EP3387743A1

EP3387743A1 - Method for compensating the nonlinear effects of a voltage inverter

Info

Publication number: EP3387743A1
Application number: EP16819607.9A
Authority: EP
Inventors: Najib ROUHANA; Serge Loudot; Edouard Negre; Nicolas PATIN
Original assignee: Renault SAS
Current assignee: Renault SAS
Priority date: 2015-12-10
Filing date: 2016-12-05
Publication date: 2018-10-17
Also published as: WO2017098125A1; FR3045240B1; FR3045240A1

Abstract

The invention relates to a method for compensating the nonlinear effects produced by the addition of a dead time in a pulse-width modulation control of a voltage inverter, the method comprising: - a transforming step (231), during which a value of said transformed current (i) is calculated (i _d,i _q) in a plane of rotation (dq); - a step of determining the sign of the current flowing through one phase; - a step of estimating a total amplitude value of the compensation voltage adapted to attenuate the harmonics of order 2N of said transformed current value; - a step of calculating (232) said compensation voltage (Δv(k)); and - a compensation step (233) in which the reference voltages used for the pulse-width modulation control are modified, by adding or subtracting said compensation voltage (Δv(k)) depending on the determined sign of said phase currents (ia, ib, ic).

Description

Method of compensating the non-linear effects of a voltage inverter

The present invention relates to a method for compensating non-linear effects produced in particular by the addition of a dead time in a pulse width modulation control of a voltage inverter powered by a continuous bus and able to provide the at least one electrical phase for supplying an electric machine with N> 1 supply phase.

In the field of electrical engineering, an inverter is frequently used to power a rotating electrical machine, for example a machine with three-phase power supply. The voltage inverter receives a DC voltage and transforms it into an AC voltage, for example into three AC voltages for supplying the phases of an electric machine.

In known manner, a voltage inverter comprises a plurality of switching arms connected in parallel between two supply lines. These are connected to the same DC voltage source. Each arm has at least two power switches, in series with a connection midpoint connected to the electrical load.

Generally, the switching arms are controlled by so-called pulse width modulation strategies, known by their acronym MLI or its English equivalent, Puise Width Modulation, PWM.

One of the main problems encountered in the variable speed drive application of electrical machines, especially in the low voltage / low speed mode, is the generation of an error voltage. The latter, caused by the non-ideal characteristics of the components that make up the power module, is the difference between the desired ideal output voltage, referred to as the reference voltage, and the actual output voltage supplied by the inverter. Due to the non-linear character of the inverter, the actual output voltage is deviated from the reference voltage.

The most important part of this non-linearity comes from the fixed time value, called dead time, added between the complementary switching signals, to avoid the simultaneous conduction, in the same inverter arm, of the power switches. In addition to the dead time added, another source of non-linearity is derived from the inherent and non-ideal characteristics of switching devices such as the open / close time (t ₀ ff) as well as the voltage drop in the state. passing from these.

The duration of opening t _on corresponding to the response time of the power switch which is none other than the time which separates the order (from the control) of the ON state of a transistor and its switching effective; and the closing time t _off corresponding to the response time of the power switch which is none other than the time which separates the order (from the control) of the OFF state of a transistor and its switching .

The influence of this non-linearity on the output of the machine torque is particularly noticeable in the low voltage / low speed regime.

This imposes a precise compensation of the generated error voltage in order to improve the performances in low speed, in particular as regards the vibrations and the acoustic nuisances, and in order to avoid the problems of instability and to reduce the losses in the electric machine.

Several compensation methods are known. A first solution is to adopt a direct type approach through predefined measurement tables. In other words, the voltages are corrected with a predetermined correction value, which is a function of the operating conditions of the machine. However, this compensation is not effective enough, because, on the one hand, the inherent characteristics of the switching devices are variable, in particular as a function of the operating speed of the electric machine, the frequency, the DC bus voltage. and temperature, and secondly, the manual measurement of these different parameters is in itself a laborious and difficult process because these parameters vary from one inverter to another.

Also known from patent document JP2006074898A a compensation method consisting in providing a compensation voltage depending on the PWM switching frequency, the DC bus voltage and the sign of the current and the system parameters (dead time duration, duration opening and closing of the switches, duty cycle specific to the voltage drop of the power switches during a MLI period). However, this method requires an exact knowledge of the system parameters.

We therefore want to find a solution to reduce the impact of the non-linear effects of the inverter on the electric machine, effectively, independent of the PWM strategy used, and transferable to different machines without it being necessary to know their intrinsic parameters.

There is provided a method for compensating non-linear effects produced by the addition of a dead time in a pulse width modulation control of a voltage inverter fed by a continuous bus and capable of supplying at least one electrical phase for supplying an electric machine with N> 1 supply phase, in particular with 3 supply phases, said compensation method comprising a step of evaluating a compensation voltage, said evaluation step comprising:

a transformation step during which values of the phase currents transformed in a rotational plane, for example in the Park plane, are calculated;

a step of determining the signs of the phase currents as a function of said transformed current values;

a step of estimating a total amplitude value of the compensation voltage in the course of which, on the one hand, a fixed amplitude value is defined as a function of the amount of dead time added and the voltage of the continuous bus, and, on the other hand, a value representing the harmonics of rank 2N of said transformed phase current value is calculated, and an amplitude value is estimated as a function of said representative value of harmonics of rank 2N;

a step of calculating said compensation voltage as a function of the sign of the phase current and of said total amplitude value; and

a compensating step during which a reference voltage of the pulse width modulation control is modified by anticipatory action, also known as feed forward, by adding or subtracting said compensation voltage following the sign of said determined phase currents. Thus, the estimation step aims at determining a total amplitude of compensation voltage making it possible to reduce the harmonics of rank 2N of said transformed phase current value.

Advantageously and in a nonlimiting manner, the estimate of the representative value of the harmonics of rank 2N is performed by an adaptive bandpass filter.

Advantageously and in a nonlimiting manner, during said step of estimating a fixed amplitude value, a value strictly dependent on the DC bus voltage, the added dead time and the PWM period used is imposed.

Advantageously and in a nonlimiting manner, during said step of estimating a total amplitude value, one defines, on the one hand, a known fixed amplitude value, and on the other hand, an error is calculated cumulative, also called cumulative sum, of a plurality of representative values of harmonics of rank 2N calculated over a period of time corresponding to a fraction 1 / 2N of fundamental period.

Advantageously and in a nonlimiting manner, said fixed amplitude value is defined as a function of the amount of dead time added, and said variable amplitude value is estimated as a function of said cumulative error.

Advantageously and in a nonlimiting manner, said variable amplitude value is estimated as a function of the variations of a plurality of cumulative errors previously calculated.

Advantageously and in a nonlimiting manner, said variable amplitude value is estimated as a function of at least one previously calculated amplitude value.

Advantageously and in a nonlimiting manner, said total amplitude value is calculated as a function of said variable amplitude value and said fixed amplitude value.

Advantageously and in a nonlimiting manner, the step of calculating said compensation voltage comprises a multiplication of said estimated total amplitude value with the sign of the determined phase current. Advantageously and in a nonlimiting manner, during the step of determining the sign of the phase current, the high frequencies of said transformed current value are filtered.

Advantageously and in a nonlimiting manner, said anticipative action compensation voltage, also known under the English name of "feed-forward", is injected to modify the reference voltages used for controlling the PWM command.

Advantageously and in a nonlimiting manner, the method is implemented for each of the N phases of said electric machine.

Other features and advantages of the invention will emerge on reading the following description of a particular embodiment of the invention, given by way of indication but not limitation, with reference to the accompanying drawings in which:

FIG. 1 is a diagram of a step of evaluating a compensation voltage of the non-linear effects of an inverter according to one embodiment of the invention;

FIG. 2 is a diagram detailing the step of determining the sign of the process current according to the embodiment of FIG. 1;

FIG. 3 is a diagram detailing the step of estimating a total amplitude value of the compensation voltage of the method according to the embodiment of FIG. 1;

FIGS. 4a and 4b are representations of bandpass filters for filtering the rank six harmonics of the phase current in a rotational plane according to the embodiment of FIG. 1;

FIG. 5 is a logic diagram showing a step of minimizing rank six harmonics of the phase current in a rotational plane according to the embodiment of FIG. 1;

FIG. 6 is a representation of a method for filtering rank six harmonics of the phase current in a rotational plane according to one embodiment of the invention.

FIG. 7 is a global representation of anticipatory action compensation. With reference to FIG. 1 and FIG. 7, a compensation method, not shown, of the non-linear effects produced by the addition of a dead time 220 in a pulse width modulation control 210 of a voltage inverter 100 fed by a continuous bus and capable of supplying at least one electrical phase for supplying an electric machine with N> 1 supply phase, comprises a step 230 of evaluating a compensating voltage A nk) effects nonlinear of the inverter 100, especially during the switching phases.

Evaluation step 230 comprises a step of determining b1 of the sign of each of the phase currents, an estimation step b2 of a value of total amplitude v4 (v) of the compensation voltage v {k) and a step 232 of calculating said compensation voltage Av {k).

In this embodiment, the method adapted for an electric machine powered by three electrical phases is described, but the person skilled in the art could adapt this method to any other number of phases adapted to power an electric machine.

Steps b1 and b2 are performed in parallel and jointly make it possible to obtain the compensation voltage at v {k).

Before each step b1 and b2, during each transformation step 231, each of the phase currents i is transformed into a rotational plane

(dq)

The transformation of a current i in the rotational plane ^) is carried out by a transformation of Park of an angle 0 _eIec , which corresponds to the mathematical tool allowing to realize a change of reference where one passes from a plane stationary to a rotating or rotational plane.

The term "stationary plane ( ^ace )" means the reference linked to the stator of the electric machine.

The term "rotational plane" means the Park plane where a three-phase system is modeled by means of a two-phase system whose two components, one which is along the axis d and the other along the axis q, do not depend on of the electric angle 0 _eIec . This rotating mark is connected to the rotor or to the magnetic field in the gap of the electric machine. The transformation of a phase current i in the rotational plane ^dq by the Park transform produces two signals ^ld and ^lq each being a component of the phase current i in the plane (dq).

Step b1, with reference to FIG. 2, makes a determination of the sign of each of the phase phase currents i of the machine.

With the aid of a low-pass filter b1 1, the high-frequency components of the signals ^ld and ^lq, which are due to the switching phenomena of the power electronics components in each arm of the inverter 100, are eliminated while at the same time impacting the phase currents ia, ib, ic. Thus, filtered ^ld and ^lq signals b12 corresponding to the phase currents in the Park plane are obtained.

The filtered phase currents ia, ib, ic b14 are then reconstructed as a function of the filtered signals ^ld and ^lq b12 using the inverse Park transform b13.

Then, in a step b15, the sign of the phase currents ia, ib, ic is determined as a function of the reconstructed currents b14.

In parallel, the step of determining the compensating voltage amplitude b2 is carried out, with reference to FIG. 3, which makes it possible to determine the total amplitude v (k) of the compensation voltage Av (k). according to the currents in the rotational plane ^d \

In a first step b21 of the determining step b2 of the total amplitude F (fe) of the compensation voltage the harmonic of rank six is estimated for each of the two components, d and q of the current vector in the rotational plane ^).

Indeed, the harmonic of rank six, which appears in the plane ( ^dq reflects the existence of undesirable nonlinear effects, and thus the presence of an error which must be compensated until minimization of the harmonic of rank six. To this end, the filtered signals ^ld and ^lq in the plane of Park through an adaptive band pass filter, so as to obtain _6d h and h _6q filtered signals b22 corresponding to harmonics of order six phase currents in the plan of Park.

Figs. 4a and 4b each show an embodiment of the adaptive bandpass filter so that its cut-off pulse adapts to changes in the rotor speed of the electric machine.

In particular, the band-pass filter is represented in FIG. 4a and is defined by a transfer function of the form:

I - F (z)

"> = - ^

With:

In which k 2 and k 1 are filter adjustment coefficients well known to those skilled in the art.

Then, a straightening step b23 is performed for the filtered signals h _6d and h _6q b22. In other words, a calculation operation of an absolute value is applied for each of the filtered signals h _6d and h _6q b22.

The _corrected h _6d and h _6q signals b24 will subsequently be called error b24. Indeed, these rectified signals b24 correspond to the error associated with the non-linear effects that we wish to compensate, in particular to the non-linearity introduced by the inverter 100.

Once the calculated signals h _6d and h _{6q have been} recalculated, a step of accumulating the error b25, called the cumulative error b25, in which the rectified signal b24 at time k is integrated over a corresponding duration is implemented. to a sixth fundamental period for a three-phase network, or 1 / 2N for an N-phase network.

Thus, a cumulated sum b25 of the rectified signal values b24 is calculated so as to obtain the total error b26.

In Park's plan, there are two harmonic components of rank six, one along the axis d and the other along the axis q. Also, we calculate in practice two cumulative errors: and Their sum resulting in a total error b26.

In order to achieve this objective, we accumulate the signals h _6d and h _6q rectified b24 every 1/6 of fundamental period, for a three-phase network, this allows us to be in phase with the harmonic of rank six.

The cumulative error b26 is reset to zero at each 1/6 fundamental period, by a trigger signal, called trigger signal, issued by one of the computers of the system.

When the trigger signal is received, the final value is recovered which corresponds to the fixed value b28 of the accumulated error in the previous period.

This fixed value b28 is stored in a buffer memory b27 so as to keep it constant for the entire duration of 1/6 of the next fundamental period.

A step of minimizing b29 of the error b28 is carried out, as a function of the accumulated fixed error value b28, making it possible to obtain the unknown and variable amplitude b30, _sr (k) which is the first component of the voltage compensation Av (k).

With reference to FIG. 5, an advantageous embodiment of a minimization step b29 of the error b28 is described.

However, a person skilled in the art could not limit himself to this way of performing this step of minimizing b29 of the error b28.

Since the method is implemented by a computer, by nature discrete, we mean by the term of time, or "step", the successive cycles of implementation of the method by the computer.

The step of minimizing b29 of the error b28 comprises an initial step b292 of supplying the values of the fixed error b28 to the present step (k), of the amplitude _ar {k - 1} b299 at the preceding step (k- 1) and a voltage adjustment value b291 defining the speed of convergence of the minimization step b29 of the error b28. At the first execution, if no amplitude value of the compensation voltage b299 at the previous step (k-1) is available, in other words, at k = 0, the variable concerned is initialized to zero.

The voltage adjustment value b291 is imposed by those skilled in the art. It can be of a fixed nature or of an adaptive nature. The higher this value b291 is, the more the method will make it possible to obtain a fast convergence, but by reducing its precision, and on the contrary the smaller this value, the more the process will allow a slow but precise convergence.

Then, we evaluate b293 the evolution of the error b28 between the present step (k) and the previous step (k-1). In other words, it is estimated whether the error increases or decreases.

The sign of the evolution of the compensation voltage amplitude b299 applied with respect to the amplitude at the previous step (k-1) is determined b294, b295. In other words, it is estimated whether the amplitude of the compensation voltage b299 at the previous steps increases or decreases.

Figure 5 shows two different logic blocks b294 and b295 but these refer to the same process step, separated in two for better understanding.

After these two steps b293, b294, b295, it is determined whether to increase or decrease the amplitude at the present pitch, as a function of the evolution of the sign of the error b293 and as a function of the sign of the evolution of the amplitude of the compensation voltage.

If the sign of the evolution of the error is positive and if the sign of the evolution of the amplitude of tension is positive, then the fixed error b28 increases.

A step of decrementing 298 of the voltage amplitude 234 is then carried out by subtracting from the amplitude value v ^ ik) at the previous step b299, the adjustment value & b291.

If the sign of the evolution of the error is positive and if the sign of the evolution of the amplitude of tension is negative, then the error b28 increases.

An incrementation step b297 of the voltage amplitude 234 is then carried out by adding to the magnitude value λ _w (fc) at previous step b299 the adjustment value b291. If the sign of the evolution of the error is negative and if the sign of the evolution of the amplitude of tension is positive then the error b28 decreases.

An incrementation step b297 of the voltage amplitude 234 is then carried out by adding the adjustment value b291 to the amplitude value _ar (k) at the previous step b299.

If the sign of the evolution of the error is negative and if the sign of the evolution of the amplitude of tension is negative then the error b28 decreases.

A step of decrementing b296 of the voltage amplitude 234 is then carried out by subtracting from the amplitude value ¾ _ffir (fc) at the previous step b299 the adjustment value & b291.

Thus we obtain at a time k a value of amplitude ¾ _a (&) at the present step to apply.

In a second step b34, the amplitude is added with a fixed amplitude value v _fix . _e (k to obtain a total amplitude value v (k).

The fixed amplitude value is calculated based on the amount of dead time added. In particular it can be determined by the following formula: v _fixed {k) = td ^* Vdc / T_MLI, in which td is the amount of dead time added, Vdc the voltage b31 of the DC bus, and T_MLI the period of the modulation signal pulse width.

In particular, it is possible to first calculate the ratio b32 td / T_MLI and then multiply it at the voltage of the DC bus b31.

Then, a calculation step 232 of the compensation voltage Av (k) is carried out at time k, during which the product is calculated between the current sign value i obtained in step b1 with the value of total amplitude v (k) determined in step b2 at the same instant.

Finally, with reference to FIG. 7, during a compensation step 233, said compensation voltage is injected at -u (k) by anticipatory action, also known by the English term "feed-forward", so that to modify the amplitude of the reference voltages 211 used for the control of the strategy, or command, MLI 210. Thus we can compensate for most of the error in open loop.

According to a second embodiment of the invention, with reference to FIG. 6, the harmonics in the Park plane are estimated with an angle of rotation ( ^{± 1} ).

Instead of using an adaptive bandpass filter to estimate the rank six harmonics in the rotational plane (1), as previously described with reference to FIGS. 4a and 4b, the principle of the planes which turn to FIG. EIEC ₀ ^ ù k = l, 2, ... _with \ _es sequences

, -50, +70, -110, +130, suitable as for example: ^elec , ^elec , ^elec , ^elec , etc.

The skilled person being able to implement this principle with his general knowledge.

Each of these transforms converts the harmonic of rank ( ^{6k ± l} ) _! which is an alternating component (AC) in the stationary plane ( ^abc into a continuous component (DC) in its own rotational plane.

For example, the transition from the stationary plane to the park plan clean

-50

at harmonic 5, at an electric angle equal to ^elec , transforms the harmonic of rank 5 into a continuous component, known by its abbreviation DC. The fundamental becomes a harmonic component of rank 6 and the harmonic 7 in the plane (abc) becomes a harmonic component of rank 12 in the work plane.

70

The same principle applies for a rotation of ^elec . The harmonic 7 becomes a DC component, the fundamental will be a harmonic component of rank 6 and the harmonic 5 becomes a harmonic component of rank 12.

According to a third embodiment of the invention, the harmonic 5 is directly estimated in the real plane. Then, using a suitable Park transformation, the rms value is calculated in the appropriate rotational plane. Thus, one works directly in the stationary plane ( ^abc ) or in the fixed plane (αβ). So just with a filter adaptive bandpass as described above with reference to FIGS. 4a and 4b, to estimate the harmonic 5. However, the following adaptations must be made:

- Rank 5 harmonic is estimated in the real or stationary plane, using an adaptive bandpass filter set at 5w _eIec electrical speed

- Once the five harmonic components are estimated, a plane transformation with an electrical rotation of

-5Θ

e ^lec . From there, the new components h _Sd and h _Sq will be DC components.

- The determination of the rms value of the resulting current will be possible by the application of / RMS,

- The "trigger" signal is sent every 1/5 of fundamental period, instead of every 1/6 of fundamental period.

According to a fourth embodiment of the invention, a digital integrator is implemented which acts on an error equal to the difference between a reference, denoted s _ref and fixed by the skilled person, and the estimated signal b26. The difficulty lies in the fact that the signal b26 does not change sign and the reference must be changed and adapted to each operating point, provided that the impact of the non-linear effects varies inversely proportional to the variation of the signal. speed (which is the picture of the tension).

On the other hand, in practice it generally exists an error, as a minimum threshold is fixed, denoted _re ^£ f _min> such that once the error frozen b28 exceeds this threshold, the integration is inhibited and the last value voltage amplitude is preserved.

The step of incrementation is defined so that it is of very fine value, which the skilled person knows how to choose.

In this way, a rapid convergence of the transient algorithm can be ensured while maintaining an acceptable accuracy of the amplitude of the steady-state compensation voltage.

Claims

1. A method of compensating nonlinear effects produced by the addition of a dead time in a pulse width modulation control of a voltage inverter (100) fed by a continuous bus and capable of supplying at least one electrical phase for supplying an electric machine with N> 1 supply phase, said compensation method comprising a step of evaluating (230) a compensation voltage (& v (k)), said evaluating step (230) comprising:

- a transformation step (231) during which one calculates values of the phase currents (ia, ib, ic) transformed ^d , ^lq ), in a rotational plane ^dq );

a step of determining (b1) the signs of the phase currents (ia, ib, ic) as a function of said transformed current values ( ^ld , ^lq ),

- an estimation step (b2) with a value of total amplitude (V {k)) of the compensation voltage in which is defined the one hand, a fixed value of amplitude _that {v _f depending on the amount of dead time added and the DC bus voltage, and on the other hand, a value representing the rank 2N harmonics of said transformed phase current value is calculated, and a value of variable amplitude (v _var ) according to said representative value of harmonics of rank 2N;

a step of calculating (232) said compensation voltage (Δ¾ &)) as a function of the sign of the phase current (i) and of said total amplitude value ((fe)); and

a compensating step (233) in which a reference voltage (21 1) of the pulse width modulation control is modified by anticipatory action, by adding or subtracting said compensation voltage k}) along the sign of said determined phase currents (ia, ib, ic) (b1).

2. compensation method according to claim 1, characterized in that the calculation of the representative value of harmonics of rank 2N is performed by an adaptive bandpass filter.

3. compensation method according to claim 1 or 2, characterized in that during said step of estimating (b2) a total amplitude value ((fe)), is defined, on the one hand, a fixed fixed amplitude value (V _fixs ), and on the other hand, a cumulative error (b25) of a plurality of representative values of the harmonics of rank 2N calculated over a period of time corresponding to one fraction 1 / 2N of fundamental period.

4. Compensation method according to claim 3, characterized in that said variable amplitude value (V _var) is estimated as a function of said cumulative error (b25).

5. Compensation method according to claim 3 or 4, characterized in that said variable amplitude value (¾, _Gr ) is estimated as a function of the variations of a plurality of accumulated errors (b25) previously calculated.

6. compensation method according to any one of claims 1 to 5, characterized in that said variable amplitude value (t ^ _fl (fc)) is estimated as a function of at least one variable amplitude value ( ¾ _Er (fe - I)) previously calculated.

7. Compensation method according to any one of claims 1 to 6, characterized in that calculates (b34) said total amplitude value ø? {. ¼}) as a function of said variable amplitude value (¾ _¾r (.¾)) and said fixed amplitude value (ï _e (fc)).

8 compensation method according to any one of claims 1 to 7, characterized in that the step of calculating (232) said compensation voltage (-av (_k)) comprises a multiplication of said value of total amplitude (V (k)) estimated with the sign of the determined phase current (i).

9 compensation method according to any one of claims 1 to 8, characterized in that during the step of determining (b1) the sign of the phase current (i), the high frequencies of said value of the transformed current vector ( ^ld , ^lq ).

10. Compensation method according to any one of claims 1 to 9, characterized in that it is implemented for each of the N phases of said electric machine.