A photovoltaic array online modeling method

Abstract

The present invention discloses a photovoltaic array online modeling method, comprising the following steps: obtaining measured data of the actual operation of photovoltaic components, the measured data including current and voltage data as well as irradiance and temperature data; preprocessing the measured data to remove abnormal data and data under low irradiance; identifying photovoltaic array model parameters using a particle swarm and gray wolf hybrid optimization algorithm to obtain photovoltaic array model parameters; selecting different functions for fitting according to the distribution of each model parameter, and establishing analytical equations between model parameters and irradiance and temperature; solving the corresponding model parameters through the measured irradiance and temperature, and establishing a real-time simulation model. The photovoltaic array online modeling method provided by the present invention can improve the global search capability of the particle swarm algorithm by utilizing the balance capability between the local and global optimal solutions of the gray wolf optimization algorithm, thereby avoiding falling into the local optimal solution; and at the same time, it has real-time performance and can identify model parameters for the I-V curve under fault conditions.

Description

Technical Field

[0001] The invention relates to an online modeling method for a photovoltaic array, and belongs to the technical field of photovoltaic power generation. Background Art

[0002] Photovoltaic arrays are the first-level power electronic devices in photovoltaic systems that directly convert solar energy into electrical energy. They are in complex outdoor environments for a long time and are affected by uncertain meteorological factors such as solar ultraviolet radiation, thermal cycles, dust, and hail. They are prone to failures, which seriously affect the power generation efficiency of photovoltaic systems. Therefore, a large number of researchers have begun to focus on the research of photovoltaic array anomaly detection, fault diagnosis, and health status assessment technologies. These technologies usually require the use of measured photovoltaic array operating electrical parameters, such as: string voltage, current, or photovoltaic array current-voltage (IV) curve, and compare and analyze with the estimated results of the photovoltaic array mathematical model. Fault detection, diagnosis, and health status assessment are achieved through the degree of deviation between the two, or effective feature parameters are further extracted through the deviation between the two for fault diagnosis. The accuracy of the photovoltaic array mathematical model will directly affect the accuracy of fault diagnosis and health status assessment, so it is necessary to further calculate the photovoltaic array model parameters to improve the accuracy of the mathematical model.

[0003] In practical applications, in order to establish the output characteristics of photovoltaic arrays under different environmental conditions, the methods for model parameter identification can be mainly divided into analytical methods, iterative methods and meta-heuristic algorithms. With the long-term operation of photovoltaic arrays, the components will gradually degrade, and the nameplate parameters of the components when leaving the factory have deviated from the actual operating state, so the accuracy of the analytical method is low. The speed and accuracy of the iterative method are better than the analytical method, but it has higher requirements on the objective function and is easily affected by the initial value, resulting in local convergence and divergence. Since the meta-heuristic algorithm can realize global search to avoid falling into local optimality, and can handle calculations without gradient constraints and nonlinear problems, it is widely used in photovoltaic array model parameter extraction to overcome the problems of analytical and iterative methods.

[0004] Particle Swarm Optimization (PSO) is a commonly used group optimization algorithm proposed by Kennedy et al. in 1995. The algorithm is mainly based on the social behavior of bird flocks, i.e. "particles" in the process of searching for food (optimal position). The initial position of the particle is randomly generated in the search space. Each particle updates its own moving speed and position during the search process and eventually finds food (i.e. the optimal position). The Grey Wolf Optimization (GWO) is a new meta-heuristic optimization algorithm derived from simulating the predation behavior of grey wolf groups in nature. In the optimization process of the Grey Wolf Optimization algorithm, the α, β, and δ wolves closest to the prey target (local optimal solution) are responsible for determining the prey target and position (global optimal solution), and approaching and tracking the prey. The general individual ω in the wolf pack follows the position of the wolf with a higher level than itself to adjust its position, forming an encirclement of the prey, chasing, and then attacking and capturing, that is, obtaining the optimal solution. The quasi-particle swarm optimization algorithm has a fast convergence speed, but its global search ability is poor and it is easy to fall into the local optimal solution. The gray wolf optimization algorithm has a good balance between the local optimal solution and the global optimal solution, and is not easy to fall into the local optimal solution. However, it mainly judges the distance between the wolf pack and the prey based on the position information of the leading wolf, resulting in a slow convergence speed. Therefore, there is an urgent need for an optimization algorithm that can organically combine the advantages of the particle swarm optimization algorithm and the gray wolf optimization algorithm to overcome the shortcomings of both and be used to identify the model parameters of the photovoltaic array. In addition, it takes time to identify the model parameters of the measured IV curve of the photovoltaic array through the identification algorithm alone, which is not real-time. Moreover, the identification algorithm cannot identify the model parameters of the IV curve under fault. If the current array fails, it cannot be modeled. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a photovoltaic array online modeling method, which can improve the global search ability of the particle swarm algorithm by utilizing the balancing ability between the local and global optimal solutions of the grey wolf optimization algorithm, thereby avoiding falling into the local optimal solution; at the same time, it has real-time performance and can identify model parameters of the IV curve under fault conditions.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A photovoltaic array online modeling method comprises the following steps:

[0008] Obtain the measured data of the actual operation of the photovoltaic modules, including current and voltage data as well as irradiance and temperature data;

[0009] Pre-process the measured data to remove abnormal data of the photovoltaic array under mismatch conditions and data under low irradiance;

[0010] Based on the current and voltage data in the preprocessed measured data, the photovoltaic array model parameters are identified using the particle swarm and gray wolf hybrid optimization algorithm to obtain the photovoltaic array model parameters; according to the distribution of each model parameter, different functions are selected for fitting through the least squares method to establish the analytical equation between the model parameters and the irradiance and temperature;

[0011] Using the established analytical equations, the corresponding model parameters are solved through the measured irradiance and temperature to build a real-time simulation model.

[0012] The current and voltage data are obtained through an inverter with integrated current and voltage curve scanning function, denoted as [I mea , V mea ], where I mea and V mea Represents current and voltage respectively.

[0013] The measured data preprocessing includes abnormal point detection method, smoothing filtering method and environmental stability detection method to obtain the current and voltage data of the photovoltaic module that meets the conditions, which is recorded as [I i，mea , V i，mea ], i = (1, ..., N), where N represents the number of current and voltage data points.

[0014] The five parameters of the photovoltaic array equivalent model are A, I ph , R s , R sh and I0, where A represents the diode ideality factor, I ph represents the avalanche breakdown current, R s and R sh They represent the equivalent series-parallel resistance of the photovoltaic array, I0 represents the diode saturation current, and the diode ideal factor A is fixed and taken as 1 before model parameter identification.

[0015] The identification of photovoltaic array model parameters using particle swarm and gray wolf hybrid optimization algorithm specifically includes the following steps:

[0016] S01, initializing the optimization algorithm parameters and the gray wolf population;

[0017] S02, after initialization, the fitness value of the individual gray wolf is calculated according to the fitness function and the corresponding error function, and the position information of the three leading wolves and other individual gray wolves is determined;

[0018] S03, using the particle position update formula of the particle swarm optimization algorithm to replace the position formula of the gray wolf optimization algorithm, iteratively calculating the position information to obtain the forward speed and position of the individual gray wolf toward the three leading wolves, and locating the final position of the individual gray wolf;

[0019] S04, determine whether the maximum number of iterations has been reached, otherwise jump to S02 and continue to execute, if yes, output the corresponding global optimal result, that is, the model parameter I ph , R s , R sh and I0.

[0020] In S01, the optimization algorithm parameters are initialized including: population number pop, maximum number of iterations Iter max , the number of parameters to be determined d, the range of parameter update coefficients, where the maximum number of iterations Iter max Indicates the termination condition of the algorithm;

[0021] Initializing the gray wolf population includes: randomly generating a gray wolf population, initializing each gray wolf position in the search space Initialize according to the following formula:

[0022]

[0023] in, and Respectively represent the upper and lower bounds of the position parameters of the i-th dimension, represents the value of the i-th dimension position parameter of the j-th gray wolf, pop represents the total number of the gray wolf population, d represents the dimension of the position parameter, and rand represents a random number in [0, 1].

[0024] In S02, the root mean square error is used as the fitness function of the hybrid optimization algorithm. First, the fitness value f(X i ), i = 1, 2, ..., pop, as shown in formulas (2) to (4):

[0025] f(X i ) = RMSE i (2)

[0026]

[0027]

[0028] Among them, the root mean square error RMSE i Represents X i The corresponding degree of deviation between the simulation data and the measured data, M represents the number of points of the measured IV curve, and X represents the unknown photovoltaic array model parameters that need to be identified, that is, A, I ph , R s , R sh and I0, V m and I mThey represent the measured voltage and current respectively; A represents the diode ideal factor, and K is the Boltzmann constant, which is 1.38×10 -23 J / K, T is the temperature of the photovoltaic cell, m is the avalanche breakdown index, and q is the electron charge constant, specifically 1.602×10 -19 C.V bd is the PN junction breakdown voltage, a is the coefficient related to the avalanche breakdown current;

[0029] Sort the fitness values ​​of the gray wolves and save the location information of the α wolf with the best fitness value in the population as X a , save the location information of the β wolf with the second best fitness value in the population as X β , save the location information of the wolf with the third best fitness in the population as X δ , the remaining gray wolf is the lowest-ranking individual in the population, gray wolf ω.

[0030] In S03, the distances D between the three wolves α, β, and δ and the other individual gray wolf ω are calculated respectively. a , D β , D δ , where the particle speed and position update formula in the particle swarm optimization algorithm replaces the wolf position update formula in the gray wolf optimization algorithm, and the inertia weight k is introduced to coordinate the global search and local search capabilities of the hybrid optimization algorithm, as follows:

[0031]

[0032] Among them, D a , D β , D δ Respectively represent the distances between the three leading wolves α, β, δ and other individual gray wolf ω, C1, C2 and C3 are random vectors, and X(t) represents the position of the current individual gray wolf ω;

[0033] C1, C2 and C3 are calculated as follows:

[0034] C=2r2·a (6)

[0035] Where r1 and r2 are random vectors between the interval [0, 1], vector a is a column vector consisting of convergence factors a, and E is a row vector whose elements are all 1. The convergence factor a is calculated as follows:

[0036]

[0037] Among them, t is the current iteration number, Iter max is the maximum number of iterations;

[0038] The particle position update formula of the particle swarm optimization algorithm is used to replace the position formula of the gray wolf optimization algorithm. The expression after replacement is as follows:

[0039]

[0040]

[0041] X(t+1)=X(t)+v(t+1) (10)

[0042] v(t) and X(t) are the speed and position of the wolf pack at the tth iteration, X(t+1) is the position information of the wolf pack after iteration, k is the inertia weight, which is a random number between [0-1], c1, c2 and c3 are the individual learning factors of the gray wolf, r1, r2 and r3 are random numbers between [0, 1], A1, A2 and A3 are coefficient vectors, and the calculation method is

[0043] A=2(r1-E)·a (11).

[0044] The model parameter R is adjusted using the power exponential function. s , R sh and irradiance G d Fitting is performed using a linear function to adjust the model parameters I ph and irradiance G d Fitting is performed and the model parameter I0 and temperature T are fitted using an exponential function.

[0045] The beneficial effects of the present invention are as follows: the present invention provides a photovoltaic array online modeling method, which adopts a PSO-GWO hybrid optimization algorithm to identify model parameters, combines the particle velocity and position iteration method in a particle swarm algorithm with the position update formula of a gray wolf optimization algorithm, thereby retaining the good balancing ability of the gray wolf optimization algorithm between a local optimal solution and a global optimal solution, and introduces an iteration method of the particle swarm algorithm, thereby improving the convergence speed of the hybrid algorithm and avoiding falling into a local optimal solution; using the measured IV curve data of the photovoltaic array, establishing the relationship between the model parameters and the measured environmental parameters, and then estimating the photovoltaic array model parameters in real time through the measured environmental parameters, completing real-time online modeling, which is time-saving and capable of identifying the model parameters of the IV curve under a fault, thereby ensuring the real-time and accuracy of photovoltaic array modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of a photovoltaic array online modeling method of the present invention;

[0047] Figure 2 It is a flow chart of the parameter identification of the PSO-GWO model in the present invention;

[0048] Figure 3is a fitting curve diagram of the photovoltaic array model parameters and the environmental parameters in the present invention;

[0049] Figure 4 It is a diagram of the IV curve modeling results in the present invention. DETAILED DESCRIPTION

[0050] The present invention is further described below. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are not intended to limit the protection scope of the present invention. Specific embodiment 1

[0052] like Figure 1 As shown, the present invention discloses a photovoltaic array online modeling method, comprising the following steps:

[0053] Step 1: Obtain the measured data of the actual operation of the photovoltaic module, which includes current and voltage data as well as irradiance and temperature data. The current and voltage data are obtained through an inverter with integrated current and voltage curve scanning function, and are recorded as [I mea , V mea ], N data points for current and voltage, where I mea and V mea Represents current and voltage respectively.

[0054] Step 2: Preprocess the measured data to remove abnormal data of the photovoltaic array under mismatch conditions and data under low irradiance. The measured data preprocessing includes abnormal point detection method, smoothing filtering method and environmental stability detection method to obtain the current and voltage data of the photovoltaic module that meets the conditions, which is recorded as [I i，mea , V i，mea ], i = (1, ..., N), where N represents the number of current and voltage data points.

[0055] Step three, based on the current and voltage data in the measured data after preprocessing, use the particle swarm and gray wolf hybrid optimization algorithm to identify the photovoltaic array model parameters and obtain the photovoltaic array model parameters. The standard particle swarm optimization algorithm (PSO) has a fast convergence speed, but poor global search ability and is prone to falling into local optimal solutions; the gray wolf optimization algorithm (GWO) has a good balance between local optimal solutions and global optimal solutions, and is not easy to fall into local optimal solutions, but it needs to judge the distance between the wolf pack and the prey based on the position information of the leading wolf, resulting in slow convergence; therefore, this patent combines the advantages of the particle swarm optimization algorithm and the gray wolf optimization algorithm, and proposes a particle swarm and gray wolf hybrid optimization algorithm PSO-GWO, which uses the balance between the local and global optimal solutions of the GWO optimization algorithm to improve the global search ability of the PSO optimization algorithm, avoids falling into local optimal solutions, and identifies the photovoltaic array model parameters, as shown in the following figure. Figure 2 As shown. For the optimization algorithm, the dimension of the optimization space will restrict the solution accuracy. Too high a dimension will often make the optimization algorithm more likely to fall into the local optimum, resulting in extremely unstable algorithm performance. Therefore, when using the optimization algorithm, the dimension of the solution space should be reduced as much as possible. In addition, the influence of the model parameters on the shape of the IV curve of the photovoltaic array is not independent of each other, that is, the solutions of the five model parameters are not independent of each other, and there is a problem of mutual coupling between the model parameters, resulting in unstable parameter extraction and coupling errors, and is not conducive to the subsequent fitting of the relationship between the model parameters and the irradiance and temperature, and the establishment of an online simulation model. Taking into account the coupling problem of the influence of the model parameters on the IV characteristic curve, in order to avoid the coupling problem between the model parameters, the present invention first targets the five parameters of the photovoltaic cell equivalent model, including [A, I ph ,R s ,R sh ,I0], where A represents the diode ideality factor, I ph represents the avalanche breakdown current, R s and R sh Represent the equivalent series-parallel resistance of the photovoltaic array, I0 represents the diode saturation current, and their respective effects on the IV characteristic curve are analyzed. The results show that the effects of the ideal factor A and the diode saturation current I0 on the IV curve are mutually coupled. The increase of the ideal factor A will make the IV curve open circuit voltage V oc However, the increase in saturation current I0 will cause the IV curve open circuit voltage V ocTherefore, the two are prone to coupling errors during parameter identification. The same IV curve may identify multiple sets of solutions corresponding to A and I0, making the model parameter identification inaccurate. In order to further ensure the accuracy and stability of model parameter identification, one of the A and I0 model parameters is first fixed before model parameter identification. A large number of experiments have shown that the diode ideal factor A generally does not change with changes in environmental conditions and can be taken as 1. Then the model parameter to be determined [I ph ,R s ,R sh ,I0] The PSO-GWO hybrid optimization algorithm proposed in this invention is used to identify model parameters.

[0056] The identification of photovoltaic array model parameters using particle swarm and gray wolf hybrid optimization algorithm specifically includes the following steps:

[0057] Step a: Initialize the optimization algorithm parameters and the gray wolf population. The optimization algorithm parameters initialization includes: population number pop, maximum number of iterations Iter max , the number of parameters to be determined d, the range of parameter update coefficients, where the maximum number of iterations Iter max Indicates the termination condition of the algorithm;

[0058] Initializing the gray wolf population includes: randomly generating a gray wolf population, initializing each gray wolf position in the search space Initialize according to the following formula:

[0059]

[0060] in, and Respectively represent the upper and lower bounds of the position parameters of the i-th dimension, represents the value of the i-th dimension position parameter of the j-th gray wolf, pop represents the total number of the gray wolf population, d represents the dimension of the position parameter, and rand represents a random number in [0, 1].

[0061] Step b, after initialization, the fitness value of the individual gray wolf is calculated according to the fitness function and the corresponding error function, and the position information of the three leading wolves and other individual gray wolves is determined. In order to accurately calculate the error between the simulated output current and the measured current, the root mean square error is used as the fitness function of the hybrid optimization algorithm. First, the fitness value f(X i ), i = 1, 2, ..., pop, as shown in formulas (2) to (4):

[0062] f(X i ) = RMSE i (2)

[0063]

[0064]

[0065] Among them, the root mean square error RMSE i Represents X i The smaller the RMSE is, the smaller the distance between the gray wolf and the prey is, indicating that the PV module modeling accuracy is higher. M represents the number of points in the measured IV curve, and X represents the unknown PV array model parameters that need to be identified, i.e., A, I ph , R s , R sh and I0, V m and I m They represent the measured voltage and current respectively; A represents the diode ideal factor, and K is the Boltzmann constant, which is 1.38×10 -23 J / K, T is the temperature of the photovoltaic cell, m is the avalanche breakdown index, and q is the electron charge constant, specifically 1.602×10 -19 C.V bd is the PN junction breakdown voltage, and a is the coefficient related to the avalanche breakdown current.

[0066] Sort the fitness values ​​of the gray wolves and save the location information of the α wolf with the best fitness value in the population as X a , save the location information of the β wolf with the second best fitness value in the population as X β , save the location information of the wolf with the third best fitness in the population as X δ , the remaining gray wolf is the lowest-ranking individual in the population, gray wolf ω.

[0067] Step c, using the particle swarm optimization algorithm particle position update formula to replace the gray wolf optimization algorithm position formula, iteratively calculate the position information to obtain the forward speed and position of the individual gray wolf toward the three leading wolves, and locate the final position of the individual gray wolf.

[0068] When the wolf pack surrounds the prey, it is necessary to determine the location of the prey (global optimal solution) and initiate hunting. In the optimization process, α, β, δ wolves (local optimal solutions) are closest to the prey target and have a better understanding of the potential location information of the prey. Calculate the distance D between the three wolves α, β, δ and the other individual gray wolf ω respectively a , D β , D δ , where the particle speed and position update formula in the particle swarm optimization algorithm replaces the wolf position update formula in the gray wolf optimization algorithm, and the inertia weight k is introduced to coordinate the global search and local search capabilities of the hybrid optimization algorithm, as follows:

[0069]

[0070] Among them, D a , D β , D δ Respectively represent the distances between the three leading wolves α, β, δ and the other individual gray wolf ω, C1, C2 and C3 are random vectors, and X(t) represents the position of the current individual gray wolf ω;

[0071] C1, C2 and C3 are calculated as follows:

[0072] C=2r2·a (6)

[0073] Where r1 and r2 are random vectors between the interval [0, 1], vector a is a column vector consisting of convergence factors a, and E is a row vector whose elements are all 1. The convergence factor a is calculated as follows:

[0074]

[0075] Among them, t is the current iteration number, Iter max is the maximum number of iterations;

[0076] At the end of each iteration, the particle position update formula of the particle swarm optimization algorithm is used to replace the position formula of the gray wolf optimization algorithm to make the hybrid optimization algorithm converge faster. The replaced expression is as follows:

[0077]

[0078]

[0079] X(t+1)=X(t)+v(t+1) (10)

[0080] v(t) and X(t) are the speed and position of the wolf pack at the tth iteration, X(t+1) is the position information of the wolf pack after iteration, k is the inertia weight, which is a random number between [0-1], c1, c2 and c3 are the individual learning factors of the gray wolf, r1, r2 and r3 are random numbers between [0, 1], A1, A2 and A3 are coefficient vectors, and the calculation method is

[0081] A=2(r1-E)·a (11).

[0082] Step d: determine whether the maximum number of iterations has been reached. Otherwise, jump to step b and continue to execute. If yes, output the corresponding global optimal result, i.e., the model parameter I ph , R s , R shStep 4: According to the distribution of each model parameter, select different functions to fit through the least squares method to establish the analytical equation between the model parameters and the irradiance and temperature. s , R sh and irradiance G d Fitting is performed using a linear function to adjust the model parameters I ph and irradiance G d Fitting is performed and the model parameter I0 and temperature T are fitted using an exponential function.

[0083] Step five, using the established analytical equations, solve the corresponding model parameters through the measured irradiance and temperature to establish a real-time simulation model. Specific embodiment 2

[0085] The method of specific embodiment 1 is specifically applied to a real scene. The data of the photovoltaic power station from June 2018 to June 2019 are divided into four seasons to optimize the unknown model parameters [I ph ,R s ,R sh ,I0] for identification;

[0086] According to the distribution of each model parameter, a suitable function is selected for fitting to establish a new analytical expression, which is then fitted using the least squares method. Figure 3 The distribution of each model parameter extracted in autumn and the corresponding fitting curve are shown. It can be seen that the model parameter I ph ,R s ,R sh It is mainly related to the environmental parameter irradiance, and the model parameter I0 is mainly related to temperature. The model parameter R is expressed by the power exponential function. s ,R sh and irradiance G d Fitting is performed using a linear function to adjust the model parameters I ph and irradiance G d Fitting is performed, and an exponential function is used to fit I0 and temperature T. The fitting formulas and corresponding coefficients of each model parameter corresponding to the four seasons of spring, summer, autumn and winter are shown in Table 1;

[0087] Table 1

[0088]

[0089] The above-mentioned online model of the photovoltaic array has been established. The established analytical equations can be used to solve the corresponding model parameters through the measured irradiance and temperature to establish a real-time and accurate simulation model. The above-mentioned established analytical expressions are used to verify the measured IV curve data of the experimental power station from July to September 2019. Seven curves at different irradiance and temperature are randomly selected. The estimated curves are similar to the measured curves as shown in Figure 4 As shown, the results show that the IV curve output of the photovoltaic array online model established by this patent has a high degree of consistency with the measured IV curve, and the modeling accuracy is high.

[0090] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A photovoltaic array online modeling method, characterized by: The following steps are involved: Obtain the measured data of the actual operation of the photovoltaic modules, including current and voltage data as well as irradiance and temperature data; Preprocess the measured data to remove abnormal data of the photovoltaic array under mismatch conditions and data under low irradiance; Based on the current and voltage data in the preprocessed measured data, the photovoltaic array model parameters are identified using the particle swarm and gray wolf hybrid optimization algorithm to obtain the photovoltaic array model parameters. The five parameters of the photovoltaic array model are A, I ph , R s , R sh and I0, where A represents the diode ideality factor, I ph represents the avalanche breakdown current, R s and R sh They represent the equivalent series-parallel resistance of the photovoltaic array, I0 represents the diode saturation current, and the diode ideal factor A is fixed to 1 before model parameter identification; The identification of photovoltaic array model parameters using particle swarm and gray wolf hybrid optimization algorithm specifically includes the following steps: S01, initializing the optimization algorithm parameters and the gray wolf population; S02, after initialization, the fitness value of the individual gray wolf is calculated according to the fitness function and the corresponding error function, and the position information of the three leading wolves and other individual gray wolves is determined. The root mean square error is used as the fitness function of the hybrid optimization algorithm. First, the fitness value f(X i ), i = 1, 2, ..., pop, as shown in formulas (2) to (4): f(X i )=RMSE i (2) Among them, the root mean square error RMSE i Represents X i The corresponding degree of deviation between the simulation data and the measured data, M represents the number of points of the measured IV curve, and X represents the unknown photovoltaic array model parameters that need to be identified, that is, A, I ph , R s , R sh and I0, V m and I m They represent the measured voltage and current respectively; A represents the diode ideal factor, and K is the Boltzmann constant, which is 1.38×10 -23 J / K, T is the temperature of the photovoltaic cell, m is the avalanche breakdown index, and q is the electron charge constant, specifically 1.602×10 -19 C.V bd is the PN junction breakdown voltage, a is the coefficient related to the avalanche breakdown current; Sort the fitness values ​​of the gray wolves and save the location information of the α wolf with the best fitness value in the population as X a , save the location information of the β wolf with the second best fitness value in the population as X β , save the location information of the wolf with the third best fitness in the population as X δ , the remaining gray wolf is the lowest-ranking individual gray wolf ω in the population; S03, using the particle position update formula of the particle swarm optimization algorithm to replace the position formula of the gray wolf optimization algorithm, iteratively calculating the position information to obtain the forward speed and position of the individual gray wolf toward the three leading wolves, and locating the final position of the individual gray wolf; S04, determine whether the maximum number of iterations has been reached, otherwise jump to S02 and continue to execute, if yes, output the corresponding global optimal result, that is, the model parameter I ph , R s , R sh and I0; According to the distribution of each model parameter, different functions are selected for fitting by the least square method to establish the analytical equation between the model parameters and the irradiance and temperature; Using the established analytical equations, the corresponding model parameters are solved through the measured irradiance and temperature to build a real-time simulation model.

2. A photovoltaic array online modeling method according to claim 1, characterized in that: The current and voltage data are obtained through an inverter with integrated current and voltage curve scanning function, denoted as [I mea , V mea ], where I mea and V mea Represents current and voltage respectively.

3. A photovoltaic array online modeling method according to claim 2, characterized in that: The measured data preprocessing includes abnormal point detection method, smoothing filtering method and environmental stability detection method to obtain the current and voltage data of the photovoltaic module that meets the conditions, which is recorded as [I i，mea , V i，mea ], i = (1, ..., N), where N represents the number of current and voltage data points.

4. A photovoltaic array online modeling method according to claim 1, characterized in that: In S01, the optimization algorithm parameters are initialized including: population number pop, maximum number of iterations Iter max , the number of parameters to be determined d, the range of parameter update coefficients, where the maximum number of iterations Iter max Indicates the termination condition of the algorithm; Initializing the gray wolf population includes: randomly generating a gray wolf population, initializing each gray wolf position in the search space i=1,2,...,pop, initialize according to the following formula: in, and Respectively represent the upper and lower bounds of the position parameters of the i-th dimension, represents the value of the i-th dimension position parameter of the j-th gray wolf, pop represents the total number of the gray wolf population, d represents the dimension of the position parameter, and rand represents a random number in [0, 1].

5. The photovoltaic array online modeling method according to claim 1, characterized in that: In S03, the distances D between the three wolves α, β, and δ and the other individual gray wolf ω are calculated respectively. a , D β , D δ , where the particle speed and position update formula in the particle swarm optimization algorithm replaces the wolf position update formula in the gray wolf optimization algorithm, and the inertia weight k is introduced to coordinate the global search and local search capabilities of the hybrid optimization algorithm, as follows: Among them, D a , D β , D δ Respectively represent the distances between the three leading wolves α, β, δ and the other individual gray wolf ω, C1, C2 and C3 are random vectors, and X(t) represents the position of the current individual gray wolf ω; C1, C2 and C3 are calculated as follows: C=2r2·a (6) Where r1 and r2 are random vectors between the interval [0, 1], vector a is a column vector consisting of convergence factors a, and E is a row vector whose elements are all 1. The convergence factor a is calculated as follows: Among them, t is the current iteration number, Iter max is the maximum number of iterations; The particle position update formula of the particle swarm optimization algorithm is used to replace the position formula of the gray wolf optimization algorithm. The expression after replacement is as follows: X(t+1)=X(t)+v(t+1) (10) v(t) and X(t) are the speed and position of the wolf pack at the tth iteration, X(t+1) is the position information of the wolf pack after iteration, k is the inertia weight, which is a random number between [0-1], c1, c2 and c3 are the individual learning factors of the gray wolf, r1, r2 and r3 are random numbers between [0, 1], A1, A2 and A3 are coefficient vectors, and the calculation method is A=2(r1-E)·a (11).

6. A photovoltaic array online modeling method according to claim 1, characterized in that: The model parameter R is adjusted using the power exponential function. s , R sh and irradiance G d Fitting is performed using a linear function to adjust the model parameters I ph and irradiance G d Fitting is performed and the model parameter I0 and temperature T are fitted using an exponential function.

Images