Photovoltaic model parameter identification method and system based on differential evolution algorithm

Abstract

The invention provides a photovoltaic model parameter identification method and system based on a differential evolution algorithm, and relates to the technical field of optimization control. The method specifically comprises the following steps: establishing a photovoltaic model, and determining unknown parameters to be identified in the photovoltaic model; taking unknown parameters to be identified in the photovoltaic model as individuals, performing individual optimization by adopting a double-strategy collaborative deployment framework combining a probability-based test vector generation strategy and a parameter adaptive scheme, and updating stagnant individuals by utilizing a gradient descent strategy in the optimization process to obtain optimal individuals; and taking the parameter value corresponding to the optimal individual as the optimal parameter identified in the photovoltaic model. According to the method, a probability-based test vector generation strategy and a parameter adaptive method are combined, and a differential evolution algorithm is improved by combining a diversity improvement strategy, so that the optimization problem of accurate identification of parameters in a photovoltaic model is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optimal control, and in particular to a photovoltaic model parameter identification method and system based on a differential evolution algorithm. BACKGROUND

[0002] Excessive fossil fuel combustion has caused many environmental problems, and the use of renewable energy, such as solar energy, is expected to be a solution to these environmental problems. The photovoltaic model system can convert solar energy into electrical energy, providing important support for the use of renewable energy. However, in practical applications, the photovoltaic model system still faces many challenges, of which high cost, instability, and low efficiency are the main problems. The performance of solar cells depends on weather conditions and physical properties, which makes the photovoltaic model system have great uncertainty. Therefore, in order to better study and optimize the photovoltaic model system, an accurate model of the photovoltaic cell needs to be established for simulation.

[0003] Fast and accurate parameter identification in the photovoltaic model is crucial for computing, control, and processing photovoltaic power generation systems. In the typical configuration of the photovoltaic model, it contains diodes and resistors. As the number of diodes in the model increases, the number of unknown parameters also increases, and the simulation accuracy of the photovoltaic model depends on these key parameters. Therefore, it is necessary to apply optimization algorithms to obtain accurate photovoltaic model parameters.

[0004] However, the parameter identification problem has the characteristics of multimodality, nonlinearity, and non-convexity. Many optimization algorithms often have problems such as premature convergence and population stagnation when applied to this model. These characteristics pose challenges to existing optimization algorithms. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application combines a probability-based test vector generation strategy and a parameter adaptive method, and based on a dual-strategy collaborative deployment framework and a diversity-improved differential evolution algorithm, a photovoltaic model parameter identification method and system based on a differential evolution algorithm are proposed, which is suitable for the optimization problem of accurate identification of parameters in the photovoltaic model.

[0006] In one aspect, the present application proposes a photovoltaic model parameter identification method based on a differential evolution algorithm, which includes the following processes:

[0007] Establish a photovoltaic model and determine the unknown parameters to be identified in the photovoltaic model;

[0008] The unknown parameters to be identified in the photovoltaic model are taken as individuals, and a dual-strategy collaborative deployment framework combining a probability-based test vector generation strategy and a parameter adaptive scheme is used for individual optimization. In the optimization process, the gradient descent strategy is used to update the stagnation individuals to obtain the optimal individuals;

[0009] The parameter value corresponding to the optimal individual is taken as the optimal parameter identified in the photovoltaic model.

[0010] Further, the photovoltaic model comprises a single-diode model, a double-diode model, a triple-diode model, and a photovoltaic module model.

[0011] The mathematical model of the single-diode model is:

[0012]

[0013] wherein is the output current of the single-diode model; represents a photo-generated current; represents a reverse saturation current of a diode ; is the charge of an electron, and ; represents a terminal voltage; represents a resistance value of a series resistor; represents an ideal diode coefficient; is the Boltzmann constant, and ; represents a Kelvin temperature of a cell; represents a resistance value of a shunt resistor; the unknown parameters to be identified in the single-diode model comprise: , , , and ;

[0014] The mathematical model of the double-diode model is:

[0015]

[0016] wherein is the output current of the double-diode model; represents a current of a diode ; represents a current of a diode ; represents a reverse saturation current of a diode ; represents a reverse saturation current of a diode ; represents an ideal diode coefficient of a diode ; represents an ideal diode coefficient of a diode ; the unknown parameters to be identified in the double-diode model comprise: , , 、 、 、 and ;

[0017] The mathematical model of the triple diode model is:

[0018]

[0019] wherein represents the output current of the triple diode model; represents the current of the diode ; represents the saturation current of the diode ; represents the ideal diode coefficient of the diode ; the unknown parameters to be identified in the triple diode model include: 、 、 、 、 、 、 、 and ;

[0020] The mathematical model of the photovoltaic module model is:

[0021]

[0022] wherein represents the output current of the photovoltaic module model; represents the number of parallel diodes; represents the number of series diodes; the unknown parameters to be identified in the photovoltaic module model include: 、 、 、 and .

[0023] Further, the specific content of the individual optimization of the dual-strategy cooperative deployment framework combining the probability-based test vector generation strategy and the parameter adaptive scheme is:

[0024] Step A1: generate an initial population with a population size of , set the maximum number of fitness evaluations , and initialize the current number of fitness evaluations ;

[0025] Step A2: Calculate the fitness value of each individual in the current population, rank the individuals based on their fitness values, calculate the selection probability of each individual based on its fitness ranking, and standardize it to obtain the standardized selection probability of that individual; simultaneously, update the current number of individuals in the fitness assessment. ;

[0026] Step A3: Initialize a memory pool to store parameter pairs that have successfully generated test vectors, and calculate the scaling factor for the current iteration based on the parameter pairs stored in the memory pool. and crossover probability The parameter pair mentioned above includes: a parameter for guiding the scaling factor. Parameters generated and used to guide crossover probabilities Parameters generated ;

[0027] Step A4: Perform a selection operation on the current population based on the standardized selection probabilities of all individuals, and apply the scaling factor based on the current iteration round. Perform mutation operations on the selected individuals, and then determine the crossover probability based on the current iteration round. Perform a crossover operation on the mutated individuals to generate the initial experimental vector for the current iteration round;

[0028] Step A5: Update the initial test vector for the current iteration based on the perturbation strategy equation to obtain the test vector for the current iteration;

[0029] Step A6: Calculate the fitness value of the trial vector in the current iteration round and update the current number of fitness evaluations. ;

[0030] Step A7: Determine whether the fitness value of the experimental vector is better than that of the target vector. If so, the experimental vector is considered successful, and the experimental vector is used as the individual after the current iteration. The parameter pairs in the memory pool are updated using the weighted Lehmer mean method. If not, the target vector is retained as the individual after the current iteration.

[0031] Step A8: Update the current population based on the individuals updated in the current iteration round, and use the gradient descent strategy to update the stagnant individuals in the current population to generate the offspring population;

[0032] Step A9: Determine the current number of fitness assessments. Is it greater than the maximum number of fitness assessments? If the offspring population is selected as the current population, the process returns to step A2; otherwise, the individual with the best fitness value in the offspring population is selected as the optimal individual.

[0033] Further, the scaling factor of the current iteration round in step A3 is calculated as follows: And the crossover probability is calculated as follows:

[0034] For any individual , a parameter pair is randomly selected from the memory pool , where represents the th position in the memory pool ; represents the th position in the memory pool ;

[0035] Based on the selected parameter pair , the current number of fitness evaluations is compared with the preset threshold , and the scaling factor of the individual is calculated; at the same time, the crossover probability of the individual is calculated.

[0036] According to the parameter constraints, the scaling factor and the crossover probability calculated for the individual are constrained to generate the scaling factor and the crossover probability of the individual in the current iteration round.

[0037] Further, the perturbation strategy equation in step A5 is as follows:

[0038]

[0039] where represents the trial vector of the individual in the current iteration round ; represents the target vector of the individual in the current iteration round ; represents the initial trial vector of the individual in the current iteration round ; represents the cosine perturbation coefficient, and has:

[0040]

[0041] where represents a randomly generated uniformly distributed random number.

[0042] Further, the method for updating the parameter pair in the memory pool by using the weighted Lehmer mean method in step A7 is as follows: ​

[0043] For all the test vectors considered successful, the scaling factor and the crossover probability of the test vector are recorded into a scaling factor set and a crossover probability set respectively;

[0044] Each parameter pair in the memory pool is evaluated on the performance in history respectively, and a weight is assigned to each parameter pair according to the evaluation result;

[0045] According to the weights of all parameter pairs in the memory pool, the weighted Lehmer mean of the scaling factor set and the weighted Lehmer mean of the crossover probability set are calculated respectively;

[0046] A parameter pair is randomly selected from the memory pool, if the scaling factor set is not empty, the scaling factor in the parameter pair is updated based on the historical information according to the weighted Lehmer mean , if the scaling factor set is empty, the current scaling factor is kept unchanged;

[0047] If the crossover probability set is not empty, the crossover probability in the parameter pair is updated based on the historical information according to the weighted Lehmer mean , if the crossover probability set is empty, the current crossover probability is kept unchanged.

[0048] Further, the specific content of generating the offspring population by updating the stagnation individuals in the current population using the gradient descent strategy in step A8 is as follows:

[0049] For any dimension in the current population, the average value and the standard deviation of the dimension are calculated, and it is determined whether the dimension converges;

[0050] For any individual in the current population, the stagnation index of the individual is calculated, and the stagnation index of the individual is compared with a predetermined threshold, and the individual whose stagnation index exceeds the predetermined threshold is regarded as a stagnation individual;

[0051] For any convergent dimension, the number of stagnation individuals in the current population under the dimension is counted , if exceeds a preset value of the number of stagnation individuals, gradient descent update is performed on all the stagnation individuals in the current population, and the offspring population is generated according to the updated individuals.

[0052] ​​In another aspect, the application provides a photovoltaic model parameter identification system based on a differential evolution algorithm, comprising:

[0053] A model construction module is configured to establish a photovoltaic model and determine unknown parameters to be identified in the photovoltaic model.

[0054] An individual optimization module is configured to obtain the unknown parameters to be identified in the photovoltaic model as individuals, perform individual optimization by using a dual-strategy co-deployment framework combining a probability-based trial vector generation strategy and a parameter adaptive scheme, and update a stagnation individual by using a gradient descent strategy during the optimization process to obtain an optimal individual.

[0055] An optimal parameter determination module is configured to determine a parameter value corresponding to the optimal individual as an optimal parameter identified in the photovoltaic model.

[0056] In a third aspect, the application provides an electronic device, comprising one or more processors and a memory configured to store instructions that, when executed by the one or more processors, cause the one or more processors to perform the photovoltaic model parameter identification method based on the differential evolution algorithm.

[0057] In a fourth aspect, the application provides a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the photovoltaic model parameter identification method based on the differential evolution algorithm.

[0058] In a fifth aspect, the application provides a computer program product comprising a computer program or instructions that, when executed by a processor, implement the photovoltaic model parameter identification method based on the differential evolution algorithm.

[0059] The above technical solutions have the following beneficial effects:

[0060] The method provides an optimization algorithm for improving the search capability of a differential evolution algorithm (DE) variant and its application in accurate identification of parameters of a photovoltaic model, i.e., a differential evolution algorithm with a bi-strategy co-deployment framework and diversity improvement (BDDE), which balances exploration and development by using a dual-strategy co-deployment framework and a diversity improvement strategy.

[0061] The method constructs a dual-strategy deployment framework, which utilizes a probability-based trial vector generation strategy and a parameter mapping method to utilize their respective advantages, wherein the population-based strategy improves the search performance of the algorithm by promoting a diversity mechanism that prevents the population from stagnating in local optimal values; the parameter-based strategy dynamically adjusts key DE parameters to ensure a balanced trade-off between early global search and late fine utilization. The method also proposes a diversity improvement strategy based on gradient descent, while measuring the diversity level and stagnation detection. The gradient descent method is used to dynamically monitor stagnation individuals with low diversity levels and apply a gradient design scheme update to guide the individuals out of stagnation and increase population diversity. Tests show that BDDE is significantly better than other existing algorithms in the parameter accurate identification application of photovoltaic models, achieving highly competitive performance.

[0062] Due to the characteristics of multi-modal, non-linear, and non-convexity of the parameter identification problem of photovoltaic models, an optimization algorithm needs to be used to accurately identify the parameters thereof. The method is directed to the above characteristics, and realizes a breakthrough from a theoretical method to an industrial landing by innovative improvement of the differential evolution algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 A flowchart of a photovoltaic model parameter identification method based on a differential evolution algorithm in the embodiment;

[0064] Figure 2 A circuit diagram of a photovoltaic model in the embodiment; in the diagram: (a) is a circuit diagram of a single-diode SDM model; (b) is a circuit diagram of a double-diode DDM model; (c) is a circuit diagram of a triple-diode TDM model; (d) is a circuit diagram of a PV component model;

[0065] Figure 3 A flowchart of a differential evolution algorithm based on a dual-strategy cooperative deployment framework and diversity improvement in the embodiment;

[0066] Figure 4 A structure diagram of a photovoltaic model parameter identification system based on a differential evolution algorithm in the embodiment. DETAILED DESCRIPTION

[0067] In order to facilitate the understanding of the present application, the specific embodiments of the present application are further described in detail below in combination with the drawings and embodiments. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0068] In the present embodiment, the software Matlab R2021a is applied on a personal computer with an Intel i7-12700k processor, a CPU frequency of 3.2 GHz and a memory capacity of 32 GB to run the BDDE algorithm for each photovoltaic model.

[0069] Embodiment 1

[0070] A photovoltaic model parameter identification method based on a differential evolution algorithm according to the present embodiment, as shown in Figure 1 , comprises the following processes:

[0071] A photovoltaic model is established, and unknown parameters to be identified in the photovoltaic model are determined.

[0072] The photovoltaic model comprises a single diode (SDM) model, a double diode (DDM) model, a triple diode (TDM) model and a photovoltaic (PV) component model.

[0073] As shown in Figure 2 , the single diode model is represented as:

[0074] The current of the diode is calculated, and is represented as:

[0075] (1)

[0076] wherein represents the reverse saturation current of the diode ; is the charge of an electron, and ; represents the terminal voltage; represents the resistance value of the series resistance; is the output current of the SDM; represents the ideal diode coefficient; is the Boltzmann constant, and ; represents the Kelvin temperature of the battery.

[0077] The shunt resistor current is calculated, and is represented as:

[0078] (2)

[0079] wherein represents the resistance value of the shunt resistor. ​

[0080] Diode-based current and shunt resistor current , the output current of the SDM is calculated as:

[0081] (3)

[0082] where represents the photo-generated current.

[0083] Substitute the expressions of and into equation (3), the output current of the SDM is calculated as:

[0084] (4)

[0085] where , , , and are five unknown parameters to be identified.

[0086] As shown in Figure 2 (b), the double-diode model is represented as:

[0087] The output current of the DDM is calculated as , which is represented as:

[0088] (5)

[0089] where represents the current of diode ; represents the current of diode ; represents the reverse saturation current of diode , which is used to describe the saturation current; represents the reverse saturation current of diode , which is used to describe the diffusion current; represents the ideal diode coefficient of diode , which is called the recombination diode factor; represents the ideal diode coefficient of diode , which is called the ideal diffusion coefficient; and are two other unknown parameters based on the SDM and the DDM.

[0090] As shown in Figure 2 (c), the triple-diode model is represented as:

[0091] The output current of the TDM is calculated as​ is:

[0092] (6)

[0093] wherein denotes the current of the diode ; denotes the saturation current of the diode ; denotes the ideal diode factor of the diode ; and denote two unknown parameters based on the DDM.

[0094] As Figure 2 (d) shows, the photovoltaic module model is represented as:

[0095] The output current of the PV module model is calculated:

[0096] (7)

[0097] wherein denotes the output current of the PV module model; denotes the number of parallel diodes; denotes the number of series diodes; , , , , a is five unknown parameters.

[0098] The unknown parameters to be identified in the photovoltaic model are taken as individuals, and a bi-strategy co-deployment framework combining a probability-based trial vector generation strategy and a parameter adaptive scheme is used for individual optimization. In the optimization process, the gradient descent strategy is used to update the stagnation individuals to obtain the optimal individual.

[0099] In the present embodiment, as Figure 3 shown, first, a bi-strategy co-deployment framework (CF) is constructed, which combines a probability-based trial vector generation strategy and a novel parameter adaptive method to take advantage of their respective advantages; secondly, the present embodiment proposes a diversity improvement strategy based on gradient descent to enrich the diversity of the population. The diversity level of the population is measured to detect the stagnation of individuals in the population, and the gradient descent method is used to guide the individuals out of stagnation and increase the diversity of the population. In addition, the population diversity of the BDDE variant is visualized for exploration and development analysis to illustrate the impact of its components.

[0100] The specific content of the individual optimization by the double-strategy cooperative deployment framework combining the probability-based test vector generation strategy and the parameter adaptive scheme is:

[0101] Step A1: generating an initial population with a population size of , denoted as , wherein represents the first individual, represents the second individual; represents the th individual, setting the maximum number of fitness evaluations , and initializing the current number of fitness evaluations .

[0102] In this embodiment, an initial population is generated, at this time, the current number of fitness evaluations is 0, the fitness function is used to calculate the population fitness value, and then based on the fitness value of the current population , the individuals are ranked, and the ranking of the individual is used as the new index value of the individual.

[0103] Step A2: calculating the fitness value of each individual in the current population, ranking the individuals according to the fitness values of all individuals in the current population, and then calculating the selection probability of each individual according to the fitness ranking of the individual and normalizing to obtain the normalized selection probability of the individual; at the same time, updating the current number of fitness evaluations .

[0104] The selection probability of the individual and the normalized selection probability of the individual are represented as:

[0105] (8)

[0106] , wherein represents the ranking of the individual; represents the selection probability of the individual with the ranking of ; represents the normalized selection probability of the individual with the ranking of ; represents the sum of the selection probabilities of all individuals in the current population.

[0107] In this embodiment, it is necessary to change the scaling factor for optimization in the early and late stages of differential evolution, that is, in the early stage of DE, the value of is increased to improve population diversity, and in the late stage, the value of is reduced to optimize the solution.

[0108] Step A3: initialize a memory pool for storing parameter pairs of successfully generated trial vectors, and calculate the scaling factor at the current iteration round based on the parameter pairs stored in the memory pool and the crossover probability .

[0109] the scaling factor at the current iteration round and the crossover probability are calculated as follows:

[0110] the parameter pair includes: parameters for guiding the scaling factor generated by the mutation operation and parameters for guiding the crossover probability generated by the crossover operation .

[0111] for any individual , a parameter pair is randomly selected from the memory pool , wherein represents the th position in the memory pool ; represents the th position in the memory pool .

[0112] In this embodiment, for the initially empty memory pool, a reasonable default value is generally set for calculation when calculating the scaling factor of the individual; at this time, the crossover probability of the individual is 0, that is, when there is no historical information, the mutation operation is completely relied on for exploration, avoiding invalid crossover.

[0113] based on the selected parameter pair , the scaling factor of the individual is calculated by comparing the current number of fitness evaluation with the preset threshold , and is expressed as:

[0114] (9)

[0115] wherein represents the threshold between the two evolution stages of generating , and in this embodiment, the threshold is taken as ; is the natural logarithm; represents a random number generated in accordance with the Cauchy distribution; is a disturbance term, and ; when the first condition is satisfied, controls the convergence, randomness is introduced into the generation of F, and the Cauchy distribution function is applied to accelerate the convergence speed.

[0116] At the same time, calculate the crossover probability of this individual. , is represented as:

[0117] (10)

[0118] in This indicates the generation of random numbers that follow a normal distribution.

[0119] Based on the parameter constraints, the scaling factor calculated for this individual and crossover probability Apply constraints to generate the scaling factor and crossover probability for the individual in the current iteration round.

[0120] The parameter constraints are expressed as follows:

[0121] (11)

[0122] (12)

[0123] in This represents the scaling factor after the individual is constrained. This represents the crossover probability of the individual after constraints. This indicates taking the maximum value; This indicates taking the minimum value.

[0124] Step A4: Perform a selection operation on the current population based on the standardized selection probabilities of all individuals, and apply the scaling factor based on the current iteration round. Perform mutation operations on the selected individuals, and then determine the crossover probability based on the current iteration round. Perform a crossover operation on the mutated individuals to generate the initial experimental vector for the current iteration round.

[0125] Step A5: Update the initial test vector for the current iteration based on the perturbation strategy equation to obtain the test vector for the current iteration.

[0126] The perturbation strategy equation is:

[0127] (13)

[0128] in Represents an individual In the current iteration round The test vector below; Represents an individual In the current iteration round The target vector is determined by the individual to be updated in the current population during each iteration. representing an individual the initial trial vector under the current iteration round; representing a cosine perturbation coefficient, and has:

[0129] (14)

[0130] wherein is a randomly generated uniformly distributed random number; when approaches 1 or -1, the mutation operation increases the search step, and when approaches 0, the mutation step becomes smaller.

[0131] Step A6: Calculate the fitness value of the trial vector under the current iteration round, and update the current number of fitness evaluations .

[0132] Step A7: Determine whether the fitness value of the trial vector is better than the fitness value of the target vector, if yes, consider that the trial vector is successful, take the trial vector as the updated individual under the current iteration round, and update the parameter pair in the memory pool using the weighted Lehmer mean method; if not, keep the target vector as the updated individual under the current iteration round.

[0133] In the present embodiment, the parameter pair in the memory pool is updated at the end of each F value generation. Based on the historical information parameter adaptive strategy, the parameter pair randomly selected in the memory pool is updated using the weighted Lehmer mean method.

[0134] The specific content of the parameter pair in the memory pool being updated using the weighted Lehmer mean method is:

[0135] For all trial vectors considered successful, the scaling factor and the crossover probability for generating the trial vector are recorded in the scaling factor set and the crossover probability set respectively.

[0136] The performance of each parameter pair in the memory pool in history is evaluated respectively, and a weight is assigned to each parameter pair according to the evaluation result.

[0137] According to the weights of all parameter pairs in the memory pool, the weighted Lehmer average of the scaling factor set is calculated using the weighted Lehmer mean method , and the weighted Lehmer average of the crossover probability set is calculated using the weighted Lehmer mean method .

[0138] A parameter pair is randomly selected from the memory pool, and if the scaling factor set ​is not empty, the scaling factor in the parameter pair is updated based on the weighted Lehmer average , and if the scaling factor set is empty, the current scaling factor is kept unchanged.

[0139] If the crossover probability set is not empty, the crossover probability in the parameter pair is updated based on the weighted Lehmer average , and if the crossover probability set is empty, the current crossover probability is kept unchanged.

[0140] (15)

[0141] (16)

[0142] wherein is the weight of the individual ; represents the calculation of the weighted Lehmer average; represents the th parameter in the scaling factor set ; represents the fitness value of the target vector of the individual in the current iteration round ; represents the fitness value of the trial vector of the individual in the current iteration round ; represents the scaling factor updated based on the historical information ; represents the current scaling factor based on the historical information ; represents the th parameter in the crossover probability set ; represents the crossover probability updated based on the historical information ; represents the current crossover probability based on the historical information .

[0143] Step A8: update the current population according to the individuals updated in the current iteration round, and update the stagnated individuals in the current population by using the gradient descent strategy to generate the offspring population.

[0144] The present embodiment proposes a diversity improvement strategy, uses the gradient descent strategy to disturb the stagnated individuals therein, and promotes the stagnated individuals to jump out of the local optimal value.

[0145] The specific steps for updating stagnant individuals in the current population using a gradient descent strategy to generate the offspring population are as follows:

[0146] For any dimension in the current population, calculate the mean and standard deviation of that dimension, and determine whether the dimension has converged.

[0147] Since the convergence of a dimension in a population depends on the relationship between the mean and the standard deviation, this implementation method applies this principle to each dimension. , together Dimension, calculate the mean and standard deviation of this dimension in the population:

[0148] (17)

[0149] (18)

[0150] in Indicates the dimension in the current population The average value; Representing dimensions The first Individual; Indicates the dimension in the current population Standard deviation; Representing dimensions Whether it converges; if the mean is greater than the standard deviation, then mark it. This indicates that the dimension may converge; otherwise, it is marked as convergent. , indicating that the dimension does not converge.

[0151] For any individual in the current population, calculate the stagnation index of that individual and compare it with a predetermined threshold. Individuals whose stagnation index exceeds the predetermined threshold are identified as stagnant individuals.

[0152] The stagnation index is calculated as follows:

[0153] (19)

[0154] in Represents an individual Stagnation indicators; Represents an individual In the current iteration round The target vector below fitness value; Represents an individual Generated test vectors The fitness value.

[0155] For any convergent dimension, count the number of stagnant individuals in that dimension in the current population. ,like If the number of stagnant individuals exceeds a preset value, then gradient descent is used to update all stagnant individuals in the current population, and a offspring population is generated based on the updated individuals.

[0156] In this embodiment, if an individual The stagnation indicator counter exceeds the predetermined threshold. Then the individual If it is not the current globally optimal solution, the number of times `counte` is called needs to be counted, defined as... If in the convergence dimension The number exceeds For stagnant individuals Perform gradient descent updates and calculate the numerical gradient:

[0157] (20)

[0158] in Representing dimensions gradient information; Indicates the gradient calculation step size; Represents a vector; It is a unit vector; For learning rate; It is a constant; It is a random number; For the updated individual; This represents the gradient vector; in this implementation, the gradient calculation step size is initialized. Using gradient information and learning rate Individuals are updated using gradient descent to ensure they remain within range. inside, use To constrain new individuals .

[0159] Step A9: Determine the current number of fitness assessments. Is it greater than the maximum number of fitness assessments? If the offspring population is selected as the current population, the process returns to step A2; otherwise, the individual with the best fitness value in the offspring population is selected as the optimal individual.

[0160] The parameter value corresponding to the best individual is used as the optimal parameter identified in the photovoltaic model.

[0161] In this embodiment, to verify the effectiveness of the photovoltaic model parameter identification method based on differential evolution algorithm, the parameters of each photovoltaic model were optimized using the above process. Then, using existing experimental datasets, the error function value of each photovoltaic model was calculated to evaluate the accuracy of the model predictions. The error function for each photovoltaic model is as follows:

[0162] The error function for the single-diode SDM model is:

[0163] (21)

[0164] wherein denotes the error function; denotes a set of parameters of the single-diode SDM model, .

[0165] The error function for the double-diode DDM model is:

[0166] (22)

[0167] wherein denotes a set of parameters of the single-diode SDM model, .

[0168] The error function for the triple-diode TDM model is:

[0169] (23)

[0170] wherein denotes a set of parameters of the single-diode SDM model, .

[0171] The error function for the photovoltaic PV component model is:

[0172] (24)

[0173] wherein denotes a set of parameters of the single-diode SDM model, .

[0174] Based on the error function of each photovoltaic model, the root mean square error value of each photovoltaic model is calculated:

[0175] (25)

[0176] wherein X is the solution vector and N is the amount of experimental data.

[0177] In the embodiment, the root mean square error values of each photovoltaic model are as follows: the root mean square error value of SDM is 9.86021877891447E-04; the root mean square error value of DDM is 9.82484851785080E-04; and the root mean square error value of TDM is 9.82995105171601E-04. In addition, the embodiment also evaluates the model of a specific type of photovoltaic module, and the root mean square error values are as follows: the root mean square error value of Photowatt-PWP201 is 2.42507486809513E-03; the root mean square error value of STP6-120 / 36 is 1.66006031250854E-02; and the root mean square error value of STM6-40 / 36 is 1.72981370994069E-03.

[0178] By comparing the above root mean square error values, it can be seen that the double diode model DDM has the highest prediction accuracy under these test conditions, as it has the smallest root mean square error value. The root mean square error values of other models are relatively high, especially the STP6-120 / 36 type, which has the highest root mean square error value, indicating that the prediction accuracy needs to be improved. These results are of great significance for selecting a suitable photovoltaic model for accurate photovoltaic system design and performance prediction, and also verify the effectiveness of the photovoltaic model parameter identification method based on the differential evolution algorithm proposed in the embodiment.

[0179] Example 2:

[0180] The embodiment proposes a photovoltaic model parameter identification system based on a differential evolution algorithm, as shown in Figure 4 , which includes:

[0181] A model construction module is used to establish a photovoltaic model and determine unknown parameters to be identified in the photovoltaic model.

[0182] An individual optimization module is used to obtain unknown parameters to be identified in the photovoltaic model as individuals, and a double-strategy collaborative deployment framework combining a probability-based trial vector generation strategy and a parameter adaptive scheme is used for individual optimization. In the optimization process, a gradient descent strategy is used to update stagnant individuals to obtain optimal individuals.

[0183] An optimal parameter determination module is used to determine the parameter values corresponding to the optimal individuals as the optimal parameters identified in the photovoltaic model.

[0184] Example 3:

[0185] The embodiment provides an electronic device, comprising one or more processors and a memory for storing instructions which, when executed by the one or more processors, cause the one or more processors to perform the photovoltaic model parameter identification method based on a differential evolution algorithm.

[0186] The electronic device can be a mobile phone, a computer or a tablet computer, and the like, comprising a memory and a processor, and the memory stores a computer program which, when executed by the processor, implements the photovoltaic model parameter identification method based on a differential evolution algorithm as described in the embodiments. It can be understood that the electronic device can further comprise an input / output (I / O) interface and a communication component.

[0187] The processor is configured to perform all or part of the steps of the photovoltaic model parameter identification method based on a differential evolution algorithm as described in the above embodiments. The memory is configured to store various types of data, which can include, for example, instructions of any application program or method in the electronic device, and application program related data.

[0188] The processor can be an Application Specific Integrated Cricuit (ASIC), a Digital Signal Processor (DSP), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic elements, and is configured to perform the photovoltaic model parameter identification method based on a differential evolution algorithm as described in the above embodiments.

[0189] Embodiment 4

[0190] The embodiment provides a computer readable storage medium storing executable instructions which, when executed, can be stored in a computer readable storage medium in the form of a software functional unit and sold or used as an independent product.

[0191] The computer software product is stored in a storage medium and comprises a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the photovoltaic model parameter identification method based on a differential evolution algorithm as described in the embodiments.

[0192] The aforementioned storage medium includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD (Secure Digital Memory Card) or a DX (an abbreviation of Memory Data Register, MDR) memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an APP (an abbreviation of Application) application store, etc., which can store a program check code, and stores a computer program thereon, the computer program being executed by a processor to implement each step of the photovoltaic model parameter identification method based on a differential evolution algorithm as described above.

[0193] Embodiment 5:

[0194] The embodiment provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the photovoltaic model parameter identification method based on a differential evolution algorithm.

[0195] Based on such understanding, the technical solution of the present application or the part of the technical solution which essentially contributes to the prior art or the part of the technical solution can be embodied in the form of a computer program product.

[0196] Each embodiment in the present application is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the difference from other embodiments.

[0197] The scope of protection of the present application is not limited to the above-described embodiments. Obviously, those skilled in the art can make various modifications and variations to the present disclosure without departing from the scope and spirit of the present disclosure. If these modifications and variations belong to the scope of the present disclosure and its equivalent technology, the present disclosure also includes these modifications and variations.

Claims

1. A photovoltaic model parameter identification method based on a differential evolution algorithm, characterized in that, The method comprises the following processes: A photovoltaic model is established, and unknown parameters to be identified in the photovoltaic model are determined; The unknown parameters to be identified in the photovoltaic model are taken as individuals, and an individual optimization is performed by using a double-strategy collaborative deployment framework combining a probability-based trial vector generation strategy and a parameter adaptive scheme, and in the optimization process, a gradient descent strategy is used to update stagnation individuals to obtain an optimal individual; Parameter values corresponding to the optimal individual are taken as optimal parameters identified in the photovoltaic model. 2.The photovoltaic model parameter identification method based on differential evolution algorithm according to claim 1, characterized in that, The photovoltaic model comprises a single-diode model, a double-diode model, a triple-diode model and a photovoltaic module model; The mathematical model of the single-diode model is: ； wherein is the output current of the single diode model; represents the photo-generated current; represents the reverse saturation current of the diode ; is the charge of one electron, and ; represents the terminal voltage; represents the resistance value of the series resistance; represents the ideal diode factor; is the Boltzmann constant, and ; represents the Kelvin temperature of the battery; represents the resistance value of the shunt resistor; the unknown parameters to be identified in the single diode model include: , , , and ; The mathematical model of the double-diode model is: ； wherein represents the output current of a double diode model; represents the current of a diode ; represents the current of a diode ; represents the reverse saturation current of a diode ; represents the reverse saturation current of a diode ; represents the ideal diode factor of a diode ; represents the ideal diode factor of a diode ; the unknown parameters to be identified in the double diode model include: , , , , , and ; The mathematical model of the triple-diode model is: ； wherein represents the output current of a triple diode model; represents the current of a diode ; represents the saturation current of a diode ; represents the ideal diode factor of a diode ; the unknown parameters to be identified in the triple diode model include: , , , , , , , and ; The mathematical model of the photovoltaic module model is: ； wherein represents an output current of a photovoltaic module model; represents a number of parallel diodes; represents a number of serial diodes; unknown parameters to be identified in the photovoltaic module model, including: , , , and . 3.The photovoltaic model parameter identification method based on differential evolution algorithm according to claim 2, characterized in that, The specific content of the individual optimization by using the double-strategy collaborative deployment framework combining the probability-based trial vector generation strategy and the parameter adaptive scheme is: Step Al: Create an initial population of size Set the maximum number of fitness evaluations and initialize the current number of fitness evaluations ; Step A2: calculate the fitness value of each individual in the current population, rank the individuals according to the fitness values of all individuals in the current population, then calculate the selection probability of each individual according to the fitness ranking of the individual and normalize the selection probability to obtain the normalized selection probability of the individual; at the same time, update the current number of fitness evaluations ; Step A3: initialize a memory pool for storing parameter pairs of successfully generated trial vectors, and calculate the scaling factor at the current iteration round based on the parameter pairs stored in the memory pool and the cross probability ; Step A4: Selection operation is performed on the current population based on the selection probability of all individuals after normalization, and the scaling factor under the current iteration round Mutation operation is performed on the selected individuals, and the crossover probability under the current iteration round is further based on Crossover operation is performed on the mutated individuals to generate the initial trial vector under the current iteration round; Step A5: updating the initial trial vector under the current iteration round according to a perturbation strategy equation to obtain a trial vector under the current iteration round; Step A6: Calculate the fitness value of the trial vector at the current iteration round and update the current number of fitness evaluations ; Step A7: judging whether the fitness value of the trial vector is better than the fitness value of the target vector, if yes, considering that the trial vector is successful, taking the trial vector as the individual updated in the current iteration round, and using a weighted Lehmer mean method to update the parameter pairs in the memory pool; if not, keeping the target vector as the individual updated in the current iteration round; Step A8: updating the current population according to the individual updated in the current iteration round, and using a gradient descent strategy to update the stagnation individuals in the current population to generate a child population; Step A9: Determine if the current number of fitness evaluations is greater than the maximum number of fitness evaluations then the offspring population is the current population and return to step A2; otherwise, select the individual from the offspring population with the best fitness value as the best individual.

4. The photovoltaic model parameter identification method based on the differential evolution algorithm according to claim 3, characterized in that, the scaling factor for the current iteration round in step A3 and the cross probability is calculated as The parameter pair includes: a scaling factor generated parameter and a crossover probability generated parameter ; For any individual , randomly select a parameter pair from the memory pool , where represents the selection of the th position in the memory pool ; represents the selection of the th position in the memory pool based on the selected parameters by comparing the current number of fitness evaluations with a predetermined threshold a scaling factor for the individual is calculated at the same time, a crossover probability for the individual is calculated ​ According to the parameter constraints, the scaling factor calculated for the individual and the crossover probability are constrained to generate the scaling factor and the crossover probability of the individual at the current iteration round.

5. The photovoltaic model parameter identification method based on differential evolution algorithm according to claim 4, characterized in that, The perturbation strategy equation in step A5 is: ； wherein denotes an individual the trial vector under the current iteration round; denotes an individual the target vector under the current iteration round; denotes an individual the initial trial vector under the current iteration round; denotes an individual the initial trial vector under the current iteration round; denotes an individual denotes a cosine perturbation coefficient, and has: ； wherein represents a randomly generated uniformly distributed random number.

6. The photovoltaic model parameter identification method based on the differential evolution algorithm according to claim 5, characterized in that, The method for updating the parameter pairs in the memory pool by using the weighted Lehmer mean method in step A7 is: For all trial vectors deemed successful, the scaling factor and crossover probability that generated the trial vector are recorded in a scaling factor set and a crossover probability set respectively. Evaluating the performance of each parameter pair in the memory pool in history respectively, and assigning a weight to each parameter pair according to the evaluation result; According to the weights of all parameter pairs in the memory pool, a weighted Lehmer mean method is used to calculate the scaling factor set weighted Lehmer average of the scaling factor set and the cross probability set weighted Lehmer average of the cross probability set ; randomly select a parameter pair from the memory pool, if the scaling factor set is not empty, then update the scaling factor in the parameter pair based on the weighted Lehmer average based on the historical information; if the scaling factor set is empty, then keep the current scaling factor unchanged; if the intersection probability set is not empty, then the intersection probability in the pair of parameters is updated based on the weighted Lehmer average , and if the intersection probability set is empty, then the current intersection probability is maintained.

7. The photovoltaic model parameter identification method based on differential evolution algorithm according to claim 6, characterized in that, The specific content of the generation of the child population by using the gradient descent strategy to update the stagnation individuals in the current population in step A8 is: For any dimension in the current population, calculating the average value and the standard deviation of the dimension, and judging whether the dimension converges; For any individual in the current population, calculating the stagnation index of the individual, and comparing the stagnation index of the individual with a predetermined threshold, and taking the individual whose stagnation index exceeds the predetermined threshold as a stagnation individual. For any converging dimension, count the number of stagnation individuals in the current population under the dimension , if the number of stagnation individuals exceeds a preset value, perform gradient descent update on all stagnation individuals in the current population, and generate a child population according to the updated individuals.

8. A photovoltaic model parameter identification system based on differential evolution algorithm, for implementing the photovoltaic model parameter identification method based on differential evolution algorithm according to any one of claims 1 to 7, characterized in that, The method comprises the following steps: A model construction module is configured to establish a photovoltaic model, and determine unknown parameters to be identified in the photovoltaic model; An individual optimization module is configured to take the unknown parameters to be identified in the photovoltaic model as individuals, and perform individual optimization by using a double-strategy collaborative deployment framework combining a probability-based trial vector generation strategy and a parameter adaptive scheme, and in the optimization process, use a gradient descent strategy to update stagnation individuals to obtain an optimal individual; An optimal parameter determination module is configured to take parameter values corresponding to the optimal individual as optimal parameters identified in the photovoltaic model.

9. An electronic device, comprising: The method comprises the following steps: One or more processors, and a memory for storing instructions which, when executed by the one or more processors, cause the one or more processors to perform the photovoltaic model parameter identification method based on the differential evolution algorithm according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores executable instructions which, when executed, cause the processor to perform the photovoltaic model parameter identification method based on the differential evolution algorithm according to any one of claims 1 to 7.

Images