CN102946208B - Non-dead-zone control optimizing method based on immune algorithm for inverter - Google Patents

Abstract

The invention provides a non-dead-zone control optimizing method based on an immune algorithm for an inverter. On the basis of various methods for reducing or eliminating dead-zone effects, a new three-level dead-zone eliminating method directing at a single-phase full bridge inverter is provided. Under the condition that precise electrical polarity detection is not required, a switch sequence is determined by dividing a reference current area. The non-dead-zone control optimizing method can achieve seamless transition non-dead-zone control near a current zero crossing point and among all areas. A pulse-width modulation (PWM) control sequence meeting a dead-zone eliminating method requirement is optimized by means of the immune algorithm, and output waveform quality of the inverter is improved. Compared with a conventional control strategy, the non-dead-zone optimal PWM control sequence not only can effectively remove dead-zone effects, but also can reduce total harmonic distortion (THD) of the output waveform of the inverter remarkably.

Description

Inverter based on immune algorithm is controlled optimization method without dead band

Technical field

The present invention relates to inverter digital Control Technology field, the inverter especially relating to based on immune algorithm is controlled optimization method without dead band.

Background technology

Along with the development of power electronic technology, PWM voltage source inverter (VSI) is widely used in the modern industry productions such as frequency control, intelligent grid and resident living.After adopting in theory PWM to control, the Resonance Wave Composition of inverter output waveforms should higher (concentrating on around the integral multiple of carrier frequency).But in practical application, semiconductor switch device has nonideal switching characteristic, make device service time be less than the turn-off time, easily cause two power device switch generation shoot through faults of homophase brachium pontis complementation, for avoiding this situation, conventionally by signal lag conducting or turn-off in advance dead band is set.But the existence in dead band will make output voltage and electric current generation nonlinear distortion, and low-order harmonic increase, motor torque produces pulsation and increases harmonic loss etc., especially low at voltage, affects even more serious when switching frequency is high.

In order to reduce the impact with deadband eliminating, a lot of experts, scholar have carried out large quantity research, mainly comprise several aspects: 1. dead area compensation, 2. Dead Time minimizes, 3. the deadband eliminating time.Dead-zone compensation method is mainly that dead zone error is carried out to voltage compensation, or the drive pulse signal PWM width of adjusting and the compensation of the time of carrying out makes actual service time consistent with desirable ON time.2007, the propositions such as Oliveira A.C. reduced the distortion of inverter output voltage by changing switching frequency and gate signal pulsewidth, but its amount of calculation is large.The impact of the change that the online compensation method of the propositions such as Urasaki N can be eliminated pressure drop on power tube switching time and power tube on dead area compensation effect, but its prerequisite be motor parameter must be accurately.The disturbance voltage vector that the propositions such as Kim S.Y cause to offset dead band by the compensation vector that superposes in first-harmonic command voltage vector, but hardware detecting circuit has been brought unreliability, complexity.What justice waits by calculate size of current under d-q coordinate system, propose to change independently respectively three brachium pontis Dead Time separately, object is that the disturbance voltage vector follow current vector synchronization being caused by dead band is rotated, but the minimum dead band permission time is to be also a variable by device self decision and its value that is in operation.The minimized algorithm of a kind of Dead Time to unnecessary switch gate electrode drive signals is forbidden in the propositions such as Choi J.S.When adopting hardware that electric current is detected or calculated, exist and introduce the error of A/D conversion and the hysteresis of detection, and owing to there being the clamper phenomenon of switching noise and the zero current of PWM, make to occur the phenomenon of a plurality of zero crossings in current detecting, be difficult to Obtaining Accurate current polarity, so dead area compensation and minimum Dead Time method just reduce dead time effect impact to a certain extent, but can not eliminate completely.The deadband eliminating time is to arrange on the three elements condition basis in dead band in analysis, propose not arrange the control strategy in dead band, its principle is: electric current is in positive-negative half-cycle, in fact same brachium pontis only has a power device and parallel diode On current, so can forbid the not power device driving signal of On current of reality, allow it always in off state, between upper and lower switch, dead band need not be set again, realization is controlled without dead band, but the method needs state-detection situation and the current polarity of each switch of Obtaining Accurate.The propositions such as L.H.Chen have been introduced current detecting hardware circuit by judging power tube parallel diode whether conducting detecting the sense of current, and after Lin Y.K, having proposed does not need to provide independent independent current source for testing circuit, but these two kinds of methods are introduced and are detected loop, also will produce the shortcoming of the aspects such as economical and reliability simultaneously.In addition this dead band removing method be not suitable for the situation of a plurality of current zero-crossing points.To there is fast conversion frequently in its sense of current when electric current is near zero point, and the sampling time delay of digital signal or the such disappearance that drives signal that more easily causes of computation delay, so the reliably impact of deadband eliminating effectively.Yang Bo etc. propose to adopt conventional combination of controlling and crossing two kinds of methods of null range dead area compensation without dead band, effectively deadband eliminating impact of the method, but the method that above-mentioned Dead Time is controlled is all based on two level PWM control strategies.

Summary of the invention

The present invention solves the existing technical problem of prior art; A kind of effective problem that has solved the zero crossing of eliminating Inverter Dead-time impact is provided, and has not needed the inverter based on immune algorithm of point-device Current polarity detection device to control optimization method without dead band.

It is to solve the existing technical problem of prior art that the present invention also has an object; Provide a kind of and can make the transition of inverter switch device on off state more level and smooth, realized the inverter based on immune algorithm of controlling without dead band of seamless transitions between each region and controlled optimization method without dead band.

It is to solve the existing technical problem of prior art that the present invention has an object again; Provide a kind of obvious minimizing inverter output waveforms total percent harmonic distortion, improved the inverter based on immune algorithm of output waveform quality and controlled optimization method without dead band.

It is to solve the existing technical problem of prior art that the present invention finally has an object; Provide a kind of and can reduce to greatest extent switching loss, and the inverter based on immune algorithm that can improve the utilance of device and extend life cycle due to switch motion number of times mean allocation is controlled optimization method without dead band.

Above-mentioned technical problem of the present invention is mainly solved by following technical proposals:

Inverter based on immune algorithm is controlled an optimization method without dead band, it is characterized in that.

Step 1, sets iterations k, and produces initial population according to encoding operation, and definition antibody is the set of number string that single-phase inverter switching mode is corresponding, and its chromosome structure can be defined as follows:

Chrom=X ₁,X ₂,X ₃,…X _N

X wherein _m(m=1,2,3 ... N) represent the on off sequence that each switch is corresponding, X _m0,1, value in 2 three values, wherein, initial population be meet encoding operation and without the constraints situation of Dead Time under random generation;

Step 2, vaccinate, is about to add in the initial antibodies population of step 1 generation as the antibody of therapeutic vaccine, with the chromosome of random generation, obtains initial antibody population;

Step 3, calculates affinity to every item chromosome in current antibody population, and each chromosomal concentration in current population;

Step 4, calculates Immune Selection functional value according to step 3 gained affinity and concentration to each antibody in current antibody population, and the antibody that Immune Selection functional value is low is removed from current antibody population by probability;

Step 5, the antibody population completing for step 4 is carried out interlace operation, to set probability, in certain two individual crosspoints, exchange mutually occurs;

Step 6, completes antibody population for step 5 and carries out mutation operation, the sudden change that a certain position on antagonist or the genic value on some locus carry out according to the variation probability of setting;

Step 7, judges whether current iteration number of times k reaches the maximum iteration time of setting, is to enter step 8, otherwise establishes current iteration number of times k=k+1, turns back to step 3 and carries out next iteration;

Step 8, each antibody in current antibody population is calculated to affinity, judge whether current acquired results restrains, and is to enter step 9, otherwise using current antibody population as the antibody of therapeutic vaccine, turn back to step 1 and regenerate and add initial antibodies population to carry out iteration;

Step 9, each antibody in current antibody population is calculated to Immune Selection functional value, according to the be eliminated on off state control sequence of 0 ~ pi/2 part in whole cycle of single-phase inverter of dead band impact or 0 ~ π part of the antibody of Immune Selection functional value maximum.

At the above-mentioned inverter based on immune algorithm, without dead band, control optimization method, in described step 1, initial population be meet encoding operation and without the constraints situation of Dead Time under random generation, according to galvanic areas, divide, its constraints is, at electric current, decline to cross null range chromogene coding can only be sequence " 2 ", Current rise to cross null range chromogene coding can only be sequence " 1 ", in transitional region, chromogene coding can only be sequence " 0 ".

At the above-mentioned inverter based on immune algorithm, without dead band, control optimization method, in described step 3, affinity is because needs consider THD and switching loss, the obtaining based on formula of affinity:

$\mathrm{Fitness}\left(x_i\right)=\frac{1}{\underset{m=0}{\overset{N-1}{\mathrm{\Σ}}}{\∫}_{\mathrm{m\Δt}}^{(m+1)\mathrm{\Δt}}|I_m\left(t\right)-I_f\left(t\right)|\mathrm{dt}}$

In formula, xi represents i bar antibody, I _m(t) represent actual current value, use I _m(t) equivalent m Δ t is to (m+1) Δ t actual current mean value during this period of time, I _f(t) represent desired electrical flow valuve.

At the above-mentioned inverter based on immune algorithm, without dead band, control optimization method, in described step 3, the acquisition methods of chromosomal concentration is: definition i bar antibody x _iwith j bar antibody x _jfor two vectors in space X, they are mapped to and in space Y, are called vector f (x by function f _i), f (x _j), so vector f (x _i), f (x _j) distance in Y space is:

$E(x_i,x_j)=\underset{0}{\overset{\∞}{\∫}}|f\left(x_i\right)-f\left(x_j\right)|\mathrm{dt}$

Antibody x _iwith the space length summation of other antibody (comprise) itself, with the space length summation of current all antibody of antibody population be:

$E\left(x_i\right)=\underset{j=1}{\overset{H}{\mathrm{\Σ}}}E(x_i,x_j)$

In formula, H is antibody sum in current antibody population; Therefore AC is defined as:

$\mathrm{Density}\left(x_i\right)=\frac{1}{E\left(x_i\right)}.$

At the above-mentioned inverter based on immune algorithm, without dead band, control optimization method, in step 4, Immune Selection function be made as:

$P\left(x_i\right)=a*\frac{\mathrm{Fitness}\left(x_i\right)}{\underset{j=1}{\overset{H}{\mathrm{\Σ}}}\mathrm{Fitness}\left(x_j\right)}+\frac{b}{\mathrm{Density}\left(x_i\right)};$

In formula, a, b are weight coefficient, and a+b=1, obtain the choice function value of all antibody, setting threshold is removed by probability the low antibody of Immune Selection functional value from current antibody population, that is: for i antibody, produce at random the random number α of a span between [0,1] _iif, i antibody will be genetic to the next generation, if i antibody, by abandoned, is not genetic to the next generation.

At the above-mentioned inverter based on immune algorithm, without dead band, control optimization method, in described step 5, the concrete operations of interlace operation are: if judge, the random probability producing of two certain gene position of chromosome is greater than crossing-over rate, and, from this gene position, two chromosomal genes exchange.

At the above-mentioned inverter based on immune algorithm, without dead band, control optimization method, in described step 6, the concrete operations of mutation operation are: if judge, the random probability producing of certain gene position of chromosome is greater than aberration rate, if former sequence is " 1 ", random mutation is " 0 " or " 2 "; Former sequence is " 0 ", and random mutation is " 1 " or " 2 "; Former sequence is " 2 ", and random mutation is " 1 " or " 0 ".

Therefore, tool of the present invention has the following advantages: 1. the present invention adopts the method for determining on off sequence by dividing reference current region, has effectively solved the problem of the zero crossing of eliminating Inverter Dead-time impact, and has not needed point-device Current polarity detection device; 2. the present invention adopts three level control strategies, makes the transition of inverter switch device on off state more level and smooth, has realized controlling without dead band of seamless transitions between each region; 3. inverter proposed by the invention is optimized switching waveform without dead zone control method by immune algorithm, obviously reduces the total percent harmonic distortion of inverter output waveforms, has improved output waveform quality; 4. because inverter proposed by the invention is only worked a period of time without dead band control optimization method each switch in one-period, other times are failure to actuate, reduce to greatest extent switching loss, and because switch motion number of times mean allocation can improve the utilance of device and extend life cycle.

Accompanying drawing explanation

Accompanying drawing 1 is single-phase inverter main circuit schematic diagram of the prior art.

Accompanying drawing 2 is brachium pontis resolving cell structural representations of equivalence of the prior art.

Accompanying drawing 3 is reference current area distribution schematic diagrames of the embodiment of the present invention.

Accompanying drawing 4 is the conventional two level PWM control signals of tradition, the conventional pwm control signal schematic diagram without dead band pwm control signal and the embodiment of the present invention.

Accompanying drawing 5 is immune algorithm flow charts of the prior art.

Accompanying drawing 6 is interlace operation schematic diagrames of the embodiment of the present invention.

Accompanying drawing 7 is mutation operation schematic diagrames of the embodiment of the present invention.

Accompanying drawing 8 is inversion operation charts of the embodiment of the present invention.

Embodiment

Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.

Embodiment:

Technical solution of the present invention can adopt computer software technology to realize automatic operational process, below in conjunction with drawings and Examples, describes technical solution of the present invention in detail.

Accompanying drawing 1 is voltage-type single-phase inverter concrete structure, and in figure, E is DC voltage, parallel voltage-stabilizing capacitor C ₁, C ₂.L is pure inductance, and R is resistance, represents that inverter output end connects resistance sense load when R ≠ 0, when R=0, represents that inverter output end connects pure inductance load.Switching device IGBT(insulated gate bipolar transistor) V ₁～V ₄being unipolarity two-valued function switch, during conducting 1, is 0, VD during shutoff ₁～VD ₄for inverse parallel fly-wheel diode.U ₀represent inverter output voltage, i represents inverter output current, and its arrow direction is reference current positive direction.

As shown in Figure 1, adopt three level mode deadband eliminating herein.As switch V ₁, V ₄conducting and V ₂, V ₃turn-off (or VD ₁, VD ₄conducting), inverter output voltage is U ₀=E; Work as V ₂, V ₃conducting and V ₁, V ₄(or VD during shutoff ₂, VD ₃conducting), output voltage is U ₀=-E; Work as V ₁, VD ₃conducting (or V ₂, VD ₄conducting) output voltage is U ₀=0.Therefore by the on off sequence X of inverter _mbe designated as:

$X_m=\left\{\begin{array}{cc}0& U_0=0\\ 1& U_0=E\\ 2& U_0=-E\end{array}\right.$

Attached Figure 2 shows that is decomposed into two unit that are mutually related by a phase brachium pontis.As i>0 and V ₁during conducting, current i is through V ₁flow through; V ₁during shutoff, current i is through fly-wheel diode VD ₂afterflow; So V under both of these case ₂all without conducting.In like manner, as i<0 and V ₂during conducting, current i is through V ₂flow through; V ₂during shutoff, current i is through VD ₁afterflow; V under both of these case ₁all without conducting.When flowing through the sense of current one timing of a brachium pontis of inverter, only need to control the on off state of an IGBT in this brachium pontis and just can control the output voltage of whole brachium pontis, this is for providing foundation without dead band.

Without dead band, control the problem that research mainly solves: detect accurately current polarity; Effectively process current zero-crossing point; Because load current likely occurs a plurality of zero crossings with respect to standard sine electric current, so solve zero crossing problem, be also the Important Problems that needs solution.The single-phase inverter proposing herein, without dead zone control method, comprises the proposition of sequence area, 1.Wu dead band, two aspects; Zero point near zone processing method.

The present invention is divided into several districts by the reference current waveform of one-period, and as shown in Figure 3, a, d are respectively zero passage district, and b, c, e, f are conventional district, it should be noted that from b to c region and e to f region, respectively have a region transition point.Table 1 is depicted as the present invention without the on off sequence table of dead band control strategy, the present invention without dead band control strategy under, when single-phase inverter is normally worked, the on off sequence X that may occur in different reference currents region _m, and corresponding break-over of device state and output voltage U ₀.

Table 1. on off sequence table

In accompanying drawing 4, signal 6～9 is depicted as three level without dead band control signal, according to galvanic areas, determines corresponding on off state.Signal 1,2 represents respectively brachium pontis V on conventional two level single-phase inverters ₁and V ₄, lower brachium pontis V ₂and V ₃switch controlling signal; Signal 4,5 represents respectively the conventional upper brachium pontis V without dead band ₁and V ₄, lower brachium pontis V ₂and V ₃switch controlling signal; Signal 6,7,8,9 represents that respectively three level proposed by the invention are without the brachium pontis V in dead band ₁, V ₂, V ₃, V ₄control signal.From control signal 4,5, can find out, routine is simple consideration upper and lower bridge arm switch self-characteristic without dead band switch control law, does not consider same corresponding switch, as above brachium pontis V ₁and V ₄switch control law also can disunity, do not need to cut-off simultaneously.Three-level pwm control method takes full advantage of the afterflow effect of fly-wheel diode, considers the synergy of same phase brachium pontis switch simultaneously, and table 1 has shown that with accompanying drawing 4 it specifically coordinates control procedure.Be described as follows:

1. conventional regional sequence

As when the large i>0 of inverter output current, likely there is on off sequence X _m=1 (U ₀=E), X _m=2 (U ₀=-E) or X _m=0 (U ₀=0).But when electric current is during at ascent stage (b region), be deadband eliminating impact, the switch control sequence of taking herein only allows to occur X _m=1 or X _m=0, can not there is X _m=2; Similarity analysis is (c region) when electric current declines, and also can not occur X _m=1, if there is X _m=2 states are also that it corresponds to fly-wheel diode VD ₂with VD ₄cooperatively interact, rather than conventional switch V ₂with V ₃action.In addition, X in the middle of switch control sequence _m=0 in zones of different corresponding on off state be all different, for example, in c region, its X _mthe=0th, V ₄with VD ₂conducting, and in b region, its X _mthe=0th, V ₁with VD ₃conducting.Take this measure to have following advantage:

(1) can cut-off situation by mean allocation switch, so both can less switch motion number of times, can extend the life-span of switch, and reduce switching loss simultaneously;

(2) can make switch seamlessly transit, as the transition of electric current from " b region-transition point-c region-d region-e region-transition point-f region-a region ", can pass through following process implementation: ' 1 ' (V ₁, V ₄)-' 0 ' (V ₁, VD ₃)-' 2 ' (VD ₂, VD ₃)-' 2 ' (V ₂, V ₃)-' 0 ' (V ₂, VD ₄)-' 1 ' (VD ₁, VD ₄)-' 1 ' (V ₁, V ₄).Can find out, switch transition is all indirectly to pass through fly-wheel diode seamless transitions;

(3) current fluctuation is less is that output waveform quality is more excellent.

2. transitional region sequence

As shown in Figure 4, when electric current is during by region b transition c, with reference to table 1, the sequence that now on off state occurs may be only ' 1 ' or ' 0 ', and corresponding current trend is respectively increase and remain unchanged than the current value of Last status.If but when transition point occurs ' 1 ', current value is in ascent stage, electric current now will enter c region, according to shown in table 1, the corresponding swtich train in c region can only be ' 2 ' or ' 0 '.According to the symmetry of electric current, its on off sequence is ' 2 ' again, and corresponding current is downward trend.When electric current relatively large to its current error of decline stage from the maximum rising.But when transition point is ' 0 ', value when electric current keeps Last status, electric current will enter c region, and its on off sequence is ' 0 ', and its current value still keeps, and with respect to the error between rise and fall, the error that transition point is ' 0 ' is relatively little.During therefore, for transition point between region, its on off sequence is ' 0 '.

3. zero passage regional sequence

By reference to current value size zoning, (this paper is with current i < ± 5%I in the present invention _maxfor crossing the critical value of null range, wherein I _maxfor current maxima), during for zero crossing, take dead area compensation scheme.

Take that to cross null range d be example, routine cross null range likely occurrence sequence ' ... 22 ... ', if there is ' 22 ' time, for routine, without dead band control strategy, its power device and diode state are respectively: V ₃with V ₄all turn-offing, is now by diode VD ₂with VD ₃afterflow, still, the anti-phase electromotive force producing due to the very little likely inductance of electric current is now not enough to reach the threshold voltage (V of diode _f):

U _o<E+2V _F

Now electric current will dam, and current waveform also will distort.In order to prevent similar these situations, the present invention proposes to utilize dead-time compensation scheme, and ' 2 ' on off state will change, namely by original diode VD ₂with VD ₃conducting is converted to V ₂with V ₃conducting, and do not allow to occur that at the on off state in early stage changing ' 0 ' (is VD ₂with V ₄conducting), can avoid V like this ₃with V ₄action simultaneously.The method is equivalent to will be to switch V when at current over-zero ₂, V ₃enough times, its ON time is to be determined by switching device and circuit self thereof.When diode is not enough to, in the situation of conducting, electric current also will be through V ₂with V ₃form loop, making current value fast-descending is 0, and magnitude of voltage is still constant, thereby output waveform can not distort.By zero passage regional sequence is converted, cross null range a, electric current is when ascent stage, and on off sequence may be only " 1 ", crosses null range d, and electric current is when the decline stage, and on off sequence may be only " 2 ".Inverter switching device sequence respective devices on off state is as shown in table 1, and electric current is being all seamless transitions from a-b-c-d-e-f-a, and power device switch shoot through phenomenon can not occur.Meanwhile, owing to crossing the delimitation of null range, that current polarity device is not needed is too accurate.

The present invention is by dividing different regions to reference current, consider a series of problems relevant to dead band such as zero crossing, transition point and switch utilance, and calculate and optimize three-level pwm sequence by immune algorithm, realize that switch seamlessly transits, effectively deadband eliminating impact, reduce output waveform THD.

Immune algorithm (Immune Algorithm, IA) is simulation living nature natural selection and Genetic Mechanisms and a kind of self adaptation global optimizing algorithm of developing.It is that state vector in state space is mapped to hereditary space through certain coding, forms genotype data sequence, and some data sequences form colony, target function value are converted to fitness function simultaneously, is used for the quality of evaluating data string.By current colony being applied to the operations such as selection, cross and variation, produce colony of new generation.Through too much generation evolution, it can reach locally optimal solution with probability 1 in theory.Wherein to have a key character be that it can process linearity and nonlinear problem to IA, and this is also the foundation that it can be optimized voltage source inverter.As previously mentioned, the method for deadband eliminating time is with traditional gate-drive control signal different.In the present invention, immune algorithm (IA) is used to optimize the PWM control sequence of eliminating switching dead time.Accompanying drawing 5 is the flow chart of immune algorithm, comprise according to encoding operation and produce initial population, vaccinate, then iteration is carried out and is calculated affinity, calculating concentration, Immune Selection, interlace operation, mutation operation, inversion operation, adds operation, judges whether individuality of new generation meets the demands, until finishing iteration after meeting the demands.

For the sake of ease of implementation, the idiographic flow of the present embodiment is described as follows:

Step 1, produces initial population according to encoding operation, and coding is the first step of IA solving practical problems.Antibody refers to the set of number string that single-phase inverter switching mode is corresponding, and its chromosome structure can be defined as follows:

Chrom=X ₀,X ₁,X ₂,…X _N-1

X wherein _m(m=0,1,2 ... N-1) represent the on off sequence that each switch is corresponding, X _m0,1, value in 2 three values.

Initial population be meet encoding operation and without the constraints situation of Dead Time under random generation.

Step 2, vaccinate, is about to add initial antibodies population as the antibody of therapeutic vaccine, with the chromosome of random generation, obtains initial antibody population.

The conventional on off sequence obtaining without strategies such as dead band PWM control or conventional three level controls is all widely used preferably inverter control sequence, perform step first at 2 o'clock and be injected in initial antibodies population using these conventional control sequences as therapeutic vaccine, as a part for initial antibodies population.Chromosome number H=500 in the initial antibodies population of the embodiment of the present invention, while specifically implementing, those skilled in the art can set initial antibodies population scale voluntarily.By vaccinate, can reduce the convergence time of immune algorithm and and improve the efficiency of immune algorithm.

Step 3, calculates affinity, comprises every item chromosome in current antibody population is calculated to affinity.

In the present embodiment, the object of immune algorithm optimization is on the basis of the impact of deadband eliminating, to reduce the THD of output waveform.Owing to directly quantizing using output harmonic wave content as target, calculating is more difficult, but the fitting degree of inverter output waveforms and standard sine wave has determined the size of its THD, therefore can be using the error of the two as the quality of calculating output waveform.Due to the symmetry of current waveform, it is specifically determined by load, in order to improve convergence and the efficiency of immune algorithm, can finally expand to a complete complete period calculating section cycle.By 1/4 (pure inductance load) or the load of 1/2(resistance sense in whole cycle of inverter output current wave) be on average divided into N equal portions (cycle T=0.02s), each short time is:

$\mathrm{\&Delta;t}=\frac{0.02}{4N}$ Or $\mathrm{\&Delta;t}=\frac{0.02}{2N}$

The expression formula of load current is:

$I_m\left(t\right)=I_{m-1}\left(\mathrm{m\&Delta;t}\right)+\frac{k_m*E-I_{m-1}\left(\mathrm{m\&Delta;t}\right)*R}{L}*(t-\mathrm{m\&Delta;t})$

Wherein:

$K_m=\left\{\begin{array}{cc}1& X_m=1\\ -1& X_m=2\\ 0& X_m=0\end{array}\right.$

N represents the gene number on each antibody, and m represents gene numbering (m=0,1,2 on antibody ... N-1), X _mrepresent switching mode corresponding in this gene position, R, L represent respectively load resistance, inductance.

By load current expression formula, the every discrete electrical flow valuve of Δ t after the time of crossing can be calculated, the more corresponding affinity value of all antibody can be calculated by the formula of calculating affinity.The affinity value of antibody is larger, illustrates that this antibody is better, and the affinity value of antibody is less, illustrates that this antibody is poorer.The formula that embodiment calculates affinity is:

$\mathrm{Fitness}\left(x_i\right)=\frac{1}{\underset{m=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{\&Integral;}_{\mathrm{m\&Delta;t}}^{(m+1)\mathrm{\&Delta;t}}|I_m\left(t\right)-I_f\left(t\right)|\mathrm{dt}}$

In formula, x _irepresent i bar antibody, I _m(t) represent actual current value, use I _m(t) equivalent m Δ t is to (m+1) Δ t actual current mean value during this period of time, I _f(t) represent desired electrical flow valuve.

Step 4, calculating concentration, comprises the calculating concentration in the following manner of every item chromosome in current population:

Making space X is the set of current all antibody of antibody population, and space Y is the set of corresponding result after each antibody effect in current antibody population, and in the present embodiment, i.e. the actual current value of inductance is flow through in representative, establishes i bar antibody x _iwith j bar antibody x _jfor two vectors in space X, they are mapped to and in space Y, are called vector f (x by function f _i), f (x _j), so vector f (x _i), f (x _j) distance in Y space is

$E(x_i,x_j)=\underset{0}{\overset{\&infin;}{\&Integral;}}|f\left(x_i\right)-f\left(x_j\right)|\mathrm{dt}$

Antibody x _iwith the space length summation of other antibody (comprise) itself, with the space length summation of current all antibody of antibody population be:

$E\left(x_i\right)=\underset{j=1}{\overset{H}{\mathrm{\&Sigma;}}}E(x_i,x_j)$

In formula, H is antibody sum in current antibody population.

Therefore embodiment is defined as AC:

$\mathrm{Density}\left(x_i\right)=\frac{1}{E\left(x_i\right)}$

Step 5, Immune Selection, comprises according to step 3 gained affinity and step 4 gained concentration each antibody in current antibody population is calculated to Immune Selection functional value, and the antibody that Immune Selection functional value is low is removed from current antibody population by probability.

Embodiment is made as Immune Selection function:

$P\left(x_i\right)=a*\frac{\mathrm{Fitness}\left(x_i\right)}{\underset{j=1}{\overset{H}{\mathrm{\&Sigma;}}}\mathrm{Fitness}\left(x_j\right)}+\frac{b}{\mathrm{Density}\left(x_i\right)}$

In formula, a, b are weight coefficient, and a+b=1, and while specifically implementing, those skilled in the art can set value according to the concrete consideration to affinity and concentration affects.In this embodiment, a=b=0.5.

By above formula, can be calculated the choice function value of all antibody, those skilled in the art as the case may be design threshold remove by probability the low antibody of Immune Selection functional value from current antibody population.The mode that embodiment adopts is: for i antibody, produce at random the random number α of a span between [0,1] _iif, i antibody will be genetic to the next generation, if i antibody, by abandoned, is not genetic to the next generation.It should be noted that the operations such as outcast antibody will can not intersect yet again, variation, inversion, be directly eliminated.

Step 6, interlace operation, interlace operation is in certain two individual crosspoints, the process of exchange mutually, the antibody that can be produced new antibody and likely be improved by interlace operation to occur with a certain probability.If judge, the random probability producing of two certain gene position of chromosome is greater than crossing-over rate, and, from this gene position, two chromosomal genes exchange.The attached interlace operation schematic diagram that Figure 6 shows that.Wherein, Chorm A, Chorm B represent respectively chromosome A, B on off sequence.

Step 7, mutation operation, variation computing is the sudden change that a certain position on antagonist or the genic value on some locus carry out according to definitive variation probability, such sudden change can further improve population diversity.If judge, the random probability producing of certain gene position of chromosome is greater than aberration rate, if former sequence is " 1 ", sports " 0 " or " 2 "; Former sequence is " 0 ", sports " 1 " or " 2 "; Former sequence is " 2 ", sports " 1 " or " 0 ".Accompanying drawing 7 is mutation operation schematic diagram.Wherein, ChormA represents chromosome A on off sequence.

Step 8, inversion operation, general load major part is formed by combining by inductance or electric capacity, and these devices all have energy storage effect, so it is not only relevant with current state, and need to consider previous state, inversion is an additional operations as chromosomal variation, be first to choose at random individual a certain position gene, by inversion probability, by itself and a rear mutual reversed position of gene, generate a new chromosome.By inversion, operate and can produce new chromosome, the attached inversion operation chart that Figure 8 shows that.Wherein, Chorm A represents chromosome A on off sequence.

Step 9, judges whether current iteration number of times k reaches the maximum iteration time of setting, is to enter step 10, otherwise establishes current iteration number of times k=k+1, turns back to step 3 and carries out next iteration.

Flow chart from immune algorithm, for a population, to first carry out Immune Selection, by the antibody carrying over after Immune Selection, successively carry out interlace operation, mutation operation, inversion operation (comprising general inversion operation and micro-inversion operates), whether the random number whether intersect, variation or inversion depending on generation is less than crossing-over rate, aberration rate, inversion rate again.If above operating procedure completes, show that IA has completed interative computation one time, as Fig. 5, iteration completes each time, turns back to step 3 and carries out next iteration, and when iterations reaches the maximum iteration time of setting, immune algorithm stops.In the present embodiment, maximum iteration time can be made as 2000 times, and while specifically implementing, those skilled in the art can according to circumstances set voluntarily.

Step 10, each antibody in current antibody population is calculated to affinity, judge whether current acquired results restrains, and is to enter step 11, otherwise using current antibody population as the antibody of therapeutic vaccine, turn back to step 1 and regenerate and add initial antibodies population to carry out iteration.

This step affinity account form with reference to step 3 affinity computing formula.If after 2000 iteration, perform step stably value of convergence of 10 gained affinity, i.e. current acquired results convergence at every turn.If result restrains, show that the algorithm that IA solves inverter optimized switching control sequence problem completes, if result does not restrain, need using this result as new vaccine, return to step 2 and repeat the process that immunity is calculated, until result convergence.Generally, if result does not restrain after carrying out first immune algorithm, then calculate by primary immune response, result can restrain.

Step 11, each antibody in current antibody population is calculated to Immune Selection functional value, according to the be eliminated on off state control sequence of 0 ~ pi/2 part (pure inductance load) in whole cycle of single-phase inverter of dead band impact or 0 ~ π part (resistance sense load) of the antibody of Immune Selection functional value maximum.

After immune algorithm completes, obtain final antibody population, by the calculating of the Immune Selection function to each antibody, can select antibody optimum in this population, this antibody is exactly the optimal solution of solving practical problems.Now the calculating of Immune Selection function is consistent with step 5, the affinity relating in calculating and concentration are calculated also consistent with step 3, step 4, because step 10 has been calculated affinity, when calculating Immune Selection function, this step can directly utilize the result of calculation of step 10.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various modifications or supplement or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.

Claims (1)

1. the inverter based on immune algorithm is controlled an optimization method without dead band, it is characterized in that,

Step 1, sets iterations k, and produces initial population according to encoding operation, and definition antibody is the set of number string that single-phase inverter switching mode is corresponding, and its chromosome structure is defined as follows:

Chrom＝X ₁,X ₂,X ₃,…X _N

X wherein _m(m=1,2,3 ... N) represent the on off sequence that each switch is corresponding, X _m0,1, value in 2 three values, wherein, initial population be meet encoding operation and without the constraints situation of Dead Time under random generation;

Step 2, vaccinate, is about to add in the initial antibodies population of step 1 generation as the antibody of therapeutic vaccine, with the chromosome of random generation, obtains initial antibody population;

Step 3, calculates affinity to every item chromosome in current antibody population, and each chromosomal concentration in current population;

Step 4, calculates Immune Selection functional value according to step 3 gained affinity and concentration to each antibody in current antibody population, and the antibody that Immune Selection functional value is low is removed from current antibody population by probability;

Step 5, the antibody population completing for step 4 is carried out interlace operation, to set probability, in certain two individual crosspoints, exchange mutually occurs;

Step 6, completes antibody population for step 5 and carries out mutation operation, the sudden change that a certain position on antagonist or the genic value on some locus carry out according to the variation probability of setting;

Step 7, judges whether current iteration number of times k reaches the maximum iteration time of setting, is to enter step 8, otherwise establishes current iteration number of times k=k+1, turns back to step 3 and carries out next iteration;

Step 8, each antibody in current antibody population is calculated to affinity, judge whether current acquired results restrains, and is to enter step 9, otherwise using current antibody population as the antibody of therapeutic vaccine, turn back to step 1 and regenerate and add initial antibodies population to carry out iteration;

Step 9, each antibody in current antibody population is calculated to Immune Selection functional value, according to the be eliminated on off state control sequence of 0～pi/2 part in whole cycle of single-phase inverter of dead band impact or 0～π part of the antibody of Immune Selection functional value maximum;

In step 1, initial population be meet encoding operation and without the constraints situation of Dead Time under random generation, according to galvanic areas, divide, its constraints is, the coding of null range chromogene excessively declining at electric current can only be sequence 2, the null range chromogene coding of crossing at Current rise can only be for sequence 1, and in transitional region, chromogene coding can only be sequence 0;

In described step 3, affinity is because needs consider THD and switching loss, the obtaining based on formula of affinity:

In formula, x _irepresent i bar antibody, I _m(t) represent actual current value, use I _m(t) equivalent m Δ t is to (m+1) Δ t actual current mean value during this period of time, I _f(t) represent desired electrical flow valuve;

In described step 3, the acquisition methods of chromosomal concentration is: definition i bar antibody x _iwith j bar antibody x _jfor two vectors in space X, they are mapped to and in space Y, are called vector f (x by function f _i), f (x _j), so vector f (x _i), f (x _j) distance in Y space is:

Antibody x _iwith the space length summation of other antibody (comprise) itself, with the space length summation of current all antibody of antibody population be:

In formula, H is antibody sum in current antibody population; Therefore AC is defined as:

In step 4, Immune Selection function is made as:

In formula, a, b are weight coefficient, and a+b=1, obtain the choice function value of all antibody, setting threshold is removed by probability the low antibody of Immune Selection functional value from current antibody population, that is: for i antibody, produce at random the random number α of a span between [0,1] _iif, i antibody will be genetic to the next generation, if i antibody, by abandoned, is not genetic to the next generation;

In described step 5, the concrete operations of interlace operation are: if judge, the random probability producing of two certain gene position of chromosome is greater than crossing-over rate, and, from this gene position, two chromosomal genes exchange;

In described step 6, the concrete operations of mutation operation are: if judge, the random probability producing of certain gene position of chromosome is greater than aberration rate, if former sequence is 1, random mutation is 0 or 2; Former sequence is 0, and random mutation is 1 or 2; Former sequence is 2, and random mutation is 1 or 0.