A method for coordinated optimization of transformer and energy storage configuration in a photovoltaic-storage supercharging station

Abstract

This invention relates to the field of photovoltaic-storage supercharging station technology, and discloses a method for the coordinated optimization configuration of transformers and energy storage in photovoltaic-storage supercharging stations. The method includes: obtaining typical daily load curves and their occurrence probabilities based on historical charging load data from multiple scenarios; establishing an upper-level planning model and a lower-level scheduling model for the coordinated optimization of transformer and energy storage configuration based on the typical daily load curves and their occurrence probabilities; considering the typical characteristics of photovoltaic-storage supercharging stations, introducing an improved quantum particle swarm optimization algorithm to solve the upper-level planning model, and using a mixed-integer linear programming solver to collaboratively solve the lower-level scheduling model, thereby obtaining the optimal configuration scheme for transformers and energy storage. By introducing a weighted K-means clustering algorithm based on peak retention, the clustering process pays more attention to the fitting accuracy during high-power periods, retaining peak period information in the clustering results, ensuring that the configuration scheme can effectively cope with the instantaneous load impact of actual photovoltaic-storage supercharging stations, and improving power supply reliability and economy.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic-storage supercharging station technology, and more specifically, to a method for the coordinated optimization configuration of transformer and energy storage in a photovoltaic-storage supercharging station. Background Technology

[0002] With the continuous growth of new energy vehicle ownership, the load structure and operating characteristics of urban power distribution networks have undergone significant changes. To meet users' demands for charging efficiency, charging stations are gradually evolving from traditional slow and fast charging to higher-power supercharging modes. On the one hand, supercharging loads are characterized by high power levels, rapid changes, and concentrated access. Direct connection to distribution transformers can easily lead to increased load fluctuations, resulting in voltage deviation, increased network losses, and equipment overload, placing higher demands on the capacity configuration and safe operation of distribution transformers. On the other hand, against the backdrop of continuously increasing demand for renewable energy consumption, introducing photovoltaic and energy storage systems into charging stations, through the coordinated operation of photovoltaic, energy storage, and charging, can alleviate the pressure on power distribution network operations to a certain extent and improve the local consumption capacity of renewable energy. Therefore, research on the capacity coordination configuration of transformers and energy storage for photovoltaic-energy storage supercharging station scenarios has engineering application value for ensuring power supply reliability and balancing investment and operational economy.

[0003] The coordinated configuration of transformers and energy storage capacity in photovoltaic-storage supercharging stations has the following shortcomings: (1) Inaccurate load forecasting for photovoltaic-storage supercharging stations: Existing load forecasting methods can be mainly divided into two categories: probabilistic statistical modeling and historical data clustering. Existing probabilistic statistical modeling usually starts from vehicle travel patterns, road traffic conditions and user charging behavior to construct a mechanism model of electric vehicle charging demand. Its model is complex and the parameters are highly subjective, and it does not fully describe temporary charging demand. On the other hand, the historical data clustering method uses actual charging load data to extract representative typical load patterns through similarity analysis. However, it is easy to overlook the need to retain extreme peak scenarios for photovoltaic-storage supercharging stations, mistakenly smoothing out the peak load characteristics, which may lead to a serious underestimation of actual peak demand in the configuration scheme.

[0004] (2) The final configuration scheme may not be practically operational: Due to the coupling relationship between transformer capacity, energy storage scale and system operation scheduling, the single-layer optimization model is difficult to characterize their mutual constraint characteristics. When the traditional two-layer optimization model is used for analysis, it is mostly unidirectional transmission. The final configuration scheme may not be applicable to the specific application scenarios of photovoltaic-storage-supercharging stations.

[0005] (3) The quantum particle swarm optimization algorithm does not consider the essential characteristics of photovoltaic-storage-supercharging stations: For example, the comprehensive energy optimization scheduling method of hybrid adaptive quantum particle swarm optimization disclosed in CN119813380A is applied to the field of energy management technology. However, the energy storage scheduling of photovoltaic-storage-supercharging stations is essentially an event-driven problem, and the start and stop of its charging piles are discrete events. If the traditional quantum particle swarm optimization algorithm is used to search in continuous space, it is difficult to directly process discrete decision variables according to the actual situation of photovoltaic-storage-supercharging stations (such as charging pile occupancy rate).

[0006] In summary, in response to the problems of large fluctuations in high-power charging load and difficulties in coordinating transformer capacity configuration with energy storage scale in photovoltaic-storage supercharging stations, there is an urgent need for a method for coordinating and optimizing the configuration of transformers and energy storage in photovoltaic-storage supercharging stations, so as to improve the economy and operational stability of photovoltaic-storage supercharging stations in different typical scenarios. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for the coordinated optimization configuration of transformer and energy storage in a photovoltaic-storage supercharging station, so as to solve the problem that the large fluctuation and inaccurate prediction of charging load in the photovoltaic-storage supercharging station makes it difficult to coordinate the configuration of transformer capacity and energy storage scale, which affects the economy and stability of the operation of the photovoltaic-storage supercharging station.

[0008] To achieve the above objectives, embodiments of the present invention provide a method for the coordinated optimization configuration of transformers and energy storage in a photovoltaic-storage supercharging station, comprising: based on historical charging load data from multiple scenarios, introducing a weighted K-means clustering algorithm based on peak retention to obtain typical daily load curves and their occurrence probabilities; based on the typical daily load curves and their occurrence probabilities, establishing an upper-level planning model and a lower-level scheduling model for the coordinated optimization of transformer and energy storage configuration; considering the typical characteristics of the photovoltaic-storage supercharging station, introducing an improved quantum particle swarm optimization algorithm to solve the upper-level planning model, and using a mixed-integer linear programming solver to collaboratively solve the lower-level scheduling model to obtain the optimal configuration scheme of transformers and energy storage.

[0009] The beneficial effects of this invention are: (1) In response to the instantaneous impact of the load on the photovoltaic-storage supercharging station, a weighted K-means clustering algorithm based on peak retention is introduced. In the traditional clustering algorithm framework, a peak weight coefficient is introduced, which makes the clustering process pay more attention to the fitting accuracy of high power periods. The peak period information in the clustering results is directly used for the constraint setting of the subsequent lower-level scheduling model, ensuring that the subsequent capacity configuration scheme can effectively cope with the load impact in the actual operation of the photovoltaic-storage supercharging station, and improving the power supply reliability and economy. (2) Considering the coupling relationship between transformer capacity, energy storage scale and system operation scheduling, as well as the typical characteristics of photovoltaic-storage supercharging station, a two-layer optimization configuration model of upper-level planning model and lower-level scheduling model is constructed to realize the collaborative optimization between planning layer and scheduling layer, which greatly improves equipment utilization efficiency and improves system economy; (3) In view of the characteristics of the photovoltaic-storage supercharging station scenario, by introducing an improved quantum particle swarm algorithm, the transformer and energy storage collaborative configuration problem of the photovoltaic-storage supercharging station can be solved efficiently and stably. At the same time, the output optimal configuration scheme meets the actual stable operation requirements of the photovoltaic-storage supercharging station, taking into account both power supply reliability and investment economy. Attached Figure Description

[0010] Figure 1 A flowchart illustrating the method for co-optimizing the configuration of transformers and energy storage in a photovoltaic-storage supercharging station; Figure 2 The SSE variation curve in a specific embodiment is used to determine the number of clusters; Figure 3 The SIL variation curve in a specific embodiment is used to determine the number of clusters; Figure 4 The CHI variation curve in a specific embodiment is used to determine the number of clusters; Figure 5 This is a schematic diagram of a two-layer model architecture; Figure 6 A flowchart illustrating the collaborative solution of a two-layer model. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0012] Please see Figure 1 To address the challenges of large fluctuations in high-power charging loads and difficulties in coordinating transformer capacity configuration with energy storage scale in photovoltaic-storage supercharging stations, this invention provides a method for the coordinated optimization of transformer and energy storage configuration in such stations. The method includes: based on historical charging load data from multiple scenarios, introducing a weighted K-means clustering algorithm based on peak retention to obtain typical daily load curves and their probabilities of occurrence; establishing an upper-level planning model and a lower-level scheduling model for the coordinated optimization of transformer and energy storage configuration based on the typical daily load curves and their probabilities of occurrence; considering the typical characteristics of photovoltaic-storage supercharging stations, introducing an improved quantum particle swarm optimization algorithm to solve the upper-level planning model, and employing a mixed-integer linear programming solver to collaboratively solve the lower-level scheduling model, thereby obtaining the optimal configuration scheme for transformer and energy storage.

[0013] Historical charging load data from various scenarios of photovoltaic-storage supercharging stations were collected. Addressing the instantaneous impact of load fluctuations at these stations, a weighted K-means clustering algorithm based on peak value retention was introduced. This algorithm generates typical daily load curves, achieving a peak value retention rate of over 96% for cluster centers. This prevents load shocks from being averaged out, providing a reliable data foundation for subsequent optimization. Based on these typical daily load curves and their corresponding probabilities of occurrence, a two-layer optimization configuration model was established, consisting of an upper-level planning model and a lower-level scheduling model for the coordinated optimization of transformer and energy storage configurations. This two-layer optimization model achieves efficient power supply and demand management. The coordinated optimization of resource decision-making and operation scheduling balances economic efficiency and power supply reliability, effectively reducing the average annual comprehensive cost of the system. By fully considering the typical characteristics of actual photovoltaic-storage supercharging stations, an improved quantum particle swarm optimization (IQPSO-ESS) algorithm is introduced to solve the upper-level planning model, which effectively improves the global search capability, convergence speed, and solution feasibility, avoiding the problems of early convergence and infeasible solutions in traditional methods. At the same time, mixed integer linear programming is used to solve the lower-level scheduling model, so that the final output optimal configuration scheme can significantly improve equipment utilization and reduce the total life cycle cost while ensuring power supply reliability.

[0014] To more clearly illustrate a method for the coordinated optimization of transformer and energy storage configuration in a photovoltaic-storage supercharging station, the present invention is described in detail below: S1. Based on historical charging load data from multiple scenarios, a weighted K-means clustering algorithm based on peak retention is introduced to obtain typical daily load curves and their occurrence probabilities. The charging load of photovoltaic-storage supercharging stations is influenced by multiple factors such as travel patterns, application scenarios, and vehicle arrival rates, exhibiting typical characteristics such as large fluctuations, short peak durations, and strong instantaneous impacts. Considering that a single-day load curve for a single scenario cannot accurately reflect the long-term operating characteristics of a charging station, directly configuring transformers and energy storage based on maximum load not only results in low equipment utilization but also leads to insufficient overall return on investment. Therefore, this application typifies historical charging loads, reducing the scale of load data while preserving the characteristics of load time-series changes and differences in different application scenarios, making it more suitable as input for subsequent optimization configuration models.

[0015] S11. Preprocess the historical charging load data of the photovoltaic-storage supercharging station; wherein, the preprocessing includes data cleaning, feature extraction and normalization in sequence; Historical charging load data of the photovoltaic-storage supercharging station was acquired, with a sampling interval of 15 minutes (96 sampling points per day). The raw historical charging load data was preprocessed as follows: S111. Perform data cleaning on the original historical charging load data; First, duplicate records with identical timestamps are removed. Then, linear interpolation is used to fill data gaps with no more than four consecutive missing sampling points. For periods with more than four consecutive missing sampling points, data from the same period (the same period of the previous week) are used for filling. Finally, values ​​exceeding the physical range (such as negative power or values ​​exceeding 150% of the charging pile's rated power) are corrected using the following formula:

[0016] In the formula, For a moment The original sampling power, in kW; The rated power of the charging pile is expressed in kW. For a moment The corrected power value is in kW.

[0017] S112. Based on the historical charging load data after data cleaning, extract the characteristic quantities that characterize the load impact characteristics; wherein, the characteristic quantities include maximum power, average power and peak duration; For each 15-minute time window (a total of T=96 time windows per day), extract the maximum power. Average power and peak duration Three characteristic quantities characterize the impact characteristics of the load; among them, maximum power For the first Heavenly Peak charging power within each time window, in kW; average power For the first Heavenly Average charging power within a time window, in kW; peak duration For the first Heavenly Within a given time window, the charging power exceeds the peak threshold. The continuous duration, in hours.

[0018] In one specific embodiment, the peak threshold can be... The power is set to 80% of the charging pile's rated power, that is: The three features of each of the 96 time windows are combined to form the feature vector matrix for that day: In the formula, , indicating the first The maximum power sequence of the day; , indicating the first The average power sequence over the days; , indicating the first The peak duration sequence of days.

[0019] S113. Normalize the extracted feature values; To eliminate the influence of different units on the clustering results, the feature vectors of each day are normalized. For the maximum power and average power sequences, min-max normalization is used to map each sequence to the [0,1] interval, as follows: ,

[0020] In the formula, , The first The maximum and minimum values ​​of the maximum power sequence; , The first The maximum and minimum values ​​of the daily average power sequence.

[0021] For the peak duration sequence, fractional scaling normalization is used, based on the maximum possible duration of the time window (15 minutes = 0.25 hours), as follows: ,

[0022] In the formula, This represents the maximum possible duration within the time window (i.e., 15 minutes, corresponding to 0.25 hours).

[0023] The normalized daily feature vector is:

[0024] This feature vector serves as the input data for subsequent weighted K-means clustering.

[0025] By simultaneously considering three dimensions of features—maximum power, average power, and peak duration—this method can more comprehensively characterize the load's impact characteristics compared to traditional methods that only use raw power values. The introduction of peak duration weighting allows the clustering results to distinguish between two different load patterns: "instantaneous spikes" and "continuous peaks," providing a more refined input for subsequent energy storage configuration.

[0026] S12. Based on the preprocessing results of historical charging load data, obtain evaluation indicators and determine the optimal number of clusters; Based on the preprocessing results of historical charging load data, namely the normalized feature vector set The input for clustering is given, where N is the total number of historical days. To determine a reasonable number of clusters... The following three evaluation indicators are used for comprehensive judgment: S121, Sum of Squared Errors (SSE): The sum of squared errors (SSE) measures the intra-cluster tightness of clustering results and is defined as:

[0027] In the formula, For the first One cluster; For the first The center vectors of each cluster; It is the Euclidean norm.

[0028] The sum of squared errors (SSE) varies with The curve increases and then decreases monotonically; the inflection point of the curve corresponds to... The value is a candidate value. The inflection point determination uses the "elbow rule": calculate adjacent... The decrease in SSE value, when the decrease first falls below the threshold. (In one specific embodiment, the threshold) When ), the corresponding This is the elbow point.

[0029] S122, Profile Indicator (SIL): The silhouette coefficient (SIL) comprehensively evaluates the intra-cluster compactness and inter-cluster separation of the clustering results. For the first... sample The contour coefficient is defined as:

[0030] In the formula, This is the average distance between a sample and other samples in the same cluster (i.e., intra-cluster dissimilarity). It is the minimum average distance between a sample and samples from other clusters (i.e., the inter-cluster dissimilarity).

[0031] The overall silhouette coefficient is for all samples Arithmetic mean:

[0032] In the formula, The range of values ​​is A higher value indicates a better clustering effect. It is generally believed that... This indicates that the clustering results are reasonable.

[0033] S123, Calinski-Harabasz index (CHI): The CHI index, based on the ratio of intra-cluster dispersion to inter-cluster dispersion, is defined as follows:

[0034] In the formula, This represents the total number of samples (i.e., the total number of historical days). The sum of squares of inter-cluster deviations As the global center; It represents the sum of squared deviations within the cluster.

[0035] A higher CHI value indicates a denser cluster and a more dispersed cluster, resulting in better clustering performance.

[0036] Based on the above sum of squared errors (SSE), silhouette coefficient (SIL), and CHI index, plots were drawn respectively. , as well as Follow A curve showing the change in value, generally .

[0037] Sure The rule is: In Identifying elbow positions on curves ,exist Find the peak position on the curve And in Find the peak or inflection point on the curve If the three indicators point to the same If the value is correct, use it directly; if there are differences, take the median or weighted average of the three.

[0038] Please see Figure 2-4 To illustrate the determination of the number of clusters in more detail, a typical photovoltaic-storage supercharging station in an industrial area is used as a specific example. Based on charging load data from August and September, a weighted K-means clustering algorithm based on peak retention is employed to divide the charging load of each scenario into typical days. Considering the three evaluation indicators SSE, CHI, and SIL, the correlation between SSE, CHI, and SIL and the number of clusters is obtained. Change curve graph.

[0039] from Figure 2-4 It can be seen that when the number of clusters increases from 1 to 2, the SSE decreases significantly; then, when increasing from 2 to 5, the rate of decrease slows down, and after 5, the rate of decrease slows down even more. According to the elbow method, A number of clusters between 2 and 5 is considered appropriate. Under the SIL and CHI indices, a cluster size of 2 results in better performance, achieving a good balance between the silhouette coefficient and the Calinski-Harabasz index. In contrast, exceeding 2 clusters does not significantly improve the indicators. Therefore, the final decision was to divide typical days into two categories, thus determining the number of clusters to be [number missing]. .

[0040] S13. Based on the extracted features and the determined optimal number of clusters, a weighted distance function is designed to perform clustering, and typical daily load curves and their corresponding probabilities of occurrence are obtained. S131. Considering the peak characteristics of the photovoltaic-storage-supercharging station, based on the importance of extracted feature quantities and peak indicators, a weighting coefficient of the feature quantities is introduced and a weighted distance function that retains the peak value is designed. Considering the extracted maximum power sequence features Directly characterizing the impact properties of the load should be given higher weight in clustering. Therefore, for the th... Sample With the Cluster centers The weighted distance is defined as:

[0041] In the formula, This represents the total number of sampling points per day.

[0042] The weighting coefficients are set according to the importance of each feature and the peak value indication, as follows:

[0043]

[0044]

[0045] In the formula, The normalized peak threshold, as in one specific embodiment, corresponds to the original threshold. The normalization result; The peak weighting coefficient for the maximum power is, for example, taken as follows in a specific embodiment: This makes the clustering process focus more on the fitting accuracy during high-power periods; The weighting factor for the peak duration is, for example, taken in a specific embodiment as... This allows periods with peak durations to receive greater attention; This is an indicator function that takes the value 1 when the condition is true and 0 otherwise.

[0046] Substituting the weighting coefficients into the weighted distance formula above, we obtain the complete weighted distance expression for preserving the peak value. According to this weighted distance function, when the maximum power in a certain time period exceeds a threshold, the error in that time period is amplified. The time error of a given period is magnified by a factor of 1 when a peak duration exists within that period. This multiplies the accuracy of the clustering process, making it more focused on the fitting accuracy during periods of load impact.

[0047] By introducing a peak weight coefficient This approach allows cluster centers to focus more on the fitting accuracy during high-power periods, avoiding the underestimation problem caused by the "averaging" of peak values ​​in traditional methods. Experimental verification based on historical data from a real photovoltaic-storage-supercharging station shows that the peak power retention rate (the ratio of the peak value of the cluster center to the peak value of the original data) of the generated typical daily curve reaches over 95%, while the traditional K-means method only achieves 75%-85%. A detailed comparison is shown in Table 1 below: Table 1: Statistical Table of Evaluation Indicators for Traditional K-means and Weighted K-means

[0048] In summary, the traditional K-means clustering algorithm, which uses Euclidean distance as a similarity measure, tends to "average" high-power peak periods, leading to an underestimation of actual load impacts in typical daily curves. This results in problems such as under-configured transformer capacity and frequent overloads during operation. This application, however, addresses the instantaneous load impacts of photovoltaic-storage-supercharging stations by introducing a weighted K-means clustering algorithm based on peak retention. This introduces a peak weight coefficient into the traditional clustering framework, making the clustering process more focused on the fitting accuracy of high-power periods. The peak period information in the clustering results is directly used to set constraints for subsequent lower-level scheduling models, ensuring that subsequent capacity configuration schemes can effectively cope with load impacts during the actual operation of photovoltaic-storage-supercharging stations, thus improving power supply reliability and economy.

[0049] S132. A weighted K-means clustering algorithm based on peak retention is used to iteratively update the typical daily load curve and its corresponding probability of occurrence based on the weighted distance function. S1321. A K-means++ initialization strategy is adopted, which involves randomly selecting the first cluster center, and selecting subsequent cluster centers with a probability proportional to the squared distance to existing centers. Specifically: First cluster center From the preprocessed feature vector set Randomly selected from the middle; for the first Cluster centers ( ), calculate the shortest weighted distance between each sample and the selected cluster centers. proportional to The probability of selecting the next cluster center is determined by the probability of the cluster.

[0050] S1322, For each preprocessed sample Based on the weighted distance expression, calculate the weighted distance between it and each cluster center. It is assigned to the category of the cluster center with the smallest distance. :

[0051] S1323, For each category The cluster centers are recalculated as the mean vector of all samples in that category:

[0052] Based on this mean vector formula, the average values ​​of the maximum power sequence, average power sequence, and peak duration sequence are calculated respectively.

[0053] The average value of the maximum power sequence is: ; The average power sequence is: ; The average value of the peak duration sequence is: .

[0054] S1324. If the change in all cluster centers is less than a preset threshold That is, satisfying: If the algorithm succeeds, it converges; otherwise, it returns to S1322 to continue iterating. Simultaneously, a maximum number of iterations is set to prevent non-convergence. For example, in a specific embodiment, a maximum number of iterations is set. .

[0055] S1325. After clustering convergence, each cluster center... This represents a typical daily load characteristic vector. The maximum power sequence within it... The power curve serves as a typical day's data source (because maximum power directly reflects load surge characteristics and is a key input for subsequent capacity allocation). The probability of this typical day's curve appearing is determined by the proportion of sample days included in this category to the total number of days.

[0056] In the formula, Total number of days in history; For the first The number of samples in each category.

[0057] Based on this ratio, we obtain The typical daily load curves and their corresponding probabilities are as follows: .

[0058] S2. Based on typical daily load curves and their occurrence probabilities, establish a two-layer model for the coordinated optimization of transformer and energy storage configurations; Please see Figure 5Because there is a coupling relationship between transformer capacity, energy storage scale, and system operation scheduling, and because photovoltaic-storage supercharging stations exhibit typical characteristics such as large fluctuations and prolonged peak durations, a two-layer optimization configuration model is constructed, comprising a planning layer (upper layer) and a scheduling layer (lower layer). This model enables collaborative optimization between the planning and scheduling layers, thereby improving the equipment utilization efficiency of photovoltaic-storage supercharging stations and enhancing the overall system's economics. Specifically: S21. With the goal of minimizing the average daily comprehensive cost of the system, construct an upper-level planning model that includes constraints on the number of energy storage units and / or transformer capacity. The decision variables in the upper-level planning model are the transformer configuration capacity and the energy storage system scale. Its objective is to minimize the system's average daily comprehensive cost, thereby improving the overall economic efficiency of the system. (System average daily comprehensive cost) Mainly due to the system's average daily operating cost and the system's average daily investment cost It consists of two parts, including the system's average daily investment cost. The optimization results for each typical daily scenario need to be weighted by probability based on the lower-level scheduling model.

[0059] S211. With the goal of minimizing the system's average daily comprehensive cost, the objective function of the upper-level planning model is established as follows:

[0060] In the formula, the average daily operating cost of the system is... The unit is: yuan / day; average daily investment cost of the system. The unit is: yuan / day.

[0061] S2111, Average Daily Operating Cost of the System The average daily operating cost of the system is obtained by weighting the typical daily load curves and their probability of occurrence. The following results were obtained by probability-weighted averaging of the operating costs for each typical daily scenario:

[0062] In the formula, The total number of typical daily scenarios is determined by step S1; For the first The probability of occurrence of a typical daily scene is obtained from step S1). For the first The daily operating cost under a typical daily scenario is derived from the optimization of the lower-level scheduling model.

[0063] No. Daily operating costs under a typical daily scenario Including the cost of purchasing electricity from the grid. Energy storage operation and maintenance costs And transformer penalty costs ,Right now Specifically: S21111, Cost of Electricity Purchased from the Grid :

[0064] In the formula, For the first A typical daytime scene Power purchased from the power grid, in kW; For a moment Time-of-use electricity pricing, in yuan / kWh; In one specific embodiment, the time interval is... ; Total number of time periods per day .

[0065] S21112, Energy Storage Operation and Maintenance Costs :

[0066] In the formula, For the first A typical daytime scene The charging power of energy storage is measured in kW. For the first A typical daytime scene The discharge power of energy storage, measured in kW; The unit cost of energy storage is yuan / kWh.

[0067] S21113, Transformer penalty cost :

[0068] In the formula, The penalty factor is dimensionless and represents the loss coefficient for the excess electricity. The transformer capacity is configured in kVA. Assuming a power factor of 1, the active power (kW) and apparent power (kVA) are equal. To obtain a positive function, a penalty is only applied when the purchased power exceeds the transformer capacity.

[0069] In the transformer penalty cost formula, a penalty factor is introduced so that when the power purchased from the grid exceeds the transformer's configured capacity, the excess power is penalized, reflecting the lost charging service revenue due to insufficient capacity.

[0070] S2112, Average Daily Investment Cost of the System Including the average daily investment cost of transformers and average daily investment cost of energy storage ,Right now ; S21121, Average Daily Investment Cost of Transformers :

[0071] In the formula, For transformer depreciation rate; The design service life of the transformer, in years; The price per unit capacity of transformers is expressed in yuan / kVA. The capacity configured for the transformer is in kVA, and it is a decision variable for the upper level. The basic capacity fee for transformers is in yuan / kVA / month.

[0072] S21122, Average Daily Investment Cost of Energy Storage :

[0073] In the formula, Price per unit capacity of energy storage, in yuan / kWh; The number of energy storage units configured for the energy storage system is a decision variable for higher-level systems. Rated capacity of a single energy storage unit, in kWh; Energy storage depreciation rate; The term refers to the lifespan of energy storage, expressed in years.

[0074] S212. Establish constraints for the upper-level planning model; The constraints of the upper-level programming model mainly concern the range of values ​​for the decision variables: The quantity constraint for energy storage units is set as follows:

[0075] In the formula, , These are the lower and upper limits for the number of energy storage units, which are determined based on factors such as site area and investment budget. The number of energy storage units is determined by the upper-level planning model.

[0076] In some embodiments, a transformer capacity constraint can also be selectively set. The configurable transformer capacity constraint is as follows:

[0077] In the formula, , These are the lower and upper limits of the transformer configuration capacity, respectively, which are determined according to the grid connection agreement.

[0078] S22. With the goal of minimizing the daily operating cost and the peak-valley difference of the net load, a lower-level scheduling model is constructed, which includes the charging and discharging power constraints of the energy storage system, the state of charge constraints of the energy storage system, and the power balance constraints. For each typical day scenario, the lower-level scheduling model optimizes the charging and discharging power of the energy storage system by minimizing daily operating costs and the net load peak-to-valley difference, while satisfying operational constraints. This lower-level scheduling model is a mixed-integer linear programming (MILP) problem, which is solved using the Gurobi solver in this embodiment.

[0079] Under the premise of meeting operational constraints, the goal is to minimize the daily operating cost and the peak-to-valley difference of the net load, thereby coordinating the requirements of economic efficiency and operational stability.

[0080] S221. With the goal of minimizing the daily operating cost and the peak-to-valley difference of the net load, establish the objective function of the lower-level planning model: The lower-level scheduling model is a multi-objective optimization, including the sub-objectives of minimizing the daily operating cost of the system and minimizing the peak-to-valley difference in net load. By introducing sub-objective weight coefficients, a linear weighting method is used to transform the two sub-objectives into a single-objective function, and dimensionless processing is performed to avoid the influence of inconsistent dimensions. Specifically: S2211, Sub-objective 1: Establish the first objective function with the goal of minimizing the daily operating cost of the system. ; System daily operating cost and system daily operating cost of the upper-level planning model Consistency, that is:

[0081] S2212, Sub-objective two: Establish a second objective function with the goal of minimizing the peak-to-valley difference in net load. ; After introducing an energy storage system, the system's net load is defined as the original charging load (EV power) plus the energy storage charging power minus the energy storage discharging power, and then minus the photovoltaic output, i.e.:

[0082] In the formula, For the first In a typical daily scenario, time The system net load is expressed in kW. For the first In a typical daily scenario, time The electric vehicle charging load, in kW, is given by a typical daily curve obtained in step S1. For the first In a typical daily scenario, time The charging power of the energy storage system, measured in kW, is a decision variable determined by the lower-level scheduling model and must satisfy certain conditions. ; For the first In a typical daily scenario, time The discharge power of the energy storage system, measured in kW, is a decision variable determined by the lower-level scheduling model and must satisfy certain conditions. ; For the first In a typical daily scenario, time The photovoltaic output is expressed in kW. This value can be predicted based on local sunshine data (such as the sunshine curve of a typical day) or obtained by clustering historical data.

[0083] Based on the system net load, the peak-to-valley difference of the net load is obtained as follows:

[0084] In the formula, This represents the maximum net load of the system after adding energy storage capacity; This represents the minimum net load of the system after adding energy storage power.

[0085] S2213. To avoid the two sub-objectives affecting the weighting effect due to different dimensions, the percentage change before and after optimization is used for dimensionless calculation. Let the sub-objective values ​​before optimization (i.e., without energy storage) be respectively... and The dimensionless sub-objectives are as follows: ,

[0086] S2214. The multi-objective function of minimizing the daily operating cost of the system and minimizing the peak-to-valley difference of the net load is transformed into a single-objective function using the linear weighting method as follows:

[0087] In the formula, The objective function is to minimize the average daily operating cost of the system. Weighting coefficients; The objective function to minimize the peak-to-valley difference in net load The weighting coefficients; where the weighting coefficients satisfy... , , ;and and Set the parameters according to the importance of the corresponding objective function.

[0088] By adjusting the weighting coefficients between the objective function of minimizing the daily operating cost of the system and the objective function of minimizing the peak-to-valley difference in net load, different emphases can be achieved between reducing the peak-to-valley difference in net load and lowering the average daily operating cost of the system, thus obtaining differentiated optimization results. That is, by adjusting... and A trade-off can be struck between economic efficiency and peak shaving capacity, increasing... They place greater emphasis on reducing the peak-to-valley difference in net load and increasing They place greater emphasis on reducing operating costs.

[0089] It should be noted that minimizing the daily operating cost and the peak-to-valley difference of the net load as the objective function is only one option for the lower-level scheduling model in this application; there can be more sub-objectives. When using the linear weighting method to transform each sub-objective into a single-objective function, weight coefficients are set for each sub-objective. And satisfy That is, the weight coefficients of each sub-objective. It can also be adjusted according to its importance.

[0090] To avoid the impact of inconsistent dimensions on the optimization results, each sub-objective is also dedimensionalized, and the percentage change in the objective function before and after optimization is used as the evaluation index. The dedimensionalized objective function is as follows:

[0091] In the formula, To optimize the first Individual target values; To optimize the first Individual target values; This represents the total number of sub-objectives.

[0092] S222. Establish constraints for the lower-level scheduling model; The lower-level scheduling model must satisfy the following operational constraints: S2221, Energy storage system charging and discharging power constraints:

[0093] In the formula, Rated power of a single energy storage unit, in kW; The number of energy storage units is determined by the upper-level planning model.

[0094] To avoid simultaneous charging and discharging, an additional constraint is established:

[0095] S2222, State of Charge (SOC) Constraints for Energy Storage: The recursive relationship for the state of charge (SOC) of energy storage is as follows:

[0096] In the formula, To improve energy storage charging efficiency; For energy storage discharge efficiency; The rated capacity of a single energy storage unit is expressed in kWh.

[0097] The upper and lower limits of the energy storage state of charge (SOC) are established as follows:

[0098] In the formula, , These are the minimum and maximum allowable values ​​for the state of charge of energy storage, respectively.

[0099] To ensure the sustainable operation of the energy storage system, the initial and final balance constraints of the energy storage state of charge (SOC) are established as follows:

[0100] In the formula, For the first In a typical daily scenario, the start time of the scheduling cycle ( The state of charge of an energy storage system represents the percentage of remaining electricity in the system at the start of the day. For the first In a typical daily scenario, the end time of the scheduling cycle ( The state of charge of the energy storage system (corresponding to the end of 24 hours) ensures that the energy storage system returns to the same state of charge as at the beginning of the day, thereby meeting the starting conditions for operation the next day and achieving "daily cycle balance" of energy storage capacity. The initial state of charge setting value for the energy storage system, for example, in one specific embodiment, is taken as... This means that the state of charge at both the start and end times is 50%, which allows for sufficient charging and discharging space for the energy storage system (neither fully charged nor completely empty), which helps extend battery life and ensures scheduling flexibility.

[0101] S2223, System power balance constraints: The power purchased by the grid, the output of photovoltaic power, the charging and discharging of energy storage and the charging load need to be balanced in real time. Therefore, the power balance constraint is established as follows:

[0102] In the formula, For the first In a typical daily scenario, time The power purchased from the power grid, in kW, is a non-negative variable representing the amount of power that the photovoltaic-storage supercharging station draws from the main power grid. For the first In a typical daily scenario, time The photovoltaic output power, in kW, is predicted from local solar irradiance data and represents the electrical energy provided by the photovoltaic power generation system to the photovoltaic-storage supercharging station at this moment. For the first In a typical daily scenario, time The discharge power of the energy storage system, measured in kW, is a decision variable in the lower-level scheduling model, representing the rate at which the energy storage battery releases electrical energy to the load. For the first In a typical daily scenario, time The electric vehicle charging load, in kW, is given by the typical daily load curve obtained in step S1, representing the total electricity demand of all charging piles. For the first In a typical daily scenario, time The charging power of the energy storage system, measured in kW, is a decision variable in the lower-level scheduling model, representing the rate at which the energy storage battery absorbs electrical energy from the system.

[0103] S3. Considering the typical characteristics of photovoltaic-storage supercharging stations, an improved quantum particle swarm optimization algorithm and a mixed integer linear programming solver are introduced to solve the two-layer model in a collaborative manner, so as to obtain the optimal configuration scheme of transformer and energy storage. Please see Figure 6 Based on the classical Quantum Particle Swarm Optimization (QPSO) algorithm, this paper introduces an adaptive adjustment mechanism for the Levy flight step size based on the charging pile occupancy rate, tailored to the characteristics of photovoltaic-storage supercharging stations. This mechanism dynamically matches the algorithm's search behavior with the actual load state of the charging station. A particle position correction operator based on the feasible region of the energy storage load is introduced to ensure that the energy storage charging and discharging plans corresponding to all candidate solutions (i.e., particles) strictly satisfy the physical constraints of the state of charge. A dynamic adjustment mechanism based on the fitness function of the net load peak-valley difference is also introduced to adaptively balance the economic objective of the overall system cost with the peak-valley difference reduction objective during the optimization process. Simultaneously, the multiple introduced mechanisms work together to enable the improved quantum particle swarm optimization algorithm (IQPSO-ESS) to efficiently and stably solve the transformer and energy storage coordinated configuration problem of photovoltaic-storage supercharging stations, outputting an optimal solution that satisfies both power supply reliability and investment economy. Details are as follows: S31. Initialize the parameters of the improved quantum particle swarm algorithm; S311. Set basic algorithm parameters: particle swarm size Maximum number of iterations Optimize variable dimensions The optimization variables include transformer capacity. (Corresponding to the first dimension, unit: kVA), Rated energy storage power (Corresponding to the second dimension, unit: kW) and rated energy storage capacity (Corresponding to the second dimension, unit: kWh).

[0104] S312, Initialize particle positions as follows: , ;

[0105] In the formula, In order to be in A random number that is uniformly distributed within the range; and The first The lower and upper limits of the dimension variable are determined based on the actual engineering constraints; For the first The particle in the first The initial position of the dimension.

[0106] S313. Initialization of individual optimal position: In the formula, For the first The particle in the first The historical best position of dimensionality.

[0107] S314. Global Optimal Position Initialization: Initial time, global optimal position... Take the position corresponding to the particle with the lowest fitness among all particles, and calculate its fitness in the following steps S32.

[0108] S32. Introduce a dynamic adjustment mechanism for the fitness function based on the peak-valley difference of net load to calculate particle fitness; For each particle, its encoded configuration parameters (transformer capacity) Rated energy storage power Rated energy storage capacity This is passed to the lower-level scheduling model. The data generated in step S1 is then passed on. For each typical daily scenario, the lower-level MILP model is solved using the Gurobi solver to obtain the daily operating cost for each scenario. and the difference between peak and valley loads .

[0109] Based on the system's average daily comprehensive cost in step S2 Peak-to-valley difference with net load The dynamic multi-objective fitness function (which considers both economy and peak shaving capability) is constructed as follows:

[0110] In the formula, The dynamic weighting coefficients are adaptively adjusted based on the number of iterations and the charging pile occupancy rate, as follows:

[0111] In the formula, As the initial weights, for ease of subsequent detailed description, this embodiment uses... ; This is the current iteration number, starting from 0; Maximum number of iterations ( ); For the sake of clarity and detailed explanation later, this embodiment uses the diversity adjustment coefficient. ; The population diversity index is the ratio of the standard deviation of particle positions to the mean, where... The standard deviation of all particle positions, The mean value of all particle positions (calculated separately in each dimension and then averaged). As a scenario adjustment factor, based on the average occupancy rate of charging piles. Sure:

[0112] in, The average occupancy rate of charging piles under the current optimized scenario (i.e., the weighted average of all typical daily scenarios) can be obtained based on historical data statistics.

[0113] When calculating fitness, for each particle, its configuration parameters are substituted into the lower-level scheduling model, and the result is obtained after Gurobi algorithm solving. and Then calculate the fitness using the formula above. .

[0114] S33. Update the individual optimal and global optimal, and calculate the local attraction point and average optimal position; S331. Update the individual optimal and global optimal; For each particle Compare current fitness Its historical best fitness :

[0115] In the formula, This represents the optimal position for the updated individual.

[0116] Global optimal position update: ,

[0117] That is, the position of the particle with the lowest fitness among all current particles is taken as the global optimal position.

[0118] S332. Calculate the local attraction point and the average optimal position; The local attraction point is calculated as follows:

[0119] in,

[0120] In the formula, , For the sake of clarity in the following detailed description, this embodiment uses the learning factor. ; , In order to be in Independent random numbers that are uniformly distributed within the range.

[0121] The average optimal position is calculated as follows:

[0122] In the formula, Indicates the first The average of the optimal positions of all individual particles in the dimension.

[0123] S34. Introduce a Levy flight step size adaptive adjustment mechanism based on charging pile occupancy rate, dynamically calculate Levy flight step size parameters, and update particle positions; S341. Based on the current optimized scenario, calculate the average charging pile occupancy rate of the photovoltaic-storage supercharging station; For the current optimized scenario (i.e., the weighted average of all typical daily scenarios), the average charging pile occupancy rate is calculated as follows:

[0124] In the formula, This represents the total number of charging stations; For a moment The number of charging stations currently charging; This represents the total number of sampling points per day (i.e., the number of time windows).

[0125] S342. Determine the Levy dynamic step size parameters based on the average charging pile occupancy rate, and generate the Levy flight step size; Based on average charging pile occupancy rate Determine the Levy dynamic step size parameters for:

[0126] Based on dynamic step size parameters The Levy flight stride was generated using the Mantegna algorithm as follows:

[0127] In the formula, , The parameters satisfy: , In the formula, This is the Gamma function.

[0128] S343. Based on the Levy flight step size, a contraction and expansion coefficient is introduced to obtain an improved particle position update formula to update the particle position. By introducing the Levy flight term into the standard QPSO position update equation, we obtain an improved position update formula:

[0129] In the formula, In order to be in Uniformly distributed random numbers, when Use a plus sign if the condition is met, otherwise use a minus sign. The contraction and expansion coefficients are adaptively adjusted.

[0130] In the formula, , ; The current particle fitness is calculated in step S32; , These are the average fitness and minimum fitness of the current population, respectively.

[0131] S35. Introduce a particle position correction operator based on the feasible region of energy storage load to correct the feasible region of load; For each particle's energy storage charge-discharge sequence, the following correction procedure is performed to ensure that the state of charge (SOC) at all times satisfies the physical constraints.

[0132] S351, The recursive law for the state of charge (SOC) of an energy storage system is as follows:

[0133] In the formula, , These represent the energy storage charging efficiency and the discharging efficiency, respectively, both of which are set to 0.95 in this embodiment. , They are time points The charging power and discharging power are expressed in kW. , where is the time interval.

[0134] Meanwhile, the state of charge (SOC) of the energy storage system satisfies the upper and lower limits of the energy storage SOC in step S2222. In this embodiment, , Furthermore, it satisfies the initial and final balance constraints of the energy storage state of charge (SOC). In this embodiment, .

[0135] S352. When calculating the feasible region of the state of charge (SOC) of the energy storage system, calculate the forward reach threshold and the backward reach threshold respectively, and obtain the comprehensive feasible threshold based on the forward reach threshold and the backward reach threshold. Forward reachable regions (recursively from the start time):

[0136]

[0137] Backward reachable domain (recursively from the termination time backward):

[0138]

[0139] Comprehensive feasible domain:

[0140]

[0141] In the above formula, , These are the upper and lower limits of the state of charge (SOC) of the energy storage system, respectively. In this embodiment, , (Same as in step S351); , These refer to the initial state of charge (SOC) and the final state of charge (SOC) of the energy storage system, respectively. In this embodiment, (Same as in step S351); , These represent the charging efficiency and discharging efficiency of the energy storage system, respectively, which are taken as 0.95 in this embodiment. The total power rating of the energy storage system, in kW, is determined by the decision variables of the upper-level planning layer. Give; , where is the time interval; This represents the total number of sampling points per day (i.e., the number of time windows). For time period index, .

[0142] S353. When performing over-limit detection and correction, compare the state of charge of the energy storage system with the boundary value of the comprehensive feasible threshold, and correct the charging and discharging power of the energy storage system according to the principle of proportionality or symmetrical method. set up Index of the time period in which the limit was exceeded ( If a time exists Make The over-limit quantity is calculated as follows:

[0143] At any moment front and back Within each time period (i.e., 1 hour before and after), the charging and discharging power is adjusted proportionally as follows:

[0144]

[0145] In the formula, In order to adjust the total number of time periods, this embodiment ; It is a decay factor that maximizes the adjustment amount at the moment of exceeding the limit and decreases towards both sides.

[0146] The adjustment amount for adjusting the charging and discharging power varies with distance. Decay: Distance exceeding the limit The closer the time period, the greater the adjustment range of its charging and discharging power; the further away from the limit, the greater the adjustment range. The further back in time the time period, the smaller the adjustment range of its charging and discharging power. This allows the correction operation to be mainly concentrated near the time point when the limit is exceeded, thus avoiding excessive disturbance to the normal scheduling that is far from the time period when the limit is exceeded.

[0147] If a time exists Make Then, the correction is made symmetrically (i.e., increasing charging or decreasing discharging) as follows:

[0148] .

[0149] S354. Based on the corrected energy storage system charging and discharging power, perform boundary truncation processing; If the charging / discharging power exceeds the physical limit after correction, it will be truncated to the boundary value:

[0150]

[0151] Repeat steps S353 and S354 to correct the process until the SOC at all times satisfies the constraints or reaches the maximum number of corrections (5 times in this embodiment). After correction, substitute the corrected charge-discharge sequence back into the lower-level scheduling model to update the particle's fitness.

[0152] S36. Based on the algorithm termination condition, perform convergence judgment and output the optimal configuration scheme including the optimal transformer capacity, the optimal energy storage rated power, and the optimal energy storage rated capacity. Determine if the algorithm meets the termination condition: or

[0153] In the formula, This represents the current iteration number. ; This represents the maximum number of iterations. For the first During the nth iteration The global optimal position of the dimension; This is the convergence accuracy threshold.

[0154] If the termination condition is met, output the globally optimal position. The optimal configuration is as follows: (Optimal transformer capacity) Optimal energy storage rated power Optimal energy storage rated capacity (where the transformer capacity is in kVA, the rated power of energy storage is in kW, and the rated capacity of energy storage is in kWh); otherwise, return to step S32 to continue iterating.

[0155] In summary, by dynamically adjusting the Levy flight step size based on the charging pile occupancy rate on the basis of the classical quantum particle swarm optimization algorithm, the algorithm can automatically adapt to the load fluctuation characteristics of the photovoltaic-storage supercharging station at different times. It enhances global exploration during peak hours and improves local convergence accuracy during off-peak hours. Through the load (SOC) feasible region correction operator, it ensures that the energy storage operation trajectory corresponding to all particles strictly meets the physical constraints, avoiding the problem of difficult parameter tuning in the traditional penalty function method. Through the dynamic weight adjustment mechanism, the algorithm adaptively balances the economic goal and the peak shaving goal during the iteration process, and automatically increases the peak shaving weight during peak hours, realizing multi-objective optimization that is adaptive to the scenario.

[0156] The Levy flight step size adaptive adjustment mechanism based on charging pile occupancy rate, the particle position correction operator based on energy storage load feasible region, and the fitness function dynamic adjustment mechanism based on net load peak-valley difference work together to improve the solution accuracy, convergence speed, and solution feasibility of the improved quantum particle swarm optimization algorithm (i.e., IQPSO-ESS algorithm) for photovoltaic-storage-supercharging station application scenarios in the capacity configuration problem of photovoltaic-storage-supercharging stations.

[0157] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0158] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0159] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0160] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0161] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0162] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing configuration of a transformer of a light storage and supercharging station in coordination with energy storage, characterized in that, include: Based on historical charging load data from multiple scenarios, a weighted K-means clustering algorithm based on peak retention is introduced to obtain typical daily load curves and their occurrence probabilities. Based on typical daily load curves and their occurrence probabilities, an upper-level planning model and a lower-level scheduling model for the coordinated optimization of transformer and energy storage configuration are established. Considering the typical characteristics of photovoltaic-storage-supercharging stations, an improved quantum particle swarm optimization algorithm is introduced to solve the upper-level planning model, and a mixed-integer linear programming solver is used to collaboratively solve the lower-level scheduling model to obtain the optimal configuration scheme of transformers and energy storage.

2. The transformer and energy storage coordination optimization configuration method of a light storage supercharging station according to claim 1, characterized in that, The introduced peak-preserving weighted K-means clustering algorithm includes: Considering the peak characteristics of the photovoltaic-storage supercharging station, weighting coefficients for each characteristic are set based on the importance of the characteristic quantities that characterize the load impact characteristics and the peak indication. By incorporating weighting coefficients into the distance function of the K-means clustering algorithm, a weighted distance function that retains peak values ​​is obtained. By iterating using the weighted distance function, typical daily load curves and their probability of occurrence are obtained.

3. The transformer and energy storage coordination optimization configuration method of a light storage supercharging station according to claim 1, characterized in that, An improved quantum particle swarm optimization algorithm is introduced to solve the upper-level programming model, including: The algorithm parameters are initialized, and a dynamic adjustment mechanism based on the fitness function of the net load peak-valley difference is introduced to calculate the particle fitness. Based on particle fitness, update the individual optimal position and the global optimal position, and calculate the local attraction point and the average optimal position.

4. The transformer and energy storage coordination optimization configuration method of a light storage supercharging station according to claim 1 or 3, characterized in that, An improved quantum particle swarm optimization algorithm is introduced to solve the upper-level programming model, including: An adaptive adjustment mechanism for Levy flight step size based on charging pile occupancy rate is introduced to dynamically calculate Levy flight step size parameters and update particle positions; A particle position correction operator based on the feasible region of the energy storage charged state is introduced to correct the feasible region of the energy storage charged state.

5. The transformer and energy storage coordination optimization configuration method of a light storage supercharging station according to claim 4, characterized in that, An adaptive adjustment mechanism for Levy flight step size based on charging pile occupancy rate is introduced to dynamically calculate Levy flight step size parameters, including: Based on the current optimization scenario, calculate the average charging pile occupancy rate of the photovoltaic-storage supercharging station; The Levy dynamic step size parameters are determined based on the average charging pile occupancy rate, and the Levy flight step size is generated. Based on the Levy flight step size, a contraction-expansion coefficient is introduced to obtain an improved particle position update formula.

6. The transformer and energy storage coordination optimization configuration method of a light storage supercharging station according to claim 4, characterized in that, The feasible region for correcting the energy storage state of charge includes: Based on the recursive law of the state of charge of the energy storage system, the forward reachability threshold and the backward reachability threshold are calculated, and the comprehensive feasible threshold is obtained. The state of charge of the energy storage system is compared with the boundary value of the comprehensive feasible threshold. The charging and discharging power of the energy storage system is corrected according to the principle of proportionality or symmetry, and boundary truncation is performed.

7. The method for coordinated optimization configuration of transformer and energy storage in a photovoltaic-storage supercharging station according to claim 1, characterized in that, The establishment of the upper-level planning model and lower-level scheduling model for the coordinated optimization of transformer and energy storage configuration includes: With the goal of minimizing the system's average daily comprehensive cost, an upper-level planning model is constructed, which includes constraints on the number of energy storage units and / or transformer capacity. With the goal of minimizing the daily operating cost and the peak-to-valley difference of the net load, a lower-level scheduling model is constructed, which includes constraints on the charging and discharging power of the energy storage system, constraints on the state of charge of the energy storage system, and constraints on power balance.

8. The method for coordinated optimization configuration of transformer and energy storage in a photovoltaic-storage supercharging station according to claim 2, characterized in that, The characteristic quantities characterizing the load impact properties include maximum power, average power, and peak duration. When normalizing feature quantities, minimum-maximum normalization is performed on the maximum power sequence and the average power sequence. The peak duration sequence is normalized using fractional scaling.

9. The method for coordinated optimization configuration of transformer and energy storage in a photovoltaic-storage supercharging station according to claim 1, characterized in that, The process of obtaining a typical daily load curve and its probability of occurrence includes: Historical charging load data is preprocessed, and evaluation indicators are obtained based on the preprocessing results to determine the optimal number of clusters. Based on the preprocessing results and the optimal number of clusters, a weighted distance function is designed for clustering, and the maximum power sequence of each cluster center is used as a typical daily load curve. Based on the number of sample days for typical daily load scenarios and the total number of days in the typical daily load curve category, the probability of occurrence of the corresponding typical daily load curve is obtained.

10. The method for coordinated optimization configuration of transformer and energy storage in a photovoltaic-storage supercharging station according to claim 1, characterized in that, The optimal configuration scheme includes the optimal transformer capacity, the optimal rated power of energy storage, and the optimal rated capacity of energy storage. The average daily comprehensive cost of the system in the upper-level planning model includes the average daily operating cost and the average daily investment cost; wherein, the average daily operating cost is obtained based on the weighted average of the typical daily load curve and its occurrence probability, and includes the cost of purchasing electricity from the grid, the cost of energy storage operation and maintenance, and the cost of transformer penalty. The objective function of the lower-level scheduling model is a single objective function obtained by dedimensionalizing the sub-objectives of minimizing the daily operating cost of the system and minimizing the peak-valley difference of the net load, introducing sub-objective weight coefficients, and transforming them using a linear weighting method.

Images