CN104967342A - Midpoint potential predictive control method for diode clamped three-level converter - Google Patents

Abstract

The invention relates to the modulation technology in the field of multi-level power electronic converters, in particular to a midpoint potential predictive control method for a diode clamped three-level converter. The disclosed midpoint potential predicative control method for the diode clamped three-level converter comprises the steps of first dividing a space vector into 12 sub intervals; introducing a predictive unbalanced assessment function which selects the optimal switch action sequence by taking into consideration the unbalanced voltage possibly generated in the action process by the action sequence of all the sub intervals while combining the actual unbalanced voltage, and achieves the purpose of balancing the midpoint potential. The method requires no setting of balance factors, and enables a substantial increase in the robustness of the midpoint potential control.

Description

The midpoint potential forecast Control Algorithm of diode-clamped three-level converter

Technical field

The present invention relates to the modulation technique in multilevel power electronic converter field, be specifically related to the midpoint potential forecast Control Algorithm of diode-clamped three-level converter.

Background technology

In recent years, the diode-clamped three-level converter advantage that capacity is large, voltage is high owing to having, current harmonic content is little, in mine, locomotive traction, the field such as steel rolling be widely used.

Two traditional level converters, due to the shortcoming that two level converters are intrinsic, as dv/dt is large, EMI is large, and switching frequency is high, and inversion efficiency is low, is not suitable for high-voltage large-capacity high-power applications.The relatively traditional two level topologys of diode-clamped three level converters have clear superiority: 1) each power device pressure-bearing is only the half of direct voltage, thus allow that there is higher electric pressure, greatly reduce volume and the cost of device simultaneously; 2) dv/dt compared with little, output current aberration rate is little, converter capacity is large.

For diode-clamped three-level converter, neutral balance problem is the intrinsic problem that it exists.The tie point of two series capacitances is exactly its current potential terminal, when fluctuation occurs the current potential of this point, has just occurred midpoint potential energy imbalance.Midpoint potential unbalance, not only can cause the distortion of AC output voltage and electric current, the voltage born when switching device also can be caused to turn off is inconsistent, once exceed the pressure-bearing scope of power device, will cause device breakdown, have a strong impact on system performance.

The neutral point voltage control method of diode-clamped three-level converter mainly contains: one is by regulating realize neutral point voltage control, as point control in balance factor method, virtual vector etc. the action time of redundancy small vector; Two is according to the mid point sense of current, inject or extract residual voltage, realize mid point ER effect to thus complete control.Above two kinds of methods are comparatively serious at midpoint potential imbalance problem, or may occur mistake when mid point electric current exists larger harmonic wave.

Summary of the invention

The object of the invention is for this topological structure of diode-clamped three-level converter, existing neutral point voltage control strategy improved, under the prerequisite ensureing neutral-point potential balance, reduces the dependence of alignment electric current, improve the robustness of neutral point voltage control.

The present invention adopts following technical scheme:

The midpoint potential Forecasting Methodology of diode-clamped three-level converter, comprises the following steps,

Step 1: set up each switch working state in converter space vector plane figure, and be 12 subintervals by space vector plane diagram root, there are four kinds of switching sequences in each subinterval;

Step 2: according to the position of output voltage vector, corresponds to one of 12 subintervals by output voltage vector, measure DC side two capacitance voltage difference DELTA u, and calculate corresponding actual uneven function f (x _act), the interval respective value of zygote and the uneven function of reality determine middle vector position;

Step 3: determine four kinds of switching sequences in subinterval, calculates the action time of each vector in four switching sequences in subinterval, if summation action time of all vectors is greater than pulse width modulation cycle in single sequence of operation, then adjusts action time each vector;

Step 4: the action time utilizing vector in the on-off action order and each sequence of operation obtained in step 3, in conjunction with mid point current i corresponding during each vector effect _k, the uneven function f (x of computational prediction _pred);

Step 5: according to unbalance voltage minimization principle, namely uneven valuation functions f (x) is minimum, selects suitable on-off action order;

Wherein, uneven valuation functions f (x)=f (x _pred)+f (x _act), f (x in formula _pred) be the uneven function of prediction, f (x _act) be actual uneven function; And wherein i in formula _kfor in selected sequence of operation, mid point current value during kth vector effect, t _kfor the time of this vector effect; F (x _act)=2C Δ u, in formula, C is DC bus capacitor value, and Δ u is DC side two capacitance voltage difference.

Further, be greater than pulse width modulation cycle to summation action time of vectors all in single sequence of operation in step 3, method of adjustment is for be adjusted to t'=t/t action time by each vector _pWM.

Further, to the sequence of operation comprising middle vector and large vector in step 4, select to add redundant vectors and action time thereof and recalculate action time of other vectors in sequence of operation, subsequently the uneven function f (x of computational prediction again _pred).

The present invention proposes the midpoint potential forecast Control Algorithm of diode-clamped three-level converter, first space vector is divided into 12 subintervals; Introduce the uneven valuation functions of prediction, all sequences of operation in each subinterval issuable unbalance voltage in mechanism is taken into account by this function, consider in conjunction with actual unbalance voltage, select optimum on-off action order, reach the object of balance midpoint potential.

The method that the present invention proposes, without the need to the design balance factor, can ensure the optimization of neutral point voltage control effect; There is good robustness, and be easy to migrate to more high level, not by the restriction of level quantity; Control algolithm accuracy is high, is easy to Project Realization, saves software and hardware resource, can ensure the neutral-point potential balance of any time.

Accompanying drawing explanation

Fig. 1 is the topological structure of diode-clamped three-level converter;

Fig. 2 is that the space vector of diode-clamped three-level converter divides subinterval;

Fig. 3 is 0 subinterval vector distribution map;

Fig. 4 is that midpoint potential forecast Control Algorithm uses, do not use again to midpoint potential deviation map when using.

Embodiment

For further illustrating each embodiment, the invention provides drawings attached.These accompanying drawings are a part for disclosure of the present invention, and it is mainly in order to illustrate embodiment, and the associated description of specification can be coordinated to explain the operation principles of embodiment.Coordinate with reference to these contents, those of ordinary skill in the art will be understood that other possible execution modes and advantage of the present invention.Assembly in figure not drawn on scale, and similar element numbers is commonly used to assembly like representation class.

Now the present invention is further described with embodiment by reference to the accompanying drawings.

The invention discloses a kind of midpoint potential forecast Control Algorithm of diode-clamped three-level converter, by selector switch sequence of operation, realize pulse width modulation algorithm.Consulting shown in Fig. 1, is the topological structure of diode-clamped three-level converter (abbreviation three-level converter).Three-level converter comprises two series capacitances, and each brachium pontis in three-phase brachium pontis is made up of 4 device for power switching and corresponding anti-paralleled diode, the neutral point clamp circuit that on each brachium pontis, two series diodes and middle switching device form in addition.

Respectively be made up of upper brachium pontis, lower brachium pontis, each brachium pontis comprises 4 power device (S _x1-S _x4, x=a, b, c), clamping capacitance is C ₁and C ₂.The topological structure analyzing three-level converter is known, and every phase brachium pontis of three-level converter all has three kinds of on off states, then three-level converter has 3 ³=27 kinds of on off states, that is:

In this embodiment, the on off state of three-level converter is S _x, x=a, b, c, wherein, S _a, S _b, S _cbe respectively a of three-level converter, three on off state variablees of b, c three-phase brachium pontis.S _x=-1 represents that the output termination N of the x phase brachium pontis of three-level converter holds, S _xthe output termination neutral point O end of=0 expression x phase brachium pontis, S _x=1 represents that the output termination P of described x phase brachium pontis holds.

24 kinds of non-zero are had, 3 kinds of zero vectors in 27 kinds of vectors of three-level converter.The size long according to Vector Mode is divided into zero vector, short vector, middle vector long vector, and zero vector does not affect output capacitance midpoint potential.Large vector effect does not affect two capacitor voltage equalizings; Middle vector influence two capacitor voltage equalizing, but middle vector only has a kind of mode of production, there is not redundancy, therefore all pressures of two electric capacity cannot be realized with it; The balance of short vector effect to two capacitance voltages has the greatest impact, and sees that short vector occurs in pairs in figure 3, is redundancy, can be used for controlling neutral balance.

The space vector of diode-clamped three-level converter can be divided into 12 subintervals, and as shown in Figure 2, interval 0-11, determines amplitude and the position of each vector according to mid point unbalance voltage.For 12 subintervals, there are four kinds of switching sequences in each subinterval, as shown in Figure 3, is for 0,1 subinterval, four kinds of on-off actions orders corresponding to this interval be (careful for 0 subinterval herein, simultaneously for guarantee from zero vector, zero vector terminates):

(000)-(110)-(100)-(000) ①

(000)-(0-1-1)-(00-1)-(000) ②

(000)-(1-1-1)-(10-1)-(000) ③

(000)-(1-1-1)-(11-1)-(000) ④

Wherein 4. sequence of operation only employs alignment current potential and does not have influential long vector and zero vector.1. and 2. sequence of operation comprises the influential short vector of alignment current potential, and this switching sequence corrects according to mid point sense of current alignment current potential imbalance.The amplitude of the short vector used in 1. by brachium pontis electric capacity C ₁magnitude of voltage determine, but the amplitude of the short vector used in is 2. by lower brachium pontis electric capacity C ₂magnitude of voltage determine.1. or 2. in the uneven situation of mid point, selector switch sequence of operation.Due to the limitation of short vector magnitude, only have when output voltage vector is in little hexagonal area, these two kinds of on-off action orders just can be selected to balance midpoint potential.If 1. or 2. sequence of operation can not meet the demands, then select only to comprise the sequence of operation of an influential middle vector of alignment current potential 3..The amplitude of middle vector in space vector distribution map and position are determined by the uneven degree of actual midpoint potential.

For carrying out midpoint potential prediction, first provide the L-expression of middle vector under α β coordinate system

$u_{\mathrm{\α}}=\sqrt{\frac{1}{6}}\left[\begin{array}{c}{u}_{Cu}\·(2\·(T_{au1}\·T_{au2})-(T_{bu1}\·T_{bu2})-(T_{cu1}\·T_{cu2}\left)\right)\\ -u_{C1}\·(2\·(T_{al1}\·T_{al2})-(T_{bl1}\·T_{bl2})-(T_{cl1}\·T_{cl2}\left)\right)\end{array}\right]$

$u_{\mathrm{\β}}=\frac{\sqrt{2}}{2}\left[\begin{array}{c}{u}_{Cu}\·\left(\right(T_{bu1}\·T_{bu2})-(T_{cu1}\·T_{cu2}\left)\right)\\ -u_{C1}\·\left(\right(T_{bl1}\·T_{bl2})-(T_{cl1}\·T_{cl2}\left)\right)\end{array}\right]$

T in formula _{x (u, l) (1,2)}value by the Determines of corresponding switch, being 1 when opening, is 0 during shutoff.

It is minimum that on-off action selection sequentially must ensure while obtaining suitable output voltage vector, make midpoint potential imbalance.Neutral point voltage control strategy should consider the impact of on-off action order alignment current potential.To any one in four kinds of sequences of operation, the uneven valuation functions of definition midpoint potential

f(x)＝f(x _pred)+f(x _act)

F (x in formula _pred) be the uneven function of prediction, f (x _act) be actual uneven function.And wherein

$f\left(x_{pred}\right)=\underset{k=1}{\overset{n}{\Σ}}{t}_k\·i_k$

I in formula _kfor in selected sequence of operation, mid point current value during kth vector effect, t _kfor the time of this vector effect.

f(x _act)＝2·C·Δu

In formula, C is capacitance, and Δ u is two capacitance voltage differences.In some cases, due to sequence of operation, 1. or 2. amplitude is limited, and its adjustment midpoint potential is unbalanced limited in one's ability, and selection index system order 3..Consider uneven valuation functions

$f\left(x\right)=\underset{k=1}{\overset{n}{\Σ}}{t}_k\·i_k+2\·C\·\mathrm{\Δ}u=0$

If when 3. order can not meet, suitable redundant vectors will be added to control midpoint potential in 3..Therefore new on-off action order is obtained.The coefficient of uneven valuation functions value is depended in sequence of operation selection 5. or 6..

(000)—(0-1-1)—(1-1-1)—(10-1)—(000) ⑤

(000)—(1-1-1)—(10-1)—(100)—(000) ⑥

The action time of the redundant vectors added can be calculated by following formula

$t_{(0-1-1),\left(100\right)}=abs\left(\frac{f\left(x\right)}{i_{(0-1-1),\left(100\right)}}\right)$

Midpoint potential forecast Control Algorithm of the present invention can be divided into following four steps to implement:

(1) judge the position of output voltage vector, and corresponded in the subinterval of 0-11; Measure unbalance voltage Δ u, and calculate corresponding actual uneven function f (x _act); The interval respective value of zygote and the uneven function of reality determine middle vector position again.

(2) determine four kinds of on-off action orders in each subinterval respectively, and calculate the action time of each vector in each sequence of operation; Judge whether summation action time of all vectors in single sequence of operation is greater than the PWM cycle, if be greater than, need to adjust action time each vector.If the action time of all vectors is greater than the PWM cycle in single sequence of operation, then need to adjust action time each vector.Adjustment mode is as follows:

If t _1-1-1+ t _10-1> t _pWM, then $\begin{array}{c}{t}_{1-1-1}^{\′}=t_{1-1-1}/t_{PWM}\\ t_{10-1}^{\′}=t_{10-1}/t_{PWM}\end{array}.$

(3) respectively to the action time of vector in the switching sequence obtained in step (2) and each order, in conjunction with mid point current i corresponding during each vector effect _k, the uneven function f (x of computational prediction _pred).For the sequence of operation comprising middle vector and large vector, select to add redundant vectors and action time thereof and recalculate action time of other vectors in sequence of operation, subsequently the uneven function f (x of computational prediction again _pred).

(4) suitable on-off action order is selected according to unbalance voltage value minimization principle.

Consult shown in Fig. 4, for the midpoint potential forecast Control Algorithm of correspondence---not to use---midpoint potential deviate during use from use.Wherein, curve A is electric capacity C ₁the voltage V at two ends _dc1, curve B is electric capacity C ₂both end voltage V _dc2.When not using midpoint potential deviate, electric capacity C ₁the voltage V at two ends _dc1with electric capacity C ₂both end voltage V _dc2between there is certain error; When using midpoint potential deviate, curve A and curve B overlap, voltage V _dc1with voltage V _dc2equal.

Although specifically show in conjunction with preferred embodiment and describe the present invention; but those skilled in the art should be understood that; not departing from the spirit and scope of the present invention that appended claims limits; can make a variety of changes the present invention in the form and details, be protection scope of the present invention.

Claims (3)

1. the midpoint potential Forecasting Methodology of diode-clamped three-level converter, is characterized in that: said method comprising the steps of,

Step 1: set up each switch working state in converter space vector plane figure, and be 12 subintervals by space vector plane diagram root, there are four kinds of switching sequences in each subinterval;

Step 2: according to the position of output voltage vector, corresponds to one of 12 subintervals by output voltage vector, measure DC side two capacitance voltage difference DELTA u, and calculate corresponding actual uneven function f (x _act), the interval respective value of zygote and the uneven function of reality determine middle vector position;

Step 3: determine four kinds of switching sequences in subinterval, calculates the action time of each vector in four switching sequences in subinterval, if summation action time of all vectors is greater than pulse width modulation cycle in single sequence of operation, then adjusts action time each vector;

Step 4: the action time utilizing vector in the on-off action order and each sequence of operation obtained in step 3, in conjunction with mid point current i corresponding during each vector effect _k, the uneven function f (x of computational prediction _pred);

Step 5: according to unbalance voltage minimization principle, namely uneven valuation functions f (x) is minimum, selects suitable on-off action order;

Wherein, uneven valuation functions f (x)=f (x _pred)+f (x _act), f (x in formula _pred) be the uneven function of prediction, f (x _act) be actual uneven function; And wherein i in formula _kfor in selected sequence of operation, mid point current value during kth vector effect, t _kfor the time of this vector effect; F (x _act)=2C Δ u, in formula, C is DC bus capacitor value, and Δ u is DC side two capacitance voltage difference.

2. the method for claim 1, is characterized in that: be greater than pulse width modulation cycle to summation action time of vectors all in single sequence of operation in step 3, method of adjustment is for be adjusted to t'=t/t action time by each vector _pWM.

3. the method for claim 1, it is characterized in that: to the sequence of operation comprising middle vector and large vector in described step 4, select to add redundant vectors and action time thereof and the action time recalculating other vectors in sequence of operation, subsequently the uneven function f (x of computational prediction again _pred).