Switched Reluctance Motor Modeling And Position Sensorless Control Base On Neural Network

Switched Reluctance Drive (SRD) system is a novel AC driving system. It isgetting more and more attention and a lot of industrial applications due to itscharacteristic of simple, reliable performance, high efficiency and large torque outputover very wide speed range. However，the Switched Reluctance Motor(SRM) has itslimitation that can’t adapt the fields of high temperature and high speed due to usingthe Position Sensor. Therefore, in this paper, to promote its application in the field ofhigh-performance, and carry out related research for the sensorless control of SRM.The paper analyses the structure and working principle of SRM in detail，and thenFinite Element analysis software has been used to analysis dynamic features for theSRM, and calculates the precise curves of flux linkages and torques，which lays afoundation for the modeling of the SRM in the following study.The switch reluctance motor model in Matlab library takes into account thenon-linear factors. Its drive control system characteristics are close to the actual controlsystem, and then the data obtained by the model is used as the measured data. Based onFinite Element data and measured data, a method that optimizes the BP neural networkstructure using Finite Element data and trains by the measured data is proposed. Bythis method, the systematic errors caused by model structure optimization base onmeasured data could be effectively avoided, and SRM model are set up by theoptimized Neural Network. The validity of this model is verified by simulation andexperimental results. Based on the model, a dynamic simulation method for theSwitched Reluctance Drive system (SRD) is proposed. The experimental results showthat the model has better characteristics.The SRM is difficult to achieve the desired performance by traditional PIDcontrol because of its strong non-linear, a variable structure and varying parameters.Thus, an adaptive PID control of SRM based on BP neural network is proposed, andsimulation comparison experiments that include the Traditional PID control, adaptivePID control with BP Neural Network show that an effective strategy is designed, whichhas response rapidly, anti-interference ability and robustness ability.Finally, a real-time system platform is built for the SRM based on DSPTMS320F2812. Reluctance experiments are preformed. Results and analysis to them are given. The hardware circuit includes a micro-controller, the power driver and theinverter circuit, the detection circuit, the power supply circuit and the other interfaces.