An inverter voltage control parameter optimization method, device, equipment, medium and product

Abstract

The application discloses an inverter voltage control parameter optimization method and device, equipment, medium and product, and relates to the technical field of inverter control parameter optimization and intelligent optimization. The method comprises the following steps: a physical model corresponding to a three-phase off-grid inverter power supply system is constructed and discrete simulation processing is performed; wherein, the three-phase phase voltage at the load end of the three-phase off-grid inverter power supply system is converted into a normalized value and the fundamental effective value is extracted, and the voltage error is obtained by comparing the fundamental effective value with a preset reference value; and the improved river horse optimization algorithm is used to iteratively optimize the optimization variables of the PI controller within a preset time window to output optimal control parameters, so that the inverter voltage control is performed. The application can realize inverter control parameter optimization, and improve the reliability and applicability of parameter setting results.

Description

Technical Field

[0001] This application relates to the field of inverter control parameter optimization and intelligent optimization technology, and in particular to an inverter voltage control parameter optimization method, device, equipment, medium and product. Background Technology

[0002] With the large-scale integration of distributed photovoltaic (PV) systems, energy storage devices, and controllable loads, the number of distributed energy units in new power systems continues to increase. Because individual distributed resources have small capacities, dispersed operating states, and varying control methods, they are difficult to directly participate in electricity market transactions and system dispatch. Therefore, aggregators are typically required to uniformly aggregate, coordinate, manage, and optimize the dispatch of distributed power sources, energy storage devices, adjustable loads, and microgrids. By integrating and controlling their subordinate resources, aggregators can not only improve the utilization efficiency of distributed resources but also enhance the adaptability of local power supply units to external grid disturbances and internal operational fluctuations. This plays a crucial role in demand response, ancillary services, power quality assurance, and improving the reliability of local power supply.

[0003] Among the resources managed by aggregators are numerous inverter-type power supply units centered on photovoltaics and energy storage. These units rely on inverters for power conversion and output control in both grid-connected and off-grid operation scenarios. Particularly in local power supply, islanded operation, or microgrid autonomous operation scenarios, inverters often play a crucial role in voltage establishment and maintenance. For three-phase off-grid inverter power supply systems, inverters need to continuously provide stable amplitude and high-quality three-phase AC voltage to local loads under complex operating conditions such as load changes, load imbalances, and even short-term faults. Therefore, the quality of inverter control directly affects the power supply quality, operational stability, and resource coordination control effectiveness of the distributed power supply units under the aggregator's management.

[0004] Currently, in the control of three-phase off-grid inverters, the voltage closed-loop control method based on sinusoidal pulse width modulation is widely used due to its relatively mature implementation, clear structure, and extensive engineering applications. In this type of control structure, the filtered output voltage at the load end is typically acquired, compared with a reference value, and the modulation ratio is adjusted using a proportional-integral controller to generate drive signals for the inverter's switching devices, thereby achieving dynamic adjustment of the output voltage. This method has advantages such as low implementation cost and relatively direct control logic; however, its control effect is highly sensitive to proportional and integral parameters. If the parameters are not set appropriately, it may lead to large start-up overshoot, significant steady-state error, long settling time, insufficient recovery capability from load abrupt changes, and even oscillations and waveform deterioration under unbalanced or fault disturbances.

[0005] Traditional PI (Proportional-Integral) parameter tuning methods often employ empirical tuning, manual trial and error, or classical tuning methods based on linear models. These methods can achieve certain results when the object model is relatively simple and the operating conditions are relatively uniform. However, for three-phase off-grid inverter systems containing switching devices, saturation limiting, filtering stages, and discrete sampling processes, their dynamic behavior often exhibits significant nonlinearity, coupling, and operating condition dependence. Especially in aggregator applications, the load conditions faced by subordinate power supply units are more diverse, and the switching of operating conditions is more frequent. Simply relying on fixed empirical parameters is insufficient to simultaneously meet the requirements of dynamic response speed, steady-state accuracy, and robustness under different scenarios. Therefore, how to achieve automatic optimization of inverter PI parameters has become a crucial technical problem for improving the control performance of inverter power supply units under aggregators.

[0006] Current research has attempted to introduce intelligent optimization algorithms, such as particle swarm optimization (PSO), into the inverter control parameter tuning process. Intelligent optimization algorithms can directly search for control parameters based on simulation or operational evaluation results without relying on an exact analytical model, making them suitable for solving optimization problems like inverter PI parameter tuning where the objective function is complex, non-differentiable, and susceptible to multiple operating conditions. However, current methods still have certain limitations in practical applications. On the one hand, basic intelligent optimization algorithms have limitations in balancing initial population distribution, global search capability, and local exploitation capability, easily leading to problems such as insufficient initial coverage, premature convergence, getting trapped in local optima, and large fluctuations in results after multiple optimization attempts. On the other hand, in inverter models containing high-frequency switching operations and discrete simulations, fitness evaluation is often accompanied by noise, plateau effects, or even simulation failures, further increasing the instability of the parameter optimization process.

[0007] Especially for aggregators, their application goals are not merely to obtain control parameters with superior performance under a single operating condition, but rather to focus on the consistency and robustness of control parameters across multiple typical operating scenarios. For example, in microgrids, distribution areas, or industrial parks under the unified management of aggregators, inverters may need to simultaneously handle sudden load increases and decreases, three-phase imbalances, local faults, and DC-side fluctuations. If the optimized parameters are only applicable to a certain operating condition and their performance significantly degrades under other conditions, it will be difficult to meet the aggregator's actual needs for unified resource scheduling and stable operation. Therefore, an inverter control parameter optimization method is needed that can balance global search capabilities, local fine-tuning capabilities, and multi-condition robustness evaluation capabilities to improve the reliability and applicability of parameter tuning results. Summary of the Invention

[0008] The purpose of this application is to provide a method, device, equipment, medium, and product for optimizing inverter voltage control parameters, which can optimize inverter control parameters and improve the reliability and applicability of parameter tuning results.

[0009] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for optimizing inverter voltage control parameters, applied in a scenario of aggregator-managed resource energy supply for a three-phase off-grid inverter power supply system with photovoltaic and energy storage as its core; the method includes: A physical model of a three-phase off-grid inverter power supply system is constructed and discrete simulation is performed. Specifically, the three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system are converted into per-unit values ​​and the fundamental effective value is extracted. The voltage error is obtained by comparing the voltage with a preset reference value. An improved Hippo optimization algorithm is adopted to iteratively optimize the PI controller's optimization variables within a preset time window based on the fitness function, so as to output the optimal control parameters. The optimal control parameters are used for inverter voltage control. The optimization variables include proportional parameters and integral parameters. The fitness function is determined based on the voltage error of the three-phase phase voltage. Discrete simulation is performed once in each iteration.

[0010] In one embodiment, the method further includes: The PI control law based on the PI controller performs amplitude limiting modulation within the inverter voltage closed-loop control structure according to the corresponding input voltage error, and outputs the modulation ratio formed after amplitude limiting; the PI control law is determined based on the optimization variables of the PI controller. The modulation ratio is multiplied by a unit sine wave to form a modulation wave, which is then compared with a triangular carrier wave to generate an SPWM pulse to control the inverter's on and off states.

[0011] In one embodiment, the improved strategies introduced by the improved hippo optimization algorithm include: chaotic mapping initialization and opposition learning, adaptive step size / weight scheduling, and embedded differential evolution mutation and crossover.

[0012] In one embodiment, an improved hippo optimization algorithm is employed to iteratively optimize the PI controller's optimization variables within a preset time window based on a fitness function, thereby outputting optimal control parameters, including: Chaotic mapping initialization and opposition learning are used to map chaotic variables to the boundary of decision variables. Opposition learning is used to construct opposition points for each individual to simultaneously evaluate the original and opposition candidates during the initialization phase, forming multiple solutions and selecting the one with the best fitness to enter the initial population; the chaotic variables include optimization variables. Based on the adaptive step size / weight scheduling, a smooth transition from exploration to development is achieved according to the initial population. Based on the aforementioned embedded differential evolution mutation and crossover mutation mechanism to escape local optima and platform stagnation in position updates, this addresses the problem of stagnation caused by the reduced differential update amplitude after population aggregation in the platform target surface or local basin that appears during parameter optimization. Specifically, new solutions are constructed through differential evolution mutation and crossover based on the embedded differential evolution mutation, and greedy selection rules and probability triggering are used to enhance jump exploration; the enhancement strategies corresponding to the enhanced jump exploration include elite-guided Levy flight local search and multi-scenario robust fitness. The optimal control parameters are determined based on the position update information obtained after the position update and the corresponding optimization variables.

[0013] In one embodiment, a new solution is constructed through differential mutation based on the embedded differential evolution mutation and crossover, and a greedy selection rule and probability triggering are employed to enhance jump exploration, specifically including: New solutions are constructed through differential mutation, and the new solutions are subjected to binomial crossover to obtain the test vector. Greedy selection rules and probability triggering are used to enhance jump exploration.

[0014] In one embodiment, the inverter voltage control parameter optimization method further includes: A simulation failure penalty mechanism is introduced, employing either bounce boundary handling or truncation with small perturbations, to ensure that the optimization process is not interrupted and infeasible solutions are automatically eliminated, thus guaranteeing the stable operation of discrete simulation.

[0015] Secondly, this application provides an inverter voltage control parameter optimization device, comprising: The discrete simulation processing module is used to construct the physical model corresponding to the three-phase off-grid inverter power supply system and perform discrete simulation processing; in this module, the three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system are converted into per-unit values ​​and the fundamental effective value is extracted, and the voltage error is obtained by comparing it with the preset reference value. The iterative optimization module is used to iteratively optimize the optimization variables of the PI controller within a preset time window using an improved Hippo optimization algorithm based on the fitness function, so as to output the optimal control parameters. The optimal control parameters are used for inverter voltage control. The optimization variables include proportional parameters and integral parameters. The fitness function is determined based on the voltage error of the three-phase phase voltage. Discrete simulation is performed once in each iteration.

[0016] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described inverter voltage control parameter optimization method.

[0017] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described inverter voltage control parameter optimization method.

[0018] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described inverter voltage control parameter optimization method.

[0019] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, apparatus, equipment, medium, and product for optimizing inverter voltage control parameters. It involves constructing a physical model of a three-phase off-grid inverter power supply system and performing discrete simulation. Specifically, the three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system are converted to per-unit values, and the fundamental effective value is extracted and compared with a preset reference value to obtain the voltage error. An improved Hippo optimization algorithm is used to iteratively optimize the optimization variables of the PI controller within a preset time window to output the optimal control parameters for inverter voltage control. The improved Hippo optimization algorithm can automatically optimize the voltage closed-loop PI controller parameters of the three-phase off-grid inverter power supply system, reducing reliance on manual experience-based tuning, improving parameter tuning efficiency and engineering applicability, and meeting the application needs of aggregators for unified management of distributed inverter resources and high-quality power supply. Therefore, this application can optimize inverter control parameters and improve the reliability and applicability of parameter tuning results. Attached Figure Description

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

[0021] Figure 1 A flowchart of a method for optimizing inverter voltage control parameters; Figure 2 This is a schematic diagram illustrating the operation process of the inverter voltage control parameter optimization method in practical applications. Figure 3 This is a diagram of a three-phase off-grid inverter architecture. Figure 4 This is a control structure diagram of a three-phase off-grid inverter; Figure 5 A schematic diagram illustrating the design process for improving the hippo optimization algorithm; Figure 6 A flowchart illustrating the improved hippo optimization algorithm; Figure 7Schematic diagram of the structural connection of the inverter voltage control parameter optimization device; Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] This application focuses on the control problem of three-phase off-grid inverters based on swarm intelligence algorithms, verifying robustness under load variations, imbalances, and fault conditions. Specifically, the Improved Hippopotamus Optimization Algorithm (IHOA) introduces three improvement strategies: chaotic mapping initialization and Opposition-Based Learning (OBL) to enhance initial coverage and global exploration capabilities; adaptive step size / weight scheduling; and embedding mutation and crossover operators from Differential Evolution (DE) to enhance the ability to escape local optima and plateaus. Furthermore, the IHOA algorithm incorporates elite-guided Lévy flight local search and multi-scenario robust fitness to further improve final performance and generalization robustness.

[0025] In one exemplary embodiment, an inverter voltage control parameter optimization method is provided, applied to an aggregater-managed resource energy supply scenario for a three-phase off-grid inverter power supply system centered on photovoltaics and energy storage. The three-phase off-grid inverter power supply system includes a photovoltaic array, battery energy storage, a DC bus, a three-phase two-level IGBT inverter, an LC low-pass filter, and three-phase loads.

[0026] This method is executed by a computer device, specifically a terminal or server, either alone or jointly. In this embodiment, the method is described using a server as an example. Figure 1 As shown, the inverter voltage control parameter optimization method includes: Step 100: Construct the physical model of the three-phase off-grid inverter power supply system and perform discrete simulation processing. This involves converting the obtained three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system to per-unit values ​​and extracting the fundamental RMS value, and comparing it with a preset reference value to obtain the voltage error.

[0027] The voltage error is used to control the inverter within the inverter voltage closed-loop control structure consisting of a PI controller and a three-phase off-grid inverter power supply system. After being limited and modulated, the voltage error is compared with a triangular carrier wave to generate a sinusoidal pulse width modulation (SPWM) pulse, which is then used to control the inverter.

[0028] In other words, based on the PI controller's PI control law, the voltage error of the corresponding input is used to perform amplitude limiting modulation within the inverter voltage closed-loop control structure, and the output is the modulation ratio formed after amplitude limiting; the PI control law is determined based on the optimization variables of the PI controller.

[0029] The modulation ratio is multiplied by a unit sine wave to form a modulation wave, which is then compared with a triangular carrier wave to generate an SPWM pulse to control the inverter's on and off states.

[0030] Specifically, the PI control law corresponding to the PI controller is: .

[0031] Based on the voltage error, the output of the PI controller is limited to form the modulation ratio: .

[0032] .

[0033] The modulation wave is formed by multiplying the modulation ratio by the unit sine wave. .

[0034] in, It is a PI control law; This is a proportional parameter; For integration parameters; For voltage error; For time; The modulation ratio is determined by... , Decide; This is the upper limit of the modulation ratio (e.g., 0.98 or 0.99), which is slightly less than 1 in actual engineering to reserve a protection margin and avoid modulation distortion; This is a saturation function (saturation limiter) used to crop and limit the input signal to a given upper and lower limit range; It is a modulated wave; Angular frequency; For phase.

[0035] By comparing the modulated waves With triangular carrier Generate SPWM pulses to control the inverter's on and off states.

[0036] Specifically, such as Figure 2 As shown, a closed-loop control structure for inverter voltage is established. The three-phase phase voltage at the load end after filtering is measured, the three-phase phase voltage is converted into per-unit value and the fundamental effective value is extracted. The voltage error is obtained by comparing it with the reference value. The voltage error is input to the PI controller. The output of the PI controller is limited to form the modulation ratio. The modulation ratio is multiplied by the unit sine signal to form the modulation wave, and compared with the triangular carrier wave to generate SPWM pulses to control the three-phase two-level IGBT inverter.

[0037] A parameter optimization model is established using the proportional and integral parameters of the PI controller as optimization variables, and the search boundary of the optimization variables is set. A fitness function is constructed, which evaluates the performance of candidate PI parameters based on the control error of the three-phase phase voltage.

[0038] Step 200: Using an improved Hippo optimization algorithm, the optimization variables of the PI controller are iteratively optimized within a preset time window based on the fitness function to output the optimal control parameters. The optimal control parameters are used for inverter voltage control; the optimization variables include proportional and integral parameters; the fitness function is determined based on the voltage error of the three-phase phase voltage; and a discrete simulation is performed in each iteration.

[0039] The fitness function includes: .

[0040] .

[0041] .

[0042] .

[0043] .

[0044] in, The fitness function; , , as well as All are weighting coefficients; Mean square error; Weighted absolute error for integration time; Maximum overshoot; The modulation ratio smoothness index; The sampling point number; For the first Simulation time for each sampling point; The discrete sampling error represents the voltage error. This represents the number of sampling points; and The modulation ratios at the th The and the first The instantaneous value of each sampling point; This represents the average three-phase error at each sampling point. , , This refers to the voltage error of the three-phase A / B / C phases.

[0045] The improved strategies introduced in the improved Hippo optimization algorithm include: chaotic mapping initialization and opposition learning, adaptive step size / weight scheduling, and embedded differential evolution mutation and crossover.

[0046] An improved hippo optimization algorithm is employed to iteratively optimize the PI controller's variables within a preset time window based on a fitness function, outputting the optimal control parameters, including: Chaotic mapping initialization and opposition learning are used to map chaotic variables to the boundary of decision variables. Opposition learning is used to construct opposition points for each individual to simultaneously evaluate the original and opposition candidates during the initialization phase, forming multiple solutions and selecting the one with the best fitness to enter the initial population. Chaotic variables include optimization variables.

[0047] Based on adaptive step size / weight scheduling, a smooth transition from exploration to development is achieved according to the initial population; wherein, the maximum number of iterations within a preset time window is set to be... The current iteration is .

[0048] Exploration factor decreasing: .

[0049] Increasing development enhancement factor: .

[0050] Location updates employ a combination of elite-guided and differential exploration methods: .

[0051] in, The exploration factor in the current iteration; To explore the maximum value of the factor; To explore the minimum value of the factor; To control the descent shape factor; Development enhancement factors under the current iteration; To develop the minimum value of the enhancement factor; To develop the maximum value of the enhancement factor; To control the rate of ascent factor; Update the information for the corresponding position in the next iteration; Update the position information for the current iteration; and All are random vectors; Indicates element-wise multiplication; These are difference terms; For elites; Individuals randomly selected in the current iteration; It is the globally optimal individual in the current iteration.

[0052] Embedded differential evolution mutation and crossover are used to break out of local optima and plateau stagnation in position updates, in order to solve the problem of plateau target surfaces or local basins in parameter optimization, and the problem of stagnation caused by the smaller differential update amplitude after population aggregation.

[0053] Based on embedded differential evolution mutation and crossover, new solutions are constructed through differential mutation, and greedy selection rules and probability triggering are adopted to enhance jump exploration. The enhancement strategies corresponding to enhanced jump exploration include elite-guided Lévy flight local search and multi-scenario robust fitness.

[0054] The optimal control parameters are determined based on the position update information obtained after the position update and the corresponding optimization variables.

[0055] As an optional implementation, embedding differential evolution mutation and crossover constructs new solutions through differential mutation, and employs greedy selection rules and probabilistic triggering to enhance jump exploration, specifically including: Constructing a new solution through difference mutation: .

[0056] in, This is a new interpretation; , and Each index is a unique random index; This is the scaling factor; , , The solution vector represents three distinct individual positions randomly selected from the population.

[0057] The experimental vector is obtained by performing binomial cross-validation on the new solution. , for dimension have , .

[0058] in, For crossover probability, It is guaranteed that at least one dimension comes from the mutation vector; Let be the component of the test vector u_i in the d-th dimension; Let v_i be the d-th dimension of the mutation vector; Let x_i be the component of the original individual in the d-th dimension.

[0059] Greedy selection rules and probability-based triggering are employed to enhance jump exploration; where, if Then use replace Differential evolution is employed with probabilistic triggering, where the triggering probability decreases with each iteration. .

[0060] The enhancement strategies include elite-guided Lévy flight for localized search and multi-scenario robustness: .

[0061] .

[0062] Let be the probability of triggering the differential evolution operation in the t-th iteration; This represents the maximum trigger probability of the differential evolution operation (used in the early stages of the algorithm). This represents the minimum trigger probability of the differential evolution operation (used in the later stages of the algorithm). This represents the maximum number of iterations within a preset time window. For the current iteration; This is the vector of the new candidate solution after elite perturbation (the new solution generated by Lévy local search); This is the optimal solution vector in the current population (the individual with the lowest / best fitness among all individuals). Scale factor; It is a constant value; Indicates element-wise multiplication; Update location information; This represents the comprehensive weighted fitness function value; For scenario index (s-th simulation scenario); The total number of scenarios (the total number of scenarios under the evaluation conditions, such as different load modes and disturbance scenarios). As weight; Let be the fitness value of the solution calculated in the s-th scenario.

[0063] As an optional implementation, the inverter voltage control parameter optimization method further includes: A simulation failure penalty mechanism is introduced, employing either bounce boundary handling or truncation with small perturbations, to ensure that the optimization process is not interrupted and infeasible solutions are automatically eliminated, thus guaranteeing the stable operation of discrete simulation.

[0064] Rebound boundary handling includes: .

[0065] Truncation plus small perturbations include: .

[0066] in, For the reason , A defined two-dimensional vector; It is the lower bound of the variable; The upper bound of the variable; As a boundary truncation function, x is constrained between the lower bound L and the upper bound U of the variable; It represents a small random perturbation.

[0067] Specifically, such as Figure 2 As shown, the improved Hippo optimization algorithm is used to iteratively optimize the variables. In each iteration, the three-phase off-grid inverter power supply system model is called for discrete simulation. The fitness value is calculated based on the simulation results, and boundary processing is performed on individuals that exceed the limits and penalties are imposed on individuals that fail in the simulation.

[0068] After the stopping condition is met, the optimal PI parameter (optimal control parameter) is output and used for inverter voltage control. Based on the optimal PI parameter, simulation verification is performed under load change, imbalance and fault conditions to evaluate the parameter robustness.

[0069] The improved hippo optimization algorithm includes chaotic mapping initialization and opposition learning initialization strategies. By generating initial individuals and their opposition individuals and evaluating their fitness, the better individuals are selected to form the initial population, thereby improving the coverage and global exploration capabilities of the initial population.

[0070] Improved Hippo optimization algorithms include adaptive step size or weight scheduling strategies, which enhance the population's exploration ability in the early stage of iteration and enhance the local development ability in the later stage of iteration, thereby achieving a dynamic balance between exploration and development during parameter search.

[0071] The improved hippo optimization algorithm incorporates mutation and crossover operators from differential evolution to perturb candidate solutions in the population, generate experimental individuals, and retain individuals with better fitness through selection operations to enhance the ability to escape local optima and alleviate platform stagnation.

[0072] The fitness function is constructed using the three-phase phase voltage error and statistically analyzed based on simulation data within a preset time window to reduce the impact of the transient process in the initial startup phase on the optimization results of the PI parameters.

[0073] The fitness function also includes a constraint evaluation on the smoothness of modulation ratio changes to suppress the adverse effects of modulation ratio jitter on the output waveform quality.

[0074] The fitness function adopts a multi-scenario robust evaluation method, which incorporates at least two of the following operating conditions into the parameter evaluation process: normal steady state, load step, unbalanced load, single-phase ground fault, and DC bus disturbance. The parameters are then comprehensively evaluated according to preset weights to improve the generalization ability and robustness of the optimized PI parameters.

[0075] Boundary handling employs either bounce boundary handling or truncation with perturbation to reduce the probability of candidate solutions clustering at the search boundary. When simulation errors, state divergence, or invalid values ​​occur during the simulation process, a preset penalty value is assigned to the corresponding candidate solution to ensure that the optimization process is not interrupted and infeasible solutions are automatically eliminated.

[0076] The above optimization method can be implemented in a discrete simulation system containing switching devices. The inverter adopts SPWM control mode. The control objective is to maintain the stability of the effective value of the three-phase phase voltage at the load end after filtering under load change, imbalance and fault disturbance conditions, and reduce the total harmonic distortion of the output voltage and current waveforms.

[0077] As can be seen, this application introduces the improved Hippo optimization algorithm into the voltage closed-loop PI control parameter tuning process of a three-phase off-grid inverter. Combined with chaotic mapping initialization, opposition learning, adaptive step size or weight scheduling, differential evolution enhancement, and multi-condition fitness evaluation, it improves the efficiency, accuracy, and robustness of parameter search, thereby improving the voltage regulation performance and waveform quality of the inverter under complex operating conditions, and meeting the aggregator's application needs for unified management of distributed inverter resources and high-quality power supply.

[0078] This application revolves around the technical route of "SPWM three-phase off-grid inverter + voltage closed-loop PI + modulation ratio as control quantity + intelligent optimization algorithm to automatically tune PI parameters + verify robustness under load change, imbalance and fault conditions". The Particle Swarm Optimization (PSO) algorithm used for PI parameter optimization is replaced with the improved Hippo Optimization Algorithm IHOA.

[0079] like Figure 3As shown, in off-grid or weak grid conditions, the inverter essentially acts as a "power shaper." The control parameters directly determine the steady-state accuracy, dynamic recovery speed, disturbance rejection and unbalance rejection capabilities, and waveform quality (such as total harmonic distortion (THD) and three-phase unbalance). However, load mutations, nonlinear loads, filter parameter deviations, DC bus fluctuations, and device non-ideals (dead zone, saturation, delay) can push the system towards stronger nonlinearity and time-varying characteristics. This can cause seemingly usable parameters to exhibit overshoot, oscillation, integral saturation, modulation ratio exceeding the limit, distortion, or even instability under certain boundary conditions. Therefore, the core value of control optimization is not to reduce the error under a single operating condition, but to use a set of achievable, interpretable, and implementable parameters to simultaneously consider stability, dynamic performance, and power quality under multiple operating conditions, while keeping the sensitivity to model uncertainties and disturbances within the acceptable engineering range. The challenge lies in the fact that inverters contain switching devices and limiting / saturation circuits, resulting in a system exhibiting strong nonlinearity, strong coupling, and multiple time scales (carrier frequency, filter resonance, fundamental frequency, RMS calculation window). Performance indicators often conflict with each other (fast response usually leads to larger overshoot or higher harmonics, while suppressing harmonics may sacrifice dynamics). Furthermore, there are constraints (modulation ratio upper limit, device current limit, thermal limitation) and evaluation criteria (which time window, which operating condition, whether to use root mean square (RMS) error or time-weighted absolute error integral performance index (ITAE) / THD / imbalance). This means that the quality of the same parameters depends on the test scenario and indicator combination, making it difficult for traditional analytical tuning and small-signal linearization to fully cover the real performance under these nonlinearities and constraints.

[0080] The challenges of optimizing inverter control parameters based on evolutionary algorithms are as follows: The fitness function typically requires time-domain simulation or experimental measurement using pulse-width modulation (PWM) switches, resulting in high cost, noise, and non-differentiability. Evaluation results are also affected by discrete step length, initial conditions, fault injection timing, RMS calculation window, and numerical errors, exhibiting randomness with slight fluctuations in multiple evaluations of the same parameter, making evolutionary search more prone to misjudgments and premature convergence. Simultaneously, the feasible region is often fragmented by stability and hard constraints (some parameter combinations directly oscillate or trigger protection / saturation), requiring the algorithm to handle boundary conditions, failure simulations, penalty design, and feasibility preservation; otherwise, a large number of evaluations will be wasted on infeasible solutions. Furthermore, inverter control indices are inherently multi-objective and mutually constraining (MSE, MSE, MEP, MEP)... (ITAE, overshoot, settling time, THD, modulation ratio jitter, peak current, etc.) How to combine these into a single fitness (or adopt multi-objective Pareto) that can guide the search without "leading" the algorithm astray requires experience and repeated trials. Finally, the optimality of parameters under one operating condition is not the same as the robust optimality under all operating conditions. If load step, imbalance, fault, and parameter deviation are all included in the evaluation, it will cause a computational explosion of "each individual having to run multiple simulations". Therefore, a difficult trade-off needs to be made between population size, number of iterations, number of scenarios and simulation time. Costs can be controlled and results can be made reproducible and transferable by using dimensionality reduction, surrogate models, parallel computing, phased / layered optimization or robust objective design.

[0081] This application focuses on the method of embedding the IHOA into the inverter control parameter tuning process, the construction of the fitness / objective function, the key points of stable implementation of parameter search in a discrete simulation system containing switching devices, and the key enhancements of IHOA compared to the basic HOA. IHOA employs at least three improvement strategies: chaotic mapping initialization and opposition learning (OBL) to enhance initial coverage and global exploration capabilities; adaptive step size / weight scheduling to achieve a controllable transition from "early exploration to later development"; and embedding differential evolution (DE) mutation and crossover operators to enhance the ability to escape local optima and plateau stagnation. Additionally, optional elite-guided Lévy flight local search and multi-scenario robust fitness can be added to further improve final performance and generalization robustness. This application uses a continuous narrative approach to present system modeling and control structure, optimization variable encoding, objective function and constraint design, IHOA update rules and pseudocode, boundary handling and simulation failure penalty mechanism, multi-condition verification organization methods, and provides suggested parameter configurations and engineering implementation considerations to support simulation reproduction and implementation.

[0082] like Figure 4As shown, at the system architecture level, an off-grid three-phase inverter power supply system typically consists of a photovoltaic array (PV), battery energy storage, a DC bus, a three-phase two-level IGBT inverter, an LC low-pass filter, and three-phase loads. The PV output is DC, which needs to be converted to AC by the inverter to supply power to the local three-phase AC loads. Due to the high-frequency harmonics and spikes generated by PWM switching, the inverter output also needs to pass through an LC filter to obtain approximately sinusoidal three-phase voltage and current waveforms. This application uses sinusoidal pulse width modulation (SPWM) as the switching strategy, with the fundamental frequency set to 50 Hz and the carrier frequency ranging from 10 kHz to 25 kHz (e.g., 25 kHz). The discrete simulation step size is typically on the order of microseconds, such as 2 µs, to capture the high-frequency switching and filtering dynamics. The control objective is to maintain the three-phase phase voltage RMS at the filtered load end at the rated value (e.g., 240V RMS phase voltage, corresponding to approximately 339V peak value) even when the load changes or even when imbalances and short-term fault disturbances occur, and to reduce the total harmonic distortion (THD) of the voltage and current waveforms so that the THD meets engineering specifications (e.g., 5%).

[0083] In terms of control structure, a voltage closed loop is adopted: the three-phase phase voltage at the load end after filtering is measured, converted to per-unit value, and then the 50 Hz fundamental effective value is extracted by the RMS module. The error is obtained by comparing it with the reference. The error is fed into the PI controller to output the modulation ratio control signal. After amplitude limiting, the modulation ratio is ensured to be less than 1 (typical upper limit 0.98-0.99). Then, the modulation ratio is multiplied by a unit sine wave to form a modulation wave, which is compared with a triangular carrier to generate an SPWM pulse. Finally, complementary drive pulses for the upper and lower bridge arms are generated to control the turn-on and turn-off of the six IGBTs. The key point is that the modulation ratio is not a fixed constant, but is dynamically adjusted by the closed-loop PI controller according to the voltage error, so as to achieve rapid compensation and maintain stable system operation under load changes, imbalances or fault disturbances.

[0084] The three-phase voltage is The reference value is The modulation ratio is Triangular carrier is .

[0085] The per-unit RMS error for each phase voltage is defined as: .

[0086] .

[0087] .

[0088] in, This is the RMS voltage of phase A after per-unit standardization. RMS = effective value, which has been normalized to the per-unit system. The per-unit RMS voltage of phase B is the same as above; The RMS voltage after per-unit scaling for phase C is the same as above.

[0089] If a unified three-phase PI parameter is used, the structure of the PI controller for each phase is identical, with only the input error differing. The PI control law is written as: .

[0090] The control output is limited to form the modulation ratio: .

[0091] The above equation satisfies .

[0092] The modulated wave is then: .

[0093] in phase The phase difference varies (A / B / C phases differ by 120°). The SPWM generation module compares... and The switching function is obtained to form the IGBT gate drive. Because... Depend on , The PI parameters determine the dynamic quality of the voltage loop, including start-up overshoot, settling time, steady-state error, load step recovery speed, imbalance and fault compensation capability, and modulation ratio smoothness. An excessively high modulation ratio may lead to oscillations or even instability. Excessive scaling may lead to integral saturation, especially when modulation ratio limiting is present, causing recovery trailing after fault clearance. Therefore, an intelligent optimization algorithm is used for offline (or semi-online) optimization. , The core task of this application is to automatically obtain PI parameters with better performance and stronger robustness under the objective function and constraints. For frequency.

[0094] Optimization modeling begins by defining decision variables and the search space. The most basic encoding is a two-dimensional vector. : Variable boundary settings combine stability and empirical tuning range settings: .

[0095] The above equation satisfies , . , Let K_p be the boundary interval, minimum and maximum values; satisfying K_(p,min)≥0 ensures a reasonable controller structure; , Let K_i be the boundary interval, minimum and maximum values; satisfying K_(i,min)≥0 ensures that the integral is not negative.

[0096] The boundary can be initially set wide and then gradually narrowed, or a more likely stable interval can be directly selected based on experience, for example... , Then, adjust the range based on whether the optimal solution approaches the boundary. If it is necessary to improve the imbalance compensation capability, it can also be expanded to three-phase independent parameters: .

[0097] in, , For the proportional and integral parameters of phase A; , For the proportional and integral parameters of phase B; , These are the proportional and integral parameters for the C phase.

[0098] However, increasing the dimensionality will significantly increase the number of simulation evaluations and computational overhead. Therefore, this application adopts a two-dimensional optimization form with uniform parameters by default.

[0099] The objective function (fitness function) needs to be stably calculated from simulation data, reflecting voltage control quality and suppressing undesirable behavior. When using mean square error (MSE), it can be defined as: .

[0100] in, The discrete sampling error represents the voltage error. This represents the number of sampling points.

[0101] For a three-phase system, the three-phase comprehensive error can be constructed first: .

[0102] .

[0103] To avoid the initial transient state having too strong an impact on the evaluation, an error calculation window can be set to ignore the initial state. Data over a time period, such as statistics within the range of 0.05 s to 0.4 s. To further enhance speed and suppress tailing, a comprehensive objective, namely the fitness function, can be constructed: .

[0104] .

[0105] .

[0106] Used to suppress high-frequency jitter in the modulation ratio, avoiding THD degradation caused by injecting high-frequency components simply to reduce RMS error. Single or combined indicators can be selected as needed without changing the main structure of the IHOA.

[0107] like Figure 5 and Figure 6 As shown, after defining the optimization variables and objective function, it is necessary to construct an IHOA suitable for this problem. Swarm intelligence metaheuristic optimization typically uses population search, iteratively updating the positions of candidate solutions to achieve a balance between exploration and exploitation and to approach the optimal solution. The objective surface for PI parameter optimization is often non-convex, non-differentiable, and may contain noise due to factors such as switching nonlinearity, amplitude limiting saturation, RMS window effect, fault injection mutations, and numerical errors in discrete simulations. Therefore, using metaheuristic algorithms is feasible. However, common problems with metaheuristic algorithms include: poor initial coverage leading to missed high-quality regions; fixed step size resulting in insufficient early exploration or difficulty in convergence later; easy getting trapped in local optima or plateauing after population diversity decays; and large variance in results from multiple runs. The goal of IHOA is to improve convergence speed, final accuracy, and repeatability stability through engineering enhancements.

[0108] The first improvement is Chaotic Map Initialization and Opposites Learning (OBL), designed to enhance initial coverage and early global exploration capabilities. Random uniform initialization may result in uneven coverage with limited samples, while chaotic sequences offer stronger ergodicity, improving population distribution uniformity. Taking the Logistic map as an example: .

[0109] in, . The next value for the chaotic sequence is obtained recursively through the Logistic mapping. The logistic mapping control parameter is μ=4. When μ=4, the sequence has completely chaotic characteristics and strong ergodicity. The current value of the chaotic sequence, with a value range of (0,1), is used to generate the initial sample.

[0110] Mapping chaotic variables to the boundary of decision variables: .

[0111] Opposites learning constructs opposing points for each individual: .

[0112] in, For the first The individual in the first The decision variable values ​​of the dimension are obtained by mapping the chaotic sequence; For the first The lower bound of the decision variable (PI control parameter boundary); For the first Upper bound of the decision variable; For the first Individual, First The opposite point of a dimension (OBL concept, a point symmetrical to the current point within the boundary interval).

[0113] During the initialization phase, both primary and opposing candidates are evaluated simultaneously, resulting in a total of [number missing]. Find the solution and select the one with the best fitness. Each solution is introduced into the initial population, thereby improving the quality and diversity of the initial solutions with a limited number of evaluations.

[0114] The second improvement is adaptive step size / weight scheduling, used to achieve a smooth transition from exploration to development. Let the maximum number of iterations be... The current iteration is Exploration factor decreasing: .

[0115] Option 2 is acceptable. 0.1 is acceptable. To control the shape of the descent.

[0116] Increasing development enhancement factor: .

[0117] 0.1 is acceptable. 1.5 is acceptable. To control the rate of ascent.

[0118] Location updates employ a combination of elite-guided and differential exploration methods: .

[0119] The individual is randomly selected in the current iteration and satisfies... individuals, This is the globally optimal individual in the current iteration. This indicates element-wise multiplication. , For example, a random vector, satisfying . Differential terms are used to promote exploration; elite terms are used to further refine the process. Promote development, and The scheduling achieves a dynamic balance between exploration and development.

[0120] The third improvement involves embedding differential evolution (DE) mutation and crossover as a mutation mechanism to escape local optima and plateau stagnation. PI parameter optimization often results in plateaus or local basins, where population aggregation leads to smaller differential update amplitudes and stagnation. DE constructs new solutions through differential mutation: .

[0121] in, Scaling factor The test vectors are obtained by binomial cross-multiplication, using distinct random indices. , for dimension have: , .in, For crossover probability, It is guaranteed that at least one dimension comes from the mutation vector. A greedy selection rule is adopted: if Then use replace To achieve integration with HOA updates, probabilistic triggering of DE can be used, with the triggering probability decreasing with each iteration: .

[0122] In the early stages, maintain a high trigger probability to enhance jump exploration, and in the later stages, reduce the trigger probability to reduce perturbations and stabilize convergence.

[0123] The enhancement strategies include elite-guided Lévy flight for local search and multi-scenario robustness. If in later iterations... To perform a Lévy perturbation, it can be written as: .

[0124] in, (Usually 1.5) This is the scaling factor. To improve robustness, multi-scenario fitness weighting can be used: The scenario set includes normal steady state, load step, unbalanced load, single-phase ground fault, DC bus disturbance, etc., with weights... Used to highlight the importance of each scenario.

[0125] To ensure stable operation in discrete simulations involving switching devices, it is necessary to clearly define boundary handling and simulation failure penalty mechanisms. Simple truncation... This can easily lead to boundary clustering; therefore, it is recommended to use boundary bounce handling. ; Alternatively, a truncation with small perturbations can be used: , wherein, is a small random perturbation. Simulation exceptions shall be captured during fitness calculation: if a simulation error, state divergence or occurs, a penalty value is returned, for example , so as to ensure that the optimization process is not interrupted and infeasible solutions are automatically eliminated.

[0126] is a truncation operation: if x<L, take L; if x>U, take U; if x is within the interval, keep x unchanged; refers to the occurrence of "Not a Number" (NaN) in the simulation output; refers to an illegal state such as "Infinity".

[0127] The inputs of IHOA for PI parameter tuning include variable boundaries and population size, that is , maximum iteration , simulation duration , fitness calculation method ( or comprehensive ), chaos and OBL parameters, adaptive scheduling parameters, and DE parameters ( ). The outputs include the optimal solution, that is , optimal fitness and convergence curve . is the optimal proportional parameter; is the optimal integral parameter.

[0128] The present application takes into account both the optimization performance of the algorithm and the characteristics of the actual controlled object, can realize intelligent optimization of inverter control parameters without relying on complex analytical modeling and repeated manual trial and error, has good versatility, scalability and popularization and application value, and can provide effective technical support for the control optimization of inverter-type distributed resources in the aggregator scenario.

[0129] The beneficial effects of the present application are as follows: Firstly, aiming at the problems of complex operating conditions, frequent load changes and high control performance requirements of distributed inverter power supplies under the aggregator management scenario, the present application proposes an inverter voltage control parameter optimization scheme based on an improved hippopotamus optimization algorithm, which can automatically optimize the parameters of the voltage closed-loop PI controller of a three-phase off-grid inverter, reduce the dependence on the manual empirical tuning method, and improve the parameter tuning efficiency and engineering applicability.

[0130] Second, this application introduces an improved hippo optimization algorithm into the inverter PI parameter optimization process. By using chaotic mapping initialization and opposition learning strategies to enhance the diversity and coverage of the initial population, it is beneficial to improve the global search effect in the early stage of optimization and reduce the risk of premature convergence caused by uneven distribution of the initial population, thereby improving the stability and reliability of the parameter optimization results.

[0131] Third, by setting an adaptive step size or weight scheduling mechanism, this application enables the optimization process to have a strong global exploration capability in the early stage and a good local development capability in the later stage, thereby improving the problem of imbalance between exploration and development in the search process of traditional intelligent optimization algorithms, and helping to improve the convergence speed and optimization accuracy of PI parameter tuning.

[0132] In one exemplary embodiment, such as Figure 7 As shown, an inverter voltage control parameter optimization device is provided, comprising: The discrete simulation processing module is used to construct the physical model corresponding to the three-phase off-grid inverter power supply system and perform discrete simulation processing. Specifically, it converts the three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system into per-unit values ​​and extracts the fundamental effective value, and compares them with preset reference values ​​to obtain the voltage error.

[0133] The iterative optimization module is used to iteratively optimize the optimization variables of the PI controller within a preset time window using an improved Hippo optimization algorithm based on the fitness function, so as to output the optimal control parameters. The optimal control parameters are used for inverter voltage control. The optimization variables include proportional parameters and integral parameters. The fitness function is determined based on the voltage error of the three-phase phase voltage. Discrete simulation is performed once in each iteration.

[0134] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 8As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores inverter voltage control parameter optimization data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the inverter voltage control parameter optimization method.

[0135] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0136] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0137] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0138] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0139] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0140] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0141] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logic devices, etc., and are not limited to these.

[0142] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0143] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for optimizing inverter voltage control parameters, characterized in that, This method is applied to aggregator-managed resource supply scenarios in three-phase off-grid inverter power supply systems centered on photovoltaics and energy storage; the methods include: A physical model of a three-phase off-grid inverter power supply system is constructed and discrete simulation is performed. Specifically, the three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system are converted into per-unit values ​​and the fundamental effective value is extracted. The voltage error is obtained by comparing the voltage with a preset reference value. An improved Hippo optimization algorithm is adopted to iteratively optimize the PI controller's optimization variables within a preset time window based on the fitness function, so as to output the optimal control parameters. The optimal control parameters are used for inverter voltage control. The optimization variables include proportional parameters and integral parameters. The fitness function is determined based on the voltage error of the three-phase phase voltage. Discrete simulation is performed once in each iteration.

2. The inverter voltage control parameter optimization method according to claim 1, characterized in that, The method further includes: The PI control law based on the PI controller performs amplitude limiting modulation within the inverter voltage closed-loop control structure according to the corresponding input voltage error, and outputs the modulation ratio formed after amplitude limiting; the PI control law is determined based on the optimization variables of the PI controller. The modulation ratio is multiplied by a unit sine wave to form a modulation wave, which is then compared with a triangular carrier wave to generate an SPWM pulse to control the inverter's on and off states.

3. The inverter voltage control parameter optimization method according to claim 1, characterized in that, The improved strategies introduced in the improved hippo optimization algorithm include: chaotic mapping initialization and opposition learning, adaptive step size / weight scheduling, and embedded differential evolution mutation and crossover.

4. The inverter voltage control parameter optimization method according to claim 3, characterized in that, An improved hippo optimization algorithm is employed to iteratively optimize the PI controller's variables within a preset time window based on a fitness function, outputting the optimal control parameters, including: Chaotic mapping initialization and opposition learning are used to map chaotic variables to the boundary of decision variables. Opposition learning is used to construct opposition points for each individual to simultaneously evaluate the original and opposition candidates during the initialization phase, forming multiple solutions and selecting the one with the best fitness to enter the initial population; the chaotic variables include optimization variables. Based on the adaptive step size / weight scheduling, a smooth transition from exploration to development is achieved according to the initial population. Based on the aforementioned embedded differential evolution mutation and crossover mutation mechanism to escape local optima and platform stagnation in position updates, this addresses the problem of stagnation caused by the reduced differential update amplitude after population aggregation in the platform target surface or local basin that appears during parameter optimization. Specifically, new solutions are constructed through differential evolution mutation and crossover based on the embedded differential evolution mutation, and greedy selection rules and probability triggering are used to enhance jump exploration; the enhancement strategies corresponding to the enhanced jump exploration include elite-guided Levy flight local search and multi-scenario robust fitness. The optimal control parameters are determined based on the position update information obtained after the position update and the corresponding optimization variables.

5. The inverter voltage control parameter optimization method according to claim 4, characterized in that, Based on the aforementioned embedded differential evolution mutation and crossover, new solutions are constructed through differential mutation, and greedy selection rules and probability triggering are employed to enhance jump exploration, specifically including: New solutions are constructed through differential mutation, and the new solutions are subjected to binomial crossover to obtain the test vector. Greedy selection rules and probability triggering are used to enhance jump exploration.

6. The inverter voltage control parameter optimization method according to claim 1, characterized in that, The inverter voltage control parameter optimization method further includes: A simulation failure penalty mechanism is introduced, employing either bounce boundary handling or truncation with small perturbations, to ensure that the optimization process is not interrupted and infeasible solutions are automatically eliminated, thus guaranteeing the stable operation of discrete simulation.

7. An inverter voltage control parameter optimization device, characterized in that, include: The discrete simulation processing module is used to construct the physical model corresponding to the three-phase off-grid inverter power supply system and perform discrete simulation processing; in this module, the three-phase phase voltages at the load end of the three-phase off-grid inverter power supply system are converted into per-unit values ​​and the fundamental effective value is extracted, and the voltage error is obtained by comparing it with the preset reference value. The iterative optimization module is used to iteratively optimize the optimization variables of the PI controller within a preset time window using an improved Hippo optimization algorithm based on the fitness function, so as to output the optimal control parameters. The optimal control parameters are used for inverter voltage control. The optimization variables include proportional parameters and integral parameters. The fitness function is determined based on the voltage error of the three-phase phase voltage. Discrete simulation is performed once in each iteration.

8. A computer device, comprising: The memory and processor contain a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the inverter voltage control parameter optimization method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the inverter voltage control parameter optimization method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the inverter voltage control parameter optimization method according to any one of claims 1-6.

Images