Photovoltaic model parameter extraction method, system, equipment and medium

Abstract

The invention discloses a photovoltaic model parameter extraction method, system and device and a medium, and relates to the technical field of photovoltaic power generation array detection, and the method comprises the steps: obtaining a parameter range of a photovoltaic model, initializing a first optimization parameter vector set of an arctic light optimizer, and obtaining a second optimization parameter vector set of an arctic light optimizer; executing an arctic light optimizer on the first optimization parameter vector set to obtain a second optimization parameter vector set; obtaining a third optimization parameter vector set according to the first optimization parameter vector set and the second optimization parameter vector set; executing a random opposite learning stage on the third optimization parameter vector set to obtain a fourth optimization parameter vector set; obtaining a fifth optimization parameter vector set according to the fourth optimization parameter vector set and the third optimization parameter vector set; executing NMs simplex on the optimal optimization parameter vector in the fifth optimization parameter vector set, replacing the first optimization parameter vector set with the fifth optimization parameter vector set, updating the initial optimal optimization parameter vector, and outputting the optimal optimization parameter vector; according to the method, the convergence precision of the arctic light optimizer is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation array detection, and in particular to a photovoltaic model parameter extraction method, system, equipment and medium. Background Art

[0002] Among many types of energy, solar energy stands out with its unique advantages and has become one of the most widely used sustainable energy sources. With the continuous advancement of technology, solar photovoltaic power generation systems have achieved rapid development. Scientists have proposed some photovoltaic models, which play an indispensable role in the entire life cycle of photovoltaic power generation systems. They are extremely critical to the simulation, control, analysis, evaluation and design of photovoltaic power generation systems. Currently, commonly used photovoltaic models include single diode model, double diode model, three diode model and photovoltaic module model. However, these models are implicit transcendental equations in mathematical form. The characteristics of this equation determine that its solution process faces great challenges, especially the extraction of unknown parameters in the equation is extremely difficult. Since these unknown parameters are crucial to accurately describe the performance and behavior of photovoltaic power generation systems, how to efficiently extract the unknown parameters in the model has always been a focus and continuous exploration of researchers in this field.

[0003] At present, many scholars use metaheuristic algorithms to solve the problem of extracting unknown parameters. Metaheuristic algorithms are a type of optimization algorithm. Because of their strong adaptability, simple principles, no need to calculate gradients and easy implementation, they have been studied by many researchers and have been widely used. For example, the Polar Lights Optimizer (PLO) is a new physical phenomenon-based metaheuristic algorithm proposed by Yuan et al. to solve the problem of extracting unknown parameters of photovoltaic models.

[0004] However, the existing meta-heuristic algorithms are difficult to open up new space, which will cause the algorithm to fall into the local optimal situation when dealing with the parameter extraction problem of the photovoltaic model, and the processing ability near the optimal value is weak, and the convergence accuracy of the algorithm is poor, which makes it difficult for the photovoltaic model constructed by the extracted parameters to accurately present the accurate relationship between current and voltage, and ultimately causes the photovoltaic model to perform poorly under different irradiance and temperature conditions. Summary of the invention

[0005] In view of the shortcomings of the prior art that the Aurora Borealis optimizer has difficulty in exploring new spaces and is prone to falling into local optimality, which makes it difficult for the photovoltaic model constructed by the extracted parameters to accurately present the accurate correlation between current and voltage, the present invention proposes a photovoltaic model parameter extraction method, system, device and medium, which help the Aurora Borealis optimizer to open up a new search space by introducing the technology of random adversarial learning, and at the same time introduce NMs simplex to help the Aurora Borealis optimizer further develop the area where the optimal optimization parameter vector is located to improve the convergence accuracy of the Aurora Borealis optimizer, thereby solving the problems existing in the prior art.

[0006] A photovoltaic model parameter extraction method comprises the following steps: Obtain the parameter range of the photovoltaic model, and initialize the first optimization parameter vector set of the Northern Lights optimizer within the parameter range; evaluate the fitness values ​​of all vectors in the first optimization parameter vector set, and sort and compare the fitness values ​​of all vectors to obtain the initial optimal optimization parameter vector; Input the first optimization parameter vector set into the Northern Lights optimizer to obtain the second optimization parameter vector set; select the third optimization parameter vector set using a greedy strategy according to the fitness values ​​corresponding to the vectors in the first optimization parameter vector set and the second optimization parameter vector set; perform random adversarial learning on the third optimization parameter vector set to generate a new solution space to obtain the fourth optimization parameter vector set; select N vectors with the best fitness values ​​from the optimization parameter vector set of size 2N composed of the fourth optimization parameter vector set and the third optimization parameter vector set to form the fifth optimization parameter vector set; input the optimization parameter vector with the largest fitness value in the fifth optimization parameter vector set into the NMs simplex algorithm to obtain a new optimal optimization parameter vector; replace the first optimization parameter vector set with the fifth optimization parameter vector set, use the new optimal optimization parameter vector to update the initial optimal optimization parameter vector, and after multiple iterative updates, output the final optimal optimization parameter vector; According to the final optimal optimization parameter vector, the unknown parameters of the photovoltaic model are obtained.

[0007] Further, the photovoltaic model includes a single diode model, a double diode model, a triple diode model and a photovoltaic module model; The single diode model is expressed as: ; ; ; ; in, is the output current of the model, is the output voltage of the model, is the light-induced current generated, Through the shunt resistor After the current, is the shunt resistor, is the diode current, is the series resistance in the circuit, Represents the ideality factor of the diode under ideal conditions, is a constant value, Indicates temperature, represents the reverse saturation current of the diode, represents the charge, then the model includes five parameters: , , , and ; The double diode model is expressed as: ; ; in, and represents the current through the diode, and represents the ideal factor; then the model contains seven unknown parameters: , , , , , and ; The three-diode model is expressed as: ; in, represents the current through the diode, represents the ideal factor; it follows that the model includes nine unknown parameters: , , , , , , , and ; The photovoltaic module model is composed of multiple single diode models and double diode models, which are expressed as: ; in, represents the number of solar cells connected in parallel, Indicates the number of solar cells connected in series.

[0008] Furthermore, the fitness value It is used to measure the error between simulated data and real data, which is expressed as: ; in, Represents real data, represents simulated data, represents the unknown parameters that can be extracted, Indicates the number of samples.

[0009] Further, executing the Northern Lights Optimizer on the first optimization parameter vector set to obtain the second optimization parameter vector set specifically includes the following steps: The second optimization parameter vector set is obtained through the auroral elliptical orbit process. The optimization parameter vector is The value on the dimension, the acquisition process is expressed as: ; ; ; ; ; ; ; in, represents the rotation speed of the aurora particles, represents the damping coefficient, is the number of evaluations of the current fitness value, is a random number from Levy distribution, is the step size coefficient, To adjust the index of Levy distribution, Represents the entire optimization parameter vector set in The average value of the dimension, represents the size of the optimization parameter vector set, represents the first optimization parameter vector set The optimization parameter vector is The value of the dimension, Indicates that the solution space is The lower limit of the dimension, Indicates that the solution space is The upper limit of the dimension, is the factor controlling the step size, and To control the rotation speed and the coefficients of the elliptical orbit, is the factor controlling the step size, represents the first The optimization parameter vector is The value of the dimension, is the maximum number of evaluations; The second optimization parameter vector is obtained by particle collision. The optimization parameter vector is The value on the dimension, the acquisition process is expressed as: ; ; in, represents the first The optimization parameter vector is The value of the dimension, represents the first optimization parameter vector set The optimization parameter vector is The value of the dimension, To control the step size, represents the first optimization parameter vector set The random optimization parameter vector is The value of the dimension, and is the random value of the control condition judgment, represents the probability of particle collision, is the number of evaluations of the current fitness value, is the total number of evaluations.

[0010] Furthermore, the third optimization parameter vector set is selected based on the first optimization parameter vector set and the second optimization parameter vector set by a greedy strategy based on the corresponding fitness values, which is expressed as: Greedy selection process: ; in, represents the third optimization parameter vector set optimization parameter vector, represents the first optimization parameter vector set optimization parameter vector, represents the first optimization parameter vector, Represents the fitness value of the parameter vector.

[0011] Furthermore, the third optimization parameter vector set is subjected to a random adversarial learning phase to open up a new solution space and obtain a fourth optimization parameter vector set. The process is expressed as follows: ; in, represents the fourth optimization parameter vector set The optimization parameter vector is The value of the dimension, and Respectively represent the third optimization parameter vector concentrated in the The maximum and minimum values ​​in a dimension, represents the third optimization parameter vector set The optimization parameter vector is The value of the dimension, To determine the parameters for adversarial learning, is the step size coefficient.

[0012] Furthermore, the construction method of the NMs simplex is expressed as: ; in, For the construction The point is The value of the dimension, and is the coefficient of the construction point, is the optimal optimization parameter vector in The value of the dimension, It is any value from 1 to D, where D is the dimension of the problem.

[0013] The present invention also includes a photovoltaic model parameter extraction system, comprising: The acquisition module is used to obtain the parameter range of the photovoltaic model and initialize the first optimization parameter vector set of the Northern Lights optimizer within the parameter range; evaluate the fitness values ​​of all vectors in the first optimization parameter vector set, and sort and compare the fitness values ​​of all vectors to obtain the initial optimal optimization parameter vector.

[0014] A computing module is used to input the first optimization parameter vector set into the Northern Lights optimizer to obtain a second optimization parameter vector set; select a third optimization parameter vector set using a greedy strategy according to the fitness values ​​corresponding to the vectors in the first optimization parameter vector set and the second optimization parameter vector set; perform random adversarial learning on the third optimization parameter vector set to generate a new solution space to obtain a fourth optimization parameter vector set; select N vectors with the best fitness values ​​from an optimization parameter vector set of size 2N composed of the fourth optimization parameter vector set and the third optimization parameter vector set to form a fifth optimization parameter vector set; input the optimization parameter vector with the largest fitness value in the fifth optimization parameter vector set into the NMs simplex algorithm to obtain a new optimal optimization parameter vector; replace the first optimization parameter vector set with the fifth optimization parameter vector set, use the new optimal optimization parameter vector to update the initial optimal optimization parameter vector, and output the final optimal optimization parameter vector after multiple iterative updates.

[0015] The parameter extraction module is used to obtain the unknown parameters of the photovoltaic model according to the final optimal optimization parameter vector.

[0016] The present invention also includes a photovoltaic model parameter extraction computer device, including: a memory, a processor and a computer program stored in the memory, and the processor implements the steps of the photovoltaic model parameter extraction method when executing the computer program.

[0017] The present invention also includes a readable storage medium, wherein the readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the steps of the photovoltaic model parameter extraction method are executed.

[0018] The present invention provides a photovoltaic model parameter extraction method, which has the following beneficial effects: The present invention performs random adversarial learning on the third optimization parameter vector set to generate a new solution space, thereby preventing the problem of slow convergence caused by excessive exploration of new areas, thereby generating a fourth optimization vector, and sorting the optimization parameter vector set composed of the fourth optimization parameter vector set and the third optimization parameter vector set to obtain a fifth optimization parameter vector, further enhancing the convergence speed of the algorithm; at the same time, the NMs simplex further develops the optimization parameter vector with the largest fitness value in the fifth optimization parameter vector set, and further develops the area near the optimal optimization parameter vector through a series of processes such as reflecting, expanding and shrinking these points in different dimensions, thereby improving the convergence accuracy of the algorithm; the method proposes to improve the Aurora Borealis optimizer by using random adversarial learning and NMs simplex, so that the improved Aurora Borealis optimizer has better performance in the parameter extraction problem of the photovoltaic model, and solves the problem that the existing Aurora Borealis optimizer is prone to fall into local optimality in the photovoltaic parameter extraction problem, which makes it difficult for the photovoltaic model constructed by the extracted parameters to accurately present the accurate correlation between current and voltage. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flowchart of a photovoltaic model parameter extraction method according to an embodiment of the present invention; Figure 2 A schematic diagram of random adversarial learning in an embodiment of the present invention; Figure 3 Schematic diagram of NMs simplex in an embodiment of the present invention; Figure 4 is an equivalent circuit diagram of a photovoltaic model in an embodiment of the present invention; wherein Figure 4 (a) represents the equivalent circuit diagram of a single diode, Figure 4 (b) shows the equivalent circuit diagram of a double diode, Figure 4 (c) shows the equivalent circuit diagram of three diodes. Figure 4 (d) represents the equivalent circuit diagram of the photovoltaic module model; Figure 5 Schematic diagram of experimental results of various algorithms in the embodiment of the present invention on a single diode model; Figure 6 Schematic diagram of experimental results of various algorithms in the embodiment of the present invention on a dual-diode model; Figure 7 Schematic diagram of experimental results of various algorithms in the embodiment of the present invention on a three-diode model; Figure 8 It is a schematic diagram of the experimental results of various algorithms in the embodiments of the present invention on the photovoltaic module model; Fig. 9 It is a box plot of the experimental results of each algorithm on a single diode model in the embodiment of the present invention; Fig.10 It is a box plot of the experimental results of various algorithms on the dual diode model in the embodiments of the present invention; Fig.11 It is a box plot of the experimental results of various algorithms on the three-diode model in the embodiments of the present invention; Fig.12 It is a box plot of the experimental results of each algorithm on the photovoltaic module model in the embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0021] The present invention proposes a photovoltaic model parameter extraction method, such as Figure 1 As shown, the method specifically comprises the following steps: S1. Obtain the parameter range of the photovoltaic model. The photovoltaic model is divided into a single diode model, a double diode model, a triple diode model and a photovoltaic module model. Figure 4 The equivalent circuit diagram of the photovoltaic model is shown.

[0022] Among them, the mathematical expression of the single diode model is as follows: (1) (2) (3) (4) in, is the output current of the model, is the output voltage of the model, is the light-induced current generated, Through the resistor After the current, is the shunt resistor, is the diode current, is the series resistance in the circuit, Represents the ideality factor of the diode under ideal conditions, The constant value is , Indicates temperature, represents the reverse saturation current of the diode, The constant value is , Represents the amount of charge, which shows that the model mainly includes five parameters: , , , and .

[0023] The mathematical representation of the double diode model is as follows: (5) (6) in, and represents the current through the diode, and represents the ideal factor; then the model contains seven unknown parameters: , , , , , and .

[0024] The mathematical expression of the three-diode model is as follows: (7) in, represents the current through the diode, represents the ideality factor. Therefore, there are nine unknown parameters in the three-diode model: , , , , , , , and .

[0025] The photovoltaic module model consists of multiple single diode models and double diode models, which has better performance and more complex structure. The mathematical expression is as follows: (8) in, represents the number of solar cells connected in parallel, Indicates the number of solar cells connected in series.

[0026] S2. Initialize the first optimization parameter vector set of the Northern Lights optimizer within the parameter range , the size of the first optimization parameter vector set is , the maximum number of evaluations .

[0027] S3. Evaluate the fitness values ​​of all vectors in the first optimization parameter vector set: , and compare to get the optimal optimization parameter vector.

[0028] Is the fitness value, which is used to measure the error between simulated data and real data. Its mathematical form is as follows: (9) in, Represents real data, represents simulated data, Indicates unknown parameters that need to be extracted. Indicates the number of samples.

[0029] S4. Execute the Northern Lights optimizer on the first optimization parameter vector set to obtain a second optimization parameter vector set. Use a greedy strategy to compare the vectors in the first optimization parameter vector set and the second optimization parameter vector set. Vectors with better fitness values ​​constitute a third optimization parameter vector set.

[0030] Execute the Northern Lights Optimizer to process the first optimization parameter vector set. The main steps of the Northern Lights Optimizer are aurora elliptical orbits, particle collisions, and greedy selection. The steps are as follows: Auroral elliptical orbit process: (10) (11) (12) (13) (14) (15) (16) in, represents the rotation speed of the aurora particles, Represents the damping coefficient, whose value is a random number uniformly distributed between 1 and 1.5. is the number of evaluations of the current fitness value, is a random number from Levy distribution, is the step size coefficient, is an index for adjusting the Levy distribution, and in the present invention, its value is 1.5, Represents the entire optimization parameter vector set in The average value of the dimension, represents the size of the population, represents the first optimization parameter vector set The optimization parameter vector is The value of the dimension, Indicates that the solution space is The lower limit of the dimension, Indicates that the solution space is The upper limit of the dimension, is a factor that controls the step size, and its value is a random number uniformly distributed between 0 and 1. and is the coefficient that controls the rotation speed and elliptical orbit, is a factor that controls the step size, and its value is a random number uniformly distributed between 0 and 1. represents the first The optimization parameter vector is The value of the dimension.

[0031] Particle collision process: (17) (18) in, represents the first The optimization parameter vector is The value of the dimension, represents the first optimization parameter vector set The optimization parameter vector is The value of the dimension, is a factor that controls the step size, and its value is a random number uniformly distributed between 0 and 1. represents the first optimization parameter vector set The random optimization parameter vector is The value of the dimension, and The random value for controlling the condition judgment is a random number uniformly distributed between 0 and 1. represents the probability of particle collision, is the number of evaluations of the current fitness value, is the total number of evaluations.

[0032] Greedy selection process: (19) in, represents the third optimization parameter vector set optimization parameter vector, represents the first optimization parameter vector set optimization parameter vector, represents the first optimization parameter vector, Represents the fitness value of the parameter vector.

[0033] S5, perform the random adversarial learning phase on the third optimization parameter vector set to open up a new solution space and obtain a solution size of The fourth optimization parameter vector set and the third optimization parameter vector set form a fourth optimization parameter vector set of size The new optimization parameter vector set is selected to obtain the best fitness value. The vectors constitute the fifth optimization parameter vector set.

[0034] Perform a randomized adversarial learning phase on the third set of optimized parameter vectors: (20) in, represents the fourth optimization parameter vector set The optimization parameter vector is The value of the dimension, and Respectively represent the third optimization parameter vector concentrated in the The maximum and minimum values ​​in a dimension, represents the third optimization parameter vector set The optimization parameter vector is The value of the dimension, To determine the parameters for adversarial learning, its value is a random number uniformly distributed between 0 and 1. is the step coefficient, and its value is a random number uniformly distributed between 0 and 1. The schematic diagram of adversarial learning is as follows Figure 2 After the fourth optimization parameter vector set is generated, the fourth optimization parameter vector set is compared with the third optimization parameter vector set to extract the first The optimization parameter vectors with better fitness values ​​form the fifth optimization parameter vector set It is worth mentioning that random adversarial learning uses the maximum and minimum values ​​in each dimension and performs adversarial learning on the third optimization parameter vector. In this process, the algorithm can explore the parameter space outside the third optimization parameter vector set, with the aim of seeking other potential areas and alleviating the problem of local optimality. In addition, adversarial learning is performed in a random manner with a probability of 0.3, which can avoid the problem of slow convergence caused by excessive exploration of new space.

[0035] S6. Execute the NMs simplex on the optimization parameter vector with the best fitness value in the fifth optimization parameter vector set to obtain a new optimal optimization parameter vector.

[0036] Perform NMs simplex on the optimal optimization parameter vector in the fifth set of optimization parameter vectors: Performing NMs simplex can enhance the convergence accuracy of the current optimal optimization parameter vector and further guide the algorithm to continue to develop the optimal optimization parameter vector. The basic principle of NMs simplex revolves around the construction of approximate search directions. Next, five basic operations are used to search and replace known points: reflection, expansion, external contraction, internal contraction, and overall contraction. The selection of the search direction does not require gradient calculation. Instead, it infers the direction based on the known value of the objective function. The process of determining a specific search point depends on the relationship between the worst point in the simplex and the center of mass of the remaining points. From a geometric point of view, in In dimensional space, NMs simplex can be constructed with vertices of a polyhedron, making it possible to identify The schematic diagram of the NMs simplex is as follows: Figure 3 , constructed as follows: (twenty one) in, For the construction The point is The value of the dimension, and are the coefficients of the construction points, and their values ​​are 0.05 and 0.00025 respectively. is the optimal optimization parameter vector in The value of the dimension, 1 to A value of is the dimension of the problem. It is worth mentioning that the construction process only changes one dimension of the optimal optimization parameter vector, and the other dimensions remain unchanged. Therefore, when the optimal optimization parameter vector is added, there will be points.

[0037] The main process is as follows: Step 1: Right The fitness values ​​of the points are sorted.

[0038] (twenty two) Step 2: Calculate the reflection point and its fitness value. If and , then Replace with , then go to step 6. If , then go to step 3. If , then proceed to step 4.

[0039] (twenty three) (twenty four) in, is the reflection coefficient, express The center of a point.

[0040] Step 3: Calculate the extension point and its fitness value. If ,but will be Replace. Otherwise, will be Replace, and then proceed to step six.

[0041] (25) in, is the expansion coefficient.

[0042] Step 4: If , then calculate the inner contraction point and its fitness value. Then if , then use replace , and proceed to step 6. Otherwise, continue to step 5. , then calculate the external contraction point and its fitness value. Then, if , then Replace with , then proceed to step 6. Otherwise, proceed to step 5.

[0043] (26) (27) in, is the shrinkage coefficient.

[0044] Step 5: Next is the description of the overall shrinkage point. It includes shrinkage Construct a new simplex from all points other than .

[0045] (28) in, is the shrinkage coefficient.

[0046] Step 6: If the termination condition is met, the search ends; otherwise, continue to the next iteration.

[0047] S7. Replace the first optimization parameter vector set with the fifth optimization parameter vector set, and update the optimal optimization parameter vector and the current evaluation times.

[0048] S8. Repeat steps 4 to 7 until the maximum number of evaluations is reached, and output the optimal optimization parameter vector.

[0049] Experimental proof: Figure 5 The average value of the root mean square error (RMSE) of the experimental results of each algorithm on the single diode model is shown. It can be seen that RONMPLO has better results than algorithms such as GOFPANM, and has a significant improvement over the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, TLBO is the teaching and learning optimizer, and PLO is the original Northern Lights optimizer).

[0050] Figure 6 The average RMSE of the experimental results of each algorithm on the dual diode model is shown. It can be seen that RONMPLO has better results than algorithms such as GOFPANM, and has a significant improvement over the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, TLBO is the teaching and learning optimizer, and PLO is the original Northern Lights optimizer).

[0051] Figure 7The average RMSE of the experimental results of each algorithm on the three-diode model is shown. It can be seen that RONMPLO has better results than algorithms such as GOFPANM, and has a significant improvement over the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, TLBO is the teaching and learning optimizer, and PLO is the original Northern Lights optimizer).

[0052] Figure 8 The average RMSE of the experimental results of each algorithm on the photovoltaic module model is shown. It can be seen that RONMPLO has better results than algorithms such as GOFPANM, and has a significant improvement over the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, TLBO is the teaching and learning optimizer, and PLO is the original Northern Lights optimizer).

[0053] Fig. 9 The box plots showing the experimental results of each algorithm on a single diode model show that RONMPLO is also very stable compared to algorithms such as GOFPANM, and is also greatly improved compared to the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, TLBO is the teaching and learning optimizer, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, and PLO is the original Northern Lights optimizer).

[0054] Fig.10The box plots showing the experimental results of various algorithms on the dual diode model show that RONMPLO is also very stable compared to algorithms such as GOFPANM, and is also greatly improved compared to the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, TLBOBSA is a hybrid backtracking algorithm and teaching and learning algorithm, GOFPANM is a pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, and PLO is the original Northern Lights optimizer).

[0055] Fig.11 The box plots showing the experimental results of various algorithms on the three-diode model show that RONMPLO is also very stable compared to algorithms such as GOFPANM, and is also greatly improved compared to the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, TLBO is the teaching and learning optimizer, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, and PLO is the original Northern Lights optimizer).

[0056] Fig.12 The box plots showing the experimental results of various algorithms on the photovoltaic module model show that RONMPLO is also very stable compared to algorithms such as GOFPANM, and is also greatly improved compared to the original PLO. (RONMPLO is the improved Northern Lights optimizer of the present invention, IJAYA is the improved victory algorithm, GOTLBO is the improved teaching and learning algorithm, MLBSA is the improved backtracking algorithm for multi-strategy learning, TLBO is the teaching and learning optimizer, GOFPANM is the pollination algorithm based on generalized adversarial learning and NMs simplex, EHHO is the enhanced Harris Hawk algorithm, BSA is the original backtracking algorithm, and PLO is the original Northern Lights optimizer).

[0057] Based on the same inventive concept, the present invention also proposes a photovoltaic model parameter extraction system, comprising: The acquisition module is used to obtain the parameter range of the photovoltaic model and initialize the first optimization parameter vector set of the Northern Lights optimizer within the parameter range; evaluate the fitness values ​​of all vectors in the first optimization parameter vector set, and sort and compare the fitness values ​​of all vectors to obtain the initial optimal optimization parameter vector.

[0058] The computing module is used to input the first optimization parameter vector set into the Northern Lights optimizer to obtain the second optimization parameter vector set; select the third optimization parameter vector set by adopting a greedy strategy according to the fitness values ​​corresponding to the vectors in the first optimization parameter vector set and the second optimization parameter vector set; perform random adversarial learning on the third optimization parameter vector set to generate a new solution space to obtain a fourth optimization parameter vector set; form a solution space of size 1 from the fourth optimization parameter vector set and the third optimization parameter vector set The optimization parameter vector is concentrated to select the best fitness value vectors form a fifth optimization parameter vector set; the optimization parameter vector with the largest fitness value in the fifth optimization parameter vector set is input into the NMs simplex algorithm to obtain a new optimal optimization parameter vector; the fifth optimization parameter vector set replaces the first optimization parameter vector set, and the new optimal optimization parameter vector is used to update the initial optimal optimization parameter vector. After multiple iterative updates, the final optimal optimization parameter vector is output.

[0059] The parameter extraction module is used to obtain unknown parameters extracted from the photovoltaic model according to the final optimal optimization parameter vector.

[0060] The present invention also provides a photovoltaic model parameter extraction computer device, comprising: a memory, a processor and a computer program stored in the memory, and the processor implements the steps of the photovoltaic model parameter extraction method when executing the computer program.

[0061] The present invention further provides a readable storage medium, wherein the readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the steps of the photovoltaic model parameter extraction method are executed.

[0062] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A photovoltaic model parameter extraction method, characterized in that: The following steps are involved: Obtain the parameter range of the photovoltaic model, and initialize the first optimization parameter vector set of the Northern Lights optimizer within the parameter range; evaluate the fitness values ​​of all vectors in the first optimization parameter vector set, and sort and compare the fitness values ​​of all vectors to obtain the initial optimal optimization parameter vector; Input the first optimization parameter vector set into the Northern Lights optimizer to obtain a second optimization parameter vector set; select a third optimization parameter vector set using a greedy strategy according to the fitness values ​​corresponding to the vectors in the first optimization parameter vector set and the second optimization parameter vector set; perform random adversarial learning on the third optimization parameter vector set to generate a new solution space to obtain a fourth optimization parameter vector set; The fourth optimization parameter vector set and the third optimization parameter vector set form a vector having a size of The optimization parameter vector is concentrated to select the best fitness value vectors to form a fifth optimization parameter vector set; input the optimization parameter vector with the largest fitness value in the fifth optimization parameter vector set into the NMs simplex algorithm to obtain a new optimal optimization parameter vector; replace the first optimization parameter vector set with the fifth optimization parameter vector set, use the new optimal optimization parameter vector to update the initial optimal optimization parameter vector, and after multiple iterative updates, output the final optimal optimization parameter vector; According to the final optimal optimization parameter vector, the unknown parameters of the photovoltaic model are obtained.

2. A photovoltaic model parameter extraction method according to claim 1, characterized in that: The photovoltaic model includes a single diode model, a double diode model, a triple diode model and a photovoltaic module model; The single diode model is expressed as: ； ； ； ； in, is the output current of the model, is the output voltage of the model, is the light-induced current generated, Through the shunt resistor After the current, is the shunt resistor, is the diode current, is the series resistance in the circuit, Represents the ideality factor of the diode under ideal conditions, is a constant value, Indicates temperature, represents the reverse saturation current of the diode, represents the charge, then the model includes five unknown parameters: , , , and ; The double diode model is expressed as: ； ； in, and represents the current through the diode, and represents the ideal factor; The three-diode model is expressed as: ； in, represents the current through the diode, represents the ideal factor; The photovoltaic module model is composed of multiple single diode models and double diode models, which are expressed as: ； in, represents the number of solar cells connected in parallel, Indicates the number of solar cells connected in series.

3. A photovoltaic model parameter extraction method according to claim 1, characterized in that: The fitness value It is used to measure the error between simulated data and real data, which is expressed as: ； in, Represents real data, represents simulated data, represents the unknown parameters that can be extracted, represents the number of samples, Represents the size of the optimization parameter vector set.

4. A photovoltaic model parameter extraction method according to claim 1, characterized in that: The method of executing the Northern Lights optimizer on the first optimization parameter vector set to obtain the second optimization parameter vector set specifically includes the following steps: The second optimization parameter vector set is obtained through the auroral elliptical orbit process. The optimization parameter vector is The value on the dimension, the acquisition process is expressed as: ； ； ； ； ； ； ； in, represents the rotation speed of the aurora particles, represents the damping coefficient, is the number of evaluations of the current fitness value, is a random number from Levy distribution, is the step size coefficient, To adjust the index of Levy distribution, Represents the entire optimization parameter vector set in The average value of the dimension, represents the size of the optimization parameter vector set, represents the first optimization parameter vector set The optimization parameter vector is The value of the dimension, Indicates that the solution space is The lower limit of the dimension, Indicates that the solution space is The upper limit of the dimension, is the factor controlling the step size, and To control the rotation speed and the coefficients of the elliptical orbit, is the factor controlling the step size, represents the first The optimization parameter vector is The value of the dimension, is the maximum number of evaluations; The second optimization parameter vector is obtained by particle collision. The optimization parameter vector is The value on the dimension, the acquisition process is expressed as: ； ； in, represents the first The optimization parameter vector is The value of the dimension, represents the first optimization parameter vector set The optimization parameter vector is The value of the dimension, is the factor controlling the step size, represents the first optimization parameter vector set The random optimization parameter vector is The value of the dimension, and is the random value of the control condition judgment, represents the probability of particle collision, is the number of evaluations of the current fitness value, is the total number of evaluations.

5. A photovoltaic model parameter extraction method according to claim 1, characterized in that: According to the fitness values ​​corresponding to the vectors in the first optimization parameter vector set and the second optimization parameter vector, a greedy strategy is adopted to select the third optimization parameter vector set, which is expressed as: Greedy selection process: ； in, represents the third optimization parameter vector set optimization parameter vector, represents the first optimization parameter vector set optimization parameter vector, represents the first optimization parameter vector, Represents the fitness value of the parameter vector.

6. A photovoltaic model parameter extraction method according to claim 5, characterized in that: The process of performing random adversarial learning on the third optimization parameter vector set to generate a new solution space and obtain the fourth optimization parameter vector set is expressed as follows: ； in, represents the fourth optimization parameter vector set The optimization parameter vector is The value of the dimension, and Respectively represent the third optimization parameter vector concentrated in the The maximum and minimum values ​​in a dimension, represents the third optimization parameter vector set The optimization parameter vector is The value of the dimension, To determine the parameters for adversarial learning, is the step size coefficient.

7. A photovoltaic model parameter extraction method according to claim 1, characterized in that: The construction method of the NMs simplex algorithm is expressed as: ； in, For the construction The point is The value of the dimension, and is the coefficient of the construction point, is the optimal optimization parameter vector in The value of the dimension, 1 to Any value of The dimension of the problem.

8. A photovoltaic model parameter extraction system, characterized in that: include: An acquisition module, used for acquiring a parameter range of a photovoltaic model, and initializing a first optimization parameter vector set of a northern lights optimizer within the parameter range; Evaluate the fitness values ​​of all vectors in the first optimization parameter vector set, and sort and compare the fitness values ​​of all vectors to obtain the initial optimal optimization parameter vector; A calculation module is used to input the first optimization parameter vector set into the Northern Lights optimizer to obtain a second optimization parameter vector set; select a third optimization parameter vector set using a greedy strategy according to the fitness values ​​corresponding to the vectors in the first optimization parameter vector set and the second optimization parameter vector set; perform random adversarial learning on the third optimization parameter vector set to generate a new solution space to obtain a fourth optimization parameter vector set; The fourth optimization parameter vector set and the third optimization parameter vector set form a vector having a size of The optimization parameter vector is concentrated to select the best fitness value vectors to form a fifth optimization parameter vector set; input the optimization parameter vector with the largest fitness value in the fifth optimization parameter vector set into the NMs simplex algorithm to obtain a new optimal optimization parameter vector; replace the first optimization parameter vector set with the fifth optimization parameter vector set, use the new optimal optimization parameter vector to update the initial optimal optimization parameter vector, and after multiple iterative updates, output the final optimal optimization parameter vector; The parameter extraction module is used to obtain the unknown parameters of the photovoltaic model according to the final optimal optimization parameter vector.

9. A photovoltaic model parameter extraction computer device, characterized in that: include: A memory, a processor and a computer program stored in the memory, wherein the processor implements the steps of the photovoltaic model parameter extraction method according to any one of claims 1 to 7 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, they are used to execute the steps of the photovoltaic model parameter extraction method according to any one of claims 1 to 7.

Images