Three-Phase Induction Motor Electrical Parameter Estimation Using Artificial Neural Networks

The three-phase induction motor is the machine most used in industry to transform electrical into mechanical energy due to its constructive robustness and simple maintenance. Knowledge of the parameters of the equivalent motor circuit can be essential for the size and design of drive control, increasing the efficiency and reliability of the system. However, these values are typically unavailable, which makes specific parameter estimation methods necessary. Techniques involving no-load and blocked rotor tests are the most common in the industry. However, these methods have the disadvantage of being laborious and time-consuming. With the evolution of data processing systems, computational techniques are increasingly commonly applied to this type of task. However, most of the proposed computational methods require specific adaptations for each machine, making the process costly and not guaranteeing convergence to feasible solutions. In this context, this work proposes a reliable and generalist method, based on artificial neural networks, to estimate the parameters of the equivalent electrical circuit of three-phase induction motors based on the data provided by the manufacturer. A total of 542 different motors were considered, with powers ranging from 0.75kW to 55kW. Two approaches are compared, the first using a single neural network to estimate all machine parameters and the second using an expert network to estimate each parameter. The results show that the expert system achieved the best results, reaching errors below 8%.