The content of the invention
Iteration is pursued it is an object of the invention to provide a kind of extrapolation with excellent stability, accuracy and rapidity
The emulation mode of the MPPT algorithm of method.
Technical solution of the invention is:
A kind of emulation mode of the MPPT algorithm of pursuit iterative method of extrapolating, it is characterized in that:Comprise the following steps:
(1) it is actual control process algorithm to use variable step extrapolation, and process is as follows:
The first step:Any step-length is given, integral approximation is calculated with trapezoid formula
Wherein:T1For the value that compound trapezoidal integeration is tried to achieve;A, b are two interval end points;F (a) is interval endpoint a correspondences
Value;F (b) is the corresponding values of interval endpoint b;
Second step:Integral approximation is asked for again by variable step extrapolation.Order
Calculate T now2n:
Wherein:H is step-length;T2nIt is that interval is divided into the timesharing such as 2n, is asked with compound trapezoidal integeration in each minizone
The value for obtaining;It is minizone left end point [xi,xi+1] midpoint;N is the n-th decile;
3rd step:For precision controlling, and if only if | R2n-Rn|<During ε, stop calculating, take R now2nIt is next step
Approximation, otherwise continues to reduce by half step-length, turns second step execution;
Wherein:R2nIt is the error of the timesharing such as 2n;RnIt is the error of the n-th grade timesharing;ε is precision;
(2) optimized using pursuit iterative method
In simulation process, when being iterated to power output every time, I is usedk, UkWhole components carry out product, obtain this
When pk, bring desired value into iterative formula again, obtain new Ik+1, Uk+1, pk+1, needs when such process causes to calculate
Retain two approximate solution pk, pk+1, waste so has been resulted in the information calculated, increase iterative process step,
Reduce arithmetic speed;Iterative process is improved with extrapolation, a new component is often calculated just new with this at once
Data replace corresponding legacy data to be iterated, then convergence rate faster, and now only needs to store a newest data
;
Wherein:IkCurrent value when being walked for K;UkMagnitude of voltage when being walked for K;pkPerformance number when being walked for K;Ik+1For K+1 is walked
When current value;Uk+1Magnitude of voltage when being walked for K+1;pk+1Performance number when being walked for K+1.
The actual algorithm flow of step (1) is:
The first step:Start
Second step:Two the endpoint values a, b and permissible accuracy ε of input interval;The photovoltaic array output of measurement k moment PV
Voltage U(k), electric current I(k), and calculate power P now(k);
3rd step:K is entered as 1, k values since 1, step-length h values are b-a;To ensure that power is last samples value
P1, using trapezoid formula T1=(b-a) [f (a)+f (b)]/2=h [f (a)+f (b]/2, obtain P now1;
4th step:Gather the voltage U at (k+1) T moment(k+1), electric current I(k+1), calculate power P(k+1);Calculate △ P now
=P(k+1)-P(k);
5th step:Judge △ P >=ε;
6th step:Such as if so, will now obtain voltage increment △ U and be entered as 0, the initial value x of step-length is entered as a+h/2, instead
It, voltage U now is entered as U+ △ U, and the initial value x of step-length is entered as x+h;
7th step:Circulation above step, material calculation h=(b-a)/2i(i=0,1,2...), precision ε=(b-a)
[P(k+1)+P(k)];
8th step:Judge △ P now>b-a;
9th step:Such as if so, voltage increment △ U are entered as-△ U;Conversely, directly performing next step;
Tenth step:Uref(k+1)=Uref(k)+△U;
11st step:Return to resampling, output;
Wherein:P1 is the initial value of power;X is the initial value of step-length;ε is precision;IkCurrent value when being walked for K;UkIt is K
Magnitude of voltage during step;pkPerformance number when being walked for K;pk+1Performance number when being walked for K+1;Ik+1Current value when being walked for K+1;Uk+1
Magnitude of voltage when being walked for K+1;K is walked for K;H is step-length;A, b are two interval end points;△ P are the variable quantity of power;△U
It is the variable quantity of voltage;Uref(k+1) reference voltage level when for K+1;UrefReference voltage level when () is K k.
The emulation mode of the MPPT algorithm of described extrapolation pursuit iterative method, it is characterized in that:The actual algorithm of step (2)
Flow is:
The first step:Voltage U when measurement kth is walkedi k, electric current
Second step:Calculate power△Pk, △ U=Ui k+1-Ui k;
3rd step:Value for variable i=1,2,3... performs the process for chasing after;
4th step:Judge △ Pk<0;
5th step:Such as if so, judge now Ui k+1>Ui k;
6th step:If meeting the four, the 5th steps simultaneously then performs Uref k+1=Uref k-△U;Otherwise perform Uref k+1=
Uref k+△U;
7th step:For variable i=n-1, n-2, n-3... carries out value, and the process caught up with is performed to variable i;
8th step:Again to voltage Uk, electric current IkCarry out assignment;
9th step:Return, and export;
Wherein:IkCurrent value when being walked for K;UkMagnitude of voltage when being walked for K;pkPerformance number when being walked for K;Ik+1For K+1 is walked
When current value;Uk+1Magnitude of voltage when being walked for K+1;pk+1Performance number when being walked for K+1;K is walked for K;△ U are the change of voltage
Amount;△ P are the variable quantity of power;Uref(k+1) reference voltage level when for K+1;UrefReference voltage level when () is K k;A, b are area
Between two end points;ε is precision.
The analogous diagram realized with conventional control algolithm of the invention and data carry out careful analysis and compare, with excellent
Stability, accuracy and rapidity.
Specific embodiment
A kind of emulation mode of the MPPT algorithm of pursuit iterative method of extrapolating, comprises the following steps:
(1) it is actual control process algorithm to use variable step extrapolation, and process is as follows:
The first step:Any step-length is given, integral approximation is calculated with trapezoid formula
Wherein:T1For the value that compound trapezoidal integeration is tried to achieve;A, b are two interval end points;F (a) is interval endpoint a correspondences
Value;F (b) is the corresponding values of interval endpoint b;
Second step:Integral approximation is asked for again by variable step extrapolation.Order
Calculate T now2n:
Wherein:H is step-length;T2nIt is that interval is divided into the timesharing such as 2n, is asked with compound trapezoidal integeration in each minizone
The value for obtaining;It is minizone left end point [xi,xi+1] midpoint;N is the n-th decile;
3rd step:For precision controlling, and if only if | R2n-Rn|<During ε, stop calculating, take R now2nIt is next step
Approximation, otherwise continues to reduce by half step-length, turns second step execution;
Wherein:R2nIt is the error of the timesharing such as 2n;RnIt is the error of the n-th grade timesharing;ε is precision;
(2) optimized using pursuit iterative method
In simulation process, when being iterated to power output every time, I is usedk, UkWhole components carry out product, obtain this
When pk, bring desired value into iterative formula again, obtain new Ik+1, Uk+1, pk+1, needs when such process causes to calculate
Retain two approximate solution pk, pk+1, waste so has been resulted in the information calculated, increase iterative process step,
Reduce arithmetic speed;Iterative process is improved with extrapolation, a new component is often calculated just new with this at once
Data replace corresponding legacy data to be iterated, then convergence rate faster, and now only needs to store a newest data
;
Wherein:IkCurrent value when being walked for K;UkMagnitude of voltage when being walked for K;pkPerformance number when being walked for K;Ik+1For K+1 is walked
When current value;Uk+1Magnitude of voltage when being walked for K+1;pk+1Performance number when being walked for K+1.
The actual algorithm flow of step (1) is:
The first step:Start
Second step:Two the endpoint values a, b and permissible accuracy ε of input interval;The photovoltaic array output of measurement k moment PV
Voltage U(k), electric current I(k), and calculate power P now(k);
3rd step:K is entered as 1, k values since 1, step-length h values are b-a;To ensure that power is last samples value
P1, using trapezoid formula T1=(b-a) [f (a)+f (b)]/2=h [f (a)+f (b]/2, obtain P now1;
4th step:Gather the voltage U at (k+1) T moment(k+1), electric current I(k+1), calculate power P(k+1);Calculate △ P now
=P(k+1)-P(k);
5th step:Judge △ P >=ε;
6th step:Such as if so, will now obtain voltage increment △ U and be entered as 0, the initial value x of step-length is entered as a+h/2, instead
It, voltage U now is entered as U+ △ U, and the initial value x of step-length is entered as x+h;
7th step:Circulation above step, material calculation h=(b-a)/2i(i=0,1,2...), precision ε=(b-a)
[P(k+1)+P(k)];
8th step:Judge △ P now>b-a;
9th step:Such as if so, voltage increment △ U are entered as-△ U;Conversely, directly performing next step;
Tenth step:Uref(k+1)=Uref(k)+△U;
11st step:Return to resampling, output;
Wherein:P1 is the initial value of power;X is the initial value of step-length;ε is precision;IkCurrent value when being walked for K;UkIt is K
Magnitude of voltage during step;pkPerformance number when being walked for K;pk+1Performance number when being walked for K+1;Ik+1Current value when being walked for K+1;Uk+1
Magnitude of voltage when being walked for K+1;K is walked for K;H is step-length;A, b are two interval end points;△ P are the variable quantity of power;△U
It is the variable quantity of voltage;Uref(k+1) reference voltage level when for K+1;UrefReference voltage level when () is K k.
The emulation mode of the MPPT algorithm of described extrapolation pursuit iterative method, it is characterized in that:The actual algorithm of step (2)
Flow is:
The first step:Voltage U when measurement kth is walkedi k, electric current
Second step:Calculate power△Pk, △ U=Ui k+1-Ui k;
3rd step:Value for variable i=1,2,3... performs the process for chasing after;
4th step:Judge △ Pk<0;
5th step:Such as if so, judge now Ui k+1>Ui k;
6th step:If meeting the four, the 5th steps simultaneously then performs Uref k+1=Uref k-△U;Otherwise perform Uref k+1=
Uref k+△U;
7th step:For variable i=n-1, n-2, n-3... carries out value, and the process caught up with is performed to variable i;
8th step:Again to voltage Uk, electric current IkCarry out assignment;
9th step:Return, and export;
Wherein:IkCurrent value when being walked for K;UkMagnitude of voltage when being walked for K;pkPerformance number when being walked for K;Ik+1For K+1 is walked
When current value;Uk+1Magnitude of voltage when being walked for K+1;pk+1Performance number when being walked for K+1;K is walked for K;△ U are the change of voltage
Amount;△ P are the variable quantity of power;Uref(k+1) reference voltage level when for K+1;UrefReference voltage level when () is K k;A, b are area
Between two end points;ε is precision.
The principle of variable step extrapolation:
, it is necessary to provide appropriate step-length in the implementation process of maximum power point control algolithmIf
It is too big that step-length takes, then precision is difficult to meet.If it is too small that step-length takes, it will cause the increase of amount of calculation in control process.
An appropriate step-length was given before system emulation, is often difficult to realize.In actual emulation, usually using the side of variable step
Method, basic thought is:During step-length halves successively, recycling quadrature formula carries out computing, until step-length folding twice
The absolute value of the integration differential after half | I2n(f)-In(f)|<The precision σ of permission, then quadrature process terminate, and by newest I2n
F () is used as new quadrature approximation.Time [a, b] is carried out into n deciles, then step-length isTnFor:
If by [a, b] interval 2n deciles, then step-length will also be changed into original half, then now step-length isAnd this
When T2nFor:
For compound quadrature formula, its algorithm is absolutely convergent, therefore, the sequence of approximation is gone out with method construct by half
T (h) is classified as,And in actual control process, usually with an approximating sequence T1,T2,
...Tn... .., goes constantly to approach exact value T, lucky sequence { Ti(i=1,2...) and generally cease manner of breathing with interval step-length
Close.But, because this convergence of algorithm speed is slower, and convergence rate how is improved in actual control process and saves calculating
Amount becomes a new problem, not enough for this point, and algorithm above is improved, and is generalized to more generally situation.Give
A fixed convergent sequenceIt is converged to f (0), a new convergence is reconstructed on the basis of this convergent sequence of numbers
Sequence can quickly converge on f (0), algorithm is obtained secondary acceleration.
The implementation process for specifically applying to this paper is as follows, carries out Taylor expansion to any given step-length h first:
If f'(0 now) ≠ 0, then nowThe rank for approaching f (0) is all O (h).Now by (3), (4)
Two formulas are improved, and obtain:
If now f " (0) ≠ 0, thenThe order of error for approaching f (0) is O (h2).I.e. new sequenceF (0) is converged on faster than old sequence f (h), by that analogy, can continue to use f1H () generation is new
f2H () sequence, makes it more quickly converge on f (0), due to
Abbreviation is carried out to (6), (7), is had:
If now f " ' (0) ≠ 0, then sequence { f2(h) } to approach the order of error of f (0) be just O (h3)。
Above reasoning process illustrates that more accurate approximation can be extrapolated using approximation well, is a kind of adding
The convergent algorithm of speed, and expansion number of times is higher, and control accuracy is also higher.
Simulation result and data analysis
System building and emulation
To realize the Optimum utilization of large-sized photovoltaic array component, domestic and international modeling of many scholars to photovoltaic cell[10-12]
And maximum power point tracing method carried out it is widely studied[13-16].It is first according to corresponding theory analysis and control algolithm block diagram
The mathematic simulated mode of photovoltaic cell is built under Simulink environment first with Matlab softwares as shown in figure 3, herein, imitating
The selection of true parameter take the parameter of the 150W solar panels that the data of manufacturer's offer produce with Pu Sheng companies be according to
According to (as shown in table 1).Variable step extrapolation pursuit iterative method is respectively adopted to be carried out emulation and compares with perturbation observation method.
Table 1:150W photovoltaic cell parameters
Figure 3 above and Fig. 4 and Fig. 5 are respectively the system simulation models built to photovoltaic cell and two kinds of MPPT algorithms.For
The superiority of the algorithm is verified, the photovoltaic array that scale is 2 × 2 is built in Matlab softwares, it is nominally most
High-power is 600W.In order to study influence of the illumination variation to its power output, following simulation analysis are done:Intensity of illumination is in 0.5s
When be reduced to 600W/m2 suddenly from 1000W/m2.
Fig. 6, Fig. 7 are exported for simulation result, and from Fig. 6, Fig. 7, variable step extrapolation pursuit iterative method is regardless of in system fortune
During row beginning still when intensity of illumination is mutated, its maximal power tracing effect all has more advantage than perturbation observation method.
Fig. 8, Fig. 9 are power partial enlarged drawing, and indicating extrapolation pursuit iterative method can be after maximum power point be traced into
Stabilization is worked at maximum power point, and perturbation observation method is constantly shaken at maximum power point, causes Partial Power to lose.
To sum up, set forth herein algorithm can well improve the stability of a system, with quick tracking effect.
Finally, varying environment change is can adapt in order to verify the algorithm, the control algolithm is applied to whole photovoltaic hair
System emulation is carried out in electric system.In order to preferably verify the superiority of the algorithm, 100KW grid-connected photovoltaic systems have been built
Simulation model, as shown in Figure 10.
Simulation result and analysis
Perturbation observation method and variable step extrapolation pursuit iterative method is respectively adopted carries out emulation and compare, Figure 11 is that given illumination is strong
Degree, illumination simulation intensity gradually reduces (rising) first, until illumination reaches minimum (height) value, then simulates due to shade or day
Tracking process of the system to peak power when illumination declines suddenly (rising) caused by gas mutation.
By Figure 12 power outputs waveform it is easy to see that although the control of perturbation observation method is simple, algorithm is also easy-to-understand,
And the requirement to hardware facility is not also high, implementation process is more convenient, and it is main as follows that it is not enough:For perturbation observation method, its
Problem most rambunctious is often the setting for disturbing step-length, and step-length is long, and the tracking velocity of system quickly, but works as intensity of illumination
Maximum power point can also occur larger fluctuation when there is acute variation, even if intensity of illumination even variation, power output change
Process has also fluctuated, and still has some drawbacks urgently to be resolved hurrily.
Analogous diagram and numerical analysis table are as follows:Figure 12 is the maximum power point output waveform under perturbation observation method, right
The table 2 answered is the concrete numerical value of iteration output.
The perturbation observation method power output of table 2
It is clear that by table 2, can be acutely shaken at maximum power point (MPP) place for perturbation observation method system, meeting
Cause a large amount of losses of power, although reduce disturbance step-length, system can be caused in the concussion reduction of MPP, that so brings is unfavorable
Factor is exactly slowing for system tracking MPP, and system response slows up.When light intensity is varied widely, disturbance observation
Method can occur the erroneous judgement to system, and the direct result that erroneous judgement brings is exactly to cause whole system to occur acutely, significantly shaking, directly
To busbar voltage collapse.
Comparison diagram 12, Figure 13 power Ps/W real-time trackings simulation curve are seen it is easy to see that being disturbed for whole photovoltaic system
The method of examining needs 0.5s just completely into steady operational status, and variable step extrapolation pursuit iterative method only needs to 0.35s and is achieved that
Steady-state operation and start-up course concussion very little.Speed is significantly faster than the toggle speed of perturbation observation method, even if illuminance abrupt variation, system
Tracking to maximum power point is not also almost shaken.When intensity of illumination gradually increases (gradually decrease), power is also put down therewith
Steady increase (steady to reduce), when intensity of illumination increases suddenly (reduce suddenly), system can react rapidly, make power output
Rapid increase (reduction), comes back on the maximum power point of now intensity of illumination output.
The variable step of table 3 extrapolation pursuit iterative method power output result
Contrast table 2, the emulation of table 3 data output result is easy to get, and perturbation observation method wants to reach and extrapolation pursuit iterative method phase
Same required precision 8 iteration of needs could be realized, and pursuit Iterative number of times of extrapolating is little, it is thus only necessary to 4 iteration
Tracking accuracy can be just fully achieved and control is required.
It is analyzed by the emulation data to above two method, it can be seen that the optimal value of two methods is identical
, it ensure that the accuracy of system tracking.The iterations of apparent variable step extrapolation pursuit iterative method is considerably less than disturbance
Observation, so as to improve the operating efficiency of whole system.
There is following some advantage compared with original algorithm:
The algorithm be it is a kind of accelerate convergent algorithm, can avoid due to when instantaneous physical quantity is detected to disturbing that system is brought
It is dynamic, reduce the dependence to system hardware, so that stability when ensureing system operation, it is ensured that while tracking velocity, Ke Yiti
In high precision.
The algorithm is applied in 100KW photovoltaic generating systems, according to simulation result it is easy to see that whole photovoltaic system
Only need to 0.35s and just come into steady-state operating condition, toggle speed of the speed considerably beyond perturbation observation method.
Change with intensity of illumination has good tracking effect to maximum power point, even if illuminance abrupt variation does not almost have yet
Concussion.Demonstrating the algorithm not only has the advantages that perturbation observation method, and with fast response time, and it is excellent that iterations is few etc.
Point.Expected purpose of design is reached, has all been shown in tracking velocity or in tracking accuracy apparent superior
Property, with application and promotional value well.