CN115618721A - Whale photovoltaic model parameter optimization method based on information sharing search strategy and NM simplex type - Google Patents

Abstract

The invention discloses a whale photovoltaic model parameter optimization method based on an information sharing search strategy and an NM simplex type. The introduced information sharing search strategy can strengthen the communication among individuals and enable the individuals to better carry out local development, the NM simplex strategy can enable an algorithm to search in a wider solution space and can obtain better solution quality, and the result proves that after the two strategies are added, the improved original whale optimization algorithm is respectively improved by 30.01%, 24.24%, 50.98% and 6.94% on a single diode model, a double diode model, a triple diode model and a PV model, and the method has higher capability of solving the problem of solar photovoltaic model parameter extraction.

Description

Whale photovoltaic model parameter optimization method based on information sharing search strategy and NM simplex type

Technical Field

The invention relates to the technical field of detection of solar photovoltaic cells and photovoltaic power generation arrays, in particular to a whale photovoltaic model parameter optimization method based on an information sharing search strategy and NM simplex type.

Background

More and more countries and regions are beginning to move from fossil energy to renewable energy in order to achieve economic sustainability. In order to cope with the increasing demand for energy, solar photovoltaic systems based on solar cells have been developed more and more rapidly in recent years. They can convert solar energy into electricity with little environmental hazard, because of the advantages of high efficiency and cleanliness, making this technology increasingly a niche in worldwide electricity production.

Scientists have applied photovoltaic technology to solar energy systems to increase their efficiency of energy conversion and the rate of loss during the conversion process. The characteristics of photovoltaic systems under different environments are simulated by adopting an equivalent circuit model, and a Single Diode Model (SDM) and a Double Diode Model (DDM) can be established through a mathematical model and have relatively simple structures, so that the two models are widely applied. However, the photovoltaic model is often exposed to outdoor harsh environment, and the occurrence of failure and performance degradation is inevitable. Therefore, it is crucial to correctly evaluate the performance of a photovoltaic panel in actual operation, how to design, evaluate and optimize a photovoltaic system efficiently. The actual performance of the photovoltaic system mainly depends on the unknown parameters, so the evaluation of the unknown parameters can better understand the performance of the photovoltaic system in actual work, but the SDM and DDM are implicit transcendental equations, and the evaluation of the unknown parameters becomes extremely difficult. How to be able to efficiently and reliably propose unknown parameters becomes very important.

In order to find a good parameter evaluation method, researchers have conducted a series of experiments such as an analytical method, a direct method, a numerical calculation method, a method of extracting parameters using a special function, a method of extracting parameters by means of a Lambert W function, and the like. However, since various constraints of the objective function greatly limit the application range of the optimization method, the method is not universal, and meanwhile, a large number of gradient calculations are required in the calculation process, which can not avoid falling into local optimization.

The group intelligent algorithm (SI) is widely concerned by researchers because of its strong applicability, is not limited by any black box problem, can find an approximately optimal solution to a problem in a reasonable time, and does not need to perform gradient computation, and has been widely studied in the fields of feature selection, engineering optimization, image segmentation, etc., such as optimizing parameters of a Photovoltaic (PV) model by a multi-population parallel co-evolutionary differential evolution (mppcced) method, obtaining better parameter identification accuracy on the Photovoltaic (PV) model by combining random learning and Simplex (NMs), and designing a fractional order chaotic set particle swarm optimization algorithm in order to model SDM, DDM, and TDM under different environments.

Although the meta-heuristic algorithm obtains good performance in application, a plurality of problems still exist, firstly, the meta-heuristic algorithm is easy to fall into local optimization, and has a further space for improving the quality and the convergence rate of a solution, secondly, the meta-heuristic algorithm is known according to the free lunch theorem, the algorithm needs to be specifically analyzed for specific problems, so that the algorithm is prompted to continuously explore and design an algorithm suitable for the problem aiming at different practical problems, therefore, two mechanisms are added into the whale optimization algorithm, and the two mechanisms supplement each other and play a better effect together. The whale optimization algorithm has very strong global search capability, can search in the whole solution space, and can determine the approximate range of the optimal solution. The information sharing search mechanism can carry out rough local search on the solution obtained in the whale optimization algorithm, and the interval where the optimal solution is located is narrowed. The NM simplex type can perform more precise search on the interval where the optimal solution is located, find out the optimal vector group and output the optimal vector group.

Disclosure of Invention

The invention provides a whale photovoltaic model parameter optimization method based on an information sharing search strategy and NM simplex type, which can efficiently obtain an optimized parameter value of a photovoltaic model and has a small error with an actually measured parameter value.

A whale photovoltaic model parameter optimization method based on an information sharing search strategy and NM simplex type comprises the following steps:

(1) Setting a parameter range of the photovoltaic model according to the structural model of the photovoltaic array, and setting the parameter range of the photovoltaic model as an initial parameter vector set;

(2) Performing global search on the initial parameter vector set by adopting a whale optimization algorithm to obtain an initial optimization parameter vector set;

(3) Performing local search on the initial optimization parameter vector set by adopting an information sharing search strategy to obtain an optimization parameter vector set;

(4) Selecting a first optimal parameter vector from the initial optimal parameter vector set obtained in the step (2) and the optimal parameter vector set obtained in the step (3) based on the fitness value by adopting a greedy strategy, and taking a parameter vector set where the first optimal parameter vector is located as a first parameter vector set;

(5) Local search is carried out on the first optimal parameter vector by adopting an NM simplex type to obtain a second optimal parameter vector;

(6) If the fitness value of the second optimal parameter vector is lower than that of the first optimal parameter vector, the second optimal parameter vector is used as the optimal parameter vector, the first optimal parameter vector in the first parameter vector set is replaced by the second optimal parameter vector to form a second parameter vector set, the second parameter vector set is input to the step (2) to carry out whale optimization algorithm global search, and if the fitness value of the second optimal parameter vector is higher than that of the first optimal parameter vector, the first optimal parameter vector is used as the optimal parameter vector, and the first parameter vector set is input to the step (2) to carry out whale optimization algorithm global search;

(7) And (5) iterating the steps (2) - (6) until an iteration time threshold is reached, and stopping iteration, wherein the obtained optimal parameter vector is used as the optimized photovoltaic model parameter vector.

The structural model of the photovoltaic array is a single-diode five-parameter model, a double-diode seven-parameter model, a three-diode nine-parameter model and a photovoltaic module model based on a five-parameter single-diode solar cell;

wherein, the output current I of the single-diode five-parameter model _L1 Comprises the following steps:

wherein, I _ph Current generated by light, I _sd Representing reverse saturation current, q being the basic charge, V _L To output a voltage, R _s Representing the series resistance, n is the ideal coefficient of the diode, k is the Boltzmann constant, T is the Kelvin temperature, R _sh Current I representing shunt resistance _sh Comprises the following steps:

five unknown parameters, i.e. [ I ] _ph ,I _sd ,R _s ,R _sh ,n]；

Output current I of double-diode seven-parameter model _L2 Comprises the following steps:

wherein seven unknown parameters of the dual-diode seven-parameter model are [ I ] _ph ,I _sd1 ,I _sd2 ,R _s ,R _sh ,n ₁ ,n ₂ ]. Wherein, I _sd1 And I _sd2 For reverse saturation current of the diode, n ₁ And n ₂ Is an ideal factor of a double diode tube;

output current I of three-diode nine-parameter model _L3 Comprises the following steps:

there are mainly nine unknown parameters in the three-diode model as [ I _ph ,I _sd1 ,I _sd2 ,I _sd3 ,R _s ,R _sh ,n ₁ ,n ₂ ,n ₃ ]. Wherein, I _sd1 ，I _sd2 And I _sd3 Is a three-diode reverse saturation current，n ₁ ，n ₂ And n ₃ Is an ideal factor of three diodes;

output current I of photovoltaic module model based on five-parameter single-diode solar cell _L4 Comprises the following steps:

wherein, the five-parameter vector of the photovoltaic module model based on the single-diode solar cell is [ I ] _ph ,I _sd ,R _s ,R _sh ,n]. Wherein I _L And V _L For actually measuring the current and voltage in the I-V curve, voltage V _L And (= KT/q), where K is boltzmann's constant and q is the basic charge amount.

And (3) carrying out global search on the initial parameter vector set by adopting a whale optimization algorithm, wherein the global search comprises the following steps:

when p < 0.5 and | A | ≧ 1, the initial optimization parameter vector set X' (t) is:

X'(t)＝x _rand -A*D

A＝2*a*r ₁ -a

a＝2-(2*t/MAX_FEs)

D＝|C*x ^* (t-1)-X(t-1)|

C＝2*r ₂

wherein x is _rand For randomly selected parameter vectors in the optimized parameter vector set or the initial parameter vector set obtained from the last iteration, the value of a decreases linearly from 2 to 0, MAX _ FEs is the maximum iteration number, t represents the current iteration number, r ₁ And r ₂ Is the random number in (0, 1), X (t-1) is the optimized parameter vector set or initial parameter vector set obtained from the last iteration, X ^* (t-1) represents the optimal parameter vector in the last iteration population, p being [0,1 ]]The random number of (2);

when p < 0.5 and | A | < 1, the initial optimization parameter vector set X' (t) is:

X'(t)＝x ^* (t-1)-A*D

when p is more than or equal to 0.5, the initial optimization parameter vector set X' (t) is as follows:

X'(t)＝X ^* (t)+D*e ^bl *cos(2πl)

wherein b is a constant and l is a random number in (0, 1).

And locally searching the initial optimized parameter vector set by adopting an information sharing search strategy to obtain an optimized parameter vector set X' (t) as follows:

wherein X '(t) is an initial optimization parameter vector set, X' _index (t) is a random parameter vector in the initial optimized parameter vector set, pa is a random number in (0, 1); j is an element of [0,1 ]]α c, β c are domain learning coefficients, d is the dimension (i.e., the number of unknowns) representing the problem being optimized.

Adopting NM simplex to carry out local search on the first optimal parameter vector to obtain a second optimal parameter vector, comprising the following steps:

(5.1) the initial simplex includes a plurality of vertices, and assigning a plurality of variable values in the first optimal parameter vector to a portion of the vertices of the initial simplex;

(5.2) determining the vertex P with the largest function value _high Second largest vertex P _sechi And the smallest vertex P _low Computing out of the vertex P _high Center P of the latter simplex _center Finally, define P _high Has a reflection point of P _refl Comprises the following steps:

P _refl ＝(1+α)P _center -αP _high

where α is called the reflection coefficient and α > 0 _low Function value of minimum vertex, y _refl As a function of the point of maximum reflection, y _sechi The value of the next largest vertex, if y _low ≤y _refl ≤y _sechi Then use P _refl Substitution of P _high Obtaining a second optimal parameter vector;

(5.3) if y _refl ≤y _low Then, the reflection point is expanded to obtain an expansion point P _exp If y is _exp ＜y _low Then P will be _exp Replacement by P _high Obtaining a second optimal parameter vector; if y _exp ≥y _low By P _refl Replacement of P _high Obtaining a second optimal parameter vector, and then returning to the step (5.2);

(5.4) when y _refl ＞y _sechi And y is _refl ≤y _high By P _refl Replacement of P _high And to P _refl Performing a shrinking operation if y _refl ＞y _high Then directly to P _high Performing a shrinking operation to obtain a shrinking vertex P _cont If y is _cont ≤y _high Then P is added _high Replacement by P _cont Obtaining a second optimal parameter vector, and then returning to the step (5.2); the formula for the calculation of the shrink vertex is as follows:

P _cont ＝βP _high +(1-β)P _center

(5.5) if y _cont ＞y _high To P is excluded from _low Compressing all vertexes except the vertex to obtain a second optimal parameter vector, i-th compressed vertex P _i Comprises the following steps:

P _i ←δP _i +(1-δ)P _low

wherein "←" represents a valuation, δ is called a compression coefficient and 0 < δ < 1, and P is recalculated _low The function value of each excepted vertex returns to the step (5.2).

Expanding the reflection point to obtain an expanded point P _exp Comprises the following steps:

P _exp ＝γP _refl +(1-γ)P _center

wherein gamma is an expansion coefficient, and gamma is more than 1.

Is directly to P _high Performing a shrinking operation to obtain a shrinking vertex P _cont Comprises the following steps:

P _cont ＝βP _high +(1-β)P _center

wherein beta is a coefficient of contraction, beta is more than 0 and less than 1.

The fitness value RMSE (X) is:

wherein N is the number of measured current data, T _i For the ith actually-measured output current, X is the optimized photovoltaic model parameter vector, f _i (. Cndot.) is the ith predicted output current.

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

the invention provides a whale optimization algorithm combining an information sharing search strategy and NM simplex. The method comprises the steps of roughly searching a parameter range in a global mode through a basic whale optimization algorithm, roughly searching a local mode through an information sharing search strategy to optimize a parameter vector set, obtaining optimized parameter vectors through a greedy strategy, finely searching the optimized parameter vectors in a local mode through an NM simplex mode, comparing fitness values of the two optimized parameter vectors, further iterating the optimized parameter vectors and the corresponding parameter vector sets, obtaining final optimized parameters of a photovoltaic array structure model after set iteration times are met, and improving accuracy of photovoltaic parameter extraction through the method. Improved whale optimization algorithms achieved 30.01%, 24.24%, 50.98% and 6.94% improvement on single diode, double diode, triple diode and PV models, respectively.

Drawings

Fig. 1 is a block flow diagram of a method for optimizing parameters of a photovoltaic model of whales based on an information sharing search strategy and an NM simplex type according to an embodiment of the present invention.

Fig. 2 is an equivalent circuit diagram of a single diode, a double diode, a triple diode nine-parameter photovoltaic module and a single diode photovoltaic cell-based photovoltaic module according to an embodiment of the present invention, where fig. 2 (a) is the equivalent circuit diagram of the single diode, fig. 2 (b) is the equivalent circuit diagram of the double diode, fig. 2 (c) is the equivalent circuit diagram of the triple diode nine-parameter photovoltaic module, and fig. 2 (d) is the equivalent circuit diagram of the single diode photovoltaic cell-based photovoltaic module.

Fig. 3 is a flowchart of a method for optimizing parameters of a photovoltaic model of whales based on an information sharing search strategy and an NM simplex type according to an embodiment of the present invention.

FIG. 4 is a diagram illustrating NM simplex reflection operation according to an embodiment of the present invention.

FIG. 5 is a schematic diagram of NM simplex expansion operation provided by an embodiment of the present invention.

FIG. 6 shows a schematic diagram of NM simplex according to an embodiment of the present invention when P is _refl Is superior to P _high Schematic diagram of the shrinking operation.

FIG. 7 shows a NM simplex having a median P value according to an embodiment of the present invention _high Is superior to P _refl Schematic diagram of the pinch operation.

FIG. 8 illustrates a compression operation (P) in NM simplex provided by an embodiment of the present invention _refl Is superior to P _high Case (c).

FIG. 9 shows an NM simplex compression operation (P) according to an embodiment of the present invention _refl P _high Is superior to P _refl Case (c).

Fig. 10 is a graph of RMSE convergence curves on a single diode for a whale optimization photovoltaic model parameter method based on an information sharing search strategy and NM simplex provided by an embodiment of the present invention (where ISNMWOA denotes the algorithm, IJAYA denotes an improved hybrid frog leap algorithm, GOTLBO denotes an optimization algorithm based on teaching and learning of opponent learning, MLBSA denotes a backtracking search algorithm of a multiple learning strategy, GOFPANM denotes a flower pollination optimization algorithm based on opponent learning and NM simplex, EHHO denotes an adaptive orthogonal harrisck eagle optimization algorithm, GWO denotes a wolf optimization algorithm, and WOA denotes a whale optimization algorithm).

Fig. 11 is a RMSE convergence graph on a double diode for the whale optimization photovoltaic model parameter method based on the information sharing search strategy and the NM simplex provided by the embodiment of the present invention (where ISNMWOA denotes the present algorithm, IJAYA denotes the improved hybrid leapfrog algorithm, GOTLBO denotes the optimization algorithm of teaching and learning based on opponent learning, MLBSA denotes the backtracking search algorithm of the multi-learning strategy, GOFPANM denotes the flower pollination optimization algorithm based on opponent learning and the NM simplex, EHHO denotes the adaptive orthogonal harris eagle optimization algorithm, GWO denotes the grayish optimization algorithm, and WOA denotes the whale optimization algorithm).

Fig. 12 is a graph of RMSE convergence curves on three diodes for a whale optimization photovoltaic model parameter method based on an information sharing search strategy and NM simplex provided by an embodiment of the present invention (where ISNMWOA denotes the algorithm, IJAYA denotes an improved hybrid frog leap algorithm, GOTLBO denotes an optimization algorithm based on teaching and learning of opponent learning, MLBSA denotes a backtracking search algorithm of a multiple learning strategy, GOFPANM denotes a flower pollination optimization algorithm based on opponent learning and NM simplex, EHHO denotes an adaptive orthogonal harrisck eagle optimization algorithm, GWO denotes a wolf optimization algorithm, and WOA denotes a whale optimization algorithm).

Fig. 13 is a RMSE convergence graph on a PV model of a whale optimization photovoltaic model parameter method based on an information sharing search strategy and an NM simplex provided by an embodiment of the present invention (where ISNMWOA denotes the present algorithm, IJAYA denotes an improved hybrid leapfrog algorithm, GOTLBO denotes an optimization algorithm of teaching and learning based on opponent learning, MLBSA denotes a backtracking search algorithm of a multi-learning strategy, GOFPANM denotes a flower pollination optimization algorithm based on opponent learning and an NM simplex, EHHO denotes an adaptive orthogonal harris eagle optimization algorithm, GWO denotes a grayish optimization algorithm, and WOA denotes a whale optimization algorithm).

Fig. 14 is a graph of I-V curve fitting of simulated experimental data and actual measured data in photovoltaic models on SDM, DDM, TDM, and PV according to the method for optimizing photovoltaic model parameters based on information sharing search strategy and NM simplex provided by the embodiment of the present invention, where fig. 14 (a) is a graph of I-V curve fitting of simulated experimental data and actual measured data in photovoltaic models on SDM, fig. 14 (b) is a graph of I-V curve fitting of simulated experimental data and actual measured data in photovoltaic models on DDM, fig. 14 (c) is a graph of I-V curve fitting of simulated experimental data and actual measured data in photovoltaic models on DDM, and fig. 14 (d) is a graph of I-V curve fitting of simulated experimental data and actual measured data in photovoltaic models on PV.

Detailed Description

The invention will be further illustrated and described with reference to specific embodiments.

The invention provides a whale optimization algorithm based on an information sharing search strategy and Nelder-Mead simplex for identifying unknown parameters of a photovoltaic model, as shown in figure 1, the algorithm comprises the following steps:

(1) Setting a parameter range of the photovoltaic model according to the structural model of the photovoltaic array, and setting the parameter range of the photovoltaic model as an initial parameter vector set;

(2) Performing global search on the initial parameter vector set by adopting a whale optimization algorithm to obtain an initial optimization parameter vector set;

(3) Performing local search on the initial optimization parameter vector set by adopting an information sharing search strategy to obtain an optimization parameter vector set;

(4) Selecting a first optimal parameter vector from the initial optimal parameter vector set obtained in the step (2) and the optimal parameter vector set obtained in the step (3) based on the fitness value by adopting a greedy strategy, and taking a parameter vector set where the first optimal parameter vector is located as a first parameter vector set;

(5) Local search is carried out on the first optimal parameter vector by adopting an NM simplex type to obtain a second optimal parameter vector;

(6) If the fitness value of the second optimal parameter vector is lower than that of the first optimal parameter vector, taking the second optimal parameter vector as an optimal parameter vector, replacing the first optimal parameter vector in the first parameter vector set with the second optimal parameter vector to form a second parameter vector set, inputting the second parameter vector set into the step (2) to perform whale optimization algorithm global search, and if the fitness value of the second optimal parameter vector is higher than that of the first optimal parameter vector, taking the first optimal parameter vector as an optimal parameter vector, and inputting the first parameter vector set into the step (2) to perform whale optimization algorithm global search;

(7) And (5) iterating the steps (2) - (6) until an iteration time threshold is reached, and stopping iteration, wherein the obtained optimal parameter vector is used as the optimized photovoltaic model parameter vector.

As shown in fig. 2, the structural models of the photovoltaic array are a single-diode five-parameter model (SDM), a two-diode seven-parameter model (DDM), a three-diode nine-parameter model (TDM), and a five-parameter single-diode solar cell based photovoltaic module model (PV module model);

wherein, FIG. 2 (a) is an equivalent circuit diagram of a single diode, which is composed of a current source connected in parallel with the diode, a shunt resistor representing leakage current and a series resistor representing load current loss, and the output current I of the five-parameter model of the single diode _L1 Comprises the following steps:

wherein, I _ph Current generated by light, I _sd Representing reverse saturation current, q is the basic charge (1.60217646X 10) ^-19 C)，V _L To output a voltage, R _s Representing the series resistance, n is the ideal coefficient of the diode, and k is the Boltzmann constant (1.3806503X 10) ²³ J/K), T is the Kelvin temperature, R _sh Indicating shunt resistance, current I of shunt resistance _sh Comprises the following steps:

five unknown parameters, i.e. [ I ] _ph ,I _sd ,R _s ,R _sh ,n]；

Fig. 2 (b) is an equivalent circuit diagram of a double diode, which is realized by adding a diode to an original single diode to characterize the complex loss of carriers in a dissipation region. Output current I of double-diode seven-parameter model _L2 Comprises the following steps:

wherein seven unknown parameters of the dual-diode seven-parameter model are [ I ] _ph ,I _sd1 ,I _sd2 ,R _s ,R _sh ,n ₁ ,n ₂ ]. Wherein, I _sd1 And I _sd2 For reverse saturation current of the diode, n ₁ And n ₂ Is an ideal factor of a double diode tube;

fig. 2 (c) is an equivalent circuit diagram of a three-diode by adding one diode to the original two-diode. Output current I of three-diode nine-parameter model _L3 Comprises the following steps:

there are mainly nine unknown parameters in the three-diode model as [ I _ph ,I _sd1 ,I _sd2 ,I _sd3 ,R _s ,R _sh ,n ₁ ,n ₂ ,n ₃ ]. Wherein, I _sd1 ，I _sd2 And I _sd3 Is a three-diode reverse saturation current, n ₁ ，n ₂ And n ₃ An ideality factor of three diodes;

fig. 2 (d) is an equivalent circuit diagram of a photovoltaic module composed of a solar cell composed of a single diode. It is an interface that converts light into electrical energy, and is generally composed of a plurality of solar cells connected in series or in parallel. The photovoltaic module we use here is made of N _p The solar cells are connected in parallel, and N _s Photovoltaic module formed by serially connecting solar cells, based on output current I of photovoltaic module model of five-parameter single-diode solar cell _L4 Comprises the following steps:

wherein, the five-parameter vector of the photovoltaic module model based on the single-diode solar cell is [ I ] _ph ,I _sd ,R _s ,R _sh ,n]. In which I _L4 And V _L Is made ofMeasuring current and voltage, voltage V, in I-V curves _L = KT/q, where K is the Boltzmann constant, typically 1.3806503X 10 ^-23 J/K, q is the basic charge quantity, and the value is 1.60217646 multiplied by 10 ^-19 C。

The fitness value RMSE (X) is:

wherein N is the number of measured current data, T _i For the ith actually-measured output current, X is the optimized photovoltaic model parameter vector, f _i (. H) is the ith predicted output current. For SDM, DDM, as shown in equations (7), (8) and (9). Wherein X in the formula (7) is represented by I _ph 、I _sd 、R _s 、R _sh And n, X in the formula (8) is represented by I _ph 、I _sd1 、I _sd2 、R _s 、R _sh 、n ₁ And n ₂ Composition, X in the formula (9) is represented by I _ph 、I _sd1 、I _sd2 、I _sd3 、R _s 、R _sh 、n ₁ 、n ₂ And n ₃ And (4) forming.

The method for optimizing photovoltaic model parameters for SDM based on information sharing search strategy and NM simplex whales is shown in FIG. 3 as follows:

step 1: and setting the search range of the model parameters according to data provided by an RTC France photovoltaic solar cell manufacturer. Parameter ranges for single and double diode solar cells: I.C. A _ph ∈[0,1](A),R _s ∈[0,0.5](Ω),R _sh ∈[1,100](Ω),n∈[1,2],I _sd1 ∈[0,1](A),I _sd2 ∈[0,1](A),n ₁ ∈[1,2],n ₂ ∈[1,2]. Parameter ranges for a photovoltaic module consisting of single diode solar cells: I.C. A _ph ∈[0,2](A),I _sd ∈[0,50](A),R _s ∈[0,2](Ω),R _sh ∈[1,2000](Ω),n∈[1,50]。

Step 2: and performing global search in a feasible solution space by adopting a basic whale optimization algorithm to obtain initial vectors of a plurality of photovoltaic model parameters. The algorithm comprises the following main parameters and steps:

number of individuals, i.e. parameter vectors: n =30;

maximum number of evaluations of algorithm termination: MAX _ FEs;

initializing parameters of the information sharing search strategy: j, alpha;

the upper bound of the parameter vector is expressed as UB, and the lower bound is expressed as LB;

the population is represented by X _i = rand. (UB-LB) + LB, i = 1., N initialize the individual, x _i Representing the ith individual in the population.

The whale optimization algorithm first calculates a, C, b, l, p at each location update.

When p < 0.5 and | A | ≧ 1, the initial optimization parameter vector set X' (t) is:

X'(t)＝x _rand -A*D (10)

wherein, X _rand For randomly selected parameter vectors from the set of optimized parameter vectors or the set of initial parameter vectors from the last iteration, a is given by the formula a =2 a r ₁ A, a's value decreases linearly from 2 to 0, calculated from a =2- (2 × t/MAX _ FEs), t represents the current number of iterations, D is D = | C × x ^* (t-1) -X (t-1) | calculated by C =2 × r ₂ Is calculated to obtain ₁ And r ₂ Is the random number in (0, 1), X (t-1) is the optimized parameter vector set or initial parameter vector set obtained from the last iteration, X ^* (t-1) represents the optimal parameter vector in the last iteration population.

When p < 0.5 and | A | < 1, the initial optimization parameter vector set X' (t) is:

X'(t)＝x ^* (t-1)-A*D (11)

when p is more than or equal to 0.5, the initial optimization parameter vector set X' (t) is as follows:

X'(t)＝x ^* (t-1)+D*e ^bl *cos(2πl) (12)

wherein b is a constant and l is a random number in (0, 1). Whales swim to the prey in a spiral shape while contracting the enclosure.

Selecting a first optimal parameter vector from the initial optimal parameter vector set obtained in the step (2) and the optimal parameter vector set obtained in the step (3) based on the fitness value by adopting a greedy strategy, and taking a parameter vector set where the first optimal parameter vector is located as a first parameter vector set;

and step 3: an information sharing search strategy is performed for each individual in the population, alternating as follows:

and locally searching the initial optimized parameter vector set by adopting an information sharing search strategy to obtain an optimized parameter vector set X' (t) as follows:

wherein X '(t) is an initial optimization parameter vector set, X' _index (t) is a random parameter vector in the initial optimized parameter vector set, pa is a random number in (0, 1); j is an element of [0,1 ]]α c, β c are domain learning coefficients, d is the dimension (i.e., the number of unknowns) representing the problem being optimized.

And 4, step 4: performing a Nelder-Mead simplex operation based on the optimal individual position. The method is a local search algorithm for solving the unconstrained optimization problem and does not require any derivative information of the objective function. For the function minimization problem of d variables, the NM method uses operations of reflection, expansion, contraction, compression, etc., by comparing the objective function values of (d + 1) vertices of the simplex, then replacing the vertex with the largest objective function value with a new point, and by gradually iterating to continuously update the simplex, the simplex will eventually approach the optimal solution to the problem:

step 4-1: the initial simplex comprises a plurality of vertexes, and a plurality of variable values in the first optimal parameter vector are assigned to partial vertexes of the initial simplex;

step 4-2: determining the vertex P with the largest function value _high Second largest vertex P _sechi And the smallest vertex P _low Calculating to remove the vertex P _high Center P of the latter simplex _center Finally, as shown in FIG. 4, P is defined _high Has a reflection point of P _refl Comprises the following steps:

P _refl ＝(1+α)P _center -αP _high (14)

where α is called the reflection coefficient and α > 0 _low Function value of minimum vertex, y _refl As a function of the point of maximum reflection, y _sechi The function value of the next largest vertex, if y _low ≤y _refl ≤y _sechi Then use P _refl Substitution of P _high Obtaining a second optimal parameter vector;

step 4-3: if y _refl ≤y _low If the reflection point is expanded to obtain an expanded point P as shown in FIG. 5 _exp If y is _exp ＜y _low Then P will be _exp Replacement by P _high Obtaining a second optimal parameter vector; if y _exp ≥y _low By using _lf P _er Replacement of P _high Obtaining a second optimal parameter vector, and then returning to the step 4-2;

step 4-4: when y is _refl ＞y _sechi And y is _refl ≤y _high By P _refl Substitution of P _high As shown in FIG. 6, and for P _refl Performing a shrinking operation if y _refl ＞y _high As shown in the figure7, then directly to P _high Performing a shrinking operation to obtain a shrinking vertex P _cont If y is _cont ≤y _high Then P will be _high Substitution to P _cont Obtaining a second optimal parameter vector, and then returning to the step 4-2; the formula for the computation of the shrink vertex is as follows:

P _cont ＝βP _high +(1-β)P _center (15)

and 4-5: if y is _cont ＞y _high To P in addition to P _low Compressing all vertexes except the vertex to obtain a second optimal parameter vector, i-th compressed vertex P _i Comprises the following steps:

P _i ←δP _i +(1-δ)P _low (16)

wherein "←" represents a valuation, δ is called a compression coefficient and 0 < δ < 1, and P is recalculated _low The function value of each excluded vertex is returned to step 4-2. Compression operation when shrink fails (P) _refl Is superior to P _high Scenario) is shown in fig. 8. Compression operation when shrink fails (P) _high Is superior to P _refl Scenario) is shown in fig. 9.

And 5: if the fitness value of the second optimal parameter vector is lower than that of the first optimal parameter vector, taking the second optimal parameter vector as an optimal parameter vector, replacing the first optimal parameter vector in the first parameter vector set with the second optimal parameter vector to form a second parameter vector set, inputting the second parameter vector set to the step 2 for whale optimization algorithm global search, and if the fitness value of the second optimal parameter vector is higher than that of the first optimal parameter vector, taking the first optimal parameter vector as an optimal parameter vector, and inputting the first parameter vector set to the step 2 for whale optimization algorithm global search;

and 6: and judging whether the current evaluation time t reaches the maximum evaluation time MAX _ FEs, if not, continuing to return to the step 2 to sequentially execute the related steps, and if so, taking the obtained optimal parameter vector as the optimized photovoltaic model parameter vector.

In the invention, a function fmisearchbnd based on Nelder-Mead simplex provided by a MATLAB toolbox is adopted, and as the optimal parameters acquired by us may not be the optimal parameters expected by us in the process of starting evaluation, a large amount of time is not needed to be spent on searching nearby, the optimal values acquired gradually approach the optimal values expected by us along with the gradual increase of the evaluation times, and at the moment, the searching nearby can be carried out for multiple times, and based on the idea, the parameter representing the maximum evaluation times in fmisearchbnd is set to be 0.1 MAX_FEs. As shown in fig. 10, fig. 11, fig. 12, fig. 13 and fig. 14 (a) - (d), and compared with table 1, the optimization effect of the present algorithm is slightly better than that of GOFPANM on each model, but the running time of the present algorithm is less, that is, the better performance is obtained in less time.

The above are preferred embodiments of the present invention, and all changes made according to the technical solutions of the present invention that produce functional effects do not exceed the scope of the technical solutions of the present invention belong to the protection scope of the present invention. TABLE 1 run time (in seconds) of different algorithms on single diode, double diode, triple diode and PV models (where ISNMWOA stands for this algorithm, IJAYA stands for improved hybrid leapfrog algorithm, GOTLBO stands for optimization algorithm of teaching and learning based on opponent learning, MLBSA stands for backtracking search algorithm of multiple learning strategies, GOFPANM stands for flower pollination optimization algorithm based on opponent learning and NM simplex, EHHO stands for adaptive orthogonal Harris eagle optimization algorithm, GWOO stands for Lansiya optimization algorithm, WOA stands for whale optimization algorithm)

TABLE 2 estimation of optimal parameters for ST40 at 25 ℃ and different irradiance

TABLE 3 estimation of optimal parameters for SM55 at 25 deg.C and different irradiance

TABLE 4 optimal parameter estimation for KC200GT at 25 deg.C with different irradiance

TABLE 5 irradiance of 1000W/m ² And optimal parameter estimation for ST40 at different temperatures

TABLE 6 irradiance of 1000W/m ² And optimal parameter estimation for SM55 at different temperatures

TABLE 7. Irradiance of 1000W/m ² And optimal parameter estimation for KC200GT at different temperatures

Claims (8)

1. A whale photovoltaic model parameter optimizing method based on an information sharing search strategy and an NM simplex model is characterized by comprising the following steps:

(1) Setting a parameter range of the photovoltaic model according to the structural model of the photovoltaic array, and setting the parameter range of the photovoltaic model as an initial parameter vector set;

(2) Performing global search on the initial parameter vector set by adopting a whale optimization algorithm to obtain an initial optimization parameter vector set;

(3) Performing local search on the initial optimization parameter vector set by adopting an information sharing search strategy to obtain an optimization parameter vector set;

(4) Selecting a first optimal parameter vector from the initial optimal parameter vector set obtained in the step (2) and the optimal parameter vector set obtained in the step (3) based on the fitness value by adopting a greedy strategy, and taking a parameter vector set where the first optimal parameter vector is located as a first parameter vector set;

(5) Local search is carried out on the first optimal parameter vector by adopting an NM simplex type to obtain a second optimal parameter vector;

(6) If the fitness value of the second optimal parameter vector is lower than that of the first optimal parameter vector, taking the second optimal parameter vector as an optimal parameter vector, replacing the first optimal parameter vector in the first parameter vector set with the second optimal parameter vector to form a second parameter vector set, inputting the second parameter vector set into the step (2) to perform whale optimization algorithm global search, and if the fitness value of the second optimal parameter vector is higher than that of the first optimal parameter vector, taking the first optimal parameter vector as an optimal parameter vector, and inputting the first parameter vector set into the step (2) to perform whale optimization algorithm global search;

(7) And (5) iterating the steps (2) - (6) until an iteration time threshold is reached, and stopping iteration, wherein the obtained optimal parameter vector is used as the optimized photovoltaic model parameter vector.

2. The method for optimizing photovoltaic model parameters of whale based on information sharing search strategy and NM simplex of claim 1, wherein the structural model of the photovoltaic array is a single-diode five-parameter model, a two-diode seven-parameter model, a three-diode nine-parameter model and a five-parameter single-diode solar cell based photovoltaic module model;

wherein, the output current I of the single-diode five-parameter model _L1 Comprises the following steps:

wherein, I _ph Current generated by light, I _sd Representing reverse saturation current, q being the basic charge, V _L To output a voltage, R _s Representing the series resistance, n is the ideal coefficient of the diode, k is the Boltzmann constant, T is the Kelvin temperature, R _sh Indicating shunt resistance, current I of shunt resistance _sh Comprises the following steps:

five unknown parameters, i.e. [ I ] _ph ,I _sd ,R _s ,R _sh ,n]；

Output current I of double-diode seven-parameter model _L2 Comprises the following steps:

wherein seven unknown parameters of the dual-diode seven-parameter model are [ I ] _ph ,I _sd1 ,I _sd2 ,R _s ,R _sh ,n ₁ ,n ₂ ]Wherein, I _sd1 And I _sd2 For reverse saturation current of the diode, n ₁ And n ₂ Is an ideal factor of a double diode tube;

output current I of three-diode nine-parameter model _L3 Comprises the following steps:

there are mainly nine unknown parameters in the three-diode model as [ I _ph ,I _sd1 ,I _sd2 ,I _sd3 ,R _s ,R _sh ,n ₁ ,n ₂ ,n ₃ ]Wherein, I _sd1 ，I _sd2 And I _sd3 Is a three-diode reverse saturation current, n ₁ ，n ₂ And n ₃ Is an ideal factor of three diodes;

output current I of photovoltaic module model based on five-parameter single-diode solar cell _L4 Comprises the following steps:

wherein, the five-parameter vector of the photovoltaic module model based on the single-diode solar cell is [ I ] _ph ,I _sd ,R _s ,R _sh ,n]In which I _L And V _L For actually measuring the current and voltage in the I-V curve, voltage V _L And (= KT/q), where K is boltzmann's constant and q is the basic charge amount.

3. The method of claim 1, wherein the whale optimization algorithm is used to perform a global search on an initial set of parameter vectors, and the method comprises:

when p < 0.5 and | A | ≧ 1, the initial optimization parameter vector set X' (t) is:

X'(t)＝x _rand -A*D

A＝2*a*r ₁ -a

a＝2-(2*t/MAX_FEs)

D＝|C*x ^* (t-1)-X(t-1)|

C＝2*r ₂

wherein x is _rand Is derived fromThe value of a of the optimized parameter vector set or the randomly selected parameter vector in the initial parameter vector set obtained from the last iteration is linearly decreased from 2 to 0, MAX _ FEs is the maximum iteration number, t represents the current iteration number, r ₁ And r ₂ Is the random number in (0, 1), X (t-1) is the optimized parameter vector set or initial parameter vector set obtained from the last iteration, X ^* (t-1) represents the optimal parameter vector in the last iteration population, p being [0,1]The random number of (2);

when p < 0.5 and | A | < 1, the initial optimization parameter vector set X' (t) is:

X'(t)＝x ^* (t-1)-A*D

when p is more than or equal to 0.5, the initial optimization parameter vector set X' (t) is as follows:

X'(t)＝x ^* (t-1)+D*e ^bl *cos(2πl)

wherein b is a constant and l is a random number in (0, 1).

4. The whale model parameter optimization method based on the information sharing search strategy and the NM simplex according to claim 1, wherein the initial optimized parameter vector set is locally searched by the information sharing search strategy to obtain an optimized parameter vector set X "(t) which is:

wherein X '(t) is an initial optimization parameter vector set, X' _index (t) is a random parameter vector in the initial optimized parameter vector set, pa is a random number in (0, 1); j is an element of [0,1 ]]α c, β c are domain learning coefficients, d is the dimension (i.e., the number of unknowns) representing the problem being optimized.

5. The method of claim 1, wherein the obtaining a second optimal parameter vector by performing a local search on the first optimal parameter vector using NM simplex, comprises:

(5.1) the initial simplex includes a plurality of vertices, and assigning a plurality of variable values in the first optimal parameter vector to a portion of the vertices of the initial simplex;

(5.2) determining the vertex P with the largest function value _high Second largest vertex P _sechi And the smallest vertex P _low Calculating to remove the vertex P _high Center P of rear simplex _center Finally, define P _high Is P _refl Comprises the following steps:

P _refl ＝(1+α)P _center -αP _high

where α is called the reflection coefficient and α > 0 _low Function value of minimum vertex, y _refl As a function of the point of maximum reflection, y _sechi The function value of the next largest vertex, if y _low ≤y _refl ≤y _sechi Then use P _refl Replacement of P _high Obtaining a second optimal parameter vector;

(5.3) if y _refl ≤y _low Then, the expansion operation is performed on the reflection point to obtain an expansion point P _exp If y is _exp ＜y _low Then P will be _exp Replacement by P _high Obtaining a second optimal parameter vector; if y _exp ≥y _low By P _refl Substitution of P _high Obtaining a second optimal parameter vector, and then returning to the step (5.2);

(5.4) when y _refl ＞y _sechi And y is _refl ≤y _high By P _refl Substitution of P _high And to P _refl Performing a shrinking operation if y _refl ＞y _high Then directly to P _high Performing a shrinking operation to obtain a shrinking vertex P _cont If y is _cont ≤y _high Then P will be _high Substitution to P _cont Obtaining a second optimal parameter vector, and then returning to the step (5.2); the formula for the calculation of the shrink vertex is as follows:

P _cont ＝βP _high +(1-β)P _center

(5.5) if y _cont ＞y _high To P is excluded from _low Compressing all vertexes except the vertex to obtain a second optimal parameter vector, i-th compressed vertex P _i Comprises the following steps:

P _i ←δP _i +(1-δ)P _low

wherein "←" represents a valuation, δ is called a compression coefficient and 0 < δ < 1, and P is recalculated _low The function value of each excepted vertex returns to the step (5.2).

6. The method of claim 5, wherein the dilation operation is performed on the reflection points to obtain dilation point P, and the dilation point P is a parameter of the optimized photovoltaic model of whale based on the information sharing search strategy and NM simplex type _exp Comprises the following steps:

P _exp ＝γP _refl +(1-γ)P _center

wherein gamma is an expansion coefficient, and gamma is more than 1.

7. The method of optimizing photovoltaic model parameters for whales based on an information sharing search strategy and NM simplex as claimed in claim 5, wherein P is directly paired with P _high Performing a shrinking operation to obtain a shrinking vertex P _cont Comprises the following steps:

P _cont ＝βP _high +(1-β)P _center

wherein beta is a shrinkage coefficient, and beta is more than 0 and less than 1.

8. The method for optimizing photovoltaic model parameters for whales based on an information sharing search strategy and NM simplex according to claim 2, wherein the fitness value RMSE (X) is:

wherein N is the number of measured current data, T _i For the ith actually-measured output current, X is the optimized photovoltaic model parameter vector, f _i (. Cndot.) is the ith predicted output current.

Images