Sensorless Control Of A Switched Reluctance Motor Driver Based On Sliding-Mode Rotor Position Observer

Switched Reluctance Motor (SRM) Driver (SRD) is an advanced electromechanical device. But the operation of SRD requires accurate information from the rotor position. A shaft encoder or resolver is usually utilized for this. The position sensor not only adds complexity, cost and size to the whole drive system, but also causes limitation for some industrial applications. This has led to attempts to find an alternative way to detect rotor position i.e. indirect rotor position sensing. This dissertation is focused on the theory of indirect rotor position sensing using voltage and current information of SRM.Firstly, a method for initial rotor position sensing in standstill and rotating situation is presented. Each SRM phase is excited with a narrow voltage pulse. The present dissertation includes a detail study of the effect of the exciting time, the motional back emf and resolution in analog to digital conversion on the precision in the proposed method. The proper exciting time is given and a conclusion is drawn that the insufficient resolution in analog to digital conversion is the primary source of position estimate error. The rotor position region can be detected by comparing the three resultant phase currents. And then the suitable resultant phase current is selected to estimate the rotor position, the method is simple and the estimate error is reduced. The method is also can be used in rotating situation, eliminating the starting hesitation for reliable indirect rotor position sensing.Secondly, a new"pulse exciting"method suitable for low speed application is described. According to the new method, sensing voltage pulse is mainly injected into the idle phase in the minimum inductance region to detect rotor position. The negative torque produced by the sensing voltage pulse can be neglected in the minimum inductance region and the efficiency of SRM is improved. In the minimum inductance region the back EMF can be neglected. And in the increasing inductance region the EMF opposes the rise of current in the phase, so the position estimation scheme is reliable. Therefore the new method is sufficiently precise even under the high back EMF effect. According to the new method, the next excited phase is sensed to detect rotor position, so the selected phase is ensured to be idle to be injected with the sensing voltage pulse.Thirdly, an analytical evaluation of the sliding-mode observer (SMO) for eliminating position sensors in SRD is presented, no extra hardware is required. The sliding-mode rotor position observer is based on the inductance mode simulated by linear piecewise function. The magnetic saturation effect is also considered. Our aim is to propose a simple and compact model of SRM that could be used for online rotor position estimation. The principle of SMO is described, the reaching and sliding conditions and fundamental of choosing the switching gains are also given. In practical applications, the mathematic model errors of SRM are inevitable. The effects caused by model errors are also analyzed. The rotor position estimate error is caused by the model errors. And the speed estimate error caused by the model errors and disturbances is also discussed. Discussions and mathematic equations for optimizing switching gains to reduce the speed estimate error are given. The quantitative relationship between acceleration and the speed estimate error is also given. Method of acceleration estimation is adopted to amend the speed estimate error.Artificial neural network (ANN) forms a very efficient mapping structure for the nonlinear SRM, and is a potential develop direction for SRM indirect rotor position sensing. This dissertation utilizes the powerful identification ability of non-linear model of ANN to implement indirect rotor position sensing. Small quantities of data are used for training the network, and large numbers of data are used for verification. Integrating with the optimize adjustment of the turn-on and turn-off angle control method, the proper phase electromagnetic information is selected for position sensing in the overlapping region of two conducting phases. The trained ANN works in the high precision region, the rotor position error is reduced.A SRD experimental facility is established and a series of experimental research are conducted to validate the indirect rotor position sensing method. Hardware implementation of a SRM controller is achieved using Texas Instrument's TMS320LF2407A Digital Signal Processor (DSP) and some peripheral devices. The converter topology is asymmetric half-bridge circuit; power device is FUJI's IGBT 2MBI100-060L. From experimental results, the position sensing method presented in this dissertation is proved to be effective, which has advantages of high precision and fast dynamic response.