Active power distribution network dynamic reconstruction optimization method for photovoltaic output fluctuation

Abstract

The invention discloses an active power distribution network dynamic reconstruction optimization method for photovoltaic output fluctuation, and the method comprises the steps: building a joint probability density model of different working conditions of photovoltaic output based on a Copula function according to the preprocessed power distribution network data, screening out photovoltaic output influence factors, and carrying out the typical fluctuation scene division; taking the safety and reliability of the power distribution network as constraint conditions, and constructing an active power distribution network dynamic reconstruction multi-objective optimization model containing photovoltaic access; solving the active power distribution network dynamic reconstruction multi-objective optimization model to obtain an optimal economy topology reconstruction result; and based on the rolling time domain, performing closed-loop optimization and real-time correction on the optimal economy topology reconstruction result, and outputting a final reconstruction optimization result. According to the method, the problems of voltage out-of-limit, branch overload and network loss rise of the active power distribution network, and insufficient adaptability and real-time performance of a traditional reconstruction scheme, which are caused by uncertainty fluctuation of photovoltaic output due to influence of meteorological factors, are solved.

Description

Technical Field

[0001] This invention belongs to the field of power grids and relates to distribution network operation optimization technology, specifically to an active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations. Background Technology

[0002] With the large-scale grid connection of distributed photovoltaic and wind power and other new energy sources, as well as the rapid growth of stochastic loads such as electric vehicles and flexible loads, the uncertainty of power output on the source side and demand on the load side of the distribution network has significantly increased, leading to frequent fluctuations in the operating status of the distribution network. Traditional static reconfiguration strategies are difficult to adapt to dynamically changing operating conditions. Most existing distribution network reconfiguration methods are based on deterministic models and do not fully consider the impact of source and load uncertainties on the reconfiguration results, which can easily lead to problems such as poor adaptability of reconfiguration schemes, high network loss rates, and increased risk of voltage exceeding limits. At the same time, some reconfiguration methods that consider uncertainties often use a single prediction model or a fixed optimization period, which cannot accurately capture the temporal fluctuation characteristics of source and loads, and have shortcomings such as insufficient optimization timeliness and high computational complexity, making it difficult to meet the actual needs of safe, economical, and stable operation of the distribution network.

[0003] Therefore, developing a power distribution network reconfiguration optimization strategy that can accurately quantify source-load uncertainty and dynamically adapt to operating conditions has significant engineering application value. Summary of the Invention

[0004] Purpose of the invention: In order to address the problem that photovoltaic power output is subject to uncertain fluctuations due to meteorological factors, which leads to voltage overruns, branch overloads, and increased network losses in active distribution networks, and the insufficient adaptability and real-time performance of traditional reconfiguration schemes, this invention provides a dynamic reconfiguration optimization method for active distribution networks that addresses photovoltaic power output fluctuations.

[0005] Technical Solution: To achieve the above objectives, this invention provides an active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations, comprising the following steps:

[0006] S1: Based on the preprocessed distribution network data, a joint probability density model of photovoltaic power output under different operating conditions is constructed using the Copula function to screen out the factors affecting photovoltaic power output;

[0007] S2: Based on the factors affecting photovoltaic power output, typical fluctuation scenarios are divided;

[0008] S3: Based on the typical fluctuation scenarios, and with the security and reliability of the distribution network as constraints, a multi-objective optimization model for dynamic reconfiguration of active distribution networks with photovoltaic access is constructed.

[0009] S4: Based on the improved adaptive particle swarm optimization algorithm, the multi-objective optimization model of dynamic reconfiguration of active distribution network is solved to obtain the optimal economic topology reconfiguration result that ensures the safety of distribution network.

[0010] S5: Based on the rolling time domain, perform closed-loop optimization and real-time correction on the optimal economic topology reconstruction result obtained in step S4, and output the final reconstruction optimization result.

[0011] Furthermore, the method for screening factors affecting photovoltaic output in step S1 includes:

[0012] A1: Constructing the reference sequence Comparison sequence matrix;

[0013] A2: Calculate the correlation coefficient based on the reference sequence and comparison sequence matrices:

[0014]

[0015] in, Let be the correlation coefficient between the k-th meteorological characteristic at time i and the photovoltaic output. The resolution coefficient, The minimum difference between the reference sequence and the comparison sequence at two levels. The maximum difference between the reference sequence and the comparison sequence at two levels;

[0016] A3: Based on the correlation coefficient Calculate the grey relational degree:

[0017]

[0018] in, Let the grey relational degree between the k-th type of meteorological characteristics and photovoltaic output be denoted as . ;

[0019] A4: Set a correlation threshold r0 to filter out r k Meteorological characteristics ≥r0 are considered key influencing factors.

[0020] Furthermore, the construction process of the joint probability density model in step S1 includes:

[0021] B1: Marginal distribution fitting:

[0022] The standardized photovoltaic power output sequence was analyzed separately. and 5 types of meteorological characteristic sequences , , , , Perform edge distribution fitting;

[0023] Considering the non-normal nature of the data, the kernel density estimation method is used to construct the marginal distribution functions of each variable, and the expression is:

[0024]

[0025] Where X is the variable to be fitted, n is the number of samples, and h is the window width. Let x be the point to be estimated, and let x be the i-th sample data point, where the Gaussian kernel function is used.

[0026] The marginal distribution functions F of each variable are obtained by kernel density estimation. P (p), F G (g), F T (t), F C (c) F V (v), F H (h) and the corresponding probability density function f P (p), f G (g), f T (t), f C (c), f V (v), f H (h);

[0027] B2: Copula function selection:

[0028] The t-Copula function is used to construct the joint distribution of photovoltaic power output and meteorological factors. Using photovoltaic power output P, ​​irradiance G, ambient temperature T, and cloud cover C as core variables, a four-dimensional t-Copula joint distribution function is constructed.

[0029]

[0030] Where u1=F P (p), u2=F G (g), u3=F T (t), u4=F C (c) represents the cumulative probability values ​​of the marginal distributions of each variable, θ is the correlation parameter matrix, and ν is the degree of freedom parameter. Let be the cumulative distribution function of the four-dimensional t-distribution. It is the inverse function of the univariate t-distribution;

[0031] B3: Solving for parameters θ and ν using the maximum likelihood estimation method:

[0032] Construct the log-likelihood function:

[0033]

[0034] in, This represents the cumulative marginal distribution probability of the first random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of photovoltaic active power output; This represents the cumulative marginal distribution probability of the second random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of irradiance. This represents the cumulative marginal distribution probability of the third random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of the ambient temperature. This represents the cumulative marginal distribution probability of the 4th random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of wind speed; The joint probability density function of the t-Copula function is expressed as:

[0035]

[0036] in, Let be the probability density function of a four-dimensional t-distribution. Let be the probability density function of a univariate t-distribution;

[0037] The optimal parameters are obtained by maximizing the log-likelihood function using the BFGS numerical optimization algorithm. and Finally, a joint probability model is constructed:

[0038] Furthermore, the process of classifying typical fluctuation scenarios in step S2 includes:

[0039] C1: Determine the number of clusters and initial parameters;

[0040] C2: Constructing the objective function:

[0041] The core of fuzzy C-means clustering is minimizing the objective function J, which is expressed as:

[0042]

[0043] in, Let x be the membership degree of the i-th sample to the j-th scene. i Let v be the key meteorological feature vector of the i-th sample. j Let be the cluster center vector of the j-th scenario. For sample x i With cluster center v j The Euclidean distance; c is the number of clusters; m is the fuzzy weighting index that controls the degree of fuzziness in the clustering results;

[0044] C3: Iteratively solve for cluster centers and membership degrees:

[0045]

[0046] C4: After the iteration converges, based on the maximum membership degree u of each sample ij,max Determine the scene category to which it belongs.

[0047] Furthermore, in step S3, the multi-objective optimization model for dynamic reconfiguration of the active distribution network combines the operational needs of the active distribution network and sets three major optimization objectives with economic optimization and safety controllability as the core: economic objective, voltage safety objective, and operational stability objective.

[0048] Furthermore, the three optimization objectives are expressed as follows:

[0049] Economic objective: Minimize network losses and switching operation costs.

[0050] The objective function is:

[0051]

[0052] Where T is the number of time periods within the optimization period; P ij,t Q ij,t U represents the active power and reactive power of branch ij at time t. i,t Let R be the voltage at node i at time t. ij Let C be the resistance of branch ij, Δt be the duration of the time interval, and C be the resistance of branch ij. s Cost per single switching operation; x ij,t Let x be the switching state of branch ij at time t. ij,t =1 indicates a closed loop, x ij,t =0 indicates disconnection; |x ij,t+1 -x ij,t | represents the change in switch state between adjacent time periods, with 1 when it changes and 0 when it remains unchanged; E is the set of branches in the distribution network;

[0053] Voltage safety objective: Minimize node voltage offset

[0054] The objective function is:

[0055]

[0056] Where N is the total number of distribution network nodes, U N The node's rated voltage; Let be the voltage offset rate of node i at time t; T is the total number of time periods within the optimization period.

[0057] Operational stability objective: Branch load rate equalization

[0058] The objective function is:

[0059]

[0060] Where |E| represents the total number of distribution network branches, and S ij,t Let S be the apparent power of branch ij at time t. ij,maxThe rated apparent power of branch ij, Let t be the average load rate of all branches in the network at time t.

[0061] Furthermore, the constraints in step S3 include:

[0062] Current constraints:

[0063]

[0064] Among them, P i,t Q i,t P represents the injected active and reactive power at node i at time t. PV,i,t Let t be the output of the photovoltaic system connected to node i at time t; Let i be the set of adjacent nodes of node i; Let be the load active power of node i at time t. Let be the reactive power of the load at node i at time t;

[0065] Node voltage constraints:

[0066]

[0067] Among them, U min and U max These are the lower and upper limits of the node voltage, respectively.

[0068] Branch capacity constraints:

[0069]

[0070] Among them, S ij,max This refers to the rated capacity of the branch circuit;

[0071] Topological constraints include switch state constraints, radial constraints, and switch operating frequency constraints.

[0072] Switch state constraint: x ij,t ∈{0,1}, meaning the switch only has two states: closed or open;

[0073] Radial constraint: The reconfigured distribution network must maintain a radial topology, with no loops or islands, meaning that all load nodes are connected to the power source through closed branches;

[0074] Switch operation frequency constraint: The number of times a single switch operates within one optimization cycle shall not exceed 3.

[0075] Photovoltaic output constraints:

[0076]

[0077] Among them, P PV,max,iThe rated output of the photovoltaic system connected to node i.

[0078] Furthermore, the process of solving the multi-objective optimization model for dynamic reconfiguration of the active distribution network based on the improved adaptive particle swarm optimization algorithm in step S4 includes:

[0079] D1: Sequentially execute the coding scheme design, chaotic initialization of the population, and selection of feasible solutions;

[0080] D2: Introducing dynamic inertia weights and adaptive learning factors;

[0081] D3: Particle velocity and position update rules: Combining the characteristics of binary encoding, design discretized velocity and position update formulas;

[0082] D4: Robustness Enhancement Strategy for Photovoltaic Fluctuations: To improve the algorithm's adaptability to photovoltaic power output fluctuations, a scenario robustness penalty factor and dynamic search step size adjustment are introduced;

[0083] D5: When the iteration termination condition is met, the optimal solution is selected.

[0084] Further, step S5 includes:

[0085] E1: Construct a rolling temporal optimization framework, including:

[0086] E1-1: Time-domain window division: Divide the total optimization period, such as 24 hours, into several continuous rolling windows, each with a duration of τ. τ is dynamically adjusted according to the photovoltaic fluctuation scenario.

[0087] E1-2: Data update within the window: Real-time acquisition of power distribution network operation data when each scrolling window starts;

[0088] E1-3: Dynamic Update of Optimization Objectives and Constraints: Based on the updated photovoltaic power output data and operating status, adjust the optimization objective weights and constraint thresholds for the current window;

[0089] E2: Perform rolling iterative optimization of the initial reconstruction scheme, specifically including:

[0090] E2-1: Solving the Model Within a Window: For each scrolling window, using the updated photovoltaic power output data and operating status data as input, the improved adaptive particle swarm optimization algorithm is called to resolve the multi-objective optimization model, obtaining the optimal reconstruction scheme within the current window. ;

[0091] E2-2: Scheme Smooth Transition Constraint: Introducing a scheme smoothing constraint to limit the amount of change in the switching states of adjacent windows:

[0092]

[0093] in, K represents the on / off state at the midpoint of the previous window. max The maximum number of switching changes allowed;

[0094] E2-3: Iterative update of the reconstruction scheme: After each scrolling window ends, output the optimal reconstruction scheme for the second half of the current window and execute it;

[0095] E3: Emergency corrections to the reconstruction plan based on real-time monitoring, including:

[0096] E3-1: Real-time monitoring of operating status: During the execution of each rolling window, key operating indicators are monitored in real time through the distribution network synchronous phasor measurement device and data acquisition and monitoring system;

[0097] E3-2: When the correction triggering conditions are met, the emergency correction mechanism is triggered;

[0098] E3-3: The fast gradient descent algorithm is used to implement local adjustments to the reconstruction scheme.

[0099] Furthermore, the local adjustment process of the reconstruction scheme in step E3-3 includes:

[0100] E3-3-1: Locating the root cause of the problem: For voltage over-limit nodes, identify the switch status of adjacent branches and determine the key adjustable switches; for branch overload, screen the parallel or series branch switches of the overloaded branch.

[0101] E3-3-2: Local optimization objective: With the goal of eliminating constraint violations as quickly as possible, the local optimization objective is set as minviolation(X), where violation(X) is the degree of constraint violation;

[0102] E3-3-3: Local adjustment of switch states: Only the state of key switches is optimized, and the switch state combination that minimizes the degree of constraint violation is found by using the gradient descent method;

[0103] E3-3-4: Scheme Verification: After modification, verify whether the distribution network topology meets the radial constraint and the island constraint. Execute the modified scheme immediately after ensuring its feasibility.

[0104] Beneficial Effects: Compared with existing technologies, this invention effectively solves the problems of voltage exceeding limits, branch overload, and increased network losses in the active distribution network caused by the uncertain fluctuations in photovoltaic output due to meteorological factors. This is achieved by constructing a complete technical system encompassing precise quantification of photovoltaic output fluctuations, dynamic modeling across multiple scenarios, efficient algorithmic solution, and rolling time-domain closed-loop optimization. Simultaneously, it overcomes the technical deficiencies of traditional reconfiguration schemes in terms of adaptability and real-time performance. First, key meteorological influencing factors on photovoltaic output are screened using grey relational analysis. Then, a joint probability density model is constructed using the t-Copula function. Finally, fuzzy C... Mean clustering is used to classify typical photovoltaic (PV) output fluctuation scenarios, accurately capturing the uncertainty characteristics of PV output under all operating conditions, and improving the adaptability of reconfiguration schemes to PV output fluctuations from the source. A multi-objective optimization model for active distribution network dynamic reconfiguration with economic, voltage safety, and operational stability objectives is specifically constructed. Differentiated constraint relaxation and objective weights are configured for different PV fluctuation scenarios to achieve a dynamic balance between safe and economical operation of the distribution network, avoiding voltage overrun and branch overload problems at the model level. The particle swarm optimization algorithm is improved by chaotic initialization, dynamic inertia weight, adaptive learning factor, and scenario robustness penalty factor to improve the algorithm's solution efficiency, convergence accuracy, and robustness to PV fluctuations, ensuring that the optimal reconfiguration scheme that satisfies all constraints is obtained quickly. A closed-loop optimization framework is constructed based on rolling time domain, dynamically adjusting the optimization window duration, updating grid operation data and PV output information in real time, and iteratively optimizing the reconfiguration scheme. An emergency correction mechanism is added, and the reconfiguration scheme is locally adjusted in real time through a fast gradient descent algorithm, effectively improving the real-time performance and dynamic adaptability of the reconfiguration scheme. After implementation, this invention can accurately adapt to the fluctuation characteristics of photovoltaic power output under all operating conditions, effectively reduce distribution network losses, avoid voltage overrun and branch overload problems, balance branch load rate, reduce the frequency of switch operation, improve the economic operation level of the active distribution network while ensuring its safe and stable operation, significantly enhance the distribution network's ability to resist disturbances caused by photovoltaic power output fluctuations, and the reconfiguration scheme meets the technical requirements of actual engineering operation of the active distribution network. Attached Figure Description

[0105] Figure 1 This is a simplified flowchart of the method of the present invention;

[0106] Figure 2 This is a flowchart illustrating the steps of the method of the present invention;

[0107] Figure 3 To improve the architecture of the IEEE 33-node system. Detailed Implementation

[0108] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0109] Example 1:

[0110] like Figure 1 and Figure 2 As shown in the figure, this embodiment provides an active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations, including the following steps:

[0111] S1: Based on the preprocessed distribution network data, a joint probability density model of photovoltaic power output under different operating conditions is constructed using the Copula function to screen out the factors affecting photovoltaic power output;

[0112] Distribution network data includes meteorological characteristics and photovoltaic power output data, including:

[0113] Historical data is collected for the area where the active distribution network is located, including time-series output data of photovoltaic power plants and corresponding meteorological monitoring data. Among these, photovoltaic output data... The unit is KW, the time resolution is 15 minutes, and n is the total duration of historical data calculated at 15-minute intervals; the meteorological data covers irradiance. The unit is W / m 2 Ambient temperature Units are in °C; cloud cover Wind speed The unit is m / s; relative humidity. The unit is %; there are five categories of potential influencing factors.

[0114] The raw data underwent preprocessing: Outliers exceeding the mean ± 3 standard deviations in photovoltaic output and meteorological data were removed using the 3σ criterion; missing data were imputed using linear interpolation to ensure data continuity; all data were standardized to eliminate dimensional differences. The standardization formula is as follows:

[0115]

[0116] in: The original data, For the corresponding data sequence, The data is standardized and its value range is [0,1].

[0117] Grey relational analysis can quantify the degree of correlation between two systems or variables, and is suitable for analyzing uncertain systems with limited sample size and incomplete information. Using standardized photovoltaic power output sequences... As a reference sequence, the standardized five types of meteorological characteristic sequences G ∗ T ∗ C ∗ V ∗ H ∗ To compare the sequences, the grey relational degree between each meteorological feature and photovoltaic output is calculated.

[0118] Methods for screening factors affecting photovoltaic power output include:

[0119] A1: Constructing the reference sequence Comparison sequence k=1,2,3,4,5, corresponding to irradiance, temperature, cloud cover, wind speed, and humidity, respectively;

[0120] A2: Calculate the correlation coefficient based on the reference sequence and comparison sequence matrices:

[0121]

[0122] in, Let be the correlation coefficient between the k-th meteorological characteristic at time i and the photovoltaic output. The resolution coefficient, The minimum difference between the reference sequence and the comparison sequence at two levels. The maximum difference between the reference sequence and the comparison sequence at two levels;

[0123] A3: Based on the correlation coefficient Calculate the grey relational degree:

[0124]

[0125] in, Let the grey relational degree between the k-th type of meteorological characteristics and photovoltaic output be denoted as . The closer the value is to 1, the more significant the impact of the meteorological characteristic on photovoltaic output.

[0126] A4: Set a correlation threshold r0 to filter out r k Meteorological characteristics ≥r0 serve as key influencing factors, providing core input variables for subsequent scene segmentation.

[0127] The process of constructing the joint probability density model includes:

[0128] B1: Marginal distribution fitting:

[0129] The standardized photovoltaic power output sequence was analyzed separately. and 5 types of meteorological characteristic sequences , , , , Perform edge distribution fitting;

[0130] Considering the non-normal nature of the data, the kernel density estimation (KDE) method is used to construct the marginal distribution functions of each variable, and the expression is:

[0131]

[0132] Where X is the variable to be fitted, n is the number of samples, and h is the window width. Let x be the point to be estimated, and let x be the i-th sample data point, where the Gaussian kernel function is used.

[0133] The marginal distribution functions F of each variable are obtained by kernel density estimation. P (p), F G (g), F T (t), F C (c) F V (v), F H (h) and the corresponding probability density function f P (p), f G (g), f T (t), f C (c), f V (v), f H (h);

[0134] B2: Copula function selection:

[0135] The t-Copula function is chosen to construct the joint distribution of photovoltaic (PV) output and meteorological factors. Its advantage lies in its ability to accurately capture the tail correlations between variables, making it suitable for scenarios with significant fluctuations in PV output under extreme weather conditions. A four-dimensional t-Copula joint distribution function is constructed using PV output P, ​​irradiance G, ambient temperature T, and cloud cover C as core variables.

[0136]

[0137] Where u1=F P (p), u2=F G (g), u3=F T (t), u4=F C (c) represents the cumulative probability values ​​of the marginal distributions of each variable, θ is the correlation parameter matrix, and ν is the degree of freedom parameter. Let be the cumulative distribution function of the four-dimensional t-distribution. It is the inverse function of the univariate t-distribution;

[0138] B3: Solving for parameters θ and ν using the maximum likelihood estimation method:

[0139] Construct the log-likelihood function:

[0140]

[0141] in, This represents the cumulative marginal distribution probability of the first random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of photovoltaic active power output; This represents the cumulative marginal distribution probability of the second random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of irradiance. This represents the cumulative marginal distribution probability of the third random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of the ambient temperature. This represents the cumulative marginal distribution probability of the 4th random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of wind speed; The joint probability density function of the t-Copula function is expressed as:

[0142]

[0143] in, Let be the probability density function of a four-dimensional t-distribution. Let be the probability density function of a univariate t-distribution;

[0144] The optimal parameters are obtained by maximizing the log-likelihood function using the BFGS numerical optimization algorithm. and Finally, a joint probability model is constructed:

[0145] ; S2: Based on the factors affecting photovoltaic power output, typical fluctuation scenarios are divided;

[0146] In this embodiment, typical fluctuation scenarios are divided based on fuzzy C-means clustering. Standardized data of key meteorological features selected from the data are used as clustering input. Combined with the corresponding photovoltaic power output fluctuation coefficients, historical operating days are divided into four typical fluctuation scenarios. The process includes:

[0147] C1: Determine the number of clusters and initial parameters;

[0148] Based on the characteristics of photovoltaic power output fluctuations and actual engineering needs, the number of clusters is set to c=4, corresponding to four scenarios: weak fluctuation, medium fluctuation, strong fluctuation, and extreme fluctuation. A fuzzy factor is set, taking the commonly used engineering value m=2. The iteration termination condition is set: the change in cluster centers between two consecutive iterations is less than ε=10. -5 Or the number of iterations reaches the maximum value N. max =100.

[0149] C2: Constructing the objective function:

[0150] The core of fuzzy C-means clustering is minimizing the objective function J, which is expressed as:

[0151]

[0152] in, Let x be the membership degree of the i-th sample to the j-th scene. i Let v be the key meteorological feature vector of the i-th sample. j Let be the cluster center vector of the j-th scenario. For sample x i With cluster center v j The Euclidean distance; c is the number of clusters, corresponding to four types of photovoltaic power output fluctuation scenarios: weak, medium, strong, and extreme; m is the fuzzy weighting index, which controls the fuzziness of the clustering results and is a key hyperparameter of FCM. The larger m is, the more fuzzy the clustering results are.

[0153] C3: Iteratively solve for cluster centers and membership degrees:

[0154]

[0155] C4: After the iteration converges, based on the maximum membership degree u of each sample ij,max Determine its corresponding scenario category; in this embodiment, based on the average meteorological characteristics of each scenario and the photovoltaic power output fluctuation coefficient, the four scenarios are named as follows:

[0156] Scenario 1: Slight fluctuations: Stable irradiance, suitable temperature, low cloud cover, photovoltaic power output fluctuation coefficient λ≤0.1;

[0157] In scenario 2, the irradiance is moderate, the temperature fluctuation is small, and the cloud cover is moderate. The photovoltaic output fluctuation coefficient λ∈[0.1-0.3];

[0158] Scenario 3: Strong fluctuations: Unstable irradiance, large temperature variations, and abundant cloud cover; photovoltaic output fluctuation coefficient λ∈[0.3-0.5];

[0159] Scenario 4: Extreme fluctuations: low and sudden irradiance, temperature deviating from the suitable range, heavy cloud cover or rain / snow, photovoltaic output fluctuation coefficient λ≥0.5.

[0160] Through the above steps, the key influencing factors of photovoltaic power output are screened and typical fluctuation scenarios are divided, providing solid data support and scenario foundation for the subsequent construction and targeted optimization of the distribution network dynamic reconfiguration model.

[0161] S3: Based on the typical fluctuation scenarios, and with the security and reliability of the distribution network as constraints, a multi-objective optimization model for dynamic reconfiguration of active distribution networks with photovoltaic access is constructed.

[0162] The multi-objective optimization model for dynamic reconfiguration of active distribution networks combines the operational needs of active distribution networks with the core objectives of economic optimization and safety controllability. It sets three major optimization objectives: economic objective, voltage safety objective, and operational stability objective.

[0163] The three optimization objectives are expressed as follows:

[0164] Economic objective: Minimize network losses and switching operation costs.

[0165] The core economic benefit of distribution network reconfiguration stems from reduced network losses, while simultaneously avoiding equipment wear and maintenance costs caused by frequent switching operations. The objective function is:

[0166]

[0167] Where T is the number of time periods within the optimization cycle, divided into 15-minute intervals, with a 24-hour optimization cycle corresponding to T=96; P ij,t Q ij,t U represents the active power and reactive power of branch ij at time t. i,t Let R be the voltage at node i at time t. ij Let C be the resistance of branch ij, and Δt be the duration of the time interval; s The cost per switch operation is expressed in yuan per operation; in this embodiment, a commonly used engineering value of 500 yuan is used. ij,t Let x be the switching state of branch ij at time t. ij,t =1 indicates a closed loop, x ij,t =0 indicates disconnection; |x ij,t+1 -x ij,t | represents the change in switch state between adjacent time periods, with 1 when it changes and 0 when it remains unchanged; E is the set of branches in the distribution network;

[0168] Voltage safety objective: Minimize node voltage offset

[0169] Fluctuations in photovoltaic output can easily cause node voltage to deviate from the rated value, or even exceed the limit. It is necessary to control the voltage deviation within the allowable range. The objective function is:

[0170]

[0171] Where N is the total number of distribution network nodes; U N U is the node's rated voltage; in this embodiment, it is taken as U. N =10kV; f2 represents the voltage offset rate of node i at time t. The smaller f2 is, the closer the voltage of the entire network is to the rated value and the better the voltage quality. T is the total number of time periods within the optimization period.

[0172] Operational stability objective: Branch load rate equalization

[0173] To avoid the unbalanced operation of the distribution network caused by photovoltaic fluctuations, resulting in some branches being overloaded and others being lightly loaded, and to improve the network's immunity to disturbances, the objective function is:

[0174]

[0175] Where |E| represents the total number of distribution network branches, and S ij,t Let S be the apparent power of branch ij at time t. ij,max The rated apparent power of branch ij, Let f3 be the average load rate of all branches in the network at time t. The smaller f3 is, the more balanced the distribution of branch load rates and the stronger the operational stability.

[0176] The constraints include:

[0177] Current constraints:

[0178]

[0179] Among them, P i,t Q i,t P represents the injected active and reactive power at node i at time t. PV,i,t Let t be the output of the photovoltaic system connected to node i at time t; Let i be the set of adjacent nodes of node i; Let be the load active power of node i at time t. Let be the reactive power of the load at node i at time t;

[0180] Node voltage constraints:

[0181]

[0182] Among them, U min and U max These are the lower and upper limits of the node voltage, respectively, to ensure that the voltage does not exceed the limits; in this embodiment... U min =0.95U N U max =1.05U N ;

[0183] Branch capacity constraints:

[0184]

[0185] Among them, S ij,max The rated capacity of the branch circuit should be set to prevent overload of the branch circuit from causing equipment damage or power outage;

[0186] Topological constraints include switch state constraints, radial constraints, and switch operating frequency constraints.

[0187] Switch state constraint: xij,t ∈{0,1}, meaning the switch only has two states: closed or open;

[0188] Radial constraint: The reconfigured distribution network must maintain a radial topology, with no loops or islands, meaning that all load nodes are connected to the power source through closed branches;

[0189] Switch operation frequency constraint: that is, the number of times a single switch is operated within one optimization cycle shall not exceed 3 times, so as to extend the service life of the equipment;

[0190] Photovoltaic output constraints:

[0191]

[0192] Among them, P PV,max,i The rated output of the photovoltaic system connected to node i is set to ensure that the photovoltaic output meets the physical limitations of the equipment.

[0193] For the four types of photovoltaic fluctuation scenarios, a scenario adaptation coefficient α is introduced. j j=1,2,3,4 correspond to weak, medium, strong, and extreme fluctuation scenarios, respectively, and the constraint slackness and target weight ratio are dynamically adjusted:

[0194] In a weak fluctuation scenario with α1=0.8, the voltage deviation rate constraint can be appropriately relaxed to ±6%, and the weight of the economic target is increased to 0.5.

[0195] In the medium-fluctuation scenario, α2=1.0: maintain the basic constraints and weight configuration;

[0196] In a scenario with strong fluctuations, α3=1.2: the voltage offset rate constraint is tightened to ±4%, and the weight of the operational stability target is increased to 0.25;

[0197] In extreme fluctuation scenarios, α4=1.5: the upper limit of branch load rate is reduced to 80%, the weight of voltage stability target is increased to 0.4, and priority is given to ensuring the safe operation of the distribution network.

[0198] By constructing the above model, targeted optimizations can be achieved for different photovoltaic fluctuation scenarios. This ensures the economic efficiency of the distribution network when photovoltaic output is stable, while strengthening safety constraints when fluctuations intensify, providing clear and feasible mathematical model support for subsequent algorithm solutions.

[0199] S4: Based on the improved adaptive particle swarm optimization algorithm, the multi-objective optimization model of dynamic reconfiguration of active distribution network is solved to obtain the optimal economic topology reconfiguration result that ensures the safety of distribution network.

[0200] The process of solving the multi-objective optimization model for dynamic reconfiguration of active distribution networks based on the improved adaptive particle swarm optimization algorithm includes:

[0201] D1: Algorithm Encoding and Population Initialization: The encoding method design, chaotic population initialization, and feasible solution selection are performed sequentially.

[0202] Encoding Design: A binary encoding scheme is adopted, with each particle corresponding to a set of distribution network switch states. The particle dimension is equal to the total number of distribution network branches |E|, and the k-th element of the particle is S. k ∈{0,1}, respectively, correspond to the switch of branch k being either open (0) or closed (1).

[0203] Chaotic Initial Population: To improve the diversity of the initial population and prevent premature convergence of the algorithm, a Logistic chaotic mapping is used to generate the initial population. The formula for the Logistic chaotic mapping is:

[0204]

[0205] Where μ is a control parameter (in this embodiment, μ=4 to ensure the mapping is in a completely chaotic state), z m ∈[0,1] represents the chaotic sequence value. The chaotic sequence value is compared with a threshold of 0.5, z m When ≥0.5, it is encoded as 1, z m When the value is less than 0.5, it is encoded as 0, and the initial particle swarm is generated.

[0206] Feasible solution selection:

[0207] After initialization, the feasibility of the topology structure corresponding to each particle is checked. Invalid particles with loops or isolated islands are removed. Invalid particles are replaced by randomly adjusting the switch states to ensure that all particles in the initial population satisfy the topological constraints. The population size is set to N. p =60.

[0208] D2: Adaptive parameter optimization design: To balance the global search and local optimization capabilities of the algorithm, dynamic inertia weights and adaptive learning factors are introduced;

[0209] Dynamic inertia weight w: The inertia weight determines the degree to which particles inherit historical velocities. It adopts a piecewise adaptive adjustment strategy and changes dynamically with the number of iterations.

[0210]

[0211] Among them, the global search w is enhanced in the early stage of iteration. max =0.9, intermediate transition w mid =0.6, local optimization is strengthened in the later stage of iteration. min =0.3, where iter is the current iteration number, and the segment node iter1 = 0.5 × iter max Maximum number of iterations iter max =200.

[0212] The adaptive learning factors include c1 and c2. c1 is the individual cognitive factor, which guides the particle to move closer to its own optimal position, and c2 is the social cognitive factor, which guides the particle to move closer to the global optimal position. A linear reverse adjustment strategy is adopted.

[0213]

[0214] Among them, c 1,max =2.0、c 1,min =0.5, c 2,max =2.0、c 2,min =0.5, ensuring that the initial iteration focuses on individual exploration, while the later iteration focuses on global convergence.

[0215] D3: Particle Velocity and Position Update Rules: Combining the characteristics of binary encoding, discretized velocity and position update formulas are designed, including:

[0216] Speed ​​updates:

[0217]

[0218] Among them, v i,d (iter) represents the velocity of particle i in the d-th dimension at the iter-th iteration, s i,d (iter) represents the position (encoded value) of particle i in the d-th dimension, pbest i,d Let gbest be the d-th dimension of the individual optimal position of particle i. d The d-th dimension represents the global optimal position, and r1 and r2 are random numbers in the interval [0,1] to increase the randomness of the search.

[0219] Position Update: The probability of converting continuous velocity into discrete position using the Sigmoid function:

[0220]

[0221] Here, rand(0,1) is a uniform random number in the interval [0,1], and the Sigmoid function ensures the probabilistic nature of position updates, balancing search diversity and convergence.

[0222] D4: Robustness Enhancement Strategy for Photovoltaic Power Output Fluctuations: To improve the algorithm's adaptability to photovoltaic power output fluctuations, a scenario robustness penalty factor and dynamic search step size adjustment are introduced, specifically including:

[0223] Robustness penalty factor β: Different penalties are imposed on particles that violate the constraints for different photovoltaic fluctuation scenarios. The penalty factor is positively correlated with the scenario fluctuation coefficient.

[0224]

[0225] Where k is the penalty coefficient, and in this embodiment k=0.8; λ j Let be the photovoltaic output fluctuation coefficient for scenario j. For particles that violate voltage constraints and branch capacity constraints, their fitness values ​​are corrected as follows:

[0226]

[0227] Here, violation represents the degree of constraint violation, which is the sum of the voltage over-limit difference and the branch overload ratio. The algorithm converges to the feasible solution region through a penalty-forced algorithm.

[0228] Dynamic search step size adjustment: based on the current population fitness variance σ fit Adjust the search step size, σ fit A larger σ indicates better population diversity, and a larger step size is used to enhance the global search; fit The smaller the value, the more convergent the population tends to be; therefore, a finer search with a small step size is employed.

[0229]

[0230] Among them, step max To maximize the search step size, in this embodiment, step max =0.5; σ fit,max This represents the fitness variance of the initial population.

[0231] D5: When the iteration termination condition is met, the optimal solution is selected.

[0232] Iteration termination condition: When the number of iterations reaches iter max =200, or the change in the global optimal fitness value over 10 consecutive iterations is less than 10. -6 The algorithm terminates when the time is right.

[0233] Optimal solution selection: After the iteration terminates, the switching state combination corresponding to the global optimal particle is verified to satisfy all constraints. The corresponding network loss, voltage deviation rate, load factor balance and other indicators are calculated to confirm that the combination is the optimal reconfiguration scheme (i.e. the optimal switching state combination) under the current photovoltaic fluctuation scenario.

[0234] S5: Based on the rolling time domain, perform closed-loop optimization and real-time correction on the optimal economic topology reconstruction result obtained in step S4, and output the final reconstruction optimization result.

[0235] Step S5 includes:

[0236] E1: Construct a rolling temporal optimization framework, including:

[0237] E1-1: Time-domain window division: Divide the total optimization period, such as 24 hours, into several continuous rolling windows, each with a duration of τ. τ is dynamically adjusted according to the photovoltaic fluctuation scenario.

[0238] For weak fluctuation scenarios, τ=4 hours; for medium fluctuation scenarios, τ=2 hours; for strong fluctuation scenarios, τ=1 hour; and for extreme fluctuation scenarios, τ=0.5 hours. The window overlap rate is set to 50% to ensure the continuity and smoothness of the optimization.

[0239] E1-2: Data Update within the Window: Upon startup of each scrolling window, real-time data collection is performed on the distribution network operation, including actual photovoltaic output data. Real-time operating status data such as node voltage and branch power are used to update the photovoltaic output forecast for the next τ hours based on current meteorological monitoring data. ;

[0240] E1-3: Dynamic Update of Optimization Objectives and Constraints: Based on the updated photovoltaic power output data and operating status, adjust the optimization objective weights and constraint thresholds for the current window;

[0241] In this embodiment, if the actual photovoltaic output deviates from the predicted value... If 10%≤δ<20%, the voltage safety target weight is increased by 10%, and the voltage deviation constraint is tightened to ±4.5%; if δ≥20%, the operational stability target weight is increased by 15%, and the branch load rate limit is reduced to 85%, prioritizing safe operation.

[0242] E2: Perform rolling iterative optimization of the initial reconstruction scheme, specifically including:

[0243] E2-1: Solving the Model Within a Window: For each scrolling window, using the updated photovoltaic power output data and operating status data as input, the improved adaptive particle swarm optimization algorithm is called to resolve the multi-objective optimization model, obtaining the optimal reconstruction scheme within the current window. ;

[0244] E2-2: Scheme Smooth Transition Constraint: To avoid frequent switching operations caused by excessive differences in reconstruction schemes between adjacent windows, a scheme smoothing constraint is introduced to limit the amount of change in the switching state of adjacent windows.

[0245]

[0246] in, K represents the on / off state at the midpoint of the previous window. max The maximum number of allowed switching changes (based on scenario settings: K for weak fluctuation scenarios) max =2, Medium Fluctuation Scenario K max =3. Strong fluctuation scenario K max=4. Extreme fluctuation scenario K max =5), ensuring the economy of switch operation and equipment safety;

[0247] E2-3: Iterative update of reconstruction scheme: After each rolling window ends, the optimal reconstruction scheme for the second half of the current window is output and executed; at the same time, the scheme is used as the initial solution for the next rolling window, reducing the number of algorithm iterations and improving optimization efficiency;

[0248] E3: Emergency corrections to the reconstruction plan based on real-time monitoring, including:

[0249] E3-1: Real-time Operational Status Monitoring: During the execution of each rolling window, key operational indicators, including node voltage, are monitored in real time through the distribution network synchronous phasor measurement device and data acquisition and monitoring system. Determine if limits are exceeded, branch power Determine if there is an overload and the rate of fluctuation in photovoltaic output. Determine the degree of volatility.

[0250] E3-2: When the correction triggering conditions are met, the emergency correction mechanism is triggered;

[0251] The modified trigger condition includes any one of the following:

[0252] Voltage over-limit trigger: Node voltage exists. or And the duration exceeds 30 seconds;

[0253] Branch overload trigger: Apparent power exists in the branch. And the duration exceeds 15 seconds;

[0254] Severe fluctuations trigger: photovoltaic power output fluctuation rate .

[0255] E3-3: Using the fast gradient descent algorithm to implement local adjustments to the reconstruction scheme:

[0256] E3-3-1: Locating the root cause of the problem: For voltage over-limit nodes, identify the switch status of adjacent branches and determine the key adjustable switches; for branch overload, screen the parallel or series branch switches of the overloaded branch.

[0257] E3-3-2: Local optimization objective: With the goal of eliminating constraint violations as quickly as possible, the local optimization objective is set as minviolation(X), where violation(X) is the degree of constraint violation;

[0258] E3-3-3: Local adjustment of switch states: Only the state of key switches is optimized (to avoid delays caused by global search). The gradient descent method is used to find the switch state combination that minimizes the degree of constraint violation, and the adjustment time is controlled within 10 seconds.

[0259] E3-3-4: Scheme Verification: After modification, verify whether the distribution network topology meets the radial constraint and the island constraint. Execute the modified scheme immediately after ensuring its feasibility.

[0260] This embodiment evaluates and provides feedback on the effectiveness of the reconstruction scheme:

[0261] Operational performance evaluation: After each rolling window ends, the core operational performance indicators within the current window are calculated, including loss reduction rate, voltage qualification rate, branch load balance, and number of switching operations, and an evaluation report is generated.

[0262] Algorithm parameter feedback optimization: Adjusting and improving the parameters of the adaptive particle swarm optimization algorithm based on evaluation results.

[0263] If the voltage qualification rate is lower than 98%, increase the coefficient k of the robustness penalty factor β to 1.0; if the network loss reduction rate is lower than expected, increase the weight of the economic target by 5%; if the number of switching operations is too high, increase the value of the switching operation cost Cs.

[0264] Based on the above, the method of the present invention is summarized in this embodiment as follows:

[0265] First, a joint probabilistic model of photovoltaic (PV) output, irradiance, ambient temperature, and cloud cover is constructed based on the Copula function to accurately quantify the uncertainty and fluctuation characteristics of PV output. Fuzzy C-means clustering is then used to select a set of typical operating conditions covering extreme fluctuation scenarios. Second, a dynamic reconfiguration mathematical model of an active distribution network with PV access is established. The optimization objectives are minimizing network losses, node voltage deviations, and switching operation costs. Power flow constraints, node voltage upper and lower limits, branch capacity constraints, and radial topology constraints are embedded to achieve multi-objective collaborative optimization. Third, an improved adaptive particle swarm optimization algorithm is designed. Chaotic initialization enhances population diversity, and dynamic inertia weights and adaptive learning factors are introduced to balance global search and local optimization capabilities. A robust penalty factor for PV fluctuation scenarios is combined to enhance the algorithm's adaptability to uncertain operating conditions. Finally, based on a rolling time-domain optimization strategy, the entire time period is divided into several continuous optimization windows. PV output prediction data and grid operating status are updated in real time, and the optimal switching combination scheme for each time period is dynamically output to ensure that the active distribution network always meets power flow constraints and node voltage limits under PV output fluctuation conditions.

[0266] Example 2:

[0267] To verify the effectiveness and efficacy of the method of the present invention, the following experiments and data analyses were conducted in this embodiment:

[0268] This embodiment uses the improved IEEE 33-node system as the object of reconstruction and optimization, such as... Figure 3 As shown, the node numbers of the improved IEEE 33-node system are 1 to 33, and the initial state of the improved IEEE 33-node system is shown in Tables 1 to 3.

[0269] Table 1 Initial State of the Improved IEEE 33-Node System

[0270]

[0271] Table 2 Initial State of Branch

[0272]

[0273] Table 3 Connection Branch Status

[0274]

[0275] In this embodiment, the results after reconstruction are compared with those before reconstruction, as shown in Table 4.

[0276] Table 4 Comparison of results before and after reconstruction

[0277]

[0278] As shown in Table 4, dynamic reconfiguration of the distribution network significantly improves operational performance by optimizing the topology. The switchgear combinations before and after reconfiguration are significantly adjusted, from the initial 9-15, 12-22, 18-33, 21-8, 25-29 to 7-8, 9-10, 14-15, 32-33, 25-29. By changing the power transmission path and power flow distribution, the network loss is effectively reduced, decreasing from 202.68kW to 139.55kW, a reduction of approximately 31.15%. This significantly improves power transmission efficiency and reduces energy loss. The minimum node voltage increases from 0.910pu to 0.937pu, and the voltage deviation is significantly reduced, approaching the rated value of 1.0pu. This effectively alleviates voltage dips, improves power quality and system stability, and fully verifies the effectiveness and feasibility of the proposed dynamic reconfiguration method for distribution networks in terms of loss reduction, efficiency improvement, and voltage support.

Claims

1. A method for dynamic reconfiguration optimization of active distribution networks to address photovoltaic power output fluctuations, characterized in that, Includes the following steps: S1: Based on the preprocessed distribution network data, a joint probability density model of photovoltaic power output under different operating conditions is constructed using the Copula function to screen out the factors affecting photovoltaic power output; S2: Based on the factors affecting photovoltaic power output, typical fluctuation scenarios are divided; S3: Based on the typical fluctuation scenarios, and with the security and reliability of the distribution network as constraints, a multi-objective optimization model for dynamic reconfiguration of active distribution networks with photovoltaic access is constructed. S4: Based on the improved adaptive particle swarm optimization algorithm, the multi-objective optimization model of dynamic reconfiguration of active distribution network is solved to obtain the optimal economic topology reconfiguration result that ensures the safety of distribution network. S5: Based on the rolling time domain, perform closed-loop optimization and real-time correction on the optimal economic topology reconstruction result obtained in step S4, and output the final reconstruction optimization result.

2. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 1, characterized in that, The screening method for factors affecting photovoltaic output in step S1 includes: A1: Constructing the reference sequence Comparison sequence matrix; A2: Calculate the correlation coefficient based on the reference sequence and comparison sequence matrices: ； in, Let be the correlation coefficient between the k-th meteorological characteristic at time i and the photovoltaic output. The resolution coefficient, The minimum difference between the reference sequence and the comparison sequence at two levels. The maximum difference between the reference sequence and the comparison sequence at two levels; A3: Based on the correlation coefficient Calculate the grey relational degree: ； in, Let the grey relational degree between the k-th type of meteorological characteristics and photovoltaic output be denoted as . ; A4: Set a correlation threshold r0 to filter out r k Meteorological characteristics ≥r0 are considered key influencing factors.

3. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 2, characterized in that, The construction process of the joint probability density model in step S1 includes: B1: Marginal distribution fitting: The standardized photovoltaic power output sequence was analyzed separately. and 5 types of meteorological characteristic sequences , , , , Perform edge distribution fitting; Considering the non-normal nature of the data, the kernel density estimation method is used to construct the marginal distribution functions of each variable, and the expression is: ； Where X is the variable to be fitted, n is the number of samples, and h is the window width. Let x be the point to be estimated, and let x be the i-th sample data point, where the Gaussian kernel function is used. The marginal distribution functions F of each variable are obtained by kernel density estimation. P (p), F G (g), F T (t), F C (c) F V (v), F H (h) and the corresponding probability density function f P (p), f G (g), f T (t), f C (c), f V (v), f H (h); B2: Copula function selection: The t-Copula function is used to construct the joint distribution of photovoltaic power output and meteorological factors. Using photovoltaic power output P, ​​irradiance G, ambient temperature T, and cloud cover C as core variables, a four-dimensional t-Copula joint distribution function is constructed. ； Where u1=F P (p), u2=F G (g), u3=F T (t), u4=F C (c) represents the cumulative probability values ​​of the marginal distributions of each variable, θ is the correlation parameter matrix, and ν is the degree of freedom parameter. Let be the cumulative distribution function of the four-dimensional t-distribution. It is the inverse function of the univariate t-distribution; B3: Solving for parameters θ and ν using the maximum likelihood estimation method: Construct the log-likelihood function: ； in, This represents the cumulative marginal distribution probability of the first random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of photovoltaic active power output; This represents the cumulative marginal distribution probability of the second random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of irradiance. This represents the cumulative marginal distribution probability of the third random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of the ambient temperature. This represents the cumulative marginal distribution probability of the 4th random variable in the i-th sample, corresponding to the cumulative marginal distribution probability of wind speed; The joint probability density function of the t-Copula function is expressed as: ； in, Let be the probability density function of a four-dimensional t-distribution. Let be the probability density function of a univariate t-distribution; The optimal parameters are obtained by maximizing the log-likelihood function using the BFGS numerical optimization algorithm. and Finally, a joint probability model is constructed: 。 4. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 3, characterized in that, The process of classifying typical fluctuation scenarios in step S2 includes: C1: Determine the number of clusters and initial parameters; C2: Constructing the objective function: The core of fuzzy C-means clustering is minimizing the objective function J, which is expressed as: ； in, Let x be the membership degree of the i-th sample to the j-th scene. i Let v be the key meteorological feature vector of the i-th sample. j Let be the cluster center vector of the j-th scenario. For sample x i With cluster center v j The Euclidean distance; c is the number of clusters; m is the fuzzy weighting index that controls the degree of fuzziness in the clustering results; C3: Iteratively solve for cluster centers and membership degrees: ； C4: After the iteration converges, based on the maximum membership degree u of each sample ij,max Determine the scene category to which it belongs.

5. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 4, characterized in that, In step S3, the multi-objective optimization model for dynamic reconfiguration of the active distribution network combines the operational needs of the active distribution network and sets three major optimization objectives with economic optimization and safety controllability as the core: economic objective, voltage safety objective, and operational stability objective.

6. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 5, characterized in that, The three optimization objectives are expressed as follows: Economic objective: Minimize network losses and switching operation costs. The objective function is: ； Where T is the number of time periods within the optimization period; P ij,t Q ij,t U represents the active power and reactive power of branch ij at time t. i,t Let R be the voltage at node i at time t. ij Let C be the resistance of branch ij, Δt be the duration of the time interval, and C be the resistance of branch ij. s Cost per single switching operation; x ij,t Let x be the switching state of branch ij at time t. ij,t =1 indicates a closed loop, x ij,t =0 indicates disconnection; |x ij,t+1 -x ij,t | represents the change in switch state between adjacent time periods, with 1 when it changes and 0 when it remains unchanged; E is the set of branches in the distribution network; Voltage safety objective: Minimize node voltage offset The objective function is: ； Where N is the total number of distribution network nodes, U N The node's rated voltage; Let be the voltage offset rate of node i at time t; T is the total number of time periods within the optimization period. Operational stability objective: Branch load rate equalization The objective function is: ； Where |E| represents the total number of distribution network branches, and S ij,t Let S be the apparent power of branch ij at time t. ij,max The rated apparent power of branch ij, Let t be the average load rate of all branches in the network at time t.

7. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 6, characterized in that, The constraints in step S3 include: Current constraints: ； Among them, P i,t Q i,t P represents the injected active and reactive power at node i at time t. PV,i,t Let t be the output of the photovoltaic system connected to node i at time t; Let i be the set of adjacent nodes of node i; Let be the load active power of node i at time t. Let be the reactive power of the load at node i at time t; Node voltage constraints: ； Among them, U min and U max These are the lower and upper limits of the node voltage, respectively. Branch capacity constraints: ； Among them, S ij,max This refers to the rated capacity of the branch circuit; Topological constraints include switch state constraints, radial constraints, and switch operating frequency constraints. Switch state constraint: x ij,t ∈{0,1}, meaning the switch only has two states: closed or open; Radial constraint: The reconfigured distribution network must maintain a radial topology, with no loops or islands, meaning that all load nodes are connected to the power source through closed branches; Switch operation frequency constraint: The number of times a single switch operates within one optimization cycle shall not exceed 3. Photovoltaic output constraints: ； Among them, P PV,max,i The rated output of the photovoltaic system connected to node i.

8. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 7, characterized in that, The process of solving the multi-objective optimization model for dynamic reconfiguration of the active distribution network based on the improved adaptive particle swarm optimization algorithm in step S4 includes: D1: Sequentially execute the coding scheme design, chaotic initialization of the population, and selection of feasible solutions; D2: Introducing dynamic inertia weights and adaptive learning factors; D3: Particle velocity and position update rules: Combining the characteristics of binary encoding, design discretized velocity and position update formulas; D4: Robustness Enhancement Strategy for Photovoltaic Fluctuations: To improve the algorithm's adaptability to photovoltaic power output fluctuations, a scenario robustness penalty factor and dynamic search step size adjustment are introduced; D5: When the iteration termination condition is met, the optimal solution is selected.

9. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 8, characterized in that, Step S5 includes: E1: Construct a rolling temporal optimization framework, including: E1-1: Time-domain window division: Divide the total optimization period, such as 24 hours, into several continuous rolling windows, each with a duration of τ. τ is dynamically adjusted according to the photovoltaic fluctuation scenario. E1-2: Data update within the window: Real-time acquisition of power distribution network operation data when each scrolling window starts; E1-3: Dynamic Update of Optimization Objectives and Constraints: Based on the updated photovoltaic power output data and operating status, adjust the optimization objective weights and constraint thresholds for the current window; E2: Perform rolling iterative optimization of the initial reconstruction scheme, specifically including: E2-1: Solving the Model Within a Window: For each scrolling window, using the updated photovoltaic power output data and operating status data as input, the improved adaptive particle swarm optimization algorithm is called to resolve the multi-objective optimization model, obtaining the optimal reconstruction scheme within the current window. ; E2-2: Scheme Smooth Transition Constraint: Introducing a scheme smoothing constraint to limit the amount of change in the switching states of adjacent windows: ； in, K represents the on / off state at the midpoint of the previous window. max The maximum number of switching changes allowed; E2-3: Iterative update of the reconstruction scheme: After each scrolling window ends, output the optimal reconstruction scheme for the second half of the current window and execute it; E3: Emergency corrections to the reconstruction plan based on real-time monitoring, including: E3-1: Real-time monitoring of operating status: During the execution of each rolling window, key operating indicators are monitored in real time through the distribution network synchronous phasor measurement device and data acquisition and monitoring system; E3-2: When the correction triggering conditions are met, the emergency correction mechanism is triggered; E3-3: The fast gradient descent algorithm is used to implement local adjustments to the reconstruction scheme.

10. The active distribution network dynamic reconfiguration optimization method for photovoltaic power output fluctuations according to claim 9, characterized in that, The local adjustment process of the reconstruction scheme in step E3-3 includes: E3-3-1: Locating the root cause of the problem: For voltage over-limit nodes, identify the switch status of adjacent branches and determine the key adjustable switches; for branch overload, screen the parallel or series branch switches of the overloaded branch. E3-3-2: Local optimization objective: With the goal of eliminating constraint violations as quickly as possible, the local optimization objective is set as minviolation(X), where violation(X) is the degree of constraint violation; E3-3-3: Local adjustment of switch states: Only the state of key switches is optimized, and the switch state combination that minimizes the degree of constraint violation is found by using the gradient descent method; E3-3-4: Scheme Verification: After modification, verify whether the distribution network topology meets the radial constraint and the island constraint. Execute the modified scheme immediately after ensuring its feasibility.

Images