Photovoltaic fluctuation adaptive power distribution network electric vehicle carrying capacity evaluation method and system

Abstract

The application discloses a photovoltaic fluctuation adaptive power distribution network electric vehicle carrying capacity evaluation method and system, and belongs to the technical field of power system planning and operation. In view of the problem that the fixed robust budget parameter in the existing robust evaluation method cannot adapt to the dynamic change of photovoltaic output, the application proposes an adaptive framework combining offline training and online application. Firstly, the historical photovoltaic output daily curve is classified through a time series clustering algorithm, and various typical fluctuation types are divided. Then, for each type, the voltage safety is taken as a hard constraint, and the maximum carrying capacity is taken as an objective, and the optimal robust budget parameter is solved offline. In online evaluation, the fluctuation type of the target day is identified, and the corresponding optimal parameter is called to construct an adaptive box-type uncertainty set, solve a two-stage robust optimization model, and output the carrying capacity evaluation result. The application realizes intelligent matching of the evaluation strategy and the photovoltaic fluctuation characteristics, and significantly improves the carrying capacity evaluation precision under the premise of ensuring safety.

Description

Technical Field

[0001] This invention relates to the field of power system analysis and planning technology, and more specifically, to a method and system for assessing the carrying capacity of electric vehicles in a distribution network that is adaptive to photovoltaic fluctuations. Background Technology

[0002] The penetration rate of distributed photovoltaic (PV) power and electric vehicles (EVs) in power distribution networks continues to increase. Distributed PV output is significantly affected by various meteorological factors such as solar irradiance and cloud cover, exhibiting significant intermittency and randomness; EV charging loads show a clustering effect in time and space. Together, these factors constitute the challenge of "dual uncertainty of source and load" in power distribution network planning. The carrying capacity of a power distribution network for EVs refers to the maximum EV charging load that the grid can accommodate while meeting all safe operating constraints. Accurately assessing this capacity is a crucial step in scientifically guiding the planning of charging infrastructure.

[0003] Currently, robust optimization is the mainstream method for assessing load capacity. This method constructs a set containing the range of uncertainties in photovoltaic output and solves for the maximum grid-connected capacity that can still operate safely under the "worst-case" scenario within this uncertainty set. The robust budget parameter Γ is a key parameter controlling the conservatism of the model. The larger the value of Γ, the wider the uncertainty set and the more conservative the assessment result; the smaller the value of Γ, the more optimistic the assessment but the security may be insufficient.

[0004] Current technologies generally use a fixed Γ value for all evaluation scenarios. However, photovoltaic output characteristics vary drastically with weather conditions: short-term large fluctuations may occur in sunny weather, requiring a larger Γ value to ensure safety; fluctuations are mild in continuous cloudy or rainy weather, where a smaller Γ value is sufficient, and a large Γ value may lead to an overly conservative evaluation. A fixed Γ value cannot adapt to this dynamic change, making it difficult to achieve a dynamically optimal balance between safety and economy in the evaluation results.

[0005] Through in-depth research and practice, this invention has discovered a deeper flaw in existing technologies: the setting of the robust budget parameter Γ is completely disconnected from the actual physical fluctuations of photovoltaic output. Whether determined empirically or through trial and error, a fixed Γ value implicitly assumes that the uncertainty of photovoltaic power generation poses the same "threat mode" to the grid under drastically different weather conditions such as sunny, cloudy, and rainy days. In reality, different fluctuation types have different impact mechanisms on voltage safety boundaries, and therefore require different safety margins. Therefore, a mechanism is urgently needed to intelligently match the robustness of grid assessment with the physical characteristics of photovoltaic fluctuations, fundamentally resolving the contradictions caused by the aforementioned "one-size-fits-all" approach. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing fixed robust budget parameter evaluation methods and provide a photovoltaic fluctuation adaptive distribution network electric vehicle carrying capacity evaluation method and system. By establishing a mapping relationship between fluctuation type and optimal robust budget parameter through offline training, the optimal parameter is adaptively called during online evaluation, which significantly improves the accuracy and economy of carrying capacity evaluation while ensuring safety.

[0007] The core concept of this invention lies in identifying and utilizing the inherent mapping relationship between "photovoltaic power output fluctuation type" and "robust budget parameters required to ensure voltage safety." This mapping knowledge base is established through offline learning, and during online evaluation, the optimal parameters are proactively invoked by sensing real-time fluctuation types, thereby achieving dynamic adaptive adjustment of the conservatism of the evaluation model. This concept transforms traditional static, passive robustness evaluation into dynamic, proactive intelligent defense.

[0008] To achieve the above objectives, the present invention adopts the following technical solution.

[0009] This invention provides a photovoltaic fluctuation adaptive method for evaluating the carrying capacity of electric vehicles in distribution networks, comprising three stages: an offline calibration stage for fluctuation types, an offline training stage for optimal robust budget parameters, and an online adaptive evaluation stage.

[0010] Offline calibration phase for fluctuation types: Based on historical solar irradiance time-series data of the distribution network area, the data is segmented by day to form a daily photovoltaic output curve. A time-series clustering algorithm is then used to divide all historical photovoltaic output daily curves into... A typical type of fluctuation, It is a natural number not less than three. Experiments have verified that when... When the value is 3, the clustering results stably correspond to three typical types: strong fluctuation type in sunny weather, medium fluctuation type in cloudy weather, and weak fluctuation type in cloudy and rainy weather, which can comprehensively cover the typical fluctuation characteristics of photovoltaic power output.

[0011] Offline training phase for optimal robust budget parameters: For each typical fluctuation type, the specific process of searching for the optimal robust budget parameters within a preset value range includes: fitting the historical daily photovoltaic output curves for that type using a probability distribution function, and generating a massive sample of photovoltaic output scenarios for that type using Monte Carlo simulation. A hard safety constraint is set that the distribution network node voltage does not exceed the limit under all historical statistical scenarios of that type. This means that the node voltage amplitude remains within a safe range of ±7% of the nominal voltage. With maximizing the electric vehicle evaluation carrying capacity of the distribution network as the optimization objective, a parameter search is performed within the preset value range of the robust budget parameters to obtain the optimal robust budget parameters for that type. For all The process iterates through all types of fluctuations to build a knowledge base that maps fluctuation types to optimal robust budget parameters.

[0012] Online adaptive evaluation phase: For the target date to be evaluated, obtain the short-term photovoltaic output forecast curve for that day, and determine its fluctuation type label by similarity matching with the cluster center curve corresponding to the season or month of the target date; retrieve the optimal robust budget parameter corresponding to the label from the mapping knowledge base. Based on the optimal parameters invoked, combined with the expected value sequence obtained by probability distribution fitting and the fluctuation amplitude sequence derived from historical statistical fluctuation boundaries, an adaptive box uncertainty set applicable to the target day is constructed; taking maximizing the total capacity of electric vehicles connected to the distribution network as the upper-level objective and taking the safe operation of the system under the worst scenario covered by the adaptive box uncertainty set as the lower-level objective, a two-stage robust optimization model is established and solved, and the carrying capacity assessment results are output.

[0013] The above scheme fundamentally changed the fixed The evaluation model obtains the optimal balance between safety and economy for each fluctuation type through offline training. During online evaluation, it adaptively matches the optimal parameters according to the fluctuation type, so that the evaluation results can automatically approach the most economic point on the safety red line under any weather conditions, systematically improving the accuracy of carrying capacity evaluation and the utilization rate of power grid assets.

[0014] Furthermore, the temporal clustering algorithm employs the dynamic time warping algorithm to calculate the similarity distance between any two daily curves, and uses... Clustering is performed using the -means++ clustering algorithm. The dynamic time warping algorithm can effectively align phase-shifted curves and accurately measure the similarity of fluctuation patterns; -means++'s clustering algorithm optimizes the selection of initial cluster centers, improving cluster stability and repeatability.

[0015] Furthermore, the probability distribution function is a beta distribution. The beta distribution is defined on a closed interval... The mathematical expression for the probability density function of a continuous probability distribution on the given surface is:

[0016]

[0017] In the formula: Let be a normalized random variable of irradiance, with a range of values ​​of . ; and These are two shape parameters of the beta distribution, both real numbers greater than zero, which together determine the shape of the distribution curve; For beta functions, defined as

[0018]

[0019] Shape parameters and The shape parameter was estimated from historical irradiance data using the maximum likelihood estimation method. Since the log-likelihood function does not have an explicit analytical solution with respect to the parameter, the Newton-Raphson iterative method was used for numerical solution. Alternatively, the method of moments could be used to obtain initial parameter values ​​before fine-tuning the iterations. To characterize the differences in irradiance statistical features across different time periods, the shape parameter was estimated separately for each scheduling period.

[0020] Based on the fitted beta probability distribution, a Monte Carlo simulation method was used to generate photovoltaic power output scenario samples, with a sampling number of no less than 5,000 to ensure statistical representativeness. The sampled irradiance samples were multiplied by the photovoltaic installation area and the photovoltaic module conversion efficiency under standard test conditions to convert them into photovoltaic power output scenario samples.

[0021] For the scenario samples generated by the Monte Carlo simulation, the high quantile and low quantile are extracted for each time period. The high quantile is taken as the 99.5th quantile, and the low quantile is taken as the 0.5th quantile. Half of the difference between the two is taken as the maximum fluctuation amplitude of photovoltaic output for that time period, denoted as . .

[0022] Adaptive Box Uncertain Set The specific mathematical expression is:

[0023] ;

[0024] The technical meanings of the symbols in the formula are as follows: For the first Variables in actual photovoltaic output over time periods; For the first The expected value of photovoltaic output for a given period is calculated by multiplying the irradiance predicted by numerical weather forecast by the photovoltaic module efficiency. The fluctuation type obtained through offline training corresponds to the target day. The optimal robust budget parameter is matched, which is exactly at the critical balance point between security and capacity maximization; For the first Maximum fluctuation amplitude of photovoltaic output during a given period; product This indicates the th parameter after using the optimal parameters. The maximum permissible unidirectional fluctuation range during the time period; As an auxiliary variable for uncertainty, in a closed interval Freely selectable values, Pick This indicates that the photovoltaic output has reached its positive upper limit. When the lower limit of the negative value is reached, take When it equals the expected value; This represents the set of scheduled time slots for a day. This uncertain set is determined by... It achieves "adaptive" control of scale—automatically adopting a larger value to ensure safety when fluctuations are severe, and automatically adopting a smaller value to reduce conservatism when fluctuations are mild.

[0025] Furthermore, in the two-stage robust optimization model, the upper layer aims to maximize the sum of the planned access capacities of all candidate nodes that allow electric vehicles to connect, with the access capacity of each node as the decision variable. The lower layer aims to minimize the total system load curtailment by adjusting the charging and discharging power of energy storage devices under the worst-case photovoltaic power output scenario covered by the adaptive box uncertainty set.

[0026] Operational safety constraints include: branch power flow constraints based on the DistFlow branch power flow model and using second-order cone relaxation treatment, the core mathematical form of which is...

[0027]

[0028] Node voltage safety constraints

[0029]

[0030] The allowable deviation is typically the nominal voltage. Energy storage device operation constraints include upper limits for charging power, upper limits for discharging power, upper and lower limits for energy storage capacity, time-series constraints related to the state of charge (SOC), and mutual exclusion constraints between charging and discharging states. The mathematical expression for the SOC time-series constraints is as follows:

[0031]

[0032] The mutual exclusion constraint of charge and discharge states introduces binary variables, which can be represented as follows:

[0033]

[0034] The model solution employs a column and constraint generation algorithm. It approximates the global optimum through iterative alternation of the main problem and subproblems. The main problem solves for the optimal access scheme and updates the lower bound based on the known set of worst-case scenarios; the subproblems utilize strong duality theory of linear programming to transform the inner-level minimization problem into its dual maximization problem, and then... The method linearizes the bilinear terms generated after transformation, converting it into a mixed-integer linear programming problem for solution. It identifies new worst-case scenarios and updates the upper bound. Iteration stops when the relative gap between the upper and lower bounds is less than a convergence threshold, which is preferably set to... Output the load-bearing capacity assessment results.

[0035] Furthermore, the method also includes online verification of safety margin: the carrying capacity assessment results are input into the distribution network digital twin simulation platform, and at least 10,000 random photovoltaic output scenarios and random load scenarios are generated based on the Monte Carlo method for probabilistic power flow calculation. The voltage over-limit probability and line overload probability are statistically analyzed. If the voltage over-limit probability is lower than the specified value, the method will determine the safety margin. This indicates that there is a sufficient safety margin.

[0036] This invention also provides a photovoltaic fluctuation adaptive electric vehicle load-bearing capacity assessment system for power distribution networks, comprising:

[0037] The data management module is used to store and manage distribution network topology parameters, historical solar irradiance data, real-time weather forecast data, load data, and carrying capacity assessment results.

[0038] An offline training engine, connected to the data management module, is used to read historical data from the data management module, perform time-series clustering of historical photovoltaic power output daily curves and search for optimal robust budget parameters, and construct and update the mapping relationship between fluctuation types and optimal robust budget parameters.

[0039] The online evaluation engine is connected to the data management module and the offline training engine respectively. It is used to receive evaluation requests and target day data, call the optimal parameters from the offline training engine, perform fluctuation type identification, adaptive box uncertainty set construction and two-stage robust optimization model solution, and output the bearing capacity evaluation results.

[0040] The human-computer interaction and visualization module is connected to the online evaluation engine and is used to provide an interface for configuring evaluation parameters and to display the load-bearing capacity evaluation results.

[0041] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the program to implement the above-described method. Additionally, a computer-readable storage medium on which a computer program stored is executed by a processor to implement the above-described method.

[0042] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0044] Figure 1 This is a flowchart of the overall process of the method of the present invention.

[0045] Figure 2 This diagram illustrates the location and capacity of electric vehicles connected to the IEEE 33-node distribution network.

[0046] Figure 3 This chart compares the maximum load capacity of electric vehicles under different scenarios. The horizontal axis represents the scenario type (Case 1, Case 2, Case 3), and the vertical axis represents the load capacity (unit: kW).

[0047] Figure 4 A comparison diagram of node voltage distribution under different scenarios. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments. The described embodiments are only a portion of the embodiments of this invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0049] like Figures 1-4 As shown, the method of the present invention includes three core stages. The first two stages are offline stages, which are executed periodically; the third stage is an online stage, which responds quickly to evaluation requests.

[0050] Phase 1: Offline calibration of fluctuation types

[0051] Based on at least three years of historical hourly solar irradiance data for the distribution network area, the data is segmented daily to form a daily photovoltaic output curve. A dynamic time warping algorithm is used to calculate the similarity distance between any two curves. Dynamic time warping finds the optimal matching path in terms of shape between two curves by nonlinearly stretching or compressing the time axis, using the cumulative distance along the matching path as a similarity measure. The specific calculation process is as follows: Let the two curve sequences be... and A cumulative distance matrix is ​​constructed, where each element is the minimum of the Euclidean distance to the current point plus three adjacent predecessor elements (left, bottom, and bottom left). The final element in the bottom right corner of the matrix represents the dynamic time-warped distance. This algorithm effectively addresses phase shift issues caused by cloud cover time drift and seasonal shifts in sunrise and sunset times.

[0052] Using dynamic time warping distance as a metric, Clustering is performed using the -means++ clustering algorithm. -means++'s clustering algorithm optimizes the initial cluster center selection strategy—the first center is randomly selected, and subsequent centers are sampled with probability based on the square of the distance to the nearest existing center, with greater probability of selection for farther distances—ensuring that the initial centers are far apart and the clustering results are stable. Pick At that time, the stable clustering results correspond to three typical types: sunny day strong fluctuation type (arched single peak, peak normalized value within) The above), fluctuating pattern in cloudy weather (multiple fluctuations, with peaks at...). to (between), cloudy and rainy with weak fluctuations (relatively flat, peak at) the following).

[0053] Phase 2: Offline training of optimal robust budget parameters

[0054] With the first Taking this type as an example, the specific process of searching for the optimal robust budget parameter within the preset value range includes: fitting the historical irradiance data of this type using a beta probability distribution, the probability density function of which is:

[0055]

[0056] In the formula: For normalized irradiance, and For shape parameters, beta function

[0057]

[0058] Shape parameters were estimated separately for each scheduling period using the maximum likelihood estimation method, and numerical solutions were obtained using the Newton-Raphson iterative method. Ten thousand Monte Carlo simulations were performed based on the fitted beta distribution. The irradiance samples were multiplied by the photovoltaic installation area and the module conversion efficiency under standard test conditions to convert them into photovoltaic output scenario samples. The samples were then extracted time-by-time. quantiles and Quantiles are calculated by taking half the difference between the two values ​​as the fluctuation amplitude. .

[0059] by exist The optimal value is searched using the golden section search method within the specified interval. Two probe points are selected for each iteration, and the carrying capacity assessment model is run on each. The assessment model aims to maximize the sum of the connected capacities of all candidate nodes. Under the sample of all historical photovoltaic output scenarios corresponding to this typical fluctuation type, the voltage amplitude of all nodes in the distribution network is within the nominal voltage range. The area is subject to strict safety constraints. The load-bearing capacity varies during the search. The parameters first rise and then fall; the peak point is the optimal robust budget parameter. Traverse all Types are used to construct a mapping knowledge base.

[0060] Phase 3: Online Adaptive Evaluation

[0061] Obtain the short-term photovoltaic output forecast curve for the target day, calculate its dynamic time-warped distance from the curves of each cluster center, and select the type with the smallest distance as the target day label. Retrieved from the knowledge base Construct an adaptive box-type uncertainty set applicable to the target date:

[0062] ;

[0063] In the formula: For the first Variables in actual photovoltaic output over time periods; Expected contribution to photovoltaic power generation; The optimal robust budget parameters for matching; This represents the maximum fluctuation range of photovoltaic power output. As an auxiliary variable for uncertainty, in Internal value; This is the set of scheduling time periods.

[0064] A two-stage robust optimization model is established. The upper layer aims to maximize the sum of the planned electric vehicle access capacity of each candidate node, with the access capacity of each node as the decision variable. The lower layer searches for the worst-case photovoltaic power output scenario in an adaptive uncertainty set, and minimizes the total load curtailment by adjusting the energy storage charging and discharging power.

[0065] The constraints include: second-order cone relaxation constraints for DistFlow branch power flow, the core form of which is...

[0066]

[0067] Node voltage safety constraints

[0068]

[0069] Voltage amplitude at Within the nominal voltage range; energy storage operation constraints. In the energy storage constraints, the state-of-charge (SOC) timing correlation constraint is expressed as follows:

[0070]

[0071] The mutual exclusion constraint of charge and discharge states introduces binary variables, which can be represented as follows:

[0072]

[0073] The column and constraint generation algorithm is used to solve the problem, and the convergence threshold is preferably set to [value missing]. The main problem is to find the optimal access scheme and update the lower bound based on a known set of scenarios. The subproblems utilize strong duality theory of linear programming to transform inner-level minimization into dual maximization, through large... The method linearizes the bilinear terms, transforming them into a mixed-integer linear programming problem, and then solves it using CPLEX or Gurobi. It identifies new worst-case scenarios and updates the upper bound. The bearing capacity assessment result is output when the gap between the upper and lower bounds is less than the convergence threshold.

[0074] Optionally, the evaluation results are input into a digital twin simulation platform to perform no less than 10,000 Monte Carlo probabilistic power flow calculations, with the voltage over-limit probability being lower than... To verify the safety margin for standards.

[0075] Evaluation System Implementation Examples

[0076] The evaluation system includes: a data management module that uses a time-series database and a relational database to store meteorological and power grid data respectively; an offline training engine that includes a time-series clustering unit, a parameter optimization unit, and a knowledge base management unit, and automatically starts training periodically; an online evaluation engine that completes type identification and parameter matching in milliseconds and model solving in minutes, and is tightly integrated with CPLEX or Gurobi solver interfaces; and a human-computer interaction and visualization module that provides a web interface for planners to configure parameters and view results.

[0077] Simulation verification and effect analysis

[0078] Simulation verification was performed on the IEEE 33-node standard distribution network test system. The system reference voltage was 12.66 kV, the reference power was 10 MVA, with 33 nodes and 32 branches. Photovoltaic access nodes were 10, 16, 20, 24, and 30, with a total installed capacity equal to [percentage missing]% of the peak load. Energy storage is configured at the aforementioned nodes, with a capacity equal to that of the photovoltaic capacity. Charge and discharge efficiency State of charge range to The photovoltaic data uses hourly measured irradiance data from a specific region over three years. Pick Monte Carlo simulation 10,000 times, convergence threshold .

[0079] Three comparison methods are set up: Comparison Method 1 adopts... (Traditional conservative and robust), compared to method two. (Industry experience value) The method of this invention adopts adaptive... (Sunny Day Appointment) Cloudy Overcast and rainy ).

[0080] Historical daily back-calculation assessment results for the whole year: The comparative method has an average annual load capacity of 1250kW, exceeding the zero-voltage limit, but is severely conservative; the comparative method two has an average annual load capacity of 1580kW, an improvement over method one. However, there were three days of voltage exceeding the limit throughout the year; the method of this invention has an average annual load capacity of 1820kW, which is higher than that of method one. Improved compared to method two Furthermore, it achieves zero voltage exceedance throughout the year, with safety comparable to Method 1 and significantly better economic performance than the two comparative methods.

[0081] The simulation results above strongly validate the core findings and technical effects of this invention. Comparing the evaluation results of Method 1 and Method 2, which represent two static strategies—"safety extremes" and "economic risk-taking"—neither can adapt to the dynamic changes in photovoltaics. In contrast, the method of this invention accurately identifies the daily fluctuation type and matches it with an offline optimized... This successfully achieved global security assurance (equivalent to the most conservative strategy, with zero overruns throughout the year) and a significant leap in global economic efficiency (broadening carrying capacity far exceeding that of the two fixed strategies). The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A photovoltaic fluctuation adaptive method for assessing the carrying capacity of electric vehicles in distribution networks, characterized in that, Includes the following steps: Offline Phase: Historical daily photovoltaic output curves are subjected to time-series clustering to classify K typical fluctuation types, where K is not less than three. For each typical fluctuation type, the voltage amplitude of distribution network nodes is maintained at a certain level under scenario samples generated based on historical data of that type through probability distribution fitting and Monte Carlo simulation, with the voltage amplitude remaining within the nominal voltage range. Within the safe range, there are hard constraints. With the goal of maximizing the total capacity of electric vehicles connected to the distribution network, the robust budget parameters are optimized within the preset value range of the robust budget parameters to obtain the optimal robust budget parameters for this typical fluctuation type. A mapping relationship between the typical fluctuation type and the optimal robust budget parameters is then constructed. Online phase: Identify the fluctuation type of the target day, call the optimal robust budget parameter corresponding to the fluctuation type of the target day from the mapping relationship, construct an adaptive box uncertainty set based on the called optimal robust budget parameter, solve a two-stage robust optimization model with maximizing electric vehicle access capacity as the upper-level objective based on the constructed adaptive box uncertainty set, and output the maximum allowable access capacity of electric vehicles for each candidate node of the distribution network on the target day.

2. The photovoltaic fluctuation adaptive method for assessing the carrying capacity of electric vehicles in distribution networks according to claim 1, characterized in that, The time-series clustering uses the dynamic time warping algorithm to calculate the similarity distance between any two daily curves, and then uses the K-means++ clustering algorithm for clustering.

3. The photovoltaic fluctuation adaptive method for assessing the carrying capacity of electric vehicles in distribution networks according to claim 1, characterized in that, The search for optimal robust budget parameters includes: fitting historical photovoltaic power output data of this type using the beta probability distribution; estimating the two shape parameters of the beta distribution using the maximum likelihood estimation method according to the scheduling period; generating photovoltaic power output scenario samples using the Monte Carlo simulation method; and performing Monte Carlo simulation sampling no less than 5,000 times.

4. The photovoltaic fluctuation adaptive method for evaluating the carrying capacity of electric vehicles in distribution networks according to claim 3, characterized in that, The adaptive box uncertainty set is constructed as follows: based on the historical photovoltaic output statistics sample corresponding to the fluctuation type of the target day, the 99.5th percentile is extracted as the high quantile and the 0.5th percentile as the low quantile for each time period, and half of the difference between the two is taken as the maximum fluctuation amplitude of photovoltaic output for that time period; the adaptive box uncertainty set The mathematical expression is: ; in, Let t be the actual photovoltaic power output variable in time period t. This represents the expected photovoltaic output for that period, calculated by multiplying the predicted irradiance from numerical weather forecasts by the photovoltaic module efficiency. The optimal robust budget parameters for the call. This represents the maximum fluctuation range of photovoltaic power output during that period. This is an auxiliary variable for uncertainty, and its value range is a closed interval. , This is the set of scheduling periods for a day.

5. The photovoltaic fluctuation adaptive method for evaluating the carrying capacity of electric vehicles in distribution networks according to claim 1, characterized in that, The two-stage robust optimization model is solved using a column and constraint generation algorithm, which iteratively approaches the global optimum through alternating main and subproblems. The subproblems utilize the strong duality theory of linear programming to transform the inner minimization problem into its dual maximization problem, and employ the Big M method to linearize the generated bilinear terms by introducing a sufficiently large positive number M, thus transforming the subproblems into mixed integer linear programming problems for solution.

6. The photovoltaic fluctuation adaptive method for assessing the carrying capacity of electric vehicles in distribution networks according to claim 1, characterized in that, The constraints of the two-stage robust optimization model include the following constraints that must be satisfied simultaneously: branch power flow constraints based on the DistFlow branch power flow model and using second-order cone relaxation treatment, the core mathematical form of which is: ; in, , These represent the active and reactive power flows of branch ij, respectively. Let be the current in branch ij. Let be the voltage at node i; Node voltage safety constraints: The node voltage amplitude is maintained at 1 / 3 of the nominal voltage. Within the safe zone; And the operational constraints of energy storage devices; the operational constraints of energy storage devices include upper limit constraints on charging power, upper limit constraints on discharging power, upper and lower limit constraints on energy storage capacity, time-series correlation constraints on state of charge, and mutual exclusion constraints on charge and discharge states; among which the time-series correlation constraints on state of charge are expressed as follows: ; in, The state of charge of the energy storage during time period t. , These refer to energy storage charging and discharging efficiency, respectively. , These represent the energy storage charging and discharging power during time period t. The time interval for scheduling; The mutual exclusion constraint of charge and discharge states is expressed as: ; in, , These are binary variables, representing the charging and discharging states respectively.

7. The photovoltaic fluctuation adaptive method for assessing the carrying capacity of electric vehicles in distribution networks according to claim 1, characterized in that, The method also includes an online safety margin verification step: inputting the carrying capacity assessment results into the distribution network digital twin simulation platform, generating random scenarios based on the Monte Carlo method to perform probabilistic power flow calculations, statistically analyzing the voltage over-limit probability and line overload probability, and if the voltage over-limit probability is less than 0.01%, then the safety margin is confirmed to be qualified, and a safety margin assessment report is generated.

8. A photovoltaic fluctuation adaptive electric vehicle load-bearing capacity assessment system for power distribution networks, characterized in that, include: The data management module is used to store and manage distribution network topology parameters, historical solar irradiance data, real-time weather forecast data, load data, and carrying capacity assessment results. An offline training engine, connected to the data management module, is used to read historical data from the data management module, perform time-series clustering of historical photovoltaic power output daily curves and search for optimal robust budget parameters, and construct and update the mapping relationship between fluctuation types and optimal robust budget parameters. The online evaluation engine, connected to both the data management module and the offline training engine, is used to receive evaluation requests and target day data. And further used for: Identify the volatility type of the target day; The optimal robust budget parameter corresponding to the fluctuation type of the target day is retrieved from the mapping relationship; Based on the invoked optimal robust budget parameters, the expected value sequence of photovoltaic output, and the maximum fluctuation amplitude sequence of photovoltaic output, an adaptive box-shaped uncertainty set is constructed; wherein, the robust budget parameters of the adaptive box-shaped uncertainty set are the invoked optimal robust budget parameters corresponding to the fluctuation type of the target day, and vary with the fluctuation type of the target day, rather than being a fixed value; the adaptive box-shaped uncertainty set is expressed as: ； In the formula: For the first Variables in actual photovoltaic output over time period For the first The expected photovoltaic output for a given period is calculated by multiplying the predicted irradiance value from numerical weather forecasts by the photovoltaic module efficiency. The optimal robust budget parameters for the target date's fluctuation type are used. For the first Maximum fluctuation range of photovoltaic power output during the period This is an auxiliary variable for uncertainty, and its value range is a closed interval. , A set of scheduling periods for a day; Based on the constructed adaptive box uncertainty set solution, a two-stage robust optimization model with the goal of maximizing electric vehicle access capacity is used to output the carrying capacity assessment results. The human-computer interaction and visualization module is connected to the online evaluation engine and is used to provide an evaluation parameter configuration interface and display the carrying capacity evaluation results, which include at least the access capacity of each node, security margin indicators, and evaluation reports.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the photovoltaic fluctuation adaptive method for evaluating the carrying capacity of electric vehicles in power distribution networks as described in 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 photovoltaic fluctuation adaptive method for evaluating the carrying capacity of electric vehicles in power distribution networks as described in any one of claims 1 to 7.

Images