Research On Least Square Estimation Of Induction Motor Parameters

Parameter identification of induction motor is the premise to realize high performance of speed regulation system for AC motor. This article studies the classical induction motor control method-vector control and direct torque control by comparison. Vector control depends on the rotor time constant or the stator resistance and other induction motor parameters to achieve the precise orientation of the magnetic field of the rotor. While the direct torque control is mainly dependent on the stator resistance to determine the position of the stator flux vector. After studying the existing induction motor parameter identification methods and making a comprehensive comparison, this article finally chooses the least squares method as the identification algorithm to analyze and resolve the outstanding problems while estimating induction motor parameters.Firstly, create a correct mathematical model of induction motor and give the deduction of identification equations. Based on the induction motor features of higher-order, nonlinearity, strong coupling and multivariate, establish the mathematical induction motor model under reasonable constraints. It uses voltage and load torque as input variables and makes mechanical torque and speed as the output variables. Then validate the model by simulation platform Matlab/Simulink. Guided by the established mathematical model and the least squares standard formula, deduce the suitable identification equation for parameter estimation.Secondly, propose appropriate solutions direct at the two outstanding issues during identifying motor parameters by least squares method. After adequate theoretical analysis and simulation, choose Chebyshev filter and Butterworth filter as digital filters under different noise conditions. In order to solve the problem of influence on identification results when the high-term derivation is discreted in identification equation, the modified Euler method is applied into digital filter solver. Local truncation error arising from the higher-derivative term in the identification equation, use the modified Euler method to solve the digital filter. Based on the autocorrelation error analysis of identification equation, choose the generalized least squares in real time method to estimate parameters. Higher estimation accuracy can be got by a combination of both methods mentioned above.Finally, use Matlab/Simulink as the simulation platform to build digital filter and least squares method simulation modules and to verify the feasibility and accuracy of design through a lot and comprehensive simulation experiments. The results show that the modified Euler method significantly reduces the identification error than the Euler method does, and does not increase the computational difficulty; generalized least squares in real time method has obvious advantages in estimation results compared to the real-time least squares method which improves the error estimation accuracy greatly.