Development and evaluation of an AI-based method for improving the energy efficiency of electric motor−inverter systems

Abstract

This study investigates the energy efficiency of an electric motor–inverter system operating under variable load conditions. The aim of the research is to develop and evaluate an AI-based supervisory method for identifying inefficient operating modes and optimizing control parameters using real-time electrical and operational data. Unlike conventional motor-control studies, which mainly focus on torque ripple, flux regulation, or speed response, the proposed approach considers the motor, inverter, load, and control algorithm as a single energy-efficiency-oriented system. The input vector of the model includes voltage, current, inverter frequency, rotational speed, electromagnetic torque, temperature, load level, power factor, and total harmonic distortion. The output variables are system efficiency, total power loss, and the classification of the operating mode as either efficient or inefficient. The proposed method was evaluated under 24 steady-state operating modes formed by four load levels and six inverter frequency values. Under the most inefficient tested operating condition, the overall system efficiency increased from 78.3% to 90.2% after AI-based optimization, while energy consumption decreased from 118 kWh to 88 kWh. In addition, the power factor improved from 0.76 to 0.94, total harmonic distortion decreased from 15% to 5%, and inverter switching losses decreased from 8% to 3%. The results confirm that the proposed method can identify inefficient operating states and support the selection of energy-efficient control parameters. The main limitation of the study is that the validation was performed using one motor - inverter configuration and mainly under steady-state conditions. Therefore, future research should include transient load changes, different motor power ratings, and embedded real-time implementation.