Hybrid Measurement and Deep Learning Framework for Harmonic Estimation in Induction Motors
Abstract
Induction motors (IMs) are widely used in industrial applications due to their robustness, low cost, and simple construction. However, the increasing integration of power electronic devices such as inverters, soft starters, and variable frequency drives (VFDs) has made modern energy systems more complex, introducing significant harmonic distortions. These harmonics alter the sinusoidal nature of supply voltages, causing additional losses, torque pulsations, overheating, and reduced efficiency. In particular, odd-order harmonics such as the 5th, 7th, 11th, and 13th adversely affect motor performance by generating reverse torques, increasing mechanical vibrations, and accelerating insulation and bearing degradation. To address this challenge, this study investigates harmonic estimation in IMs using a combination of real measurements and artificial intelligence. Voltage signals were acquired from motors under star-delta, soft starter, and VFD-fed conditions at an industrial facility. To augment the dataset and model real noise environments, conditional Generative Adversarial Networks (cGANs) were employed to generate synthetic signals at varying signal-to-noise ratios. A feedforward neural network was then trained with these real and synthetic signals to estimate the amplitudes of key harmonics. The proposed model, optimized using the Adam optimization algorithm significantly improved estimation accuracy, reducing the Mean Absolute Error (MAE) from 0.9008 to 0.2993 and the Root Mean Squared Error (RMSE) from 1.0195 to 0.4123. The proposed framework also achieved very accurate estimation of the 5th, 7th, 11th, and 13th harmonics compared to ground-truth measurements. These results demonstrate the potential of combining real-world measurements, synthetic data generation, and machine learning regression for accurate harmonic characterization in IMs, contributing to improved diagnostics, monitoring, and efficiency in industrial energy systems.