CN106169910A - Photovoltaic cell parameter identification method based on group hunting algorithm - Google Patents

Abstract

The invention discloses a kind of photovoltaic cell parameter identification method based on group hunting, for the photovoltaic cell parameter of identification Single-Stage Grid Connected Solar Inverter System, comprise the following steps: obtain photovoltaic array output voltage and output electric current and the output voltage of photovoltaic cell and output electric current；Build photovoltaic cell mechanism model, determine object function and need the parameter of identification；Initialization model parameter, determines finder, follower and rogue；Update finder, follower and the position of rogue；Calculate the target function value that new position is corresponding, redistribute finder, follower and rogue；Output parameter identification result.The present invention passes through randomness and the local optimal searching of finder of rogue, optimal speed is fast, maintaining global search simultaneously always, efficiently solve optimization problem and be absorbed in the problem of local minimum, the optimization problem for multi-modal high dimensional nonlinear function has advantage clearly.

Description

Photovoltaic cell parameter identification method based on group hunting algorithm

Technical field

The present invention relates to a kind of photovoltaic cell parameter identification method based on group hunting algorithm, belong to energy technology field.

Background technology

In recent years, photovoltaic power generation technology achieves with the power generation characteristics of its uniqueness and develops widely and application.And photovoltaic electric Pond is the important component part of photovoltaic array in photovoltaic generating system, and photovoltaic cell output characteristic can directly affect whole photovoltaic and send out The overall output characteristics of electricity system, calculates the output of photovoltaic generating system, it was predicted that the generated energy of photovoltaic plant, energy exactly Enough more reasonably arrange unit output, contribute to being incorporated into the power networks and dispatching of large-scale photovoltaic electricity generation system.Can so setting up The model of accurate description photovoltaic cell output characteristic is to carry out photovoltaic system to correlational studyes such as effect on power system analyses and to imitate Genuine basis.

In terms of photovoltaic cell model, Chinese scholars is done a lot of work, is summed up and mainly has three kinds of models: light The photovoltaic module mould that photovoltaic assembly U-I characteristic model (mechanism model), engineering simplified model and consideration partial phantom block Type.Wherein, mechanism model is due to clear concept, and characteristic is more consistent with actual measurement, is therefore accepted by most researcheres.

And actually, traditional photovoltaic cell parameter identification and power forecasting method, how based on photovoltaic cell factory The data that family provides.But, after being photovoltaic array by photovoltaic cell connection in series-parallel, each cell operating status in array also differs Cause, now utilize data that producer provides cannot the operational factor of accurate recognition photovoltaic array, also cannot calculate to a nicety photovoltaic The output of array.

The most conventional parameter identification method has method of least square, neural network algorithm and particle cluster algorithm etc. Deng.Tradition parameter identification method based on method of least square realizes simple, but this algorithm is adapted to solve system of linear equations, photovoltaic The parameter identification problem of battery can not describe with linear relationship significantly, and therefore method of least square has very in this problem of solution Big limitation；Parameter identification method convergence rate based on neural network algorithm is slow, is easily absorbed in local optimum；Based on population The parameter identification method of algorithm has optimization time length, is easily absorbed in precocity, realizes the problems such as cumbersome.

Summary of the invention

The technical issues that need to address of the present invention are to provide a kind of photovoltaic cell parameter identification method based on group hunting.

The present invention uses following technical proposals:

A kind of photovoltaic cell parameter identification method based on group hunting, described parameter identification method is used for identification single stage type light The photovoltaic cell of photovoltaic grid-connected system, described photovoltaic cell is by m string n and forms photovoltaic array: comprise the following steps:

Step 1: obtain photovoltaic array output voltage U_L1With output electric current I_L1And the output voltage U of photovoltaic cell_LAnd output Electric current I_L, by forming the most step by step:

Step 1-1: read the output voltage U of described photovoltaic array_L1With output electric current I_L1；

Step 1-2: the output voltage of described photovoltaic cell is U_L=U_L1/ m, output electric current is I_L=I_L1/n；

Step 2: build photovoltaic module mechanism model:

$I_L=I_{ph.ref}-I_{o.ref}\exp \{\[\frac{q(U_L+I_L{R}_s)}{nKT}\]-1\}-\frac{(U_L+I_L{R}_s)}{R_{sh}}---\left(1\right)$

Wherein, U_L、I_LIt is respectively output voltage and output electric current, I_ph.refFor the photogenerated current under standard test condition, I_o.refFor the diode reverse saturation current under standard test condition, n is diode quality factor, R_sFor series resistance, R_shFor Parallel resistance, T is the absolute temperature of battery, and K is Boltzmann constant, and K=1.380*10^ (-23), q are electron charge, q= 1.680*10^(-19)；

Step 3: arrange object function be fitness be root-mean-square error R_MES；

$R_{MSE}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{\[f(U_L,I_L,x)\]}^2}---\left(2\right)$

Wherein, x=[I_ph.ref,I_o.ref,R_s,R_sh, n] and it is the described photovoltaic cell parameter that needs identification；N is measurement data The number of sampled point；Fitness is:

$f(U_L,I_L,x)=I-(I_{ph.ref}-I_{o.ref}\exp \{\[\frac{q(U_L+I_L{R}_s)}{nKT}\]-1\}-\frac{(U_L+I_L{R}_s)}{R_{sh}})---\left(3\right)$

Wherein, I is the actual measurement output electric current of photovoltaic array；

Step 4: initialize: group scale a, maximum transfer distance l are set_max, steering locking angle α_max, maximum iteration time Maximum attempts try_number that MAXGEN, finder update；Initial position { the X of all members_i, (i=1,2, ... a) and angle { ψ_i}X_i=(x₁,x₂,...,x_a),ψ_i=(ψ_i1,ψ_i2,...,ψ_i(a-1)), wherein, x_iFor described photovoltaic cell Need the parameter of identification；Finder's update times is 0；Iterations is 0；

Step 5: calculate the target function value of each member, chooses the minimum member of target function value as finder, and presses Randomly selecting the 80% of non-finder member according to default ratio is follower, and remaining is rogue；

Step 6: finder, follower and rogue are carried out location updating；It is made up of step in detail below:

Step 6-1: the search behavior of finder；By forming the most step by step:

Step 6-1-1: finder's update times adds 1；Finder starts search from current location, then in search volume The front of current location, left side, right side are scanned respectively, update position respectively in three directions, then count respective adaptation Degree, preferably puts adaptive value as new finder, and three directions update according to equation below respectively:

Front:

Right:

Left:

Wherein, X^k _pFor the position of finder in kth time iteration；r₁Be a meansigma methods be 0, standard deviation be 1 normal state divide The random number of cloth；r₂It is generally evenly distributed in the random number of (0,1)；θ_maxFor maximum transfer angle, it is scalar；D_p ^k={ D_i ^kIs The direction of search, D_i ^k=(d_i1 ^k,d_i2 ^k,...,d_im ^k), the direction of search is a unit vector, by search angle calculation, process As follows:

If the fitness of three positions is not all better than current fitness, change scanning angle the most according to the following formula, the most at random Take a little；

Step 6-1-2: judge that whether finder's update times is equal to try_number；If it is, turn to step 6-1-3； Otherwise turn to step 6-1-1；

Step 6-1-3: the search angle of finder is updated to

Step 6-1-4: judge whether group members position exceeds and set interval, if it is, turn to step 6-1-5；No Then turn to step 6-2；

Step 6-1-5: the position of random initializtion group members in search volume；

Step 6-2: the location updating of follower such as following formula:

$X_i^{k+1}=X_i^k+r_3(X_p^k-X_i^k)$

Wherein, X_i ^kFor i-th follower position in kth time iteration, r₃For being evenly distributed on the random number of (0,1)；

Step 6-3: the location updating of rogue is as follows:

Wherein, l_w=α * r₁l_max, wherein, r₁Be a meansigma methods be 0, standard deviation is the random number of the normal distribution of 1, α For the steering angle of rogue in iterative process；X_w ^kIt it is the w rogue position in kth time iteration；

Step 7: calculate the target function value R after group members location updating_MESValue；

Step 8: judge that iterations, whether equal to maximum iteration time MAXGEN, turns to step 9 if not, otherwise proceeds to Step 10；

Step 9: iterations adds 1；Turn to step 5,

Step 10: the position of finder is exported as the identification result of battery parameter.

Use and have the beneficial effects that produced by technique scheme:

1, the present invention passes through randomness and the local optimal searching of finder of rogue, and optimal speed is quickly.

2, the present invention addition by rogue, makes whole iterativecurve continuous decrease, efficiently solves optimization problem The problem being absorbed in local minimum, and find out optimal solution in whole search volume rapidly.

3, the present invention ensures that in search procedure group members is searched in the range of limiting, if search is jumped out during border, just returns Initial position to search volume, it is to avoid out-of-limit, improves the precision of result.

4, the present invention has advantage clearly for the optimization problem of multi-modal high dimensional nonlinear function.

Accompanying drawing explanation

Fig. 1 is photovoltaic cell mechanism model principle schematic；

Fig. 2 is photovoltaic cell parameter identification method flow chart based on group hunting algorithm.

Detailed description of the invention

The present invention is further detailed explanation with detailed description of the invention below in conjunction with the accompanying drawings.

As it is shown in figure 1, photovoltaic cell is equivalent to general P-N junction, the model of battery can use backward diode and electric current The parallel circuit in source comes equivalent.Photovoltaic cell mathematical model is set up, with model according to photovoltaic cell mechanism model principle schematic For object of study, push over out the photogenerated current I under unknown parameter to be identified in mechanism model, i.e. standard test condition_ph.ref, Diode reverse saturation current I under standard test condition_o.ref, diode quality factor n, series resistance R_s, parallel resistance R_sh, And with root-mean-square error R_MSEFor object function.

As in figure 2 it is shown, a kind of photovoltaic cell parameter identification method based on group hunting, described parameter identification method is used for distinguishing Knowing the photovoltaic cell of Single-Stage Grid Connected Solar Inverter System, described photovoltaic cell is by m string n form photovoltaic array: include with Lower step:

Step 1: obtain photovoltaic array output voltage U_L1With output electric current I_L1And the output voltage U of photovoltaic cell_LAnd output Electric current I_L, by forming the most step by step:

Step 1-1: read the output voltage U of described photovoltaic array_L1With output electric current I_L1；

Step 1-2: the output voltage of described photovoltaic cell is U_L=U_L1/ m, output electric current is I_L=I_L1/n；

Step 2: build photovoltaic module mechanism model:

$I_L=I_{ph.ref}-I_{o.ref}\exp \{\[\frac{q(U_L+I_L{R}_s)}{nKT}\]-1\}-\frac{(U_L+I_L{R}_s)}{R_{sh}}---\left(1\right)$

Wherein, U_L、I_LIt is respectively output voltage and output electric current, I_ph.refFor the photogenerated current under standard test condition, I_o.refFor the diode reverse saturation current under standard test condition, n is diode quality factor, R_sFor series resistance, R_shFor Parallel resistance, T is the absolute temperature of battery, and K is Boltzmann constant, and K=1.380*10^ (-23), q are electron charge, q= 1.680*10^(-19)；

Step 3: arrange object function be fitness be root-mean-square error R_MES；

$R_{MSE}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{\[f(U_L,I_L,x)\]}^2}---\left(2\right)$

Wherein, x=[I_ph.ref,I_o.ref,R_s,R_sh, n] and it is the described photovoltaic cell parameter that needs identification；N is measurement data The number of sampled point；Fitness is:

$f(U_L,I_L,x)=I-(I_{ph.ref}-I_{o.ref}\exp \{\[\frac{q(U_L+I_L{R}_s)}{nKT}\]-1\}-\frac{(U_L+I_L{R}_s)}{R_{sh}})---\left(3\right)$

Wherein, I is the actual measurement output electric current of photovoltaic array；

Step 4: initialize: group scale a, maximum transfer distance l are set_max, steering locking angle α_max, maximum iteration time Maximum attempts try_number that MAXGEN, finder update；Initial position { the X of all members_i, (i=1,2, ... a) and angle { ψ_i}X_i=(x₁,x₂,...,x_a),ψ_i=(ψ_i1,ψ_i2,...,ψ_i(a-1)), wherein, x_iFor described photovoltaic cell Need the parameter of identification；Finder's update times is 0；Iterations is 0；

Step 5: calculate the target function value of each member, chooses the minimum member of target function value as finder, and with It is follower that machine chooses remaining 80%, and the most remaining 20% is rogue；

Step 6: finder, follower and rogue are carried out location updating；It is made up of step in detail below:

Step 6-1: the search behavior of finder；By forming the most step by step:

Step 6-1-1: finder's update times adds 1；Finder starts search from current location, then in search volume The front of current location, left side, right side are scanned respectively, update position respectively in three directions, then count respective adaptation Degree, preferably puts adaptive value as new finder, and three directions update according to equation below respectively:

Front:

Right:

Left:

Wherein, X^k _pFor the position of finder in kth time iteration；r₁Be a meansigma methods be 0, standard deviation be 1 normal state divide The random number of cloth；r₂It is generally evenly distributed in the random number of (0,1)；θ_maxFor maximum transfer angle, it is scalar；D_p ^k={ D_i ^kIs The direction of search, D_i ^k=(d_i1 ^k,d_i2 ^k,...,d_im ^k), the direction of search is a unit vector, by search angle calculation, process As follows:

If the fitness of three positions is not all better than current fitness, change scanning angle the most according to the following formula, the most at random Take a little；

Step 6-1-2: judge that whether finder's update times is equal to try_number；If it is, turn to step 6-1-3； Otherwise turn to step 6-1-1；

Step 6-1-3: the search angle of finder is updated to

Step 6-1-4: judge whether group members position exceeds and set interval, if it is, turn to step 6-1-5；No Then turn to step 6-2；

Step 6-1-5: the position of random initializtion group members in search volume；

Step 6-2: the location updating of follower such as following formula:

$X_i^{k+1}=X_i^k+r_3(X_p^k-X_i^k)$

Wherein, X_i ^kFor i-th follower position in kth time iteration, r₃For being evenly distributed on the random number of (0,1)；

Step 6-3: the location updating of rogue is as follows:

Wherein, l_w=α * r₁l_max, wherein, r₁Be a meansigma methods be 0, standard deviation is the random number of the normal distribution of 1, α For the steering angle of rogue in iterative process；X_w ^kIt it is the w rogue position in kth time iteration；

Step 7: calculate the target function value R after group members location updating_MESValue；

Step 8: judge that iterations, whether equal to maximum iteration time MAXGEN, turns to step 9 if not, otherwise proceeds to Step 10；

Step 9: iterations adds 1；Turn to step 5,

Step 10: the position of finder is exported as the identification result of battery parameter.

Claims (1)

1. a photovoltaic cell parameter identification method based on group hunting, described parameter identification method is used for identification stage photovoltaic single The photovoltaic cell parameter of grid-connected system, described photovoltaic cell is by m string n and forms photovoltaic array: it is characterized in that: include Following steps:

Step 1: obtain photovoltaic array output voltage U_L1With output electric current I_L1And the output voltage U of photovoltaic cell_LWith output electric current I_L, by forming the most step by step:

Step 1-1: read the output voltage U of described photovoltaic array_L1With output electric current I_L1；

Step 1-2: the output voltage of described photovoltaic cell is U_L=U_L1/ m, output electric current is I_L=I_L1/n；

Step 2: build photovoltaic module mechanism model:

$I_L=I_{ph.ref}-I_{o.ref}\exp \{\[\frac{q(U_L+I_L{R}_s)}{nKT}\]-1\}-\frac{(U_L+I_L{R}_s)}{R_{sh}}---\left(1\right)$

Wherein, U_L、I_LIt is respectively output voltage and output electric current, I_ph.refFor the photogenerated current under standard test condition, I_o.refFor Diode reverse saturation current under standard test condition, n is diode quality factor, R_sFor series resistance, R_shFor electricity in parallel Resistance, T is the absolute temperature of battery, and K is Boltzmann constant, and K=1.380*10^ (-23), q are electron charge, q=1.680* 10^(-19)；

Step 3: arrange object function be fitness be root-mean-square error R_MES；

$R_{MSE}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{\[f(U_L,I_L,x)\]}^2}---\left(2\right)$

Wherein, x=[I_ph.ref,I_o.ref,R_s,R_sh, n] and it is the described photovoltaic cell parameter that needs identification；N is adopting of measurement data The number of sampling point, fitness is:

$f(U_L,I_L,x)=I-(I_{ph.ref}-I_{o.ref}\exp \{\[\frac{q(U_L+I_L{R}_s)}{nKT}\]-1\}-\frac{(U_L+I_L{R}_s)}{R_{sh}})---\left(3\right)$

Wherein, I is the actual measurement output electric current of photovoltaic array；

Step 4: initialize: group scale a, maximum transfer distance l are set_max, steering locking angle α_max, maximum iteration time Maximum attempts try_number that MAXGEN, finder update；Initial position { the X of all members_i, (i=1,2, ... a) and angle { ψ_i}X_i=(x₁,x₂,...,x_a),ψ_i=(ψ_i1,ψ_i2,...,ψ_i(a-1)), wherein, x_iFor described photovoltaic cell Need the parameter of identification；Finder's update times is 0；Iterations is 0；

Step 5: calculate the target function value of each member, chooses the minimum member of target function value as finder, and according in advance If ratio randomly select the 80% of non-finder member for follower, remaining is rogue；

Step 6: finder, follower and rogue are carried out location updating；It is made up of step in detail below:

Step 6-1: the search behavior of finder；By forming the most step by step:

Step 6-1-1: finder's update times adds 1；Finder starts search from current location, then current in search volume The front of position, left side, right side are scanned respectively, update position respectively in three directions, then count respective fitness, Adaptive value is preferably put as new finder, and three directions update according to equation below respectively:

Front:

Right:

Left:

Wherein, X^k _pFor the position of finder in kth time iteration；r₁Be a meansigma methods be 0, standard deviation is the normal distribution of 1 Random number；r₂It is generally evenly distributed in the random number of (0,1)；θ_maxFor maximum transfer angle, it is scalar；D_p ^k={ D_i ^kFor searching for Direction, D_i ^k=(d_i1 ^k,d_i2 ^k,...,d_im ^k), the direction of search is a unit vector, and by search angle calculation, process is such as Under:

If the fitness of three positions is not all better than current fitness, changes scanning angle the most according to the following formula, take the most at random a little；

Step 6-1-2: judge that whether finder's update times is equal to try_number；If it is, turn to step 6-1-3；Otherwise Turn to step 6-1-1；

Step 6-1-3: the search angle of finder is updated to

Step 6-1-4: judge whether group members position exceeds and set interval, if it is, turn to step

6-1-5；Otherwise turn to step 6-2；

Step 6-1-5: the position of random initializtion group members in search volume；

Step 6-2: the location updating of follower such as following formula:

$X_i^{k+1}=X_i^k+r_3(X_p^k-X_i^k)$

Wherein, X_i ^kFor i-th follower position in kth time iteration, r₃For being evenly distributed on the random number of (0,1)；

Step 6-3: the location updating of rogue is as follows:

Wherein, l_w=α * r₁l_max, wherein, r₁Be a meansigma methods be 0, standard deviation is the random number of the normal distribution of 1, and α is for repeatedly The steering angle of rogue during Dai；X_w ^kIt it is the w rogue position in kth time iteration；

Step 7: calculate the target function value R after group members location updating_MESValue；

Step 8: judge that iterations, whether equal to maximum iteration time MAXGEN, turns to step 9 if not, otherwise proceeds to step 10；

Step 9: iterations adds 1；Turn to step 5,

Step 10: the position of finder is exported as the identification result of battery parameter.