Method for optimizing photovoltaic model parameters based on information sharing search strategy and nm simplex whale optimization

Abstract

The application discloses a whale optimization photovoltaic model parameter method based on an information sharing search strategy and an NM simplex, the algorithm combines the information sharing strategy and the NM simplex strategy, and improves the performance of the original whale optimization algorithm.The introduced information sharing search strategy can strengthen the communication between individuals, and enable the individuals to better develop locally, and the NM simplex strategy can enable the algorithm to search in a wider solution space, and can obtain better solution quality, and results prove that after the two strategies are added, the improved original whale optimization algorithm obtains 30.01%, 24.24%, 50.98% and 6.94% improvements on a single diode, a double diode, a triple diode and a PV model respectively, and the method has higher ability to solve the solar photovoltaic model parameter extraction.

Description

Technical Field

[0001] This invention relates to the field of detection technology for solar photovoltaic cells and photovoltaic power generation arrays, specifically to a method for optimizing photovoltaic model parameters based on an information-sharing search strategy and the NM simplex. Background Technology

[0002] In order to achieve sustainable economic development, more and more countries and regions are shifting from fossil fuels to renewable energy. To meet the ever-increasing demand for energy, solar photovoltaic (PV) systems based on solar cells have seen rapid development in recent years. These systems can convert solar energy into electricity with minimal environmental impact. Their high efficiency and cleanliness have enabled this technology to gradually gain a foothold in global electricity production.

[0003] Scientists have applied photovoltaic (PV) technology to solar energy systems to improve energy conversion efficiency and reduce losses during the conversion process. We use equivalent circuit models to simulate the characteristics of PV systems under different environments. Since single-diode (SDM) and dual-diode (DDM) models can be established through mathematical models and have relatively simple structures, these two models are widely used. However, PV models are often exposed to harsh outdoor environments, making failures and performance degradation inevitable. Therefore, effectively designing, evaluating, and optimizing PV systems, and correctly assessing the performance of PV panels in practical operation, is crucial. The actual performance of a PV system mainly depends on its unknown parameters; therefore, evaluating these unknown parameters allows us to better understand the performance of the PV system in actual operation. However, SDM and DDM are implicit transcendental equations, making the evaluation of unknown parameters extremely difficult. Therefore, finding efficient and reliable methods to extract unknown parameters becomes very important.

[0004] To find a good parameter evaluation method, researchers have conducted a series of experiments, including analytical methods, direct methods, numerical methods, methods for extracting parameters using special functions, and methods for extracting parameters using the Lambert-W function. However, the various constraints of the objective function greatly limit the scope of application of such optimization methods, making them lack universality. At the same time, the computation process requires a large amount of gradient calculation, which inevitably leads to getting trapped in local optima.

[0005] Swarm intelligence (SI) algorithms have attracted widespread attention from researchers due to their advantages such as strong applicability, no limitation by any black box problem, ability to find near-optimal solutions to problems in a reasonable time and no need for gradient calculation. They have been extensively studied in fields such as feature selection, engineering optimization, and image segmentation. For example, the parameters of photovoltaic (PV) models have been optimized by using the multi-swarm parallel co-evolutionary differential evolution (MPPCEDE) method. By combining stochastic learning and simplex (NMs), relatively good parameter recognition accuracy has been achieved in PV models. In order to model SDM, DDM, and TDM in different environments, a fractional-order chaotic ensemble particle swarm optimization algorithm has been designed.

[0006] While metaheuristic algorithms have demonstrated excellent performance in applications, several challenges remain. First, they are prone to getting trapped in local optima, and there is room for improvement in both solution quality and convergence speed. Second, according to the "no free lunch" theorem, algorithms must be analyzed on a problem-by-problem basis. This necessitates continuous exploration and the design of algorithms tailored to different practical problems. To address this, we incorporated two mechanisms into the whale optimization algorithm. These two mechanisms complement each other, resulting in better performance. The whale optimization algorithm possesses strong global search capabilities, capable of searching the entire solution space and determining the approximate range of the optimal solution. The information-sharing search mechanism performs a coarse local search on the solutions obtained from the whale optimization algorithm, narrowing down the interval containing the optimal solution. The NM simplex provides a more refined search within the interval containing the optimal solution, identifying and outputting the optimal vector set. Summary of the Invention

[0007] This invention provides a method for optimizing photovoltaic model parameters based on an information-sharing search strategy and the NM simplex. This method can obtain the optimized parameter values ​​of the photovoltaic model relatively efficiently, and the error between the optimized parameter values ​​and the measured parameter values ​​is small.

[0008] A method for optimizing photovoltaic model parameters based on an information-sharing search strategy and the NM simplex includes:

[0009] (1) Based on the structural model of the photovoltaic array, set the parameter range of the photovoltaic model and set the parameter range of the photovoltaic model as the initial parameter vector set;

[0010] (2) The initial optimized parameter vector set is obtained by performing a global search on the initial parameter vector set using the whale optimization algorithm;

[0011] (3) The initial optimization parameter vector set is obtained by performing a local search using an information-sharing search strategy;

[0012] (4) Using a greedy strategy, select the first optimal parameter vector from the initial optimized parameter vector set obtained in step (2) and the optimized parameter vector set obtained in step (3) based on the fitness value, and take the parameter vector set where the first optimal parameter vector is located as the first parameter vector set;

[0013] (5) The second optimal parameter vector is obtained by using the NM simplex to perform a local search on the first optimal parameter vector;

[0014] (6) If the fitness value of the second optimal parameter vector is lower than the fitness value of the first optimal parameter vector, then the second optimal parameter vector is used as the optimal parameter vector, and the first optimal parameter vector in the first parameter vector set is replaced with the second optimal parameter vector to form the second parameter vector set. The second parameter vector set is input to step (2) for global search of the whale optimization algorithm. If the fitness value of the second optimal parameter vector is higher than the fitness value of the first optimal parameter vector, then the first optimal parameter vector is used as the optimal parameter vector, and the first parameter vector set is input to step (2) for global search of the whale optimization algorithm.

[0015] (7) Iteration steps (2)-(6) until the iteration number threshold is reached, then stop the iteration and use the obtained optimal parameter vector as the optimized photovoltaic model parameter vector.

[0016] The structural models of photovoltaic arrays include a single-diode five-parameter model, a dual-diode seven-parameter model, a triple-diode nine-parameter model, and a photovoltaic module model based on a five-parameter single-diode solar cell.

[0017] Among them, the output current I of the single diode five-parameter model L1 for:

[0018]

[0019] Among them, I ph I is the current generated by light. sd Represents the reverse saturation current, q is the fundamental charge, and V L For the output voltage, R s Represents the series resistance, n is the ideal coefficient of the diode, k is the Boltzmann constant, T is the Kelvin temperature, and R... sh This represents the shunt resistor, and the current I flowing through it is... sh for:

[0020]

[0021] Five unknown parameters, namely [I ph ,I sd ,R s ,R sh [,n];

[0022] Output current I of the dual diode seven-parameter model L2 for:

[0023]

[0024] Among them, the seven unknown parameters of the dual-diode seven-parameter model are [I ph ,I sd1 ,I sd2 ,R s ,R sh [,n1,n2]. Where, I sd1 and I sd2 n1 and n2 are the reverse saturation current of the diode, and n1 and n2 are the ideality factors of the dual diode.

[0025] The output current I of the three-diode nine-parameter model L3 for:

[0026]

[0027] In the three-diode model, there are nine main unknown parameters [I]. ph ,I sd1 ,I sd2 ,I sd3 ,R s ,R sh [n1, n2, n3]. Where, I sd1 I sd2 and I sd3 n1, n2, and n3 are the reverse saturation currents of the three diodes, and n1, n2, and n3 are the ideality factors of the three diodes.

[0028] The output current I of a photovoltaic module model based on a five-parameter single-diode solar cell. L4 for:

[0029]

[0030] Among them, the five-parameter vector of the photovoltaic module model based on a single-diode solar cell is [I ph ,I sd ,R s ,R sh ,n]. Where I L and V L To measure the current and voltage in the IV curve, voltage V L =KT / q, where K is the Boltzmann constant and q is the elementary charge.

[0031] The whale optimization algorithm is used to perform a global search on the initial parameter vector set, including:

[0032] When p < 0.5 and |A| ≥ 1, the initial optimization parameter vector set X'(t) is:

[0033] X'(t)=x rand -A*D

[0034] A = 2*a*r1 - a

[0035] a = 2 - (2*t / MAX_FEs)

[0036] D = |C*x * (t-1)-X(t-1)|

[0037] C = 2 * r²

[0038] Where, x rand The parameter vector is randomly selected from the optimized parameter vector set or the initial parameter vector set obtained from the previous iteration. The value of 'a' decreases linearly from 2 to 0. MAX_FEs is the maximum number of iterations, t represents the current iteration number, r1 and r2 are random numbers in (0,1), X(t-1) is the optimized parameter vector set or the initial parameter vector set obtained from the previous iteration, and x... * (t-1) represents the optimal parameter vector in the population in the previous iteration, and p is a random number in [0, 1].

[0039] When p < 0.5 and |A| < 1, the initial optimization parameter vector set X'(t) is:

[0040] X'(t)=x * (t-1)-A*D

[0041] When p ≥ 0.5, the initial optimization parameter vector set X'(t) is:

[0042] X'(t)=X * (t)+D*e bl *cos(2πl)

[0043] Where b is a constant and l is a random number in (0,1).

[0044] The optimal parameter vector set X(t) is obtained by performing a local search on the initial optimal parameter vector set using an information-sharing search strategy:

[0045]

[0046] Where X'(t) is the initial set of optimization parameter vectors, x' index(t) is a random parameter vector in the initial optimization parameter vector set, pa is a random number in (0,1); J∈[0,1], αc and βc are the domain learning coefficients, d represents the dimension of the problem being optimized (i.e., the number of unknowns), .* represents the dot product, and Levy is Levy flight.

[0047] The second optimal parameter vector is obtained by locally searching the first optimal parameter vector using the NM simplex, including:

[0048] (5.1) The initial simplex consists of multiple vertices. The values ​​of multiple variables in the first optimal parameter vector are assigned to some vertices of the initial simplex.

[0049] (5.2) Determine the vertex P with the largest function value. high The second largest vertex P sechi and the smallest vertex P low Calculate the value of excluding vertex P high The center P of the simplex after center Finally, define P high The reflection point is P refl for:

[0050] P refl =(1+α)P center -αP high

[0051] Where α is called the reflection coefficient and α > 0, y low The function value of the smallest vertex, y refl The function value at the point of greatest reflection, y sechi The function value of the second largest vertex, if y low ≤y refl ≤y sechi Then use P refl Replace P high The second optimal parameter vector is obtained;

[0052] (5.3) If y refl ≤y low Then, by performing an expansion operation on the reflection point, the expanded point P is obtained. exp If y exp <y low Then P exp Replace with P high The second optimal parameter vector is obtained; if y exp ≥y low , using P refl Replace P high Then, obtain the second optimal parameter vector and return to step (5.2);

[0053] (5.4) When y refl >ysechi And y refl ≤y high , using P refl Replace P high and for P refl Perform a contraction operation if y refl >y high Then directly target P high Perform a contraction operation to obtain the contracted vertex P. cont If y cont ≤y high Then P high Replace with P cont The second optimal parameter vector is obtained, and then the process returns to step (5.2); the formula for calculating the contracted vertex is as follows:

[0054] P cont =βP high +(1-β)P center

[0055] (5.5) If y cont >y high , except for P low The second optimal parameter vector is obtained by compressing all vertices except the i-th vertex. i for:

[0056] P i ←δP i +(1-δ)P low

[0057] Where "←" indicates assignment, δ is called the compression coefficient and 0 < δ < 1, P is recalculated. low Return to step (5.2) for the function value of each vertex except the one mentioned above.

[0058] The expansion point P is obtained by performing an expansion operation on the reflection point. exp for:

[0059] P exp =γP refl +(1-γ)P center

[0060] Where γ is the expansion coefficient, and γ > 1.

[0061] Directly to P high Perform a contraction operation to obtain the contracted vertex P. cont for:

[0062] P cont =βP high +(1-β)P center

[0063] Where β is the contraction coefficient, 0 < β < 1.

[0064] The fitness value RMSE(X) is:

[0065]

[0066] Where N is the number of measured current data points, and T i Let f be the i-th measured output current, X be the optimized photovoltaic model parameter vector, and f be the output current. i (·) represents the i-th predicted output current.

[0067] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0068] This invention provides a whale optimization algorithm that combines an information-sharing search strategy with the NM simplex. The basic whale optimization algorithm performs a coarse global search of the parameter range. Then, the information-sharing search strategy performs a coarse local search on the global search to optimize the parameter vector set. A greedy strategy is used to obtain the optimized parameter vector, followed by a fine local search using the NM simplex to optimize the parameter vector. The fitness values ​​of the two optimized parameter vectors are compared, and the superior optimized parameter vector and its corresponding parameter vector set are further iterated. After a set number of iterations, the final optimized parameters for the photovoltaic array structure model are obtained. This method improves the accuracy of photovoltaic parameter extraction. The improved whale optimization algorithm achieves performance improvements of 30.01%, 24.24%, 50.98%, and 6.94% on single-diode, dual-diode, triple-diode, and PV models, respectively. Attached Figure Description

[0069] Figure 1 A flowchart illustrating the method for optimizing photovoltaic model parameters based on an information-sharing search strategy and the NM simplex, provided for specific embodiments of the present invention.

[0070] Figure 2 is The present invention provides nine parameters for single-diode, dual-diode, and triple-diode photovoltaic modules, as well as equivalent circuit diagrams of photovoltaic modules based on single-diode photovoltaic cells, in specific embodiments thereof. Figure 2 a) is the equivalent circuit diagram of a single diode. Figure 2 b) is the equivalent circuit diagram of the two diodes. Figure 2 c) is the equivalent circuit diagram of three diodes with nine parameters. Figure 2 d) is the equivalent circuit diagram of a photovoltaic module based on a single diode photovoltaic cell.

[0071] Figure 3 The flowchart illustrates a method for optimizing photovoltaic model parameters using a whale model based on an information-sharing search strategy and the NM simplex, provided for a specific embodiment of the present invention.

[0072] Figure 4 A schematic diagram of NM simplex reflection operation provided for a specific embodiment of the present invention.

[0073] Figure 5 A schematic diagram of the NM simplex expansion operation provided for a specific embodiment of the present invention.

[0074] Figure 6 The NM simplex provided for specific embodiments of the present invention, when P refl Better than P high A schematic diagram of the contraction operation.

[0075] Figure 7 The NM simplex provided for specific embodiments of the present invention, when P high Better than P refl A schematic diagram of the contraction operation.

[0076] Figure 8 The NM simplex compression operation (P) provided for specific embodiments of the present invention refl Better than P high (The situation).

[0077] Figure 9 The NM simplex compression operation (P) provided for specific embodiments of the present invention refl P high Better than P refl (The situation).

[0078] Figure 10 The RMSE convergence curve of the whale optimization photovoltaic model parameter method based on information sharing search strategy and NM simplex provided in the specific embodiments of the present invention on a single diode (where ISNMWOA represents the algorithm, IJAYA represents the improved hybrid frog jumping algorithm, GOTLBO represents the teaching and learning optimization algorithm based on opposition learning, MLBSA represents the backtracking search algorithm with multiple learning strategies, GOFPANM represents the flower pollination optimization algorithm based on opposition learning and NM simplex, EHHO represents the adaptive orthogonal Harris Eagle optimization algorithm, GWO represents the gray wolf optimization algorithm, and WOA represents the whale optimization algorithm).

[0079] Figure 11The RMSE convergence curve of the whale optimization photovoltaic model parameter method based on information sharing search strategy and NM simplex provided in the specific embodiments of the present invention on a dual diode is shown (where ISNMWOA represents the algorithm, IJAYA represents the improved hybrid frog jumping algorithm, GOTLBO represents the teaching and learning optimization algorithm based on opposition learning, MLBSA represents the backtracking search algorithm with multiple learning strategies, GOFPANM represents the flower pollination optimization algorithm based on opposition learning and NM simplex, EHHO represents the adaptive orthogonal Harris Eagle optimization algorithm, GWO represents the gray wolf optimization algorithm, and WOA represents the whale optimization algorithm).

[0080] Figure 12 The RMSE convergence curve of the whale optimization photovoltaic model parameter method based on information sharing search strategy and NM simplex provided in the specific embodiments of the present invention on a three-diode grid (where ISNMWOA represents the algorithm, IJAYA represents the improved hybrid frog jumping algorithm, GOTLBO represents the teaching and learning optimization algorithm based on opposition learning, MLBSA represents the backtracking search algorithm with multiple learning strategies, GOFPANM represents the flower pollination optimization algorithm based on opposition learning and NM simplex, EHHO represents the adaptive orthogonal Harris Eagle optimization algorithm, GWO represents the gray wolf optimization algorithm, and WOA represents the whale optimization algorithm).

[0081] Figure 13 The RMSE convergence curve of the whale optimization photovoltaic model parameter method based on information sharing search strategy and NM simplex provided in the specific embodiments of the present invention on the PV model (where ISNMWOA represents the algorithm, IJAYA represents the improved hybrid frog jumping algorithm, GOTLBO represents the teaching and learning optimization algorithm based on opposition learning, MLBSA represents the backtracking search algorithm with multiple learning strategies, GOFPANM represents the flower pollination optimization algorithm based on opposition learning and NM simplex, EHHO represents the adaptive orthogonal Harris Eagle optimization algorithm, GWO represents the gray wolf optimization algorithm, and WOA represents the whale optimization algorithm).

[0082] Figure 14 The figure shows the IV curve fitting between simulated experimental data and actual measured data in photovoltaic models on SDM, DDM, TDM, and PV, based on the information-sharing search strategy and NM simplex method for whale-based optimization of photovoltaic model parameters, as provided in the specific embodiments of the present invention. Figure 14 (a) is a graph showing the IV curve fitting between simulated experimental data and actual measured data in the photovoltaic model on SDM. Figure 14 (b) is a graph showing the fitting of the IV curves between the simulated experimental data and the actual measured data in the photovoltaic model on the DDM. Figure 14 (c) is a graph showing the fitting of the IV curves between the simulated experimental data and the actual measured data in the photovoltaic model on the DDM. Figure 14 (d) is a graph showing the fitting of the IV curve between the simulated experimental data and the actual measured data in the photovoltaic model on the PV. Detailed Implementation

[0083] The present invention will be further described and illustrated below with reference to specific embodiments.

[0084] This invention provides a whale optimization algorithm based on an information-sharing search strategy and the Nelder-Mead simplex for identifying unknown parameters in photovoltaic models, such as... Figure 1 As shown, it includes:

[0085] (1) Based on the structural model of the photovoltaic array, set the parameter range of the photovoltaic model and set the parameter range of the photovoltaic model as the initial parameter vector set;

[0086] (2) The initial optimized parameter vector set is obtained by performing a global search on the initial parameter vector set using the whale optimization algorithm;

[0087] (3) The initial optimization parameter vector set is obtained by performing a local search using an information-sharing search strategy;

[0088] (4) Using a greedy strategy, select the first optimal parameter vector from the initial optimized parameter vector set obtained in step (2) and the optimized parameter vector set obtained in step (3) based on the fitness value, and take the parameter vector set where the first optimal parameter vector is located as the first parameter vector set;

[0089] (5) The second optimal parameter vector is obtained by using the NM simplex to perform a local search on the first optimal parameter vector;

[0090] (6) If the fitness value of the second optimal parameter vector is lower than the fitness value of the first optimal parameter vector, then the second optimal parameter vector is used as the optimal parameter vector, and the first optimal parameter vector in the first parameter vector set is replaced with the second optimal parameter vector to form the second parameter vector set. The second parameter vector set is input to step (2) for global search of the whale optimization algorithm. If the fitness value of the second optimal parameter vector is higher than the fitness value of the first optimal parameter vector, then the first optimal parameter vector is used as the optimal parameter vector, and the first parameter vector set is input to step (2) for global search of the whale optimization algorithm.

[0091] (7) Iteration steps (2)-(6) until the iteration number threshold is reached, then stop the iteration and use the obtained optimal parameter vector as the optimized photovoltaic model parameter vector.

[0092] like Figure 2The structural models of photovoltaic arrays are shown to be a single-diode five-parameter model (SDM), a dual-diode seven-parameter model (DDM), a three-diode nine-parameter model (TDM), and a photovoltaic module model based on a five-parameter single-diode solar cell.

[0093] Among them, Figure 2 a) is the equivalent circuit diagram of a single diode, which consists of a current source connected in parallel with the diode, a shunt resistor representing the leakage current, and a series resistor representing the load current loss. The output current I of the five-parameter model of the single diode is... L1 for:

[0094]

[0095] Among them, I ph I is the current generated by light. sd Represents the reverse saturation current, where q is the fundamental charge (1.60217646 × 10⁻⁶). -19 C), V L For the output voltage, R s Represents the series resistance, n is the ideality coefficient of the diode, and k is the Boltzmann constant (1.3806503 × 10⁻⁶). 23 J / K), T is Kelvin temperature, R sh This represents the shunt resistor, and the current I flowing through it is... sh for:

[0096]

[0097] Five unknown parameters, namely [I ph ,I sd ,R s ,R sh [,n];

[0098] Figure 2 b) is the equivalent circuit diagram of the dual diode, which is achieved by adding a diode to the original single diode to characterize the complex losses of carriers in the dissipation region. The output current I of the seven-parameter dual diode model is... L2 for:

[0099]

[0100] Among them, the seven unknown parameters of the dual-diode seven-parameter model are [I ph ,I sd1 ,I sd2 ,R s ,R sh [,n1,n2]. Where, I sd1 and I sd2n1 and n2 are the reverse saturation current of the diode, and n1 and n2 are the ideality factors of the dual diode.

[0101] Figure 2 c) shows the equivalent circuit diagram of the three diodes, which is achieved by adding a diode to the original two-diode circuit. The output current I of the three-diode nine-parameter model is... L3 for:

[0102]

[0103] In the three-diode model, there are nine main unknown parameters [I]. ph ,I sd1 ,I sd2 ,I sd3 ,R s ,R sh [n1, n2, n3]. Where, I sd1 I sd2 and I sd3 n1, n2, and n3 are the reverse saturation currents of the three diodes, and n1, n2, and n3 are the ideality factors of the three diodes.

[0104] Figure 2 d) is the equivalent circuit diagram of a photovoltaic module composed of solar cells made of single diodes. It is the interface for converting light into electrical energy and is typically composed of multiple solar cells connected in series or parallel. The photovoltaic module we are using here is composed of N... p A number of solar cells are connected in parallel, and N s A photovoltaic module consisting of several solar cells connected in series, based on a photovoltaic module model with a five-parameter single-diode solar cell, has an output current I... L4 for:

[0105]

[0106] Among them, the five-parameter vector of the photovoltaic module model based on a single-diode solar cell is [I ph ,I sd ,R s ,R sh ,n]. Where I L4 and V L To measure the current and voltage in the IV curve, voltage V L = KT / q, where K is the Boltzmann constant, typically taken as 1.3806503 × 10⁻⁶. -23 J / K, where q is the elementary charge, with a value of 1.60217646 × 10⁻⁶. -19 C.

[0107] The fitness value RMSE(X) is:

[0108]

[0109] Where N is the number of measured current data points, and T i Let f be the i-th measured output current, X be the optimized photovoltaic model parameter vector, and f be the output current. i (·) represents the i-th predicted output current. For SDM, DDM, as shown in equations (7), (8), and (9), X in equation (7) is determined by I. ph I sd R s R sh Composed of I and n, X in formula (8) is composed of I ph I sd1 I sd2 R s R sh It consists of n1 and n2, and X in formula (9) is composed of I ph I sd1 I sd2 I sd3 R s R sh It consists of n1, n2 and n3.

[0110] For SDM, a search strategy based on information sharing and a whale optimization method for photovoltaic model parameters based on NM simplex are proposed, such as Figure 3 As shown below:

[0111] Step 1: Based on the data provided by RTC France, the photovoltaic solar cell manufacturer, set the search range for the model parameters. For single / dual diode solar cells, the parameter range is: I ph ∈[0,1](A),R s ∈[0,0.5](Ω),R sh ∈[1,100](Ω),n∈[1,2],I sd1 ∈[0,1](A),I sd2 ∈[0,1](A), n1∈[1,2], n2∈[1,2]. For the parameter range of a photovoltaic module composed of a single diode solar cell: I ph ∈[0,2](A),I sd ∈[0,50](A),R s ∈[0,2](Ω),R sh ∈[1,2000](Ω),n∈[1,50).

[0112] Step 2: A basic whale optimization algorithm is used to perform a global search within the feasible solution space to obtain initial vectors for several photovoltaic model parameters. The main parameters and steps involved in the algorithm are as follows:

[0113] The number of individuals, i.e., the number of parameter vectors: N = 30;

[0114] The maximum number of evaluations required for the algorithm to terminate: MAX_FEs;

[0115] Parameters for initializing the information sharing search strategy: J, alpha;

[0116] The upper bound of the parameter vector is denoted as UB, and the lower bound is denoted as LB;

[0117] The population is represented by X, and is composed of x i = rand·(UB-LB)+LB, i=1,...,N Initialize individuals, x i This represents the i-th individual in the population.

[0118] The whale optimization algorithm first calculates A, C, b, l, p when each position is updated.

[0119] When p < 0.5 and |A| ≥ 1, the initial optimization parameter vector set X'(t) is:

[0120] X'(t)=x rand -A*D (10)

[0121] Among them, X rand A is a parameter vector randomly selected from the optimized parameter vector set or the initial parameter vector set obtained from the previous iteration. A is calculated using the formula A = 2*a*r1 - a, where the value of a decreases linearly from 2 to 0. A is calculated using a = 2 - (2*t / MAX_FEs), where t represents the current iteration number. D is calculated using D = |C*x * (t-1)-X(t-1)| is calculated, C is calculated from C=2*r2, r1 and r2 are random numbers in (0,1), X(t-1) is the optimized parameter vector set or the initial parameter vector set obtained from the previous iteration, x * (t-1) represents the optimal parameter vector in the population during the previous iteration.

[0122] When p < 0.5 and |A| < 1, the initial optimization parameter vector set X'(t) is:

[0123] X'(t)=x * (t-1)-A*D (11)

[0124] When p ≥ 0.5, the initial optimization parameter vector set X'(t) is:

[0125] X'(t)=x *(t-1)+D*e bl *cos(2πl) (12)

[0126] Where b is a constant and l is a random number in (0,1). The whale swims in a spiral towards its prey while simultaneously tightening its encirclement.

[0127] A greedy strategy is adopted to select the first optimal parameter vector from the initial optimized parameter vector set obtained in step (2) and the optimized parameter vector set obtained in step (3) based on the fitness value, and the parameter vector set containing the first optimal parameter vector is taken as the first parameter vector set.

[0128] Step 3: Implement an information-sharing search strategy for each individual in the population, with the iteration method as follows:

[0129] The optimal parameter vector set X(t) is obtained by performing a local search on the initial optimal parameter vector set using an information-sharing search strategy:

[0130]

[0131] Where X'(t) is the initial set of optimization parameter vectors, x' index (t) is a random parameter vector in the initial optimization parameter vector set, pa is a random number in (0,1); J∈[0,1], αc and βc are the domain learning coefficients, d represents the dimension of the problem being optimized (i.e., the number of unknowns), .* represents the dot product, and Levy is Levy flight.

[0132] Step 4: Perform Nelder-Mead simplex operations based on the optimal individual position. This method is a local search algorithm for solving unconstrained optimization problems and does not require any derivative information of the objective function. For minimizing a function with d variables, the NM method uses operations such as reflection, expansion, contraction, and compression. It compares the objective function values ​​of the (d+1) vertices of the simplex, then replaces the vertex with the largest objective function value with a new point. Through iterative updates, the simplex eventually approximates the optimal solution to the problem.

[0133] Step 4-1: The initial simplex consists of multiple vertices. Assign multiple variable values ​​from the first optimal parameter vector to some vertices of the initial simplex.

[0134] Step 4-2: Determine the vertex P with the largest function value. high The second largest vertex P sechi and the smallest vertex P low Calculate the value of excluding vertex P high The center P of the simplex after center ,like Figure 4 As shown, P is finally defined. high The reflection point is Prefl for:

[0135] P refl =(1+α)P center -αP high (14)

[0136] Where α is called the reflection coefficient and α > 0, y low The function value of the smallest vertex, y refl The function value at the point of greatest reflection, y sechi The function value of the second largest vertex, if y low ≤y refl ≤y sechi Then use P refl Replace P high The second optimal parameter vector is obtained;

[0137] Step 4-3: If y refl ≤y low ,like Figure 5 As shown, the expanded point P is obtained by performing an expansion operation on the reflection point. exp If y exp <y low Then P exp Replace with P high The second optimal parameter vector is obtained; if y exp ≥y low ,use lf P er Replace P high We obtain the second optimal parameter vector and then return to step 4-2;

[0138] Step 4-4: When y refl >y sechi And y refl ≤y high , using P refl Replace P high ,like Figure 6 As shown, and for P refl Perform a contraction operation if y refl >y high ,like Figure 7 As shown, then directly apply P high Perform a contraction operation to obtain the contracted vertex P. cont If y cont ≤y high Then P high Replace with P cont We obtain the second optimal parameter vector, and then return to step 4-2; the formula for calculating the contracted vertex is as follows:

[0139] P cont =βP high +(1-β)Pcenter (15)

[0140] Steps 4-5: If y cont >y high , except for P low The second optimal parameter vector is obtained by compressing all vertices except the i-th vertex. i for:

[0141] P i ←δP i +(1-δ)P low (16)

[0142] Where "←" indicates assignment, δ is called the compression coefficient and 0 < δ < 1, P is recalculated. low Except for the function value of each vertex, return to step 4-2. The compression operation (P) occurs when shrinkage fails. refl Better than P high (Scenario) such as Figure 8 As shown. Compression operation (P) when contraction fails. high Better than P refl (Scenario) such as Figure 9 As shown.

[0143] Step 5: If the fitness value of the second optimal parameter vector is lower than that of the first optimal parameter vector, then the second optimal parameter vector is used as the optimal parameter vector, and the first optimal parameter vector in the first parameter vector set is replaced with the second optimal parameter vector to form a second parameter vector set. The second parameter vector set is input into Step 2 for global search using the whale optimization algorithm. If the fitness value of the second optimal parameter vector is higher than that of the first optimal parameter vector, then the first optimal parameter vector is used as the optimal parameter vector, and the first parameter vector set is input into Step 2 for global search using the whale optimization algorithm.

[0144] Step 6: Determine whether the current number of evaluations t has reached the maximum number of evaluations MAX_FEs. If not, return to step 2 and execute the relevant steps in sequence. If the maximum number of evaluations has been reached, use the obtained optimal parameter vector as the optimized photovoltaic model parameter vector.

[0145] In this invention, we use the Nelder-Mead simplex function `fminsearchbnd` provided by the MATLAB toolbox. Since the optimal parameters we obtain at the beginning of the evaluation process may not be the optimal parameters we expect, we don't need to spend a lot of time searching the vicinity. As the number of evaluations gradually increases, the optimal value gradually approaches the expected optimal value. At this point, we can search the vicinity multiple times. Based on this idea, in this invention, we set the parameter representing the maximum number of evaluations in `fminsearchbnd` to 0.1 * MAX_FEs. Figure 10 , Figure 11 , Figure 12 , Figure 13 and Figure 14 As shown in (a)-(d), and in conjunction with Table 1 for comparison, the optimization effect of this algorithm is slightly better than GOFPANM on each model, but the algorithm takes less time to run, that is, it achieves better performance in less time.

[0146] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention are all within the protection scope of the present invention. Table 1. Running time (in seconds) of different algorithms on single diode, dual diode, triple diode and PV models (where ISNMWOA represents the present algorithm, IJAYA represents the improved hybrid frog jumping algorithm, GOTLBO represents the teaching and learning optimization algorithm based on opposition learning, MLBSA represents the backtracking search algorithm with multiple learning strategies, GOFPANM represents the flower pollination optimization algorithm based on opposition learning and NM simplex, EHHO represents the adaptive orthogonal Harris Eagle optimization algorithm, GWO represents the gray wolf optimization algorithm, and WOA represents the whale optimization algorithm)

[0147]

[0148] Table 2. Optimal parameter estimates for ST40 at 25°C and different irradiance levels.

[0149]

[0150] Table 3. Optimal parameter estimates for SM55 at 25°C and different irradiance levels.

[0151]

[0152]

[0153] Table 4. Optimal parameter estimates for KC200GT at 25°C and different irradiance levels

[0154]

[0155] Table 5. At an irradiance of 1000 W / m 2 Optimal parameter estimation for ST40 under different temperatures

[0156]

[0157]

[0158] Table 6. At an irradiance of 1000 W / m 2 Optimal parameter estimation of SM55 under different temperatures

[0159]

[0160] Table 7. At an irradiance of 1000 W / m 2 Optimal parameter estimation for KC200GT under different temperatures

[0161]

[0162]

Claims

1. A method for optimizing photovoltaic model parameters using a whale model based on an information-sharing search strategy and the NM simplex, characterized in that, include: (1) Based on the structural model of the photovoltaic array, set the parameter range of the photovoltaic model and set the parameter range of the photovoltaic model as the initial parameter vector set; (2) The initial optimized parameter vector set is obtained by performing a global search on the initial parameter vector set using the whale optimization algorithm; (3) The initial set of optimized parameter vectors is searched locally using an information-sharing search strategy to obtain the set of optimized parameter vectors; (4) Using a greedy strategy, select the first optimal parameter vector from the initial optimized parameter vector set obtained in step (2) and the optimized parameter vector set obtained in step (3) based on the fitness value, and take the parameter vector set where the first optimal parameter vector is located as the first parameter vector set; (5) The second optimal parameter vector is obtained by using the NM simplex to perform a local search on the first optimal parameter vector; (6) If the fitness value of the second optimal parameter vector is lower than that of the first optimal parameter vector, then the second optimal parameter vector is used as the optimal parameter vector, and the first optimal parameter vector in the first parameter vector set is replaced with the second optimal parameter vector to form the second parameter vector set. The second parameter vector set is input to step (2) for global search of the whale optimization algorithm. If the fitness value of the second optimal parameter vector is higher than that of the first optimal parameter vector, then the first optimal parameter vector is used as the optimal parameter vector, and the first parameter vector set is input to step (2) for global search of the whale optimization algorithm. (7) Iteration steps (2)-(6) until the iteration number threshold is reached, then stop the iteration and use the obtained optimal parameter vector as the optimized photovoltaic model parameter vector; The second optimal parameter vector is obtained by locally searching the first optimal parameter vector using the NM simplex, including: (5.1) The initial simplex consists of multiple vertices. The values ​​of multiple variables in the first optimal parameter vector are assigned to some vertices of the initial simplex. (5.2) Determine the vertex with the largest function value The second largest vertex and the smallest vertex Calculate the removal of vertices The center of the simplex after Finally defined The reflection point is for: ,in, This is called the reflection coefficient and , The function value of the smallest vertex. The function value at the point of greatest reflection. The function value of the second largest vertex, if Then use replace The second optimal parameter vector is obtained; (5.3) If Then, by performing an expansion operation on the reflection point, the expanded point is obtained. ,if Then Replace with The second optimal parameter vector is obtained; if ,use replace Then, obtain the second optimal parameter vector and return to step (5.2); (5.4) When and ,use replace and to Perform a contraction operation, if Then directly to Perform a contraction operation to obtain the contraction vertex. ,if Then Replace with The second optimal parameter vector is obtained, and then the process returns to step (5.2). The formula for calculating the contracted vertex is as follows: ; (5.5) If , except The second optimal parameter vector is obtained by compressing all vertices except the first one. i Compressed vertices for: ,in," " indicates assignment, This is called the compressibility factor and Recalculate Return to step (5.2) for the function value of each vertex except the one mentioned above.

2. The method for optimizing photovoltaic model parameters based on information sharing search strategy and NM simplex according to claim 1, characterized in that, The structural models of photovoltaic arrays include a single-diode five-parameter model, a dual-diode seven-parameter model, a triple-diode nine-parameter model, and a photovoltaic module model based on a five-parameter single-diode solar cell; among them, the output current of the single-diode five-parameter model is... for: ,in, The electric current generated by light, Represents the reverse saturation current. For the basic electric charge, For output voltage, This represents the series resistance. The ideal coefficient of the diode, Boltzmann constant, T Kelvin temperature, Indicates shunt resistance for: Five unknown parameters, namely Output current of the dual-diode seven-parameter model for: Among them, the seven unknown parameters of the dual-diode seven-parameter model are: ,in, and This is the reverse saturation current of the diode. and Ideal factor for dual diodes; output current of the three-diode nine-parameter model. for: The three-diode model has nine main unknown parameters. ,in, , and This refers to the reverse saturation current of the three diodes. , and The ideal factor for a three-diode diode; Output current of a photovoltaic module model based on a five-parameter single-diode solar cell for: Among them, the five-parameter vector of the photovoltaic module model based on a single-diode solar cell is: ,in and To measure the current and voltage in the IV curve, voltage Where K is the Boltzmann constant, q This represents the fundamental charge.

3. The method for optimizing photovoltaic model parameters based on information sharing search strategy and NM simplex according to claim 1, characterized in that, The whale optimization algorithm is used to perform a global search on the initial parameter vector set, including: when and At that time, the initial optimization parameter vector set for: in, The parameter vectors are randomly selected from the optimized parameter vector set obtained from the previous iteration or the initial parameter vector set. The value decreases linearly from 2 to 0. The maximum number of iterations is given, and t represents the current number of iterations. and yes Random numbers in the data, The set of optimized parameters or the set of initial parameters obtained from the previous iteration. Let p represent the optimal parameter vector in the previous iteration of the population, where p is a random number in the range [0, 1]. and At that time, the initial optimization parameter vector set for: ,when At that time, the initial optimization parameter vector set for: in, b It is a constant. l for Random numbers in the array.

4. The method for optimizing photovoltaic model parameters based on information sharing search strategy and NM simplex according to claim 1, characterized in that, An information-sharing search strategy is used to perform a local search on the initial set of optimization parameter vectors to obtain the optimized parameter vector set. for: in, For the initial set of optimization parameter vectors, For the initial set of optimization parameter vectors, a random parameter vector. yes Random numbers in the data; , , For the domain learning coefficient, It represents the dimension of the problem being optimized (that is, the number of unknowns). This represents the dot product. It was Levi's flight.

5. The method for optimizing photovoltaic model parameters based on information sharing search strategy and NM simplex according to claim 1, characterized in that, The expansion point is obtained by performing an expansion operation on the reflection point. for: in, The expansion coefficient, .

6. The method for optimizing photovoltaic model parameters based on an information-sharing search strategy and the NM simplex according to claim 1, characterized in that, Directly Perform a contraction operation to obtain the contraction vertex. for: in, It is the coefficient of shrinkage. .

7. The method for optimizing photovoltaic model parameters based on information sharing search strategy and NM simplex according to claim 2, characterized in that, fitness value for: Among them, among them, The number of measured current data points. T i For the first i The measured output current This is the optimized photovoltaic model parameter vector. For the first i One predicted output current.

Images