Photovoltaic cell model parameter identification method and system based on improved pollination algorithm

Abstract

This invention discloses a method and system for identifying photovoltaic cell model parameters based on an improved flower pollination algorithm, belonging to the field of photovoltaic cell modeling. It includes: receiving settings for the type of the photovoltaic cell model and the parameters of the flower pollination algorithm; determining the composition of parameters to be identified based on the set photovoltaic cell model type, using these as the positions of individual pollen individuals, and initializing the pollen population; iterating the population using the improved flower pollination algorithm, and outputting the global optimal solution and the corresponding pollen positions; during population iteration, when updating the pollen individual positions globally for pollination, the parameter controlling the pollination intensity uses a Gaussian distributed random number; when updating the pollen individual positions locally for pollination, the positions of two pollen individuals in the population that are different from the current pollen come from different clusters. This invention uses a Gaussian distribution to represent long-distance pollination in the flower pollination process and a clustering evolution strategy to represent short-distance pollination, effectively improving convergence speed and solution accuracy.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic cell modeling, and more specifically, relates to a method and system for identifying photovoltaic cell model parameters based on an improved flower pollination algorithm. Background Technology

[0002] Global warming and the increasing frequency of extreme weather events have made the necessity of clean energy increasingly important to humanity. Due to its widespread availability, harmlessness, and sustainability, solar energy has become one of the major renewable energy sources and is also receiving growing attention as an important alternative to fossil fuels. To ensure that photovoltaic power generation efficiently converts solar energy into electrical energy, more and more researchers have conducted in-depth studies on the parameter identification of photovoltaic cell models. Significant progress has been made in the past few decades in accurately modeling photovoltaic cells using mathematical models to simulate their characteristics.

[0003] Currently, widely accepted photovoltaic (PV) cell models fit observed current-voltage (IV) data through an equivalent circuit composed of diodes. These models primarily include single-diode and dual-diode models. Accurate identification of PV cell model parameters is crucial not only for evaluating PV cell performance under shading and different temperatures but also for PV cell design optimization and quality control. However, existing PV cell model parameter identification methods largely employ deterministic optimization methods from operations research. Deterministic optimization methods typically require the optimization problem to be convex or differentiable, and the optimization results obtained through deterministic methods often deviate significantly from the ideal results. Heuristic algorithms, on the other hand, treat the optimization problem as a black box, disregarding its specific form, and offer advantages such as ease of implementation and high solution efficiency.

[0004] Research has found that the Flower Pollination Algorithm (FPA), a popular heuristic algorithm, suffers from slow convergence speed and low solution accuracy in model parameter identification. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for identifying photovoltaic cell model parameters based on an improved flower pollination algorithm, thereby solving the problems of slow convergence speed and low solution accuracy in existing photovoltaic cell model parameter identification.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a method for identifying photovoltaic cell model parameters based on an improved flower pollination algorithm, the method comprising:

[0007] S1. Receive the settings for the photovoltaic cell model type and flower pollination algorithm parameters, determine the composition of the parameters to be identified based on the set photovoltaic cell model type, use them as the position of individual pollen, and initialize the pollen population;

[0008] S2. An improved flower pollination algorithm is used for population iteration, and the global optimal solution and the corresponding pollen location are output as the parameter identification results of the photovoltaic cell model;

[0009] In the improved flower pollination algorithm, when updating the position of individual pollen individuals globally during population iteration, the parameter controlling the pollination intensity uses a Gaussian distributed random number. When updating the position of individual pollen individuals locally during pollination, the positions of two pollen individuals in the population that are different from the current pollen come from different clusters. The clusters are two sets divided by sorting the individual fitness from small to large.

[0010] Preferably, during global pollination, the pollen location The update formula is as follows:

[0011]

[0012] in, G is the position vector of the i-th individual evaluated by the (t+1)-th function. t The global optimal individual position corresponds to the position of the individual with the lowest fitness in the current population. Gauss(0,α) represents a Gaussian random number with a mean of 0 and a standard deviation of α.

[0013] Preferably, during local pollination, the pollen location The update formula is as follows:

[0014]

[0015] in, Let i be the position vector of the i-th individual evaluated by the (t+1)-th function, B be the cluster set with smaller fitness, and C be the cluster set with larger fitness, and i1 and i2 be the indices of individuals randomly selected from B and C, respectively.

[0016] Preferably, after obtaining the location of the new individual, an individual is randomly selected from the cluster set C with higher fitness and a local search is performed according to the chaotic elite strategy, as follows:

[0017] If the ratio of the current number of evaluations to the maximum number of evaluations is greater than a random number in the interval [0,1], then a chaotic random number z is generated. l =z l ·η·(1-z l-1 And perform a local search at the globally optimal location. Otherwise, the individual's historical best position is used. Among them, zl Let z0 be the value of the l-th chaotic sequence, and z0 be an initial random number uniformly distributed in the interval [0,1], and η be a random number between [3.57,4.0]. Let i be the position of the 3rd individual in the t-th iteration in the j-th dimension. Let i3 be the historical best position of the i3th individual in the j-th dimension, which is evaluated by the function of the t-th iteration. i3 is the randomly selected index position of the individual, and step is a random number distributed in the interval [0,1].

[0018] Preferably, after obtaining the new individual position, an adaptive boundary handling strategy is used to adjust unreasonable positions in the individual evolution process, as follows:

[0019] like but

[0020] otherwise,

[0021] in, Let Lb be the position of the i-th individual in the j-th dimension during the t-th iteration. j Let Ub be the lower bound of the j-th dimension of the problem space. j Let φ be the upper bound of the j-th dimension of the problem space, and let φ be the median value of the upper and lower boundaries of the problem.

[0022] Preferably, if the photovoltaic cell model type is a single diode model, then the parameter to be identified is X = {I} ph ,I sd ,R s ,R sh If the photovoltaic cell model type is a dual-diode model, then the parameter to be identified is X = I. ph ,I sd1 ,R s ,R sh ,n1,I sd2 ,n2};

[0023] Among them, I ph Indicates the photovoltaic current generated by solar energy; I sd Represents the diode short-circuit current; R s and R sh These represent the series and parallel resistances, respectively; n represents the diode's ideality factor; I sd1 and I sd2 n1 and n2 represent the saturation currents of the first and second diodes, respectively; n1 and n2 represent the ideality factors of the first and second diodes, respectively.

[0024] Preferably, the fitness function is the root mean square error of the observed current and the calculated current.

[0025] To achieve the above objectives, in a second aspect, the present invention provides a photovoltaic cell model parameter identification system based on an improved flower pollination algorithm, comprising: a processor and a memory; the memory being used to store computer execution instructions; and the processor being used to execute the computer execution instructions, causing the method described in the first aspect to be executed.

[0026] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0027] To address the shortcomings of slow convergence and low accuracy in photovoltaic (PV) cell model parameter identification using flower pollination algorithms, this invention proposes a method and system for PV cell model parameter identification based on an improved flower pollination algorithm. The core idea is to use a Gaussian distribution to represent long-distance pollination and a clustering evolution strategy to represent short-distance pollination, effectively improving the convergence speed and accuracy of the flower pollination algorithm in PV model parameter identification. Furthermore, an adaptive boundary handling strategy is employed to adjust unreasonable positions during the individual evolution process; and a chaotic elite learning strategy is used to improve the performance of weaker individuals, further enhancing convergence speed and accuracy. Attached Figure Description

[0028] Figure 1 A flowchart of a photovoltaic cell model parameter identification method based on an improved flower pollination algorithm provided by the present invention;

[0029] Figure 2(a) shows the box-shaped statistical charts drawn after the single diode model of the present invention and the comparative method were run 30 times respectively;

[0030] Figure 2(b) shows the box-shaped statistical charts drawn after the present invention and the comparative method were run 30 times in the dual diode model, respectively;

[0031] Figure 3(a) shows the current-voltage curve generated after obtaining the model parameters using the method of the present invention under the single diode model provided in the embodiment of the present invention;

[0032] Figure 3(b) shows the power-voltage curve generated after obtaining the model parameters using the method of the present invention under the single diode model provided in the embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0034] like Figure 1As shown, this invention proposes a method for identifying photovoltaic cell model parameters based on an improved flower pollination algorithm, specifically including the following steps:

[0035] (1) Set the parameters of the flower pollination algorithm and the type of photovoltaic cell model to be identified. The parameters of the flower pollination algorithm include: conversion probability p, population size N, and maximum number of target evaluations Fes. max And the fitness function F(). After determining the type of photovoltaic cell model to be identified, the composition of the parameters to be identified, the total number of individual dimensions D, and the upper and lower limits of the problem space can be determined.

[0036] In this invention, the fitness function F() uses the root mean square error of the observed current and the calculated current.

[0037] (2) Randomly initialize the population Pop in the problem space.

[0038] Pollen individuals in a population (Pop) are represented as follows:

[0039]

[0040] in, Let Ub be the position of the i-th individual in the j-th dimension during the t-th iteration, representing the j parameter values ​​to be identified in the problem. In this invention, each individual can be represented as a position vector composed of all the parameters to be identified; j Lb is the upper bound of the j-th dimension of the problem space. j is the lower bound of the j-th dimension of the problem space; r1 is a random number uniformly distributed in the interval [0,1]. N represents the number of individuals in the population; D represents the total number of dimensions for any individual, and in the problem, it represents the number of all parameters to be identified.

[0041] (3) Assess the fitness of each individual's location in the population and update the individual locations in the population.

[0042] The individual location is updated in the following way:

[0043]

[0044] in, Let the (t+1)th function evaluate the position vector of the i-th individual. The position vector of the i-th individual is evaluated by the (t-1)-th function. The fitness value of the i-th individual is evaluated using the (t+1)-th function.

[0045] (4) Update the global best position and the historical best position of an individual in the current population based on the individual fitness of the population.

[0046] The global optimal individual position is the position of the individual with the lowest fitness in the current population, while the historical optimal individual position is the position of any individual in the population up to the current occurrence of the individual with the lowest fitness.

[0047] The global optimal position and the individual historical optimal position are expressed as follows:

[0048]

[0049] Among them, G t Let the function evaluate the globally optimal individual position vector in the current population for the t-th time. That is, the optimal identification parameter vector output by the algorithm for the problem after the search is completed; Evaluate the current individual using the t-th function. Corresponding to the individual's historical optimal position vector,

[0050] (5) Based on the conversion probability p, choose between global pollination or local pollination.

[0051] Global pollination uses a vector combination of the current individual, the global best individual, and Gaussian distributed random numbers to generate a new individual position.

[0052] Local pollination employs a clustering evolution strategy to randomly combine the current individual, the set of individuals in the better subpopulation B, and the set of individuals in the worse subpopulation C to generate a new individual position.

[0053] Global pollination and local pollination are expressed as follows:

[0054]

[0055] Where Gauss(0,α) represents a Gaussian random number with a mean of 0 and a standard deviation of α; B and C represent the sets of better and worse individuals, respectively, sorted by fitness from smallest to largest according to the clustering idea, where B is the first N / 2 better individuals and C is the last N / 2 worse individuals; i1 and i2 are the indices of individuals randomly selected from B and C, respectively; r3 and r4 are random numbers uniformly distributed in the interval [0,1].

[0056] (6) Adaptive boundary processing strategy is used to check the boundaries of new individuals. If the individual position of the population exceeds the problem boundary, the destruction value of exceeding the nearest problem boundary is recorded. If the destruction value exceeds the preset threshold, the individual is fixed at the problem boundary. Otherwise, the individual is mapped into the problem space with the problem boundary as the center of symmetry.

[0057] The adaptive boundary handling strategy is expressed as follows:

[0058]

[0059]

[0060] Where φ is the median value of the upper and lower boundaries of the problem, φ=(Ub j +b j )×0.5.

[0061] (7) Randomly select an individual from the set of individuals C of the poor subpopulation and perform a local search according to the chaotic elite strategy.

[0062] The chaotic elite strategy is as follows: if the ratio of the current number of evaluations to the maximum number of evaluations is greater than a random number in the interval [0,1], then a chaotic random number is generated and a local search is performed at the global optimal position; otherwise, the individual's historical optimal position is adopted.

[0063] The chaotic elite strategy is expressed as follows:

[0064] z l+1 = l ··(1- l )

[0065]

[0066] Where η is a random number between [3.57, 4.0]; z l Let z0 be the value of the l-th chaotic sequence, and z5 be an initial random number uniformly distributed in the interval [0,1]. The function evaluates the position of the i3th historical best individual in the j-th dimension for the t-th iteration, where i3 is a randomly selected individual index position. step is a random number distributed in the interval [0,1].

[0067] (8) Determine whether the search stopping condition has been met. If it is met, output the global optimal position as the final solution. Otherwise, go to step (3) for repeated iterative calculation.

[0068] The search stops when the maximum number of iterations is reached or the maximum number of function evaluations is reached.

[0069] To verify the effectiveness of the model proposed in this invention, this embodiment uses a photovoltaic cell with a diameter of 27 mm at a light irradiance of 1000 W / m². 2 The 26 pairs of voltage and current data measured at 33℃ were used as the research dataset.

[0070] The diode model used is as follows:

[0071] 1) Single diode model:

[0072]

[0073] Among them, I ph Indicates the photovoltaic current generated by solar energy; I sd This represents the diode short-circuit current; q represents the elementary charge, q = 1.60217646 × 10⁻⁶. -19 C;V L and I L These represent the output voltage and current values, respectively; R s and R sh represents the series and parallel resistances respectively; n represents the diode ideality factor; k represents the Boltzmann constant, k = 1.3806503 × 10⁻⁶. -23 J / K; T represents Kelvin temperature; X represents the position parameter vector to be optimized in the single diode model.

[0074] 2) Dual-diode model:

[0075]

[0076] Among them, I sd1 and I sd2 represents the saturation current of the first and second diodes, respectively; n1 and n2 represent the ideality factors of the first and second diodes, respectively. X represents the position parameter vector to be optimized in the dual-diode model.

[0077] 3) Objective function evaluation

[0078] For the two models above, this embodiment uses the root mean square error (RMSE) as the objective function to measure the overall error of measuring and calculating the current:

[0079]

[0080] In the formula, K represents the number of voltage-current measurements.

[0081] To demonstrate the reliability of the method of this invention, the Backtracking Search Algorithm (BSA), the Performance-Oriented JAYA Algorithm (PGJAYA), the Marine Predator Algorithm (MPA), and the Primitive Flower Pollination Algorithm (FPA) were selected for comparison. All methods were set to a population size of 30, a maximum number of evaluations of 30,000, and were run independently 30 times.

[0082] The following analysis, based on relevant data, examines the method and framework provided by this invention:

[0083] Table 1

[0084]

[0085] The best results in Table 1 are highlighted in bold. As shown in Table 1, in both single-diode and dual-diode models, the method of this invention achieves more accurate and stable results than other comparative methods across various statistical indicators. Specifically, in the single-diode model, PGJAYA achieves similar results to the method of this invention in terms of optimal results, but considering the median, mean, and standard deviation, the method of this invention performs better overall than PGJAYA. In the dual-diode model, the method of this invention yields results similar to other comparative methods, but with improvements. Furthermore, for the photovoltaic cell model parameter identification problem, even a small improvement is significant. In conclusion, the method of this invention provides a more reliable solution than other comparative methods for both single-diode and dual-diode model parameter identification problems.

[0086] As shown in Figure 2(a), for the single-diode model, the data distribution of the other comparative methods, except for the PGJAYA method and the method of this invention, exhibits varying degrees of fluctuation. Furthermore, the MPA method and the PGJAYA method show different numbers of outliers, while the data distribution of the method of this invention is dense and stable. From the overall distribution of the box plots, the method of this invention demonstrates excellent reliability.

[0087] As shown in Figure 2(b), for the dual-diode model, the data distribution of the other comparative methods, except for the PGJAYA method and the method of this invention, exhibits varying degrees of fluctuation. Furthermore, the MPA method and the PGJAYA method show different numbers of outliers, while the data distribution of the method of this invention is dense and stable. From the overall distribution of the box plots, the method of this invention demonstrates excellent reliability.

[0088] Table 2

[0089]

[0090] As shown in Table 2, the absolute error values ​​for current are between -2.50741270E-03 and 1.61722183E-03, and the absolute error values ​​for power are between -1.46257383E-03 and 9.01327437E-04. These results further demonstrate the accuracy of the method of the present invention in parameter identification.

[0091] As shown in Figures 3(a) and 3(b), the calculated data obtained through the model parameters fits the observed data very well. The results demonstrate that the method of this invention can accurately identify photovoltaic model parameter identification problems and is an effective solution tool.

[0092] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying photovoltaic cell model parameters based on an improved flower pollination algorithm, characterized in that, The method includes: S1. Receive the settings for the photovoltaic cell model type and flower pollination algorithm parameters, determine the composition of the parameters to be identified based on the set photovoltaic cell model type, use them as the position of individual pollen, and initialize the pollen population; S2. An improved flower pollination algorithm is used for population iteration, and the global optimal solution and the corresponding pollen location are output as the parameter identification results of the photovoltaic cell model; In the improved flower pollination algorithm, when updating the position of individual pollen in global pollination during population iteration, the parameter controlling the pollination intensity is a Gaussian distributed random number. When updating the position of individual pollen in local pollination, the positions of two pollen in the population that are different from the current pollen come from different clusters. The clusters are two sets divided by sorting the individual fitness from small to large. During local pollination, the pollen location The update formula is as follows: in, For the first Evaluation of the second-order function The position vector of each individual For clusters with low fitness, For cluster sets with high fitness, and From respectively and The index corresponding to a randomly selected individual in the dataset; After obtaining the new individual's location, it is selected from the cluster set with higher fitness. A random individual is selected from the pool for a local search using a chaotic elite strategy, as follows: If the ratio of the current number of evaluations to the maximum number of evaluations is greater than a random number in the interval [0,1], then a chaotic random number is generated. And perform a local search at the globally optimal location. Otherwise, the individual's historical best position is used. ,in, For the first The values ​​of the sub-chaotic sequence, initial value A random number uniformly distributed in the interval [0,1]. A random number between [3.57, 4.0]; For the first The second iteration The individual in the first The position of the dimension For the first Evaluation of the second-order function The individual in the first The individual's historical best position in the dimension. For randomly selected individual index positions, A random number distributed in the interval [0,1]. After obtaining the new individual position, an adaptive boundary handling strategy is used to adjust unreasonable positions in the individual's evolutionary process, as follows: in, For the first The second iteration The individual in the first The position of the dimension For the problem space The lower bound of the dimension, For the problem space The upper limit of dimensions, This is the midpoint between the upper and lower boundaries of the problem.

2. The method as described in claim 1, characterized in that, During global pollination, the pollen location The update formula is as follows: in, The globally optimal individual position corresponds to the position of the individual with the lowest fitness in the current population. This indicates that the mean is 0 and the standard deviation is 0. Gaussian distributed random numbers.

3. The method as described in claim 1, characterized in that, If the photovoltaic cell model type is a single diode model, then the parameters to be identified are: If the photovoltaic cell model type is a dual-diode model, then the parameters to be identified are: ; in, This represents photovoltaic current generated by solar energy; Indicates the diode short-circuit current; and These represent series and parallel resistors, respectively. Indicates the diode's ideality factor; and These represent the saturation currents of the first and second diodes, respectively. and These represent the ideal factors of the first and second diodes, respectively.

4. The method as described in claim 1, characterized in that, The fitness function is the root mean square error of the observed current and the calculated current.

5. A photovoltaic cell model parameter identification system based on an improved flower pollination algorithm, characterized in that, include: Processor and memory; The memory is used to store computer-executed instructions; The processor is configured to execute the computer execution instructions, causing the method described in any one of claims 1 to 4 to be executed.

Images