CN103346676A - Control system of high frequency switching power supply for Cz silicon single crystal furnace and control method - Google Patents

Abstract

The invention discloses a control system of a high frequency switching power supply for a Cz silicon single crystal furnace and a control method. The control method includes the steps of combining self-adaptive fuzzy control and sliding mode variable structure control, and aiming at a high-frequency switch heating electric power for the Cz silicon single crystal furnace to design a power controller based on a self-adaptive fuzzy control and sliding mode variable structure control method. The control system of the high frequency switching power supply for the Cz silicon single crystal furnace and the control method play respective advantages of fuzzy control and sliding mode variable structure control, further improve dynamic performance of the system, have good robustness for input voltage or load disturbance, and relieve or avoid vibration which generally happens to a sliding mode.

Description

Cz monocrystal stove control system and the control method of high frequency switch power

Technical field

The invention belongs to the Cz monocrystal stove and use high frequency switch power control technology field, the control system that relates to a kind of high frequency switch power, be specifically related to a kind of Cz monocrystal stove with the control system of high frequency switch power, the invention still further relates to the method that adopts above-mentioned control system to control.

Background technology

Industry monocrystalline silicon (no matter being solar level or IC level) is to adopt Cz(Czochralski more than 90%) the method growth, namely become silicon liquid to become monocrystalline silicon by seed crystal guiding pulling growth in the temperature environment about 1420 ℃ unmelted polycrystalline silicon under vacuum environment, these monocrystalline silicon are widely used in solar energy and the most basic material of integrated circuit industry conduct.Heating power supply is heart and the power of straight pulling silicon single crystal furnace, and its control precision, reliability, conversion efficiency etc. are that the monocrystalline silicon industry is enhanced productivity and the basis of benefit.The switch heating power supply that with IGBT is power device utilizes the IGBT power semiconductor as switching tube, utilizes the break-make of high-frequency pulsed width modulation (PWM) technology or high-frequency impulse frequency modulating technology (PFM) control electric current to form high-frequency pulse current; Under the help of inductance (high frequency transformer), by the stable low-voltage, high-current of diode full-wave rectification output.Such power supply has advantages such as conversion efficiency height, power factor height, volume are little, good stability.At present, silicon monocrystal growth just develops towards high-purity, high integrality, high uniformity and major diameter direction, this control precision to Switching Power Supply has proposed requirements at the higher level, need improve the interference rejection ability of system by advanced person's control algolithm, be that heater provides electric energy to obtain high-precision DC power supply.

Conventional PID control method needs the long adjusting time when load changing and input voltage disturbance, dynamic characteristic is relatively poor, causes the output power of power supply fluctuation, influences the monocrystalline quality; Simultaneously, because the non-linear characteristics of high frequency switch power self adopt traditional PID to control the part that also comes with some shortcomings, for example when being operated in the bigger working point of duty ratio, apparent in view hyperharmonic vibration can appear in output voltage, is unfavorable for the steady operation of Switching Power Supply.

Sliding mode variable structure control method at first makes the state trajectory of system move to the diverter surface of suitably choosing by control, make the state path along the diverter surface progressive motion to balance point then, in a single day system enters the sliding mode motion, under certain condition, just interference and parameter perturbation have consistency to external world.Do not need precise analytic model and parameter Estimation owing to become structure control, so this method has that algorithm is simple, good in anti-interference performance, advantages such as canbe used on line easily, is applicable to uncertain non-linear multivariable control object.Non-linear characteristics at Switching Power Supply can be incorporated into Non-Linear Control Theories such as sliding moding structure in the control strategy of high frequency switch power.But switching characteristic is discontinuous in itself, so there is the buffeting problem in Sliding Mode Variable Structure System.In Sliding mode variable structure control, introduce fuzzy control theory, the influence of buffeting can be weakened even eliminate, further improve the performance of sliding formwork control.

Summary of the invention

The purpose of this invention is to provide a kind of Cz monocrystal stove control system of high frequency switch power, it is relatively poor to have solved existing control system dynamic characteristic, causes the output power of power supply fluctuation, influences the problem of monocrystalline quality.

Another object of the present invention provides the method that adopts above-mentioned control system that the Cz monocrystal stove is controlled with high frequency switch power.

The technical solution adopted in the present invention is, the Cz monocrystal stove control system of high frequency switch power, comprise the analog signal isolation collection change-over circuit, wave digital lowpass filter, adaptive fuzzy sliding mode controller, PWM generation module and the IGBT isolated drive circuit that connect successively, analog signal is isolated the collection change-over circuit, the IGBT isolated drive circuit is connected with the power section main circuit respectively, and analog signal is isolated the collection change-over circuit and is connected with LED display module, keyboard and host computer by soft-core processor;

Another technical scheme of the present invention is, the Cz monocrystal stove is specifically implemented according to following steps with the control method of high frequency switch power:

Step 1: Switching Power Supply power section main circuit importation incoming transport voltage, through the front-end filtering rectification, carried out voltage transformation after full-bridge inverting process and the high frequency transformer effect, after the rectifying and wave-filtering of rear end, obtain output dc voltage;

Step 2: set up model according to controlled device, obtain the key parameter in the model: input voltage V _DC, load inductance L _f, load capacitance C _o, load resistance R, original edge voltage loses the ratio R with load current _d

Step 3: according to the modelling adaptive fuzzy sliding mode controller in the step 2, determine coefficient and control law among the switching function s, obtain closed-loop control system;

Step 4: analog signal is isolated the d. c. voltage signal of gathering 1 output of change-over circuit acquisition step, give the adaptive fuzzy sliding mode controller that step 3 obtains as feedback signal, adaptive fuzzy sliding mode controller is obtained controlled quentity controlled variable, and output PWM driving signal amplifies to the IGBT isolated drive circuit;

Step 5: the duty ratio of the driving signal controlling full-bridge inverter after the amplification, when output voltage is higher than set point, reduce the service time of IGBT isolated drive circuit, reduce duty ratio; When output voltage is lower than set point, increase the service time of IGBT isolated drive circuit, rising duty ratio, thereby the turning on and off of control switch pipe, it is steady to regulate output voltage.

Characteristics of the present invention also are,

Step 2 is wherein set up model according to controlled device, obtains the key parameter in the model, specifically implements according to following steps:

The input voltage of definition filter is V _LC-in, input current is I _LC-in, output voltage is V _o, output current is I _o

The transfer function of output filter is:

$H_f=\frac{V_o}{V_{\mathrm{LC}-\mathrm{in}}}=\frac{R//\frac{1}{C_{o}s}}{R//\frac{1}{C_{o}s}+L_{f}s}=\frac{1}{L_f{\mathrm{<figure-callout\; id="2"\; label="C\; o\; s"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">C</figure-callout>}}_o{s}^{}{}^2+\frac{L_f}{R}s+1},$

When load was R, the input impedance of output filter was:

$Z_f=\frac{V_{\mathrm{LC}-\mathrm{in}}}{I_{\mathrm{LC}-\mathrm{in}}}=\frac{I_{\mathrm{LC}-\mathrm{in}}\&times;(L_{f}s+R//\frac{1}{C_{o}s})}{I_{\mathrm{LC}-\mathrm{in}}}=L_{f}s+R//\frac{1}{C_{o}s}=\frac{R+L_{f}s+L_f{C}_o{\mathrm{<figure-callout\; id="2"\; label="Rs"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">Rs</figure-callout>}}^2}{{\mathrm{RC}}_o{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">s</figure-callout>}}^2+1},$

When load was R, the output impedance of output filter was:

$Z_n=\frac{V_o}{I_o}=\frac{V_o}{\frac{V_o}{R}+\frac{V_o}{1/C_{o}s}+\frac{V_o}{L_{f}s}}=\frac{1}{\frac{1}{R}+C_{o}s+\frac{1}{L_{f}s}}=\frac{L_{f}s}{L_f{\mathrm{<figure-callout\; id="2"\; label="C\; o\; s"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">C</figure-callout>}}_o{s}^{}{}^2+\frac{L_f}{R}s+1},$

The transfer function of phase-shifting full-bridge ZVS converter is:

$G_{\mathrm{vd}}=H_{f}n{V}_{\mathrm{DC}}\frac{Z_f}{Z_f+R_d},$

Try to achieve:

$G_{\mathrm{vd}}\left(s\right)=\frac{{\mathrm{nV}}_{\mathrm{DC}}}{L_f{\mathrm{<figure-callout\; id="2"\; label="C\; o\; s"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">C</figure-callout>}}_o{s}^{}{}^2+(\frac{L_f}{R}+R_d{C}_o)s+\frac{R_d}{R}+1},$

Wherein, n is the no-load voltage ratio of transformer, V _DCBe input voltage, L _fBe load inductance, C _oBe load capacitance, R _dBe the ratio of original edge voltage loss with load current, R is load resistance.

Step 3 is wherein determined coefficient and control law among the switching function s according to the modelling adaptive fuzzy sliding mode controller in the step 2, obtains closed-loop control system, specifically implements according to following steps:

If fuzzy system is made of the fuzzy rule of following IF-THEN form:

${\mathrm{\&epsiv;}}^i=\mathrm{<figure-callout\; id="1"\; label="IF\; x"\; filenames="HDA00003308217200011.png,HDA00003308217200012.png"\; state="\{\{state\}\}">IF</figure-callout>}{x}_{}{}_1\mathrm{is}{A}_1^j\mathrm{and}\&CenterDot;\&CenterDot;\&CenterDot;x_n\mathrm{is}{A}_n^j\mathrm{THEN\; y\; is}{B}^j,$

Adopt product inference machine, the average ambiguity solution device of monodrome fuzzy device and center, then fuzzy system is output as:

$y\left(x\right)=\frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{y}^j\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)}{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)},$

Wherein,

Be x _iMembership function;

Introduce vectorial ξ (x), following formula becomes:

y(x)=θ ^Tξ(x)，

Wherein, θ=[y ¹Y ^m] ^T, ξ (x)=[ξ ¹(x) ... ξ ^m(x)] ^T

$\mathrm{\&xi;}\left(x\right)=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)}{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)},$

Adopt fuzzy system to approach ε, then ds (k) becomes:

$\mathrm{ds}\left(k\right)=-\widehat{\mathrm{\&epsiv;}}\mathrm{T\; sgn}\left(s\left(k\right)\right)-\mathrm{qTs}\left(k\right),$

$\widehat{\mathrm{\&epsiv;}}\left(x\right|{\mathrm{\&theta;}}_{\mathrm{\&epsiv;}})={{\mathrm{\&theta;}}_{\mathrm{\&epsiv;}}}^T\mathrm{\&xi;}\left(x\right),$

ξ (x) is fuzzy vector, parameter θ _ε ^TChange according to adaptive law, the design adaptive law is:

${\stackrel{\&CenterDot;}{\mathrm{\&theta;}}}_{\mathrm{\&epsiv;}}=-{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="HDA00003308217200011.png,HDA00003308217200012.png"\; state="\{\{state\}\}">r</figure-callout>}}_1\mathrm{s\&xi;}\left(x\right)u,$

Definition Lyapunove function:

R wherein ₁Be positive constant, then:

Wherein,

With formula

Substitution formula u (k)=(C _eB) ^-1=(C _eR (k+1)-C _eAx (k)-s (k)-ds (k)):

$\stackrel{\&CenterDot;}{V}=-k\left|s\right|+\mathrm{s\&omega;},$

$\stackrel{\&CenterDot;}{V}\&le;0,$

Adopt continuous function S _δReplace sgn (s):

$S_{\mathrm{\&delta;}}=\frac{s}{\left|s\right|+\mathrm{\&delta;}},$

δ=δ ₀+δ ₁||e||，

Wherein, δ ₀And δ ₁Be two positive constants.

Step 4 adaptive fuzzy sliding mode controller is wherein obtained controlled quentity controlled variable, specifically implements according to following steps:

Step 1): it is poor that the d. c. voltage signal that utilization obtains and set point are done, and obtains the derivative of deviation e and deviation

Substitution

Try to achieve sliding-mode surface equation s;

Step 2): with the derivative of sliding-mode surface equation s and sliding-mode surface equation

Equation u (k)=(C of substitution control rate u _eB) ^-1=(C _eR (k+1)-C _eAx (k)-s (k)-ds (k)),

Wherein ds (k)=-ε Tsgn (s (k))-qTs (k), try to achieve control rate u;

Step 3): with the derivative of sliding-mode surface equation s and sliding-mode surface equation Via adaptive fuzzy sliding mode controller (3) computing, obtain the size of auto-adaptive parameter ε, the substitution control rate

u(k)=(C _eB) ^-1=(C _eR(k+1)-C _eAx(k)-s(k)-ds(k))，

Wherein ds (k)=-ε Tsgn (s (k))-qTs (k), thereby the value of regulating control rate u;

Step 4): with the discrete state equations x of system (k+1)=Ax (the k)+Bu (k) of the output valve substitution modeling gained of control rate u, x ∈ R ⁿ, u ∈ R ⁿThereby the PWM of regulation voltage level, and output in real time drives signal and amplifies to the IGBT isolated drive circuit.

The invention has the beneficial effects as follows, utilize Sliding mode variable structure control system parameters perturbation and external disturbance to be had the advantage of very strong robustness, dynamic characteristic and the robustness of raising system when reference voltage variation and load disturbance utilize adaptive fuzzy control method to overcome the buffeting that exists in the sliding formwork control simultaneously.The Switching Power Supply control method that this Adaptive Fuzzy Control combines with sliding formwork control is according to the switch work period, dynamically error is revised, dynamically compensate the size of controlled quentity controlled variable, be conducive to guarantee approx that system moves along diverter surface, reduce the systematic steady state error, reach the purpose that weakens and even eliminate high dither, shorten settling time, improve the dynamic quality of system.

Description of drawings

Fig. 1 is the structural representation of Cz monocrystal stove of the present invention with the control system of high frequency switch power;

Fig. 2 is the phase-shifting full-bridge ZVZCS converter small-signal model that adopts in the inventive method step 2;

Fig. 3 is at the fuzzy sliding mode tracking control fundamental diagram of high frequency switch power in the inventive method step 3;

Fig. 4 is that the fuzzy sliding mode tracking control device is realized block diagram in the inventive method;

Fig. 5 is the flow chart of control method of the present invention;

System started when Fig. 6 was 60V for voltage setting value in the embodiment of the invention Adaptive Fuzzy Sliding Mode Control and sliding formwork control response waveform comparison diagram, wherein, last figure is Adaptive Fuzzy Sliding Mode Control figure as a result, figure below is sliding formwork control figure as a result;

Fig. 7 is Adaptive Fuzzy Sliding Mode Control and the sliding formwork control response waveform comparison diagram when duty cycle property changes between 5 Ω and 10 Ω in the embodiment of the invention, and wherein, last figure is Adaptive Fuzzy Sliding Mode Control figure as a result, and figure below is sliding formwork control figure as a result;

Fig. 8 is Adaptive Fuzzy Sliding Mode Control and the sliding formwork control response waveform comparison diagram when voltage setting value periodically changes between 60V and 80V in the embodiment of the invention, and wherein, last figure is Adaptive Fuzzy Sliding Mode Control figure as a result, and figure below is sliding formwork control figure as a result;

Fig. 9 is respectively 1mH for inductance value in the embodiment of the invention, 3mH, and the Adaptive Fuzzy Sliding Mode Control during 5mH and sliding formwork control response waveform comparison diagram, wherein, last figure is Adaptive Fuzzy Sliding Mode Control figure as a result, figure below is sliding formwork control figure as a result;

The adjustment process schematic diagram of auto-adaptive parameter ε when Figure 10 is 60V for set point in the embodiment of the invention.

Among the figure, 1.IGBT isolated drive circuit, 2.PWM generation module, 3. adaptive fuzzy sliding mode controller, 4. analog signal isolation collection change-over circuit, 5. wave digital lowpass filter.

Embodiment

The present invention is described in detail below in conjunction with embodiment.

At the load of Cz monocrystal stove and the characteristics determined, power main circuit topological structure and main power device parameter are determined; The structure of controller and the control algolithm of employing have determined the performance of whole power supply.The Cz monocrystal stove of the present invention structure of the control system of high frequency switch power, as shown in Figure 1, comprise analog signal isolation collection change-over circuit 4, wave digital lowpass filter 5, adaptive fuzzy sliding mode controller 3, PWM generation module 2 and the IGBT isolated drive circuit 1 that connects successively.Analog signal is isolated being connected with the power section main circuit respectively of collection change-over circuit 4, IGBT isolated drive circuit 1, and analog signal is isolated collection change-over circuit 4 and also is connected with LED display module, keyboard and host computer by soft-core processor.

Wherein analog signal is isolated the collections of gathering change-over circuit 4 main realization power supply outputs, for controller provides feedback information; Power inverter partly adopts the mode of phase-shifting full-bridge, and PWM generation module 2 produces 4 tunnel phase-shift PWM ripples, exports by the final voltage of control duty ratio change power supply according to the output of controller.IGBT isolated drive circuit 1 is protected the reliable turn-off that PWM amplifies to guarantee IGBT simultaneously and is adopted anti-IGBT power device here.These are peripheral slave parts of controller, in case after these designed, the final response performance of power supply depended primarily on designed control algolithm.

The Cz monocrystal stove of the present invention control method of high frequency switch power, as shown in Figure 5: specifically implement according to following steps:

Step 1: Switching Power Supply power section main circuit importation incoming transport voltage, through the front-end filtering rectification, carried out voltage transformation after full-bridge inverting process and the high frequency transformer effect, after the rectifying and wave-filtering of rear end, obtain output dc voltage;

Step 2: set up model according to controlled device, obtain the key parameter in the model, comprise input voltage V _DC, load inductance L _f, load capacitance C _o, load resistance R, original edge voltage loses the ratio R with load current _d

Specifically implement according to following steps:

Phase-shifting full-bridge power supply ZVZCS converter is the Buck code converter, and its small-signal model as shown in Figure 2.There is duty-cycle loss in this circuit, and its equivalent duty ratio is:

$D_{\mathrm{eff}}=D-\frac{2n{L}_r}{V_{\mathrm{DC}}}[{2I}_f-\frac{V_o}{L_f}(1-D)\frac{T}{2}]---\left(1\right)$

As seen D _EffBe subjected to input voltage V _DC, former Buck circuit duty ratio D and load current I _fInfluence, so there are three kinds of disturbances in its small-signal model, its size is respectively:

${\hat{d}}_v=\frac{{4\mathrm{nL}}_r{f}_s{I}_f}{{V_{\mathrm{DC}}}^2}\widehat{V_{\mathrm{DC}}}$

${\hat{d}}_d=(1-\frac{L_r}{L_f}{n}^2{D}_{\mathrm{eff}})\hat{d}---\left(2\right)$

${\hat{d}}_i=\frac{{4\mathrm{nL}}_r{f}_s}{V_{\mathrm{DC}}}\widehat{i_f}$

The input voltage of definition filter is V _LC-in, input current is I _LC-in, output voltage is V _o, output current is I _o

The transfer function of output filter is:

$H_f=\frac{V_o}{V_{\mathrm{LC}-\mathrm{in}}}=\frac{R//\frac{1}{C_{o}s}}{R//\frac{1}{C_{o}s}+L_{f}s}=\frac{1}{L_f{C}_o{s}^2+\frac{L_f}{R}s+1}---\left(3\right)$

When load was R, the input impedance of output filter was:

$Z_f=\frac{V_{\mathrm{LC}-\mathrm{in}}}{I_{\mathrm{LC}-\mathrm{in}}}=\frac{I_{\mathrm{LC}-\mathrm{in}}\&times;(L_{f}s+R//\frac{1}{C_{o}s})}{I_{\mathrm{LC}-\mathrm{in}}}=L_{f}s+R//\frac{1}{C_{o}s}=\frac{R+L_{f}s+L_f{\mathrm{<figure-callout\; id="2"\; label="C\; o\; Rs"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">C</figure-callout>}}_o{\mathrm{Rs}}^{}{}^2}{{\mathrm{<figure-callout\; id="2"\; label="RC\; o\; s"\; filenames="HDA00003308217200011.png"\; state="\{\{state\}\}">RC</figure-callout>}}_o{s}^{}{}^2+1}---\left(4\right)$

When load was R, the output impedance of output filter was:

$Z_n=\frac{V_o}{I_o}=\frac{V_o}{\frac{V_o}{R}+\frac{{\mathrm{<figure-callout\; id="1"\; label="V\; o"\; filenames="HDA00003308217200011.png,HDA00003308217200012.png"\; state="\{\{state\}\}">V</figure-callout>}}_o}{1/C_{o}s}+\frac{V_o}{L_{f}s}}=\frac{1}{\frac{1}{R}+C_{o}s+\frac{1}{L_{f}s}}=\frac{L_{f}s}{L_f{C}_o{s}^2+\frac{L_f}{R}s+1}---\left(5\right)$

Therefore, from controlled quentity controlled variable d to output V _o, the transfer function of phase-shifting full-bridge ZVS converter is:

$G_{\mathrm{vd}}=H_{f}n{V}_{\mathrm{DC}}\frac{Z_f}{Z_f+R_d}---\left(6\right)$

Formula (3) (4) substitution (6) formula is tried to achieve:

$G_{\mathrm{vd}}\left(s\right)=\frac{{\mathrm{nV}}_{\mathrm{DC}}}{L_f{C}_o{s}^2+(\frac{L_f}{R}+R_d{C}_o)s+\frac{R_d}{R}+1}---\left(7\right)$

Wherein, n is the no-load voltage ratio of transformer, V _DCBe input voltage, L _fBe load inductance, C _oBe load capacitance, R _dBe the ratio of original edge voltage loss with load current, R is load resistance.

After the model that has obtained controlled device, by the control method of adaptive fuzzy sliding mode, just can realize the output voltage control of high frequency switch power main circuit.

Step 3: according to the modelling adaptive fuzzy sliding mode controller 2 in the step 2, determine coefficient and control law among the switching function s, obtain closed-loop control system;

Specifically implement according to following steps:

Adaptive Fuzzy Sliding Mode Control is that Adaptive Fuzzy Control and traditional sliding formwork control are combined, and the advantage of the two is combined closely.Fuzzy sliding mode tracking control has two advantages, and the one, the control target transfers the sliding formwork function to from tracking error, to make sliding formwork function s be zero as long as apply control, tracking error is with progressive arrival zero point.The 2nd, for the high order system more than the second order, in routine control input should for

And the input of fuzzy sliding mode tracking control

All the time be two-dimentional, as shown in Figure 3.The meaning of fuzzy control is its soft and smooth control signal, the chattering phenomenon that alleviates or avoided general sliding formwork to occur.

1. sliding mode controller design

If the discrete system state equation is as follows:

x(k+1)=Ax(k)+Bu(k),x∈R ⁿ,u∈R ⁿ （8）

Position command is r (k), the derivative of position command dr (k), R=[r (k), dr (k)], R ₁=[r (k+1), dr (k+1)].

R (k) and r (k+1) adopt the method for linear extrapolation to predict:

r(k+1)=2r(k)-r(k-1),dr(k+1)=2dr(k)-dr(k-1) （9）

Switching function is

s(k)=C _e[r(k)-x(k),dr(k)-dx(k)] ^T （10）

C wherein _e=[c, 1].

Discrete control rate based on index convergence rate:

u(k)=(C _eB) ^-1=(C _eR(k+1)-C _eAx(k)-s(k)-ds(k)) （11）

Wherein ds (k)=-ε Tsgn (s (k))-qTs (k).

Under fixing condition of sampling time, ε value has determined the amplitude of controller buffeting.

Define a Lyaounov function:

$V=\frac{1}{2}{s}^2---\left(12\right)$

At second-order system, have

Then have:

$\stackrel{\&CenterDot;}{V}=s\stackrel{\&CenterDot;}{s}---\left(13\right)$

Bring formula (11) into formula (13), wherein

Provable

So the stable of sliding formwork control law must be demonstrate,proved.

2. adaptive fuzzy sliding mode controller design

If the ε in the formula (11) is difficult for determining, can adopts fuzzy system

Replace ε, realize Adaptive Fuzzy Sliding Mode Control.

If fuzzy system is made of the fuzzy rule of following IF-THEN form:

${\mathrm{\&epsiv;}}^i=\mathrm{IF}{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="HDA00003308217200011.png,HDA00003308217200012.png"\; state="\{\{state\}\}">x</figure-callout>}}_1\mathrm{is}{A}_1^j\mathrm{and}\&CenterDot;\&CenterDot;\&CenterDot;x_n\mathrm{is}{A}_n^j\mathrm{THEN\; y\; is}{B}^j$

Adopt product inference machine, the average ambiguity solution device of monodrome fuzzy device and center, then fuzzy system is output as:

$y\left(x\right)=\frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{y}^j\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)}{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)}---\left(14\right)$

Wherein, Be x _iMembership function.

Introduce vectorial ξ (x), formula (14) becomes:

y(x)=θ ^Tξ(x) （15）

Wherein, θ=[y ¹Y ^m] ^T, ξ (x)=[ξ ¹(x) ... ξ ^m(x)] ^T

$\mathrm{\&xi;}\left(x\right)=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)}{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)}---\left(16\right)$

In working control, ε often is difficult for determining that control rate (11) is difficult to realize, adopts fuzzy system to approach ε, and then ds (k) becomes:

$\mathrm{ds}\left(k\right)=-\widehat{\mathrm{\&epsiv;}}\mathrm{T\; sgn}\left(s\left(k\right)\right)-\mathrm{qTs}\left(k\right)---\left(17\right)$

$\widehat{\mathrm{\&epsiv;}}\left(x\right|{\mathrm{\&theta;}}_{\mathrm{\&epsiv;}})={{\mathrm{\&theta;}}_{\mathrm{\&epsiv;}}}^T\mathrm{\&xi;}\left(x\right)---\left(18\right)$

ξ (x) is fuzzy vector, parameter θ _ε ^TChange according to adaptive law.The design adaptive law is:

${\stackrel{\&CenterDot;}{\mathrm{\&theta;}}}_{\mathrm{\&epsiv;}}=-{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="HDA00003308217200011.png,HDA00003308217200012.png"\; state="\{\{state\}\}">r</figure-callout>}}_1\mathrm{s\&xi;}\left(x\right)u---\left(19\right)$

Definition Lyapunove function:

R wherein ₁Be positive constant, then:

Wherein,

Formula (20) substitution control rate (11) is got:

$\stackrel{\&CenterDot;}{V}=-k\left|s\right|+\mathrm{s\&omega;}---\left(22\right)$

According to fuzzy approximation theory, adaptive fuzzy system can realize that approximate error ω is very little.Therefore:

$\stackrel{\&CenterDot;}{V}\&le;0---\left(23\right)$

In order to reduce buffeting, adopt continuous function S _δReplace sgn (s):

$S_{\mathrm{\&delta;}}=\frac{s}{\left|s\right|+\mathrm{\&delta;}}---\left(24\right)$

δ=δ ₀+δ ₁||e|| （25）

Wherein, δ ₀And δ ₁Be two positive constants, so designed adaptive law is stable.

In order to reach higher arithmetic speed and the real-time of control, native system adopts the three class pipeline structure when the FPGA hardware platform is realized above-mentioned algorithm, be respectively input obfuscation, fuzzy reasoning output and sliding formwork control, three links are carried out computing simultaneously, reach the optimum operation efficient of algorithm.

The membership values of variable generally adopts look-up table to obtain, traditional look-up table namely is the degree of membership of storing corresponding each the fuzzy element of each domain element in FPGA respectively, need more FPGA ROM resource like this, and after inquiry, need the degree of membership of each fuzzy element correspondence is carried out record, waste the FPGA resource, influence the operational efficiency of program.The present invention proposes comparatively directly degree of membership query method, here adopt Triangleshape grade of membership function, because each domain element is 0.5 or 1 greater than 0 degree of membership, namely be under the jurisdiction of two fuzzy languages at most, therefore at each domain element design 6Bit(2*3Bit) ROM cell store it and be subordinate to situation, with domain language in triad (000～110) expression 7, XXX is represented as sky, and namely this value only is under the jurisdiction of preceding 3Bit.The domain element value can directly obtain it and is subordinate to language and membership function value as searching the address, has so not only saved resource but also has simplified search procedure.

The input s of fuzzy controller, Be two variablees, each variable comprises 7 fuzzy elements, and therefore, inference rule is 49, rule list adopts the mode of look-up table to design equally, for taking minimum resource, selects for use the ROM of 49*4bit as rule list, address wire is 6, and high 3 is the fuzzy language value of input variable S, and low 3 are

Fuzzy language value.Because fuzzy element adopts 000～110 binary system to represent that successively therefore after the previous stage streamline was finished the inquiry of fuzzy variable degree of membership, what directly will obtain was subordinate to element as the address in program, can find corresponding reasoning output in ROM.

The implementation procedure of sliding mode controller as shown in Figure 4.Be input as deviation e and deviation derivative

After constituting switching function s it is exported, get equation s and the derivative thereof of switching function

As the fuzzy controller input, fuzzy controller output controlled quentity controlled variable, the controlled quentity controlled variable of returning and deviation e and deviation derivative

The switching function s that must make new advances after the computing, duty ratio is controlled by the s that continues the computing renewal, has finally reached the control switch pipe and has opened shutoff, makes system's output voltage stabilization in set point.

Step 4: analog signal is isolated the d. c. voltage signal of gathering 1 output of change-over circuit 4 acquisition step, give the controller part as feedback signal, the adaptive fuzzy sliding mode controller 3 that controller partly utilizes step 3 to obtain is obtained controlled quentity controlled variable, and concrete computational process is:

1. it is poor to utilize magnitude of voltage that collecting part obtains and set point to do, and obtains the derivative of deviation e and deviation

Substitution

Try to achieve the equation on sliding formwork plane.

2. with the derivative of sliding-mode surface equation s and sliding-mode surface equation

The equation of substitution control rate u is tried to achieve control rate u.

3. with the derivative of sliding-mode surface equation s and sliding-mode surface equation

Via the fuzzy controller computing, obtain the size of auto-adaptive parameter ε, regulate the value of control rate u.

4. with system's discrete state equations of the output valve substitution modeling gained of control rate u, thus regulation voltage level in real time, and output PWM drives signal to 1 amplification of IGBT isolated drive circuit;

Step 5: the duty ratio of the driving signal controlling full-bridge inverter after the amplification.When output voltage is higher than set point, reduce the service time of IGBT isolated drive circuit 1, reduce duty ratio; When output voltage is lower than set point, increase the service time of IGBT isolated drive circuit 1, rising duty ratio, thereby the turning on and off of control switch pipe, it is steady to regulate output voltage, reaches the purpose of regulating electric power output voltage or power.

Embodiment

In order to verify the effect of above-mentioned adaptive fuzzy sliding mode controller, at TDR-100 type single crystal growing furnace full-bridge inverting high frequency switch power, parameter is chosen as follows: switching function is Wherein, c=4.5 in the control rate, q=50.Power input voltage V=380V, switching frequency f=20kHz.Resistance 5 Ω of load, load inductance is 1mH.

Fig. 6 controls when voltage setting value is 60V comparative result figure as a result for Adaptive Fuzzy Sliding Mode Control and sliding formwork.Wherein, last figure is fuzzy sliding mode tracking control figure as a result, and figure below is sliding formwork control figure as a result.To arrive the stable time be 0.005s to the Adaptive Fuzzy Sliding Mode Control mode as seen from the figure, and it is 0.015s that the sliding formwork control mode arrives the stable time, and obviously as can be seen sliding formwork control exist more serious buffeting constantly starting.

Fig. 7 is the control of Adaptive Fuzzy Sliding Mode Control and sliding formwork when the variation of duty cycle property between 5 Ω and 10 Ω, and fuzzy sliding mode tracking control and sliding formwork are controlled figure as a result.Wherein, last figure is fuzzy sliding mode tracking control figure as a result, and figure below is sliding formwork control figure as a result.As seen from the figure in 0.04s and the 0.08s moment, set load value and change saltus step, bigger fluctuation appears in sliding formwork control response waveform, reaches stable state again through 0.002s, the fuzzy sliding mode tracking control response wave shape then fluctuates less, reaches stable state again through 0.001s.

When Fig. 8 periodically changed between 60V and 80V for voltage setting value, fuzzy sliding mode tracking control and sliding formwork control is figure as a result.Wherein, last figure is fuzzy sliding mode tracking control figure as a result, and figure below is sliding formwork control figure as a result.At 0.04s and 0.08 constantly, voltage setting value changes as can be seen, and it is stable that fuzzy sliding mode tracking control reaches behind 0.002s rapidly again, and sliding formwork regulating and controlling time and sliding formwork regulating and controlling time phase difference are little, but shake is stronger.

Fig. 9 is respectively 1mH for inductance value, 3mH, and during 5mH, the startup response wave shape of Adaptive Fuzzy Sliding Mode Control and the control of sliding formwork control comparison diagram.Wherein, last figure is fuzzy sliding mode tracking control figure as a result, and figure below is sliding formwork control figure as a result.Two kinds of control methods are when inductance value increases as can be seen among the figure, and response speed is slack-off.But fuzzy sliding mode tracking control is controlled than sliding formwork, and it is very fast that response speed is wanted, and the buffeting amplitude is less, and system still keeps good performance.

Figure 10 describes under the modified fuzzy sliding mode controlling method, and set point is 60V, the adjustment process of auto-adaptive parameter ε, and the starter system oscillation phase, parameter ε carries out self adaptation to be regulated, and reaches steady operation moment ε in system and reaches stationary value too, finishes adjustment process.

By above-mentioned comparing result as can be seen, Adaptive Fuzzy Sliding Mode Control combines the strong adaptability of sliding formwork control and the characteristics that fuzzy control does not need the accurate equation of system compared to fuzzy sliding mode tracking control.Buffeting reduces, and shorten start-up time, and under the sudden change situation of system load and reference voltage, the adjusting time is shorter simultaneously, and overshoot is lower, and under modified fuzzy sliding mode controlling method, the ripple of system's waveform is less in addition, so the robustness of control method is stronger.Under the situation of circuit parameter disturbance, can both reach stable state, and disturbance has consistency to circuit parameter after arriving stable state.

Claims (6)

1.Cz the monocrystal stove control system of high frequency switch power, it is characterized in that, comprise the analog signal isolation collection change-over circuit (4), wave digital lowpass filter (5), adaptive fuzzy sliding mode controller (3), PWM generation module (2) and the IGBT isolated drive circuit (1) that connect successively, described analog signal is isolated collection change-over circuit (4), IGBT isolated drive circuit (1) is connected with the power section main circuit respectively.

2. described Cz monocrystal stove according to claim 1 is characterized in that with the control system of high frequency switch power, and described analog signal is isolated collection change-over circuit (4) and is connected with LED display module, keyboard and host computer by soft-core processor.

3.Cz the monocrystal stove control method of high frequency switch power, it is characterized in that, adopt the Cz monocrystal stove control system of high frequency switch power, the structure of described control system is: comprise the analog signal isolation collection change-over circuit (4) that connects successively, wave digital lowpass filter (5), adaptive fuzzy sliding mode controller (3), PWM generation module (2) and IGBT isolated drive circuit (1), described analog signal is isolated collection change-over circuit (4), IGBT isolated drive circuit (1) is connected with the power section main circuit respectively, and described analog signal is isolated collection change-over circuit (4) by soft-core processor and LED display module, keyboard and host computer connect; Described control method, specifically implement according to following steps:

Step 1: Switching Power Supply power section main circuit importation incoming transport voltage, through the front-end filtering rectification, carried out voltage transformation after full-bridge inverting process and the high frequency transformer effect, after the rectifying and wave-filtering of rear end, obtain output dc voltage;

Step 2: set up model according to controlled device, obtain the key parameter in the model: input voltage V _DC, load inductance L _f, load capacitance C _o, load resistance R, original edge voltage loses the ratio R with load current _d

Step 3: according to the modelling adaptive fuzzy sliding mode controller (2) in the step 2, determine coefficient and control law among the switching function s, obtain closed-loop control system;

Step 4: analog signal is isolated the d. c. voltage signal of gathering 1 output of change-over circuit (4) acquisition step, give the adaptive fuzzy sliding mode controller (3) that step 3 obtains as feedback signal, adaptive fuzzy sliding mode controller (3) is obtained controlled quentity controlled variable, and output PWM driving signal amplifies to IGBT isolated drive circuit (1);

Step 5: the duty ratio of the driving signal controlling full-bridge inverter after the amplification, when output voltage is higher than set point, reduce the service time of IGBT isolated drive circuit (1), reduce duty ratio; When output voltage is lower than set point, increase the service time of IGBT isolated drive circuit (1), rising duty ratio, thereby the turning on and off of control switch pipe, it is steady to regulate output voltage.

4. Cz monocrystal stove according to claim 3 is characterized in that with the control method of high frequency switch power described step 2 is set up model according to controlled device, obtains the key parameter in the model, specifically implements according to following steps:

The input voltage of definition filter is V _LC-in, input current is I _LC-in, output voltage is V _o, output current is I _o

The transfer function of output filter is:

$H_f=\frac{V_o}{V_{\mathrm{LC}-\mathrm{in}}}=\frac{R//\frac{1}{C_{o}s}}{R//\frac{1}{C_{o}s}+L_{f}s}=\frac{1}{L_f{C}_o{s}^2+\frac{L_f}{R}s+1},$

When load was R, the input impedance of output filter was:

$Z_f=\frac{V_{\mathrm{LC}-\mathrm{in}}}{I_{\mathrm{LC}-\mathrm{in}}}=\frac{I_{\mathrm{LC}-\mathrm{in}}\&times;(L_{f}s+R//\frac{1}{C_{o}s})}{I_{\mathrm{LC}-\mathrm{in}}}=L_{f}s+R//\frac{1}{C_{o}s}=\frac{R+L_{f}s+L_f{C}_o{\mathrm{Rs}}^2}{{\mathrm{RC}}_o{s}^2+1},$

When load was R, the output impedance of output filter was:

$Z_n=\frac{V_o}{I_o}=\frac{V_o}{\frac{V_o}{R}+\frac{V_o}{1/C_{o}s}+\frac{V_o}{L_{f}s}}=\frac{1}{\frac{1}{R}+C_{o}s+\frac{1}{L_{f}s}}=\frac{L_{f}s}{L_f{C}_o{s}^2+\frac{L_f}{R}s+1},$

The transfer function of phase-shifting full-bridge ZVS converter is:

$G_{\mathrm{vd}}=H_{f}n{V}_{\mathrm{DC}}\frac{Z_f}{Z_f+R_d},$

Try to achieve:

$G_{\mathrm{vd}}\left(s\right)=\frac{{\mathrm{nV}}_{\mathrm{DC}}}{L_f{C}_o{s}^2+(\frac{L_f}{R}+R_d{C}_o)s+\frac{R_d}{R}+1},$

Wherein, n is the no-load voltage ratio of transformer, V _DCBe input voltage, L _fBe load inductance, C _oBe load capacitance, R _dBe the ratio of original edge voltage loss with load current, R is load resistance.

5. Cz monocrystal stove according to claim 3 is with the control method of high frequency switch power, it is characterized in that, described step 3 is according to the modelling adaptive fuzzy sliding mode controller (2) in the step 2, determine coefficient and control law among the switching function s, obtain closed-loop control system, specifically implement according to following steps:

If fuzzy system is made of the fuzzy rule of following IF-THEN form:

${\mathrm{\&epsiv;}}^i=\mathrm{IF}{x}_1\mathrm{is}{A}_1^j\mathrm{and}\&CenterDot;\&CenterDot;\&CenterDot;x_n\mathrm{is}{A}_n^j\mathrm{THEN\; y\; is}{B}^j,$

Adopt product inference machine, the average ambiguity solution device of monodrome fuzzy device and center, then fuzzy system is output as:

$y\left(x\right)=\frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{y}^j\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)}{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)},$

Wherein, Be x _iMembership function;

Introduce vectorial ξ (x), following formula becomes:

y(x)=θ ^Tξ(x)，

Wherein, θ=[y ¹Y ^m] ^T, ξ (x)=[ξ ¹(x) ... ξ ^m(x)] ^T

$\mathrm{\&xi;}\left(x\right)=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)}{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\mathrm{\&mu;}{A}_i^j\left(x_i\right)\right)},$

Adopt fuzzy system to approach ε, then ds (k) becomes:

$\mathrm{ds}\left(k\right)=-\widehat{\mathrm{\&epsiv;}}\mathrm{T\; sgn}\left(s\left(k\right)\right)-\mathrm{qTs}\left(k\right),$

$\widehat{\mathrm{\&epsiv;}}\left(x\right|{\mathrm{\&theta;}}_{\mathrm{\&epsiv;}})={{\mathrm{\&theta;}}_{\mathrm{\&epsiv;}}}^T\mathrm{\&xi;}\left(x\right),$

ξ (x) is fuzzy vector, parameter θ _ε ^TChange according to adaptive law, the design adaptive law is:

${\stackrel{\&CenterDot;}{\mathrm{\&theta;}}}_{\mathrm{\&epsiv;}}=-r_1\mathrm{s\&xi;}\left(x\right)u,$

Definition Lyapunove function:

R wherein ₁Be positive constant, then:

Wherein,

With formula

Substitution formula u (k)=(C _eB) ^-1=(C _eR (k+1)-C _eAx (k)-s (k)-ds (k)):

$\stackrel{\&CenterDot;}{V}=-k\left|s\right|+\mathrm{s\&omega;},$

$\stackrel{\&CenterDot;}{V}\&le;0,$

Adopt continuous function S _δReplace sgn (s):

$S_{\mathrm{\&delta;}}=\frac{s}{\left|s\right|+\mathrm{\&delta;}},$

δ＝δ ₀+δ ₁||e||，

Wherein, δ ₀And δ ₁Be two positive constants.

6. Cz monocrystal stove according to claim 3 is characterized in that described step 4 adaptive fuzzy sliding mode controller (3) is obtained controlled quentity controlled variable, specifically implements according to following steps with the control method of high frequency switch power:

Step 1): it is poor that the d. c. voltage signal that utilization obtains and set point are done, and obtains the derivative of deviation e and deviation

Substitution

Try to achieve sliding-mode surface equation s;

Step 2): with the derivative of sliding-mode surface equation s and sliding-mode surface equation Equation u (k)=(C of substitution control rate u _eB) ^-1=(C _eR (k+1)-C _eAx (k)-s (k)-ds (k)),

Wherein ds (k)=-ε Tsgn (s (k))-qTs (k), try to achieve control rate u;

Step 3): with the derivative of sliding-mode surface equation s and sliding-mode surface equation

Via adaptive fuzzy sliding mode controller (3) computing, obtain the size of auto-adaptive parameter ε, the substitution control rate

u(k)=(C _eB) ^-1=(C _eR(k+1)-C _eAx(k)-s(k)-ds(k))，

Wherein ds (k)=-ε Tsgn (s (k))-qTs (k), thereby the value of regulating control rate u;

Step 4): with the discrete state equations x of system (k+1)=Ax (the k)+Bu (k) of the output valve substitution modeling gained of control rate u, x ∈ R ⁿ, u ∈ R ⁿThereby the PWM of regulation voltage level, and output in real time drives signal and amplifies to IGBT isolated drive circuit (1).

Images