Method and device for predicting line loss in distribution network

Abstract

The present invention provides a method and device for predicting line loss of a distribution network, and belongs to the field of electric power engineering technology. The method comprises: obtaining characteristic data sets of a distribution network to be predicted in a first time period and a second time period, respectively, as well as meteorological data and holiday data of the time period to be predicted; projecting the characteristic data set of the first time period to a first coordinate system to obtain a first test set, and projecting the characteristic data set of the second time period to a second coordinate system to obtain a second test set; wherein the first coordinate system is a preset number of characteristic vectors of a divergence matrix composed of the characteristic data set of the first time period, and the second coordinate system is a preset number of characteristic vectors of a divergence matrix composed of the characteristic data set of the second time period; inputting the meteorological difference and holiday difference between the first test set, the second test set, the time period to be predicted, and the second time period and the first time period into a trained line loss prediction model, and performing line loss prediction on the distribution network. The present invention can improve the accuracy of line loss prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power engineering, and in particular to a method and device for predicting line loss in a distribution network. Background Art

[0002] In the development of power systems, line loss is a comprehensive indicator that reflects the power grid equipment, production operation and management level of power supply enterprises. In the process of power station operation and maintenance, the accurate prediction of line loss is particularly important for the operation of power stations. It not only affects the accuracy of power consumption declaration of industrial enterprise parks, but also provides necessary data support for reducing line loss in the future.

[0003] At present, the common method for predicting line losses in distribution networks is to predict line losses based on the historical total output power and total power load, or to cluster the historical power attribute information according to a pre-set radius and predict line losses through the cluster center of each cluster.

[0004] However, due to the large amount of power attribute information, how to improve the accuracy of distribution network line loss prediction has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The embodiment of the present invention provides a method and device for predicting line loss in a distribution network, so as to solve the problem that the current line loss prediction method is inaccurate.

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting line loss in a distribution network, comprising:

[0007] Acquire characteristic data sets of the distribution network to be predicted in the first time period and the second time period, as well as meteorological data and holiday data of the time period to be predicted, wherein the first time period is a historical time period with a preset interval adjacent to the time period to be predicted, and the second time period is a time period in the same period of the year before the time period to be predicted;

[0008] Projecting the feature data set of the first time period into the first coordinate system to obtain the first test set, and projecting the feature data set of the second time period into the second coordinate system to obtain the second test set; wherein the first coordinate system is a preset number of feature vectors of the scatter matrix composed of the feature data set of the first time period, and the second coordinate system is a preset number of feature vectors of the scatter matrix composed of the feature data set of the second time period;

[0009] The meteorological differences and holiday differences between the first test set, the second test set, the time period to be predicted, the second time period and the first time period are respectively input into the trained line loss prediction model to predict the line loss of the distribution network.

[0010] In a possible implementation, the feature data set of the first period includes n features;

[0011] Projecting the feature data set of the first period into the first coordinate system to obtain a first test set includes:

[0012] Standardizing the matrix composed of the feature data set of the first period, and determining a first scatter matrix of the feature data set of the first period after the standardization;

[0013] Calculate the eigenvalues ​​of the first scatter matrix and the eigenvectors corresponding to each eigenvalue;

[0014] Based on the size of the eigenvalues ​​of the first scatter matrix, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first n eigenvectors are formed into a first coordinate system;

[0015] A dimension reduction process is performed on the standardized feature data set of the first period based on the first coordinate system to obtain a first test set.

[0016] In a possible implementation, the feature data set of the second period includes n features;

[0017] Projecting the feature data set of the second period into the second coordinate system to obtain a second test set includes:

[0018] Standardizing the matrix composed of the feature data set of the second period, and determining a second scatter matrix of the feature data set of the first period after the standardization;

[0019] Calculate the eigenvalues ​​of the second scatter matrix and the eigenvectors corresponding to each eigenvalue;

[0020] Based on the size of the eigenvalues ​​of the second scatter matrix, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first n eigenvectors form a second coordinate system;

[0021] The standardized feature data set of the second period is subjected to dimensionality reduction processing based on the second coordinate system to obtain a second test set.

[0022] In a possible implementation, the line loss prediction model includes a prediction module and a correction module, and the correction module includes a weather correction module and a holiday correction module;

[0023] The meteorological correction module is used to determine a first correction factor based on the meteorological characteristics of the time period to be predicted, the meteorological characteristics of the second time period, and the meteorological characteristics of the first time period;

[0024] The holiday correction module is used to determine the second correction factor based on the holiday data of the time period to be predicted and the holiday data of the first time period;

[0025] The first correction factor and the second correction factor are used to correct the prediction result of the prediction module to obtain the final prediction result of the line loss prediction model.

[0026] In one possible implementation, the line loss prediction model is trained based on a historical data set of at least two years of the distribution network to be predicted. The historical data set includes actual line loss in the substation, meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section.

[0027] In one possible implementation, the training process of the line loss prediction model is as follows:

[0028] Determine a first input data set and a second input data set of training samples, wherein the first input data set is obtained by projecting the first historical data set onto a projection coordinate system of the first historical data set, and the projection coordinate system of the first historical data set is a preset number of eigenvectors of a divergence matrix of the first historical data set; the second input data set is obtained by projecting the second historical data set onto a projection coordinate system of the second historical data set, and the projection coordinate system of the second historical data set is a preset number of eigenvectors of a divergence matrix of the second historical data set; the first historical data set is a historical data set within a time period of a preset interval adjacent to the forecast period, and the second historical data set is a historical data set within the same period of the year before the forecast period;

[0029] Determine a meteorological data difference set and a holiday difference set corresponding to the first input data set and the second input data set;

[0030] Based on the first input data set, the second input data set, the meteorological data difference set and the holiday difference set, the neural network model is trained to obtain a line loss prediction model.

[0031] In a possible implementation, the first input data set is based on normalizing the first historical data set, and the normalized data constitutes the first training data set; the divergence matrix of the first training data set is calculated, and the eigenvalues ​​and eigenvectors of the divergence matrix are obtained; based on the size of the eigenvalues ​​of the divergence matrix of the first training data set, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first 7 eigenvectors constitute the projection coordinate system of the first historical data set; based on the projection coordinate system of the first historical data set, the first training data set is subjected to dimensionality reduction processing to obtain the first input data set;

[0032] The second input data set is based on normalizing the second historical data set, and the normalized data constitute the second training data set; the divergence matrix of the second training data set is calculated, and the eigenvalues ​​and eigenvectors of the divergence matrix are obtained; based on the size of the eigenvalues ​​of the divergence matrix of the second training data set, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first 7 eigenvectors form the projection coordinate system of the second historical data set; based on the projection coordinate system of the second historical data set, the second training data set is subjected to dimensionality reduction processing to obtain the second input data set.

[0033] In a possible implementation manner, after obtaining the characteristic data sets of the distribution network to be predicted in the first time period and the second time period respectively, the method further includes:

[0034] Missing values ​​and / or outliers are processed for the feature data sets of the first period and the second period.

[0035] In a possible implementation, the feature data set includes meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity, and cable cross-section.

[0036] In a second aspect, an embodiment of the present invention provides a device for predicting line loss in a distribution network, comprising:

[0037] An acquisition module is used to acquire characteristic data sets of the distribution network to be predicted in a first time period and a second time period, as well as meteorological data and holiday data of the time period to be predicted, wherein the first time period is a historical time period with a preset interval adjacent to the time period to be predicted, and the second time period is a time period in the same period of the year before the time period to be predicted;

[0038] A projection module, used to project the feature data set of the first time period into a first coordinate system to obtain a first test set, and project the feature data set of the second time period into a second coordinate system to obtain a second test set; wherein the first coordinate system is a preset number of feature vectors of the scatter matrix composed of the feature data set of the first time period, and the second coordinate system is a preset number of feature vectors of the scatter matrix composed of the feature data set of the second time period;

[0039] The prediction module is used to input the meteorological differences and holiday differences between the first test set, the second test set, the time period to be predicted, the second time period and the first time period into the trained line loss prediction model to predict the line loss of the distribution network.

[0040] The embodiment of the present invention provides a method and device for predicting line loss of a distribution network. First, characteristic data sets of a distribution network to be predicted in a first time period and a second time period are obtained, as well as meteorological data and holiday data of the time period to be predicted. Then, the characteristic data set of the first time period is projected to the first coordinate system to obtain a first test set, and the characteristic data set of the second time period is projected to the second coordinate system to obtain a second test set. Finally, the first test set, the second test set, and the meteorological difference and holiday difference between the time period to be predicted and the second time period and the first time period are input into a trained line loss prediction model to predict line loss of the distribution network. By using the historical data adjacent to the time period to be predicted and the historical data of the same period of the previous year of the time period to be predicted, as well as the meteorological difference and holiday difference between the time period to be predicted and the second time period and the first time period, the trained line loss prediction model is used to predict line loss, which can fully consider the influence of meteorology and holidays on line loss. In addition, the line loss of the time period to be predicted is predicted by using the historical data adjacent to the time period to be predicted and the historical data of the same period of the previous year of the time period to be predicted, so that the line loss can be predicted more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a flow chart of an implementation of a method for predicting line loss in a distribution network provided by an embodiment of the present invention;

[0043] Figure 2 It is a structural schematic diagram of a device for predicting line loss in a distribution network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0046] Figure 1 The implementation flow chart of the method for predicting line loss in a distribution network provided by the embodiment of the present invention is described in detail as follows:

[0047] Step S110: Acquire characteristic data sets of the distribution network to be predicted in the first time period and the second time period, respectively, as well as meteorological data and holiday data of the time period to be predicted.

[0048] The first period is a historical period with a preset interval adjacent to the period to be predicted, and the second period is a period of the same period of the year before the period to be predicted. The preset interval here can be 6 months, 3 months, 1 month, or the same as the interval of the period to be predicted, which is not limited here, and the preset interval is less than 12 months. And the last period in the first period is the period immediately adjacent to the period at the beginning of the period to be predicted.

[0049] Line loss is an unavoidable power loss phenomenon in the distribution network. Line loss is line loss, specifically the energy loss caused when electric energy flows through the transmission line.

[0050] For example, if the period to be predicted is April to June 2024, then the first period can be January to March 2024, and the second period can be April to June 2023.

[0051] Meteorological data may include temperature, humidity, and wind speed. Meteorological data for the forecast period can be obtained through the meteorological forecast platform. Holidays include weekends and national statutory holidays. Holiday data can be determined through the holiday calendar. Holidays and meteorological data will seriously affect the load and power of the distribution network.

[0052] In some embodiments, in order to accurately predict the line loss of the distribution network, the feature data set may include meteorological data, average operating years of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross section. Each feature in the feature data set has a certain correlation with the line loss, so the line loss can be predicted by these features.

[0053] The meteorological data of the first and second periods, the average operating life of the equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section data are all real data measured in history and can be found through historical records. The recording frequency can be the average data of each day or the average data of multiple days, which can be set according to the specific application scenario and is not limited here.

[0054] In this embodiment, the meteorological data may include temperature, humidity, and wind speed.

[0055] In some embodiments, after obtaining the characteristic data sets of the first time period and the second time period, as well as the meteorological data and holiday data of the time period to be predicted, these data need to be processed, such as processing missing values ​​or outliers, to improve the accuracy of the original input data.

[0056] Step S120: project the feature data set of the first time period into the first coordinate system to obtain a first test set, and project the feature data set of the second time period into the second coordinate system to obtain a second test set.

[0057] The first coordinate system is a preset number of characteristic vectors of the scatter matrix composed of the characteristic data set of the first period. The second coordinate system is a preset number of characteristic vectors of the scatter matrix composed of the characteristic data set of the second period.

[0058] In some embodiments, since the feature data set not only includes multiple features, but each feature also includes data from multiple time periods, the feature data set is a high-dimensional data set. If the high-dimensional data set is used directly, the amount of data analysis will increase, which will increase the difficulty of prediction. In addition, it is impossible to obtain the data correlation between each feature data in these feature data sets to improve the accuracy of prediction. In order to further explore the correlation of feature data in the feature data set and extract accurate power fingerprint data for prediction, it is necessary to process the feature data set. The specific processing process is as follows:

[0059] For example, the feature data set for the first period includes 7 features, and the feature data set for the second period includes 7 features. These 7 features are meteorological data, average equipment operating years, holiday data, three-phase load data, power supply radius, distribution transformer capacity, and cable cross section. Each feature includes data from m periods.

[0060] Since the types of data in the feature data set are different, the feature data set needs to be standardized first. The matrix composed of the feature data set of the first period is standardized, and the first scatter matrix of the feature data set of the first period after the standardized processing is determined.

[0061] For example, after the standardized processing, the feature data set M1 of the first period is {x1, x2, x3, x4, x5, x6, x7}, and the covariance between the feature data sets of each dimension is calculated to form a covariance matrix Cov (x1, x2, x3, x4, x5, x6, x7). Cov is a covariance calculation function, which represents the correlation degree of two feature data.

[0062] Then, the first divergence matrix Q1 of the feature data set M1 of the first time period after the standardized processing is determined, Q1 = (m-1) × Cov (x1, x2, x3, x4, x5, x6, x7), m means that each feature contains data of m time periods.

[0063] Next, the eigenvalues ​​of the first scatter matrix Q1 and the eigenvectors corresponding to each eigenvalue are calculated. Based on the size of the eigenvalues ​​of the first scatter matrix Q1, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small. The first 7 eigenvectors form the first coordinate system R1. Alternatively, the number of eigenvalues ​​to be retained can be determined by the cumulative contribution rate of the eigenvalues, and the eigenvectors corresponding to the eigenvalues ​​can be selected according to the retained eigenvalues.

[0064] Finally, the dimension reduction process is performed on the standardized feature data set M1 of the first period based on the first coordinate system R1 to obtain a first test set S1, where S1 = R1 × M1.

[0065] Similarly, after the standardized processing, the feature data set M2 = {y1, y2, y3, y4, y5, y6, y7} of the second period is used to calculate the covariance between the feature data sets of each dimension to form a covariance matrix Cov (y1, y2, y3, y4, y5, y6, y7). Cov is a covariance calculation function, which represents the correlation between two feature data.

[0066] Then, the second scatter matrix Q2 of the feature data set M2 of the second time period after the standardized processing is determined, Q2 = (m-1) × Cov (y1, y2, y3, y4, y5, y6, y7), m means that each feature contains data of m time periods.

[0067] Next, the eigenvalues ​​of the second scatter matrix Q2 and the eigenvectors corresponding to each eigenvalue are calculated. Based on the size of the eigenvalues ​​of the second scatter matrix Q2, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small. The first 7 eigenvectors form the second coordinate system R2. Alternatively, the number of eigenvalues ​​to be retained can be determined by the cumulative contribution rate of the eigenvalues, and the eigenvectors corresponding to the eigenvalues ​​can be selected according to the retained eigenvalues.

[0068] Finally, the dimension reduction process is performed on the standardized feature data set M2 of the second period based on the second coordinate system R2 to obtain a second test set S2, where S2 = R2×M2.

[0069] Step S130: input the meteorological differences and holiday differences between the first test set, the second test set, the time period to be predicted, the second time period and the first time period into the trained line loss prediction model to predict the line loss of the distribution network.

[0070] In some embodiments, the line loss prediction model includes a prediction module and a correction module, and the correction module includes a weather correction module and a holiday correction module.

[0071] Since holidays and meteorological data will seriously affect the load and power of the distribution network, and thus have a greater impact on line losses, it is necessary to consider the impact of holidays and meteorological data on line loss prediction.

[0072] In this embodiment, the meteorological correction module is used to determine the first correction factor based on the meteorological characteristics of the time period to be predicted, the meteorological characteristics of the second time period, and the meteorological characteristics of the first time period.

[0073] Specifically, based on the meteorological characteristics of the first time period and the meteorological characteristics of the time period to be predicted, a first difference between the first time period and the time period to be predicted is determined. Based on the meteorological characteristics of the second time period and the meteorological characteristics of the time period to be predicted, a second difference between the second time period and the time period to be predicted is determined. Then, based on the first difference and the second difference, a first correction factor is determined. Exemplarily, different weights can be assigned to the first difference and the second difference, respectively, and the respective weights of the first difference and the second difference are obtained based on training sample training.

[0074] The meteorological characteristics here include temperature, humidity and wind speed, so the first difference is a fitting curve of the temperature difference, humidity difference and wind speed difference, and the second difference is also a fitting curve of the temperature difference, humidity difference and wind speed difference.

[0075] In this embodiment, the holiday correction module is used to determine the second correction factor based on the holiday data of the time period to be predicted, the holiday data of the first time period and the holiday data of the second time period.

[0076] Specifically, based on the holiday data of the first time period and the holiday data of the time period to be predicted, a third difference between the first time period and the time period to be predicted is determined. Based on the holiday data of the second time period and the holiday data of the time period to be predicted, a fourth difference between the second time period and the time period to be predicted is determined. Then, based on the third difference and the fourth difference, a second correction factor is determined. Exemplarily, different weights can be assigned to the third difference and the fourth difference, respectively, and the respective weights of the third difference and the fourth difference are obtained based on training sample training.

[0077] In some embodiments, the line loss prediction model is trained based on a historical data set of at least 2 years of the distribution network to be predicted. The historical data set includes actual line loss in the substation area, meteorological data, average operating years of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross section.

[0078] For example, obtain the actual line loss, meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section of the distribution network to be predicted from January 2020 to December 2022.

[0079] A training set is constructed, and the training set includes training samples and test samples. The training set includes a first historical data set, a second historical data set, a meteorological data difference set, and a holiday difference set, wherein the meteorological data difference set includes a first meteorological data difference set between the period to be predicted and the first historical data set, and a second meteorological data difference set between the prediction period and the second historical data set. The meteorological data and holidays of the period to be predicted are also known. The holiday difference includes a first holiday difference set between the period to be predicted and the first historical data set, and a second holiday difference set between the prediction period and the second historical data set.

[0080] The time period corresponding to the first historical data set is a time period adjacent to the period to be predicted, and the time period corresponding to the second historical data set is a time period of the same period of the year before the period to be predicted.

[0081] To facilitate understanding, the method of constructing one of the samples in the training set is explained. Since the training set needs to include many samples, they will not be described here one by one.

[0082] For example, the line loss in the period to be predicted is November 2022, and the first historical data is the meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section in October 2022. The second historical data is the meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section in November 2021. Then, based on the first historical data, the second historical data and the period to be predicted, the meteorological data difference and the holiday difference are determined. It should be noted that the meteorological data difference here includes 2 meteorological data differences, including the difference between the meteorological data in November 2022 and the meteorological data in October 2022, and the difference between the meteorological data in November 2022 and the meteorological data in November 2021. The holiday difference also includes two differences, the difference between the meteorological data in November 2022 and the holiday difference in October 2022, and the difference between the meteorological data in November 2022 and the holiday difference in November 2021.

[0083] Based on the same construction method, many groups of historical data are constructed to form training samples, thereby improving the training accuracy of the constructed line loss prediction model.

[0084] After obtaining the first historical data set and the second historical data set, it is necessary to perform normalization processing on the first historical data set and the second historical data set respectively to obtain the first training data set and the second training data set.

[0085] Then, a first training scatter matrix corresponding to the first training data set and a second training scatter matrix corresponding to the second training data set are calculated respectively. The calculation method of the scatter matrix is ​​the same as the calculation method of the scatter matrix in step S120, which will not be repeated here.

[0086] Next, the eigenvalues ​​of the first training scatter matrix and the second training scatter matrix and the eigenvectors corresponding to each eigenvalue are obtained respectively. The eigenvalues ​​of the first training scatter matrix and the eigenvalues ​​of the second training scatter matrix are sorted from large to small respectively. Based on the sorting of the eigenvalues ​​of the first training scatter matrix, the first 7 eigenvectors are selected to form the projection coordinate system of the first historical data set. Based on the sorting of the eigenvalues ​​of the second training scatter matrix, the first 7 eigenvectors are selected to form the projection coordinate system of the second historical data set.

[0087] Finally, the first training data set is processed by dimensionality reduction based on the projection coordinate system of the first historical data set to obtain a first input data set, and the second training data set is processed by dimensionality reduction based on the projection coordinate system of the second historical data set to obtain a second input data set.

[0088] After obtaining the first input data set, the second input data set, the meteorological data difference set and the holiday difference set, these data can be used to train the constructed neural network model.

[0089] The meteorological data difference set includes a first meteorological data difference set and a second meteorological data difference set, and the holiday difference set includes a first holiday difference set and a second holiday difference set.

[0090] The neural network model may include an input layer, a multi-layer LSTM unit and an output layer, wherein the multi-layer LSTM unit includes a prediction module and a correction module, and the correction module includes a weather correction module and a holiday correction module.

[0091] Therefore, when predicting line losses, not only the damage to power data is taken into account, but also the impact of weather, holidays and equipment aging on line losses is fully considered, so that the distribution network line losses can be predicted more accurately.

[0092] The prediction method provided by the embodiment of the present invention first obtains the characteristic data sets of the distribution network to be predicted in the first time period and the second time period, as well as the meteorological data and holiday data of the time period to be predicted, then projects the characteristic data set of the first time period to the first coordinate system to obtain the first test set, and projects the characteristic data set of the second time period to the second coordinate system to obtain the second test set. Finally, the first test set, the second test set, the meteorological difference and holiday difference between the time period to be predicted and the second time period and the first time period are input into the trained line loss prediction model to predict the line loss of the distribution network. By using the historical data adjacent to the time period to be predicted and the historical data of the same period of the previous year of the time period to be predicted, as well as the meteorological difference and holiday difference between the time period to be predicted and the second time period and the first time period, the trained line loss prediction model is used to predict the line loss, which can fully consider the influence of meteorology and holidays on the line loss. In addition, the line loss of the time period to be predicted is predicted by using the historical data adjacent to the time period to be predicted and the historical data of the same period of the previous year of the time period to be predicted, so that the line loss can be predicted more accurately.

[0093] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0094] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0095] Figure 2 The schematic diagram of the structure of the device for predicting line loss of a distribution network provided by an embodiment of the present invention is shown. For the convenience of explanation, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0096] like Figure 2 As shown, the distribution network line loss prediction device 200 includes:

[0097] The acquisition module 210 is used to acquire characteristic data sets of the distribution network to be predicted in the first time period and the second time period, as well as meteorological data and holiday data of the time period to be predicted, wherein the first time period is a historical time period with a preset interval adjacent to the time period to be predicted, and the second time period is a time period in the same period of the year before the time period to be predicted;

[0098] A projection module 220 is used to project the feature data set of the first time period into a first coordinate system to obtain a first test set, and to project the feature data set of the second time period into a second coordinate system to obtain a second test set; wherein the first coordinate system is a preset number of feature vectors of the scatter matrix composed of the feature data set of the first time period, and the second coordinate system is a preset number of feature vectors of the scatter matrix composed of the feature data set of the second time period;

[0099] The prediction module 230 is used to input the meteorological differences and holiday differences between the first test set, the second test set, the time period to be predicted, the second time period and the first time period into the trained line loss prediction model to predict the line loss of the distribution network.

[0100] In a possible implementation, the feature data set of the first period includes n features;

[0101] The projection module 220 is used to perform a standardization process on the matrix composed of the feature data set of the first period, and determine a first scatter matrix of the feature data set of the first period after the standardization process;

[0102] Calculate the eigenvalues ​​of the first scatter matrix and the eigenvectors corresponding to each eigenvalue;

[0103] Based on the size of the eigenvalues ​​of the first scatter matrix, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first n eigenvectors are formed into a first coordinate system;

[0104] A dimension reduction process is performed on the standardized feature data set of the first period based on the first coordinate system to obtain a first test set.

[0105] In a possible implementation, the feature data set of the second period includes n features;

[0106] The projection module 220 is used to perform a normalization process on the matrix composed of the feature data set of the second period, and determine a second scatter matrix of the feature data set of the first period after the normalization process;

[0107] Calculate the eigenvalues ​​of the second scatter matrix and the eigenvectors corresponding to each eigenvalue;

[0108] Based on the size of the eigenvalues ​​of the second scatter matrix, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first n eigenvectors form a second coordinate system;

[0109] The standardized feature data set of the second period is subjected to dimensionality reduction processing based on the second coordinate system to obtain a second test set.

[0110] In a possible implementation, the line loss prediction model includes a prediction module and a correction module, and the correction module includes a weather correction module and a holiday correction module;

[0111] The meteorological correction module is used to determine a first correction factor based on the meteorological characteristics of the time period to be predicted, the meteorological characteristics of the second time period, and the meteorological characteristics of the first time period;

[0112] The holiday correction module is used to determine the second correction factor based on the holiday data of the time period to be predicted and the holiday data of the first time period;

[0113] The first correction factor and the second correction factor are used to correct the prediction result of the prediction module to obtain the final prediction result of the line loss prediction model.

[0114] In one possible implementation, the line loss prediction model is trained based on a historical data set of at least two years of the distribution network to be predicted. The historical data set includes actual line loss in the substation, meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section.

[0115] In one possible implementation, the training process of the line loss prediction model is as follows:

[0116] Determine a first input data set and a second input data set of training samples, wherein the first input data set is obtained by projecting the first historical data set onto a projection coordinate system of the first historical data set, and the projection coordinate system of the first historical data set is a preset number of eigenvectors of a divergence matrix of the first historical data set; the second input data set is obtained by projecting the second historical data set onto a projection coordinate system of the second historical data set, and the projection coordinate system of the second historical data set is a preset number of eigenvectors of a divergence matrix of the second historical data set; the first historical data set is a historical data set within a time period of a preset interval adjacent to the forecast period, and the second historical data set is a historical data set within the same period of the year before the forecast period;

[0117] Determine a meteorological data difference set and a holiday difference set corresponding to the first input data set and the second input data set;

[0118] Based on the first input data set, the second input data set, the meteorological data difference set and the holiday difference set, the neural network model is trained to obtain a line loss prediction model.

[0119] In a possible implementation, the first input data set is based on normalizing the first historical data set, and the normalized data constitutes the first training data set; the divergence matrix of the first training data set is calculated, and the eigenvalues ​​and eigenvectors of the divergence matrix are obtained; based on the size of the eigenvalues ​​of the divergence matrix of the first training data set, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first 7 eigenvectors constitute the projection coordinate system of the first historical data set; based on the projection coordinate system of the first historical data set, the first training data set is subjected to dimensionality reduction processing to obtain the first input data set;

[0120] The second input data set is based on normalizing the second historical data set, and the normalized data constitute the second training data set; the divergence matrix of the second training data set is calculated, and the eigenvalues ​​and eigenvectors of the divergence matrix are obtained; based on the size of the eigenvalues ​​of the divergence matrix of the second training data set, the eigenvectors corresponding to all eigenvalues ​​are sorted from large to small, and the first 7 eigenvectors form the projection coordinate system of the second historical data set; based on the projection coordinate system of the second historical data set, the second training data set is subjected to dimensionality reduction processing to obtain the second input data set.

[0121] In a possible implementation, the acquisition module is used to process missing values ​​and / or abnormal values ​​of the feature data sets of the first time period and the second time period.

[0122] In a possible implementation, the feature data set includes meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity, and cable cross-section.

[0123] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0124] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0125] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned distribution network line loss prediction method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.

[0126] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for predicting line loss in a distribution network, characterized in that: include: Acquire characteristic data sets of the distribution network to be predicted in the first time period and the second time period, respectively, as well as meteorological data and holiday data of the time period to be predicted, wherein the first time period is a historical time period with a preset interval adjacent to the time period to be predicted, and the second time period is a time period in the same period of the year before the time period to be predicted; the characteristic data set of the first time period includes n features, and the characteristic data set of the second time period includes n features; the characteristic data set includes meteorological data, average operating years of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross section; Standardize the matrix composed of the feature data set of the first period, and determine the first scatter matrix of the feature data set of the first period after the standardization; calculate the eigenvalues ​​of the first scatter matrix and the eigenvectors corresponding to each eigenvalue; based on the size of the eigenvalues ​​of the first scatter matrix, sort the eigenvectors corresponding to all the eigenvalues ​​from large to small, and form the first n eigenvectors into a first coordinate system; perform dimensionality reduction processing on the feature data set of the first period after the standardization based on the first coordinate system to obtain a first test set; Standardize the matrix composed of the feature data set of the second period, and determine the second scatter matrix of the feature data set of the second period after the standardization; calculate the eigenvalues ​​of the second scatter matrix and the eigenvectors corresponding to each eigenvalue; based on the size of the eigenvalues ​​of the second scatter matrix, sort the eigenvectors corresponding to all the eigenvalues ​​from large to small, and form the first n eigenvectors into a second coordinate system; perform dimensionality reduction processing on the feature data set of the second period after the standardization based on the second coordinate system to obtain a second test set; The first test set, the second test set, the meteorological difference between the time period to be predicted and the second time period, the meteorological difference between the time period to be predicted and the first time period, the holiday difference between the time period to be predicted and the second time period, and the holiday difference between the time period to be predicted and the first time period are input into the trained line loss prediction model to perform line loss prediction on the distribution network.

2. The method for predicting line loss in a distribution network according to claim 1, characterized in that: The line loss prediction model includes a prediction module and a correction module, and the correction module includes a weather correction module and a holiday correction module; The meteorological correction module is used to determine a first correction factor based on the meteorological characteristics of the time period to be predicted, the meteorological characteristics of the second time period, and the meteorological characteristics of the first time period; The holiday correction module is used to determine a second correction factor based on the holiday data of the time period to be predicted, the holiday data of the first time period, and the holiday data of the second time period; The first correction factor and the second correction factor are used to correct the prediction result of the prediction module to obtain the final prediction result of the line loss prediction model.

3. The method for predicting line loss in a distribution network according to claim 2, characterized in that: The line loss prediction model is trained based on a historical data set of at least 2 years in the history of the distribution network to be predicted, and the historical data set includes actual line loss in the substation, meteorological data, average operating life of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross-section.

4. The method for predicting line loss in a distribution network according to claim 3, characterized in that: The training process of the line loss prediction model is as follows: Determine a first input data set and a second input data set of training samples, wherein the first input data set is obtained by projecting the first historical data set onto a projection coordinate system of the first historical data set, and the projection coordinate system of the first historical data set is a preset number of eigenvectors of a divergence matrix of the first historical data set; the second input data set is obtained by projecting the second historical data set onto a projection coordinate system of the second historical data set, and the projection coordinate system of the second historical data set is a preset number of eigenvectors of a divergence matrix of the second historical data set; the first historical data set is a historical data set within a time period of a preset interval adjacent to a prediction period, and the second historical data set is a historical data set for a time period of the same period of the year before the prediction period; Determine a meteorological data difference set and a holiday difference set corresponding to the first input data set and the second input data set; Based on the first input data set, the second input data set, the meteorological data difference set and the holiday difference set, a neural network model is trained to obtain the line loss prediction model.

5. The method for predicting line loss in a distribution network according to claim 4, characterized in that: The first input data set is based on normalizing the first historical data set, and the normalized data constitutes a first training data set; Calculate the scatter matrix of the first training data set, and obtain the eigenvalues ​​and eigenvectors of the scatter matrix; sort the eigenvectors corresponding to all eigenvalues ​​from large to small based on the eigenvalues ​​of the scatter matrix of the first training data set, and form the projection coordinate system of the first historical data set with the first 7 eigenvectors; perform dimensionality reduction processing on the first training data set based on the projection coordinate system of the first historical data set to obtain the first input data set; The second input data set is based on normalizing the second historical data set, and the normalized data constitutes a second training data set; Calculate the scatter matrix of the second training data set, and obtain the eigenvalues ​​and eigenvectors of the scatter matrix; based on the size of the eigenvalues ​​of the scatter matrix of the second training data set, sort the eigenvectors corresponding to all eigenvalues ​​from large to small, and form the projection coordinate system of the second historical data set with the first 7 eigenvectors; perform dimensionality reduction processing on the second training data set based on the projection coordinate system of the second historical data set to obtain the second input data set.

6. The method for predicting line loss in a distribution network according to claim 1, characterized in that: After obtaining the characteristic data sets of the distribution network to be predicted in the first time period and the second time period respectively, the method further includes: Missing values ​​and / or outlier processing are performed on the feature data sets of the first time period and the second time period.

7. A device for predicting line loss in a distribution network, characterized in that: include: An acquisition module is used to acquire feature data sets of the distribution network to be predicted in a first time period and a second time period, respectively, as well as meteorological data and holiday data of the time period to be predicted, wherein the first time period is a historical time period with a preset interval adjacent to the time period to be predicted, and the second time period is a time period in the same period of the year before the time period to be predicted; the feature data set of the first time period includes n features, and the feature data set of the second time period includes n features; the feature data set includes meteorological data, average operating years of equipment, holiday data, three-phase load data, power supply radius, distribution transformer capacity and cable cross section; A projection module is used to perform standardization processing on the matrix composed of the feature data set of the first time period, and determine the first scatter matrix of the feature data set of the first time period after the standardization processing; calculate the eigenvalues ​​of the first scatter matrix and the eigenvectors corresponding to each eigenvalue; based on the size of the eigenvalues ​​of the first scatter matrix, sort the eigenvectors corresponding to all the eigenvalues ​​from large to small, and form the first n eigenvectors into a first coordinate system; perform dimensionality reduction processing on the feature data set of the first time period after the standardization processing based on the first coordinate system to obtain a first test set; The projection module is further used to perform standardization processing on the matrix composed of the feature data set of the second time period, and determine the second scatter matrix of the feature data set of the second time period after the standardization processing; calculate the eigenvalues ​​of the second scatter matrix and the eigenvectors corresponding to each eigenvalue; based on the size of the eigenvalues ​​of the second scatter matrix, sort the eigenvectors corresponding to all the eigenvalues ​​from large to small, and form the first n eigenvectors into a second coordinate system; perform dimensionality reduction processing on the feature data set of the second time period after the standardization processing based on the second coordinate system to obtain a second test set; The prediction module is used to input the first test set, the second test set, the meteorological difference between the time period to be predicted and the second time period, the meteorological difference between the time period to be predicted and the first time period, the holiday difference between the time period to be predicted and the second time period, and the holiday difference between the time period to be predicted and the first time period into the trained line loss prediction model to predict the line loss of the distribution network.

Images