Neural network based approach for the computation of harmonic power in a real-time microprocessor-based vector control for an induction motor drive

With increasing harmonic pollution in the power system, real-time monitoring and analysis of harmonic variation have become important. A neural networks based approach has been demonstrated to be a very interesting tool for this purpose. However, initial conditions are critical for real-time implementation. A selective harmonic tracking algorithm is proposed to calculate initial conditions for the neural network. In this paper, a fully digital algorithm to measure and evaluate harmonic power using a neural network together with a selective harmonic tracking method is applied to an efficient variable-speed vector-controlled 1.5 kW induction motor drive.