A method for determining medium voltage distribution network access based on load peak-valley coupling

Abstract

The present invention proposes a method for determining access to a medium-voltage distribution network based on load peak-valley coupling, including: collecting load data of the medium-voltage distribution network and preprocessing them separately; collecting load data of the medium-voltage distribution network and preprocessing them separately using time series as input; constructing a dynamic load coupling access point optimization model to handle dynamic load changes in the medium-voltage distribution network; constructing a load-topology adaptive optimization model in the final access point selection; constructing an adaptive dynamic load-topology adjustment model according to the real-time load fluctuation and topological structure adjustment of the distribution network; long-term optimization and system evaluation, while combining the load peak-valley fluctuation characteristics of the medium-voltage distribution network, and through long-term evaluation and feedback optimization, load management and energy transmission efficiency are performed. The present invention not only solves the problems of inaccurate load prediction, inflexible access point selection and unstable power grid operation in the prior art, but also provides effective technical support for the construction and optimization of future smart grids.

Description

Technical Field

[0001] The present invention belongs to the technical field of load peak-valley coupling, and in particular relates to a method for determining access to a medium-voltage distribution network based on load peak-valley coupling. Background Art

[0002] With the continuous development of the power system and the increasing load demand of the power grid, the importance of medium-voltage distribution network in power transmission and distribution has become increasingly prominent. The medium-voltage distribution network connects the high-voltage transmission network with the low-voltage distribution network, and is responsible for the key task of transmitting electricity from the main part of the power grid to the end user. However, with the large-scale access of renewable energy (such as wind power and solar energy) and the development of distributed power generation systems, the traditional medium-voltage distribution network faces a series of new challenges. First, these renewable energy sources are intermittent and unstable, resulting in a significant increase in the peak and valley differences in the power grid load, especially during periods of large fluctuations in power demand. The peak and valley values ​​of the power grid load change dramatically, causing unbalanced power transmission. In addition, the traditional design and operation of distribution networks are mostly based on fixed load forecasting and static optimization strategies. This method shows great limitations when the load changes dramatically, and it is difficult to respond to dynamically changing load demands in real time.

[0003] Existing load forecasting methods and access optimization strategies mainly rely on linear regression and time series models of historical data, with low prediction accuracy, especially when the load changes are complex, the error is large, and it is difficult to accurately predict future load peaks and valleys. On the other hand, existing access point optimization methods are usually based on preset rules or fixed optimization models, which are difficult to adjust flexibly and cannot adapt to real-time changes in loads and diversified energy access needs. These methods usually do not consider the actual topological structure of the power grid and its dynamic characteristics, resulting in unreasonable access point selection and power distribution, increasing the energy loss of power grid operation and the load pressure of equipment. In addition, traditional methods lack effective control over power grid stability and voltage fluctuations, especially under extreme conditions of peak load and valley load, which can easily lead to voltage instability and power grid safety risks.

[0004] Therefore, the existing technology has significant deficiencies in coping with load peak-valley changes, optimizing access point selection, and maintaining grid stability. The present invention aims to overcome the above technical defects and achieve more efficient grid operation and more stable power supply through a method for determining medium-voltage distribution network access based on load peak-valley coupling, combined with advanced prediction algorithms and topology optimization technology. Summary of the invention

[0005] The purpose of the present invention is to design a method for determining medium-voltage distribution network access based on load peak-valley coupling, which can not only accurately predict the load peak-valley changes of the power grid, but also flexibly optimize the access points based on the prediction results, and consider the dynamic characteristics of the power grid topology to maximize the utilization efficiency of power resources and the overall stability of the power grid.

[0006] In order to achieve the above object, the present invention provides a method for determining medium voltage distribution network access based on load peak-valley coupling, the method comprising the following steps:

[0007] The load data of the centralized pressure distribution network is preprocessed respectively, and the preprocessed data is feature expanded, the expanded features are combined into a data set and a time series is constructed to obtain time series data containing load data features; wherein the load data includes load power, voltage and external factors at each time point;

[0008] A load peak-valley prediction model based on a multi-scale dependency memory network is constructed. The time series is used as input, and load peak-valley prediction is performed by simultaneously processing load changes at different time scales. The multi-scale dependency memory network includes a short-term dependency module and a long-term dependency module. The short-term dependency module uses the information of the most recent time points in the sliding window and inputs it into the long-short-term memory network to capture short-term load fluctuations within approximately 1 day. The long-term dependency module is based on the historical load data of the previous few days and is input into the multi-scale dependency memory network through the largest time window to capture periodic fluctuations.

[0009] Among them, in the load peak and valley prediction model based on the multi-scale dependency memory network, the implicit state output by the short-term dependency module and the multi-scale features output by the long-term dependency module are merged in the fusion layer to form the final node load feature representation, and the data after feature fusion is mapped to the prediction of the target load value through the fully connected layer;

[0010] A dynamic load coupling access point optimization model is constructed to handle dynamic load changes in the medium voltage distribution network. At the same time, the energy loss of access point selection is minimized according to the load peak and valley prediction. At the same time, voltage constraints and line capacity constraints are added to the dynamic load coupling access point optimization model. The voltage constraint keeps the voltage within a reasonable range and can automatically adjust when the voltage fluctuates greatly; the line capacity constraint ensures that the selection of the access point does not exceed the maximum transmission capacity of the line load. Finally, the dynamic load coupling access point optimization model is solved to obtain the final access point selection; wherein, the objective function of the dynamic load coupling access point optimization model is expressed as follows:

[0011] ,

[0012] in, Indicates The impedance of the line, is the current flowing through the line. This current is related to the node load characteristic representation. The model determines the load of each node through the node load characteristic representation and calculates the power flow and loss on the line accordingly.

[0013] In the final access point selection, a load-topology adaptive optimization model is constructed to minimize the energy loss of the medium-voltage distribution network by dynamically adjusting the line. At the same time, a dynamic line reconstruction mechanism is introduced in the dynamic adjustment line to dynamically optimize the enabled or disabled state of the line by adjusting the elements in the connection matrix to achieve balanced load distribution and reduce energy loss under different load conditions. In order to prevent grid instability, voltage constraints and line capacity constraints are further added, and finally the optimized topology structure is output; wherein, the objective function of the load-topology adaptive optimization model is expressed as follows:

[0014] ,

[0015] in, Indicates line Impedance; Indicates that the line The current; represents the regularization term, which is used to suppress the frequent switching of the line; Represents the weight of the regularization term, controlling the flexibility of topology adjustment;

[0016] Construct an adaptive dynamic load-topology adjustment model based on the real-time load fluctuation and topology structure adjustment of the distribution network, combine the load forecasting, access point optimization and topology optimization results, and improve the operation efficiency of the power grid and ensure the stability of the medium-voltage distribution network by adaptively adjusting the access points and power transmission paths;

[0017] Based on long-term optimization and system evaluation, combined with the peak-to-valley fluctuation characteristics of the medium-voltage distribution network, load management and energy transmission efficiency are carried out through long-term evaluation and feedback optimization.

[0018] Furthermore, the preprocessing includes Z-score processing, missing value filling processing and maximum-minimum normalization processing;

[0019] The Z-score processing is calculated based on the mean and standard deviation, and determines whether it is an outlier by setting a threshold. At the same time, the detected outliers are corrected by weighted average based on adjacent time points;

[0020] The missing value filling process adopts a dynamic time backtracking method to fill the missing values ​​using the pattern similarity of the load data at the previous moment;

[0021] The maximum-minimum normalization process normalizes all data to within the scope of

[0022] The feature expansion process constructs interactive features by performing feature engineering on external factors;

[0023] The time series construction will use the sliding window method with a window length of Set to 96 time points for construction.

[0024] Furthermore, the time series data containing load data characteristics are divided into training set, validation set and test set, and the weighted mean square error is used to construct the loss function to adapt to the non-uniformity of peak and valley load changes.

[0025] Further, the voltage constraint is expressed as follows:

[0026] ,

[0027] in, Indicates access node The voltage, and are the minimum and maximum voltage values ​​allowed, and N represents the total number of nodes;

[0028] The line capacity constraint is expressed as follows:

[0029]

[0030] in, Indicates that the line The power, Indicates the maximum capacity of the line, and M indicates the total number of lines.

[0031] Furthermore, the dynamic load coupling access point optimization model is solved by mixed integer linear programming, and the selection of access points is set as a binary variable , indicating that at the node Whether to connect to distributed power supply or load is indicated as follows:

[0032] ,

[0033] in, Indicates the final access point selection.

[0034] Furthermore, the dynamic line reconstruction mechanism is expressed as follows:

[0035] ,

[0036] in, Indicates the activation status between lines;

[0037] The status of each line is determined by load forecasting and access points, and is adjusted in real time through topology optimization. When a load peak occurs, low-impedance lines are activated to reduce transmission losses, while some high-impedance lines are disabled.

[0038] Furthermore, the load-topology adaptive optimization model introduces a topology change penalty term in the objective function according to the fact that too frequent topology reconstruction leads to unstable system operation, which is used to control the frequent adjustment of the topology structure, and is expressed as follows:

[0039] ,

[0040] in, Represents the connection matrix before and after topology adjustment The difference is the change of the circuit breaker status; It represents the penalty coefficient that controls the frequency of topology adjustment.

[0041] Furthermore, the adaptive dynamic load-topology adjustment model minimizes the comprehensive operation loss of the system by adjusting the topology structure and access point selection in real time. The objective function of the adaptive dynamic load-topology adjustment model is expressed as follows:

[0042] ,

[0043] in, Indicates line Impedance, controlling energy loss in power transmission; Indicates that the line The current depends on the current load And the line status after topology adjustment ; Indicates the deviation between the current load forecast value and the actual load value; represents the penalty term, which is used to suppress frequent topology adjustments and control the stability of the system; and Indicates the adjustment parameter.

[0044] Furthermore, the long-term optimization objective function of the long-term optimization and system evaluation framework is expressed as follows:

[0045] ,

[0046] in, Indicates time The energy loss at each moment reflects the impact of load peak and valley fluctuations on energy consumption in the power grid after topology adjustment. represents the adjustment cost, which indicates the switching and reconfiguration cost of the system due to frequent topology reconfiguration and access point adjustment, especially considering the high-frequency adjustment caused by load fluctuation in the medium-voltage distribution network; represents the long-term stability constraint; , and are weight parameters respectively, which balance energy consumption, adjustment cost and long-term stability according to the actual operation requirements of the power grid.

[0047] Furthermore, the long-term optimization and system evaluation framework designs voltage balance and load stability evaluation mechanisms for long-term operation. , which is expressed as follows:

[0048] ,

[0049] in, and Respectively represent nodes In time The maximum and minimum voltage values ​​within.

[0050] The beneficial technical effects of the present invention are at least as follows:

[0051] The innovation of the present invention lies in the proposal of a comprehensive method for determining access to a medium-voltage distribution network, which solves several key problems in the prior art by integrating load peak and valley prediction and grid topology optimization. First, the present invention adopts a load peak and valley prediction method based on an advanced time series model (such as a Transformer model), which can more accurately capture the seasonal and cyclical changes in the load, as well as the characteristics of sudden fluctuations, and improve the accuracy of load prediction, especially when the grid load changes are complex and diverse. Compared with traditional linear regression and simple time series models, this method not only improves the accuracy of prediction, but also significantly reduces the dependence on historical data and error sensitivity.

[0052] Secondly, the present invention introduces the mixed integer linear programming (MILP) method in access point optimization, which can handle multiple optimization objectives and complex physical constraints at the same time, flexibly respond to different scenarios of load peak and valley changes, dynamically adjust access point selection, and improve the rationality of power distribution and the operation efficiency of the power grid. In addition, the present invention combines Graph Neural Networks (GNN) to intelligently optimize the topology of the distribution network. By learning the complex relationship between the nodes and lines of the power grid, the optimal node connection method is found, which reduces the energy loss in the power transmission process and effectively controls the voltage fluctuation and the stability of the power grid.

[0053] Finally, the method of the present invention has the ability of real-time monitoring and dynamic adjustment. Through real-time data monitoring and feedback mechanism, it can continuously optimize the load forecasting model and access strategy to ensure that the power grid always maintains efficient operation and stable power supply under the conditions of load fluctuations and topology changes. This comprehensive solution not only solves the problems of inaccurate load forecasting, inflexible access point selection and unstable power grid operation in the prior art, but also provides effective technical support for the construction and optimization of future smart grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0055] Figure 1 The present invention is a flowchart of a method for determining medium voltage distribution network access based on load peak-valley coupling according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0057] In one or more embodiments, Figure 1 As shown, a method for determining medium voltage distribution network access based on load peak-valley coupling is disclosed, and the method includes steps S1 to S6, including:

[0058] S1. Collect the load data of the medium voltage distribution network and preprocess them respectively, and expand the features of the preprocessed data at the same time, form a data set with the expanded features and construct a time series to obtain time series data containing the load data features; wherein the load data includes the load power, voltage and external factors at each time point.

[0059] First, the present invention comprehensively collects and preprocesses the load data of the medium voltage distribution network to cope with the special volatility and multi-dimensional characteristics of the data. The load data includes the load power at each time point ( ),Voltage( ) and external factors ( ), such as temperature, humidity, weather conditions and holiday markings, which significantly affect the load changes of the power grid. In order to ensure the uniformity of the data, all these data sources are first aligned and merged according to the timestamp to form a structured multidimensional dataset. The structure of the dataset is ,in is the timestamp of each time point, is the load power value, is the voltage value, and It is an external environmental factor.

[0060] During the data collection process of the power system, sensors may fail or communication may be interrupted, which may lead to abnormal values ​​and missing values ​​in the data. These situations will significantly affect the accuracy of the subsequent prediction model. Therefore, the present invention first detects and repairs the abnormal values ​​in the data set. Assume that the load data The distribution of is approximately normal distribution. The present invention determines whether there is an outlier by calculating the Z-score of each data point. The formula of Z-score is:

[0061] ,

[0062] in, and The load data are By setting the threshold , to determine whether it is an outlier. For the detected outliers, the present invention does not simply discard them, but uses a weighted average based on adjacent time points to correct them:

[0063] ,

[0064] here and are weight values ​​set according to the proximity of time points, usually This correction method can ensure the retention of load fluctuation trend and is particularly suitable for situations with frequent fluctuations in power load.

[0065] For missing values ​​in the data caused by sensor failure or communication problems, using a simple interpolation method may lose the time series characteristics of load fluctuations. The present invention adopts a missing value filling method based on dynamic time warping (DTW), which uses the pattern similarity of the load data at previous moments to fill the missing values. Specifically, assuming that the time point corresponding to the current missing value is , then use before The data of each time point is filled by weighting the load pattern similarity:

[0066] ,

[0067] in is the weight calculated based on the similarity of load patterns, It's time point and This method can adaptively fill in missing values ​​according to the similarity of time series, and is particularly suitable for dealing with scenarios with frequent load fluctuations.

[0068] After the outlier correction and missing value filling are completed, in order to further improve the availability of the data, the present invention performs normalization processing on the data set. and voltage The magnitude of is different. If used directly, it may affect the convergence of the model. The present invention adopts the maximum-minimum normalization method to normalize all data to Within the scope of:

[0069] ,

[0070] For external factors data , such as temperature and humidity, also use a similar normalization method to make them conform to a unified scale to ensure the stability of subsequent prediction model inputs.

[0071] Considering that load data is affected by multiple factors, simply using raw data may not be able to fully capture the complexity. Therefore, the present invention performs feature expansion to generate more features that can characterize load fluctuations. By performing feature engineering on external factors such as temperature and humidity, interactive features are constructed. For example, the cross term of temperature and humidity and the square term of temperature, etc.:

[0072] ,

[0073] This feature expansion can enhance the model's ability to capture the complexity of load changes, especially when environmental conditions change drastically, and can better predict load peak and valley fluctuations.

[0074] Finally, the present invention constructs a time series for the processed data set to meet the input requirements of the subsequent model. Assuming that the present invention hopes to predict the load peak and valley changes in the next 24 hours, the present invention will use the sliding window method with a window length of Set to 96 time points (corresponding to 24 hours). For each time point , before extraction of the present invention The data of time points is taken as input:

[0075] ,

[0076] This sliding window method can capture the information of historical load fluctuations, thus providing powerful input features for load peak and valley prediction.

[0077] Through the above data preprocessing method, the present invention ensures the high quality and availability of power load data, and at the same time, through the steps of outlier correction, missing value filling, feature expansion and time series construction, fully considers the complexity and volatility of power load data. This lays a solid foundation for subsequent load peak and valley prediction, enabling the prediction model to accurately respond to the dynamic changes and external influences of power load.

[0078] S2. Construct a load peak and valley prediction model based on a multi-scale dependent memory network, take the time series as input, and perform load peak and valley prediction by simultaneously processing load changes at different time scales.

[0079] Specifically, the present invention proposes a model based on capturing time series features for the peak-to-valley changes in the load of the medium-voltage distribution network. The goal of this solution is to make full use of the time series data generated in step 1 ( ), combined with environmental factors and historical load data, to build an accurate load peak and valley prediction model to solve the complexity of dynamic load fluctuations. Since load changes are not only affected by historical loads, but also by the multi-factor coupling of the relationship between grid nodes and the external environment, the present invention incorporates a multi-layer dependent long-short-term memory structure into the model design, and optimizes the feature selection in a targeted manner to ensure the accuracy and stability of the results.

[0080] Further, using the time series data from step 1 As input, in order to ensure that the model can capture complex temporal dependencies and nonlinear fluctuation characteristics, the present invention adopts a multi-level time-dependent modeling strategy.

[0081] Furthermore, the present invention proposes a prediction scheme based on a multi-scale dependency memory network (MDMN). MDMN aims to achieve more refined load peak and valley prediction by simultaneously processing load changes at different time scales. The model is composed of two main modules in terms of architecture:

[0082] Short-term dependency module: used to capture short-term load fluctuations within approximately 1 day. This module uses the information of the latest time points in the sliding window and inputs it into the short-term prediction network. For this purpose, the present invention uses an adjusted long short-term memory network (LSTM) structure to process these short-term data inputs. The model formula is:

[0083] ,

[0084] in, is the implicit state at the current time point, capturing the past The main function of this module is to capture the sudden fluctuations in the power grid load, especially the short-term peak-to-valley changes during the morning and evening peak hours.

[0085] Long-term dependency module: used to identify long-term load trends and cyclical changes over multiple days or weeks. The long-term dependency module is based on historical load data from the previous few days and uses a relatively large window input to capture cyclical fluctuations. The model formula is:

[0086] ,

[0087] in, It is a multi-scale dependency feature, covering the load pattern of the previous 7 days. The long-term dependency module handles medium- and long-term fluctuations in load by expanding the window length and period capture, and combined with the periodic operation law of the power grid, it is particularly suitable for dealing with load peak and valley fluctuations caused by holidays and seasonal changes.

[0088] Furthermore, the implicit state of the short-term dependency module output Multi-scale features of long-term dependency modules The fusion layer merges the features to form the final load feature representation. The feature-fused data is mapped to the prediction of the target load value through the fully connected layer:

[0089] ,

[0090] in, For The load forecast results at the time, and is the weight matrix of short-term and long-term features, In this way, the present invention can utilize the short-term sudden fluctuation and long-term trend change of the load to ensure that the prediction result has a higher accuracy in the peak-to-valley change.

[0091] Furthermore, the complete data set generated in step 1 is divided into a training set, a validation set, and a test set according to time periods. As for the selection of the loss function, the present invention uses the weighted mean square error (WMSE) to adapt to the non-uniformity of peak and valley load changes. The formula is as follows:

[0092] ,

[0093] in, is the weight related to the load peak-to-valley variation amplitude, is the actual load value, is the predicted value. Weight The design is to emphasize the prediction accuracy during peak and valley periods, making the model pay more attention to moments when load changes drastically.

[0094] Through the above design, MDMN can combine the short-term fluctuations and long-term trends of power grid load, solve the complex dynamic fluctuation problems in load peak and valley forecasting, and accurately identify load peak and valley points in the forecast. This multi-level dependent feature capture mechanism ensures that the load forecast results can adapt to the dynamic needs of the power grid and provide reliable basic data for subsequent access point optimization.

[0095] S3. Construct a dynamic load-coupled access point optimization model to handle dynamic load changes in the medium-voltage distribution network, and minimize the energy loss of access point selection based on load peak and valley prediction.

[0096] Specifically, use the future load peak and valley data predicted in step 2 As a key input, combined with historical load data , voltage data And environmental factors These input data are used to describe the dynamic changes of each access point and its surrounding load. In particular It represents the load trend in the future. Therefore, the introduction of this data enables access point selection to cope with future load fluctuations and prevent local overload or voltage instability caused by uneven load.

[0097] In addition to load-related data, the model also incorporates information about the grid’s topology and physical constraints, including substation capacities. , Line impedance Etc. These data, combined with load forecasts, enable the model to optimize the distribution of access points to suit the actual operating conditions of the grid.

[0098] Furthermore, the core goal of access point optimization is to minimize the energy loss during power transmission, which is particularly important in medium-voltage distribution networks, especially in scenarios where load peaks and valleys fluctuate significantly. Load fluctuations are frequent, and the loss of power transmission is unstable, so it is necessary to dynamically select access points to match load changes. The objective function of the present invention is to minimize the loss on the transmission line, and the formula is:

[0099] ,

[0100] in, Indicates The impedance of the line, is the current flowing through the line. This current is related to the node load prediction The model is related to The load at each node is determined and based on this the power flows and losses on the line are calculated.

[0101] In view of the actual situation of the distribution network, the present invention specially designs access point optimization in scenarios with large load fluctuations, that is, adjusting the distribution of access points at load peaks to reduce transmission losses during peak periods. Traditional methods cannot dynamically respond to such fluctuations, and DLCAO achieves this optimization through deep integration with load forecast data.

[0102] Furthermore, in the medium voltage distribution network, the voltage stability of each node must be strictly maintained within a certain range to avoid the voltage being too high or too low due to the selection of the access point. To ensure this, the model introduces a voltage constraint, which is expressed as:

[0103] ,

[0104] in Indicates access node The voltage, and are the minimum and maximum voltage values ​​allowed. This constraint ensures that when the access point is optimized, the system voltage will not be unstable due to the access to distributed power sources or loads, especially when the load fluctuates, the system can still keep the voltage within a reasonable range. Voltage Fluctuation and Load Forecasting Closely related, DLCAO uses load peak and valley prediction to enable access points to automatically adjust when voltage fluctuations are large.

[0105] This design specifically takes into account the patented load peak-valley coupling problem. Unlike traditional static access solutions, DLCAO can dynamically respond to voltage changes with the support of load forecasting, ensuring that the system always operates within a safe voltage range.

[0106] Furthermore, since the power grid line has capacity limitations, the selection of access points must ensure that the line load does not exceed its maximum transmission capacity. The present invention designs the line capacity constraint conditions:

[0107] ,

[0108] in, To pass the line The power, is the maximum capacity of the line. Prediction by node load The DLCAO calculates the load of each line in the future period in advance through load peak and valley prediction to ensure that the access point selection will not cause line overload. This capacity constraint design, combined with load peak and valley prediction, makes access point selection more intelligent and predictable, and can consider load changes in advance to avoid line overload risks.

[0109] Furthermore, the access point optimization problem is solved by mixed integer linear programming (MILP). In this model, the choice of access point is set as a binary variable , indicating that at the node Whether to connect to distributed power supply or load:

[0110] ,

[0111] By optimizing , DLCAO based on load forecast As well as voltage and current constraints, the optimal access point location is dynamically selected to maximize the operating efficiency of the power grid and minimize energy loss. Since the peak and valley changes of the load have been predicted in advance, the model can intelligently adjust the access point to keep the system in the optimal state at different time periods.

[0112] The special design of DLCAO solves the core problem in medium-voltage distribution networks, namely the impact of dynamic changes in load and load peak-valley coupling on access point selection. Through the deep coupling of load forecasting and access point selection, the present invention ensures efficient operation of the power grid under complex load conditions, especially during periods of severe load fluctuations, which can significantly reduce the energy loss of the power grid and improve operational stability.

[0113] S4. In the final access point selection, a load-topology adaptive optimization model is constructed to minimize the energy loss of the medium-voltage distribution network by dynamically adjusting the lines.

[0114] Specifically, in this step, the present invention proposes an innovative load-topology adaptive optimization model (Load-Topology Adaptive Optimization, LTAO), which aims to dynamically optimize the topology structure according to the load peak-valley changes in the medium-voltage distribution network. LTAO is not only consistent with the load peak-valley prediction in the previous step (step 2 ) and access point optimization (access point selection in step 3 ), and also fully considers the actual operation requirements of the distribution network under dynamic load changes, especially under different load conditions, by adjusting the topology to optimize the power transmission path and reduce transmission losses. Through this optimization, the energy efficiency of the system and the stability of the power grid are ensured, especially in scenarios with severe load fluctuations.

[0115] Specifically, the input to the topology optimization model is the results of the previous steps, including:

[0116] Access point selection : This is the access point variable that has been optimized in step 3, indicating the node Whether it is selected as the access point of distributed generation or load.

[0117] Load peak and valley prediction : This is the load forecast value output from step 2, which reflects the load change trend of each node in the future time period and can capture the peak and valley values ​​of the future load.

[0118] Topology and physical constraints: This includes node locations , Line connection matrix And node capacity , Line impedance And other actual power grid operation data.

[0119] The integration of these data enables the present invention to adjust the topology of the power grid in real time based on load forecasting results and access point selection, so that the power transmission path can adapt to dynamic load changes.

[0120] Furthermore, in order to cope with the peak and valley fluctuations of the load, the core goal of topology optimization is to minimize the energy loss in the power transmission process while ensuring the voltage stability of the system and the load balance of the line. Specifically, at the peak load, the line is dynamically adjusted to ensure that the load can be transmitted through the low-impedance line, thereby reducing the transmission loss. The objective function of loss minimization can be expressed as:

[0121] ,

[0122] in, Indicates line The impedance of a line usually represents the physical properties of the line. Indicates that the line The current is predicted by the node load and access point selection Decide. Represents a regularization term, which is used to suppress frequent switching of lines. This design is a special consideration for actual scenarios to avoid frequent adjustments of the topology when the load fluctuation is small, which may lead to system instability. Represents the weight of the regularization term, which controls the flexibility of topology adjustment. The present invention can control the frequency of line reconstruction and ensure the adaptability of the system when the load peaks and valleys change.

[0123] Furthermore, in order to ensure uniform distribution of load in the topological network, the present invention introduces a dynamic line reconstruction mechanism. This mechanism adjusts the connection matrix The elements in the circuit dynamically optimize the enabled or disabled state of the circuit to achieve balanced load distribution and reduce energy loss under different load conditions. The reconstruction of the circuit is based on load peak and valley prediction. , and adjust the connectivity matrix by :

[0124] ,

[0125] The status of each line is determined by the load forecast and access point The topology is optimized and adjusted in real time. When a load peak occurs, low-impedance lines are activated to reduce transmission losses, while some high-impedance lines are disabled.

[0126] In order to ensure that the load of each node can be reasonably distributed, the present invention introduces the following load balancing constraints:

[0127] ,

[0128] in, It is a slave node To Node The power flow, Is a node Through this constraint, the present invention ensures that the load demand of each node in the system can be met through the activated lines.

[0129] Furthermore, in order to prevent topology adjustment from causing grid instability, LTAO also introduces constraints on node voltage and line capacity:

[0130] Voltage constraint: voltage at each node It needs to be kept within the specified safety range to prevent the voltage from being too high or too low and causing damage to the power grid. This constraint is expressed as:

[0131] ,

[0132] Voltage Fluctuation and Load Forecasting Directly related, LTAO maintains voltage stability when load peaks and valleys fluctuate by dynamically adjusting the topology.

[0133] Line capacity constraint: To prevent line overload, the transmission power of each line must be less than its maximum capacity :

[0134]

[0135] in, It is through the line The actual power transmission is achieved by combining the load prediction value By combining the topology reconstruction, the present invention ensures that the line always operates within a safe transmission capacity range under load peak and valley fluctuation conditions, thus avoiding line overload.

[0136] Furthermore, in order to further balance the flexibility of topology adjustment and the stability of the system, the present invention introduces an innovative topology change penalty term in the objective function. In actual power grids, too frequent topology reconstruction may lead to unstable system operation, so the present invention introduces a penalty term , used to control frequent adjustments of the topology:

[0137] ,

[0138] in, Represents the connection matrix before and after topology adjustment The difference is the change in the line switch state. Indicates the penalty coefficient for controlling the frequency of topology adjustment. , which can suppress frequent topology adjustments, so that the model will not frequently readjust the line when there is a small load fluctuation. This penalty term ensures the stability of the system, especially when the load fluctuates slightly, the topology will not be adjusted too frequently, thereby reducing unnecessary system switching operations.

[0139] Furthermore, the solution of the LTAO model adopts mixed integer nonlinear programming (MINLP). In this optimization problem, the line state matrix , load distribution variables and load current etc. are all decision variables. The model is solved by the following steps:

[0140] Topology adjustment: through load forecasting and access point selection , dynamically adjust the line state matrix , enable or disable different lines.

[0141] Load distribution: Based on the adjusted topology, the load on each line is redistributed to ensure balanced transmission of load in the system and minimize energy loss.

[0142] Penalty Constraint: Adjusting Penalty by Introducing Topology , control the frequency of topology reconstruction and avoid frequent adjustments of the system.

[0143] LTAO and load peak and valley prediction and access point optimization The deep integration of the network dynamically adjusts the topology of the power grid to adapt to the load peak and valley changes. The model not only reduces transmission losses by reconfiguring the lines, but also introduces a penalty term for topology adjustment to avoid instability caused by frequent system adjustments. Through this innovative design, LTAO provides an efficient and stable power transmission optimization solution under the complex load conditions of the medium-voltage distribution network.

[0144] .S5. Construct an adaptive dynamic load-topology adjustment model. According to the real-time load fluctuation and topology structure adjustment of the distribution network, combined with the load forecasting, access point optimization and topology optimization results, the access points and power transmission paths are adaptively adjusted to improve the operation efficiency of the power grid and ensure the stability of the medium-voltage distribution network.

[0145] Specifically, in this step, the present invention proposes an adaptive dynamic load-topology adjustment model (ADLTA) based on a patent scenario, which aims to optimize the operation status of the power grid in real time. ADLTA targets the real-time load fluctuation and topology structure adjustment of the distribution network, combines load forecasting, access point optimization and topology optimization results, and improves the operation efficiency of the power grid and ensures the stability of the system by adaptively adjusting the access points and power transmission paths.

[0146] Furthermore, the input of the ADLTA model mainly comes from the output of the previous steps:

[0147] Topology Optimization Output :This is the topology optimized in step 4, indicating the activation states between the lines ( ) and the power transmission of each line ( ).

[0148] Load Forecasting : From step 2, it represents the load peak and valley prediction for a specific period in the future, ensuring that the present invention can adjust the system in advance to cope with load fluctuations.

[0149] Access point selection : The access point selection obtained in step 3 represents the node Whether to connect to distributed power sources or loads.

[0150] These input data bring together the system's historical status, load forecasts, and dynamic topology information, enabling the model to make adaptive adjustments based on real-time changes to avoid system overload, energy waste, or voltage imbalance.

[0151] Furthermore, the core goal of ADLTA is to minimize the comprehensive operating loss of the system and ensure the stable operation of the power grid by adjusting the topology and access point selection in real time. In order to achieve this goal, the present invention defines the following real-time adjustment objective function:

[0152] ,

[0153] in, Indicates line The impedance controls the energy loss of power transmission. Indicates that the line The current depends on the current load And the line status after topology adjustment . Indicates the deviation between the current load forecast value and the actual load value. The present invention introduces a load tracking mechanism through this item to ensure that the system can adjust in real time to cope with load changes. represents the penalty term, which is used to suppress frequent topology adjustments and control the stability of the system. and are adjustment parameters, which are used to balance the accuracy of load tracking and the flexibility of topology adjustment. Through this optimization goal, the system can make adaptive adjustments in the case of real-time load fluctuations and topology changes, reduce transmission losses and improve the operation stability of the system.

[0154] Furthermore, the load tracking mechanism in ADLTA relies on the difference between real-time load monitoring and forecasting. Assume that the load forecast value for each time period in the system is , and the actual load value is , the model will dynamically adjust the access point and topology according to this difference. The present invention designs the following adjustment mechanism:

[0155] Access point adaptive adjustment: based on load prediction With actual load The model will dynamically adjust the enabled and disabled states of the access points. If the load of a node exceeds its safe capacity, the present invention will balance the load by dynamically enabling or disabling the access points. For example, if the node If the load is too high, the present invention will dynamically shut down the access point , redistribute the load:

[0156] ,

[0157] Adaptive adjustment of topology: When the load peaks and valleys vary greatly, ADLTA will dynamically adjust the topology and optimize the transmission path by reconfiguring the line status. is the initial line state matrix, It is a matrix adjusted in real time according to load changes. Indicates whether the line is enabled:

[0158] ,

[0159] Through this adjustment, the present invention can ensure that the load is transmitted through the optimal path in the system, thereby reducing the energy loss of the system.

[0160] Furthermore, in order to ensure the voltage stability of the system, ADLTA introduces voltage balance constraints to ensure that load fluctuations do not cause system voltage instability. The specific constraints are as follows:

[0161] ,

[0162] in, Is a node The voltage across the load is predicted The topology structure is dynamically adjusted to ensure that the voltage is always within a stable range. Through voltage constraints, ADLTA can dynamically adjust the line when the load fluctuates to ensure that the voltage does not exceed the predetermined range.

[0163] Furthermore, in actual scenarios, too frequent adjustments may have a negative impact on the system. Therefore, the present invention introduces an adaptive frequency control mechanism to control the frequency of topology adjustment and access point switch state changes to prevent instability caused by frequent system switching. The specific formula is as follows:

[0164] ,

[0165] This penalty term controls the frequency of line switch state changes, Indicates the status difference before and after the line adjustment, by adjusting the parameters To reduce the impact of frequent adjustments on system stability.

[0166] Furthermore, the optimization solution of ADLTA is based on the adaptive adjustment of real-time data and historical load data, and is solved using mixed integer nonlinear programming (MINLP). The solution steps are as follows:

[0167] Real-time load monitoring: load data monitored in real time and load forecasting The access points and topology can be adjusted dynamically based on the differences.

[0168] Dynamic adjustment solution: adjust the topology structure according to the optimization objective function and access point selection , while considering constraints such as voltage balance and frequency control.

[0169] Feedback and update: The load forecasting model and topology optimization strategy are regularly updated through the feedback mechanism to ensure that the system can be adaptively adjusted in long-term operation.

[0170] The ADLTA model solves the dynamic optimization problem of medium-voltage distribution networks under load fluctuations through real-time optimization and adaptive adjustment. It combines load forecasting, access point optimization, and topology optimization to reduce energy loss, improve system stability, and prevent instability caused by frequent system adjustments by dynamically adjusting access points and power transmission paths. ADLTA is particularly suitable for scenarios with frequent load fluctuations, ensuring efficient operation of the power grid under different load conditions.

[0171] S6. Based on long-term optimization and system evaluation, combined with the peak-to-valley fluctuation characteristics of the medium-voltage distribution network, load management and energy transmission efficiency are carried out through long-term evaluation and feedback optimization.

[0172] Specifically, after a period of operation, the overall optimization effect of the power grid needs to be evaluated, and the models and strategies need to be periodically improved and optimized.

[0173] In the long-term optimization and system evaluation stage, the present invention designs a long-term adaptive system evaluation and optimization framework (Long-term Adaptive System Evaluation and Optimization, LASEO) specifically for medium-voltage distribution networks. This scheme combines the outputs of the previous steps, especially considers the load peak-valley fluctuation characteristics of the medium-voltage distribution network, and achieves continuous and efficient operation of the system through long-term evaluation and feedback optimization. LASEO specifically targets the challenges brought about by the dynamic changes of load and topology in such power grids, and achieves better load management and energy transmission efficiency through optimization models and adjustment strategies.

[0174] Furthermore, the input of the LASEO framework mainly comes from the optimization results of the previous steps and is specially designed for the characteristics of the medium-voltage distribution network:

[0175] In the long-term optimization and system evaluation stage, the present invention designs a long-term adaptive system evaluation and optimization framework (Long-term Adaptive System Evaluation and Optimization, LASEO) specifically for medium-voltage distribution networks. This scheme combines the outputs of the previous steps, especially considers the load peak-valley fluctuation characteristics of the medium-voltage distribution network, and achieves continuous and efficient operation of the system through long-term evaluation and feedback optimization. LASEO specifically targets the challenges brought about by the dynamic changes of load and topology in such power grids, and achieves better load management and energy transmission efficiency through optimization models and adjustment strategies.

[0176] Furthermore, the input of the LASEO framework mainly comes from the optimization results of the previous steps and is specially designed for the characteristics of the medium-voltage distribution network:

[0177] Real-time optimization results : This is the topology adjustment result after real-time optimization in step 5, which represents the activation status of each line and the power flow.

[0178] Historical load forecast : From step 2, it describes the load changes in each time period in the future and is used for long-term optimization evaluation.

[0179] Actual load value :The load data obtained from the actual monitoring system is used to evaluate the accuracy of load forecasting and to perform feedback optimization based on the characteristics of long-term load fluctuations.

[0180] In this input system, LASEO closely combines load fluctuations and system topology to meet the special needs of the distribution network, ensuring that long-term optimization not only focuses on the stability of short-term operation, but also comprehensively considers the complex problems brought about by future load changes.

[0181] Furthermore, the long-term optimization goal of LASEO combines the peak and valley characteristics of the medium-voltage distribution network and the multi-dimensional consideration of system energy consumption, focusing on balancing energy loss and long-term stability of the system. Its optimization objective function is:

[0182] ,

[0183] in, Indicates time The energy loss at the moment reflects the impact of load peak and valley fluctuations on energy consumption in the power grid after topology adjustment. It optimizes the line activation state and load forecasting to calculate. is the adjustment cost of the system, which represents the switching and reconfiguration cost of the system due to frequent topology reconfiguration and access point adjustment, especially considering the high-frequency adjustment caused by load fluctuations in the medium-voltage distribution network. It is a long-term stability constraint of the system, which measures the voltage stability of the system under load fluctuations, node capacity constraints, etc., to ensure that long-term optimization not only focuses on energy consumption, but also ensures that the voltage and capacity of the power grid always remain within a reasonable range. , and are weight parameters, respectively, to balance energy consumption, adjustment cost and long-term stability according to the actual grid operation requirements. Through this optimization goal, LASEO can effectively adapt to long-term load changes and ensure that the system always operates in a state of low energy consumption and high stability.

[0184] Furthermore, LASEO introduces a long-term evaluation and feedback mechanism to optimize access points and topology adjustment strategies by regularly evaluating the system's operating status and load forecast error. Since the load fluctuations of the medium-voltage distribution network are large, especially the drastic changes in peak and valley loads will affect the long-term stability of the system, the present invention quantitatively evaluates the load forecast error and energy loss:

[0185] ,

[0186] in, is the number of nodes in the grid, is the load forecast value, is the actual load value. By calculating the prediction error, the system can adjust the access point selection in time. , and re-evaluate topology adjustments , ensuring that the system can flexibly adjust to long-term load fluctuations.

[0187] LASEO also designed a periodic adjustment mechanism for scenarios with large load fluctuations in the distribution network, so that the system can review the overall operating status in each time window (such as weekly or monthly) and globally optimize the access point and line status. The specific design is as follows:

[0188] Access point long-term adjustment: Evaluate the continuous operation efficiency of access points based on long-term load changes. If the access point of a node is continuously inefficient over a long period of time, the system will automatically adjust Enabled status:

[0189] ,

[0190] Long-term optimization of line status: evaluate the long-term operation of each line, especially high-load frequency and high-energy consumption lines. Optimize the use of high-energy-consuming lines, and give priority to activating low-energy-consuming and efficient power transmission paths:

[0191] ,

[0192] Furthermore, LASEO has designed a voltage balance and load stability evaluation mechanism for the long-term operation of the distribution network, especially considering the impact of voltage fluctuations on the system when the load fluctuates greatly. The voltage stability is quantified by the following formula:

[0193] ,

[0194] and Respectively represent nodes In time By regularly evaluating voltage fluctuations, LASEO can identify potential voltage imbalance problems when load peaks and valleys vary greatly, and then and topology Adjustments keep the system stable.

[0195] In addition, for the long-term system health status, LASEO will also monitor the operation status of the power grid based on the load capacity of the node to ensure that the load of each node does not exceed its safety capacity. This evaluation mechanism ensures that the equipment will not be overloaded or fail due to load fluctuations during the long-term operation of the power grid.

[0196] Furthermore, the long-term optimization solution of LASEO is based on historical data and long-term feedback mechanism, and is optimized through mixed integer linear programming (MILP). The specific solution steps include:

[0197] Long-term optimization of energy loss: Analyze energy loss trends based on historical load fluctuations and dynamically adjust access points and line status , achieving long-term optimization with minimized energy losses.

[0198] Periodic evaluation and feedback: In each optimization cycle (e.g., monthly), key indicators such as the system's energy consumption, voltage stability, and load fluctuation are reviewed, and global adjustments are made to the topology and access point configuration for the next cycle.

[0199] Global health monitoring and long-term stability: Regularly check the voltage fluctuations and load balance of each node in the power grid to ensure that the system is in an efficient and stable operating state for a long time.

[0200] Through this design, the LASEO framework improves the energy efficiency and stability of the power grid through long-term evaluation and feedback optimization strategies based on the characteristics of load fluctuations in medium-voltage distribution networks. This solution not only integrates the optimization results of the previous steps, but also combines the long-term trend of load peak and valley fluctuations. Through the optimization of access points and topology structures, it provides an adaptive long-term optimization solution that can cope with complex power grid operation needs. In the continuous system evaluation, LASEO ensures that the power grid can maintain efficient and stable operation for a long time through periodic adjustments, and achieves global optimization through energy loss and voltage fluctuation control.

[0201] These are just some preferred embodiments of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and making equivalent changes according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for determining medium voltage distribution network access based on load peak-valley coupling, characterized in that: The method comprises the following steps: Collect the load data of the medium voltage distribution network and pre-process them respectively, and expand the features of the pre-processed data, form the expanded features into a data set and construct a time series to obtain time series data containing the load data features; wherein the load data includes the load power, voltage and external factors at each time point; The load data of the medium voltage distribution network is collected and preprocessed respectively. The time series is used as the input of the multi-scale dependency memory network, and the load peak and valley prediction is performed by simultaneously processing the load changes of different time scales; wherein, the multi-scale dependency memory network includes a short-term dependency module and a long-term dependency module. The short-term dependency module uses the information of the latest several time points in the sliding window and inputs it into the long short-term memory network to capture the short-term load fluctuation within approximately 1 day; the long-term dependency module is based on the historical load data of the previous few days and is input into the multi-scale dependency memory network through the largest time window to capture the periodic fluctuation; Among them, in the load peak and valley prediction model based on the multi-scale dependency memory network, the implicit state output by the short-term dependency module and the multi-scale features output by the long-term dependency module are merged in the fusion layer to form the final node load feature representation, and the data after feature fusion is mapped to the prediction of the target load value through the fully connected layer; A dynamic load coupling access point optimization model is constructed to handle dynamic load changes in the medium voltage distribution network. At the same time, the energy loss of access point selection is minimized according to the load peak and valley prediction. At the same time, voltage constraints and line capacity constraints are added to the dynamic load coupling access point optimization model. The voltage constraint keeps the voltage within a reasonable range and can automatically adjust when the voltage fluctuates greatly; the line capacity constraint ensures that the selection of the access point does not exceed the maximum transmission capacity of the line load. Finally, the dynamic load coupling access point optimization model is solved to obtain the final access point selection; wherein, the objective function of the dynamic load coupling access point optimization model is expressed as follows: ， in, Indicates The impedance of the line, is the current flowing through the line. This current is related to the node load characteristic representation. The model determines the load of each node through the node load characteristic representation and calculates the power flow and loss on the line accordingly. In the final access point selection, a load-topology adaptive optimization model is constructed to minimize the energy loss of the medium-voltage distribution network by dynamically adjusting the line. At the same time, a dynamic line reconstruction mechanism is introduced in the dynamic adjustment line to dynamically optimize the enabled or disabled state of the line by adjusting the elements in the connection matrix to achieve balanced load distribution and reduce energy loss under different load conditions. In order to prevent grid instability, voltage constraints and line capacity constraints are further added, and finally the optimized topology structure is output; wherein, the objective function of the load-topology adaptive optimization model is expressed as follows: ， in, Indicates line Impedance; Indicates that the line The current; represents the regularization term, which is used to suppress the frequent switching of the line; Represents the weight of the regularization term, controlling the flexibility of topology adjustment; Construct an adaptive dynamic load-topology adjustment model based on the real-time load fluctuation and topology structure adjustment of the distribution network, combine the load forecasting, access point optimization and topology optimization results, and improve the operation efficiency of the power grid and ensure the stability of the medium-voltage distribution network by adaptively adjusting the access points and power transmission paths; Carry out long-term optimization and system evaluation, combine the peak and valley fluctuation characteristics of the load of the medium-voltage distribution network, and carry out load management and energy transmission efficiency through long-term evaluation and feedback optimization.

2. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 1, characterized in that: The preprocessing includes Z-score processing, missing value filling processing and maximum-minimum normalization processing; The Z-score processing is calculated based on the mean and standard deviation, and determines whether it is an outlier by setting a threshold. At the same time, the detected outliers are corrected by weighted average based on adjacent time points; The missing value filling process adopts a dynamic time backtracking method to fill the missing values ​​using the pattern similarity of the load data at previous moments; The maximum-minimum normalization process normalizes all data to within the scope of The feature expansion process constructs interactive features by performing feature engineering on external factors; The time series construction will use the sliding window method with a window length of Set to 96 time points for construction.

3. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 1, characterized in that: The time series data containing load data characteristics are divided into training set, validation set and test set, and the weighted mean square error is used to construct the loss function to adapt to the non-uniformity of peak and valley load changes.

4. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 1, characterized in that: The voltage constraint is expressed as follows: ， in, Indicates access node The voltage, and are the minimum and maximum voltage values ​​allowed, and N represents the total number of nodes; The line capacity constraint is expressed as follows: ， in, Indicates that the line The power, Indicates the maximum capacity of the line, and M indicates the total number of lines.

5. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 4, characterized in that: The dynamic load coupling access point optimization model is solved by mixed integer linear programming, and the selection of access points is set as a binary variable. , indicating that at the node Whether to connect to distributed power supply or load is indicated as follows: ， in, Indicates the final access point selection.

6. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 1, characterized in that: The dynamic line reconstruction mechanism is expressed as follows: ， in, Indicates the activation status between lines; The status of each line is determined by load forecasting and access points, and is adjusted in real time through topology optimization. When a load peak occurs, low-impedance lines are activated to reduce transmission losses, while some high-impedance lines are disabled.

7. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 6, characterized in that: The load-topology adaptive optimization model introduces a topology change penalty term in the objective function according to the fact that too frequent topology reconstruction leads to unstable system operation, which is used to control the frequent adjustment of the topology structure, and is expressed as follows: ， in, Represents the connection matrix before and after topology adjustment The difference is the change of the circuit breaker status; It represents the penalty coefficient that controls the frequency of topology adjustment.

8. The method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 1, characterized in that: The adaptive dynamic load-topology adjustment model minimizes the comprehensive operation loss of the system by adjusting the topology structure and access point selection in real time. The objective function of the adaptive dynamic load-topology adjustment model is expressed as follows: ， in, Indicates line Impedance, controlling energy loss in power transmission; Indicates that the line The current depends on the current load And the line status after topology adjustment ; Indicates the deviation between the current load forecast value and the actual load value; represents the penalty term, which is used to suppress frequent topology adjustments and control the stability of the system; and Indicates the adjustment parameter.

9. The method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 1, characterized in that: The long-term optimization objective function of the long-term optimization and system evaluation framework is expressed as follows: ， in, Indicates time The energy loss at each moment reflects the impact of load peak and valley fluctuations on energy consumption in the power grid after topology adjustment. represents the adjustment cost, which indicates the switching and reconfiguration cost of the system due to frequent topology reconfiguration and access point adjustment, especially considering the high-frequency adjustment caused by load fluctuation in the medium-voltage distribution network; represents the long-term stability constraint; , and are weight parameters respectively, which balance energy consumption, adjustment cost and long-term stability according to the actual operation requirements of the power grid.

10. A method for determining medium voltage distribution network access based on load peak-valley coupling according to claim 9, characterized in that: The long-term optimization and system evaluation framework designs voltage balance and load stability evaluation mechanisms for long-term operation. , which is expressed as follows: ， in, and Respectively represent nodes In time The maximum and minimum voltage values ​​within.

Images