators and fill the missing data values ​​with the median to obtain standard distribution network line loss data and first line loss characteristic indicator data. The first line loss characteristic index data is normalized to obtain standard line loss characteristic index data. The distribution network line loss prediction model construction module is used to construct an LSTM neural network model based on the LSTM algorithm, train the LSTM neural network model according to the training dataset and the test dataset, and determine the optimal distribution network line loss prediction model. The reasonable confidence interval calculation module for line loss is used to obtain the line loss characteristic index data to be predicted for different types of distributed power sources, input the line loss characteristic index data to be predicted into the optimal distribution network line loss prediction model, and output and calculate the reasonable confidence interval for line loss of distribution network for different types of distributed power sources based on the first distribution network line loss prediction data corresponding to different types of distributed power sources. The distribution network line loss anomaly assessment module is used to perform a first clustering of the first distribution network line loss prediction data based on the improved fuzzy C-means algorithm to obtain a first clustering result, and then perform a second clustering of the first clustering result based on the reasonable confidence interval of the line loss to obtain suspected abnormal feeders. The module calculates the anomaly coefficient of the suspected abnormal feeder and assesses and obtains the distribution network line loss anomaly degree of the suspected abnormal feeder based on the anomaly coefficient.

8. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method for assessing line loss anomalies in distribution network feeders as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the line loss anomaly assessment method for distribution network feeders as described in any one of claims 1 to 6.

Images