Identification and control of switched reluctance motors using artificial neural networks and a priori knowledge
The identification and control of switched reluctance motors (SRMs) is of primary interest in the area of controls because of the SRM's reliability and high torque to weight ratio. As desirable as SRMs may be, they are not without drawbacks. The most notable difficulty being the highly nonlinear way that static torque varies with rotor position and stator current. It is hoped that by applying artificial neural networks (ANNs) to the problem of SRM identification and control along with novel structural modifications in the form of gray layers, a new control strategy can be achieved. A decoupled extended Kalman filter is applied to the parameter estimation problem involved in adjusting the ANN weights. A modification to gradient descent that accounts for periodicity in a plant is also presented. Gray layers appropriate to the SRM state identification and control are developed. An extensive differential equation model of a three phase doubly salient SRM is derived and then simplified for use in the simulated implementation. This model is compared to previous work in order to increase confidence in the derived model. A series-parallel identification model is used to identify the states of the switched reluctance motor. Results verify that ANNs that have been modified with a gray layer are able to more accurately represent the theoretical differential equations that govern the operation of the switched reluctance motor states. Simulations were performed to address the problem of SRM control with ANNs. Both velocity control and position control schemes were considered. Main results pertaining to velocity control affirm the proposition that ANNs can effectively be used to control SRMs. In all situations the control objectives were met in an expected fashion. Position control was also shown to be viable, however more research could be done with both control schemes.
Garside, Jeffrey Jay, "Identification and control of switched reluctance motors using artificial neural networks and a priori knowledge" (1996). Dissertations (1962 - 2010) Access via Proquest Digital Dissertations. AAI9717060.