Power distribution network reconstruction optimization method and system, computer equipment and storage medium

Abstract

The invention relates to the technical field of power distribution network reconstruction, and discloses a power distribution network reconstruction optimization method and system, computer equipment and a storage medium, and the method comprises the steps: building a power distribution network reconstruction model with the minimization of the total operation cost of a power distribution network as a target function and the power grid security constraint as a constraint condition; adopting an improved random branch strategy to calculate an optimal strategy weight parameter, and configuring the optimal strategy weight parameter to a model optimization solver; and solving the power distribution network reconstruction model by adopting the configured model optimization solver to obtain an optimal reconstruction scheme of the power distribution network. According to the method, the search space of the branch and bound algorithm can be reduced, the solving efficiency of the reconstruction model is remarkably improved, accurate and efficient power distribution network reconstruction optimization is achieved, the high efficiency and reliability of a power system under various operation conditions are ensured, and a solution with higher adaptability and high efficiency is provided for power distribution network reconstruction optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network reconstruction, and in particular to a distribution network reconstruction optimization method, system, computer equipment and storage medium. Background Art

[0002] Distribution network reconfiguration technology refers to the use of technology to improve reliability, reduce line losses, balance loads, and improve voltage quality by changing the topology of the distribution network. It is an important optimization method in power systems. Currently, the optimization problem of distribution network reconfiguration is usually modeled as a mixed integer quadratic programming model, and a solver is used to solve the model to achieve distribution network reconfiguration optimization. However, with the increase in renewable energy penetration and the expansion of the distribution network scale, traditional solution methods have encountered multiple bottlenecks in solving the problem, which limits the application of the model in large-scale, high-complexity scenarios. Specifically, on the one hand, as the scale of the distribution network increases, the complexity of the model increases, and the high-complexity model leads to a sharp increase in the amount of computation, which increases the difficulty of solving the model. On the other hand, traditional solution methods mostly rely on fixed heuristic rules and lack the ability to dynamically adapt to the characteristics of the problem, which prolongs the solution time and makes it difficult to meet the current real-time optimization requirements of distribution network reconfiguration. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention provides a distribution network reconstruction optimization method, system, computer equipment and storage medium, which can solve the computational bottleneck problem of traditional solution methods when dealing with high-complexity and large-scale distribution network reconstruction problems, thereby improving the solution efficiency of the model, meeting the real-time nature of distribution network reconstruction optimization, and ensuring the high efficiency and reliability of power system operation.

[0004] In a first aspect, the present invention provides a distribution network reconstruction optimization method, the method comprising: A distribution network reconstruction model is established with the minimization of the total operating cost of the distribution network as the objective function and the grid security constraint as the constraint condition; Using an improved random branching strategy, calculating optimal strategy weight parameters, and configuring the optimal strategy weight parameters to a preset model optimization solver, wherein the model optimization solver is a solver using a branch and bound algorithm; Using the configured model optimization solver to solve the distribution network reconstruction model, and obtain the optimal reconstruction plan of the distribution network; The step of adopting the improved random branching strategy to calculate the optimal strategy weight parameters includes: Acquiring real-time multi-source data of the distribution network, and performing data processing and feature extraction on the real-time multi-source data to obtain a feature vector; Inputting the feature vector into a preset strategy weight prediction model to obtain initial strategy weight parameters, wherein the strategy weight prediction model is constructed based on an extreme random tree model; According to the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameters.

[0005] Furthermore, with the minimization of the total operating cost of the distribution network as the objective function and the grid security constraint as the constraint condition, the steps of establishing the distribution network reconstruction model include: Obtaining a total operating cost of the distribution network based on the power generation cost, line loss cost, and switch operation cost, and taking minimization of the total operating cost of the distribution network as the objective function; The constraints include power balance constraint, power generation output constraint, line power flow constraint, voltage amplitude constraint, switch operation constraint, load dispatch constraint and minimum opening and closing time constraint; A distribution network reconstruction model is obtained according to the objective function and the constraint conditions.

[0006] Furthermore, the objective function is expressed as follows: Where E represents the objective function, represents the minimization function, t represents the time period, T represents the time period set, g represents the generator set, G represents the generator set, C g Indicates the power generation cost of the generator set per unit time, P g,t represents the power generated by the generator set in time period t, represents a line, L represents a line set, represents the loss cost coefficient of the line, represents the power flow of the line in time period t, k represents the switching device, K represents the set of switching devices, S k represents the closing operation cost of the switchgear, represents the decision variable for the closing operation of the switching device in time period t, R k represents the cost of breaking operations of the switchgear, Represents the decision variable for the breaking operation of the switching device in time period t.

[0007] Furthermore, the step of acquiring real-time multi-source data of the distribution network, and performing data processing and feature extraction on the real-time multi-source data to obtain a feature vector includes: Acquire real-time multi-source data of the distribution network and pre-process the real-time multi-source data, wherein the real-time multi-source data includes load data, line data, and topology data, and the pre-processing includes missing exception processing and data standardization; Feature extraction is performed on the pre-processed real-time multi-source data to obtain a feature vector, which includes load features, line features, topology features, and structural features.

[0008] Furthermore, the training steps of the strategy weight prediction model include: Acquire historical multi-source data of the distribution network, and perform data processing and feature extraction on the historical multi-source data to obtain a historical feature vector; Constructing a strategy candidate pool including multiple branch strategies based on the historical multi-source data, and initializing strategy weight parameters corresponding to each branch strategy; Based on each set of strategy weight parameters, a branch and bound algorithm is run in multiple preset distribution network reconstruction scenarios to obtain the exploration cost under each distribution network reconstruction scenario; Calculating an expected exploration cost based on the exploration cost, and taking minimization of the expected exploration cost as an optimization goal; The optimization objective is solved by using a Bayesian optimization algorithm to obtain the historical optimal strategy weight parameters; A data set is constructed based on the historical feature vectors and the historical optimal strategy weight parameters, and the strategy weight prediction model is trained using the data set.

[0009] Furthermore, the step of determining the optimal strategy weight parameter by adopting a dynamic weight fusion mechanism based on the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter includes: Calculating feature similarities between the feature vector and each historical feature vector, selecting a historical feature vector corresponding to the feature vector based on the feature similarities, and determining corresponding historical optimal strategy weight parameters based on the corresponding historical feature vectors; The initial strategy weight parameter and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter.

[0010] Furthermore, the step of performing weighted summation of the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter to obtain the optimal strategy weight parameter includes: Determining a weight fusion coefficient based on a ratio between a preset baseline exploration cost and a current exploration cost corresponding to the initial strategy weight parameter; According to the weight fusion coefficient, the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter.

[0011] In a second aspect, the present invention provides a distribution network reconstruction optimization system, the system comprising: A model building module is used to establish a distribution network reconstruction model with the minimization of the total operating cost of the distribution network as the objective function and the grid security constraint as the constraint condition; A weight configuration module is used to calculate the optimal strategy weight parameters using an improved random branching strategy, and configure the optimal strategy weight parameters to a preset model optimization solver, wherein the model optimization solver is a solver using a branch and bound algorithm; An optimization solution module, configured to solve the distribution network reconstruction model using a configured model optimization solver to obtain an optimal distribution network reconstruction solution; The weight configuration module is further configured to obtain real-time multi-source data of the power distribution network, and perform data processing and feature extraction on the real-time multi-source data to obtain a feature vector; Inputting the feature vector into a preset strategy weight prediction model to obtain initial strategy weight parameters, wherein the strategy weight prediction model is constructed based on an extreme random tree model; According to the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameters.

[0012] In a third aspect, an embodiment of the present invention further provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0013] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0014] The present invention provides a distribution network reconfiguration optimization method, system, computer device, and storage medium. By constructing a strategy candidate pool and accurately predicting and dynamically adjusting strategy weight parameters based on Bayesian optimization and extreme random tree models, the present invention can reduce the search space of the branch-and-bound algorithm and significantly improve the efficiency of solving the reconfiguration model, thereby achieving accurate and efficient distribution network reconfiguration optimization. The present invention ensures the efficiency and reliability of the power system under various operating conditions, providing a more adaptable and efficient solution for distribution network reconfiguration optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 1 is a flow chart of a distribution network reconstruction optimization method according to an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the distribution network reconstruction optimization system according to an embodiment of the present invention; Figure 3 is an internal structural diagram of a computer device according to an embodiment of the present invention; Reference numerals: 10. Model building module; 20. Weight configuration module; 30. Optimization solution module. DETAILED DESCRIPTION

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0017] See also Figure 1 A distribution network reconstruction optimization method proposed in the first embodiment of the present invention includes steps S10 to S30: Step S10, establishing a distribution network reconstruction model with minimization of the total operating cost of the distribution network as the objective function and grid security constraints as the constraints; Step S20: using an improved random branching strategy to calculate optimal strategy weight parameters, and configuring the optimal strategy weight parameters to a preset model optimization solver, wherein the model optimization solver is a solver using a branch and bound algorithm; Step S30: using the configured model optimization solver to solve the distribution network reconstruction model and obtain an optimal distribution network reconstruction solution.

[0018] In the present invention, the main goal of distribution network reconstruction is to optimize the distribution network operation structure by changing the switch state in the network, thereby minimizing the total operating cost of the distribution network. In addition to minimizing the total operating cost, the distribution network reconstruction also needs to meet the safety constraints of the grid operation. Therefore, the steps of constructing the distribution network reconstruction model in this embodiment include: Obtaining a total operating cost of the distribution network based on the power generation cost, line loss cost, and switch operation cost, and taking minimization of the total operating cost of the distribution network as the objective function; The constraints include power balance constraint, power generation output constraint, line power flow constraint, voltage amplitude constraint, switch operation constraint, load dispatch constraint and minimum opening and closing time constraint; A distribution network reconstruction model is obtained according to the objective function and the constraint conditions.

[0019] In this embodiment, the total operating cost of the distribution network includes power generation cost, line loss cost, and switch operation cost. Based on minimizing the total operating cost of the distribution network, the objective function shown below can be obtained: Where E represents the objective function, represents the minimization function, t represents the time period, T represents the time period set, g represents the generator set, G represents the generator set, C g Indicates the power generation cost of the generator set per unit time, P g,t represents the power generated by the generator set in time period t, represents a line, L represents a line set, represents the loss cost coefficient of the line, represents the power flow of the line in time period t, k represents the switching device, K represents the set of switching devices, S k represents the closing operation cost of the switchgear, represents the decision variable for the closing operation of the switching device in time period t, R k represents the cost of breaking operations of the switchgear, Represents the decision variable for the breaking operation of the switching device in time period t.

[0020] The goal of distribution network reconfiguration is to minimize the total operating cost of the system. However, during the optimization process, a series of constraints need to be considered to ensure the safe, stable, and feasible operation of the system. Therefore, based on the consideration of grid operation safety, this embodiment adds multiple constraints to the above objective function, including: ① Power balance constraint The power balance constraint means that the power demand of each power node must be met, and all the generated power and load demands in the power grid, as well as the power flow of the lines, must satisfy the power balance relationship. The specific expression is: in, represents the power generation of the i-th power node in time period t, is the load demand of the i-th power node in time period t, is the set of lines connected to power node i, is the power flow from power node i to power node j in time period t, N represents the set of power nodes, and T represents the set of time periods.

[0021] ② Generator output constraints Generator output constraint means that the generator power output cannot exceed its capacity upper limit or fall below its minimum output limit. The specific expression is: in, and are the minimum and maximum output of the generator respectively, is the output of the generator in time period t.

[0022] ③ Line power flow constraints Line power flow constraint means that in order to prevent overload, the power flow of each line must meet its maximum transmission capacity. The specific expression is: in, For the line The power flow in time period t is, For the line Maximum power transmission capability.

[0023] ④ Voltage amplitude constraint The voltage amplitude constraint means that the voltage amplitude of each node needs to be kept within a certain range to ensure the stable operation of the power grid. The voltage amplitude constraint can be expressed as: in, represents the voltage amplitude of power node i in time period t, and They are the minimum and maximum allowable voltage values ​​respectively.

[0024] ⑤Switch operation constraints Switching operation constraints refer to the constraints that node or line switching operations must follow during distribution network reconstruction. For line switching operations, the following constraints must be met: If a line is opened during time period t, the power flow through the line must be zero: If a line is closed in time period t, the power flow of the line is not zero and needs to satisfy the power flow constraint of the line: in, For the line The switching state in time period t, Indicates line Be disconnected, Indicates line connect.

[0025] ⑥ Load scheduling constraints Load scheduling constraints mean that the load of the power grid must be scheduled according to actual demand. Load scheduling constraints usually take into account load volatility and forecast errors, and are specifically in the form of: in, is the load demand of the i-th power node in time period t, is the maximum load demand of the i-th power node in time period t.

[0026] ⑦Minimum opening and closing time constraints Minimum opening and closing time constraints mean that in order to avoid frequent switching operations, the reconstruction of the distribution network also requires limiting the operating frequency of the switchgear. Generally speaking, the minimum opening and closing time interval of the switchgear is a specific value to prevent excessive operation. The specific form is: in, is the time interval between opening and closing of the switching device, It is the minimum time interval between opening and closing.

[0027] Based on the aforementioned objective function and constraints, a distribution network reconstruction model is constructed. By solving the distribution network reconstruction model, a distribution network reconstruction solution can be obtained. The conventional approach to solving the distribution network reconstruction model is to model the distribution network reconstruction optimization problem as a mixed integer quadratic programming (MIQP) problem and then solve it using a branch-and-bound algorithm. However, as the scale of distribution networks continues to expand, the distribution network reconstruction problem involves a large number of integer variables and complex constraints. The corresponding MIQP model also grows in size, significantly increasing the difficulty of solving the model. This causes traditional branch-and-bound algorithms to face the problem of a large search tree and slow solution speed during the solution process.

[0028] To improve the model solving efficiency, this embodiment provides a method for optimizing the strategy weight of the branch and bound algorithm based on an improved random branching strategy. By calculating the optimal strategy weight parameter and configuring it in the algorithm solver, the model solving efficiency of the solver is improved. The specific steps for calculating the optimal strategy weight parameter include: Acquiring real-time multi-source data of the distribution network, and performing data processing and feature extraction on the real-time multi-source data to obtain a feature vector; Inputting the feature vector into a preset strategy weight prediction model to obtain initial strategy weight parameters, wherein the strategy weight prediction model is constructed based on an extreme random tree model; According to the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameters.

[0029] In this embodiment, the operating data of the current distribution network is first obtained, and data processing and feature extraction are performed to obtain the feature vector of the distribution network. The operating data of the distribution network is multi-source data, including load data, line data and topology data, such as load data and power data of power nodes, line branch parameters, topology of the distribution network, switch status in the line, etc. For these real-time multi-source data, in addition to feature extraction, missing exception processing and data standardization are also required. For missing data, missing values ​​are filled according to different data types. For example, for continuous data such as load data and branch parameters, interpolation can be used to fill missing values ​​to ensure data continuity. For discrete data such as switch status, mode filling is used, that is, the missing values ​​are filled with the value with the highest frequency in the field. For abnormal data that exceeds the reasonable range, the abnormal values ​​can be adjusted according to the actual physical meaning, or smoothed by sliding average, filtering, etc., so as to eliminate invalid data caused by measurement errors or temporary failures and improve data reliability. Then the processed data is standardized, for example, by using the minimum-maximum normalization method to scale the data to the [0,1] interval, thereby eliminating dimensional differences. Finally, the pre-processed real-time multi-source data is subjected to feature extraction to obtain a feature vector. In this embodiment, the extracted features include load features, line features, topological features, and structural features, wherein the load features include node load demand, peak load proportion, load fluctuation rate, etc., the line features include line impedance, active power, reactive power, current value, etc., the topological features include node degree, connection relationship, number of loops, number of connected subgraphs, etc., and the structural features include comprehensive load index and branch utilization rate, wherein the comprehensive load index is a weighted index generated by combining the historical load curve with the real-time demand, that is, the weighted sum of the historical load and the real-time demand load, and the branch utilization rate is the ratio of the actual power flow to the maximum capacity of the line, reflecting the line load status. It should be noted here that this embodiment provides preferred features, not specific limitations. In actual applications, correlation analysis or feature importance evaluation can be used to select key features.

[0030] This embodiment enhances the modeling capability of the model for complex constraints of the distribution network by constructing a high-dimensional feature matrix containing original features and constructed features. The high-dimensional feature matrix composed of various feature vectors will be input into a preset strategy weight prediction model to perform strategy weight prediction. The strategy weight prediction model is preferably constructed using an extreme random tree model. In order to improve the prediction accuracy of the strategy weight prediction model, this embodiment provides a method for training the strategy weight prediction model. The specific steps include: Acquire historical multi-source data of the distribution network, and perform data processing and feature extraction on the historical multi-source data to obtain a historical feature vector; Constructing a strategy candidate pool including multiple branch strategies based on the historical multi-source data, and initializing strategy weight parameters corresponding to each branch strategy; Based on each set of strategy weight parameters, a branch and bound algorithm is run in multiple preset distribution network reconstruction scenarios to obtain the exploration cost under each distribution network reconstruction scenario; Calculating an expected exploration cost based on the exploration cost, and taking minimization of the expected exploration cost as an optimization goal; The optimization objective is solved by using a Bayesian optimization algorithm to obtain the historical optimal strategy weight parameters; A data set is constructed based on the historical feature vectors and the historical optimal strategy weight parameters, and the strategy weight prediction model is trained using the data set.

[0031] The key to training the strategy weight prediction model in this embodiment lies in the construction of a dataset. The dataset used for model training includes feature data and corresponding label values. The feature data is constructed from historical multi-source data of the distribution network. That is, historical multi-source data of the distribution network is obtained, and historical feature vectors are obtained through data processing and feature extraction. The specific steps are the same as those for processing real-time multi-source data and will not be repeated here. The label values ​​in the dataset are calculated using a strategy candidate pool based on Bayesian optimization.

[0032] Specifically, based on historical multi-source data, that is, based on the historical operating status of the distribution network, a strategy candidate pool for the distribution network reconstruction problem is constructed. The strategy candidate pool includes multiple branch strategies. These branch strategies cover a variety of strategies for dealing with complex constraints in the distribution network reconstruction problem. The strategy candidate pool includes at least strong branch strategy, pseudo-cost branch strategy, optimal constraint branch strategy and hybrid branch strategy. Among them, the strong branch strategy gives priority to variables that have a significant impact on the objective function, such as the line switch status; the pseudo-cost branch strategy predicts the branch effect based on historical pseudo-cost, such as the charging and discharging decision of energy storage equipment; the optimal constraint branch strategy selects variables for constraints that are seriously out of limit, such as voltage amplitude exceeding the limit; and the hybrid branch strategy is a dynamic combination of the above strategies. Then, the weights of each strategy in the strategy candidate pool are initialized. The initialization steps include: presetting the strategy weight parameters of each branch strategy, then traversing all possible weight combinations, and finally screening the combinations that meet the normalization conditions to obtain a strategy weight parameter set, that is, the set contains multiple groups of strategy weight parameters, each group of strategy weight parameters corresponds to a branch strategy combination, where the normalization condition is that the sum of the weight parameters of each strategy in the combination is 1, and the strategy weight parameters reflect the priority of the strategy. Assume that each type of branch strategy in the strategy candidate pool contains multiple strategies, and each branch strategy has a different strategy weight parameter preset, such as the weight parameter of strong branch strategy 1 is 0.5, the weight parameter of strong branch strategy 2 is 0.6, the weight parameter of pseudo-cost branch strategy 1 is 0.4, the weight parameter of pseudo-cost branch strategy 2 is 0.3, the weight parameter of optimal constraint branch strategy 1 is 0.2, the weight parameter of optimal constraint branch strategy 2 is 0.1, the weight parameter of hybrid branch strategy 1 is 0.1, the weight parameter of hybrid branch strategy 2 is 0.3, etc. Then, the strategy combinations that meet the normalization conditions are screened out, such as combination 1 includes: strong branch strategy 1, pseudo-cost branch strategy 1, hybrid branch strategy 1; combination 2 includes: strong branch strategy 1, pseudo-cost branch strategy 2, optimal constraint branch strategy 2 and hybrid branch strategy 1; combination 3 includes: strong branch strategy 2 and pseudo-cost branch strategy 1. In other words, the strategy combinations in this embodiment are not required to cover all strategy types, that is, the number of strategy types in the combination can be dynamically adjusted to balance search efficiency and strategy diversity. Preferably, the initialization of the strategy weight parameters can be achieved by using a grid search algorithm or a Bayesian optimization algorithm.

[0033] For each set of policy weight parameters in the initialized policy candidate pool, a branch-and-bound algorithm is run in multiple pre-set distribution network reconfiguration scenarios. These scenarios refer to distribution networks of varying sizes, including small-scale (e.g., a 33-node distribution network), medium-scale (e.g., a 269-node distribution network), and large-scale (e.g., a 1,362-node distribution network). Multiple instances of the distribution network reconfiguration optimization problem are generated for each scenario to simulate different operating conditions. For each set of policy weight parameters in the policy candidate pool, the parameters are configured in the branch-and-bound solver and the algorithm is executed, i.e., an independent tree search experiment is performed. To ensure statistical accuracy, multiple independent tree search experiments are performed for each set of policy weight parameters, each simulating a different network topology optimization scenario.

[0034] When executing a branch-and-bound algorithm in a distribution network reconfiguration scenario, branch nodes typically correspond to specific switch state combinations or topological configurations in network topology optimization. Therefore, when the algorithm reaches a branch node, a random number is used to determine the branching strategy for that node. Specifically, a random number U is generated from a uniform distribution, ranging from (0, 1). The probability intervals for each branching strategy in each set of strategy weight parameters are determined by its strategy weight parameter. The corresponding branching strategy is selected based on the probability interval that U falls within. Taking combination 1 as an example, the strategy weights for strong branching strategy 1, pseudo-cost branching strategy 1, and hybrid branching strategy 1 are 0.5, 0.4, and 0.1, respectively. By setting the probability intervals for each branching strategy based on its strategy weight, we can conclude that when the random number U is in the range [0, 0.5), strong branching strategy 1 is selected; when the random number U is in the range [0.5, 0.9), pseudo-cost branching strategy 1 is selected; and when the random number U is in the range [0.9, 1], hybrid branching strategy 1 is selected.

[0035] Based on the selected branching strategy, the integer variable of the current branch node is branched to generate child nodes. The total number of nodes generated by the branch path is then recorded, which is the total exploration cost of the branch and bound algorithm under this branch path. For each simulation experiment, it is necessary to record the total number of nodes, solution time, and strategy weight parameter combination used in the branch path. For each set of strategy weight parameters, the average number of nodes in different instances is calculated, and the average number of nodes is used as the expected exploration cost. For example, if combination 1 generated an average of 1500 nodes in 50 experiments, the expected exploration cost of this combination is 1500. Through experiments, a strategy performance mapping table can be obtained, which contains the expected exploration cost and average solution time for each set of strategy weight parameters in different scenarios.

[0036] In order to iteratively optimize the policy weights, this embodiment takes the minimization of the expected exploration cost as the optimization goal, adopts the Bayesian optimization algorithm for solution, uses the mapping table as the input data of the Bayesian optimization algorithm, and drives the weight parameters to be iteratively updated. Through the Bayesian optimization algorithm, the optimal weight that meets the optimization goal can be found, that is, the historical optimal policy weight parameters under each scenario are obtained. When performing the Bayesian optimization iteration, based on the data in the mapping table, a Gaussian process posterior distribution model of the optimization goal is established, and the expected improvement is used as the acquisition function to select the potential optimal weights. Then, the new weight combination is evaluated through each round of iteration, and the Gaussian process model is updated until the model converges. Finally, the optimal policy weight combination under the scenario is obtained, that is, the historical optimal policy weight parameters are obtained. Since the above-mentioned historical feature vectors also represent a certain distribution network scenario, the historical feature vectors are used as feature data, and the historical optimal policy weight parameters under the corresponding scenario are used as label data to construct a data set for training the policy weight prediction model.

[0037] In this example, the policy weight prediction model is constructed using an extreme randomized tree model. By introducing random feature selection and a node splitting mechanism, the extreme randomized tree model can effectively process high-dimensional data and reduce the risk of overfitting, thereby improving classification accuracy. The policy weight prediction model is then trained using the aforementioned training set. The specific training steps can be found in the conventional model training steps and will not be detailed here.

[0038] According to the above-mentioned data set construction process, when reconstructing and optimizing the current distribution network, it can be understood as processing a new problem instance. Therefore, by inputting the feature vector generated by real-time multi-source data into the trained strategy weight prediction model, the corresponding strategy weight parameters can be predicted. In order to ensure the efficiency and adaptability of the strategy weight parameters in practical applications, this embodiment introduces a parameter dynamic optimization mechanism. Through the corresponding historical optimal strategy weight parameters, the initial strategy weight parameters predicted by the model are weighted and fused, thereby realizing dynamic optimization of the parameters. When selecting the corresponding historical optimal strategy weight parameters, the historical optimal strategy weight parameters under the scenario can be selected according to the scenario corresponding to the current feature vector. The specific steps include: Calculating feature similarities between the feature vector and each historical feature vector, selecting a historical feature vector corresponding to the feature vector based on the feature similarities, and determining corresponding historical optimal strategy weight parameters based on the corresponding historical feature vectors; The initial strategy weight parameter and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter.

[0039] In this embodiment, the feature similarity between the feature vector and each historical feature vector is first calculated. The feature similarity can be represented by Euclidean distance or cosine distance. Then, the historical feature vector corresponding to the maximum value of the feature similarity is selected, and the historical optimal strategy weight parameter corresponding to the historical feature vector is used as the historical optimal strategy weight parameter of the feature vector. Finally, according to the preset weight fusion coefficient, the feature vector and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter. Its expression is: Where w f represents the optimal strategy weight parameter, β represents the weight fusion coefficient, w p Represents the initial strategy weight parameter, w h Represents the historical optimal strategy weight parameter.

[0040] It should be noted that, since the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters in this embodiment are actually weight parameter combinations, there may be inconsistencies in the branch strategies contained in these two combinations. When performing weighted summation, if a branch strategy in the initial strategy weight parameters does not exist in the historical optimal strategy weight parameters, the historical weight parameters corresponding to the branch strategy are set to zero, that is, when performing weighted fusion, the consistency of the branch strategies needs to be maintained. The branch strategy here refers to the branch strategy type. For example, the predicted parameter combination contains strong branch strategy 1, and the historical parameter combination contains strong branch strategy 2. At this time, both strategies are strong branch strategies. Therefore, weighted fusion can be performed. If the historical parameter combination does not contain any strong branch strategies, the w corresponding to the strong branch strategy 1 is set to zero. h Set to zero. Due to the inconsistency of branch strategies, the optimal strategy weight parameters obtained after weighting may not satisfy the situation where the sum of the weights is 1. Therefore, after weighted fusion, the weighted strategy weight parameters need to be normalized so that the final optimal strategy weight parameters meet the normalization conditions.

[0041] In another preferred embodiment, in order to balance the fusion ratio of prediction weights and historical weights, the present invention also provides a method for dynamically adjusting the weight fusion coefficient, which specifically includes: Determining a weight fusion coefficient based on a ratio between a preset baseline exploration cost and a current exploration cost corresponding to the initial strategy weight parameter; According to the weight fusion coefficient, the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter.

[0042] In this embodiment, the weight fusion coefficient is dynamically calculated based on the performance of the real-time solution. The performance is expressed in terms of detection cost. The weight fusion coefficient can be expressed as: Where C b represents the baseline exploration cost, C c represents the current exploration cost, To prevent division by zero, set a very small constant, such as The baseline exploration cost is determined based on the average number of nodes in the current scenario, which is the expected exploration cost in the above dataset.

[0043] Since the weight fusion coefficient should be less than or equal to 1, when the current exploration cost is less than the baseline exploration cost, β is 1. In this case, the optimal strategy weight parameter is the initial strategy weight parameter. When β is small, the stability of the algorithm solution can be guaranteed by the historical weight. After obtaining the optimal strategy weight parameter, it is configured into the heuristic algorithm and pruning strategy of the model optimization solver to guide the solution process, thereby improving the efficiency and quality of the solution. The model optimization solver can use the Gurobi solver or the CPLEX solver, and is not specifically limited here.

[0044] Furthermore, this embodiment updates the strategy weight prediction model through online learning, such as recording the performance of the strategy weight parameters in terms of exploration cost and solution time as feedback information, and regularly uses the newly collected feedback information to perform incremental training on the strategy weight prediction model to improve the adaptability and accuracy of the model.

[0045] The effectiveness of the optimization method provided by the present invention is verified through comparative experiments. The experiment uses several typical distribution network reconstruction scenarios and generates multiple MIQP models under different network scales, device numbers, and renewable energy access ratios to comprehensively evaluate the performance of the present invention in terms of solution efficiency and performance improvement.

[0046] The experiment mainly includes the following scenarios: Scenario 1: Small-scale distribution network reconstruction. This scenario features varying numbers of generators and energy storage devices, and considers various renewable energy integration ratios. This scenario primarily serves to verify the effectiveness and adaptability of the method in smaller-scale networks.

[0047] Scenario 2: Medium-scale distribution network reconstruction. This scenario covers a variety of complex distribution network topologies, introduces different energy storage configurations, and incorporates a high proportion of renewable energy. This scenario is used to test the method's performance in medium-scale networks and its ability to handle complex constraints.

[0048] Scenario 3: Large-Scale Distribution Network Reconfiguration. This scenario involves a larger number of devices, a higher proportion of renewable energy access, and a more complex network topology. This scenario aims to evaluate the efficiency and performance of the method when dealing with large-scale, highly complex networks.

[0049] The experiment mainly compares the solution time of two solution methods: one is to use the traditional optimization solver (Gurobi) to directly solve the MIQP model, and the other is to use the optimization method provided by this invention to solve it, and records the solution time of the two respectively.

[0050] For scenario 1, the distribution network parameters for Example 1 were set as follows: 33 nodes, 13 generators, 126 switches, and 5 energy storage devices. The experimental results showed that the Gurobi solver took 298.7 seconds to solve, while the proposed method took 218.3 seconds.

[0051] For scenario 2, the distribution network parameters for Example 2 were set as follows: 269 nodes, 127 generators, 3898 switches, and 34 energy storage devices. The experimental results showed that the Gurobi solver took 598.4 seconds to solve, while the proposed method took 348.9 seconds to solve.

[0052] For scenario three, the distribution network parameters for Example 3 were set as follows: 1,362 nodes, 378 generators, 26,672 switches, and 131 energy storage devices. The experimental results showed that the Gurobi solver took 1,498.5 seconds to solve, while the proposed method took 898.2 seconds.

[0053] According to the above experimental results, the optimization method provided by the present invention shows better solution efficiency than traditional optimization solvers in various calculation examples, and as the network scale, the number of generator sets and the number of energy storage devices increase, the advantages of the optimization method of the present invention become more obvious.

[0054] See also Figure 2 Based on the same inventive concept, a distribution network reconstruction optimization system proposed in a second embodiment of the present invention includes: A model building module 10 is used to establish a distribution network reconstruction model with minimization of the total operating cost of the distribution network as the objective function and grid security constraints as the constraints; A weight configuration module 20 is used to calculate the optimal strategy weight parameters using an improved random branching strategy, and configure the optimal strategy weight parameters to a preset model optimization solver, wherein the model optimization solver is a solver using a branch and bound algorithm; An optimization solution module 30 is configured to solve the distribution network reconstruction model using a configured model optimization solver to obtain an optimal distribution network reconstruction solution; The weight configuration module 20 is further configured to obtain real-time multi-source data of the power distribution network, and perform data processing and feature extraction on the real-time multi-source data to obtain a feature vector; Inputting the feature vector into a preset strategy weight prediction model to obtain initial strategy weight parameters, wherein the strategy weight prediction model is constructed based on an extreme random tree model; According to the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameters.

[0055] The technical features and effects of the distribution network reconstruction and optimization system proposed in the embodiment of the present invention are the same as those of the method proposed in the embodiment of the present invention and are not further described here. Each module in the above-mentioned distribution network reconstruction and optimization system can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each of the above modules.

[0056] In addition, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0057] See also Figure 3 , an internal structure diagram of a computer device in one embodiment, the computer device can specifically be a terminal or a server. The computer device includes a processor, a memory, a network interface, a display, and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a distribution network reconstruction optimization method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0058] It can be understood by those skilled in the art that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have the same component arrangement.

[0059] In addition, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when the computer program is executed by a processor.

[0060] In summary, an embodiment of the present invention proposes a distribution network reconstruction optimization method, system, computer device and storage medium. The method takes minimizing the total operating cost of the distribution network as the objective function and the power grid security constraint as the constraint condition to establish a distribution network reconstruction model; adopts an improved random branching strategy to calculate the optimal strategy weight parameters, and configures the optimal strategy weight parameters to a preset model optimization solver, which is a solver using a branch and bound algorithm; adopts the configured model optimization solver to solve the distribution network reconstruction model to obtain the optimal reconstruction scheme of the distribution network; wherein, the step of using the improved random branching strategy to calculate the optimal strategy weight parameters includes: obtaining real-time multi-source data of the distribution network, and performing data processing and feature extraction on the real-time multi-source data to obtain a feature vector; inputting the feature vector into a preset strategy weight prediction model to obtain an initial strategy weight parameter, and the strategy weight prediction model is constructed based on an extreme random tree model; according to the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameter. The present invention constructs a strategy candidate pool and accurately predicts and dynamically adjusts the strategy weight parameters based on Bayesian optimization and extreme random tree models, effectively reducing the search space of the branch and bound algorithm and significantly improving the solution efficiency of the reconstruction model. The present invention overcomes the computational bottleneck of traditional methods in dealing with high-complexity and large-scale distribution network reconstruction problems. Through accurate and efficient distribution network reconstruction optimization, the efficiency and reliability of the power system under various operating conditions are ensured, providing a more adaptable and efficient solution for distribution network reconstruction optimization.

[0061] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the various technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0062] The above-described embodiments merely represent several preferred implementations of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art could make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the scope of protection of the claims.

Claims

1. A distribution network reconstruction optimization method, characterized in that: include: A distribution network reconstruction model is established with the minimization of the total operating cost of the distribution network as the objective function and the grid security constraint as the constraint condition; Using an improved random branching strategy, calculating optimal strategy weight parameters, and configuring the optimal strategy weight parameters to a preset model optimization solver, wherein the model optimization solver is a solver using a branch and bound algorithm; Using the configured model optimization solver to solve the distribution network reconstruction model, and obtain the optimal reconstruction plan of the distribution network; The step of adopting the improved random branching strategy to calculate the optimal strategy weight parameter includes: Acquiring real-time multi-source data of the distribution network, and performing data processing and feature extraction on the real-time multi-source data to obtain a feature vector; Inputting the feature vector into a preset strategy weight prediction model to obtain initial strategy weight parameters, wherein the strategy weight prediction model is constructed based on an extreme random tree model; According to the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameters.

2. The distribution network reconstruction optimization method according to claim 1, characterized in that: The steps of establishing a distribution network reconstruction model with minimization of the total operating cost of the distribution network as the objective function and grid security constraints as the constraints include: Obtaining a total operating cost of the distribution network based on the power generation cost, line loss cost, and switch operation cost, and taking minimization of the total operating cost of the distribution network as the objective function; The constraints include power balance constraint, power generation output constraint, line power flow constraint, voltage amplitude constraint, switch operation constraint, load dispatch constraint and minimum opening and closing time constraint; A distribution network reconstruction model is obtained according to the objective function and the constraint conditions.

3. The distribution network reconstruction optimization method according to claim 2, characterized in that: The objective function is expressed as follows: Where E represents the objective function, represents the minimization function, t represents the time period, T represents the time period set, g represents the generator set, G represents the generator set, C g Indicates the power generation cost of the generator set per unit time, P g,t represents the power generated by the generator set in time period t, represents a line, L represents a line set, represents the loss cost coefficient of the line, represents the power flow of the line in time period t, k represents the switching device, K represents the set of switching devices, S k represents the closing operation cost of the switchgear, represents the decision variable for the closing operation of the switching device in time period t, R k represents the cost of breaking operations of the switchgear, Represents the decision variable for the breaking operation of the switching device in time period t.

4. The distribution network reconstruction optimization method according to claim 1, characterized in that: The step of acquiring real-time multi-source data of the distribution network, and performing data processing and feature extraction on the real-time multi-source data to obtain a feature vector includes: Acquire real-time multi-source data of the distribution network and pre-process the real-time multi-source data, wherein the real-time multi-source data includes load data, line data, and topology data, and the pre-processing includes missing exception processing and data standardization; Feature extraction is performed on the pre-processed real-time multi-source data to obtain a feature vector, which includes load features, line features, topology features, and structural features.

5. The distribution network reconstruction optimization method according to claim 1, characterized in that: The training steps of the strategy weight prediction model include: Acquire historical multi-source data of the distribution network, and perform data processing and feature extraction on the historical multi-source data to obtain a historical feature vector; Constructing a strategy candidate pool including multiple branch strategies based on the historical multi-source data, and initializing strategy weight parameters corresponding to each branch strategy; Based on each set of strategy weight parameters, a branch and bound algorithm is run in multiple preset distribution network reconstruction scenarios to obtain the exploration cost under each distribution network reconstruction scenario; Calculating an expected exploration cost based on the exploration cost, and taking minimization of the expected exploration cost as an optimization goal; The optimization objective is solved by using a Bayesian optimization algorithm to obtain the historical optimal strategy weight parameters; A data set is constructed based on the historical feature vectors and the historical optimal strategy weight parameters, and the strategy weight prediction model is trained using the data set.

6. The distribution network reconstruction optimization method according to claim 5, characterized in that: The step of determining the optimal strategy weight parameter by adopting a dynamic weight fusion mechanism based on the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter comprises: Calculating feature similarities between the feature vector and each historical feature vector, selecting a historical feature vector corresponding to the feature vector based on the feature similarities, and determining corresponding historical optimal strategy weight parameters based on the corresponding historical feature vectors; The initial strategy weight parameter and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter.

7. The distribution network reconstruction optimization method according to claim 6, characterized in that: The step of performing weighted summation of the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter to obtain the optimal strategy weight parameter comprises: Determining a weight fusion coefficient based on a ratio between a preset baseline exploration cost and a current exploration cost corresponding to the initial strategy weight parameter; According to the weight fusion coefficient, the initial strategy weight parameter and the corresponding historical optimal strategy weight parameter are weighted and summed to obtain the optimal strategy weight parameter.

8. A distribution network reconstruction optimization system, characterized in that: include: A model building module is used to establish a distribution network reconstruction model with the minimization of the total operating cost of the distribution network as the objective function and the grid security constraint as the constraint condition; A weight configuration module is used to calculate the optimal strategy weight parameters using an improved random branching strategy, and configure the optimal strategy weight parameters to a preset model optimization solver, wherein the model optimization solver is a solver using a branch and bound algorithm; An optimization solution module, configured to solve the distribution network reconstruction model using a configured model optimization solver to obtain an optimal distribution network reconstruction solution; The weight configuration module is further configured to obtain real-time multi-source data of the power distribution network, and perform data processing and feature extraction on the real-time multi-source data to obtain a feature vector; Inputting the feature vector into a preset strategy weight prediction model to obtain initial strategy weight parameters, wherein the strategy weight prediction model is constructed based on an extreme random tree model; According to the initial strategy weight parameters and the corresponding historical optimal strategy weight parameters, a dynamic weight fusion mechanism is adopted to determine the optimal strategy weight parameters.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images