A method, device, medium and equipment for parameter identification of photovoltaic system

Abstract

The present invention discloses a parameter identification method, device, medium and equipment for a photovoltaic system, and relates to the technical field of photovoltaic system parameter identification. The present invention uses Gaussian process regression to establish a highly simulated photovoltaic system proxy model, which improves the accuracy of the evaluation while reducing the overhead. In the optimization process of parameter identification, the rime optimization algorithm and distance-based adaptive search are combined to search for better individuals. Specifically, in order to cope with multi-dimensional parameter identification, the rime optimization algorithm can simulate the growth process of rime to freely explore the global parameter space and take into account local exploration, optimize each individual in the search population of the identified parameters, and optimize the optimal individual through distance-based adaptive search, thereby improving the search capability of the rime optimization algorithm, and using mutation-assisted ontology search to jump out of the local optimum. The present invention improves the adaptability of photovoltaic system parameter identification to dynamic changes in the environment, avoiding falling into the local optimum or generating oscillations.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic system parameter identification, and in particular to a photovoltaic system parameter identification method, device, medium and equipment. Background Art

[0002] As a clean, efficient, and renewable energy source, solar energy is gaining increasing attention and application worldwide. Solar photovoltaic systems convert solar energy directly into electricity through photoelectric conversion technology, producing no pollutants. As the core component of the system, the conversion efficiency and service life of solar cells directly determine the performance of the entire photovoltaic system.

[0003] To improve the power prediction accuracy, fault diagnosis capabilities, and operational stability of photovoltaic systems, mathematical modeling and optimization algorithms are often used to determine photovoltaic system parameters. In the field of photovoltaic system parameter identification, key parameters such as photocurrent, reverse saturation current, ideality factor, series resistance, and shunt resistance must be accurately identified. These parameters directly affect the accuracy of fitting the solar cell's output characteristic curve and the accuracy of predicting the system's power generation efficiency.

[0004] In the existing technology, some studies have used swarm intelligence algorithms to identify photovoltaic system parameters, especially the agent-assisted evolutionary algorithm that has developed rapidly in recent years. By constructing efficient agent models such as Gaussian process regression and support vector regression, it can significantly save computing costs while maintaining high optimization accuracy.

[0005] However, existing parameter identification methods still face key challenges. The photovoltaic system parameter optimization process must simultaneously address the conflicting requirements of local precision search and global rapid convergence. In practical applications, the IV characteristic curve of solar cells has multiple extreme value regions and drifts with changing environmental conditions. Existing parameter identification methods lack adaptability to dynamic environmental changes and are prone to falling into local optimality or oscillation. This makes it impossible to obtain stable and accurate parameter identification results under variable operating conditions and periods of rapid changes in light intensity. Summary of the Invention

[0006] Based on this, it is necessary to provide a parameter identification method, device, medium and equipment for a photovoltaic system to address the above technical problems.

[0007] The present invention adopts the following technical solutions:

[0008] The present invention provides a parameter identification method for a photovoltaic system, comprising:

[0009] Obtaining the measured current and voltage curve of the photovoltaic system's current operating condition to be predicted;

[0010] Generate multiple sets of parameters to be identified in the parameter space of each parameter to be identified of the photovoltaic system, and construct a proxy model of the photovoltaic system based on each set of parameters to be identified and their corresponding real current-voltage curves;

[0011] Initialize the individuals in the search population using each set of parameters to be identified, predict the predicted current-voltage curve of each individual in the search population using the proxy model, and determine the current predicted fitness value of each individual based on the deviation between the measured current-voltage curve and each predicted current-voltage curve;

[0012] Optimize each individual using the rime optimization algorithm based on the current predicted fitness value of each individual; determine the current optimal individual in the search population, determine the adaptive update step size based on the distance between the current optimal individual and each current individual, and update the current optimal individual;

[0013] If the updated current optimal individual is better than the current optimal individual, the current optimal individual is replaced with the updated current optimal individual; otherwise, the updated current optimal individual is mutated, and the mutated result is added to the search population for the next round of search optimization. After multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current predicted working conditions.

[0014] The present invention provides a parameter identification device for a photovoltaic system, comprising:

[0015] An acquisition module is used to obtain the measured current and voltage curve of the photovoltaic system's current working condition to be predicted;

[0016] A fitting module is used to generate multiple groups of parameters to be identified in the parameter space of each parameter to be identified of the photovoltaic system, and to construct a proxy model of the photovoltaic system based on each group of parameters to be identified and their corresponding real current-voltage curves;

[0017] An initialization module is used to initialize individuals in the search population using each set of parameters to be identified, predict the predicted current-voltage curve of each individual in the search population using a proxy model, and determine the current predicted fitness value of each individual based on the deviation between the measured current-voltage curve and each predicted current-voltage curve;

[0018] The optimization module is used to optimize each individual through the rime optimization algorithm according to the current predicted fitness value of each individual; determine the current best individual in the search population, determine the adaptive update step size based on the distance between the current best individual and the current individuals, and update the current best individual;

[0019] The iterative module is used to replace the current optimal individual with the updated current optimal individual if the updated current optimal individual is better than the current optimal individual; otherwise, the updated current optimal individual is mutated and the mutated result is added to the search population for the next round of search optimization. After multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current working condition to be predicted.

[0020] The present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the parameter identification method of the photovoltaic system is implemented.

[0021] The present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the parameter identification method of the photovoltaic system is implemented.

[0022] At least one of the above technical solutions adopted by the present invention can achieve the following beneficial effects:

[0023] The present invention combines the rime optimization algorithm and distance-based adaptive search to search for better individuals in the optimization process of photovoltaic system parameter identification. Specifically, in order to cope with multi-dimensional parameter identification, the rime optimization algorithm can simulate the growth process of rime to freely explore the global parameter space while taking into account local exploration, and optimize each individual in the search population of the parameters to be identified. At the same time, the optimization process of the rime optimization algorithm has a high correlation with the optimal individual. Therefore, the present invention further proposes to search and optimize the optimal individual based on the distance between the optimal individual and each individual, further improving the optimization search capability of the parameters to be identified. When the updated optimal individual is inferior to the optimal individual before the update, the optimization of the parameters to be identified may fall into a local optimum. To this end, the present invention uses mutation search to assist in jumping out of the local optimum and attempting to find a better solution in the solution space. The optimization strategy for photovoltaic system parameter identification of the present invention avoids falling into a local optimum or generating oscillations, improves adaptability to dynamic environmental changes, and can obtain stable parameter identification results even during periods of rapid changes in light intensity under variable working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0025] Figure 1 A schematic flow chart of a parameter identification method for a photovoltaic system provided by the present invention;

[0026] Figure 2 A schematic diagram of a rime algorithm search provided by the present invention;

[0027] Figure 3 A schematic diagram of the algorithm flow of an agent-assisted enhanced rime optimization algorithm provided by the present invention;

[0028] Figure 4 An equivalent circuit diagram of a photovoltaic model provided by the present invention;

[0029] Figure 5 A schematic diagram of the average value of RMSE of experimental results of various algorithms provided by the present invention on a single diode model;

[0030] Figure 6 A schematic diagram showing the average value of RMSE of experimental results of various algorithms provided by the present invention on a dual-diode model;

[0031] Figure 7 A schematic diagram of the average value of RMSE of the experimental results of various algorithms provided by the present invention on a three-diode model;

[0032] Figure 8 This is a schematic diagram of a parameter identification device for a photovoltaic system provided by the present invention. DETAILED DESCRIPTION

[0033] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and corresponding drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] Currently, solar photovoltaic systems are subject to multiple influences during actual operation, including light intensity, temperature fluctuations, and component aging. Complex coupling relationships exist between components, resulting in highly nonlinear input-output characteristics. Existing proxy models are unable to adequately fit these complex nonlinear relationships, leading to large parameter identification errors under variable operating conditions. For example, when encountering intermittent light conditions caused by rainy weather, models built using traditional methods often fail to accurately reflect the dynamic changes in cell characteristics. Furthermore, the photovoltaic system parameter optimization process must simultaneously address the conflicting requirements of local precision search and global rapid convergence. In practical applications, the IV characteristic curve of solar cells exhibits multiple extreme value regions and drifts with changing environmental conditions. Existing parameter identification methods lack adaptability to dynamic environmental changes and are prone to falling into local optima or oscillations. This makes it difficult to obtain stable and accurate parameter identification results under variable operating conditions, especially during periods of rapid light intensity fluctuations.

[0035] Researchers have proposed a variety of improvement schemes, including hybrid agent models, adaptive parameter adjustment, and parallel computing architectures. Taking the Rime algorithm as an example, although it performs well on standard test functions, it still suffers from significant drawbacks in actual engineering applications, such as insufficient population diversity and decreased accuracy in high-dimensional optimization. Therefore, developing a new enhanced agent-assisted optimization algorithm has important theoretical value and engineering significance. An ideal algorithm should possess the following key features: 1. The ability to construct a high-precision agent model that accurately captures the complex characteristics of the objective function; 2. An intelligent exploration-exploitation balance mechanism that can adaptively adjust the search strategy according to the different stages of the optimization process; and 3. Greater adaptability to dynamic environments.

[0036] Such an algorithm will significantly improve the efficiency of solving complex optimization problems, provide more reliable technical support for photovoltaic system engineering applications, and promote the further development of clean energy technology.

[0037] The technical solutions provided by various embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0038] Figure 1 The following is a flow chart of a parameter identification method for a photovoltaic system according to the present invention, which specifically includes the following steps:

[0039] S101: Obtaining a measured current and voltage curve of a photovoltaic system under a current working condition to be predicted.

[0040] S102: generating multiple groups of parameters to be identified in the parameter space of each parameter to be identified of the photovoltaic system, and constructing a proxy model of the photovoltaic system according to each group of parameters to be identified and their corresponding real current-voltage curves.

[0041] S103: Initialize individuals in the search population using each set of parameters to be identified, predict the predicted current-voltage curve of each individual in the search population using the proxy model, and determine the current predicted fitness value of each individual based on the deviation between the measured current-voltage curve and each predicted current-voltage curve.

[0042] S104: Optimize each individual using the rime optimization algorithm according to the current predicted fitness value of each individual; determine the current optimal individual in the search population, determine the adaptive update step size according to the distance between the current optimal individual and the current individuals, and update the current optimal individual.

[0043] S105: If the updated current optimal individual is better than the current optimal individual, the current optimal individual is replaced with the updated current optimal individual; otherwise, the updated current optimal individual is mutated, and the mutated result is added to the search population for the next round of search optimization. After multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current working condition to be predicted.

[0044] For the sake of convenience, the following description will only be based on the server as the execution subject. The server mentioned in the present invention can be a server set up on a business platform, or a device such as a desktop computer or a laptop computer that can execute the solution of the present invention.

[0045] The parameter identification method of the photovoltaic system proposed in the present invention realizes efficient identification of photovoltaic system parameters by integrating the proxy model and intelligent optimization technology.

[0046] First, an initial sample set covering the critical parameter space can be generated, and a surrogate model can be constructed using Gaussian process regression to rapidly fit the nonlinear IV characteristics of photovoltaic systems in complex environments (such as cloudy conditions and temperature fluctuations). During the parameter identification optimization process, the surrogate model can dynamically guide the search for the parameters to be evaluated. The specific optimization can be achieved by combining the rime optimization algorithm and distance-based adaptive search. The rime algorithm primarily optimizes all individuals in the search population, while the distance-based adaptive search optimizes the best individual in the search population. When the optimized best individual outperforms the unoptimized one, a true IV curve measurement of the optimized best individual is triggered, i.e., true fitness value verification. If verification passes, the best individual is updated; otherwise, mutation is used to escape the local optimum (e.g., parameter traps caused by component aging).

[0047] The parameter identification process of this invention continuously adds new variants to the search population, combining adaptive step-size adjustment to achieve parameter identification tracking optimization, ultimately outputting the optimal parameter set that minimizes the root mean square error (RMSE) between the measured and model IV curves. This closed-loop process, combining "agent-accelerated evaluation - intelligent dynamic search - online verification and update," reduces computational costs while ensuring accuracy. This effectively addresses the engineering challenges faced by traditional methods in dynamic environments, such as slow convergence and the tendency to fall into local optimality.

[0048] Generally, when performing parameter identification of a photovoltaic system, the measured current and voltage curve of the photovoltaic system's current working condition to be predicted can be obtained first. The measured current and voltage curve can be used as a basis for judging the true adaptability during the parameter identification and optimization process.

[0049] Subsequently, in one or more embodiments of the present invention, parameter ranges for each unidentified parameter of the photovoltaic system can be obtained, and multiple initial samples can be generated within the parameter space corresponding to the parameter ranges of each unidentified parameter. Each initial sample constitutes a set of unidentified parameters. The unidentified parameters may include: photocurrent, series resistance, output current, reverse saturation current, shunt resistance, etc. Based on these initial samples, a proxy model of the photovoltaic system can be constructed, thereby reducing the complexity of fitness assessment. Furthermore, these initial samples can be used to form a search population.

[0050] Specifically, in one or more embodiments of the present invention, to accurately capture the nonlinear variations in battery characteristics under different environmental conditions (e.g., clear sky, cloudy sky, and shadow), the server may divide the parameter ranges of the various dimensions of the photovoltaic system into pre-defined intervals. Using Latin hypercube sampling, the server generates multiple sets of parameters to be identified within the parameter space corresponding to the parameter ranges of each parameter to be identified, thereby covering the critical parameter space. For example, an initial sample set DB of N = 100 may be generated to ensure uniform coverage of the parameter ranges of the parameters to be identified.

[0051] Based on the generated initial samples, the true current-voltage curve of the photovoltaic system corresponding to each initial sample can be further determined, and then the initial samples and the corresponding true current-voltage curves can be fitted through Gaussian process regression to construct a proxy model of the photovoltaic system. Subsequently, the proxy model can be used to quickly fit the nonlinear IV characteristics of the photovoltaic system in complex environments (such as cloudy and temperature changes).

[0052] Specifically, the real IV curve can be measured for each sample point, or a candidate solution set can be constructed through numerical calculation. Among them, the candidate solution X The definition is: each line is a photovoltaic cell parameter combination, dimension ( The corresponding number of single, double, and triple diode model parameters. X The corresponding output y Satisfies formula 1: .

[0053] in, is an unknown continuous function, which represents the mapping relationship from each parameter to be identified to the real IV characteristic. Here it can be considered as It is the mapping of each parameter to be identified to IV characteristics to be learned by Gaussian process regression. is a system with a mean of zero and a variance of In Gaussian process regression, it is assumed that is a random process that defines the input features Therefore, we can The prior distribution of is modeled to derive the posterior distribution, thereby predicting the fitness value.

[0054] In one or more embodiments of the present invention, the Matérn function can be used as the kernel function of Gaussian process regression to fit each set of parameters to be identified and the corresponding real current-voltage curve to obtain a proxy model of the photovoltaic system. This function has good fitting capabilities for two typical nonlinear relationships in photovoltaic systems, including continuous non-smooth characteristics (such as the inflection point of the curve caused by temperature mutation) and the length scale parameter. Adaptive adjustment of local sensitivity ( The smaller it is, the more sensitive it is to the mutation response). It is specifically expressed by the following formula:

[0055] .

[0056] in, represents the distance between input data points, is a positive real number called the length scale parameter, is the modified Bessel function. Is a non-negative real number, the present invention can take the value Therefore, the following formula can be obtained from the above formula:

[0057] .

[0058] After constructing the proxy model of the photovoltaic system, a search population can be further formed based on the initial sample. Individuals in the search population can be initialized directly using the initial sample, and certain screening can also be performed to form a search population based on the elite individuals in the initial sample. That is, in one or more embodiments of the present invention, the proxy model can be used to predict the predicted current-voltage curve of each individual in the search population. Based on the deviation between the measured current-voltage curve and each predicted current-voltage curve, the current predicted fitness value of each individual is determined, and the groups of parameters to be identified are sorted in descending order according to their predicted fitness values; thereby, according to the preset sample selection coefficient, selection is performed from the sorted groups of parameters to be identified from front to back, and individuals in the search population are initialized using the selected groups of parameters to be identified. Through the strategy of evolutionary sampling, the initial sample is transformed into an elite group to obtain a search population with a better initial condition for the parameters to be identified, thereby improving the convergence efficiency of the search for the parameters to be identified.

[0059] After obtaining the search population, the server can first optimize each individual through the rime optimization algorithm according to the current predicted fitness value of each individual. Figure 2 This is a schematic diagram of a rime algorithm search in the present invention. Figure 3 This is a schematic diagram of the algorithm flow of an agent-assisted enhanced rime optimization algorithm in the present invention. Figure 3 middle, represents the current optimal individual after update, represents the current optimal individual, Indicates the current worst individual.

[0060] The rime optimization algorithm is divided into three key elements: a soft rime search strategy, a hard rime penetration mechanism, and a forward greedy selection mechanism. Soft rime exhibits strong randomness during growth, with particles freely covering most surfaces but growing slowly in the same direction. The algorithm simulates the randomness and coverage characteristics of soft rime by guiding a specific update formula. The soft rime search strategy is shown below:

[0061] ;

[0062] ; ; .

[0063] As the number of iterations increases, θ 、 β and E will gradually change. t represents the current iteration number, and T It represents the maximum number of iterations allowed. w Always equal to 5. Indicates the i The number of mist particles (corresponding to individuals) j The updated position of each fog particle (corresponding to the individual parameters to be identified), and It means the optimal mist particle j A fog particle. is a random number in the range (-1,1), and h is also a random number, and its value range is (0,1). Indicates the i The mist particles are j The upper limit of fog particles, Indicates the i The mist particles are j The lower limit of fog particles.

[0064] Compared to soft rime, which has a clearer and more regular growth pattern, when rime particles gradually condense into hard rime, their consistent growth direction allows various rime agents to easily penetrate it. This phenomenon is known as hard rime penetration. To simulate this phenomenon, the rime optimization algorithm proposes a hard rime penetration mechanism, aiming to further expand the solution space and enhance the algorithm's local exploration capabilities. The specific update formula for rime particles in the hard rime penetration mechanism is shown below:

[0065] .

[0066] Where, is a random number whose value range is between (-1,1). It represents the current mist particles Normalized value of the predicted fitness value.

[0067] The forward greedy selection mechanism operates by comparing the change in particle fitness before and after an update, an improvement on the traditional greedy selection method. If the updated value is better, it replaces the previous value and updates the solutions of both particles simultaneously. Furthermore, it uses the best fitness value and corresponding solution to update the global best solution. This mechanism ensures that the population continues to move towards the global optimum with each iteration. Specifically, see the following formula:

[0068] ;

[0069] .

[0070] Where, For the i The predicted fitness value corresponding to the position of each fog particle, For the i The predicted fitness value corresponding to the updated position of each fog particle, For the i The position of the mist particles, For the i The updated position of the fog particles, is the optimal position of mist particles, is the predicted fitness value corresponding to the position of the optimal mist particle.

[0071] After updating the search population using the Rime Optimization Algorithm, the optimal individual in the population can be further optimized using a distance-based adaptive search. This search continuously attempts to find better solutions in the parameter space of photovoltaic cells in real environments (hereinafter referred to as the solution space) and optimizes the current optimal individual by dynamically adjusting the step size and search direction.

[0072] Specifically, in one or more embodiments of the present invention, the server may first determine an initial update step length based on the average of the distances between the current optimal individual and each current individual. The server then updates the current optimal individual based on the initial update step length and updates the initial update step length using the following formula:

[0073] , ;

[0074] , , ;

[0075] in, is the current optimal individual, is the current optimal individual after update, is the adaptive update step size, For the j The initial update step size of the parameters to be identified in dimensions, is the direction vector matrix j List, is the predicted fitness value of the current optimal individual after update, is the predicted fitness value of the current optimal individual, is the update step size obtained after updating the initial update step size. It can be considered as the mapping relationship learned by the above-mentioned proxy model, that is, the predicted fitness values ​​before and after the update are determined by the proxy model.

[0076] The values ​​of Alfa and Beta directly affect the rate of change of the step size. Choosing a larger Alfa can quickly increase the step size, helping the algorithm to quickly expand the search space during the exploration phase. Choosing a negative Beta, on the other hand, can reduce the step size when needed, helping to refine the search when approaching the optimal solution. Therefore, in this paper, Alfa is set to 20 and Beta is set to -0.5.

[0077] This strategy can evaluate the current best individual through the above proxy model to obtain the predicted fitness value of the current best individual. It can also determine the predicted fitness values ​​of all current individuals through the above proxy model and sort them. According to the average distance between all sorted individuals and the current best individual, the step size of the parameter to be identified in each dimension is calculated. In each iteration, according to the current best individual X and step length h Get the updated current optimal individual . The update step size is dynamically adjusted according to the changes in the objective function. If a better solution is found in a certain iteration, the current optimal individual is replaced:

[0078] .

[0079] If the updated value of the predicted fitness value of the current optimal individual is small enough, it can be considered that the algorithm has converged to the optimal solution. The minimum value can be defined by the following formula:

[0080] .

[0081] During this search, the strategy will re-search the solution space multiple times, starting from different parameter combinations based on the current optimal individual. Each re-search provides a new starting point, preventing the strategy from falling into a local optimum and increasing the likelihood of finding the optimal solution.

[0082] After the adaptive update step size is used to update the current optimal individual, the optimization direction can be further adjusted based on the comparison between the updated current optimal individual and the one before the update.

[0083] In one or more embodiments of the present invention, as described above, the server can determine the predicted fitness value of the updated current optimal individual using a proxy model of the photovoltaic system based on the measured current and voltage curves of the current operating condition to be predicted. If the updated optimal individual is better, the current optimal individual is replaced with the updated current optimal individual. Otherwise, the updated current optimal individual can be mutated and the mutated result added to the search population for the next round of search optimization. Thus, after multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current operating condition to be predicted.

[0084] In order to more accurately evaluate and obtain excellent individuals, in one or more embodiments of the present invention, if the predicted fitness value of the updated current optimal individual is better than the current optimal individual, the updated current optimal individual and the current optimal individual can be further measured through the mathematical model of the photovoltaic system to obtain the measured current-voltage curves corresponding to the updated current optimal individual and the current optimal individual respectively; according to the measured current-voltage curve of the current working condition to be predicted and the measured current-voltage curves corresponding to the updated current optimal individual and the current optimal individual respectively, the true fitness values ​​corresponding to the updated current optimal individual and the current optimal individual respectively are determined; if the true fitness value of the updated current optimal individual is better than the current optimal individual, the current optimal individual is replaced with the updated current optimal individual.

[0085] Furthermore, in one or more embodiments of the present invention, if the true predicted fitness value of the updated current optimal individual is worse than that of the current optimal individual, the true fitness value of the current worst individual can be determined by measuring the mathematical model of the photovoltaic system; if the true fitness value of the updated current optimal individual is better than that of the current worst individual, the current worst individual is replaced with the updated current optimal individual; and the updated current optimal individual is mutated; if the predicted fitness value of the updated current optimal individual is worse than that of the current optimal individual, the true fitness value evaluation is no longer performed, and the updated current optimal individual is mutated, that is, the mutation-assisted local exploration strategy is used to explore potential candidate solutions.

[0086] Mutation-assisted local search aims to find a better solution in the solution space. The algorithm uses Cauchy and Gaussian mutations to generate new candidate solutions, and can also update the solution in real time according to its fitness value. This method combines the advantages of increased diversity, assisted local search, and dynamic solution update. Cauchy mutation introduces larger changes, which helps to escape the local optimal solution, while Gaussian mutation can be fine-tuned in the local area. The strategy of dynamically updating the solution is based on the fitness value of the candidate solution, so that the better solution can be retained and utilized, thereby accelerating the convergence process. In addition, the evaluation and ranking based on the agent model can effectively reduce the computational cost of the algorithm and improve the search efficiency. Among them, the updated current optimal individual can be mutated according to the Cauchy mutation through the following formula:

[0087] .

[0088] in, are the parameters to be identified of the updated current optimal individual, is the central location parameter of the Cauchy distribution, is the scale parameter of the Cauchy distribution.

[0089] Alternatively, the updated current optimal individual can be mutated according to Gaussian mutation using the following formula:

[0090] .

[0091] in, is the mean of the Gaussian distribution, is the standard deviation of the Gaussian distribution.

[0092] The specific mutation to be selected can be determined by temporarily generating a threshold range within which the random number falls. Cauchy distribution and Gaussian distribution are used to generate random numbers for mutation operations and to generate new candidate solutions.

[0093] Based on multiple rounds of iterative optimization, the search ends when the preset maximum number of runs of the algorithm is reached or the value of the current optimal individual is small enough (less than the preset threshold).

[0094] Finally, the current optimal individual at the end of the search is used as the parameters to be identified for the photovoltaic system under the current working condition to be predicted and output.

[0095] For the mathematical model of photovoltaic system, the solar photovoltaic model system consists of single diode model, double diode model and triple diode model. Figure 4 This is an equivalent circuit diagram of a photovoltaic model in the present invention, where (a), (b), and (c) correspond to the circuit diagrams of a single diode, a double diode, and a triple diode, respectively. Indicates the flow through the parallel resistor The shunt current is 100%, and the other specific parameters are explained in the following formula. A single diode typically consists of a power supply in parallel with the diode, a parallel resistor to represent the leakage current, and a series resistor to compensate for the load current loss. The formula for a single diode is as follows:

[0096] .

[0097] in, represents the output current, represents the photogenerated current, is the reverse saturation current in the single diode model, represents the parallel resistance, represents the series resistance, is the output voltage, represents the ideality factor of the diode, represents the Boltzmann constant ( ), represents the Kelvin cell temperature ( ).

[0098] The double diode increases the number of diodes from one to two. The formula is as follows:

[0099] .

[0100] Where, represents the current of the first diode, represents the current of the second diode. represents the diode diffusion factor, represents the diode recombination factor.

[0101] The formula for three diodes is shown below:

[0102] .

[0103] Where, represents the current of the first diode, represents the current of the second diode, represents the current of the third diode. represents the diode diffusion factor, represents the diode recombination factor, represents the third diode ideality factor.

[0104] The core of the optimization problem is to transform the actual problem into a mathematical model and solve the optimal parameters by minimizing the objective function. In the present invention, when the predicted fitness value is determined by the proxy model or the real fitness value is determined by the mathematical model, the root mean square error (RMSE) between the measured data (the measured current-voltage curve) and the model calculated value (the current-voltage curve determined by the proxy model or the mathematical model) can be used as the objective function. Since this function is a nonlinear transcendental function, its solution is challenging - this type of function does not follow the conventional rules of algebra and calculus, and often requires special optimization methods. Therefore, the key to the present invention is to search for the optimal parameter vector that minimizes the objective function. X , that is, the parameters to be identified.

[0105] .

[0106] Where, is the total number of measurement data points, For measurement data, Calculate values ​​for the model, For the i The error function for each data point.

[0107] Figure 5 Schematic diagram of the average value of RMSE of experimental results of various algorithms in the present invention on a single diode model, Figure 5 "F1" means Figure 5 For the single-diode model, we can see that SRIME has better results than other algorithms and is significantly improved compared to RIME (SRIME represents the improved method of the present invention, while SA-ADEA, CALSAPSO, MS-MOEA, SAEA-RFSS, and RIME are all proxy-assisted model optimization algorithms proposed by other studies). Figure 6 Schematic diagram of the average value of RMSE of the experimental results of each algorithm in the present invention on the dual-diode model, Figure 6 "F2" means Figure 6 For the dual-diode model, we can see that SRIME has better results than other algorithms and is significantly improved compared to RIME. Figure 7 Schematic diagram of the average value of RMSE of the experimental results of each algorithm in the present invention on the three-diode model, Figure 7 "F3" means Figure 7 Corresponding to the three-diode model, we can see that SRIME has better results than other algorithms and is greatly improved compared to RIME.

[0108] based on Figure 1The photovoltaic system parameter identification method shown in the figure first employs an evolutionary sampling strategy to perform an elitist transformation on the initial population. Secondly, the method uses a Gaussian process regression model to establish a highly realistic proxy model. This highly realistic proxy model predicts and evaluates the fitness of the parameter identification optimization process, improving the accuracy of the evaluation while reducing overhead, thereby enhancing the efficiency of population updates. Finally, the present invention combines the rime optimization algorithm and distance-based adaptive search to search for better individuals in the optimization process of photovoltaic system parameter identification. Specifically, in order to cope with multi-dimensional parameter identification, the rime optimization algorithm can simulate the growth process of rime to freely explore the global parameter space while taking into account local exploration, and optimize each individual in the search population of the parameters to be identified. At the same time, the optimization process of the rime optimization algorithm has a high correlation with the optimal individual. Therefore, the present invention further proposes to search and optimize the optimal individual based on the distance between the optimal individual and each individual, further improving the optimization search capability of the parameters to be identified. When the updated optimal individual is inferior to the optimal individual before the update, the optimization of the parameters to be identified may fall into a local optimum. To this end, the present invention uses mutation search to assist in jumping out of the local optimum and attempting to find a better solution in the solution space. The optimization strategy for photovoltaic system parameter identification of the present invention avoids falling into a local optimum or generating oscillations, improves adaptability to dynamic environmental changes, and can obtain stable parameter identification results even during periods of rapid changes in light intensity under variable working conditions.

[0109] When applying the photovoltaic system parameter identification method provided by the present invention, it is not necessary to Figure 1 The steps are executed in the order shown. The specific execution order of the steps can be determined according to needs, and the present invention does not limit this.

[0110] The above is a parameter identification method for a photovoltaic system provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding parameter identification device for a photovoltaic system, such as Figure 8 shown.

[0111] Figure 8 A schematic diagram of a parameter identification device for a photovoltaic system provided by the present invention, comprising:

[0112] An acquisition module 201 is used to obtain a measured current and voltage curve of the photovoltaic system's current working condition to be predicted;

[0113] A fitting module 202 is configured to generate multiple sets of parameters to be identified within the parameter space of each parameter to be identified of the photovoltaic system, and to construct a proxy model of the photovoltaic system based on each set of parameters to be identified and their corresponding real current-voltage curves;

[0114] Initialization module 203, configured to initialize individuals in the search population using each set of parameters to be identified, predict the predicted current-voltage curve of each individual in the search population using the proxy model, and determine the current predicted fitness value of each individual based on the deviation between the measured current-voltage curve and each predicted current-voltage curve;

[0115] The optimization module 204 is used to optimize each individual using the rime optimization algorithm according to the current predicted fitness value of each individual; determine the current best individual in the search population, determine the adaptive update step size based on the distance between the current best individual and each current individual, and update the current best individual;

[0116] The iterative module 205 is used to replace the current optimal individual with the updated current optimal individual if the updated current optimal individual is better than the current optimal individual; otherwise, the updated current optimal individual is mutated and the mutated result is added to the search population for the next round of search optimization. After multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current working condition to be predicted.

[0117] The specific definitions of the photovoltaic system parameter identification device can be found in the definitions of the photovoltaic system parameter identification method described above and will not be repeated here. Each module in the photovoltaic system parameter identification device described above can be implemented in whole or in part through software, hardware, or a combination thereof. Each of these modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0118] The present invention also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 A parameter identification method for a photovoltaic system is provided.

[0119] The present invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, it may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 A parameter identification method for a photovoltaic system is provided.

[0120] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes in the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0121] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, 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 the present invention.

Claims

1. A method for identifying parameters of a photovoltaic system, characterized in that: include: Obtaining the measured current and voltage curve of the photovoltaic system's current operating condition to be predicted; Generate multiple sets of parameters to be identified in the parameter space of each parameter to be identified of the photovoltaic system, and construct a proxy model of the photovoltaic system based on each set of parameters to be identified and their corresponding real current-voltage curves; Initialize the individuals in the search population using each set of parameters to be identified, predict the predicted current-voltage curve of each individual in the search population using the proxy model, and determine the current predicted fitness value of each individual based on the deviation between the measured current-voltage curve and each predicted current-voltage curve; Optimize each individual using the rime optimization algorithm based on the current predicted fitness value of each individual; determine the current optimal individual in the search population, determine the adaptive update step size based on the distance between the current optimal individual and each current individual, and update the current optimal individual; If the updated current optimal individual is better than the current optimal individual, the current optimal individual is replaced with the updated current optimal individual; otherwise, the updated current optimal individual is mutated, and the mutated result is added to the search population for the next round of search optimization. After multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current predicted working conditions.

2. The photovoltaic system parameter identification method according to claim 1, characterized in that: The method of determining the adaptive update step size based on the distance between the current optimal individual and each current individual and updating the current optimal individual specifically includes: Determine the initial update step size based on the average value of the distance between the current optimal individual and each current individual; The current optimal individual is updated according to the initial update step size through the following formula, and the initial update step size is updated: ， ； ， ， ； in, is the current optimal individual, is the current optimal individual after update, is the adaptive update step size, For the j The initial update step size of the parameters to be identified in dimensions, is the direction vector matrix j List, is the fitness value of the current optimal individual after update, is the fitness value of the current optimal individual, is the update step size obtained after updating the initial update step size.

3. The photovoltaic system parameter identification method according to claim 1, wherein: If the updated current best individual is better than the current best individual, the current best individual is replaced by the updated current best individual, specifically including: According to the measured current and voltage curves of the current working condition to be predicted, the fitness value of the updated current optimal individual is determined through the proxy model of the photovoltaic system; If the predicted fitness value of the updated current optimal individual is better than that of the current optimal individual, the updated current optimal individual and the current optimal individual are measured using the photovoltaic system mathematical model to obtain the measured current-voltage curves corresponding to the updated current optimal individual and the current optimal individual, respectively; based on the measured current-voltage curve of the current working condition to be predicted and the measured current-voltage curves corresponding to the updated current optimal individual and the current optimal individual, respectively, the true fitness values ​​corresponding to the updated current optimal individual and the current optimal individual are determined; If the true fitness value of the updated current optimal individual is better than that of the current optimal individual, the current optimal individual is replaced by the updated current optimal individual.

4. The photovoltaic system parameter identification method according to claim 3, wherein: Otherwise, the updated current optimal individual is mutated, specifically including: If the updated true fitness value of the current best individual is worse than that of the current best individual, the true fitness value of the current worst individual is determined by measuring the photovoltaic system mathematical model; If the true fitness value of the updated current best individual is better than that of the current worst individual, the current worst individual is replaced by the updated current best individual; and the updated current best individual is mutated; If the predicted fitness value of the updated current optimal individual is worse than that of the current optimal individual, the updated current optimal individual is mutated.

5. The photovoltaic system parameter identification method according to claim 1, wherein: The mutating of the updated current optimal individual specifically includes: The updated current optimal individual is mutated according to the Cauchy mutation using the following formula: ； Alternatively, the updated current optimal individual is mutated according to Gaussian mutation using the following formula: ； in, are the parameters to be identified of the updated current optimal individual, is the central location parameter of the Cauchy distribution, is the scale parameter of the Cauchy distribution, is the mean of the Gaussian distribution, is the standard deviation of the Gaussian distribution.

6. The photovoltaic system parameter identification method according to claim 1, wherein: The generating of multiple groups of parameters to be identified within the parameter space of each parameter to be identified of the photovoltaic system specifically includes: According to the preset interval division of the parameter range of each dimension of the photovoltaic system to be identified, multiple groups of parameters to be identified are generated in the parameter space corresponding to the parameter range of each parameter to be identified by Latin hypercube sampling.

7. The photovoltaic system parameter identification method according to claim 1, wherein: The method of constructing a proxy model of the photovoltaic system based on each set of parameters to be identified and their corresponding real current-voltage curves specifically includes: The Matérn function is used as the kernel function of Gaussian process regression to fit each group of parameters to be identified and their corresponding real current-voltage curves to obtain a proxy model of the photovoltaic system.

8. The photovoltaic system parameter identification method according to claim 1, wherein: The initialization of the search population individuals by using each set of parameters to be identified specifically includes: The predicted current-voltage curve of each individual in the search population is predicted by the surrogate model, and the current predicted fitness value of each individual is determined according to the deviation between the measured current-voltage curve and each predicted current-voltage curve, so as to sort each group of parameters to be identified in descending order according to the predicted fitness value of each group of parameters to be identified; According to the preset sample selection coefficient, the parameters to be identified are selected from the sorted groups from front to back, and the individuals in the search population are initialized by the selected groups of parameters to be identified.

9. A parameter identification device for a photovoltaic system, characterized in that: include: An acquisition module is used to obtain the measured current and voltage curve of the photovoltaic system's current working condition to be predicted; A fitting module is used to generate multiple groups of parameters to be identified in the parameter space of each parameter to be identified of the photovoltaic system, and to construct a proxy model of the photovoltaic system based on each group of parameters to be identified and their corresponding real current-voltage curves; An initialization module is used to initialize individuals in the search population using each set of parameters to be identified, predict the predicted current-voltage curve of each individual in the search population using a proxy model, and determine the current predicted fitness value of each individual based on the deviation between the measured current-voltage curve and each predicted current-voltage curve; The optimization module is used to optimize each individual through the rime optimization algorithm according to the current predicted fitness value of each individual; determine the current best individual in the search population, determine the adaptive update step size based on the distance between the current best individual and the current individuals, and update the current best individual; The iterative module is used to replace the current optimal individual with the updated current optimal individual if the updated current optimal individual is better than the current optimal individual; otherwise, the updated current optimal individual is mutated and the mutated result is added to the search population for the next round of search optimization. After multiple rounds of iterative optimization, the current optimal individual is used as the parameters to be identified for the photovoltaic system under the current working condition to be predicted.

10. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 8 when executing the program.

Images