CN105160395A - Efficiency optimizing inertia variation particle swarm method of resonant electric energy transmission device - Google Patents

Abstract

The invention discloses an efficiency optimizing inertia variation particle swarm method of a resonant electric energy transmission device, belonging to the field of searching methods for system transmission efficiency in a magnetic coupling wireless electric energy transmission system. The method is characterized in that the particle swarm size in a common particle swarm algorithm is separately set into a maximum particle swarm size (i) Nmax(/i)=30 and a minimum particle swarm size (i) Nmin(/i)=2, the particle swarm size is generally increased along the iteration frequency and is gradually reduced along the inertia curve, the redundancy particles are removed, the algorithm is simplified, and the later convergence speed of the algorithm is accelerated. The particle swarm optimization algorithm provides basis for particle size selection, and the convergence speed is accelerated at the later stage of searching, and the algorithm searching time is shortened.

Description

The improving efficiency inertial change particle swarm optimization of resonant mode electric energy dispensing device

Technical field

The invention belongs to magnetic coupling wireless power transmission technical field, particularly relate to the finding method field of system transfers efficiency in magnetic coupling radio energy transmission system, be specifically related to a kind of improving efficiency inertial change particle swarm optimization of resonant mode electric energy dispensing device.

Background technology

Delivery of electrical energy is the major issue that academia pays close attention to always, and non-contact power supply technique is the focus studied in recent years.Wireless power transmission mode mainly contains 3 kinds: the first is induction; The second is microwave radio formula; The third magnetic coupling is resonant, and three kinds of modes respectively have its advantage.The resonant delivery of electrical energy basic thought of magnetic coupling realizes based on magnetic coupling resonance principle, and when power supply excitation frequency reaches certain value, whole system is in resonant condition, now can realize wireless high-effect Energy Transfer.The resonant wireless power transmission mode of magnetic coupling has long transmission distance than first kind of way, the advantage that through-put power is large compared with the second way, obtained great concern in recent years, but this technology is also in the starting stage, especially to the research of its transfer efficiency related fields be always lack.

Magnetic coupling radio energy transmission system efficiency is different at different electrical power excitation frequency point place, and its efficiency-frequency curve is an one-dimensional functions.For a system, distance between dispatch coil is fixed, its transfer efficiency function there will be one or two extreme points along with the change of excitation frequency, this just makes general algorithm (hill-climbing algorithm, simulated annealing etc.) easily be absorbed in local optimum and miss global optimum.Particle cluster algorithm can provide a kind of ways of addressing this issue, have superiority although general particle cluster algorithm compares when solving general function optimization problem, but be aimed at the resonant radio energy of magnetic coupling and go out communication system, when there is the situation of an extreme point in system, algorithm there will be of short duration stagnation behavior in the search later stage, can not Fast Convergent, expend time in longer; And for algorithm itself, the conference of crossing of population scale setting causes algorithm to carry out unnecessary calculating, less scale then causes algorithm directly to miss global optimum, even can not find extreme point, general population scale is located between 20-40, but its population scale accurately choose all the time all according to individual when dealing with problems ceaselessly hit and miss experiment out, very blindly.For above situation, be badly in need of finding a kind of optimization method for the feature of magnetic coupling radio energy transmission system own to find problem with resolution system efficiency.Therefore, the feature how being directed to magnetic coupling wireless power supply system designs a kind of optimization method and system maximal efficiency and corresponding Frequency point can be found rapidly to be necessary.

Summary of the invention

The technical matters that the present invention solves there is provided a kind of improving efficiency inertial change particle swarm optimization of resonant mode electric energy dispensing device, the method is taked to reduce in the mode of similar inertial curve gradually with iterations increase to population scale, mainly solve conventional particle colony optimization algorithm in magnetic coupling radio energy transmission system in searching process, there will be the problem that the phenomenon of of short duration stagnation and the population scale of this algorithm own choose, particle cluster algorithm is restrained rapidly, finds the optimal value of system effectiveness fast.

The present invention adopts following technical scheme for solving the problems of the technologies described above, the improving efficiency inertial change particle swarm optimization of resonant mode electric energy dispensing device, it is characterized in that: by the separately setting of the population scale in general particle cluster algorithm, be respectively maximum population scale Nmax=30 and smallest particles group scale Nmin=2, the mode of population scale along with iterations increase along inertial curve reduces gradually, and its concrete implementation step is:

(1), initialization algorithm, comprise setting particle populations dimension D, maximum iteration time MaxNum, limits particle maximal rate v simultaneously _max, initialization inertia weight w;

(2), directly setting population maximum-norm Nmax be 30 and population smallest size Nmin be 2, the speed v of random initializtion particle and the position of particle, setting primary group scale is maximum-norm Nmax=30, initialization iterations t=1;

(3), fitness function is adopted calculate the fitness function value f of each particle of current population _i, f _irepresent the fitness function value of i-th particle, wherein ω=2 π f _r, f _rfor current excitations frequency, ω is the angular frequency of excitation power supply, and M is the mutual inductance between transmitting and receiving coil, L ₁, L ₂for transmitting coil and receiving coil inductance, C ₁, C ₂for electric capacity, R _sfor the internal resistance of source, R _lfor pull-up resistor, R ₁, R ₂for resistance in loop;

(4), f is used _i-bestrepresent the optimal-adaptive degree functional value that i-th particle searches when the t time iteration, use f _i-gbestrepresent when the t time iteration, the optimal-adaptive degree functional value that all particles searches, before particle cluster algorithm starts iteration, setting f _i-best=0, f _i-gbest=0, by the particle fitness function value f obtained in step (3) _iwith individual extreme value f _i-bestand global extremum f _i-gbestcompare, if f _i≤ f _i-best, so f _i-best=f _i, p _i=x _i, p _irepresent that fitness function value is f _i-bestparticle position, x _ibe be f to fitness function value _ithe position of particle, if f _i≤ f _i-gbest, so f _i-gbest=f _i, p _g=x _i, p _gthat in particle populations, global optimum is f _i-gbestparticle position;

(5), by formula Npresent=Nmax* (e ^-(t/MaxNum)) ⁿupgrade population scale, wherein Npresent is the current scale of population, Nmax is maximum population scale, MaxNum is maximum iteration time, t is current iteration number of times, n is the power exponent controlling population scale Changing Pattern, is regulated the speed degree of population scale change, by formula by parameter n $v_i^{t+1}=w*v_i^t+c_1*rand*(p_i-x_i^t)+c_2*rand*(p_g-x_i^t)$ And formula $x_i^{t+1}=x_i^t+v_i^{t+1}$ Upgrade speed and the position of each particle, then make iterations t=t+1, turn to step (6), wherein v _i ^t+1represent the speed of t+1 iteration i-th particle, v _i ^trepresent the speed of current the t time iteration i-th particle, c ₁and c ₂represent Studying factors, the random number between rand representative [01], p _irepresent that fitness function value is f _i-bestparticle position, p _gthat in particle populations, global optimum is f _i-gbestparticle position, x _i ^t+1represent t+1 iteration i-th particle position, x _i ^trepresent the t time iteration, i-th particle current location, w represents inertia weight;

(6), according to formula calculate the variance sum of particle fitness function value, f _avgfor the mean value of all particles fitness function value, if wherein there is (f _i-f _avg) >1, then a=max (f _i-f _avg), otherwise, a=1, judge variance whether equal 0 or particle cluster algorithm whether reach maximum iteration time, if not, then turn to step (3), if it is turn to step (7);

(7) the global optimum p, will searched _gas output, p _gthat in particle populations, global optimum is f _i-gbestparticle position, the frequency values that the optimal value namely searched is corresponding;

(8), load current i is detected with current sensor ₂peak value, if Δ be setting maximum current peak fluctuation range, i _2maxfor load current peak, i _2maxa k kth current cycle current peak that () is load, i _2max(k+1) be the current peak of kth+1 current cycle of load, judge | i _2max(k+1) |-| i _2max(k) | whether > Δ is set up, if judged result is yes, then turn to step (1), particle cluster algorithm is restarted, if judged result is no, particle cluster algorithm turns to step (7).

The fitness function that particle cluster algorithm of the present invention adopts changes along with the change of mutual inductance between transmitting and receiving coil, only has the mutual inductance first determined between launching and receiving coil, fitness function is made to become only relevant with excitation frequency function, then search for particle cluster algorithm, wherein fitness function is the function of efficiency and frequency.This particle swarm optimization algorithm in actual applications by excitation frequency used during the initial excitation system of detection, and then calculates the mutual inductance of two coils.The present invention takes to reduce in the mode of similar inertial curve gradually with iterations increase to population scale, mainly solve conventional particle group algorithm in magnetic coupling radio energy transmission system in searching process, there will be the problem that the phenomenon of of short duration stagnation and the population scale of this algorithm own choose, make particle cluster algorithm go convergence, find system effectiveness optimal value fast.This particle swarm optimization algorithm sets algorithm and restarts condition, and when the change of distance load current being detected, algorithm is restarted, and again searches for the frequency of maximum efficiency value and its correspondence.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of particle swarm optimization algorithm of the present invention;

Fig. 2 is the result analogous diagram of general particle swarm optimization algorithm;

Fig. 3 is the result analogous diagram of particle swarm optimization algorithm of the present invention;

Fig. 4 is that population scale of the present invention increases reduction figure with iterations.

Specific implementation method

Describe particular content of the present invention by reference to the accompanying drawings in detail.The present invention is mainly for magnetic coupling radio energy transmission system, and use improved Particle Swarm Algorithm, particle scale is reduced, and algorithm can find efficiency maximum point and its corresponding frequencies fast.Illustrate below by way of specific instantiation and emulate with Matlab.Optimization method flow process is shown in Fig. 1, and the technical solution adopted in the present invention is: the improving efficiency inertial change particle swarm optimization of resonant mode electric energy dispensing device, and concrete steps are:

(1), initialization algorithm, comprise setting particle populations dimension D=1, maximum iteration time MaxNum=200, limits particle maximal rate v simultaneously _max, initialization inertia weight w;

(2), directly setting population maximum-norm Nmax be 30 and population smallest size Nmin be 2, the speed v of random initializtion particle and the position of particle.Setting primary group scale is maximum-norm Nmax=30, and initialization iterations t=1, at present, the setting of population scale does not have unified rule, usually sets according to optimizing object and personal experience.It is Nmax=30 that this algorithm only needs directly to set population maximum-norm, can solve the various situations of resonant mode electric energy dispensing device improving efficiency.Set smallest size in algorithm, make population scale be reduced to smallest size Nmin by maximum-norm Nmax gradually with the increase of iterations, Nmin=2 in this algorithm;

(3), fitness function is adopted calculate the fitness function value f of each particle of current population _i, f _irepresent the fitness function value of i-th particle, wherein ω=2 π f _r, f _rcurrent excitations frequency, ω is the angular frequency of excitation power supply, and M is the mutual inductance between transmitting and receiving coil, L ₁, L ₂for transmitting coil and receiving coil inductance, C ₁, C ₂for electric capacity, R _sfor the internal resistance of source, R _lfor pull-up resistor, R ₁, R ₂for resistance in loop.This algorithm is first by current excitations frequency f _rand system of equations $\left\{\begin{array}{c}{\stackrel{\·}{U}}_{L_1}={\mathrm{j\ωL}}_1{\stackrel{\·}{I}}_1\\ {\stackrel{\·}{U}}_{L_1}=j\mathrm{\ω}M\left({\stackrel{\·}{I}}_1-{\stackrel{\·}{I}}_2\right)\\ j\mathrm{\ω}M{\stackrel{\·}{I}}_1={\stackrel{\·}{I}}_2{Z}_2\\ Z_2=R_2+R_L+j\left({\mathrm{\ωL}}_2-\frac{1}{{\mathrm{\ωC}}_2}\right)\\ \mathrm{\ω}=2{\mathrm{\πf}}_r\end{array}\right.$ Derive the mutual inductance M between transmitting and receiving coil, $M=\frac{({\mathrm{\ωL}}_2-\frac{1}{{\mathrm{\ωC}}_2})\[{(R_2+R_L)}^2+{({\mathrm{\ωL}}_2-\frac{1}{{\mathrm{\ωC}}_2})}^2\]}{2\mathrm{\ω}(R_2+R_L)},$ And then derive fitness function, the fitness function that this algorithm adopts is the function of efficiency and mutual inductance M.Shown system of equations can carry out analytical derivation out according to basic circuit theorem to whole system.Wherein represent coil L ₁voltage, for input current, load current, the fitness function that this algorithm adopts is the function of efficiency and mutual inductance M, so the distance change between two coils, when causing M also can change, fitness function also can change, at this moment M can be obtained according to current voltage excitation frequency and system of equations, the fitness function under further certainty annuity current distance;

(4), f is used _i-bestrepresent the optimal-adaptive degree functional value that i-th particle searches when the t time iteration, use f _i-gbestrepresent when the t time iteration, the optimal-adaptive degree functional value that all particles searches.Before algorithm starts iteration, setting f _i-best=0, f _i-gbest=0, by the particle fitness function value f obtained in step (3) _iwith individual extreme value f _i-bestand global extremum f _i-gbestcompare, if f _i≤ f _i-best, so f _i-best=f _i, p _i=x _i; p _irepresent that fitness function value is f _i-bestparticle position, x _ibe be f to fitness function value _ithe position of particle, if f _i≤ f _i-gbest, so f _i-gbest=f _i, p _g=x _i; p _gthat in particle populations, global optimum is f _i-gbestparticle position;

(5), by formula N _present=Nmax* (e ^-(t/MaxNum)) ⁿupgrade population scale, wherein Npresent is the current scale of population, and Nmax is maximum population scale, and MaxNum is maximum iteration time, t is current iteration number of times, n is the power exponent controlling population scale Changing Pattern, by parameter n adjustable population scale change speed degree, through great many of experiments, population can be made during n=2.3 to advise reduce gradually by the mode being similar to inertial curve, algorithm can Fast Convergent, reduces Riming time of algorithm, has larger advantage.By formula $v_i^{t+1}=w*v_i^t+c_1*rand*(p_i-x_i^t)+c_2*rand*(p_g-x_i^t)$ And formula $x_i^{t+1}=x_i^t+v_i^{t+1}$ Upgrade speed and the position of each particle, then make iterations t=t+1, turn to step (6), wherein v _i ^t+1represent t+1 iteration i-th particle rapidity, v _i ^trepresent current the t time iteration i-th particle rapidity, c ₁and c ₂represent this experiment of Studying factors c ₁=2, c ₂random number between=2, rand representative [01].P _irepresent that fitness function value is f _i-bestparticle position, p _gthat in particle populations, global optimum is f _i-gbestparticle position, x _i ^t+1represent t+1 iteration i-th particle position, x _i ^trepresent the t time iteration, i-th particle current location, w represents inertia weight.This algorithm makes particle cluster algorithm no longer occur of short duration stagnation behavior in the search later stage, and algorithm can Fast Convergent;

(6), according to formula calculate the variance sum of particle fitness function value, f _avgfor the mean value of all particles fitness function value, if wherein there is (f _i-f _avg) >1, then a=max (f _i-f _avg), otherwise, a=1.Judge variance whether equal 0 or algorithm whether reach maximum iteration time, if not, then turn to step (3), if it is turn to step (7);

(7) the global optimum p searched, is exported _g, p _gthat in particle populations, global optimum is f _i-gbestparticle position, the frequency values that the optimal value namely searched is corresponding.

(8), load current i is detected with current sensor ₂peak value, if Δ be setting maximum current peak fluctuation range, i _2maxfor detected load current peak, i _2maxa k kth current cycle current peak that () is load, i _2max(k+1) be the current peak of kth+1 current cycle of load.Judge | i _2max(k+1) |-| i _2max(k) | whether > Δ is set up, if judged result is yes, then turn to step (1), algorithm is restarted; If judged result is no, algorithm branches step (7).

In order to very clearly understand the advantage of this algorithm, give the analogous diagram of general particle cluster algorithm and particle swarm optimization algorithm of the present invention respectively in figs. 2 and 3, Fig. 4 is that population scale increases reduction figure with iterations.

The optimizing result figure of the general particle cluster algorithm of Fig. 2, in figure, curve is efficiency and the frequency function image of magnetic coupling radio energy transmission system, and in figure, pentagram is searched optimal value, and the Frequency point corresponding to maximum efficiency is 13544961.6816Hz.

Fig. 3 is the optimizing result figure of particle swarm optimization algorithm of the present invention, its optimum configurations is identical with general particle cluster algorithm, can find out, comparing with general algorithm, when its Search Results is identical, 4.617000 seconds its algorithm used times, and general particle cluster algorithm is 10.265 seconds, this particle swarm optimization algorithm expends time in and decreases 55% than general particle cluster algorithm, has saved the time.

Fig. 4 is the process that population scale reduces gradually along with iterations increase, and its reduction mode is inertia decline curve.

Embodiment above describes ultimate principle of the present invention, principal character and advantage; the technician of the industry should understand; the present invention is not restricted to the described embodiments; what describe in above-described embodiment and instructions just illustrates principle of the present invention; under the scope not departing from the principle of the invention; the present invention also has various changes and modifications, and these changes and improvements all fall in the scope of protection of the invention.

Claims (1)

1. the improving efficiency inertial change particle swarm optimization of resonant mode electric energy dispensing device, it is characterized in that: by the separately setting of the population scale in general particle cluster algorithm, be respectively maximum population scale Nmax=30 and smallest particles group scale Nmin=2, the mode of population scale along with iterations increase along inertial curve reduces gradually, and its concrete implementation step is:

(1), initialization algorithm, comprise setting particle populations dimension D, maximum iteration time MaxNum, limits particle maximal rate v simultaneously _max, initialization inertia weight w;

(2), directly setting population maximum-norm Nmax be 30 and population smallest size Nmin be 2, the speed v of random initializtion particle and the position of particle, setting primary group scale is maximum-norm Nmax=30, initialization iterations t=1;

(3), fitness function is adopted calculate the fitness function value f of each particle of current population _i, f _irepresent the fitness function value of i-th particle, wherein ω=2 π f _r, f _rfor current excitations frequency, ω is the angular frequency of excitation power supply, and M is the mutual inductance between transmitting and receiving coil, L ₁, L ₂for transmitting coil and receiving coil inductance, C ₁, C ₂for electric capacity, R _sfor the internal resistance of source, R _lfor pull-up resistor, R ₁, R ₂for resistance in loop;

(4), f is used _i-bestrepresent the optimal-adaptive degree functional value that i-th particle searches when the t time iteration, use f _i-gbestrepresent when the t time iteration, the optimal-adaptive degree functional value that all particles searches, before particle cluster algorithm starts iteration, setting f _i-best=0, f _i-gbest=0, by the particle fitness function value f obtained in step (3) _iwith individual extreme value f _i-bestand global extremum f _i-gbestcompare, if f _i≤ f _i-best, so f _i-best=f _i, p _i=x _i, p _irepresent that fitness function value is f _i-bestparticle position, x _ibe be f to fitness function value _ithe position of particle, if f _i≤ f _i-gbest, so f _i-gbest=f _i, p _g=x _i, p _gthat in particle populations, global optimum is f _i-gbestparticle position;

(5), by formula Npresent=Nmax* (e ^-(t/MaxNum)) ⁿupgrade population scale, wherein Npresent is the current scale of population, Nmax is maximum population scale, MaxNum is maximum iteration time, t is current iteration number of times, n is the power exponent controlling population scale Changing Pattern, is regulated the speed degree of population scale change, by formula by parameter n $v_i^{t+1}=w*v_i^t+c_1*rand*(p_i-x_i^t)+c_2*rand*(p_g-x_i^t)$ And formula $x_i^{t+1}=x_i^t+v_i^{t+1}$ Upgrade speed and the position of each particle, then make iterations t=t+1, turn to step (6), wherein represent the speed of t+1 iteration i-th particle, represent the speed of current the t time iteration i-th particle, c ₁and c ₂represent Studying factors, the random number between rand representative [01], p _irepresent that fitness function value is f _i-bestparticle position, p _gthat in particle populations, global optimum is f _i-gbestparticle position, represent t+1 iteration i-th particle position, represent the t time iteration, i-th particle current location, w represents inertia weight;

(6), according to formula calculate the variance sum of particle fitness function value, f _avgfor the mean value of all particles fitness function value, if wherein there is (f _i-f _avg) >1, then a=max (f _i-f _avg), otherwise, a=1, judge variance whether equal 0 or particle cluster algorithm whether reach maximum iteration time, if not, then turn to step (3), if it is turn to step (7);

(7) the global optimum p, will searched _gas output, p _gthat in particle populations, global optimum is f _i-gbestparticle position, the frequency values that the optimal value namely searched is corresponding;

(8), load current i is detected with current sensor ₂peak value, if Δ be setting maximum current peak fluctuation range, i _2maxfor load current peak, i _2maxa k kth current cycle current peak that () is load, i _2max(k+1) be the current peak of kth+1 current cycle of load, judge | i _2max(k+1) |-| i _2max(k) | whether > Δ is set up, if judged result is yes, then turn to step (1), particle cluster algorithm is restarted, if judged result is no, particle cluster algorithm turns to step (7).