## AC machine control: Robust and sensorless control by parameter independency

##### Doctoral thesis

##### View/Open

##### Date

2009##### Metadata

Show full item record##### Collections

- Institutt for elkraftteknikk [1642]

##### Abstract

In this thesis it is first presented how robust control can be used to give AC motor drive systems competitive dynamic performance under parameter variations. These variations are common to all AC machines, and are a result of temperature change in the machine, and imperfect machine models. This robust control is, however, dependent on sensor operation in the sense that the rotor position is needed in the control loop. Elimination of this control loop has been for many years, and still is, a main research area of AC machines control systems.
An integrated PWM modulator and sampler unit has been developed and tested. The sampler unit is able to give current and voltage measurements with a reduced noise component. It is further used to give the true derivative of currents and voltages in the machine and the power converter, as an average over a PWM period, and as separate values for all states of the power converter. In this way, it can give measurements of the currents as well as the derivative of the currents, at the start and at the end of a single power inverter state. This gave a large degree of freedom in parameter and state identi_cation during uninterrupted operation of the induction machine. The special measurement scheme of the system achieved three main goals:By avoiding the time frame where the transistors commutate and the noise in the measurement of the current is large, filtering of the current measurement is no longer needed.The true derivative of the current in the machine is can be measured with far less noise components. This was extended to give any separate derivative in all three switching states of the power converter.Using the computational resources of the FPGA, more advanced information was supplied to the control system, in order to facilitate sensorless operation, with low computational demands on the DSP.
As shown in the papers, this extra information was first used to estimate some of the states of the machine, in some or all of the speed regions of the machine. The information was then combined to increase the dynamic performance of the sensorless operation of induction machines using the DTC algorithm, even under temperature variations.
The scope of this thesis is to develop and test strategies for improving the performance of motor drive systems, when subjected to parameter variations in the machine. First, this is performed by modifying the controller towards a more robust controller, while later a special sampler is developed, in order to estimate machine parameters on-line. This development is shown in the following papers: In the first paper, a complete motor drive system is built, and an H1 current controller is implemented, instead of a previously designed PI controller with decoupling. The results show that this controller is able to perform similar to a classical PI-controller, even when subjected to parameter variations, at the cost of increased computational demands.The second paper presents a form of robust decoupling for a PI-controller as an alternative to the higher-order H1 controller from the first paper. Although there is no speed input to the decoupling network, rotor position feedback from a resolver is still needed for the FOC to work correctly.The special sampler is introduced in the third paper. Here the sampler is used to estimate the rotor ux angle, based on measurements of the derivative of the machine currents in speci_c parts of the PWM-period. The estimator shows good performance. The estimation principle is based on measurements during the zero-period of the inverter, though. This gives poor performance in the upper speed region, when the zero-period1 of the inverter is small or non-existent.In the fourth paper, the zero-state as well as the two active states of the inverter are used to estimate the rotor speed in a machine. This gives the opportunity of estimating machine parameters in the whole speed region, except around standstill, and the results show good performance, both static and dynamic.A combination of the DTC-algorithm and the sampler is presented in paper 5. Here, the sampler is used with a di_erent scheme to estimate the stator resistance, which a_ects the torque and speed estimation. The tests show that the sampler correctly identi_es the change in stator resistance and in this way makes it possible to achieve constant torque and speed, using the DTC-algorithm, even under stator resistance change.The overall contribution to knowledge in this thesis is that a machine control system can be made less dependent on parameter variations during operations of the machine, without interfering with the operation of the power inverter. This can be done by carefully designing the controller, or in the case of sensorless operation, by means of the specialized sampler developed as a part of the work presented.
1The zero-period refer to the state where all the lower switches in the switching matrix are in the same state, and all the upper switches are in the same state, but opposite to the lower switches, connecting the outputs of the power converter either to the high or low potential of the DC-bus.