A distribution network intelligent reconstruction method and system

Abstract

The present invention discloses a method and system for intelligent reconstruction of a distribution network, the method comprising: collecting historical operation data and real-time monitoring data sets of the distribution network, performing multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix; based on the high-dimensional feature matrix, parsing the topological structure of the distribution network, mining the potential association between each node and its neighboring nodes, identifying the importance of key nodes and line vulnerability information in the topological structure; using the importance of key nodes and line vulnerability information, combined with environmental information and historical load trends, using a bidirectional long short-term memory network to perform load forecasting, and generating a load distribution plan, and performing load reconstruction on the distribution network nodes according to the load distribution plan. By using the embodiments of the present invention, the accuracy of load forecasting can be improved, and efficient reconstruction of the distribution network can be achieved based on a dynamic load distribution plan, which helps to improve the intelligence, flexibility and reliability of the distribution network.

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution network, and in particular to a method and system for intelligent reconstruction of a power distribution network. Background Art

[0002] The distribution network is an important part of the power system, responsible for safely and efficiently distributing electricity from the transmission network to each user. With the development of social economy and the application of new technologies, the operating environment of the distribution network is becoming increasingly complex, facing many challenges such as load fluctuations, changes in user demand, and uncertainties brought about by the access of renewable energy. Therefore, how to improve the intelligence level of the distribution network and improve its operational flexibility and reliability has become a hot topic in the current power industry research and is also a problem that needs to be solved urgently. Summary of the invention

[0003] The purpose of the present invention is to provide a distribution network intelligent reconstruction method and system to address the deficiencies in the prior art, improve the accuracy of load forecasting, and achieve efficient reconstruction of the distribution network based on a dynamic load distribution scheme, which helps to improve the intelligence, flexibility and reliability of the distribution network.

[0004] An embodiment of the present application provides a distribution network intelligent reconstruction method, the method comprising:

[0005] Collect historical operation data and real-time monitoring data sets of the distribution network, and perform multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix;

[0006] Based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified;

[0007] By utilizing the importance of the key nodes and the line vulnerability information, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load forecasting and generate a load distribution plan, and the distribution network nodes are load reconstructed according to the load distribution plan.

[0008] Optionally, the historical operation data and real-time monitoring data set of the distribution network are collected, and the collected data set is subjected to multi-dimensional dynamic optimization processing to generate a high-dimensional feature matrix, including:

[0009] Obtain historical operation data and real-time monitoring data sets from the actual operation platform of the distribution network. The data sets contain multiple time series indicators. Each indicator is recorded based on a timestamp to form a multi-dimensional time series set.

[0010] Use dynamic time warping technology to align the timestamps of historical data and real-time data in a multidimensional time series set to form a continuous and consistent time series. Use a multivariate interpolation algorithm to complete any missing or incomplete data points in the time series to generate complete time series data.

[0011] The density peak clustering algorithm is used to identify abnormal data in time series data, and the abnormal data is smoothed by a method based on local weighted regression;

[0012] The time-varying components are extracted from the smoothed time series using wavelet packet transform, and the spectrum of the time-varying components is analyzed by short-time Fourier transform to generate time-frequency features.

[0013] The spatial correlation of distribution network nodes is captured in the topological structure through graph convolutional network, and the Chebyshev polynomial approximation is used to expand the graph Laplacian matrix to extract spatial features.

[0014] By calculating the mutual information between features and applying the recursive feature elimination strategy for feature selection to eliminate redundant features, the Gaussian mixture model is used to model the probability distribution of different feature spaces, and feature fusion is achieved through Bayesian inference to generate a unified feature representation;

[0015] Kernel ridge regression is applied to regularize the feature representation, and feature normalization is performed through logarithmic transformation and smooth maximum scaling to obtain the final high-dimensional feature matrix.

[0016] Optionally, the topological structure of the distribution network is parsed based on the high-dimensional feature matrix, potential associations between each node and its neighboring nodes are mined, and the importance of key nodes and line vulnerability information in the topological structure are identified, including:

[0017] A graph embedding network is used to map the high-dimensional feature matrix into a low-dimensional latent space to preserve the neighboring relationship and global structure information of the nodes;

[0018] Using Gaussian random field method, the potential association probability between nodes is calculated in low-dimensional latent space;

[0019] Acquire a feature vector of each node from the high-dimensional feature matrix, and calculate a potential value of each node in the topological structure of the distribution network according to the feature vector and the potential association probability;

[0020] Determining a key node in a topological structure according to the potential value, wherein the potential value of the key node indicates the degree of importance of the key node;

[0021] Based on the maximum flow minimum cut theory, the line vulnerability analysis of the distribution network is performed, the cutting edge weight of each line is calculated, and the fragile line in the topology structure is identified according to the cutting edge weight. The cutting edge weight of the fragile line represents the degree of line vulnerability. Optionally, the importance of the key nodes and the line vulnerability information are combined with environmental information and historical load trends, and a bidirectional long short-term memory network is used to perform load forecasting, and a load distribution plan is generated. According to the load distribution plan, the load of the distribution network nodes is reconfigured, including:

[0022] The importance of the key nodes and the line vulnerability information are combined with the environmental characteristics and the historical load trends to construct a new input feature matrix;

[0023] The new input feature matrix is ​​input into the pre-trained bidirectional long short-term memory network, and the multi-step predicted load value is output through the network's time step aggregation;

[0024] Based on the multi-step predicted load value and combined with the physical constraints of the power grid, a linear programming model is established, wherein the objective function of the linear programming model is to minimize the global load deviation and optimize the load balance of key nodes;

[0025] The linear programming model is solved by using an optimization solver to generate an optimal load distribution plan, and the distribution network is reconstructed based on the optimal load distribution plan.

[0026] Another embodiment of the present application provides a distribution network intelligent reconstruction system, the system comprising:

[0027] The acquisition module is used to collect historical operation data and real-time monitoring data sets of the distribution network, and perform multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix;

[0028] A parsing module, used to parse the topological structure of the distribution network based on the high-dimensional feature matrix, to mine the potential association between each node and its neighboring nodes, and to identify the importance of key nodes and line vulnerability information in the topological structure;

[0029] The reconstruction module is used to utilize the importance of the key nodes and the line vulnerability information, combine the environmental information and the historical load trend, use the bidirectional long short-term memory network to perform load forecasting, and generate a load distribution plan, and reconstruct the load of the distribution network nodes according to the load distribution plan.

[0030] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when running.

[0031] Yet another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described above.

[0032] Compared with the prior art, the present invention provides a distribution network intelligent reconstruction method, which collects historical operation data and real-time monitoring data sets of the distribution network, performs multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix; based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its adjacent nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified; the importance of key nodes and line vulnerability information are used, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load prediction, and a load distribution plan is generated, and the load of the distribution network nodes is reconstructed according to the load distribution plan, thereby improving the accuracy of load prediction, and realizing efficient reconstruction of the distribution network based on the dynamic load distribution plan, which is helpful to improve the intelligence, flexibility and reliability of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A hardware structure block diagram of a computer terminal for a distribution network intelligent reconstruction method provided by an embodiment of the present invention;

[0034] Figure 2 A schematic diagram of a flow chart of a distribution network intelligent reconstruction method provided by an embodiment of the present invention;

[0035] Figure 3 A schematic diagram of the structure of a distribution network intelligent reconstruction system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0037] The embodiment of the present invention firstly provides a method for intelligent reconstruction of a distribution network, which can be applied to electronic devices, such as computer terminals, specifically ordinary computers, etc.

[0038] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal of a distribution network intelligent reconstruction method provided by an embodiment of the present invention. Figure 1 As shown, the computer terminal may include one or more ( Figure 1Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the computer terminal may also include a transmission device 106 for communication functions and an input and output device 108. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0039] The memory 104 can be used to store software programs and modules of application software, such as program instructions / modules corresponding to the distribution network intelligent reconstruction method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0040] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0041] See also Figure 2 The embodiment of the present invention provides a method for intelligent reconstruction of a distribution network, which may include the following steps:

[0042] S201, collecting historical operation data and real-time monitoring data sets of the distribution network, and performing multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix;

[0043] By collecting and organizing the historical operation data and real-time monitoring data of the distribution network, a multi-dimensional data set is constructed. These data sets can include multiple parameters such as voltage, current, load, ambient temperature, etc., and through dynamic optimization processing, the original data is converted into a high-dimensional feature matrix for subsequent analysis and calculation. The generation of a high-dimensional feature matrix lays the foundation for the subsequent topology analysis and load prediction. Through a multi-dimensional data set, the state of the distribution network can be fully reflected, thereby providing more sufficient information for subsequent models. Especially in complex distribution networks, the construction of this feature matrix can improve the system's ability to identify abnormal situations and predict future load changes, thereby improving the safety and stability of power supply.

[0044] Specifically, historical operation data and real-time monitoring data sets can be obtained from the actual operation platform of the distribution network. The data set contains multiple time series indicators, each of which is recorded based on a timestamp to form a multi-dimensional time series set;

[0045] This step is the first step of data collection, which involves obtaining various indicator data related to the operation of the distribution network from the actual operation platform. These data are usually marked with timestamps, which means that each indicator has a corresponding time record, thus forming a complete time series set. Through the record of time series, the operation characteristics and change patterns of the distribution network in different time periods can be effectively captured. This process can help with subsequent data processing and analysis, especially when identifying and evaluating the load changes of the distribution network and its impact on power supply, time series data is crucial.

[0046] The distribution network monitoring system can be used to link the database and write query statements to filter out the required historical and real-time monitoring indicators. At the same time, data acquisition equipment is used to ensure the accuracy and timeliness of the data. The data is stored in CSV or database format for subsequent processing.

[0047] Use dynamic time warping technology to align the timestamps of historical data and real-time data in a multidimensional time series set to form a continuous and consistent time series. Use a multivariate interpolation algorithm to complete any missing or incomplete data points in the time series to generate complete time series data.

[0048] Dynamic Time Warping (DTW) technology is an effective method for aligning time series, which can handle time delays or inconsistencies caused by different sampling frequencies. Completing missing data is a key step in ensuring the integrity of the data set. Through multivariate interpolation algorithms, missing values ​​can be estimated to construct consistent time series data. By aligning different time series, errors caused by time differences can be eliminated and the availability and consistency of data can be improved. This process ensures that all data are analyzed under a unified time base in subsequent analyses, which helps to improve the accuracy and effectiveness of subsequent algorithms.

[0049] The DTW algorithm can be implemented in a feature extraction tool, and time series alignment can be performed using a programming language such as Python. Subsequently, multivariate interpolation methods such as Lagrange interpolation or spline interpolation are used to handle missing values ​​in the data and generate complete time series data.

[0050] The density peak clustering algorithm is used to identify abnormal data in time series data, and the abnormal data is smoothed by a method based on local weighted regression;

[0051] The density peak clustering algorithm can effectively identify abnormal values ​​or outliers in the data set. The identified abnormal data may affect the subsequent analysis, so it is necessary to smooth it through local weighted regression to reduce the impact of these abnormal values ​​on the overall data. By identifying and processing abnormal data, the accuracy and reliability of the data can be ensured, and misleading analysis caused by abnormal data can be reduced. This process lays a good data foundation for subsequent feature extraction and analysis.

[0052] Clustering software (such as Sci-kit Learn) can be used to implement density peak clustering, and then the identified abnormal data can be smoothed using a local weighted regression model (such as LOESS). The smoothed time series is output in preparation for the subsequent steps. From the smoothed time series, the time-varying components are extracted using wavelet packet transform, and the time-varying components are spectrally analyzed using short-time Fourier transform to generate time-frequency features;

[0053] Wavelet packet transform is an effective time-frequency analysis tool that can decompose signals and extract time-varying features. Through short-time Fourier transform, spectrum analysis can be performed on specific time windows to further enhance the representation of time series features. The extraction of time-frequency features can reveal potential patterns and changing trends in the data, providing richer information for subsequent data modeling and learning. The acquisition of such complex features helps to improve the accuracy and robustness of load forecasting.

[0054] You can use libraries such as MATLAB or Python's NumPy and SciPy to implement wavelet packet transform and short-time Fourier transform to analyze the smoothed time series data, extract time-frequency features and generate the corresponding feature matrix.

[0055] The spatial correlation of distribution network nodes is captured in the topological structure through graph convolutional network, and the Chebyshev polynomial approximation is used to expand the graph Laplacian matrix to extract spatial features.

[0056] Graph convolutional networks (GCNs) are highly efficient in processing graph-structured data and are particularly suitable for capturing spatial relationships between nodes. By using Chebyshev polynomials to perform feature expansion on the graph Laplacian matrix, the spatial features of the nodes can be effectively extracted, enhancing the model's ability to learn graph data. By sorting out the relationships between nodes in the distribution network, key nodes and key lines can be better identified, providing a direct basis for subsequent analysis and optimization. Effectively extracting spatial features helps improve the optimization and decision-making capabilities of the distribution network.

[0057] The graph convolutional network can be implemented using a deep learning framework (such as PyTorch Geometric or TensorFlow), and the Chebyshev polynomial method can be integrated into the model to perform feature learning and spatial feature extraction of the graph Laplacian matrix.

[0058] By calculating the mutual information between features and applying the recursive feature elimination strategy for feature selection to eliminate redundant features, the Gaussian mixture model is used to model the probability distribution of different feature spaces, and feature fusion is achieved through Bayesian inference to generate a unified feature representation;

[0059] Mutual information calculation is used to evaluate the correlation between features. Through the recursive feature elimination strategy, it can effectively filter out important features from high-dimensional features and reduce the complexity of the model. At the same time, the Gaussian mixture model can provide modeling for the probability distribution of the feature space, and Bayesian inference is used to integrate these features to form a unified representation. The process of feature selection and fusion helps to improve the performance of the model, reduce the risk of overfitting, and improve the generalization ability of the system. By streamlining the features, we can focus more on the important factors that affect load forecasting and improve the interpretability of the model.

[0060] Machine learning tools (such as Scikit-learn) can be used for mutual information and recursive feature elimination, data modeling can be performed by implementing Gaussian mixture models, and various features can be integrated with Bayesian inference to finally form a compact feature representation for subsequent use.

[0061] Kernel ridge regression is applied to regularize the feature representation, and feature normalization is performed through logarithmic transformation and smooth maximum scaling to obtain the final high-dimensional feature matrix.

[0062] Kernel ridge regression is a regularization technique that can effectively improve the stability and generalization ability of the model. By performing logarithmic transformation and smooth maximum scaling on the features, all features can be learned at a uniform scale to avoid unstable model training due to eigenvalue differences. This step ensures the quality and consistency of the feature matrix, providing a solid foundation for subsequent model training. Standardized data helps accelerate model convergence and improve prediction accuracy and reliability.

[0063] In the feature processing stage, the Scikit-learn library can be used to implement kernel ridge regression, followed by logarithmic transformation through Numpy and smooth maximum scaling to generate the final high-dimensional feature matrix, ready to be input into subsequent algorithms.

[0064] S202, based on the high-dimensional feature matrix, analyzing the topological structure of the distribution network, mining the potential association between each node and its neighboring nodes, and identifying the importance of key nodes and line vulnerability information in the topological structure;

[0065] This step analyzes the high-dimensional feature matrix to gain an in-depth understanding of the relationship between each node in the distribution network. The high-dimensional feature matrix contains a variety of power operation-related data, which can reveal the role of each node in the power grid and their connectivity and interaction with adjacent nodes. By identifying key nodes, it is possible to determine which nodes have the greatest impact on the stability and reliability of the power grid, while line vulnerability information helps to identify potential fault points or weak links so that measures can be taken in advance. The main significance of this step is to enhance the safety and reliability of the distribution network. By understanding the importance of nodes and weak links in the network, more effective maintenance plans and operation strategies can be formulated to reduce the probability of failures and improve the continuity and safety of power supply. This provides a reliable data foundation for the intelligent transformation of the power grid and helps to achieve efficient management and scheduling of the power grid.

[0066] Specifically, a graph embedding network may be used to map the high-dimensional feature matrix into a low-dimensional latent space to preserve the neighboring relationship and global structural information of the nodes;

[0067] Graph embedding network is a technology that maps graph structured data to low-dimensional space. By maintaining the proximity relationship and global structural information between nodes, subsequent analysis is more efficient and convenient. High-dimensional feature matrices can better represent node relationships through graph embedding, which helps to extract important structural features. By mapping high-dimensional features to low-dimensional space, computational complexity can be reduced and the training and reasoning speed of the model can be improved. At the same time, preserving the relationship between nodes helps capture the structural information of the graph, which is crucial for subsequent potential association analysis and feature extraction.

[0068] Graph embedding methods such as DeepWalk or Node2Vec can be used to take the high-dimensional feature matrix as input, generate node sequences through techniques such as random walks, and then train them through the Skip-Gram or CBOW model to obtain the representation of the nodes in the low-dimensional space.

[0069] Using Gaussian random field method, the potential association probability between nodes is calculated in low-dimensional latent space;

[0070] Gaussian random fields are a method for modeling spatial data, suitable for capturing the interdependence between nodes in a low-dimensional latent space. In this way, the potential association probability between each node can be calculated to understand the degree of mutual influence between nodes. This step helps to build a relationship model between nodes in the distribution network, revealing the importance and interaction of each node in the topology. The calculation of this association probability provides a quantitative basis for subsequent key node identification and line vulnerability analysis.

[0071] The feature vector of the node can be set in the low-dimensional space, and the maximum likelihood estimation method of Gaussian random field can be used to calculate the potential association probability by using the similarity between node features, by constructing the covariance matrix and solving the corresponding probability distribution.

[0072] Acquire a feature vector of each node from the high-dimensional feature matrix, and calculate a potential value of each node in the topological structure of the distribution network according to the feature vector and the potential association probability;

[0073] In this process, the feature vector of each node is first extracted from the high-dimensional feature matrix. Then, the potential value of each node is calculated using the potential association probability calculated previously and the node feature vector. This potential value reflects the importance and influence of the node in the distribution network. The calculated potential value enables the effective identification of key nodes in the distribution network, so that these nodes can be focused on in the management and dispatch of the power grid, thereby improving the operational safety and reliability of the power grid.

[0074] The potential value of a node can be calculated by weighted summation based on the node's feature vector and potential association probability. For example, the potential value can be defined as the weighted average of the association probability of the node's feature vector and its neighboring nodes, which can comprehensively consider the characteristics of the node itself and its relationship with its neighboring nodes.

[0075] According to the potential value, the key nodes in the topological structure are determined, and the potential value of the key node indicates the degree of importance of the key node; in this step, the calculated potential value is used for threshold determination to identify the most critical nodes in the distribution network. The definition of the key node depends on the level of its potential value. The higher the potential value of the node, the greater its impact on the operation of the entire power grid. Identifying key nodes helps to prioritize monitoring and maintenance of these nodes when resources are limited, so as to improve the stability and risk resistance of the power grid. By effectively managing key nodes, the risk of failures and the impact on the operation of the power grid can be reduced.

[0076] An appropriate threshold can be set, such as selecting the top 10% of nodes in potential value as key nodes. By comparing the potential values, these nodes are identified using a sorting algorithm and marked for subsequent analysis and processing.

[0077] Based on the maximum flow minimum cut theory, the line vulnerability analysis of the distribution network is carried out, the cutting edge weight of each line is calculated, and the fragile lines in the topological structure are identified according to the cutting edge weight. The cutting edge weight of the fragile line represents the degree of line vulnerability.

[0078] The maximum flow and minimum cut theory provides an important analytical framework for network flow. By solving the relationship between the flow and the cut edge in the network, the vulnerable lines in the distribution network can be identified. By calculating the cut edge weight of each line, the vulnerability of the line can be evaluated. Analyzing the vulnerability of the line can help the power dispatching center and the operation and maintenance team identify potential fault points in advance, so that these lines can be given priority during maintenance and repair, thereby preventing faults and ensuring the stability of power supply.

[0079] The distribution network can be regarded as a flow network, with source and sink points set, and the maximum flow calculated by network flow algorithms (such as Ford-Folkerson algorithm), and the weight of each line is further calculated using the cutting theorem. Lines with smaller weights are regarded as vulnerable lines, and corresponding maintenance decisions are made.

[0080] S203, using the importance of the key nodes and the line vulnerability information, combined with environmental information and historical load trends, using a bidirectional long short-term memory network to perform load forecasting, and generate a load distribution plan, and reconstruct the load of the distribution network nodes according to the load distribution plan.

[0081] The core of this step is to combine the key nodes identified in the distribution network and their vulnerability information with external environmental characteristics (such as meteorological and seasonal changes) and historical load trends to construct new feature inputs. Through the bidirectional long short-term memory network (Bi-LSTM), these features are analyzed in time series to achieve accurate prediction of future loads. The generated load distribution scheme can effectively reconstruct node loads based on the prediction results, thereby optimizing the load distribution of the distribution network.

[0082] This step effectively utilizes information inside and outside the system, enhances the accuracy of load forecasting, and helps achieve smarter grid dispatching and load management. By reasonably reconfiguring the loads of distribution network nodes, the reliability and efficiency of power supply can be improved, and the operational risks of the power grid under extreme conditions can be reduced. At the same time, special treatment of key nodes and vulnerable lines helps maintain the stability of the power system and ensure the safety and sustainability of power supply.

[0083] Specifically, the importance of the key nodes and the line vulnerability information can be combined with the environmental characteristics and historical load trends to construct a new input feature matrix;

[0084] This step involves integrating various key information, such as the importance of key nodes, line vulnerability, environmental characteristics (such as real-time monitoring data such as temperature and humidity), and historical load trends (such as time series of historical load data) to form a new input feature matrix. This matrix will provide comprehensive background information for subsequent load forecasting. By constructing a comprehensive input feature matrix, the actual operation status of the power grid can be more comprehensively reflected, ensuring that the bidirectional long short-term memory network takes into account various influencing factors when making load forecasts, thereby improving the accuracy and reliability of the forecast.

[0085] The importance scores of key nodes, line vulnerability indicators, environmental characteristics and historical load trend data can be extracted from the database, and data preprocessing techniques (such as normalization and standardization) can be used to integrate these data into a unified matrix structure for subsequent input into the prediction model.

[0086] The new input feature matrix is ​​input into the pre-trained bidirectional long short-term memory network, and the multi-step predicted load value is output through the network's time step aggregation;

[0087] In this step, the constructed input feature matrix is ​​input as data into the bidirectional long short-term memory network. The model will use its stored historical information and bidirectional structure to process time series data for multi-step prediction. The prediction result will be the power grid load value at multiple time points in the future. The use of a bidirectional long short-term memory network can better capture the long-term and short-term dependencies in the time series. The output multi-step predicted load value can provide a prediction basis for power grid managers, helping them to prepare load scheduling and resource allocation in advance, and improve the operation efficiency of the power grid.

[0088] The input feature matrix can be input into the bidirectional LSTM model in the format of a time series, and the model's training parameters can be used for forward and backward propagation calculations to ultimately output load forecast values ​​for multiple time steps. The training process can be optimized through appropriate activation functions (such as ReLU) and loss functions (such as mean square error).

[0089] Based on the multi-step predicted load value and combined with the physical constraints of the power grid, a linear programming model is established, wherein the objective function of the linear programming model is to minimize the global load deviation and optimize the load balance of key nodes;

[0090] The predicted load value is combined with the constraints of the power grid (such as line capacity, maximum load of nodes, etc.) to construct a linear programming model. The goal of this model is to minimize the deviation of the overall load of the power grid by optimizing load distribution, while taking into account the load balance of key nodes to ensure the stability and safety of power grid operation. By constructing a linear programming model, the load of the power grid can be systematically managed to avoid the risk of line overload or failure caused by load imbalance. This method effectively improves the resource allocation efficiency of the power grid, allowing the power grid to maintain a stable operating state under load fluctuations.

[0091] The objective function and constraints of linear programming can be determined, a mathematical model can be constructed, and an optimization solver (such as the simplex method or the interior point method) can be used to solve the problem to obtain the optimal load distribution solution. It is ensured that the output solution can achieve effective load reconstruction within the physical constraints.

[0092] The linear programming model is solved by using an optimization solver to generate an optimal load distribution plan, and the distribution network is reconstructed based on the optimal load distribution plan.

[0093] The optimization solver is used to solve the linear programming model established above to obtain the optimal load distribution plan. This plan will guide the load reconstruction of the distribution network nodes and optimize the power distribution. By solving the optimal load distribution plan, it can ensure that the load distribution of the power grid is more reasonable during operation, thereby improving the overall efficiency and stability of the power grid. An effective load reconstruction mechanism can reduce power losses and equipment failures caused by uneven loads and enhance the reliability of the power grid.

[0094] A suitable linear programming solver (such as CPLEX or Gurobi) can be selected to input the established linear programming model for solving and obtain the optimal load distribution plan. Subsequently, the load distribution of each node in the distribution network is adjusted according to the plan and the load reconstruction strategy is implemented.

[0095] It can be seen that the historical operation data and real-time monitoring data sets of the distribution network are collected, and the collected data sets are dynamically optimized in multiple dimensions to generate a high-dimensional feature matrix; based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified; using the importance of key nodes and line vulnerability information, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load forecasting, and a load distribution plan is generated, and the load of the distribution network nodes is reconstructed according to the load distribution plan, thereby improving the accuracy of load forecasting, and realizing efficient reconstruction of the distribution network based on the dynamic load distribution plan, which helps to improve the intelligence, flexibility and reliability of the distribution network.

[0096] Another embodiment of the present invention provides a distribution network intelligent reconstruction system, see Figure 3 , the system may include:

[0097] The acquisition module 301 is used to collect historical operation data and real-time monitoring data sets of the distribution network, and perform multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix;

[0098] The parsing module 302 is used to parse the topological structure of the distribution network based on the high-dimensional feature matrix, explore the potential association between each node and its neighboring nodes, and identify the importance of key nodes and line vulnerability information in the topological structure;

[0099] The reconstruction module 303 is used to utilize the importance of the key nodes and the line vulnerability information, combined with the environmental information and the historical load trend, to use the bidirectional long short-term memory network to perform load forecasting, and to generate a load distribution plan, and to perform load reconstruction on the distribution network nodes according to the load distribution plan.

[0100] It can be seen that the historical operation data and real-time monitoring data sets of the distribution network are collected, and the collected data sets are dynamically optimized in multiple dimensions to generate a high-dimensional feature matrix; based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified; using the importance of key nodes and line vulnerability information, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load forecasting, and a load distribution plan is generated, and the load of the distribution network nodes is reconstructed according to the load distribution plan, thereby improving the accuracy of load forecasting, and realizing efficient reconstruction of the distribution network based on the dynamic load distribution plan, which helps to improve the intelligence, flexibility and reliability of the distribution network.

[0101] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0102] Specifically, in this embodiment, the above storage medium may be configured to store a computer program for performing the following steps:

[0103] S201, collecting historical operation data and real-time monitoring data sets of the distribution network, and performing multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix;

[0104] S202, based on the high-dimensional feature matrix, analyzing the topological structure of the distribution network, mining the potential association between each node and its neighboring nodes, and identifying the importance of key nodes and line vulnerability information in the topological structure;

[0105] S203, using the importance of the key nodes and the line vulnerability information, combined with environmental information and historical load trends, using a bidirectional long short-term memory network to perform load forecasting, and generate a load distribution plan, and reconstruct the load of the distribution network nodes according to the load distribution plan.

[0106] It can be seen that the historical operation data and real-time monitoring data sets of the distribution network are collected, and the collected data sets are dynamically optimized in multiple dimensions to generate a high-dimensional feature matrix; based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified; using the importance of key nodes and line vulnerability information, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load forecasting, and a load distribution plan is generated, and the load of the distribution network nodes is reconstructed according to the load distribution plan, thereby improving the accuracy of load forecasting, and realizing efficient reconstruction of the distribution network based on the dynamic load distribution plan, which helps to improve the intelligence, flexibility and reliability of the distribution network.

[0107] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0108] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0109] Specifically, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0110] S201, collecting historical operation data and real-time monitoring data sets of the distribution network, and performing multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix;

[0111] S202, based on the high-dimensional feature matrix, analyzing the topological structure of the distribution network, mining the potential association between each node and its neighboring nodes, and identifying the importance of key nodes and line vulnerability information in the topological structure;

[0112] S203, using the importance of the key nodes and the line vulnerability information, combined with environmental information and historical load trends, using a bidirectional long short-term memory network to perform load forecasting, and generate a load distribution plan, and reconstruct the load of the distribution network nodes according to the load distribution plan.

[0113] It can be seen that the historical operation data and real-time monitoring data sets of the distribution network are collected, and the collected data sets are dynamically optimized in multiple dimensions to generate a high-dimensional feature matrix; based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified; using the importance of key nodes and line vulnerability information, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load forecasting, and a load distribution plan is generated, and the load of the distribution network nodes is reconstructed according to the load distribution plan, thereby improving the accuracy of load forecasting, and realizing efficient reconstruction of the distribution network based on the dynamic load distribution plan, which helps to improve the intelligence, flexibility and reliability of the distribution network.

[0114] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the drawings. Any changes made according to the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the protection scope of the present invention.

Claims

1. A distribution network intelligent reconstruction method, characterized in that: The method comprises: Collecting historical operation data and real-time monitoring data sets of the distribution network, and performing multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix; the historical operation data and real-time monitoring data sets of the distribution network are collected, and the collected data sets are performed multi-dimensional dynamic optimization processing to generate a high-dimensional feature matrix, including: Obtain historical operation data and real-time monitoring data sets from the actual operation platform of the distribution network. The data sets contain multiple time series indicators. Each indicator is recorded based on a timestamp to form a multi-dimensional time series set. Use dynamic time warping technology to align the timestamps of historical data and real-time data in a multidimensional time series set to form a continuous and consistent time series. Use a multivariate interpolation algorithm to complete any missing or incomplete data points in the time series to generate complete time series data. The density peak clustering algorithm is used to identify abnormal data in time series data, and the abnormal data is smoothed by a method based on local weighted regression; The time-varying components are extracted from the smoothed time series using wavelet packet transform, and the spectrum of the time-varying components is analyzed by short-time Fourier transform to generate time-frequency features. The spatial correlation of distribution network nodes is captured in the topological structure through graph convolutional network, and the Chebyshev polynomial approximation is used to expand the graph Laplacian matrix to extract spatial features. By calculating the mutual information between features and applying the recursive feature elimination strategy for feature selection to eliminate redundant features, the Gaussian mixture model is used to model the probability distribution of different feature spaces, and feature fusion is achieved through Bayesian inference to generate a unified feature representation; Applying kernel ridge regression to regularize the feature representation, and performing feature standardization through logarithmic transformation and smooth maximum scaling to obtain a final high-dimensional feature matrix; Based on the high-dimensional feature matrix, the topological structure of the distribution network is analyzed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified; By utilizing the importance of the key nodes and the line vulnerability information, combined with environmental information and historical load trends, a bidirectional long short-term memory network is used to perform load forecasting and generate a load distribution plan, and the distribution network nodes are load reconstructed according to the load distribution plan.

2. The method according to claim 1, characterized in that: Based on the high-dimensional feature matrix, the topological structure of the distribution network is parsed, the potential association between each node and its neighboring nodes is mined, and the importance of key nodes and line vulnerability information in the topological structure are identified, including: A graph embedding network is used to map the high-dimensional feature matrix into a low-dimensional latent space to preserve the neighboring relationship and global structure information of the nodes; Using Gaussian random field method, the potential association probability between nodes is calculated in low-dimensional latent space; Acquire a feature vector of each node from the high-dimensional feature matrix, and calculate a potential value of each node in the topological structure of the distribution network according to the feature vector and the potential association probability; Determining a key node in a topological structure according to the potential value, wherein the potential value of the key node indicates the degree of importance of the key node; Based on the maximum flow minimum cut theory, the line vulnerability analysis of the distribution network is carried out, the cutting edge weight of each line is calculated, and the fragile lines in the topological structure are identified according to the cutting edge weight. The cutting edge weight of the fragile line represents the degree of line vulnerability.

3. The method according to claim 2, characterized in that The method utilizes the importance of the key nodes and the line vulnerability information, combines environmental information and historical load trends, uses a bidirectional long short-term memory network to perform load forecasting, and generates a load distribution plan, and performs load reconstruction on the distribution network nodes according to the load distribution plan, including: The importance of the key nodes and the line vulnerability information are combined with the environmental characteristics and the historical load trends to construct a new input feature matrix; The new input feature matrix is ​​input into the pre-trained bidirectional long short-term memory network, and the multi-step predicted load value is output through the network's time step aggregation; Based on the multi-step predicted load value and combined with the physical constraints of the power grid, a linear programming model is established, wherein the objective function of the linear programming model is to minimize the global load deviation and optimize the load balance of key nodes; The linear programming model is solved by using an optimization solver to generate an optimal load distribution plan, and the distribution network is reconstructed based on the optimal load distribution plan.

4. A distribution network intelligent reconstruction system, characterized in that: The system comprises: The acquisition module is used to collect historical operation data and real-time monitoring data sets of the distribution network, and perform multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix; the acquisition module is used to collect historical operation data and real-time monitoring data sets of the distribution network, and perform multi-dimensional dynamic optimization processing on the collected data sets to generate a high-dimensional feature matrix, including: Obtain historical operation data and real-time monitoring data sets from the actual operation platform of the distribution network. The data sets contain multiple time series indicators. Each indicator is recorded based on a timestamp to form a multi-dimensional time series set. Use dynamic time warping technology to align the timestamps of historical data and real-time data in a multidimensional time series set to form a continuous and consistent time series. Use a multivariate interpolation algorithm to complete any missing or incomplete data points in the time series to generate complete time series data. The density peak clustering algorithm is used to identify abnormal data in time series data, and the abnormal data is smoothed by a method based on local weighted regression; The time-varying components are extracted from the smoothed time series using wavelet packet transform, and the spectrum of the time-varying components is analyzed by short-time Fourier transform to generate time-frequency features. The spatial correlation of distribution network nodes is captured in the topological structure through graph convolutional network, and the Chebyshev polynomial approximation is used to expand the graph Laplacian matrix to extract spatial features. By calculating the mutual information between features and applying the recursive feature elimination strategy for feature selection to eliminate redundant features, the Gaussian mixture model is used to model the probability distribution of different feature spaces, and feature fusion is achieved through Bayesian inference to generate a unified feature representation; Applying kernel ridge regression to regularize the feature representation, and performing feature standardization through logarithmic transformation and smooth maximum scaling to obtain a final high-dimensional feature matrix; A parsing module, used to parse the topological structure of the distribution network based on the high-dimensional feature matrix, to mine the potential association between each node and its neighboring nodes, and to identify the importance of key nodes and line vulnerability information in the topological structure; The reconstruction module is used to utilize the importance of the key nodes and the line vulnerability information, combine the environmental information and the historical load trend, use the bidirectional long short-term memory network to perform load forecasting, and generate a load distribution plan, and reconstruct the load of the distribution network nodes according to the load distribution plan.

5. The system according to claim 4, characterized in that The analysis module is specifically used for: A graph embedding network is used to map the high-dimensional feature matrix into a low-dimensional latent space to preserve the neighboring relationship and global structure information of the nodes; Using Gaussian random field method, the potential association probability between nodes is calculated in low-dimensional latent space; Acquire a feature vector of each node from the high-dimensional feature matrix, and calculate a potential value of each node in the topological structure of the distribution network according to the feature vector and the potential association probability; Determining a key node in a topological structure according to the potential value, wherein the potential value of the key node indicates the degree of importance of the key node; Based on the maximum flow minimum cut theory, the line vulnerability analysis of the distribution network is carried out, the cutting edge weight of each line is calculated, and the fragile lines in the topological structure are identified according to the cutting edge weight. The cutting edge weight of the fragile line represents the degree of line vulnerability.

6. The system according to claim 5, characterized in that The reconstruction module is specifically used for: The importance of the key nodes and the line vulnerability information are combined with the environmental characteristics and the historical load trends to construct a new input feature matrix; The new input feature matrix is ​​input into the pre-trained bidirectional long short-term memory network, and the multi-step predicted load value is output through the network's time step aggregation; Based on the multi-step predicted load value and combined with the physical constraints of the power grid, a linear programming model is established, wherein the objective function of the linear programming model is to minimize the global load deviation and optimize the load balance of key nodes; The linear programming model is solved by using an optimization solver to generate an optimal load distribution plan, and the distribution network is reconstructed based on the optimal load distribution plan.

7. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 3 when executed.

8. An electronic device, comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 3.

Images