阅读说明：本技术 基于全局最大功率点跟踪和混合优化算法的光伏在线参数辨识方法 (Photovoltaic online parameter identification method based on global maximum power point tracking and hybrid optimization algorithm ) 是由 陈志聪 罗林禄 吴丽君 程树英 林培杰 于 2021-06-25 设计创作，主要内容包括：本发明涉及一种基于全局最大功率点跟踪和混合优化算法的光伏在线参数辨识方法,首次提出一种突变点检测算法用于动态工作点全局最大功率点跟踪过程的波形突变时刻检测,并将GMPPT对应的波形转换成静态关键段I-V特性曲线,并提出一种新的混合法即基于量子粒子群算法和列文伯格——马夸尔特算法的光伏模型在线参数辨识方法,该方法实现逆变器并网发电过程中GMPPT扫描的识别并提取关键段I-V特性曲线,结合量子粒子群算法强大的全局搜索能力和列文伯格——马夸尔特算法强大的局部搜索能力,进一步提高了光伏模型参数辨识的速度、精度、稳定性、可靠性和收敛性。(The invention relates to a photovoltaic online parameter identification method based on global maximum power point tracking and hybrid optimization algorithm, firstly provides a catastrophe point detection algorithm for waveform catastrophe moment detection in the dynamic working point global maximum power point tracking process, converting the waveform corresponding to GMPPT into a static key section I-V characteristic curve, and providing a new hybrid method, namely a photovoltaic model online parameter identification method based on quantum particle swarm optimization and Levenberg-Marquardt algorithm, the method realizes the identification of GMPPT scanning in the grid-connected power generation process of the inverter and extracts a key section I-V characteristic curve, and further improves the speed, the precision, the stability, the reliability and the convergence of the parameter identification of the photovoltaic model by combining the strong global search capability of a quantum particle group algorithm and the strong local search capability of a Levenberg-Marquardt algorithm.)

1. A photovoltaic online parameter identification method based on global maximum power point tracking and a hybrid optimization algorithm is characterized by comprising the following steps: the method comprises the following steps:

step S1: according to the time sequence voltage and current waveforms of the dynamic working points of the photovoltaic inverter, identifying the maximum power point tracking step by adopting a mutation point detection algorithm, recording the waveform mutation moment, and extracting the voltage and current waveforms under the stable state;

step S2: identifying a global maximum power point tracking process, namely a GMPPT waveform according to the voltage step change characteristics in the voltage and current waveforms, and extracting a key section I-V characteristic curve of the GMPPT corresponding to the waveform;

step S3: obtaining an I-V characteristic curve according to global maximum power point tracking extraction and the number N of series and parallel solar cells of a photovoltaic array_sAnd N_pSelecting a single-diode five-parameter model for parameter identification, and extracting five electrical parameters of photocurrent, single-diode reverse saturation current, ideal factors, equivalent series resistance and equivalent parallel resistance in the model according to the search range of the electrical parameters of the model;

step S4: carrying out global search on the photovoltaic model parameters by adopting a quantum particle group intelligent optimization algorithm, and obtaining an optimal photovoltaic model parameter initial value vector;

step S5: performing further local search by adopting a Levenberg-Marquardt algorithm and using the optimal photovoltaic model parameter vector obtained in the step S4 as an initial search parameter;

step S6: performing equivalent single-diode model parameter extraction on the two types of photovoltaic data through a mixed quantum particle group algorithm and a Levenberg-Marquardt algorithm, namely a QPSO-LM algorithm; the two types of photovoltaic data are static and complete I-V characteristic curves under different actual measurement conditions and key section I-V curves extracted in the dynamic working point global maximum power point tracking process.

2. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein:

the mutation point detection algorithm in the step S1 takes every 200 continuous photovoltaic array working voltages as a set according to the collection frequency of the dynamic working point data, randomly selects a point to divide the set into two parts, calculates the residual error between each point on both sides and the average value of each part, and finds the mutation point and mutation when the total residual error reaches the minimum valueThe point detection algorithm is shown as follows:when J takes the minimum value, record V_r＝kAnd the time node is the voltage and current waveform mutation point moment of the dynamic working point.

3. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein: in the global maximum power point tracking process in the step S2, a disturbance observation method is adopted for the operation of the inverter, and the value range of the voltage step of the disturbance is 0.2V to 4V; disturbing the output voltage of the photovoltaic array according to the voltage step 4V between the lowest working voltage of the inverter 80V and the normal working voltage of the photovoltaic array 120V to realize the GMPPT process, recording the corresponding voltage and current according to the catastrophe point moment, and extracting the I-V curve of the key section of the GMPPT process.

4. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein: in the step S3, the photovoltaic model is a single-diode five-parameter model; the mathematical model of the single diode photovoltaic module is as follows:

the mathematical model of a single diode photovoltaic array is as follows:

wherein, I_tAnd V_tCurrent and voltage values in the actually measured I-V curve are obtained; k is Boltzmann constant 1.3806503 x 10^-23J/K, q is the basic charge amount 1.60217646X 10^-19C; the single-diode photovoltaic model has five parameter vectors of [ I ]_phI_s，n，B_s，R_p]In which I_phIs photocurrent, I_sFor single diode reverse saturation current, n is the single diode idealisation factor, R_sEquivalent series resistance, R_pIs an equivalent parallel resistance, N_sAnd N_pThe number of cells connected in series and in parallel, respectively.

5. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein:

the algorithm in steps S4 and S5 finds an optimal set of parameter vectors x within the search range of a given parameter vector such that the following formula is shown:

and taking the sum variance SSE as an objective function of a hybrid optimization algorithm, namely a QPSO-LM algorithm, and when the sum variance SSE is minimum, optimally fitting the actually measured I-V curve and the calculated simulation I-V curve, namely, expressing the optimally fitting of the actually measured I-V curve and the calculated simulation I-V curve.

6. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein:

the hybrid optimization algorithm, i.e. the QPSO-LM algorithm, is optimized by, for reference data, the parameter vector range of its single-diode cell model: i is_ph∈[0，1]，I_s∈[0，1]，n∈[1，2]，R_s∈[0，0.5]，R_p∈[0，100](ii) a Single diode photovoltaic module: i is_ph∈[0，2]，I_s∈[0，50]，n∈[1，50]，R_s∈[0，2]，R_p∈[0，2000]。

7. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein: step S4 specifically includes the following steps:

step S41: according to Is a D-dimensional parameter vector of the search space, D is the number of parameters to be optimized, LB^j，UB^jUpper and lower bounds of the parameter, respectively;

step S42: setting the population number N of particles_pCalculating the fitness value of each particle in the particle swarm by the maximum iteration number Itermax, wherein the calculation formula of the fitness value is an objective function, namely the value of the sum variance SSE;

step S43: initializing the iterative times Iter of the particles to be 1, and entering an iterative search process;

step S44: calculating the average value m of the best position of particle history_best，

The calculation formula is as follows:

m is the population size, p_{best_i}An extremum representing an ith particle of the current iteration;

step S45: updating the formula according to the particle position:

calculating the optimal solution of the current particle in the search space;

step S46: calculating the fitness values of the population and the individuals, and judging whether the iteration number Iter is less than or equal to the maximum iteration number set by the algorithmIf Itermax, adding 1 to the current iteration number Iter and re-entering the step S44; otherwise, the fitness values obtained by the last iterative computation are sequenced, and the optimal initial solution vector X is output₀。

8. The photovoltaic online parameter identification method based on the global maximum power point tracking and hybrid optimization algorithm according to claim 1, wherein: the step S5 specifically includes the following steps:

step S51: the optimal initial solution vector X obtained according to step S4₀Calculating an objective function SSE and denoted as f (X), and a vector F (X) which is a difference between an actual measured value and a calculated value at a solution vector X, and calculating a Jacobian matrix J (X) of F (X);

step S52: setting iteration parameter Iter to be 0, calculating a maximum objective function value MaxFunEvals and a maximum iteration number MaxIter; an optimal optimality tolerance value tol;

step S53: the gradient vector G (x) of f (x) is calculated by the following formula: g (x) 2j (x)^TF (x); the Hessian matrix H (x), H (x) for f (x) is calculated as follows: h (x) 2j (x)^TJ (x) +2q (x); wherein the content of the first and second substances,D_i(x) For each F_i(x) The Hessian matrix of;

step S54: the algorithm performs iterations according to the equation: (J (x)_k)^TJ(x_k)+λ_kI)d_k＝-J(x_k)^TF(x_k) Calculating the search direction d_kThe direction is the solution of the linear least squares problem;

step S55: detecting a boundary, judging whether an iteration point x is in a constraint boundary, if the iteration point is out of the boundary, projecting the step to a nearest feasible point by an algorithm, and modifying the iteration point x into P (x) by the algorithm, wherein the updating method of P (x) comprises the following steps: p (x) LB if x < LB; p (x) UB if x > UB; p (x) x otherwise;

step S56: whether the objective function SSE, i.e. f (x), satisfies the algorithm termination condition 1:or termination condition 2: if not, Iter continues to add 1 and enters step S54 for iterative calculation until a termination condition is reached, and a final result is output.

Technical Field

The invention relates to the technical field of detection of solar cells and photovoltaic arrays, in particular to a photovoltaic online parameter identification method based on global maximum power point tracking and a hybrid optimization algorithm.

Background

In order to deal with the increasingly serious problems of environmental pollution, climate deterioration, fossil energy depletion and the like, solar energy has attracted extensive attention as a renewable and widely available clean energy. The photovoltaic array is used as the core of a photovoltaic power generation device and generally consists of photovoltaic modules connected in series and parallel. However, the photovoltaic module and the array usually work in a complex outdoor environment, and from the viewpoint of long-term operation, the performance of the photovoltaic module is not only reduced along with the increase of the service life, which results in the change of model parameters along with time, but also generates faults under the influence of a severe natural environment. Therefore, the modeling and parameter identification of the I-V characteristics of the dynamic working point of the photovoltaic module under the offline actual measurement condition and the grid-connected power generation condition have important application value and practical significance for the overall performance evaluation and real-time photovoltaic fault diagnosis of the power generation system.

The equivalent models of the photovoltaic module and the array are mainly divided into a single-diode five-parameter model and a double-diode seven-parameter model, and accurate and rapid identification of model parameters is the key of photovoltaic modeling. The mutation point detection method provided by the invention realizes the modeling of the grid-connected on-line I-V characteristic by extracting the static I-V characteristic from the dynamic waveform; the parameter extraction method can be used for extracting the key section I-V characteristic curve of the static actual measurement I-V characteristic curve and the dynamic working point, can extract corresponding photovoltaic array model parameters, and can update the photovoltaic array model on line by combining the grid-connected MPPT scanning process of the inverter so as to effectively evaluate the actual working condition of the photovoltaic power station.

The existing photovoltaic model parameter extraction methods can be roughly divided into three types, namely an analytical method, a numerical optimization method and a mixed method of the analytical method and the numerical optimization method. The analytic method is mainly used for solving the parameters of the photovoltaic model based on a few key point data (such as open-circuit voltage, short-circuit current, maximum power point voltage and current and temperature coefficient) given or actually measured by a photovoltaic component manufacturer by constructing an explicit equation of the photovoltaic model. Although this method is simple and less in calculation amount, the accuracy of the model parameters is poor, and the method is susceptible to the accuracy and noise of the key point data. To overcome the shortcomings of analytical methods, various deterministic and stochastic numerical optimization methods have been proposed in succession, which accurately extract model parameters by minimizing the root mean square error of simulated and measured I-V curves. Deterministic numerical optimization methods include the Newton-Raphson method, pattern search methods, and the like. Although the methods have high convergence speed and small calculation amount, the methods are easy to fall into local optimal values, and the accuracy of model parameters is easily influenced by a search starting point and is relatively unstable. The random numerical optimization method for extracting the parameters of the photovoltaic model mainly comprises differential evolution, a genetic algorithm, a bee colony algorithm, a particle swarm algorithm, a pollen propagation algorithm and the like. The algorithms have strong global search capability, but have large calculation amount and low convergence speed, and are difficult to be suitable for real-time parameter extraction. In order to utilize the advantages of such methods and overcome the disadvantages thereof, some hybrid methods have been proposed, including cuckoo combined simplex method, analytic method combined simplex method, artificial bee colony combined with confidence domain algorithm, etc. The above method has some problems, including the following aspects: firstly, few algorithms proposed in the past are transplanted to a hardware platform for parameter extraction, and the main reason is that the convergence rate is not essentially improved, and the algorithm complexity is high due to excessive iteration times of the algorithm and cannot be realized on the hardware platform; secondly, the algorithm proposed in the past is only suitable for a static complete I-V characteristic curve, and the online extraction of parameters in the grid-connected power generation process cannot be realized. At present, on-line parameter identification methods of photovoltaic models based on mutation point detection and mixed quantum particle group algorithm and Levenberg-Marquardt algorithm are not found in published documents and patents, the mutation point detection method is also used for extracting key section I-V curves in the GMPPT scanning process of the photovoltaic grid-connected inverter for the first time, and meanwhile, the mixed optimization algorithm is also used for extracting parameters of the photovoltaic models for the first time.

Disclosure of Invention

In view of the above, the present invention provides a photovoltaic online parameter identification method based on global maximum power point tracking and hybrid optimization algorithm, so as to overcome the defects of the prior art, thereby implementing grid-connected online parameter extraction and improving the performance of photovoltaic model parameter identification.

The invention is realized by adopting the following scheme: a photovoltaic online parameter identification method based on global maximum power point tracking and a hybrid optimization algorithm comprises the following steps:

step S1: according to the time sequence voltage and current waveforms of the dynamic working points of the photovoltaic inverter, identifying the maximum power point tracking step by adopting a mutation point detection algorithm, recording the waveform mutation moment, and extracting the voltage and current waveforms under the stable state;

step S2: identifying a global maximum power point tracking process, namely a GMPPT waveform according to the voltage step change characteristics in the voltage and current waveforms, and extracting a key section I-V characteristic curve of the GMPPT corresponding to the waveform;

step S3: obtaining an I-V characteristic curve according to global maximum power point tracking extraction and the number N of series and parallel solar cells of a photovoltaic array_sAnd N_pSelecting a single-diode five-parameter model for parameter identification, and extracting five electrical parameters of photocurrent, single-diode reverse saturation current, ideal factors, equivalent series resistance and equivalent parallel resistance in the model according to the search range of the electrical parameters of the model;

step S4: carrying out global search on the photovoltaic model parameters by adopting a quantum particle group intelligent optimization algorithm, and obtaining an optimal photovoltaic model parameter initial value vector;

step S5: performing further local search by adopting a Levenberg-Marquardt algorithm and using the optimal photovoltaic model parameter vector obtained in the step S4 as an initial search parameter;

step S6: performing equivalent single-diode model parameter extraction on the two types of photovoltaic data through a mixed quantum particle group algorithm and a Levenberg-Marquardt algorithm, namely a QPSO-LM algorithm; the two types of photovoltaic data are static and complete I-V characteristic curves under different actual measurement conditions and key section I-V curves extracted in the dynamic working point global maximum power point tracking process.

Further, the mutation point detection algorithm in step S1 uses every 200 consecutive photovoltaic array operating voltages as a set according to the collection frequency of the dynamic operating point data, randomly selects a point to divide the set into two parts, calculates the residual error between each point on both sides and the average value of each part, and finds the mutation point when the total residual error reaches the minimum value, where the mutation point detection algorithm is as follows:

when J takes the minimum value, record V_r＝kAnd the time node is the voltage and current waveform mutation point moment of the dynamic working point.

Further, in the global maximum power point tracking process in step S2, a disturbance observation method is adopted for the operation of the inverter, and the value range of the voltage step of the disturbance is 0.2V to 4V. Disturbing the output voltage of the photovoltaic array according to the voltage step 4V between the lowest working voltage of the inverter 80V and the normal working voltage of the photovoltaic array 120V to realize the GMPPT process, recording the corresponding voltage and current according to the catastrophe point moment, and extracting the I-V curve of the key section of the GMPPT process.

Further, the photovoltaic model in the step S3 is a single-diode five-parameter model; the mathematical model of the single diode photovoltaic module is as follows:

the mathematical model of a single diode photovoltaic array is as follows:

wherein, I_tAnd V_tFor the sum of the currents in the measured I-V curvesA voltage value; k is Boltzmann constant 1.3806503 x 10^-23J/K, q is the basic charge amount 1.60217646X 10^-19C; the single-diode photovoltaic model has five parameter vectors of [ I ]_ph,I_s,n,R_s,R_p]In which I_phIs photocurrent, I_sFor single diode reverse saturation current, n is the single diode idealisation factor, R_sEquivalent series resistance, R_pIs an equivalent parallel resistance, N_sAnd N_pThe number of cells connected in series and in parallel, respectively.

Further, the algorithm in steps S4 and S5 finds an optimal set of parameter vectors x within the search range of a given parameter vector such that the following formula is shown:

and taking the sum variance SSE as an objective function of a hybrid optimization algorithm (QPSO-LM), and when the sum variance SSE is minimum, optimally fitting the actually measured I-V curve and the calculated simulation I-V curve, namely, expressing the optimally fitting of the actually measured I-V curve and the calculated simulation I-V curve.

Further, the hybrid algorithm (QPSO-LM) operates by, for a given range of parameter vectors, for the reference data, the parameter vector range of its single diode cell model: i is_ph∈[0,1],I_s∈[0,1],n∈[1,2],R_s∈[0,0.5],R_p∈[0,100](ii) a Single diode photovoltaic module: i is_ph∈[0,2],I_s∈[0,50],n∈[1,50],R_s∈[0,2],R_p∈[0,2000]。

Further, step S4 specifically includes the following steps:

step S41: according toIs a D-dimensional parameter vector of the search space, D beingOptimizing the number of parameters, LB^j,UB^jUpper and lower bounds of the parameter, respectively;

step S42: setting the population number N of particles_pCalculating the fitness value of each particle in the particle swarm by the maximum iteration number Itermax, wherein a fitness value calculation formula is an objective function, namely the value of the sum variance SSE;

step S43: initializing the iterative times Iter of the particles to be 1, and entering an iterative search process;

step S44: calculating the average value m of the best position of particle history_bestThe calculation formula is as follows:

m is the population size, p_{best_i}An extremum representing an ith particle of the current iteration;

step S45: updating the formula according to the particle position:

calculating the optimal solution of the current particle in the search space;

step S46: calculating the fitness values of the population and the individuals, judging whether the iteration number Iter is less than or equal to the maximum iteration number IterMax set by the algorithm, if so, adding 1 to the current iteration number Iter and re-entering the step S44; otherwise, the fitness values obtained by the last iterative computation are sequenced, and the optimal initial solution vector X is output₀。

Further, the step S5 specifically includes the following steps:

step S51: the optimal initial solution vector X obtained according to step S4₀Calculating an objective function SSE and denoted as f (X), and a vector F (X) which is a difference between an actual measured value and a calculated value at a solution vector X, and calculating a Jacobian matrix J (X) of F (X);

step S52: setting iteration parameter Iter to be 0, calculating a maximum objective function value MaxFunEvals and a maximum iteration number MaxIter; an optimal optimality tolerance value tol;

step S53: the gradient vector G (x) of f (x) is calculated by the following formula: g (x) 2j (x)^TF (x); the Hessian matrix H (x), H (x) for f (x) is calculated as follows: h (x) 2j (x)^TJ (x) +2q (x); wherein the content of the first and second substances,D_i(x) For each F_i(x) The Hessian matrix of;

step S54: the algorithm performs iterations according to the equation: (J (x)_k)^TJ(x_k)+λ_kI)d_k＝-J(x_k)^TF(x_k) Calculating the search direction d_kThe direction is the solution of the linear least squares problem;

step S55: detecting a boundary, judging whether an iteration point x is in a constraint boundary, if the iteration point is out of the boundary, projecting the step to a nearest feasible point by an algorithm, and modifying the iteration point x into P (x) by the algorithm, wherein the updating method of P (x) comprises the following steps: p (x) LB if x < LB; p (x) UB if x > UB; p (x) x otherwise;

step S56: whether the objective function SSE, i.e. f (x), satisfies the algorithm termination condition 1:or termination condition 2: if not, Iter continues to add 1 and enters step S54 for iterative calculation until a termination condition is reached, and a final result is output.

Compared with the prior art, the invention has the following beneficial effects:

according to the method, a key section I-V characteristic curve in the GMPPT scanning process of the inverter is extracted by identifying the sudden change time of the time sequence voltage and current waveform of the dynamic working point; a photovoltaic model online parameter identification method of a hybrid quantum particle swarm optimization algorithm verifies a complete I-V characteristic curve of a standard data set and an actually measured data set under a static condition and a key section I-V characteristic curve extracted in a dynamic working point GMPPT process, and selects an optimal initial point by adopting a quantum particle swarm algorithm (QPSO), so that the problem of slow convergence caused by the influence of the selection of the initial point on the basis of a Levenberg-Marquardt algorithm (LM) is solved. In conclusion, compared with the existing photovoltaic model parameter identification algorithm, the method provided by the invention realizes photovoltaic grid-connected online modeling and greatly improves the speed, precision, reliability, convergence and stability of photovoltaic model parameter extraction.

Drawings

Fig. 1 is a general flowchart of a photovoltaic model online parameter identification method according to an embodiment of the present invention.

Fig. 2 is an equivalent circuit diagram of a photovoltaic module and a photovoltaic array or string in a single diode model according to an embodiment of the present invention, where fig. 2(a) is an equivalent model diagram of a single diode module, and fig. 2(b) is an equivalent model of a single diode array or string.

Fig. 3 is a schematic diagram of a photovoltaic array and a dynamic operating point voltage and current acquisition system and an inverter according to an embodiment of the present invention, where fig. 3(a) is a 3 × 6 photovoltaic array diagram, fig. 3(b) is a physical diagram of a data acquisition system, and fig. 3 (c) is a physical diagram of a photovoltaic inverter.

FIG. 4 is a diagram illustrating a detection diagram of a sudden change point of a voltage/current waveform during GMPPT of a dynamic operating point of an inverter, where in FIG. 4(a) is irradiance 582W/m²The detection of the mutation point under the working condition is shown schematically, and the irradiance is 748W/m in FIG. 4(b)²And (3) a schematic diagram of mutation point detection under the working condition. Fig. 5 is a graph of I-V curves and fits of critical segments extracted from a GMPPT scanning process based on mutation point detection according to an embodiment of the present invention, where fig. 5(a) is a graph of I-V curves of critical segments extracted from a GMPPT scanning process under different irradiances, and fig. 5(b) is a graph of I-V curve fits of critical segments. Fig. 6 is a comparison graph of convergence rates of parameter extraction results in a single diode model based on reference data according to an embodiment of the present invention and a conventional algorithm, wherein fig. 6(a) is a comparison graph of convergence rates of different algorithms on an rtc.france data set, and fig. 6(b) is a comparison graph of convergence rates of different algorithms on a Photowatt-PWP201 data set. FIG. 7 is a diagram showing the result of extracting parameters of the actually measured photovoltaic array under different conditions by using a single-diode model according to an embodiment of the present invention, where FIG. 7(a) is a diagram of parameter extraction, i.e., I-V characteristic curve fitting, of the actually measured photovoltaic array, and FIG. 7(b) is a diagram of actually measured lightA P-V curve fit plot of the volt array.

Detailed Description

The invention is further explained below with reference to the drawings and the embodiments.

It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.

The embodiment provides a photovoltaic online parameter identification method based on global maximum power point tracking and a hybrid optimization algorithm, which comprises the following steps:

step S1: according to the time sequence voltage and current waveforms of the dynamic working points of the photovoltaic inverter, identifying the maximum power point tracking step by adopting a mutation point detection algorithm, recording the waveform mutation moment, and extracting the voltage and current waveforms under the stable state;

step S2: identifying a global maximum power point tracking process, namely a GMPPT waveform according to the voltage step change characteristics in the voltage and current waveforms, and extracting a key section I-V characteristic curve of the GMPPT corresponding to the waveform;

step S3: obtaining an I-V characteristic curve according to global maximum power point tracking extraction and the number N of series and parallel solar cells of a photovoltaic array_sAnd N_p(ii) a Selecting a single-diode five-parameter model for parameter identification, and extracting five electrical parameters of photocurrent, single-diode reverse saturation current, an ideal factor, equivalent series resistance and equivalent parallel resistance in the model according to the search range of the electrical parameters of the model;

step S4: carrying out global search on the photovoltaic model parameters by adopting a quantum particle group intelligent optimization algorithm, and obtaining an optimal photovoltaic model parameter initial value vector;

step S5: performing further local search by adopting a Levenberg-Marquardt algorithm and using the optimal photovoltaic model parameter vector obtained in the step S4 as an initial search parameter;

step S6: performing equivalent single-diode model parameter extraction on the two types of photovoltaic data through a mixed quantum particle group algorithm and a Levenberg-Marquardt algorithm, namely a QPSO-LM algorithm; the two types of photovoltaic data are static and complete I-V characteristic curves under different actual measurement conditions and key section I-V curves extracted in the dynamic working point global maximum power point tracking process.

In this embodiment, the mutation point detection algorithm in step S1 uses every 200 consecutive photovoltaic array operating voltages as a set according to the collection frequency of the dynamic operating point data, randomly selects a point to divide the set into two parts, calculates the residual error between each point on both sides and the average value of each part, and when the total residual error reaches the minimum value,

then the mutation point can be found, and the mutation point detection algorithm is shown as the following formula:

when J takes the minimum value, record V_r＝kAnd the time node is the voltage and current waveform mutation point moment of the dynamic working point.

In this embodiment, in the global maximum power point tracking process in step S2, a disturbance observation method is adopted for the operation of the inverter, and the value range of the disturbance voltage is 0.2V to 4V; disturbing the output voltage of the photovoltaic array according to the voltage step 4V between the lowest working voltage of the inverter 80V and the normal working voltage of the photovoltaic array 120V to realize the GMPPT process, recording the corresponding voltage and current according to the catastrophe point moment, and extracting the I-V curve of the key section of the GMPPT process.

In this embodiment, the photovoltaic model in step S3 is a single-diode five-parameter model; the mathematical model of the single diode photovoltaic module is as follows:

the mathematical model of a single diode photovoltaic array is as follows:

wherein, I_tAnd V_tCurrent and voltage values in the actually measured I-V curve are obtained; k is Boltzmann constant 1.3806503 x 10^-23J/K, q is the basic charge amount 1.60217646X 10^-19C; the single-diode photovoltaic model has five parameter vectors of [ I ]_ph,I_s,n,R_s,R_p]In which I_phIs photocurrent, I_sFor single diode reverse saturation current, n is the single diode idealisation factor, R_sEquivalent series resistance, R_pIs an equivalent parallel resistance, N_sAnd N_pThe number of cells connected in series and in parallel, respectively.

In the present embodiment, the algorithm in steps S4 and S5 finds an optimal set of parameter vectors x within the search range of a given parameter vector such that the following formula is shown:

and taking the sum of squares error SSE as an objective function of a hybrid optimization algorithm (QPSO-LM), and when the value of the sum of squares error SSE is minimum, optimally fitting the actually measured I-V curve and the calculated simulation I-V curve, namely, indicating that the actually measured I-V curve and the calculated simulation I-V curve are optimally fitted.

In the present embodiment, the hybrid algorithm (QPSO-LM) is implemented byGiven the range of parameter vectors, for the reference data, the range of parameter vectors for its single diode cell model: i is_ph∈[0,1],I_s∈[0,1],n∈[1,2], R_s∈[0,0.5],R_p∈[0,100](ii) a Single diode photovoltaic module: i is_ph∈[0,2],I_s∈[0,50],n∈[1,50], R_s∈[0,2],R_p∈[0,2000]。

In this embodiment, step S4 specifically includes the following steps:

step S41: according to

Is a D-dimensional parameter vector of the search space, D is the number of parameters to be optimized, LB^j,UB^jUpper and lower bounds of the parameter, respectively;

step S42: setting the population number N of particles_pCalculating the fitness value of each particle in the particle swarm by the maximum iteration number Itermax, wherein a fitness value calculation formula is an objective function, namely the value of the sum variance SSE;

step S43: initializing the iterative times Iter of the particles to be 1, and entering an iterative search process;

step S44: calculating the average value m of the best position of particle history_best，

The calculation formula is as follows:

m is the population size, p_{best_i}An extremum representing an ith particle of the current iteration;

step S45: updating the formula according to the particle position:

calculating the optimal solution of the current particle in the search space;

step S46: calculating the fitness values of the population and the individuals, judging whether the iteration number Iter is less than or equal to the maximum iteration number IterMax set by the algorithm, if so, adding 1 to the current iteration number Iter and re-entering the step S44; otherwise, the fitness values obtained by the last iterative computation are sequenced, and the optimal initial solution vector X is output₀。

In this embodiment, the step S5 specifically includes the following steps:

step S51: the optimal initial solution vector X obtained according to step S4₀Calculating an objective function SSE and denoted as f (X), and a vector F (X) which is a difference between an actual measured value and a calculated value at a solution vector X, and calculating a Jacobian matrix J (X) of F (X);

step S52: setting iteration parameter Iter to be 0, calculating a maximum objective function value MaxFunEvals and a maximum iteration number MaxIter; an optimal optimality tolerance value tol;

step S53: the gradient vector G (x) of f (x) is calculated by the following formula: g (x) 2j (x)^TF (x); the Hessian matrix H (x), H (x) for f (x) is calculated as follows: h (x) 2j (x)^TJ (x) +2q (x); wherein the content of the first and second substances,D_i(x) For each F_i(x) The Hessian matrix of;

step S54: the algorithm performs iterations according to the equation: (J (x)_k)^TJ(x_k)+λ_kI)d_k＝-J(x_k)^TF(x_k) Calculating the search direction d_kThe direction is the solution of the linear least squares problem;

step S55: detecting a boundary, judging whether an iteration point x is in a constraint boundary, if the iteration point is out of the boundary, projecting the step to a nearest feasible point by an algorithm, and modifying the iteration point x into P (x) by the algorithm, wherein the updating method of P (x) comprises the following steps: p (x) LB if x < LB; p (x) UB if x > UB; p (x) xothermwise;

step S56: whether the objective function SSE, i.e. f (x), satisfies the algorithm termination condition 1:or termination condition 2: if not, Iter continues to add 1 and enters step S54 for iterative calculation until a termination condition is reached, and a final result is output.

As shown in fig. 3 to 7, preferably, the embodiment provides a mutation point detection algorithm for detecting a waveform mutation time in a global maximum power point tracking process of a dynamic working point, converts a waveform corresponding to GMPPT into a static key segment I-V characteristic curve, and provides a new hybrid method, that is, a photovoltaic model online parameter identification method based on a quantum-behaved particle swarm algorithm and a levenberg-marquardt algorithm, which realizes GMPPT scanning identification and key segment I-V characteristic curve extraction in an inverter grid-connected power generation process, and further improves speed, accuracy, stability, reliability and convergence of parameter identification of a photovoltaic model in combination with a strong global search capability of the quantum-behaved particle group algorithm and a strong local search capability of the levenberg-marquardt algorithm.

A photovoltaic online parameter identification method based on a global maximum power point tracking and hybrid optimization algorithm (QPSO-LM) comprises the following steps: step S1: according to the time sequence voltage and current waveforms of the dynamic working points of the photovoltaic inverter, identifying the maximum power point tracking step by adopting a mutation point detection algorithm, recording the waveform mutation moment, and extracting the voltage and current waveforms under the stable state; step S2: identifying the waveform of a global maximum power point tracking process (GMPPT) according to the voltage step change characteristics in the voltage and current waveform, and extracting a key section I-V characteristic curve of the GMPPT corresponding waveform; step S3: obtaining an I-V characteristic curve according to global maximum power point tracking extraction and the number N of series and parallel solar cells of a photovoltaic array_sAnd N_pSelecting a proper photovoltaic equivalent model and a search range of model electrical parameters; step S4: carrying out global search on the parameters of the photovoltaic model by adopting a quantum particle group intelligent optimization algorithm, and obtaining an optimal initial value vector of the parameters of the photovoltaic model; step S5: adopting Levenberg-Marquardt algorithm and using the optimal photovoltaic model parameter vector obtained in the step S4 as an initial searchParameters, performing further local search; step S6: the single diode model parameter extraction is carried out on two types of photovoltaic data (static complete I-V characteristic curves under different actual measurement conditions, and key section I-V curves extracted in the dynamic working point global maximum power point tracking process) through the hybrid optimization algorithm.

The specific flow diagram is shown in fig. 1. Fig. 2 is a single-diode equivalent circuit model of the photovoltaic array module or the battery of the present embodiment, wherein fig. 2a is a single-diode module equivalent model, and fig. 2b is a single-diode array or string equivalent model. The mutation point detection refers to the moment of detecting voltage mutation in the GMPPT scanning process of the dynamic working point; the photovoltaic model refers to a single-diode five-parameter model; the photovoltaic model parameter refers to photocurrent I_phDiode reverse saturation current I_sDiode idealisation factor n, equivalent series resistance R_sAnd an equivalent parallel resistance R_p。

The mathematical model of the single diode photovoltaic module is as follows:

a mathematical model of a single diode photovoltaic array or string is as follows:

wherein I_tAnd V_tCurrent and voltage values in the actually measured I-V curve are obtained; k is Boltzmann constant (1.3806503 × 10)^-23J/K), q is the basic charge amount (1.60217646X 10)^-19C) (ii) a The single-diode photovoltaic model has five parameter vectors of [ I ]_ph,I_s,n,R_s,R_p]In which I_phIs photocurrent, I_sFor single diode reverse saturation current, n is the single diode idealisation factor, R_sEquivalent series resistance, R_pIs an equivalent parallel resistance, N_sAnd N_pThe number of cells connected in series and in parallel, respectively. The embodiment is providedThe algorithm is developed by finding an optimal set of parameter vectors x within the search range of a given parameter vector such that the following equation is shown:

in the embodiment, the Sum of Squares Error (SSE) is used as an objective function of the hybrid optimization algorithm, and when the value is the minimum, the best fit between the actually measured I-V curve and the calculated simulation I-V curve is obtained, that is, the best fit between the actually measured I-V curve and the calculated simulation I-V curve is represented.

In step S1, first, low-pass filtering is performed on the voltage and current waveforms of the dynamic working point to filter out a 50Hz power frequency interference signal, then, according to the acquisition frequency of the dynamic working point data, every 200 continuous photovoltaic array working voltages are used as a set, one point is randomly selected to divide the set into two parts, residual errors between each point on both sides and the average value of each part are calculated, and when the total residual error reaches a minimum value, a mutation point can be found, wherein a mutation point detection algorithm is shown as the following formula:when J takes the minimum value, record V_r＝kThe time node is the time of the voltage and current waveform mutation point of the dynamic working point; extracting voltage and current waveforms under a steady state;

step S2: a global maximum power point scanning process is a disturbance observation method adopted for the work of an inverter, namely a GMPPT process realized by disturbing the output voltage of a photovoltaic array according to a voltage step length between the lowest working voltage of the inverter 80V and the normal working voltage of the photovoltaic array 120V, corresponding voltage and current are recorded according to the time of a catastrophe point, and a key section I-V curve from 80V to 120V is extracted;

step S3: the specific model parameters and algorithm out-of-parameter ranges are set as follows, for the reference data of a static complete I-V curve,single diode cell model: i is_ph∈[0,1],I_s∈[0,1],n∈[1,2],

R_s∈[0,0.5],R_p∈[0,100]. Single diode photovoltaic module: i is_ph∈[0,2],I_s∈[0, 50],n∈[1,50],R_s∈[0,2],R_p∈[0,2000]. Number of bee colony N of QPSO algorithm_p15, 780, and 200, which are the maximum iteration count maxter of the LM algorithm. For a static complete actual measurement I-V curve test instrument model of PROVA-1011 of Taiwan Taishi, for a key section I-V curve test instrument model extracted from voltage and current waveforms in the process of tracking the global maximum power point of an actual measurement dynamic working point, the model is Gude Wei GW3000-NS, and the photovoltaic model parameter search range of the experimental data is I_ph:[05N_P](A),I_s:[01N_p](μA), n:[1N_s2N_s],R_s:[01N_s/N_p](Ω),R_p:[0N_s*100](Ω), wherein N_sNumber of solar cells connected in series for the photovoltaic model (photovoltaic array/module string/module), N_pThe number of the solar cells connected in parallel in the photovoltaic array/module string/module is shown.

Step S4: and carrying out global search on the parameters of the photovoltaic model by adopting a quantum particle swarm optimization algorithm, and obtaining an optimal initial value vector of the parameters of the photovoltaic model.

Step S41: according toIs a D-dimensional parameter vector of the search space, D is the number of parameters to be optimized, LB^j,UB^jUpper and lower bounds of the parameter, respectively;

step S42: setting the population number N of particles_pCalculating the fitness value of each particle in the particle swarm by the maximum iteration number Itermax;

step S43: initializing the iterative times Iter of the particles to be 1, and entering an iterative search process;

step S44: calculating the average value m of the best position of particle history_bestThe calculation formula is as follows:

m is the population size, p_{best_i}An extremum representing an ith particle of the current iteration;

step S45: updating the formula according to the particle position:

calculating the optimal solution of the current particle in the search space;

step S46: calculating the fitness values of the population and the individuals, judging whether the iteration number Iter is less than or equal to the maximum iteration number IterMax set by the algorithm, if so, adding 1 to the current iteration number Iter and re-entering the step S44; otherwise, the fitness values obtained by the last iterative computation are sequenced, and the optimal initial solution vector X is output₀。

Step S5: performing further local search by adopting a Levenberg-Marquardt algorithm and using the optimal photovoltaic model parameter vector obtained in the step S4 as an initial search parameter;

step S51: the optimal initial solution vector X obtained according to step S4₀Calculating an objective function SSE and denoted as f (X), and a vector F (X) which is a difference between an actual measured value and a calculated value at a solution vector X, and calculating a Jacobian matrix J (X) of F (X);

step S52: setting iteration parameter Iter to be 0, calculating a maximum objective function value MaxFunEvals and a maximum iteration number MaxIter; an optimal optimality tolerance value tol;

step S53: the gradient vector G (x) of f (x) is calculated by the following formula: g (x) 2j (x)^TF (x); the Hessian matrix H (x), H (x) for f (x) is calculated as follows: h (x) 2j (x)^TJ (x) +2q (x); wherein the content of the first and second substances,D_i(x) For each F_i(x) The Hessian matrix of;

step S54: the algorithm performs iterations, based onFormula (II): (J (x)_k)^Tw(x_k)+λ_kI)d_k＝-J(x_k)^TF(x_k) Calculating the search direction d_kThe direction is the solution of the linear least squares problem;

step S55: detecting a boundary, judging whether an iteration point x is in a constraint boundary, if the iteration point is out of the boundary, projecting the step to a nearest feasible point by an algorithm, and modifying the iteration point x into P (x) by the algorithm, wherein the updating method of P (x) comprises the following steps: p (x) LB if x < LB; p (x) UB if x > UB; p (x) x otherwise;

step S56: whether the objective function SSE, i.e. f (x), satisfies the algorithm termination condition 1:or termination condition 2: if not, Iter continues to add 1 and enters step S54 for iterative calculation until a termination condition is reached, and a final result is output.

Preferably, in this example, two types of I-V curves (a static complete I-V characteristic curve under different actual measurement conditions, a key segment I-V curve extracted from a voltage current waveform in a dynamic working point global maximum power point tracking process) of the actually measured photovoltaic array under different illuminance and temperature are fitted to extract parameters of the single diode photovoltaic model, and the results are shown in tables 1 to 2, fig. 5 and fig. 7. As can be seen from the Root Mean Square Error (RMSE) of curve fitting in tables 1-2, the method provided in this embodiment can perform accurate curve fitting, and fully embodies the accuracy of this embodiment.

TABLE 1 results of static complete I-V curve fitting and parameter extraction under different illumination and temperature

TABLE 2 key segment I-V curve fitting and parameter extraction results in GMPPT process of dynamic working points under different illumination and temperature

Table 3 comparison table of photovoltaic module parameter extraction results under single diode model based on reference data for this embodiment and existing algorithm

Pararneter extraction	Iph	Io(μA)	n	Rs	Rp	RMSE(Explicit function)	RMSE(Implicit function)
								Proposed method	0.7607755	0.32302	1.48118	0.036377	53.71852	9.8602E-04	7.7539E-04
GS-INMS	0.7608	0.323	1.4812	0.0363	53.7185	9.8602E-04	7.7543E-04
								DE(Three-point)	0.76072	0.31911	1.47986	0.03629	54.19241	1.0733E-03	8.1291E-04
CSO	0.76078	0.323	1.48118	0.03638	53.7185	9.8602E-04	7.7544E-04
								ISCE	0.760776	0.32302	1.48118	0.03638	53.7185	9.8602E-04	7.7543E-04
EHA-NMS	0.760776	0.32302	1.48118	0.03638	53.7185	9.8602E-04	7.7543E-04
								STLBO	0.76078	0.32302	1.48114	0.03638	53.7187	9.8602E-04	7.8059E-04
Rcr-IJADE	0.760776	0.32302	1.48118	0.03638	53.7185	9.8602E-04	7.7543E-04
								ABC-TRR	0.760776	0.32302	1.48118	0.03638	53.7185	9.8602E-04	7.7543E-04
NM-MPSO	0.76078	0.32306	1.4812	0.03638	53.7222	9.8602E-04	7.7550E-04
								TLABC	0.76078	0.32302	1.48118	0.03638	53.7164	9.8602E-04	7.7542E-04
LMSA	0.7608	0.3185	1.4798	0.0364	53.3264	9.86E-04	7.8079E-04
								PCE	0.760776	0.323021	1.481074	0.03638	53.7185	9.8602E-04	8.0947E-04
PSO	0.76077	0.32454	1.48165	0.03636	53.855	9.8606E-04	7.7612E-04
								BMO	0.76077	0.32479	1.48173	0.03636	53.8716	9.8608E-04	7.7621E-04
MABC	0.760779	0.321323	1.481385	0.03639	53.4	9.8610E-04	1.7275E-03
								ABC	0.7608	0.3251	1.4817	0.0364	53.6433	9.8620E-04	8.3343E-04
GOTLBO	0.76078	0.331552	1.48382	0.03627	54.1154	9.8744E-04	7.7988E-04
								GGHS	0.76092	0.3262	1.48217	0.03631	53.0647	9.9097E-04	7.8146E-04
ABSO	0.7608	0.30623	1.47583	0.03659	52.2903	9.9124E-04	7.7368E-04
								IGHS	0.76077	0.34351	1.4874	0.03613	53.2845	9.9306E-04	8.2116E-04
HS	0.7607	0.305	1.4754	0.0366	53.5946	9.95E-04	7.7850E-04
								CPSO	0.7607	0.4	1.5033	0.0354	59.012	1.3900E-03	1.0255E-03

The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.