Research of speed estimator based on wavelet neural network adjusted by ant colony optimization

In order to improve low speed performance of sensor-less direct torque control (DTC) system, an self-adaptive ant colony optimization (AACO) is proposed to optimize Speed Estimator (SE) based on wavelet neural network (WNN) in this paper. The ant colony applying space grids and adaptive changing enduring coefficient will be used to optimize SE based on WNN. Draw many grids in the general area of scale, position and weight of WNN. Different information is released when ants move in different space grids. Adjust the adaptive enduring coefficient rho according to grid point´s target function value. To change relevant information will influence the moving direction of next manpower ants. Shortening variable´s range according to the information in cycle, the cycle will be halted until space between grids less than enactment precision epsiv. We compare the on-line identification results of SE based on WNN optimized with grads descend algorithm for validating identification precision of SE based on WNN optimized with AACO. Simulation experimental results indicate that, compared with genetic algorithm (GA), the introduction of AACO to SE based on WNN can reduce hidden nodes, quicken convergence speed, enhance SE´s precision of on-line identification consequently, and improve low speed performance of sensor-less DTC system effectively.