Multi-stage planning method, system and equipment for power distribution network and storage medium

Abstract

The invention relates to the technical field of monitoring analysis, and discloses a power distribution network multi-stage planning method, system and device and a storage medium, and the method comprises the steps: obtaining historical load data of a power distribution network, carrying out the feature extraction of the historical load data, and generating a medium and long term load prediction data set; performing feature extraction on the medium-and-long-term load prediction data set and performing multi-stage division by using a clustering algorithm to obtain an age limit division result of each planning stage; constructing a planning model according to the age limit division result; solving the planning model through an intelligent optimization algorithm to obtain a planning scheme; and simulation verification is carried out on each planning stage based on the planning scheme, a planning feasibility report is generated, and the planning efficiency and the planning quality are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of monitoring and analysis technology, and in particular to a method, system, equipment and storage medium for multi-stage planning of power distribution networks. Background Technology

[0002] Continued population growth and industrial market competition have led to increased electricity consumption and load demand. Power grid companies need to take responsibility for meeting this increased demand, and internal distribution network planners must develop corresponding incremental distribution network planning schemes to accommodate the continuous expansion of the distribution network system. Ultimately, the investment and construction of distribution networks within the distribution network system involves high costs, requiring careful planning and operation by power grid companies and investors. Distributed power sources play a pivotal role in identifying problems and providing economical and clean electricity in incremental distribution network planning.

[0003] In related technologies, the existing distribution network planning phase division relies on manual experience to preset fixed cycles, lacks a dynamic adjustment mechanism based on the load growth mutation rate, and is difficult to capture load mutation characteristics, thereby reducing planning efficiency and planning quality, and there are areas for improvement. Summary of the Invention

[0004] In view of the existing problems mentioned above, a multi-stage planning method, system, equipment and storage medium for power distribution networks is proposed.

[0005] Therefore, this invention provides a method, system, device, and storage medium for multi-stage planning of power distribution networks to solve the problems of existing power distribution network planning stages relying on manual experience and preset fixed periods, lacking a dynamic adjustment mechanism based on the rate of load growth mutation, making it difficult to capture load mutation characteristics, and reducing planning efficiency and quality.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a multi-stage planning method for a power distribution network, comprising: Historical load data of the distribution network is acquired, and features are extracted from the historical load data to generate a medium- and long-term load forecast dataset. Feature extraction is performed on the medium- and long-term load forecast dataset, and a clustering algorithm is used to divide it into multiple stages to obtain the year division results for each planning stage. Construct a planning model based on the stated year division results; The planning model is solved using an intelligent optimization algorithm to obtain a planning scheme; Based on the proposed planning scheme, simulation verification is performed for each planning stage, and a planning feasibility report is generated.

[0007] As a preferred embodiment of the multi-stage planning method for distribution networks described in this invention, generating a medium- to long-term load forecast dataset includes: Seasonal fluctuation characteristics were extracted from historical load data based on time series, and a load growth trend prediction model was constructed. Spatiotemporal correlation features are extracted from historical load data to obtain historical load spatiotemporal feature data; Historical load spatiotemporal characteristics data are input into the load growth trend prediction model to generate a medium- and long-term load prediction dataset.

[0008] The beneficial effects of this preferred technical solution are that it can accurately capture load change patterns, improve the accuracy of medium- and long-term load forecasting, and provide a reliable basis for distribution network planning.

[0009] As a preferred embodiment of the multi-stage planning method for power distribution networks described in this invention, the year division results for each planning stage include: Load growth characteristics are extracted from the medium- and long-term load forecast dataset, and the load growth rate for each time period is calculated using a linear regression algorithm based on a sliding window. Construct a set of load growth rate feature vectors based on the load growth rate corresponding to each time period; The initial cluster centers in the clustering algorithm are dynamically optimized using the silhouette coefficient method. Iterative clustering analysis is performed on each load growth feature vector in the load growth feature vector set, and the year division results for each planning stage are determined by combining the membership degree.

[0010] The beneficial effects of this preferred technical solution are that it can accurately divide the planning stages, effectively capture sudden load changes, and improve the scientific nature of the planning.

[0011] As a preferred embodiment of the multi-stage planning method for power distribution networks described in this invention, the iterative clustering analysis of each load growth rate feature vector in the load growth rate feature vector set includes: The load growth rate feature vector set is standardized, and principal component analysis is used to reduce the dimension of each load growth rate feature vector in the load growth rate feature vector set to generate a principal component vector set. Determine the initial range of cluster numbers and calculate the silhouette coefficient value corresponding to each cluster number. Based on the silhouette coefficient value corresponding to each cluster number, determine the optimal number of partitions in the planning stage. Initialize cluster centers, and calculate the similarity between each vector in the principal component vector set and the initial cluster centers using Mahalanobis distance to generate the initial distribution matrix of cluster centers; Iterative clustering analysis is performed based on the initial distribution matrix of cluster centers. The membership probability of each vector in the principal component vector set to the cluster of each planning stage is dynamically adjusted, and a dynamic membership threshold is set to filter high-confidence data samples. Based on the spatiotemporal distribution data corresponding to the high-confidence data samples, the boundary years for each planning stage are fitted and corrected, and the year division results for each planning stage are output.

[0012] The advantages of this preferred technical solution are that it can optimize the clustering effect, accurately determine the phase years, and enhance the adaptability of the planning.

[0013] As a preferred embodiment of the multi-stage planning method for power distribution networks described in this invention, the construction of the planning model includes: The objective function is to minimize the present value of the total cost of the distribution network over its entire lifecycle. The constraints include distributed power flow coupling constraints, node voltage opportunity constraints, candidate node installation capacity constraints, and capacity upper limit constraints, constructing a multi-dimensional constraint system. A planning model is constructed using an objective function and a multi-dimensional constraint system.

[0014] As a preferred embodiment of the multi-stage planning method for power distribution networks described in this invention, solving the planning model includes: Initialize the swarm size and number of iterations for the particle swarm optimization algorithm, and set the shrinkage factor to generate the initial solution set; Based on the objective function of the planning model, a multi-objective fitness function is constructed, and the fitness of each particle in the initial solution set is evaluated to obtain the fitness score of each particle and generate a particle fitness ranking list. The shrinkage factor is dynamically adjusted based on the particle fitness ranking list, and the historical optimal solution and global optimal solution of the particle swarm are iteratively updated based on the fitness score of each particle to generate an optimal solution set distribution cloud map. The population dispersion index of the optimized solution set distribution cloud map is monitored. When the population dispersion index is lower than the preset threshold, the chaotic mapping perturbation strategy is triggered and a chaotic sequence is generated. The particle positions in the high-density region are dynamically perturbed based on the chaotic sequence. The optimal compromise solution for each planning stage is identified from the optimal solution set distribution cloud map, and the planning scheme is output based on the optimal compromise solution; The planning scheme includes optimized access locations and capacity configuration schemes for distributed power sources at each planning stage.

[0015] As a preferred embodiment of the multi-stage planning method for power distribution networks described in this invention, generating a planning feasibility report includes: A multi-stage planning verification model is constructed based on the planning scheme, and the capacity configuration parameters and access node coordinate data corresponding to each planning stage are imported. Continuous power flow calculation is performed based on a multi-stage planning verification model to obtain the power flow distribution map, node voltage deviation data and line load rate index of the distribution network under each planning stage, and to generate an evaluation dataset. Based on the evaluation dataset, the planning schemes are quantitatively scored from multiple dimensions, and a comprehensive analysis and ranking are performed based on the scoring results to generate a planning feasibility report.

[0016] Secondly, the present invention provides a multi-stage planning system for a power distribution network, comprising: The feature extraction module is used to acquire historical load data of the distribution network, extract features from the historical load data, and generate a medium- and long-term load forecast dataset. The phase division module is used to extract features from the medium- and long-term load forecast dataset and use a clustering algorithm to divide it into multiple phases, so as to obtain the year division results for each planning phase. The model building module is used to build a planning model based on the year division results; The model solving module is used to solve the planning model using intelligent optimization algorithms to obtain the planning scheme; The report generation module is used to perform simulation verification for each planning stage based on the planning scheme and generate a planning feasibility report.

[0017] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the aforementioned multi-stage planning method for a power distribution network.

[0018] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the aforementioned multi-stage planning method for a power distribution network.

[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a planning model based on the year division results of each planning stage, with the minimization of the present value of the total cost of the distribution network throughout its entire life cycle as the core objective function. It then uses a particle swarm optimization algorithm with a shrinkage factor to solve the planning model and outputs optimized access locations and capacity configuration schemes for distributed power sources corresponding to each planning stage based on the solution results. Furthermore, it performs simulation verification for each planning stage based on these optimized access locations and capacity configuration schemes, generates a planning feasibility report, and distributes the report to the distribution automation system. This effectively reduces the occurrence of situations where a fixed period is preset based on manual experience, lacking a dynamic adjustment mechanism based on the rate of load growth mutation. It also effectively captures load mutation characteristics, thereby significantly improving planning efficiency and quality. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the overall process logic of a multi-stage planning method for power distribution networks provided in one embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Example 1, referring to Figure 1 As an embodiment of the present invention, a multi-stage planning method for a power distribution network is provided, comprising: S100: Acquire historical load data of the distribution network, extract features from the historical load data, and generate a medium- and long-term load forecast dataset; S200: Extract features from the medium- and long-term load forecast dataset and use clustering algorithms to divide it into multiple stages to obtain the year division results for each planning stage; S300: Construct a planning model based on the year division results; S400: Solve the planning model using intelligent optimization algorithms to obtain the planning scheme; S500: Based on the planning scheme, perform simulation verification for each planning stage and generate a planning feasibility report.

[0024] It should be noted that through scientific load forecasting, precise stage division, reasonable model construction, efficient algorithm solution, and rigorous simulation verification, the scientific nature and accuracy of distribution network planning have been achieved, effectively improving planning efficiency and quality, reducing planning costs, enhancing the adaptability and reliability of the distribution network, and providing a strong guarantee for the sustainable development of the distribution network.

[0025] In this embodiment of the invention, step S100 includes the following sub-steps A1-A3; In A1: Seasonal fluctuation characteristics are extracted from historical load data based on time series, and a load growth trend prediction model is constructed; In A2: Spatiotemporal correlation features are extracted from historical load data to obtain spatiotemporal feature data of historical load; In A3: Input historical load spatiotemporal characteristic data into the load growth trend prediction model to generate a medium- and long-term load prediction dataset.

[0026] Specifically, smart meter clusters are deployed at various nodes of the power distribution network, such as substations and distribution transformers. The smart meters can collect historical load data of each node in real time and accurately, including information such as active power and reactive power, and record load data every 15 minutes.

[0027] Based on time series analysis, the collected historical load data is processed and analyzed on an annual basis to examine load fluctuations in different seasons. In summer, the load typically increases significantly due to the extensive use of air conditioning and other cooling equipment. In winter, the use of heating equipment may lead to a different load variation pattern. By extracting seasonal fluctuation characteristics and combining them with other relevant factors, such as economic development trends and population growth, a load growth trend prediction model is constructed. The load growth trend prediction model can be based on machine learning algorithms, such as neural networks and regression analysis, to predict future load growth trends.

[0028] This study analyzes the spatial correlation characteristics of load data, examining historical load data for each node in the distribution network to identify the interrelationships in load changes between different nodes. For example, some nodes may exhibit similar or correlated load changes due to geographical proximity or overlapping power supply areas. By extracting spatiotemporal correlation features, historical load spatiotemporal characteristic data is obtained. This data contains comprehensive information about the load in both time and space. This spatiotemporal characteristic data is then input into a constructed load growth trend prediction model. Based on this data, the model further predicts the load situation over a longer period, such as 5 or 10 years, generating a medium- to long-term load forecast dataset. This dataset contains the predicted load values ​​for each node at different points in time.

[0029] It should be noted that by collecting load data in real time through smart meter groups, and combining time series analysis and spatiotemporal correlation feature extraction, a load growth trend prediction model can be accurately constructed, providing high-precision data support for medium- and long-term load forecasting and effectively improving the scientificity and accuracy of distribution network planning.

[0030] In this embodiment of the invention, step S200 includes the following sub-steps B1-B3; In B1: Load growth characteristics are extracted from the medium- and long-term load forecast dataset, and the load growth rate corresponding to each time period is calculated based on the linear regression algorithm with a sliding window. In B2: Construct a set of load growth rate feature vectors based on the load growth rate corresponding to each time period; In B3: The initial cluster centers in the clustering algorithm are dynamically optimized using the silhouette coefficient method, and iterative clustering analysis is performed on each load growth feature vector in the load growth feature vector set. The membership degree is then used to determine the year division results for each planning stage.

[0031] Specifically, load growth characteristics are extracted from the generated medium- and long-term load forecast dataset. A sliding window linear regression algorithm is used to divide the time period into multiple time intervals, the length of which can be set according to actual needs, such as one year. Within each time interval, the load change trend over time is fitted using a linear regression algorithm, and the load growth rate within that time interval is calculated using a sliding window linear regression algorithm. The linear regression model is expressed as: in, This is the load value. For time variables, The intercept represents the initial value of the load. The slope represents the growth rate of the load, expressed as: in, For the first At a certain point in time, For the first Load values ​​corresponding to each time point This represents the number of time points within the sliding window.

[0032] The load growth rate represents the rate at which load changes over time and is used to capture load growth trends. A sliding window divides a time series into multiple overlapping or non-overlapping time periods, and performs linear regression fitting within each window; the slope of each window represents the load growth rate.

[0033] For example, for a given node, the load growth rate is calculated to be 5% during the first time period (year 1 to year 2) using linear regression. This process is repeated to calculate the load growth rate for each time period. Based on the load growth rate, a set of load growth feature vectors is constructed, with each vector representing the load growth feature within a time period.

[0034] In this embodiment of the invention, the clustering algorithm includes the k-means clustering algorithm; The k-means clustering algorithm is a commonly used clustering method, but it is sensitive to the selection of initial cluster centers. To improve the accuracy of clustering, the silhouette coefficient method is used to dynamically optimize the initial cluster centers.

[0035] The silhouette coefficient method evaluates the quality of clustering by calculating the distance from each sample point to its corresponding cluster center and the distance to other cluster centers.

[0036] The initial cluster center combination with the largest silhouette coefficient is selected, and iterative cluster analysis is performed on each vector in the load growth characteristic vector set to group vectors with similar load growth characteristics into one class. During the clustering process, the membership degree is used to determine which cluster each vector belongs to. If a load growth characteristic vector belongs to a certain class with a membership degree greater than 0.7, it is considered to belong to that class. Based on the clustering results and the planning requirements of the distribution network, the year division results of each planning stage are determined, and time periods with similar load growth are divided into a planning stage, and the start and end years of each planning stage are determined.

[0037] It should be noted that from the collection of historical load data at each node of the distribution network to the determination of the year division results for each planning stage, accurate prediction and reasonable planning of the distribution network load have been achieved, providing an important reference for the construction and operation of the distribution network.

[0038] In this embodiment of the invention, after completing steps B1-B3, step S200 also includes steps B4-B8. In B4: The load growth rate feature vector set is standardized, and principal component analysis is used to reduce the dimension of each load growth rate feature vector in the load growth rate feature vector set to generate a principal component vector set. In B5: Determine the initial range of cluster numbers and calculate the silhouette coefficient value corresponding to each cluster number. Based on the silhouette coefficient value corresponding to each cluster number, determine the optimal number of partitions for the planning stage. In B6: Initialize cluster centers, and calculate the similarity between each vector in the principal component vector set and the initial cluster centers using Mahalanobis distance to generate the initial distribution matrix of cluster centers; In B7: Iterative clustering analysis is performed based on the initial distribution matrix of cluster centers, dynamically adjusting the membership probability of each vector in the principal component vector set to the cluster of each planning stage, and setting a dynamic membership threshold to filter high-confidence data samples; In B8: The boundary years for each planning stage are fitted and corrected based on the spatiotemporal distribution data corresponding to the high-confidence data samples, and the year division results for each planning stage are output.

[0039] Specifically, since each feature vector in the load growth feature vector set may have different dimensions and value ranges, standardization is required to eliminate the impact of these differences on subsequent analysis. For the feature of load growth rate, the growth rate may be higher in some time periods and lower in others. Through standardization, it is transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1. After standardization, principal component analysis is used to reconstruct the load growth feature vector set by dimensionality reduction. The core idea of ​​principal component analysis is to transform multiple related variables into a few unrelated composite variables (principal components). Principal components can retain as much information as possible from the original data. The original load growth characteristic vector may contain information in multiple dimensions, such as the growth rate in different time periods and the trend of growth rate changes. Through principal component analysis, it can reduce the dimensionality to a few main dimensions and generate a set of principal component vectors. This not only reduces the dimensionality of the data and the computational complexity, but also highlights the main features of the data.

[0040] Before performing cluster analysis, an initial range for the number of clusters needs to be determined, such as from 2 to 10. For each cluster number within this range, the corresponding silhouette coefficient value is calculated as follows: The overall silhouette coefficient is the average of the silhouette coefficients of all sample points, expressed as: in, For the first Profile coefficients for each sample point For sample points The average distance to all other sample points within the same cluster. For sample points The average distance to all sample points in the nearest cluster. This is the overall profile coefficient. This represents the total number of sample points.

[0041] The silhouette coefficient is a metric used to evaluate clustering performance. It measures how closely a sample point is clustered with other points within the same cluster and how well it is separated from points in other clusters. A silhouette coefficient value closer to 1 indicates a better clustering effect, while a value closer to -1 indicates a worse clustering effect.

[0042] For example, when the number of clusters is 3, the calculated silhouette coefficient is 0.6; when the number of clusters is 4, the silhouette coefficient is 0.7. By comparing the silhouette coefficient values ​​corresponding to different numbers of clusters, the number of clusters with the largest silhouette coefficient value is selected as the optimal number of partitions in the planning stage. Therefore, the silhouette coefficient value is the largest when the number of clusters is 4, so the optimal number of partitions is 4.

[0043] Specifically, after determining the optimal number of partitions, the cluster centers are initialized. Several principal component vectors can be randomly selected as the initial cluster centers. The Mahalanobis distance is used to calculate the similarity between each vector in the set of principal component vectors and the initial cluster centers. The Mahalanobis distance is expressed as: in, Principal component vectors, The cluster center vector, The Mahalanobis distance, For transpose operation, It is the inverse of the covariance matrix.

[0044] Mahalanobis distance takes into account the covariance structure of the data and can more accurately measure the distance between sample points. For example, after calculating the Mahalanobis distance between each vector in the principal component vector set and the initial cluster center, an initial distribution matrix of cluster centers is generated. The initial distribution matrix of cluster centers records the distance information between each vector and each cluster center. Based on the initial distribution matrix of cluster centers, iterative clustering analysis begins. In each iteration, the membership probability of each vector to a cluster at each planning stage is dynamically adjusted according to the distance between the vector and the cluster center. If a vector is closer to a cluster center, its membership probability to that cluster is higher. Simultaneously, a dynamic membership threshold is set. When the membership probability of a vector to a cluster exceeds this threshold, it is considered a high-confidence data sample. The dynamic membership threshold is set to 0.8. When the membership probability of a vector to a cluster reaches 0.8 or higher, it is selected. Through continuous iteration, the cluster centers and membership probabilities are updated until the clustering results converge.

[0045] Methods such as linear regression and polynomial fitting are used to fit the relationship between load growth rate and time, determine the start and end years of each planning stage, and find that the load growth rate is relatively stable in a certain period of time, while the load growth rate changes significantly in another period of time. Based on the changes, the dividing point of the planning stage is determined. Finally, the year division results of each planning stage are output, providing a specific time framework and guidance for the planning and construction of the distribution network.

[0046] It should be noted that by standardizing the process and using principal component analysis to reduce dimensionality, the clustering effect is optimized, the planning phase years are accurately defined, and the scientific nature and adaptability of the planning are improved, providing strong guidance for the construction of power distribution networks.

[0047] In this embodiment of the invention, step S300 includes the following sub-steps C1-C3; In C1: The objective function is to minimize the present value of the total cost of the distribution network over its entire lifecycle. In C2: the constraints include distributed power flow coupling constraints, node voltage opportunity constraints, candidate node installed capacity constraints, and capacity upper limit constraints, constructing a multi-dimensional constraint system; In C3: a planning model is constructed using an objective function and a multi-dimensional constraint system.

[0048] Specifically, the planning model takes minimizing the present value of the total cost of the distribution network over its entire lifecycle as its objective function and includes multiple constraints, such as distributed generation power flow coupling constraints, node voltage opportunity constraints, candidate node installed capacity constraints, and capacity ceiling constraints. These constraints ensure that the planning scheme strikes a balance between technical feasibility and economic optimality.

[0049] The objective function is expressed as: in, For the first The total annual cost includes investment costs, operation and maintenance costs, etc. The discount rate is used to calculate the present value. The total number of years in the planning period; The constraints include: Distributed power flow coupling constraints are expressed as: The node voltage machine constraint is expressed as: The candidate node installation capacity constraint is expressed as: The capacity limit constraint is expressed as: in, For the active power output of distributed power sources, Active power purchased from the main power grid For the active power of the load, For network loss active power, For nodes voltage amplitude, This is the minimum allowable voltage value. This is the maximum allowable voltage. For nodes Distributed power supply installation capacity, For nodes Maximum installed capacity of distributed power sources This represents the upper limit of the total installed capacity of distributed power sources.

[0050] The planning model integrates distributed generation power flow coupling constraints, node voltage opportunity constraints, candidate node installation capacity constraints, and capacity upper limit constraints, constructing a multi-dimensional constraint system that enables the planning model to more comprehensively and accurately reflect the actual operation of the distribution network and various constraints. It comprehensively considers the access of distributed generation with multiple aspects such as the security, stability, and power quality of the distribution network, providing a scientific basis for distribution network planning. By reasonably setting constraints, the planning model can select the optimal solution from numerous planning schemes, realize the rational layout and capacity configuration of distributed generation, improve the operating efficiency and economic benefits of the distribution network, and promote the large-scale application of renewable energy and the sustainable development of the power system.

[0051] In this embodiment of the invention, step S400 includes the following sub-steps D1-D6; In D1: Initialize the swarm size and number of iterations for the particle swarm algorithm, and set the shrinkage factor to generate the initial solution set; In D2: Based on the objective function of the planning model, a multi-objective fitness function is constructed, the fitness of each particle in the initial solution set is evaluated, the fitness score of each particle is obtained, and a particle fitness ranking list is generated. In D3: The shrinkage factor is dynamically adjusted according to the particle fitness ranking list, and the historical best solution and global best solution of the particle swarm are iteratively updated based on the fitness score of each particle to generate an optimal solution set distribution cloud map; In D4: Monitor the population dispersion index of the optimized solution set distribution cloud map. When the population dispersion index is lower than the preset threshold, trigger the chaotic mapping perturbation strategy and generate a chaotic sequence. Based on the chaotic sequence, dynamically perturb the particle positions in the high-density region. In D5: The optimal compromise solution for each planning stage is identified from the optimal solution set distribution cloud map, and the planning scheme is output based on the optimal compromise solution; In D6: The planning scheme includes the optimized access location and capacity configuration scheme of distributed power sources for each planning stage.

[0052] Specifically, the initial solution set includes determining the swarm size (i.e., the number of particles) of the particle swarm algorithm based on factors such as the complexity of the planning model and computational resources. The swarm size is set to 50 particles. Simultaneously, the number of iterations is set, such as 100 iterations, to control the algorithm's runtime and search depth. The shrinkage factor is an important parameter in the particle swarm algorithm, used to adjust the particle velocity update formula, affecting the particle's search range and convergence speed. A suitable shrinkage factor is set, such as 0.8. When initializing the particle swarm, each particle represents a possible planning scheme, and its position and velocity are randomly generated. Through the above operations, the initial solution set is generated, which is a set of particles containing different planning schemes.

[0053] The particle fitness ranking list includes the present value of total cost over the entire lifecycle objective function in the planning model, which is one of the important indicators for evaluating the quality of the planning scheme. There may also be other objectives, such as reducing carbon emissions and improving power supply reliability. Combining the above objectives, a multi-objective fitness function is constructed. The fitness function can be a weighted sum of indicators such as the present value of total cost over the entire lifecycle, carbon emissions, and power supply reliability. For each particle in the initial solution set, the planning scheme it represents is substituted into the multi-objective fitness function for calculation to obtain the particle's fitness score. The higher the fitness score, the better the planning scheme. The fitness scores of all particles are ranked to generate a particle fitness ranking list for subsequent analysis and operation.

[0054] The multi-objective fitness function is expressed as: in, For the fitness function value, The weighting factor for the present value of total cost over the entire lifecycle. The present value of total cost over the entire lifecycle. The weighting factor for carbon emissions. For carbon emissions, This is a weighting factor for power supply reliability. For power supply reliability.

[0055] The optimized solution set distribution cloud map involves dynamically adjusting the contraction factor based on the particle fitness ranking list. When the particle swarm performs well (i.e., there are many particles with high fitness scores), the contraction factor can be appropriately reduced to accelerate convergence; conversely, when the search performance is poor, the contraction factor is increased to expand the search range. For each particle, its historical optimal solution and the global optimal solution of the entire particle swarm are recorded. In each iteration, the historical optimal solution and the global optimal solution are updated based on the particle's fitness score. Through multiple iterations, the particle swarm gradually converges to the vicinity of the optimal solution, generating the optimized solution set distribution cloud map. This map illustrates the distribution of particles in the solution space, reflecting the diversity and convergence of the optimized solution set.

[0056] Specifically, the volume dispersion index is used to measure the degree of dispersion of the particle swarm in the solution space. The swarm dispersion index is obtained by calculating the distance between particles. When the swarm dispersion index is lower than the threshold, it indicates that the particle swarm is too concentrated and may be trapped in a local optimum. At this time, the chaotic mapping perturbation strategy is triggered. The threshold is set according to the fitting of historical data, for example, 0.1 or 0.05, which represents the minimum allowable dispersion of the particle swarm.

[0057] Population dispersion is measured by calculating the average Euclidean distance between particle positions and expressed as: in, As a measure of population dispersion, For the number of particles, To solve for the dimension of space, For the first The particle in the first The position of the dimension For the first The mean of the dimension.

[0058] Chaotic mapping is a mathematical model with randomness and ergodicity. Chaotic sequences are generated by chaotic mapping and applied to the particle positions in high-density regions of the solution set distribution cloud map. The positions of the particles are dynamically perturbed, causing them to escape local optima, expanding the search range, and increasing the probability of the algorithm finding the global optimum.

[0059] In the optimal solution set distribution cloud map, the optimal compromise solution for each planning stage is found from the particle swarm optimization based on the multi-objective fitness function and planning requirements. The optimal compromise solution achieves a good balance among multiple objectives, ensuring a low present value of total cost over the entire cycle while maintaining a certain level of power supply reliability and low carbon emissions. Based on the optimal compromise solution, the optimal access location and capacity configuration scheme for distributed power sources in each planning stage is determined. In a specific planning stage, a certain capacity of solar and wind power generation equipment is connected to specific locations in the distribution network to meet the power demand and planning objectives of that stage. The resulting planning scheme, containing detailed information for each planning stage, provides guidance for the construction and operation of the distribution network.

[0060] It should be noted that by optimizing the planning scheme through particle swarm optimization, and by using a multi-objective fitness function to comprehensively consider factors such as the present value of total cost over the entire cycle, power supply reliability, and carbon emissions, the shrinkage factor and chaotic mapping perturbation strategy are dynamically adjusted to effectively balance global search and local optimization capabilities. This allows for the efficient finding of the optimal compromise solution at each planning stage, providing a scientific basis for the optimized access location and capacity configuration of distributed power sources, and significantly improving the scientific and economic efficiency of distribution network planning.

[0061] In this embodiment of the invention, step S500 includes the following sub-steps E1-E3; In E1: Construct a multi-stage planning verification model based on the planning scheme, and import the capacity configuration parameters and access node coordinate data corresponding to each planning stage; In E2: Continuous power flow calculation is performed based on a multi-stage planning verification model to obtain the power flow distribution map, node voltage deviation data and line load rate index of the distribution network under each planning stage, and to generate an evaluation dataset. In E3: Based on the evaluation dataset, the planning schemes are quantitatively scored in multiple dimensions, and the scoring results are comprehensively analyzed and ranked to generate a planning feasibility report.

[0062] Specifically, based on the generated planning scheme, which includes information such as the optimized access location and capacity configuration of distributed power sources at each planning stage, a multi-stage planning verification model is constructed. The multi-stage planning verification model simulates the actual operation of the distribution network at different planning stages. For a distribution network planned in three phases, the characteristics of network structure and power distribution in each phase will be reflected. The capacity configuration parameters corresponding to each planning phase, such as the capacity of distributed power sources and the capacity of energy storage devices, as well as the coordinate data of access nodes, i.e. the specific location information of distributed power sources accessing the distribution network, will be imported into the multi-phase planning verification model. Continuous power flow calculation is a method for analyzing power flow distribution under different operating conditions of a power system. It utilizes a multi-stage planning verification model to perform continuous power flow calculations for each planning stage, taking into account the interactions between various power sources, loads, and lines in the distribution network. When distributed generation sources are connected to the distribution network, they affect the power flow distribution. The calculation obtains power flow distribution maps for each planning stage, which visually demonstrate the flow of electricity within the distribution network, including current and voltage distribution. Simultaneously, it obtains node voltage deviation data, i.e., the difference between the actual node voltage and the rated voltage, which is crucial for evaluating the power supply quality of the distribution network. Furthermore, it calculates the line load factor, reflecting the ratio of the actual load on a line to its rated load. The power flow distribution maps, node voltage deviation data, and line load factor are compiled into an evaluation dataset. This evaluation dataset comprehensively records the operating characteristics of the distribution network at each planning stage, providing rich information for subsequent evaluations.

[0063] Based on the evaluation dataset, the planning scheme is quantitatively scored from multiple dimensions. From the power supply reliability dimension, the score is based on the node voltage deviation data, with a higher score for a smaller voltage deviation. From the line operation safety dimension, the score is based on the line load rate index, with a higher score for a load rate within a reasonable range. Each dimension is assigned a corresponding weight, and the comprehensive score of the planning scheme is calculated based on the weight.

[0064] Based on the scoring results, a comprehensive analysis and ranking of schemes at different planning stages or different planning schemes are conducted. The characteristics and reasons of schemes with high and low scores are analyzed, and the advantages and disadvantages of the planning schemes are identified. The analysis results are compiled into a planning feasibility report. The report not only includes the scoring situation of each planning stage and the comprehensive analysis results, but also puts forward improvement suggestions and measures, providing decision-makers with detailed reference for judging the feasibility of planning schemes and whether the schemes need to be adjusted.

[0065] It should be noted that by using a multi-stage planning verification model and continuous power flow calculation, the power supply reliability, line operation safety and other multi-dimensional indicators of the planning scheme are comprehensively evaluated, providing decision-makers with detailed reference information, helping to optimize and adjust the scheme, and ensuring the scientific nature and feasibility of the distribution network planning.

[0066] The above is an illustrative scheme of a multi-stage planning method for a distribution network according to this embodiment. It should be noted that the technical solution of this multi-stage planning system for a distribution network and the technical solution of the above-described multi-stage planning method for a distribution network belong to the same concept. For details not described in detail in the technical solution of the multi-stage planning system for a distribution network in this embodiment, please refer to the description of the technical solution of the above-described multi-stage planning method for a distribution network.

[0067] The multi-stage planning system for the power distribution network in this embodiment includes: The feature extraction module is used to acquire historical load data of the distribution network, extract features from the historical load data, and generate a medium- and long-term load forecast dataset. The phase division module is used to extract features from the medium- and long-term load forecast dataset and use a clustering algorithm to divide it into multiple phases, so as to obtain the year division results for each planning phase. The model building module is used to build a planning model based on the year division results; The model solving module is used to solve the planning model using intelligent optimization algorithms to obtain the planning scheme; The report generation module is used to perform simulation verification for each planning stage based on the planning scheme and generate a planning feasibility report.

[0068] This embodiment also provides a computer device suitable for multi-stage planning of power distribution networks, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a multi-stage power distribution network planning method as described in the above embodiments.

[0069] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a multi-stage planning method for a power distribution network as proposed in the above embodiments.

[0070] The storage medium proposed in this embodiment and the method for implementing multi-stage planning of power distribution networks proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0071] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computing device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-stage planning method for power distribution networks, characterized in that, include: Historical load data of the distribution network is acquired, and features are extracted from the historical load data to generate a medium- and long-term load forecast dataset. Feature extraction is performed on the medium- and long-term load forecast dataset, and a clustering algorithm is used to divide it into multiple stages to obtain the year division results for each planning stage. Construct a planning model based on the stated year division results; The planning model is solved using an intelligent optimization algorithm to obtain a planning scheme; Based on the proposed planning scheme, simulation verification is performed for each planning stage, and a planning feasibility report is generated.

2. The multi-stage planning method for a power distribution network as described in claim 1, characterized in that, The generation of medium- and long-term load forecast datasets includes: Seasonal fluctuation characteristics were extracted from historical load data based on time series, and a load growth trend prediction model was constructed. Spatiotemporal correlation features are extracted from historical load data to obtain historical load spatiotemporal feature data; Historical load spatiotemporal characteristics data are input into the load growth trend prediction model to generate a medium- and long-term load prediction dataset.

3. The multi-stage planning method for a power distribution network as described in claim 2, characterized in that, The timeframes for each planning phase include: Load growth characteristics are extracted from the medium- and long-term load forecast dataset, and the load growth rate for each time period is calculated using a linear regression algorithm based on a sliding window. Construct a set of load growth rate feature vectors based on the load growth rate corresponding to each time period; The initial cluster centers in the clustering algorithm are dynamically optimized using the silhouette coefficient method. Iterative clustering analysis is performed on each load growth feature vector in the load growth feature vector set, and the year division results for each planning stage are determined by combining the membership degree.

4. The multi-stage planning method for a power distribution network as described in claim 3, characterized in that, Iterative clustering analysis of each load growth rate feature vector in the set of load growth rate feature vectors includes: The load growth rate feature vector set is standardized, and principal component analysis is used to reduce the dimension of each load growth rate feature vector in the load growth rate feature vector set to generate a principal component vector set. Determine the initial range of cluster numbers and calculate the silhouette coefficient value corresponding to each cluster number. Based on the silhouette coefficient value corresponding to each cluster number, determine the optimal number of partitions in the planning stage. Initialize cluster centers, and calculate the similarity between each vector in the principal component vector set and the initial cluster centers using Mahalanobis distance to generate the initial distribution matrix of cluster centers; Iterative clustering analysis is performed based on the initial distribution matrix of cluster centers. The membership probability of each vector in the principal component vector set to the cluster of each planning stage is dynamically adjusted, and a dynamic membership threshold is set to filter high-confidence data samples. Based on the spatiotemporal distribution data corresponding to the high-confidence data samples, the boundary years for each planning stage are fitted and corrected, and the year division results for each planning stage are output.

5. The multi-stage planning method for a power distribution network as described in claim 4, characterized in that, The planning model includes: The objective function is to minimize the present value of the total cost of the distribution network over its entire lifecycle. The constraints include distributed power flow coupling constraints, node voltage opportunity constraints, candidate node installation capacity constraints, and capacity upper limit constraints, constructing a multi-dimensional constraint system. A planning model is constructed using an objective function and a multi-dimensional constraint system.

6. A multi-stage planning method for a power distribution network as described in claim 1 or 5, characterized in that, Solving the planning model includes: Initialize the swarm size and number of iterations for the particle swarm optimization algorithm, and set the shrinkage factor to generate the initial solution set; Based on the objective function of the planning model, a multi-objective fitness function is constructed, and the fitness of each particle in the initial solution set is evaluated to obtain the fitness score of each particle and generate a particle fitness ranking list. The shrinkage factor is dynamically adjusted based on the particle fitness ranking list, and the historical optimal solution and global optimal solution of the particle swarm are iteratively updated based on the fitness score of each particle to generate an optimal solution set distribution cloud map. The population dispersion index of the optimized solution set distribution cloud map is monitored. When the population dispersion index is lower than the preset threshold, the chaotic mapping perturbation strategy is triggered and a chaotic sequence is generated. The particle positions in the high-density region are dynamically perturbed based on the chaotic sequence. The optimal compromise solution for each planning stage is identified from the optimal solution set distribution cloud map, and the planning scheme is output based on the optimal compromise solution; The planning scheme includes optimized access locations and capacity configuration schemes for distributed power sources at each planning stage.

7. The multi-stage planning method for a power distribution network as described in claim 6, characterized in that, The generated planning feasibility report includes: A multi-stage planning verification model is constructed based on the planning scheme, and the capacity configuration parameters and access node coordinate data corresponding to each planning stage are imported. Continuous power flow calculation is performed based on a multi-stage planning verification model to obtain the power flow distribution map, node voltage deviation data and line load rate index of the distribution network under each planning stage, and to generate an evaluation dataset. Based on the evaluation dataset, the planning schemes are quantitatively scored from multiple dimensions, and a comprehensive analysis and ranking are performed based on the scoring results to generate a planning feasibility report.

8. A multi-stage planning system for a power distribution network, employing the multi-stage planning method for a power distribution network as described in any one of claims 1-7, characterized in that, include: The feature extraction module is used to acquire historical load data of the distribution network, extract features from the historical load data, and generate a medium- and long-term load forecast dataset. The phase division module is used to extract features from the medium- and long-term load forecast dataset and use a clustering algorithm to divide it into multiple phases, so as to obtain the year division results for each planning phase. The model building module is used to build a planning model based on the year division results; The model solving module is used to solve the planning model using intelligent optimization algorithms to obtain the planning scheme; The report generation module is used to perform simulation verification for each planning stage based on the planning scheme and generate a planning feasibility report.

9. A computer device, comprising: A memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of a multi-stage planning method for a power distribution network according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of a multi-stage planning method for a power distribution network as described in any one of claims 1 to 7.

Images