ISSN (Print) : 2278-8948, Volume-2, Issue-2, 2013 37 Neural-Network-Based Parameter Estimations of Induction Motors Madhavi H. Nerkar & B. E. Kushare Department of electrical engg., K.K.W.I.E.E.R, Nashik, Maharashtra State, India E-mail : madhavi.nerkar@gmail.com & be_kushare@rediffmail.com Abstract – Accurate estimation of parameters during transient and steady state is required for controlling of Induction motor. Artificial neural networks (ANNs) based online identification of induction motor parameters are presented. ANNs such as feed forward network is used to develop an ANN as a memory for remembering the estimated parameters and for computing the parameters during transients. Simulations and experimental results are presented for induction motors. Index Terms—Artificial neural networks (ANNs), Induction motor, Mutual inductance, Observer system, Parameter estimation I. INTRODUCTION The Induction motor is a nonlinear multivariable dynamic system with parameters that vary with temperature, frequency, saturation, and operating point. Considering that induction motors are widely used in industrial applications, these parameters have a significant effect on the accuracy and efficiency of the motors and, ultimately, the overall system performance. Therefore, it is essential to develop algorithms for online parameter estimation of the induction motor. Such algorithms can be performed in real time because of the progress in the use of digital signal processors (DSPs) and microelectronics. In this paper ANN based parametric estimation method is developed. The Luenberger observer system is implemented for flux estimation, and the speed observer system is utilized for rotor-speed estimation. The rotor parameters are the most important parameters for the control of the induction motor drives. The rotor resistance can change up to 150% over the entire operation. A number of methods that are both for the detection of the rotor parameter and the prevention of its variation have been discussed.In the rotor parameter estimation is proposed by estimating the rotor temperature. This is based on the fact that the temperature influences the fundamental frequency component of the terminal voltage for a given input current. This is a tedious process because the temperature of the rotor windings has to be measured every time. Many published research papers have shown the effect of motor parameters on the quality of flux and speed estimation in vector control of rotation systems[3]. In this paper, the online identification of the rotor resistance and mutual inductance techniques based on ANN is presented. Artificial neural networks (ANNs) can be used to identify and control the nonlinear dynamic systems because they can approximate a wide range of nonlinear functions to any desired degree of accuracy. Moreover, they can be implemented in parallel and, therefore, shorter computational time can be achieved. In addition, they have immunity to harmonic ripples and have fault-tolerant capabilities. Since the 1990s, several investigations into the applications of neural networks in the field of electrical machines and power electronics have appeared. In recent years, the use of ANN in modulation systems, in breakdown detection, in control, in the estimation of state variables, and in the identification of induction- motor parameters. The use of ANN has been tried for estimating the rotor angular speed. Among the methods used, it is possible to note two types of ANN designs. One is based on the machine model, and the other one uses stator currents and voltages for direct speed estimation. In the proposed solution, the neural networks are used to develop an associated system for remembering the calculated values and for computing these values during the transients. Therefore, this paper considers the online identification of machine parameters in a sensor less control system. II. MATHEMATICAL MODEL AND NONLINEAR CONTROL SYSTEM OF INDUCTION MOTOR The model of a squirrel-cage induction motor expressed as a set of differential equations for the stator- current and rotor-flux vector components presented in a stationary coordinate system are as follows