Power distribution network data processing method and system considering node space layout
Through the combination of convolutional neural network and long-term memory network, the node space characteristics of distribution network data are extracted, and the problems of data loss and delay in distribution network are solved, and more efficient and accurate data processing is achieved, suitable for high proportions The distribution network with renewable energy access.
Patent Information
- Application Number
- CN202411007901.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-15
AI Technical Summary
The loss and delay of distribution network data lead to inaccuracy of data, affecting monitoring and operational decisions. The efficiency and accuracy of traditional methods need to be improved under the conditions of high proportion of renewable energy grid connection.
The method of combining convolutional neural network and long-term memory network is used to extract the node spatial characteristics of distribution network data, perform missing value filling and outlier value detection, and build a data processing model based on spatial characteristics.
It improves the efficiency and accuracy of data processing in the distribution network, provides more reliable data support, significantly improves the stability and accuracy of data, and is suitable for distribution networks with high proportion of renewable energy access.
Smart Images

Figure CN118970898A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power data processing, and in particular relates to a distribution network data processing method and system taking into account node spatial layout. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] With the development of communication technology, the level of informatization, automation and interaction of smart distribution networks has been improved, and smart distribution networks have gradually become information-physical systems in which physical systems and information systems are deeply coupled and interdependent. A large amount of power grid data has poured into the distribution network, forming a data link at the information layer, which constitutes the data foundation for ensuring the transparency of distribution network operation. However, most advanced control strategies need to rely on reliable power grid measurement data to achieve control goals, which emphasizes the importance of accurate power grid data.
[0004] According to the inventors, the data collection, transmission, storage and use processes of distribution networks are full of challenges. Network problems, equipment failures and severe weather conditions may lead to the loss and delay of distribution network data, thereby affecting the reliability of distribution network measurement data, which in turn affects the accuracy of distribution network monitoring, the control of converter equipment and the decision-making of distribution network operators. Therefore, ensuring the accuracy of data sets is a major challenge facing distribution networks, and data preprocessing is the key to improving the reliability of power grid data and enhancing the ability to regulate distribution networks. Although existing data quality improvement methods have been able to improve data accuracy, due to the complexity of power grid topology information and the surge in the amount of smart grid data, the efficiency and accuracy of traditional methods need to be improved. Summary of the invention
[0005] To solve the above problems, the present invention proposes a distribution network data processing method and system that considers the spatial layout of nodes. In order to fully consider the impact of a high proportion of renewable energy grid connection on distribution network data, the present invention fully considers the time series and spatial characteristics of distribution network data, and combines the spatial layout of distribution network nodes to improve the efficiency and accuracy of distribution network data processing.
[0006] According to some embodiments, a first solution of the present invention provides a distribution network data processing method considering the node spatial layout, which adopts the following technical solutions:
[0007] A distribution network data processing method considering node spatial layout, comprising:
[0008] Obtain distribution network data;
[0009] The node spatial features of the acquired distribution network data are extracted through a convolutional neural network to obtain a feature matrix of the distribution network data;
[0010] The node space characteristics in the characteristic matrix of distribution network data obtained by long short-term memory network analysis are used to fill missing values and detect outliers in distribution network data, thus completing the processing of distribution network data.
[0011] As a further technical limitation, in the process of extracting the node spatial characteristics of the distribution network data, the longitude and latitude coordinates of the distribution network nodes are determined according to the node position information in the acquired distribution network data, and the adjacency matrix used to represent the connection relationship between the distribution network nodes and the distance matrix used to represent the distance between the distribution network nodes are obtained in combination with the distribution network structure diagram. A convolutional neural network is used to extract the spatial information between the distribution network nodes in the obtained adjacency matrix and distance matrix to complete the extraction of the spatial characteristics of the distribution network data nodes.
[0012] Furthermore, the characteristic matrix of the distribution network data obtained includes at least node space characteristics and node time characteristics of the distribution network, and the node time characteristics are related to the moment of the acquired distribution network data; the node time characteristics are used to represent the row elements in the characteristic matrix, and the node space characteristics are used to represent the column elements in the characteristic matrix, so as to obtain the characteristic matrix of the distribution network data.
[0013] As a further technical limitation, the node spatial characteristics of the distribution network data are related to the relationship between adjacent nodes in spatial geographical locations.
[0014] As a further technical limitation, before extracting the node spatial features, the acquired distribution network data is preprocessed, and a preliminary analysis of the distribution network data is performed through the preprocessing to reduce abnormal data and irrelevant data in the distribution network data.
[0015] Furthermore, the preprocessing includes at least standardization processing, missing value processing, outlier detection and denoising processing.
[0016] According to some embodiments, a second solution of the present invention provides a distribution network data processing system considering the node spatial layout, which adopts the following technical solution:
[0017] A distribution network data processing system considering node spatial layout, comprising:
[0018] An acquisition module, configured to acquire power distribution network data;
[0019] An extraction module is configured to extract node spatial features of the acquired distribution network data through a convolutional neural network to obtain a feature matrix of the distribution network data;
[0020] The processing module is configured to use the node space features in the feature matrix of the distribution network data obtained by long short-term memory network analysis to fill missing values and detect abnormal values of the distribution network data, so as to complete the processing of the distribution network data.
[0021] According to some embodiments, a third solution of the present invention provides a computer-readable storage medium, which adopts the following technical solution:
[0022] A computer-readable storage medium stores a program, which, when executed by a processor, implements the steps of the distribution network data processing method considering the node spatial layout as described in the first solution of the present invention.
[0023] According to some embodiments, a fourth solution of the present invention provides an electronic device, which adopts the following technical solution:
[0024] An electronic device comprises a memory, a processor and a program stored in the memory and running on the processor, wherein when the processor executes the program, the steps in the distribution network data processing method considering the node spatial layout as described in the first scheme of the present invention are implemented.
[0025] According to some embodiments, a fifth solution of the present invention provides a computer program product, which adopts the following technical solution:
[0026] A computer program product includes software codes, wherein the program in the software codes executes the steps in the distribution network data processing method considering the node spatial layout as described in the first embodiment of the present invention.
[0027] Compared with the prior art, the present invention has the following beneficial effects:
[0028] The present invention fully considers the spatial layout of distribution network nodes, constructs a data processing model based on spatial characteristics, and performs data preprocessing of the power system through the constructed model, which can realize the three data processing processes of standardization processing, missing value processing, and outlier detection; it provides reference and reliable data support for the data preprocessing of the distribution network connected to distributed photovoltaics. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] The drawings in the specification that constitute a part of this embodiment are used to provide a further understanding of this embodiment. The schematic embodiments of this embodiment and their descriptions are used to explain this embodiment and do not constitute improper limitations on this embodiment.
[0030] Figure 1 This is a flow chart of a distribution network data processing method considering node spatial layout in Embodiment 1 of the present invention;
[0031] Figure 2 Schematic diagram of the structure of the convolutional neural network in the first embodiment of the present invention;
[0032] Figure 3 Schematic diagram of the calculation principle of the spatiotemporal feature matrix in the first embodiment of the present invention;
[0033] Figure 4 Schematic diagram of the structure of the long short-term memory neural network in the first embodiment of the present invention;
[0034] Figure 5 This is a schematic diagram of a CNN+LSTM model based on spatial characteristics in Embodiment 1 of the present invention;
[0035] FIG6( a ) is a schematic diagram of the spatial structure of a distribution network in Embodiment 1 of the present invention;
[0036] FIG6( b ) is a curve diagram of distribution network data in Embodiment 1 of the present invention;
[0037] FIG. 7( a ) is a schematic diagram showing the relationship between the true value and the padded value after mean padding and the discrete relationship between the true value and the padded value after mean padding in Embodiment 1 of the present invention;
[0038] FIG7( b ) is a schematic diagram of the relationship between the real value and the padded value after CNN data padded and the discrete relationship between the real value and the padded value after CNN data padded in Embodiment 1 of the present invention;
[0039] FIG7( c ) is a schematic diagram of the relationship between the real value and the padded value after data padded using the CNN+LSTM model based on spatial characteristics and the relationship between the real value and the padded value after data padded using the CNN+LSTM model based on spatial characteristics in Embodiment 1 of the present invention;
[0040] FIG8( a ) is a schematic diagram of the 3σ outlier detection result in Embodiment 1 of the present invention;
[0041] FIG8( b ) is a schematic diagram of the detection result of an isolated forest outlier in Embodiment 1 of the present invention;
[0042] Figure 8(c) is a schematic diagram of the outlier detection results of the CNN+LSTM model based on spatial characteristics in Example 1 of the present invention. DETAILED DESCRIPTION
[0043] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0044] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.
[0045] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.
[0046] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0047] Embodiment 1
[0048] Embodiment 1 of the present invention introduces a distribution network data processing method taking into account the spatial layout of nodes.
[0049] This embodiment provides a distribution network data processing method that takes into account the spatial layout of nodes, combines the time series and spatial characteristics of distribution network data, and utilizes the spatial layout of distribution network nodes to improve the accuracy of time series data processing.
[0050] This embodiment fully considers the spatial characteristics of the distribution network with a high proportion of new energy access, and proposes an improved convolutional neural network-long short-term memory network distribution network data processing method. By constructing a data prediction model based on the improved convolutional neural network and combining the long short-term memory network to extract data time series features, the accuracy of distribution network data processing is improved.
[0051] It should be noted that this embodiment utilizes the spatial layout characteristics of the distribution network nodes, which is more advantageous than average estimation and traditional CNN methods in data filling tasks. The model also makes full use of the advantages of convolutional neural networks and long short-term memory networks to achieve better processing of time series data by combining spatial features.
[0052] In this embodiment, the Figure 1 The distribution network data processing method considering the node spatial layout shown in FIG. 1 includes the following specific steps:
[0053] Step S1: formatting the original distribution network data;
[0054] Step S2: Input the processed data into the CNN network to extract the spatial features of the distribution node data;
[0055] Step S3: Input the data processed by CNN into the LSTM layer to predict the value of the time series at the next moment;
[0056] Step S4: Denormalize to obtain the final filled data or find abnormal data.
[0057] As one or more implementation modes, in step S01, this implementation can complete data collection from various nodes of the power system through various sensors, instruments, and detection equipment. The key to data collection is to accurately and comprehensively obtain the node voltage, current, load rate and other data of the power grid for subsequent analysis and processing; effectively solve the data anomaly problems caused by unavoidable factors such as external environment, equipment abnormalities, network fluctuations, etc. during the data collection process, and improve the stability and reliability of the data.
[0058] This embodiment implements preliminary processing and analysis of the data collected by various sensor devices in the smart grid, thereby reducing the negative impact of data anomalies and irrelevant data on data analysis and ensuring the effectiveness of the initial feature matrix. In this link, data preprocessing mainly includes standardization, missing value processing, outlier detection, and denoising. Specifically,
[0059] Standardization processing includes sorting node data, adjusting the three-phase sequence of data, and unifying the format to ensure consistency between different data.
[0060] Missing value processing fills or deletes missing data. Common methods include mean filling, median filling, interpolation filling, etc., which utilize the periodic characteristics of power grid data and use the same time value of the previous day to fill in the missing data position. In traditional power systems, the reliability is high, but in the distribution network with a high proportion of new energy access, the difference in daytime data is large, and the node data is increasingly affected by weather and load, resulting in reduced reliability of interpolation filling. Mean filling often uses the mean of the data before and after the missing value to fill. When multiple consecutive data are missing, the accuracy of mean filling is insufficient.
[0061] Outlier detection is to analyze and detect outliers in the original data. Common methods include 3σ outlier detection method and isolation forest outlier detection method. However, in actual engineering, there is a problem of inaccurate detection. Data near the peak are mostly regarded as outliers, but judging from the periodicity, these peaks may not be outliers. Outliers near the mean are difficult to detect, which is insufficient in engineering applications.
[0062] It should be noted that the spatial characteristics of power grid measurement data are reflected in the relationship between adjacent nodes in terms of spatial geographical location. Generally speaking, the power grid data of a node often affects the data of its adjacent nodes. For example, in a regional photovoltaic distribution network, the weather conditions of nodes with close spatial locations will similarly affect the power of photovoltaic power generation, and the power consumed by the loads of the distribution network is more similar. Local power grid failures have a greater impact on the data measurement of adjacent nodes.
[0063] As one or more implementation modes, in step S02, the following Figure 2The convolutional neural network CNN shown is used to extract and analyze the spatial features of electrical data in the distribution network.
[0064] Convolutional neural network is a multi-layer supervised learning network, which is used to process data with grid-like structure, including time series data and image data; it consists of convolutional layer, pooling layer and fully connected layer. CNN has the characteristic of weight sharing, especially in the field of image, it has achieved good results by strengthening the local connection mode of nodes between adjacent layers in the neural network to mine the spatial local correlation information of natural images.
[0065] The CNN structure in this embodiment includes two special neural network layers: one is the convolution layer, where the input of each neuron is connected to the previous layer for local feature extraction; the other is the pooling layer, which is used for further feature extraction. The number of convolution layers and pooling layers varies in different scenarios. In this embodiment, the input feature map is first formed into a convolution layer containing many features; the features of the convolution layer are then subsampled and pooled through the pooling layer; finally, the neurons in the pooling layer are fully connected.
[0066] The voltage and current data of a node in the power grid are correlated with the data of the adjacent nodes. Since CNN can effectively process data with local structures, this embodiment uses CNN to extract the spatial features of the power grid data and construct Figure 3 The two-dimensional feature matrix containing time information and spatial information is shown, where the horizontal direction of the feature matrix is the node data sorted in time order, and the vertical direction represents the geographical location order of the nodes; the constructed feature matrix is: Among them, s i (tn) is the grid data S of node i at time tn i =[s i (tn)s i (t-(n+1)) … s i (t-1)] represents the data of node i at the previous n moments.
[0067] As one or more implementation modes, in step S03, the long short-term memory neural network is an improved model based on the recurrent neural network (RNN). The model established by the traditional RNN is difficult to handle the problem of long-term dependency, because gradient vanishing and gradient explosion are prone to occur during the training process. The basic unit of the LSTM hidden layer is a special cell structure instead of a traditional neuron node. The inflow and outflow of information and the update of the previous state are realized through the input gate, output gate, and forget gate in the cell structure. It is this special cell structure that enables LSTM to solve the problem of gradient vanishing or gradient explosion in RNN. The cell structure of LSTM is as follows: Figure 4As shown in the figure, each LSTM cell has three gates: input gate, output gate and forget gate. The calculation formulas of each gate are:
[0068] f t =σ(W f ·[h t-1 ,x t ]+b f );
[0069] i ι =σ(W i ·[h ι-1 ,x ι ]+bi);
[0070] o t =σ(W o ·[h t-1 ,x t ]+b o );
[0071] h t =o t tanh(C t );
[0072]
[0073]
[0074] Among them, σ represents the sigmoid function as the activation function; f represents the forget gate, i represents the input gate, and o t represents the output gate, C represents the cell state, represents the current input cell state, and h is the cell output.
[0075] By combining the weight sharing characteristics of CNN and the memory characteristics of LSTM, the spatiotemporal characteristics of power grid data can be extracted. The CNN model and the LSTM model are combined, such as Figure 5 As shown, by analyzing the temporal and spatial characteristics of node data, the filling of electrical missing data and outlier detection can be better achieved.
[0076] As one or more implementation modes, in step S04, the obtained spatiotemporal characteristic matrix of the distribution network is input, processed by the convolution layer, transmitted to the LSTM layer, processed by a fully connected layer, and the result is output to complete the data processing of the distribution network.
[0077] Based on the CNN-LSTM model proposed in this embodiment, outlier detection is performed, and the existing data is used to complete the training. The outlier detection is performed on the data to be detected, and the point where the difference between the true value and the next point data predicted by the model exceeds the threshold is found and marked as an outlier; the limitations of the isolation forest and 3σ detection methods in practical engineering applications are solved, and the accuracy of outlier detection is significantly improved.
[0078] Case Analysis
[0079] This example uses a North China measured data set to verify the effectiveness of the proposed method.
[0080] In combination with the spatiotemporal characteristics of the distribution network nodes, the distribution network data is processed. This embodiment uses the distribution network data of a certain area with photovoltaic power generation access as the research object; each node in the area collects 96 data every day, divided into 15 minutes as the minimum unit. The three-phase voltage data of seven nodes from December 11 to 14, 2023 are used as the research object, with a total of 4*1152 data.
[0081] This embodiment uses two evaluation indicators to measure the accuracy of the data filling model, the root mean square error RMSE and the mean absolute error MAPE as evaluation indicators, and the calculation formulas of RMSE and MAPE are respectively:
[0082]
[0083]
[0084] Where n is the sample size, y i is the ith actual value, is the i-th filling value; the smaller the RMSE and MAPE values are, the more accurate the model filling is.
[0085] Combining the node distribution of the distribution network and the data of each node, it is found that there is a certain correlation between adjacent nodes. Figure 6(a) and Figure 6(b) are the spatial structure diagram of the distribution network and the 4-day measurement data of 7 adjacent nodes in the distribution network, respectively. It can be found that there is a strong correlation between adjacent nodes, and the similarity analysis is performed on the data of these seven nodes.
[0086] The data correlation between adjacent nodes is analyzed by the Pearson correlation coefficient. The calculation formula of the Pearson correlation coefficient is: The obtained Pearson correlation coefficients are shown in Table 1.
[0087] Table 1 Pearson correlation coefficient
[0088]
[0089]
[0090] By calculating the Pearson correlation coefficient, it is found that the average value of the Pearson coefficient between adjacent nodes is 0.89, which has a strong correlation. From the table, it can be found that the Pearson correlation coefficient between adjacent nodes is larger. As the distance between nodes increases, the Pearson correlation coefficient gradually decreases. This shows that the closer the nodes are in spatial distance, the higher the correlation is, and the node data has a higher similarity.
[0091] This embodiment designs a data filling case and selects nodes containing photovoltaic power sources and typical loads as research objects, which can better reflect the characteristics of the distribution network node data and are located near the center of the distribution network and have more spatial correlation with other nodes. 16 missing points are selected from the data for data filling.
[0092] The CNN+LSTM data filling model based on spatial characteristics includes the convolution layer of the CNN model, the pooling layer of the CNN model, the network layer number setting of the LSTM model and the parameter adjustment. The CNN+LSTM data filling model based on spatial characteristics has 1 CNN convolution layer, a convolution kernel size of 3*3 matrix, 1 pooling layer, 50 LSTM neurons, and the back propagation process uses the Adam optimization algorithm. The initial learning rate is 0.005, the learning rate reduction factor is 0.1, and 1200 trainings are performed.
[0093] The CNN+LSTM data filling model based on spatial characteristics is compared with the mean filling model and the CNN data filling model. The mean filling model uses a total of 10 values before and after the missing data point as a reference, takes the mean as the filling data, and fills in from the beginning to the end in chronological order. If there is no value in the reference data, the reference is selected forward and backward.
[0094] The CNN data filling model only considers a single node for training, and the parameters are selected as much as possible to be the same as the CNN+LSTM data filling model based on spatial characteristics, where the number of CNN convolution layers is 1, the convolution kernel size is a 3*1 matrix; the number of pooling layers is 1; the initial learning rate is 0.005, the learning rate reduction factor is 0.1, and 1200 trainings are performed. The experimental results of the above three models are shown in Figure 7(a), Figure 7(b) and Figure 7(c), respectively; it can be obtained that among the three models of data filling, the real value and the filling value of the CNN+LSTM data filling based on spatial characteristics have a significantly higher fit, and the MAPE and RMSE of the three models are shown in Table 2.
[0095] Table 2 Experimental results of case 1
[0096]
[0097] It can be seen from Table 2 that the CNN+LSTM data filling model based on spatial characteristics has obvious advantages in accuracy.
[0098] This node data is selected as the research object, and the CNN+LSTM outlier detection based on spatial characteristics is compared with the 3σ outlier detection and the isolation forest outlier detection. Combined with the actual project situation, the outlier detection standard adopted by the 3σ outlier detection is: Then we think that y i is an outlier.
[0099] In the isolation forest outlier detection, the proportion of outliers is set to 5%. The test results of the two methods are shown in Figure 8(a) and Figure 8(b), and the outliers are circled in the figures. It can be seen that both the σ outlier detection results and the isolation forest outlier detection methods have defects in judging whether the peak is an outlier. Most of the data near the peak are regarded as outliers, but based on periodicity, these peaks may not be outliers. The outliers near the mean are difficult to detect, which is insufficient in engineering applications.
[0100] To solve this problem, this embodiment uses the CNN+LSTM outlier detection method based on spatial characteristics to detect outliers on the same data. The parameter settings remain unchanged, and the predicted value is compared with the true value. When the difference exceeds 0.4V, the point is considered to be an outlier point. The experimental results are shown in Figure 8(c). The results show that this method solves the problem of too many peaks being judged as outliers in outlier detection, and can detect outliers near the mean, which is more in line with the judgment criteria for abnormal data.
[0101] The CNN+LSTM model based on spatial characteristics proposed in the present invention has higher accuracy than the model that only considers temporal characteristics, and has obvious advantages in data filling and outlier processing.
[0102] Based on the CNN+LSTM model based on spatial characteristics proposed in this invention, and based on the data in actual projects, a complete data preprocessing program that can be used for power systems is built. The program better implements the three parts of standardization processing, missing value processing, and outlier detection. The program provides a reference for the data preprocessing of the distribution network connected to distributed photovoltaics, and provides more reliable data support for the next step of data processing.
[0103] Embodiment 2
[0104] Embodiment 2 of the present invention introduces a distribution network data processing system that takes into account the spatial layout of nodes.
[0105] A distribution network data processing system considering node spatial layout, comprising:
[0106] An acquisition module, configured to acquire power distribution network data;
[0107] An extraction module is configured to extract node spatial features of the acquired distribution network data through a convolutional neural network to obtain a feature matrix of the distribution network data;
[0108] The processing module is configured to use the node space features in the feature matrix of the distribution network data obtained by long short-term memory network analysis to fill missing values and detect abnormal values of the distribution network data, so as to complete the processing of the distribution network data.
[0109] The detailed steps are the same as the distribution network data processing method considering the node spatial layout provided in Example 1, and will not be repeated here.
[0110] Embodiment 3
[0111] Embodiment 3 of the present invention provides a computer-readable storage medium.
[0112] A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in the distribution network data processing method considering the node spatial layout as described in the first embodiment of the present invention.
[0113] The detailed steps are the same as the distribution network data processing method considering the node spatial layout provided in Example 1, and will not be repeated here.
[0114] Embodiment 4
[0115] A fourth embodiment of the present invention provides an electronic device.
[0116] An electronic device comprises a memory, a processor and a program stored in the memory and running on the processor, wherein when the processor executes the program, the steps in the distribution network data processing method considering the node spatial layout as described in the first embodiment of the present invention are implemented.
[0117] The detailed steps are the same as the distribution network data processing method considering the node spatial layout provided in Example 1, and will not be repeated here.
[0118] Embodiment 5
[0119] Embodiment 5 of the present invention provides a computer program product.
[0120] A computer program product includes software codes, wherein the program in the software codes executes the steps in the distribution network data processing method considering the node spatial layout as described in the first embodiment of the present invention.
[0121] The detailed steps are the same as the distribution network data processing method considering the node spatial layout provided in Example 1, and will not be repeated here.
[0122] The above description is only a preferred embodiment of the present embodiment and is not intended to limit the present embodiment. For those skilled in the art, the present embodiment may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present embodiment shall be included in the protection scope of the present embodiment.
Claims
1. A distribution network data processing method considering node spatial layout, characterized in that: include: Obtain distribution network data; The node spatial features of the acquired distribution network data are extracted through a convolutional neural network to obtain a feature matrix of the distribution network data; The node space characteristics in the characteristic matrix of distribution network data obtained by long short-term memory network analysis are used to fill missing values and detect outliers in distribution network data, thus completing the processing of distribution network data.
2. A distribution network data processing method considering node spatial layout as described in claim 1, characterized in that: In the process of extracting the node spatial features of the distribution network data, the longitude and latitude coordinates of the distribution network nodes are determined according to the node location information in the obtained distribution network data. The adjacency matrix used to represent the connection relationship between the distribution network nodes and the distance matrix used to represent the distance between the distribution network nodes are obtained in combination with the distribution network structure diagram. A convolutional neural network is used to extract the spatial information between the distribution network nodes in the obtained adjacency matrix and distance matrix to complete the extraction of the spatial features of the distribution network data nodes.
3. A distribution network data processing method considering node spatial layout as described in claim 2, characterized in that: The obtained characteristic matrix of the distribution network data includes at least the node space characteristics and node time characteristics of the distribution network, and the node time characteristics are related to the time of the obtained distribution network data; the row elements in the characteristic matrix are represented by the node time characteristics, and the column elements in the characteristic matrix are represented by the node space characteristics, so as to obtain the characteristic matrix of the distribution network data.
4. A distribution network data processing method considering node spatial layout as described in claim 1, characterized in that: The node spatial characteristics of the power distribution network data are related to the relationship between adjacent nodes in spatial geographical locations.
5. A distribution network data processing method considering node spatial layout as claimed in claim 1, characterized in that: Before extracting the node spatial features, the acquired distribution network data is preprocessed, and a preliminary analysis of the distribution network data is performed through the preprocessing to reduce abnormal data and irrelevant data in the distribution network data.
6. A distribution network data processing method considering node spatial layout as described in claim 5, characterized in that: The preprocessing includes at least standardization processing, missing value processing, outlier detection and denoising processing.
7. A distribution network data processing system considering node spatial layout, characterized in that: include: An acquisition module, configured to acquire power distribution network data; An extraction module is configured to extract node spatial features of the acquired distribution network data through a convolutional neural network to obtain a feature matrix of the distribution network data; The processing module is configured to use the node space features in the feature matrix of the distribution network data obtained by long short-term memory network analysis to fill missing values and detect abnormal values of the distribution network data, so as to complete the processing of the distribution network data.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the distribution network data processing method considering the node spatial layout as described in any one of claims 1 to 6 are implemented.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the steps of the distribution network data processing method considering the node spatial layout as described in any one of claims 1-6 are implemented.
10. A computer program product comprising software code, characterized in that The program in the software code executes the steps of the distribution network data processing method considering the node spatial layout as described in any one of claims 1-6.
Citation Information
Cited By
Topology identification method and system for low-voltage distribution network based on deep learning
CN119627914A
