DE19640591C1 - Torque regulation method for induction machine - Google Patents

Abstract

The torque regulation method determines the actual flux value and the sector of the complex flux plane in which the flux vector lies. The flux amplitude is compared with a required value and the calculated machine torque is compared with a required torque, for calculating a voltage vector for reducing the disparities between the compared values. The flux characteristic during the switching in of the machine with the calculated voltage vector is used for providing a zero vector, used for the remainder of the remainder of the cycle time .

Description

The invention relates to a method for flow and torque control Rotary field machine according to claim 1.

Task

The flow and the torque of a three-phase machine are supposed to pass through with a defined cycle time direct torque control can be controlled independently. The middle turn torque should correspond to the torque setpoint as quickly as possible. Thereby the average switching frequency is approximately constant at that permitted by the pulse-controlled inverter (PWR) be kept in order to keep the torque ripple low. This is particularly the case with highly dynamic drives are desirable if the converter and machine losses a play a subordinate role. In order to keep the expenditure on analog peripheral components low hold and to be able to carry out automatic commissioning and parameter adjustments, is an implementation as a technical program in a microcomputer with as much as possible little peripheral for analog signal processing advantageous.

State of the art

The invention relates to a method and a device for controlling a rotation Field machine according to the preamble of claim 1.

For feeding induction machines whose momentum is in a wide speed range can be set highly dynamic to desired values, PWR are usually with constant on DC voltage used. For a machine with three strands (u, v, w) and two Switches per line result in eight different permissible switching states. these can after transformation into a two-strand system (a, b) are represented as space pointers. Six of these voltage space pointers (SR) span a regular hexagon on the two others are zero vectors (NV), and therefore represent a terminal short circuit of the machine. To control the machine, these eight possible switching states of the PWR must alternate be selected so that the desired operating behavior results. Procedures that To meet highly dynamic requirements, determine with the magnetic flux in the machine. A method for regulating the torque is then based on this. The Differentiation between flux-forming and torque-forming quantities was made by [6] and [1] introduced.

Known procedures

The numerous processes can be divided into three large groups:

1. Method with modulators

This group includes the classic field-oriented control in analog Technology according to [6], [1] or as an implementation with a microcomputer [5]. This procedure determine from measured currents, terminal voltages, rotor position and speed the rotor flow space pointer. Its angular position is used for transformation into a field-oriented Coordinate system (x, y) used. The torque results from the perpendicular to the Rotor flux lying stator current component and the amount of flow. Through separate river and Torque controllers are the setpoints for flux-forming and torque-forming current components determined. From these, the - mostly linear - current regulators form a target Stress space pointer, characterized by the amount and angle of stress. This SR will then from a modulator, e.g. B. pulse width modulator (PWM), by alternating Switching of the eight possible switching states of the PWR simulated. Pulse width modulators  typically work with a constant period. This will be unique after the permissible limits of the converter or other technical criteria such as B. the losses fixed.

The linear current regulator with a downstream modulator can also be caused by current hysteria reseregler to be replaced.

2. Direct procedure

Direct processes generally work in a flow-oriented manner. Such a method and such a device is known from [4]. The from measured currents and terminals Values for flux and torque are determined by switching regulators (Two / three-point controller, with / without hysteresis) compared with the setpoints. With the logical Outputs of the comparators determine the next switching state directly via switching tables placed. These procedures include both direct self-regulation [3] and that Takahashi and Noguchi method [11].

A special form of direct procedure is [7]. From a combination of Weighting functions ("fuzzy logic") for flux and torque become switching states selected and their switching times determined proportionally over a period. This method is based on the preamble of claim 1.

3. Predictive procedures

In this method, a tolerance range is specified for flux and torque. As The next switching state is the one selected that has the maximum dwell time in the pre given tolerance range. This is based on the current state of the machine a prediction of the current curve is necessary until the tolerance limits are reached (e.g. [7], [10], overview in [9]).

Discussion of the procedures listed 1. Method with modulators

These methods typically require a lot of signal processing Field orientation. The processes for rotor flux monitoring are reasonable also very sensitive to parameters [12]. The field-oriented control is sensitive Orientation error. The method is widely used because a Mo allows relatively high switching frequencies with reasonable computing effort and the related benefits such as. B. has low torque pulsations.

With a setpoint for one period of the PWM, the next - usually up to six - are switching actions set. This results in a long dead time in relation to the switching frequency, until the next setpoint can be taken into account.

A current hysteresis controller can react faster to setpoint jumps than linear current controllers with downstream PWM. However, while the PWM usually has a con Constant switching frequency, this can depend on a current hysteresis controller Operating condition fluctuate greatly. To take advantage of the switching frequency permitted by the PWR, the hysteresis width must be adjusted so that the desired average switching fre quenz results.

Since the modulator does not directly involve the flux and torque controller in either case, the resulting switching cycles are usually not optimal. The controllers can do not take into account the switching character of the PWR with fixed SR.  

2. Direct procedure

Each switching operation is based on the current operating state and the setpoints for flow and Torque derived individually. Hysteresis comparators are used for this. With analog Kompara Setpoint jumps have an immediate effect on the comparator results and the results thereof determined switching state. This results in low dead time and high dynamics. The are contrasted by the typical disadvantages of analog circuit technology. Commissioning and Adjusting parameters is complex. There are complex controller structures and computing functions can only be realized with digital systems. As with the current hysteresis controllers also hangs here the switching frequency is very dependent on the operating state. That is why, to an approximation To achieve a constant switching frequency, it is necessary to adjust the hysteresis width.

In the case of implementations on microcomputer systems, too, the fast Passing through the controller input to the switching action a high dynamic. The advantage of Flexibility is offset by the disadvantage of a longer dead time than in analog implementations about. Especially for machines that are designed for highly dynamic drives and therefore have low leakage inductances, the current and thus the rotation instantaneous rates of rise are so high that the hysteresis bands are unacceptably large must be designed to be adhered to within a computing cycle. This problem caused by the computing time can be solved by more powerful microprocessors or reduce signal processors. The effort involved does not appear to be relatively high and expensive.

In the method according to [7], the weighting function connects the flow controller with the torque regulator. Both the stationary and the dynamic behavior of the torque control are linked to the flow control. An SR selected by the flow control can higher torque fluctuations than z. B. in [11], [3], where also the tolerance bandwidth is set. The multiple weighting (flux, torque, duty cycle) is also connected with considerable computing effort.

3. Predictive procedures

Predictive methods require a relatively high signal processing effort in the micro calculator, especially if the flux and torque control in rotor flux-oriented Coordinates. The length of stay in the tolerance area can vary greatly. A Be drive with an approximately constant switching frequency is therefore associated with additional effort. Depending on The operating status can be very short. The microcomputer must therefore be fast be. The selection of a chip that appears favorable for the length of stay in the tolerance zone space pointer is not necessarily the best choice for the overall control behavior.

The method according to the invention Basic principle

Various observers are known in [12] with which the stator flow of the machine can be determined. For the method according to the invention, it is easiest to start from a description of the machine in fixed coordinates:

With an integration process, e.g. B. the second-order Runge-Kutta method according to [2], this system of differential equations can be solved by integration. Depending on the requirements Implementation is also important in terms of the accuracy and nature of the drive system of a state observer is conceivable (e.g. according to [12] by transforming the equations, see that the flows occur as state variables). The determination is essential for the procedure the derivatives of the rivers (as in Eq. 1) or in the following the torque.

The torque is calculated

The stator flux is intended on a circular path with a radius of ψ _to be maintained. The flow change can be described by the equation:

_s = ∫ (uR _{s s} ) dt (3)

If one neglects the voltage drop across the stator resistance, this is simplified Equation to:

_s = ∫ dt (4)

Fig. 1 illustrates the output voltages provided by the inverter in fixed coordinates (a, b): six SR ({1} to {6}) and two NV ({0}, {7}).

In FIG. 2, the desired circular flux flow and an example of a momentary value of the flux is located. Depending on the position of the river area pointer of the machine, you can assign the SR properties to be selected. The available SR drive the river space pointer according to Eq. 4 in different directions. A zero space pointer changes the flow only slightly. The component of the SR running tangential to the circle influences the torque, the radial component influences the flow. This results in the assignment given in FIG. 2.

The circular path of the river area pointer can be divided into sectors, e.g. B. six as in Fig. 2. The sector in which the flow space pointer is currently located is obtained by comparing the two flow components ψ _a and ψ _b by means of comparators. An angle calculation with the arc tangent function and subsequent sector assignment with the modulo function is also conceivable.

The control deviation of the torque is also fed to a comparator. At medium and high speeds it makes sense to use the sign of the mechanical  To use angular velocity instead of the comparator output. In this case, large torque deviations that cannot be corrected in a control cycle the comparator output is inverted to select highly torque changing SR. The control deviation of the flow is fed to a flow comparator.

The SR is selected depending on the comparator results (A ₀ , A ₁ ) according to the following table:

Selection of the voltage space pointer

Selection of the voltage space pointer

The method is based on the idea that normally an SR and an NV can always be switched alternately in order to keep the mean value of the torque equal to the target value. The switching operations are each calculated for a cycle time Δt _cycle . Normally, exactly one SR and exactly one NV, i.e. two switching states, are present over a cycle time.

Fig. 3 illustrates the process of a computing cycle. The duty cycles .DELTA.t _{SR of} the SR and .DELTA.t _{NV of} the NV are determined so that the torque average corresponds to the setpoint over the cycle time. This is exactly the case in FIG. 3 if the oppositely hatched areas are of the same size. The flow control is carried out by two-point control by z from cycle to cycle. B. alternately flow-building and river-breaking torque-forming SR can be selected.

A sufficiently short cycle time allows a linear approximation of the torque curve for the averaging. The mean torque value is calculated from:

M _cycle = (M _SR .Δt _SR + M _NV . (Δt _cycle -Δt _SR )) / Δt _cycle (5)

One obtains from the linear approximation

M _SR = (M (t _n ) + M (t _n + Δ _SR )) / 2 (6)

and

M _NV = (M (t _n + Δt _SR ) + M (t _n + Δt _cycle )) / 2 (7)

The torque values M (t _n + Δt _SR ) and M (t _n + Δt _cycle ) are most easily determined by linear prediction. M _SR is obtained at the switching time

M (t _n + Δt _SR ) = M (t _n ) + _SR .Δt _SR (8)

and at the end of the _cycle M (t _n + Δt _cycle ) = M (t _n ) + _SR .Δt _SR + _NV . (Δt _cycle -Δt _SR ) (9)

The derivatives of the rivers must be used anyway to integrate the system of differential equations according to Eq. 1 can be calculated. From this, the derivative of the torque can be calculated using the product rule:

With Eq. 10 one calculates M (t _n +) and sets it for M _SR in Eq. 8 a. Then you set Eq. 8 and 9 in Eq. 6 and 7 and these in turn in Eq. 5 a. The target value M target _is aimed for as the mean value for M. Resolved according to the switch-on time of the SR and only allowing times Δt _SR <Δt _cycle , you get:

When switching on an NV, the derivative of the torque _NV changes only slightly with a short cycle time. Therefore, one can simply calculate with the flows calculated in the last cycle and their derivatives (t _n -) and for _NV in Eq. Insert 11. Another possibility is to determine _NV from the torque values at the time the last NV was switched on and off, as well as its duty cycle.

The NV to be switched on after the SR (or in exceptional cases NV) is selected so that as few inverter branches as possible have to be switched. The NV thus results depending on the previous switching status according to the following table:

Assignment of the zero vectors to the previous switching state

Assignment of the zero vectors to the previous switching state

Exception handling

The procedure is only waived if the switch-on time of the SR or the NV becomes impermissibly short on switching and has a switching status applied over the entire cycle time.

At certain operating points, the duty cycle based on the mean torque value the river-forming SR become so short over several cycles that the one selected according to Table 1 ten SR did not adequately build the river weakened by the NV. If the river falling below a critical limit, the selection of a strong as special treatment flux-building vector required according to the following table:  

Strong flow-supporting voltage space pointer

Strong flow-supporting voltage space pointer

By accepting short torque peaks, Δt _{SR can be determined} for the strongly flow-supporting SR so that the flow amount at t _n + Δt _{SR reaches} the setpoint:

The flow amount is derived using the chain rule:

This exception can also be done by switching on two SRs and one NV, i.e. three Switching states within a cycle, are treated to continue the torque to keep the mean exactly. If the switches of the selected SR in different PWR strands occur, the increase in switching frequency is not critical for the converter, because the currents in this exceptional case are clearly below the nominal current.

Advantages of the invention

The method according to the invention combines a number of advantages of the methods (1, 2, 3) listed above, without having their disadvantages with regard to the objective:

• a) By defining a cycle time for two successive switching operations, the Switching frequency kept constant. It is also possible to set the cycle time to certain Adapt operating states to z. B. a certain total switching frequency.
• b) The defined cycle time allows a simple design of time-equidistant working digital controller at a higher level, e.g. B. flux and torque setpoint determination. The dead time is easy to calculate.
• c) The hysteresis which determines the switching frequency in direct processes and their aftermath guidance can be dispensed with because the switching frequency is determined by the cycle time becomes.
• d) By predicting the torque curve, model calculation of the machine and more explicitly Calculation of the switching times are no additional analog comparators and Regulator required.
• e) The calculation of two successive switching states (SR and NV) enables one of them very short (only slightly longer than the interrupt latency of the microphone computer or minimum switch-on time of the semiconductors).
• f) The inclusion of the machine model calculation or an observer in the prediction saves computing time because the model sizes are not only used to determine the operation state, but also be used in the controller.
• g) In particular, the properties under a), b), c) and e) predestine the process for Implementation as a technical program in a microcomputer.

Embodiment Overview

The method is based on an implementation as a technical program in a microcomputer system designed with appropriate peripherals.

Fig. 4 shows the structure of the overall arrangement with motor, inverter, data acquisition and control process in the microcomputer. The function blocks (block) ( 1 ), ( 2 ), ( 4 ), ( 5 ) and ( 6 ) show devices which correspond to the prior art. Claim 1 relates to the devices in blocks ( 3 ) and ( 7 ). Claim 2 relates to the device in block ( 8 ).

Fig. 5 shows a flow chart of the microcomputer implemented technical program for realizing the embodiment.

Special features of the implementation of the exemplary embodiment

In the exemplary embodiment, as indicated in FIG. 4, the measurement variables were not fed back because the parameters of the system are known with sufficient accuracy to demonstrate the function of the method with the model implemented in block ( 1 ). To determine the mechanical angular velocity, you can use Eq. 1 add the motion differential equation:

procedure

The control routine is implemented as an interrupt program in a signal processor. The control interrupt is released after an initialization procedure. The controller program first magnetizes the machine to the nominal flow. This is done _to M = 0. Subsequently, for detecting the function of the method at different rotational speeds and torque requirements are decelerated. In the simplest case, speed control is possible in a background loop by using the model speed. Fig. 6 shows the results of a test run.

List of figures

Fig. 1 voltage space pointer provided by the inverter

Fig. 2 Flußortskurve desired, for example, effect of the voltage space vector,

Fig. 3 linear approximated torque curve over a cycle,

Fig. 4 structure diagram of the embodiment,

Fig. 5 flowchart of the embodiment;

Fig. 6 trial run of the embodiment.

bibliography

[1] Blaschke, F .: The field orientation process for controlling the induction machine, Dissertation Braunschweig 1974.

[2] Bronstein, I.N., Semendjajew, K.A .: Taschenbuch der Mathematik, 20th edition, publisher Harri Deutsch 1981.

[3] Depenbrock, M .: Method and device for controlling a three-phase machine, Publication EP 0 179 356 B1.

[4] Forsell, H .: power converter drive, published application DE-A 23 18 602, April 13, 1973.

[5] Gabriel, R .: Field-oriented control of an asynchronous machine with a microcomputer, Dissertation Braunschweig 1982.

[6] Hasse, K .: On the dynamics of speed-controlled drives with converter-fed Asynchronous squirrel-cage machines, dissertation Darmstadt 1969.

[7] Hofmann, Wilfried, Krause Michael: Procedure for self-regulation of an inverter fed three-phase machine, published application DE-42 25 397 A1, July 29, 1992.

[8] Holtz, J, Stadtfeld, S .: A Predictive Controller for the Stator Current Vector of AC Machines Fed from a Switched Voltage Source, IPEC Tokyo 1983, pp. 1665-1675.

[9] Leonhard, W .: 30 Years Space Vectors, 20 Years Field Orientation, 10 Years Digital Signal Processing with Controlled AC-Drives, a Review, Part 1 & 2, EPE Journal, Vol. 1, no. 1 & 2, Jul. & Oct. 1991.

[10] Stadtfeld, S .: The method for controlling the pulse converter-fed asynchronous machine through prediction and optimization in real time, dissertation Wuppertal 1987.

[11] Takahashi, I., and Noguchi, T .: "A New Quick-Response and High-Efficiency Control Strategy of an Induction Motor ", IEEE Transactions of Industry Applications, Vol. IA- 22, No. 5, Sept./Oct. 1986).

[12] Zägelein, W .: Speed control of the asynchronous motor using an ob eight with low parameter sensitivity, dissertation Erlangen 1984.

Claims (3)

1. Method for controlling a three-phase machine that is fed with an impressed DC voltage (U-) via a pulse inverter (PWR), with the following steps:

• a) Forming the actual flow value
• b) Subdivision of the complex river level into sectors and determination of the sector in which the river area pointer lies.
• c) Comparison of the amount of the actual flow value with the desired flow value
• d) Calculation of the machine torque
• e) Comparison of the actual torque value with the torque setpoint
• f) Determination of a voltage space vector which is suitable for reducing the deviations calculated under steps c) and e)
• g) Define a cycle time
  characterized in that
• h) the curve of the torque when the voltage space vector determined under step f) and a zero vector is switched on is predicted in order to derive from it
• i) to determine the duty cycle of the voltage space vector determined in step f) in such a way that when the zero vector is switched on, the torque mean value for the rest of the cycle time is equal to the torque setpoint over the entire cycle time.

2. The method according to claim 1, characterized in that in the case of a strong flow target value deviation as a result of claim 1, step c) a flow regulating voltage space pointer is preferred over the result from claim 1, step f). 3. The method according to claim 1, characterized in that the cycle time is determined so that

• a) the permissible total switching frequency of the converter is generally used
  and
• b) the permissible switching frequency of a single strand is not exceeded at low speeds.