Self-tuning Torque Control of Induction Motors for High Performance Applications

A cooperation between Danfoss A/S Transmission Division and AUC Department of Control Engineering concerning clarification of the potential of modern control theory used in a standard drive system has been continued in the present project. The purpose of the work is stated in the following: · To analyze and develop strategies for torque control of induction motors well suited for automatic tuning. · To analyze and develop methods for automatic tuning of the applied controllers. · To develop robust methods for adaptive field oriented control. · To test the final concept on different motors. The last item is fulfilled by building up a laboratory setup for test of the methods developed on several motors of different types. The test bench is very flexible, allowing easy changing of motors and load and it is possible by software to exchange the motor and load with a software module simulating the motor/load system. Because the control system unchanged by switching between control of the simulator and the real motor, many fault sources are eliminated by this idea. In the situation with adaptive control with on-line estimated motorparameters leading to a situation with "exact agreement" between experiment and theory (simulation) - or nights without sleep. The stator phase voltages are impressed by a standard power device with optical interface to the switch-transistors. Because the stator voltages are not measured directly but only given by the on/off times of the transistors and the DC-link voltage a non-linear model of the inverter giving the relation between turn-on times and voltages is developed. A dynamic model of the induction motor based on space phasors is described. The model in a reference frame fixed to the rotor magnetizing current is analyzed in detail and extended with a model for magnetic saturating. The parameters in this non-linear model of the motor and inverter are determined by impressing some special designed stator voltage signals and measuring the stator currents. A s something new in this context a robust current controller is determined by relay experiment before starting the parameter estimation process, leading to experiments under full current control. The effect of magnetic saturation is analyzed and new methods for estimation of the mutual inductance and the reference value for the magnetizing current is develop. In all experiments, two of the tree phases are given the same potential, i.e., no net torque is generated and the motor is at standstill. The motor parameters found by system identification at standstill are used for tuning of the controllers in a field oriented control system. A new field weakening function computed for correct voltage limitation is developed and used together with over-modulation, rated torque at rated speed is shown to be achievable without voltage limitation. On standard motors the field is not a measurable variable and because the whole concept of field oriented control needs precise knowledge of especially the field angle, a good motor model is necessary. Because system identification at standstill gives good estimates of all motor parameters, only those parameters varying because of temperature and saturation have to be estimated on-line. Due to the inverter non-lineatity, conventional methods based on recursive least square methods only give reliable results with measured stator voltages. Because the stator voltages are not measured the parameter estimation problem is solved without voltage measurement by using theory of Model Reference Adaptive Systems combined with the theory of parameter estimation assuming amplitude limited noise. The method, here called Model Adaptive, may be interpreted as an adaptive "regulator" with dead-zone, tuning the parameter method, here called Model Adaptation, may be interpreted as an adaptive "regulator" with dead-zone, tuning the parameters in the mathematical model of the system so that a given performance function is minized. High performance rotor flux observers given in the literature (often called closed-loop observers) all use stator current and voltage measurements. In systems without converters for measuring these voltages only simple feed-forward observers based on the current equation are seen. In the described methods both amplitude and angle of the field are estimated. If however one concentrates only on the field angle, which is the absolutely most critical parameter concerning field oriented control, the problem may be solved without voltage measurements by a completely new adaptive approach leading to a method called Field Angle Adaptive. The method is verified both theoretically and experimentally and is characterised by simplicity robustness. Based on correct estimated field angle, methods for on-line estimation of the field amplitude (Supervised Field Magnitude Adaptive) and the rotor resistance (Supervised Rotor Resistance Adaptation) have been developed. Robust performance of the methods is shown experimentally on several different motors and by simulation. Because the methods are based on supervision of correct performance of the new Field Angle Adaptation method they are not before seen in the literature. The side effect of correct estimated values of the field amplitude and the rotor resistance is a precise torque estimator with a fast dynamically response and an on-line motor temperature supervisor based on the variation of rotor resistance.