Flexible DC power distribution network fault recovery optimization method and system based on improved genetic algorithm, computer terminal and readable storage medium

Abstract

The invention relates to a flexible DC power distribution network fault recovery optimization method and system based on an improved genetic algorithm, a computer terminal and a readable storage medium, and is suitable for a high-proportion wind and light access scene. The method comprises the following steps: constructing a multi-scene recovery model fusing wind power and photovoltaic uncertainty, and generating a typical wind and light power scene by adopting Monte Carlo sampling; establishing a recovery optimization model by taking the maximum load recovery amount and the minimum network loss as targets; based on voltage source converter control constraints, an improved genetic algorithm is designed, and optimization efficiency and solution feasibility are improved through adaptive crossover, variation and parallel evaluation. According to the method, the recovery capability and the operation stability of the flexible direct current power distribution network under the fault can be improved, the method is suitable for a power distribution system with high new energy permeability, and compared with other fault recovery methods and traditional optimization methods, the method is more effective in improving the load recovery amount and reducing the active loss of the network.

Description

Technical Field

[0001] This invention relates to the field of power system automation and intelligent dispatch, specifically to a fault recovery optimization method for flexible DC distribution networks based on an improved genetic algorithm. It is particularly applicable to multi-scenario recovery control and intelligent optimization of flexible DC distribution systems under scenarios with high proportions of wind power and photovoltaic and other new energy access. It also relates to a corresponding system, computer terminal and computer-readable storage medium. Background Technology

[0002] With the continued advancement of the "dual carbon" target, the penetration rate of renewable energy sources such as wind power and photovoltaics in distribution networks is rapidly increasing, driving the restructuring of new distribution system structures and the evolution of operating modes. Flexible DC distribution networks, with their advantages in renewable energy-friendly integration, efficient transmission, and flexible power flow control, are gradually becoming an important technological path to support high-proportion renewable energy consumption. However, with the large-scale integration of distributed wind and solar power, the uncertainty of distribution system operation has significantly increased. When the system encounters faults, it is prone to problems such as large voltage fluctuations, power imbalances, and obstructed recovery paths, seriously affecting power supply reliability and grid resilience. Especially under the flexible DC network structure, factors such as renewable energy volatility, the control response characteristics of voltage source converters (VSCs), load level differences, and the allocation of regulatory resources make system fault recovery and scheduling more complex.

[0003] In recent years, scholars both domestically and internationally have conducted extensive research on the distribution network restoration problem in high-penetration renewable energy scenarios. Some studies focus on improving regional load recoverability through islanding and distributed power source collaborative optimization strategies, while others introduce demand-side response and smart soft-switching (Soft Open Point, SOP) as active control measures to expand the restoration boundary and alleviate local power gaps. Building on this, some studies have also attempted to incorporate the control boundary and operating mode of voltage source converters into restoration optimization modeling to construct a coordinated control mechanism, thereby improving restoration efficiency and power supply stability. Meanwhile, to address the large-scale nonlinearity, multi-objective conflicts, and uncertainties in the system, heuristic optimization algorithms such as genetic algorithms, particle swarm optimization, and gray wolf algorithms have been widely applied to solving restoration models, achieving certain results in improving computational efficiency and expanding the feasible solution space. Furthermore, in recent years, some scholars have introduced the uncertainty of wind and solar power output into restoration scheduling models, employing Monte Carlo sampling or probabilistic scenario generation techniques to improve the robustness of restoration strategies.

[0004] Despite the positive progress made in the field of distribution network restoration, the following shortcomings still exist in practical scenarios for flexible DC distribution networks: First, the coupling between wind and solar power fluctuations and restoration scheduling mechanisms is insufficient, making it difficult to form a dynamic response modeling framework; second, the efficiency of existing optimization algorithms in solving large-scale flexible DC systems under multiple fault scenarios still needs improvement. Therefore, there is an urgent need for an integrated restoration method that takes into account both wind and solar power fluctuation modeling and efficient optimization algorithm design to improve the fault recovery capability of flexible DC distribution networks in the context of high penetration of new energy sources. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, the purpose of this invention is to solve the following technical problems in the fault recovery process of existing flexible DC distribution networks: 1) The fluctuation of wind and solar power output is not sufficiently modeled, making it difficult to cope with the recovery uncertainty under high-penetration renewable energy conditions; 2) Existing optimization algorithms are inefficient in solving complex constraints and multi-scenario recovery problems, making it difficult to meet the requirements of rapid engineering recovery. Therefore, this invention provides a flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm. Through multi-scenario modeling and collaborative optimization using an improved genetic algorithm, it achieves efficient and reliable recovery of flexible DC distribution networks under high-penetration wind and solar conditions. It also provides a corresponding system, computer terminal, and computer-readable storage medium.

[0006] According to one aspect of the present invention, a fault recovery optimization method for flexible DC distribution networks based on an improved genetic algorithm is provided, comprising:

[0007] Multi-scenario wind and solar power modeling: A multi-scenario power input model reflecting the uncertainty and fluctuation of wind and solar power is constructed by using typical day segmentation and Monte Carlo sampling methods;

[0008] Recovery optimization model construction: With the goal of maximizing the system load recovery and minimizing the power loss of the entire network, a multi-scenario recovery optimization model for flexible DC distribution network is established, embedding constraints such as the adjustment boundary, voltage support capability and control mode selection logic of VSC converter;

[0009] Algorithm solution strategy design: An improved genetic algorithm is introduced, and an adaptive crossover and mutation probability adjustment mechanism, a constraint penalty mechanism, and a parallel fitness evaluation mechanism for landscape scenes are adopted to improve the model solution efficiency and the feasibility of the solution;

[0010] Recovery strategy output: Outputs the optimal load recovery scheme, converter control commands and flexible switch configuration strategies to realize power supply recovery of flexible DC distribution network under fault scenarios.

[0011] According to another aspect of the present invention, a flexible DC distribution network fault recovery optimization system based on an improved genetic algorithm is provided, comprising:

[0012] The data acquisition module is used to collect power grid structure parameters, wind and solar power output forecast data, load demand data, and VSC converter operating parameters, and to construct a standardized input dataset.

[0013] The scene generation module is used to generate multiple sets of wind and solar power output scenes with uncertainties based on typical daily load curves and Monte Carlo sampling methods, and form a multi-scene power matrix as model input;

[0014] The model building module is used to establish a multi-scenario recovery optimization model with the objective functions of maximum load recovery and minimum network loss, and integrates various engineering constraints such as VSC control capability, voltage and current constraints and smart soft switching (SOP) capacity limit;

[0015] An optimization solution module is used to run an improved genetic algorithm to iteratively solve the optimization model. This module includes an adaptive crossover and mutation mechanism, a constraint penalty mechanism, and a parallel fitness evaluation mechanism to improve convergence efficiency and solution feasibility.

[0016] The recovery strategy output module is used to output the optimal recovery strategy results, including node load recovery status, VSC operation mode, and SOP power configuration scheme.

[0017] According to a third aspect of the present invention, a computer terminal is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can be used to perform the methods described above in the present invention, or to run the system described above in the present invention.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, can be used to perform the methods described above in the present invention, or to run the system described above in the present invention.

[0019] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0020] This invention enhances the adaptability of the recovery strategy to the uncertainty of new energy output by modeling and generating wind and solar power fluctuations. By introducing a VSC control mechanism, it achieves multi-dimensional constraint coordination and flexible control of the converter during the recovery process. Furthermore, by improving the genetic algorithm, it significantly improves the solution efficiency and feasible solution quality in complex multi-scenario recovery problems. Simulation results show that the proposed method can achieve higher load recovery rates and lower network losses under various fault and high-penetration scenarios, demonstrating good engineering adaptability and practical value. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the present invention will be further described below with reference to the accompanying drawings. The accompanying drawings are used to schematically illustrate the basic structure and process of the present invention and do not constitute any limitation on the present invention.

[0022] Figure 1 This is a flowchart of a flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm in one embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of the modified IEEE 33-node system topology with DC feeder in one embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram comparing the convergence curves of three algorithms in one embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the modified IEEE 33-node system topology containing a DC ring network in one embodiment of the present invention;

[0026] Figure 5 This is a framework diagram of a flexible DC distribution network fault recovery optimization system based on an improved genetic algorithm in one embodiment of the present invention. Detailed Implementation

[0027] The embodiments of the present invention are described in detail below: These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

[0028] like Figure 1 As shown, one embodiment of the present invention provides a fault recovery optimization method for flexible DC distribution networks based on an improved genetic algorithm, applicable to DC distribution systems with high wind and solar power penetration. This method constructs a multi-scenario recovery modeling framework that takes into account the uncertainties of wind and solar power. Based on typical day segmentation and Monte Carlo sampling, wind and solar power scenarios are generated to characterize their dynamic impact on the recovery path and power balance. A recovery optimization model integrating VSC control capabilities is established, comprehensively incorporating engineering constraints such as active / reactive power regulation boundaries, voltage support capabilities, and control mode switching logic. A fault recovery algorithm framework adapted to the fluctuating characteristics of wind and solar power is designed, introducing an improved genetic algorithm (IGA). Adaptive mutation and scenario parallelism mechanisms enhance the algorithm's feasibility and convergence efficiency.

[0029] Specifically, the flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm provided in this embodiment may include the following steps:

[0030] Step 1: Construct a multi-scenario wind and solar power input model. The model uses a typical day segmentation method and Monte Carlo sampling technology to generate multiple power scenarios that characterize the volatility of wind and solar power.

[0031] Step 2: Establish a flexible DC distribution network fault recovery optimization model with the objective function of maximizing the total system load recovery and minimizing the total network power loss, and consider engineering constraints such as the active / reactive power regulation boundary, voltage support capability, and control mode switching logic of the voltage source converter (VSC).

[0032] Step 3: Design an improved genetic algorithm to solve the optimization model. Improve the genetic algorithm by using an adaptive crossover probability and mutation probability adjustment mechanism, a constraint penalty function mechanism, and a parallel fitness evaluation mechanism for landscape scenes.

[0033] Step 4: After the objective function converges or reaches the preset number of iterations, output the optimal fault recovery strategy.

[0034] In some preferred embodiments, step 1 above involves constructing a multi-scenario wind and solar power input model. This model employs a typical day segmentation method and Monte Carlo sampling technology to generate multiple power scenarios characterizing the volatility of wind and solar power, including:

[0035] The analysis of typical fault recovery scenarios is based on the dimensions of fault scope, affected area, and resource availability, and the fault recovery scenarios of flexible DC distribution networks are divided into three typical scenarios, as shown in Table 1.

[0036] Table 1 Classification of Fault Recovery Scenarios

[0037]

[0038] The wind and solar power fluctuation modeling and scene generation method can classify the typical daily types of wind and solar power generation characteristics based on actual meteorological data or empirical statistics, as shown in Table 2.

[0039] Table 2 Typical Day Types

[0040]

[0041] To reflect the time-varying and uncertain nature of wind and solar power output and to provide input support for multi-scenario fault recovery models of flexible DC distribution networks, the wind and solar volatility modeling and scenario generation method employs Monte Carlo sampling (MCS) to construct multi-scenario wind / solar power output sequences based on typical days. The steps include:

[0042] 1) Determine the probability distribution function of wind and solar power output. Wind speed and solar irradiance are the main driving factors for wind and solar power output. Wind speed v can be approximately followed by a Weibull distribution, with the probability density function as follows:

[0043]

[0044] In the formula, k is the shape parameter and λ is the scale parameter, which are usually obtained based on historical data statistics or regional wind resource standards.

[0045] Irradiance G: can be approximated by a normal distribution or a triangular distribution.

[0046]

[0047] In the formula, μ and σ are the mean and standard deviation of irradiance, respectively, and can be fitted by hourly intervals.

[0048] 2) Power Conversion Model. The sampled wind speed and irradiance values ​​are converted into wind power and photovoltaic power output through the following output model. The power output of the wind turbine exhibits a typical nonlinear S-shaped characteristic with respect to wind speed, which can be expressed as:

[0049]

[0050] In the formula, v in v rated v out These are the cut-in, rated, and cut-out wind velocities of the fan, respectively.

[0051] Photovoltaic power output is mainly related to solar irradiance (radiance), exhibiting a bell-shaped curve distribution. Under standard illumination, photovoltaic power output can be approximated as follows:

[0052] P pv (t)=P rated ·f irrad (t)·η pv

[0053] In the formula, f irrad (t) represents the normalized irradiance curve per unit time, η pv Photovoltaic conversion efficiency.

[0054] 3) Construct a multi-scenario power output sequence set. By independently sampling wind speed v and irradiance G N times, construct multi-scenario power sequences for wind power and photovoltaic power:

[0055]

[0056] In the formula, i represents the i-th scene sample; t represents the time slice; and T is the total number of simulation time periods. Let represent the wind power and solar power output at time t in this scenario, respectively.

[0057] 4) Typical scenario extraction and dimensionality reduction. To reduce the computational burden and improve the solvability of the restoration model, K-means clustering can be used for all N samples to select K << N representative scenarios, which form the final optimized input set Ω = {s1, s2, ..., s K}. Each representative scenario s k corresponds to a wind power-photovoltaic combined output sequence The finally constructed scenario set will be used as the dynamic input parameter in the restoration optimization model to participate in the evaluation and decision-making of the objective function and power constraints.

[0058] In some preferred embodiments, in the above step 2, a flexible DC distribution network fault restoration optimization model is established, with the maximization of the total system load restoration and the minimization of the total network power loss as the objective function, and considering engineering constraint conditions such as the active / reactive power regulation boundary, voltage support ability, and control mode switching logic of the voltage source converter (VSC), including:

[0059] 1) Construct the objective function. Model the fault restoration as a multi-objective optimization problem, with the maximization of the system load restoration and the minimization of the network active power loss as the core objectives, and construct a comprehensive objective function F through a linear weighted strategy:

[0060] maxF = w1·F1 + w2·(-F2)

[0061] Where w1, w2 ∈ [0, 1] are weight factors, satisfying w1 + w2 = 1; -F2 means that minimizing the network active power loss is equivalent to maximizing its opposite.

[0062] The maximization of the load restoration includes: Define the load node set as N L , and the rated load of node i ∈ N L is P i load , and its restoration state is represented by the binary variable δ i ∈ {0, 1}, where δ i = 1 means that node i is restored to power supply, and δ i = 0 means that node i is out of power supply. In scenario s ∈ Ω, the total restored load of the system is:

[0063]

[0064] Take the weighted expectation of the restored load under all scenarios as the optimization objective:

[0065]

[0066] Where π s is the probability weight.

[0067] The minimization of network active power loss includes: assuming the set of DC distribution network branches is E. DC Branch (i,j)∈E DC The resistance is R ij The current is The total line loss in scenario s is:

[0068]

[0069] The expected network loss across all scenarios is used as the second optimization objective:

[0070]

[0071] 2) Constructing Model Constraints. To ensure power quality and system stability during fault recovery of the flexible DC distribution network, the constraints introduced into the recovery optimization model include:

[0072] Node power balance constraint, for any DC node i∈N DC Its power balance should satisfy:

[0073]

[0074] In the formula, Let (j,i) be the power of branch (j,i) in scene s; It provides power to the distributed wind and solar power sources connected to node i; The load power depends on the recovery state. This constraint ensures the active power supply and demand balance of the node in any scenario.

[0075] Node voltage constraints: DC node voltages must be within the allowable operating range.

[0076]

[0077] Branch current constraints: The current in each DC branch must meet its thermal stability upper limit.

[0078]

[0079] Furthermore, the branch power can be represented by a linear relationship between voltage and current:

[0080]

[0081] The VSC output power limit, for any VSC node m, should satisfy its power regulation boundary in each scenario:

[0082]

[0083] Considering the actual operation mode switching, the following control mode logic consistency must be met: When VSC is in PQ control mode, its (P,Q) output must be specified; when it is in U... dc -Q control mode requires maintaining a stable DC bus voltage; mode switching rules can be implemented via binary variable α. m,k This means that the following conditions are met:

[0084]

[0085] Load recovery status logic constraints, whether the load has been restored is determined by the binary variable δ. i (s) Control measures must be implemented in accordance with node voltage, current capacity, and other conditions. To ensure that recoverable nodes have sufficient power supply support, the following constraints can be added:

[0086]

[0087] That is, when a node is restored to power, its actual available power must be greater than or equal to the load value.

[0088] Uncertainty boundary constraints for renewable energy output: Based on Monte Carlo sampling results, wind and solar power have been generated as fixed inputs for each scenario, but the following confidence boundary conditions can still be introduced:

[0089]

[0090] This boundary can be determined based on typical daily empirical curves or statistical analysis of wind and solar prediction errors, thereby enhancing the robustness of the model.

[0091] The constraints of the SOP are as follows: Suppose the SOP is installed on branch (i,j), connecting two adjacent nodes i and j, and the adjustable power flow is... SOP active, reactive, and capacity constraints:

[0092]

[0093] In the formula, Depends on converter capacity; positive values ​​indicate power supply from node i to node j, negative values ​​the opposite; capacity constraints limit the actual operating point to fall within S... SOP The PQ capability circle is within a radius of 1.

[0094] In some preferred embodiments, step 3 above, which involves designing an improved genetic algorithm to solve the optimization model, improves the genetic algorithm through an adaptive crossover probability and mutation probability adjustment mechanism, a constraint penalty function mechanism, and a parallel fitness evaluation mechanism for landscape scenes, including:

[0095] To address the challenge of efficiently solving multi-scenario fault recovery optimization models for flexible DC distribution networks under conditions of high variable dimensionality, complex constraints, and strongly nonlinear objective functions, an Integral Genetic Algorithm (IGA) with good feasibility and convergence was designed based on the traditional genetic algorithm (GA). Considering the mixed variable characteristics (continuous + binary) in the model, a real-integer hybrid encoding structure was adopted. Specifically, floating-point encoding was directly used for node voltage, current, and VSC power; while for load recovery state δ... i and VSC control mode selection α m,k Binary variable encoding is used. Each individual's chromosome is represented as:

[0096] Individual=[δ1,δ2,...,U1,...,I ij ,...,P VSC Q VSC ,...,α m,k ]

[0097] The fitness function is based on a multi-objective synthesis function F. By transforming a complex set of constraints into an optimization function for unified solution, convergence efficiency is improved. The minimization objective is:

[0098]

[0099] In the formula, γ represents the penalty for violating the constraint:

[0100]

[0101] In the formula, g r (X) represents the r-th inequality constraint; R is the total number of constraints; λ r This is the penalty weighting factor.

[0102] To improve the stability and convergence performance of the algorithm, the following improvement strategies are proposed:

[0103] Adaptive crossover and mutation probabilities, dynamically adjusting operation probabilities based on individual fitness and generation:

[0104]

[0105] In the formula, f represents the current fitness of the individual; min f max This represents the current population extreme value; g / g max P represents the current generational ratio; c P m For adaptive crossover and mutation probabilities.

[0106] Elite retention strategy: retain the top N with the highest fitness in each generation. e Individuals directly enter the next generation, avoiding the loss of the optimal solution.

[0107] Constraint-guided mutation targets individuals that violate constraints, prioritizing minor perturbations at the location of the violation variable to increase the proportion of feasible solutions.

[0108] Parallel scene evaluation mechanisms are needed because individual fitness evaluation is time-consuming in multi-scene models. Introducing a parallel scene mechanism, by distributing the calculation tasks of K scene fitness for each individual to multi-threaded parallel processing, can accelerate the calculation speed and adapt to large-scale landscape scene restoration modeling.

[0109] The pseudocode for the improved genetic algorithm (IGA) is shown in Algorithm 1.

[0110]

[0111] Based on the same inventive concept, an embodiment of the present invention also provides a flexible DC distribution network fault recovery optimization system based on an improved genetic algorithm.

[0112] Specifically, such as Figure 5 As shown in the figure, the flexible DC distribution network fault recovery optimization system based on an improved genetic algorithm provided in this embodiment may include the following modules:

[0113] The data acquisition module is used to collect power grid structure parameters, wind and solar power output forecast data, load demand data, and VSC converter operating parameters, and to construct a standardized input dataset.

[0114] The scene generation module is used to generate multiple sets of wind and solar power output scenes with uncertainties based on typical daily load curves and Monte Carlo sampling methods, and form a multi-scene power matrix as model input;

[0115] The model building module is used to establish a multi-scenario recovery optimization model with the objective functions of maximum load recovery and minimum network loss, and integrates various engineering constraints such as VSC control capability, voltage and current constraints and smart soft switching (SOP) capacity limit;

[0116] An optimization solution module is used to run an improved genetic algorithm to iteratively solve the optimization model. This module includes an adaptive crossover and mutation mechanism, a constraint penalty mechanism, and a parallel fitness evaluation mechanism to improve convergence efficiency and solution feasibility.

[0117] The recovery strategy output module is used to output the optimal recovery strategy results, including node load recovery status, VSC operation mode, and SOP power configuration scheme.

[0118] The flexible DC distribution network fault recovery optimization system based on an improved genetic algorithm provided in the above embodiments of the present invention mainly includes the following steps in its working process:

[0119] Step 1: Utilize the data acquisition module, which supports the data awareness and preprocessing module throughout the entire fault recovery optimization process. Its functions include, but are not limited to, the following:

[0120] The power grid structure parameter acquisition unit is used to obtain the basic structure information of the current flexible DC distribution network from the dispatch master station or SCADA system. This information mainly includes: node number and node type (load node, power generation node, VSC node, etc.); topology connection relationship (DC feeder, ring network branch); parameters such as branch resistance, admittance, and current capacity; access location and equipment number of SOP and VSC; real-time / short-term operating status (switch status, communication availability, etc.). The acquired information is used to construct the topology map and branch parameter matrix of the power grid, providing input basis for power balance constraints and power flow calculations in the recovery model.

[0121] The wind and solar power output prediction data acquisition unit interfaces with wind farms and solar power stations or new energy prediction systems to acquire the following information: predicted output (power) curves of distributed wind turbines and solar modules, with configurable time granularity (e.g., 15 minutes or 1 hour); predicted values ​​of meteorological factors such as wind speed, irradiance, and temperature; uncertainty intervals and confidence levels (e.g., 95% confidence interval) for wind and solar predictions; and a historical data archiving interface for building probability distribution models and typical daily power patterns. The output data from this unit is used for subsequent multi-scenario wind and solar power modeling and Monte Carlo sampling modules.

[0122] The load forecasting and monitoring unit is used to collect or forecast the load demand of each node in the distribution network, including: load forecast values ​​(active load / controllable load) of different nodes; load level classification information (level 1, level 2, interruptible, etc.); load time-varying characteristic parameters (such as daily peak, average, load factor); real-time monitoring data of current load and historical load curves; and standardized input format through standardization processing (such as normalization, maximum and minimum scaling) for use by subsequent optimization algorithms.

[0123] The converter operation parameter acquisition unit collects the operating boundaries and control attributes of all VSC (Voltage Source Converter) devices in the flexible DC distribution network, including: maximum / minimum output active power; maximum / minimum output reactive power; configurable control modes (PQ control, Udc-Q control, Uac-θ control); supported dynamic adjustment capabilities and control response time; current converter operating status (online, offline, fault, etc.); voltage regulation range and reference voltage value. This data is a key input for modeling the VSC regulation boundary constraints and control mode selection logic.

[0124] The data preprocessing and standardization unit performs the following on the above-mentioned data types: missing value handling and time alignment; normalization and standardization (such as Min-Max scaling); and establishes a data input vector matrix with a unified structure for use by the recovery modeling and optimization algorithm module.

[0125] Step 2: Using the scenario generation module, generate typical output scenarios reflecting the power fluctuations of wind and solar power, and organize them into a standardized multi-scenario input matrix for use by the subsequent recovery and optimization model. This module mainly includes the following functional units:

[0126] The typical daily load division and time slice construction unit divides the wind and solar power output into segments according to the typical daily operation curve, dividing it into multiple representative time slices (such as early morning, morning, afternoon, evening, etc.). Statistical characteristic (mean, variance) parameters are constructed for each time slice to characterize the typical changing trends of photovoltaic irradiance and wind speed at different times.

[0127] The probability distribution fitting and parameter estimation unit, based on historical meteorological observation data (such as wind speed, solar irradiance, temperature, etc.), performs statistical analysis on each meteorological factor and fits it to the corresponding probability distribution model. For example, wind speed is fitted to a Weibull distribution, and solar irradiance is fitted to a normal distribution. These probability models are used to simulate the volatility and uncertainty of wind and solar resources.

[0128] The Monte Carlo sampling unit, based on the fitted probability distribution function, uses the Monte Carlo method to perform random sampling, generating N sets of wind speed and irradiance scene time series:

[0129]

[0130] In the formula, i represents the i-th scene sample; t represents the time slice; and T is the total number of simulation time periods. Let represent the wind power and solar power output at time t in this scenario, respectively.

[0131] The wind-solar power conversion unit, based on the sampled wind speed and irradiance time series, combined with wind power and photovoltaic power output models, calculates the power output within the corresponding time slot, forming the wind power output P. w (v) and photovoltaic output P pv (t):

[0132]

[0133] P pv (t)=P rated ·f irrad (t)·η pv

[0134] In the formula, v in vrated , v out are the cut-in, rated and cut-out wind speeds of the wind turbine respectively; f irrad (t) is the normalized irradiance curve at the unit time, and η pv is the photovoltaic conversion efficiency.

[0135] Construct a multi-scenario output sequence set unit. By independently sampling the wind speed v and irradiance G N times, construct the multi-scenario power sequences of wind power and photovoltaic:

[0136]

[0137] In the formula, i represents the i-th scenario sample; t represents the time slice; T is the total number of simulated time periods; respectively represent the wind power and photovoltaic outputs at the t-th moment in this scenario.

[0138] Typical scenario extraction and dimensionality reduction unit. Use K-means clustering for all N samples to select K << N representative scenarios, and form the final optimized input set Ω = {s1, s2, L, s K}. Each representative scenario s k corresponds to a wind power-photovoltaic combined output sequence

[0139] Step 3, use the model construction module. The objective function of the fault recovery optimization model is composed of two parts: "maximizing the load recovery amount" and "minimizing the network loss", and is constructed by the linear weighting method. The specific expression is:

[0140]

[0141] In the formula, w1, w2 ∈ [0, 1] are weight factors, satisfying w1 + w2 = 1; π s is the probability weight; the set of DC distribution network branches is E DC , and the resistance of the branch (i, j) ∈ E DC is R ij , and the current is Ω is the set of wind-solar power scenarios; the set of load nodes is N L , and the rated load of node i ∈ N L is P i load , and its restoration state is represented by the binary variable δ i ∈ {0, 1}, where δ i = 1 means that node i is restored to power supply, and δ i = 0 means that node i is out of power supply;

[0142] The engineering constraints in the fault recovery optimization model include the active and reactive power adjustment boundaries of the voltage source converter (VSC), the voltage support range, and the control mode switching logic, specifically including:

[0143] Under any scenario s, the active power output of the VSC converter and reactive power output It should meet the following requirements:

[0144]

[0145] VSC control mode controls the mode variable α. m,k ∈{0,1} means that:

[0146]

[0147] Step 4: Using the optimization solution module, design an improved genetic algorithm by using an adaptive crossover probability P. c And the probability of mutation P m The fitness values ​​of individuals in the current population are dynamically adjusted, and the specific calculation formula is as follows:

[0148]

[0149] In the formula, f represents the current fitness of the individual; min f max This represents the current population extreme value; g / g max P represents the current generational ratio; c P m For adaptive crossover and mutation probabilities;

[0150] A constraint penalty function mechanism is adopted to increase the proportion of feasible solutions searched by the algorithm by imposing penalties on individuals that violate the constraints. The specific penalty function γ is:

[0151]

[0152] In the formula, g r (X) represents the r-th inequality constraint; λ r R is the penalty weighting factor, and R is the total number of constraints;

[0153] The parallel fitness evaluation mechanism for wind and solar power scenarios refers to executing fitness calculation tasks in parallel for each wind and solar power scenario during the fitness calculation process, so as to improve the efficiency of the algorithm.

[0154] Fitness calculation is employed with a multi-objective synthesis function F as the core. By transforming a complex set of constraints into an optimization function for unified solution, convergence efficiency is improved. The fitness function is:

[0155]

[0156] In the formula, γ represents the penalty for violating the constraint.

[0157] Step 5: Using the recovery strategy output module, after completing the fault recovery optimization solution, a complete control strategy is output in a structured manner based on the optimal solution. Instructions can be issued to the scheduling master station or edge controller. This module includes the following functional units:

[0158] The node recovery status determination unit is used to analyze the recovery decision variables of each load node in the optimization results and determine whether it participates in the recovery.

[0159] The VSC operating mode configuration unit is used to extract and parse the operating mode, target output power and control commands of each VSC in the optimization results;

[0160] The SOP power allocation and capacity control unit generates a power flow configuration scheme based on the adjustment path and capacity usage of the SOP during the recovery process.

[0161] The path reconstruction and control command packaging unit intelligently reconstructs the power supply path and packages all control commands into a unified package for transmission.

[0162] It should be noted that the steps in the method provided by the present invention can be implemented using the corresponding components in the system. Those skilled in the art can refer to the technical solution of the system to implement the steps of the method, and can also refer to the technical solution of the method to implement the composition of the system. That is, the embodiments in the system and the embodiments in the method can be understood as preferred examples of each other, which will not be elaborated here.

[0163] The following simulation example further illustrates the effectiveness of the technical solution provided by the above embodiments of the present invention.

[0164] Two types of fault systems are constructed: one containing a DC feeder and the other containing a DC ring network.

[0165] For systems with DC feeder faults, the system setup is performed first. Based on the traditional IEEE 33-node distribution network, a hybrid AC / DC flexible distribution network is formed by introducing flexible DC feeders, renewable energy access points, and VSC interfaces, such as... Figure 2As shown, it has the following main features: 1) All feeders at nodes 26-33 and 19-22 are converted into DC lines; 2) The DC feeders are connected to the AC backbone network through the VSC interfaces at nodes 2 and 6, undertaking the AC / DC power conversion task; 3) An AC / DC interconnection branch is set up between nodes 12 and 22, configured with SOP to achieve multi-path control capability; 4) Wind power resources are connected at node 25, and photovoltaic modules are connected at nodes 18 and 33 respectively, forming a high-penetration wind-solar distributed power supply structure. Assume a three-phase permanent fault occurs between nodes 3 and 4, causing all loads in the areas of nodes 4-18 and 26-33 to lose power; nodes 26-33 (DC side) are primary loads; nodes 4-18 are secondary loads; and loads in other areas are tertiary loads. After fault isolation is completed, the SOP automatically activates, assisting in restoring power supply to the secondary and primary load areas through bidirectional power flow control.

[0166] Then, a comparative analysis of the algorithms was conducted. To evaluate the optimization performance of the proposed IGA algorithm, a comparative analysis was performed using the basic genetic algorithm (GA) and the particle swarm optimization (PSO) algorithm. Figure 3 The optimization and convergence process of three algorithms in typical landscape scenes is demonstrated. Figure 3 In the model, IGA achieves significant load recovery in earlier generations, and its final fitness value is superior to GA and PSO, indicating that it has better global solution search capabilities while simultaneously maximizing load recovery and minimizing network losses. GA converges slowly, with large oscillations, and is prone to getting trapped in local optima. PSO decreases rapidly in the early stages but lacks stability and struggles to improve in the later stages. The convergence performance of these three heuristic algorithms differs significantly in the typical flexible DC distribution network multi-scenario recovery model. The fundamental reason lies in their inherent differences in search strategies, constraint handling mechanisms, and population update methods. IGA introduces an adaptive crossover / mutation probability control strategy based on individual fitness and iteration generations, which can flexibly adjust the search intensity according to the algorithm's running stage. It also introduces an elite retention mechanism to prevent the optimal solution from being lost due to mutation operations, thereby increasing diversity in the early stages, avoiding getting trapped in local optima, and enhancing local search in the later stages, thus improving the quality of solutions and convergence speed. In contrast, the crossover and mutation probabilities in traditional GA are fixed values, which easily get trapped in local optima in the early stages and lack fine-grained search capabilities in the later stages, resulting in slow and unstable overall convergence. PSO updates position and velocity based on the position of the individual and the global optimum, resulting in rapid initial convergence. However, it lacks global jump capability and is prone to getting trapped in local optima when facing complex problems with multiple extrema of the objective function and multiple scenarios.

[0167] Finally, a comparative analysis of fault recovery under different scenarios was conducted. To test the impact of different typical day scenarios on the robustness of the algorithms, four algorithms—IGA, GA, PSO, and mixed integer programming (MIP)—were selected for comparison in two key indicators: recovery load and average network loss, under two typical weather scenarios: normal day and cloudy / rainy day. The results are shown in Table 3.

[0168] Table 3 Comparison of Fault Recovery in Different Scenarios

[0169]

[0170]

[0171] Table 3 shows the following comparative analysis: 1) Under normal day conditions, photovoltaic output is stable, recovery resources are sufficient, and the proposed IGA performs best, with its recovery load approaching the extreme value of MIP and excellent network loss control. This is because the use of adaptive operators and constraint guidance mechanisms improves the quality and convergence speed of its solution, while traditional GA is prone to getting trapped in local optima, has low convergence accuracy, and insufficient network loss control; PSO has weaker performance, but due to the lack of a fine local search mechanism, its network loss is relatively high; the recovery load of MIP is slightly lower than that of IGA, but it is suitable for small-scale systems, and its time consumption increases significantly when there are many scenarios, limiting its practicality. 2) In the scenario of cloudy and rainy days, when the output of photovoltaic power is low and resources are insufficient, IGA still maintains stable performance. Its robust adaptation mechanism in multiple scenarios enables it to find high-quality recovery solutions when the output of wind and solar power fluctuates greatly, making full use of limited wind power and VSC resources. The feasible solution of GA decreases significantly, making it difficult to find feasible recovery strategies quickly. PSO loss is relatively high, and particles tend to gather in the boundary region, which leads to increased voltage offset and increased loss. MIP solution is better, but it is limited by model size and computing resources.

[0172] For systems with DC loop faults, the system setup is performed first. To improve system resilience and SOP control space, a DC loop is constructed by connecting VSC between nodes 18 and 33, as follows: Figure 4 As shown in the diagram. With this configuration, the SOP (Single-Point Power Utilization) possesses multi-path power transfer capabilities when regulating power flow. Especially when the end node loses power, it can obtain supporting voltage from the reverse path, helping to alleviate the problem of large-capacity loads' single-point dependence on the SOP. Then, under the same fault condition, fault recovery optimization calculations are performed.

[0173] Then, SOP capacity sensitivity analysis was performed. Based on the above DC ring network, keeping the fault type unchanged, the SOP capacity was set to 1, 2, 3, 4, and 5 MVA respectively. Recovery optimization was performed using IGA, and the statistical indicators included the recovery load and system network loss, as shown in Table 4.

[0174] Table 4. Load recovery and network loss under different SOP capacities

[0175]

[0176] As shown in Table 4, the system's load recovery capability significantly improved during the construction of the DC ring network and the gradual increase of the SOP capacity. With the SOP capacity increasing from 1 MVA to 5 MVA, the system load recovery rate increased from 95.8% to 100%. Simultaneously, the network loss level increased from 79.54 kW to 85.70 kW. Although network losses and control complexity increased slightly during the capacity increase, they remained within an acceptable range, and the overall power quality and operational stability of the system were not negatively affected. When the SOP capacity was less than 3 MVA, due to limited power supply capacity, the system still needed to perform partial disconnection operations on low-priority loads at the end. With the increase in capacity, the coordinated control capability of the converter and SOP was fully released, enabling the recovery of both important loads and high-capacity load nodes.

[0177] Finally, a sensitivity analysis of wind and solar power penetration is conducted. In flexible DC distribution networks, wind and solar power are core renewable energy sources, and their integration ratio directly affects the system's power supply and demand balance and voltage stability. Therefore, it is necessary to conduct a sensitivity analysis of wind and solar power penetration. Based on Figure 4 The network topology, using IGA, is analyzed under wind and solar penetration rates of 30%, 40%, 50%, 60%, and 70% of the total wind and solar capacity as a percentage of peak load. The statistical indicators, system recovery capability, network loss level, and algorithm feasible solution rate are shown in Table 5.

[0178] Table 5 Comparison of Indicators under Different Wind and Solar Penetration Rates

[0179]

[0180] As shown in Table 5: 1) Load recovery capability shows a trend of first increasing and then decreasing. When the wind and solar penetration rate is less than 50%, the output of new energy sources plays a supporting role in power supply recovery, and the load recovery amount gradually increases. However, when the penetration rate exceeds 60%, the system power fluctuation increases significantly, the voltage stability of local nodes decreases, and some recovery paths fail, resulting in a decrease in the recovery amount. 2) System network loss increases with the increase of penetration rate. The main reason is that the distribution of wind and solar power output becomes randomized, which leads to a longer power dispatch path, a decrease in load reconfiguration capability, an increase in VSC adjustment frequency, and an increase in overall energy loss. 3) The feasible solution rate decreases significantly. Under the condition of 70% penetration rate, about 20% of the scenarios fail to find a recovery solution that meets all constraints due to severe power imbalance, voltage exceeding limits, etc., indicating that the robustness of system recovery decreases.

[0181] In one embodiment of the present invention, a computer terminal is also provided, comprising: a memory; a processor; and a computer program stored in the memory and executable on the processor.

[0182] When the processor executes the computer program, it is configured to perform the steps of the method described in any of the above embodiments of the present invention, or to run the functional modules of the system.

[0183] Optionally, the memory can be used to store the computer program, computer instructions, and related data for implementing the method. The memory may include: volatile memory, such as random access memory (RAM), including static random access memory (SRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), etc.; and non-volatile memory, such as flash memory, etc.

[0184] The processor executes the computer program to implement various steps of the flexible DC distribution network fault recovery optimization method, including wind and solar fluctuation scenario generation, recovery model construction, improved genetic algorithm optimization solution, and recovery strategy output. Specific functions can be found in the descriptions of the above method and device embodiments. The processor and memory can be integrated or separated via a system bus. In some embodiments, the processor and memory can be integrated into a single chip (such as an ASIC) and deployed in a scheduling control terminal, edge computing node, or intelligent master station control device.

[0185] In another embodiment, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to implement the methods described in any of the above embodiments of the present invention, or to run the functional modules of the system.

[0186] The computer-readable storage medium may include: computer storage media, such as disks, optical disks, flash drives, solid-state drives, ROM, RAM, etc.; and communication media, including any medium that enables program transmission in a wired or wireless manner, such as carrier waves, optical signals, electrical signals, etc.

[0187] The computer-readable storage medium can be coupled to a processor, enabling the processor to read program instructions from it and perform recovery and optimization operations, or to write control results into the storage medium. The processor and storage medium can be encapsulated in embedded systems, edge intelligent devices, cloud servers, or other computing terminals to achieve real-time policy control of flexible DC power distribution systems.

[0188] The above embodiments of the present invention provide a flexible DC distribution network fault recovery optimization method, system, computer terminal, and computer-readable storage medium based on an improved genetic algorithm. This method integrates dynamic modeling of wind and solar power output, multi-scenario recovery optimization modeling, VSC control capability constraint expression, and a collaborative optimization strategy based on an improved genetic algorithm. It effectively addresses issues such as uncertain recovery paths, high computational complexity, and tight coupling of control resources in the context of high-penetration renewable energy. By constructing multiple wind and solar scenarios using Monte Carlo sampling and typical days, the adaptability of the scheduling strategy to output fluctuations is improved. The engineering feasibility of the strategy is enhanced by introducing VSC control boundaries, SOP capacity limits, and fault node state variables to construct the recovery optimization model. Furthermore, the feasibility of the solution and the algorithm's solution efficiency are significantly improved by introducing an adaptive genetic algorithm and a scenario parallel mechanism. Simulation results show that the embodiments of the present invention have significant advantages over traditional methods in terms of load recovery rate, network loss level, and computational efficiency, demonstrating good engineering adaptability and promotional value.

[0189] All matters not covered in the above embodiments of the present invention are common knowledge or common technology known to those skilled in the art.

[0190] The specific embodiments of the present invention have been described in detail above. It should be understood that the described embodiments are merely one of the preferred embodiments of the present invention and do not constitute a limitation on the scope of protection of the present invention. Those skilled in the art can make various modifications, substitutions, or variations within the scope defined by the claims without departing from the essence of the present invention, and all such modifications, substitutions, or variations should be included within the scope of protection of the present invention.

Claims

1. A flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm, characterized in that, Includes the following steps: A multi-scenario wind and solar power input model is constructed. The model uses a typical day segmentation method and Monte Carlo sampling technology to generate multiple power scenarios that characterize the volatility of wind power and photovoltaic power. An optimization model for fault recovery of flexible DC distribution network is established, with the objective function being to maximize the total load recovery of the system and minimize the total power loss of the network. Engineering constraints such as the active / reactive power regulation boundary, voltage support capability, and control mode switching logic of voltage source converter (VSC) are also considered. An improved genetic algorithm is designed to solve the optimization model. The improved genetic algorithm includes: an adaptive crossover probability and mutation probability adjustment mechanism, a constraint penalty function mechanism, and a parallel fitness evaluation mechanism for landscape scenes. After the objective function converges or reaches the preset number of iterations, the optimal fault recovery strategy is output.

2. The flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm according to claim 1, characterized in that, The generation of the wind and solar power scenarios employs a multi-scenario simulation method based on Monte Carlo sampling, specifically including: Wind speed and irradiance are modeled as Weibull and normal distributions, respectively, with probability density functions as follows: In the formula, k is the shape parameter, λ is the scale parameter, and μ and σ are the mean and standard deviation of irradiance, respectively; Using the Monte Carlo sampling method, N sets of time series samples of wind speed and irradiance are generated from the above probability distribution, and then substituted into the wind power and photovoltaic power output models to obtain the wind and solar power scene set. in: In the formula, t represents the time slice, and η pv For photovoltaic conversion efficiency, f irrad (t) represents the normalized irradiance curve per unit time, and s = 1, 2, ..., N represents the s-th scene number.

3. The flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm according to claim 1, characterized in that, The objective function of the fault recovery optimization model consists of two parts: "maximizing load recovery" and "minimizing network loss," and is constructed using a linear weighting method, specifically expressed as follows: In the formula, w1 and w2 ∈ [0,1] are weighting factors, satisfying w1 + w2 = 1; π s The probability weights are E; the set of DC distribution network branches is E. DC Branch (i,j)∈E DC The resistance is R ij The current is Ω represents the set of wind and solar power scenarios; N represents the set of load nodes. L Node i∈N L The rated load is P i load Its recovery state is determined by the binary variable δ i ∈{0,1} represents, where δ i =1 indicates that node i has resumed power supply, δ i =0 indicates that node i is out of power; The engineering constraints in the optimization model include the active and reactive power regulation boundaries of the voltage source converter (VSC), the voltage support range, and the control mode switching logic, specifically including: Under any scenario s, the active power output of the VSC converter and reactive power output It should meet the following requirements: VSC control mode controls the mode variable α. m,k ∈{0,1} means that: Each converter can only use one control mode at any given time, and the mode switching is matched with the DC-side node voltage support strategy.

4. The flexible DC distribution network fault recovery optimization method based on an improved genetic algorithm according to claim 1, characterized in that, The adaptive crossover probability P in the improved genetic algorithm c And the probability of mutation P m The fitness values ​​of individuals in the current population are dynamically adjusted, and the specific calculation formula is as follows: In the formula, f represents the current fitness of the individual; f min f max This represents the current population extreme value; g / g max P represents the current generational ratio; c P m For adaptive crossover and mutation probabilities; The constraint penalty function mechanism increases the proportion of feasible solutions searched by the algorithm by imposing penalties on individuals that violate constraints. Specifically, the penalty function γ is: In the formula, g r (X) represents the r-th inequality constraint; λ r R is the penalty weighting factor, and R is the total number of constraints; The parallel fitness evaluation mechanism for wind and solar power scenarios refers to executing fitness calculation tasks in parallel for each wind and solar power scenario during the fitness calculation process, so as to improve the efficiency of the algorithm. The fitness calculation is based on a multi-objective comprehensive function F. By transforming a complex set of constraints into an optimization function for unified solution, the convergence efficiency is improved. The fitness function is: In the formula, γ represents the penalty for violating the constraint.

5. An optimization system applied to the flexible DC distribution network fault recovery optimization method based on the improved genetic algorithm as described in claim 1, characterized in that, include: The data acquisition module is used to collect power grid structure parameters, wind power and photovoltaic output forecast data, load demand data, and VSC converter operating parameters, and to build a standardized input dataset. The scene generation module is used to generate multiple sets of wind and solar power output scenes with uncertainties based on typical daily load curves and Monte Carlo sampling methods, and form a multi-scene power matrix as model input; The model building module is used to establish a multi-scenario recovery optimization model with the maximum load recovery amount and minimum network loss as the objective functions, and integrates various engineering constraints such as VSC control capability, voltage and current constraints and smart soft switching (SOP) capacity limit; An optimization solution module is used to run an improved genetic algorithm to iteratively solve the optimization model. This module includes an adaptive crossover and mutation mechanism, a constraint penalty mechanism, and a parallel fitness evaluation mechanism to improve convergence efficiency and solution feasibility. The recovery strategy output module is used to output the optimal recovery strategy results, including node load recovery status, VSC operation mode, and SOP power configuration scheme.

6. A computer terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When executing the computer program, the processor is configured to perform the method of any one of claims 1 to 4, or to run the system of claim 5.

7. A computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to perform the method of any one of claims 1 to 4, or to run the system of claim 5.

Images