Position Sensorless Control System Of Swiched Reluctance Motor Based On Neural Network

Switched Reluctance Motor with its simple structure, low cost, highoperating efficiency, strong fault torlerance, and control of a flexible and widespeed range, has been widely used in electric drive vehicles, householdappliances, general industrial, textile machinery, power transmission and otherfields. The increase of the position sensor limits its scope of application, asensorless position detection method is proposed.This dissertation first introduces the basic structure and working principle ofthe switched relectance motor, after analysising the ideal linear model, theposition sensorless detection scheme of phase excitation pulses to non-workingphase is proposed, which can be drawn from the rotor position in response to thecurrent function. Analysising of the adwantages and disadvanges of the BP (BackPropagation) neural network model, a genetic algorithm combined with BPneural network-GA-BP is proposed, which improved the problem of lowconvergence and the difficult to determine the optimal initial weights andthresholds. Therefore, the back propagagtion neural network model combinedwith genetic algorithm is established, in which the response current as the input,the rotor position as the output. The MATLAB software is used to programming.The switched reluctance motor drive and control circuits are designed, and teston a3000W12/8pole motor, analysis of the operating characteristics of themotor. The results show that the design of the sensorless position switchedreluctance motor drive control system method is simple to achieve, meeting themeasurement accuracy, and with stability motor performance, confirming thefeasibility of the designed system.