An artificial neural network based harmonic distortions estimator for grid- connected power converter-based applications

Abstract

Grid-connected solar Photovoltaic (PV) systems are predicted to cause significant harmonic distortions in today’s power networks due to the increase utilization of power conversion systems widely recognized as harmonic sources. Estimating the actual harmonic emissions of a certain harmonic source can be a challenging task, especially with multiple harmonic sources connected, changes in the system’s characteristic impedance, and the intermittent nature of renewable resources. A method based on an Artificial Neural Network (ANN) system including the location-specific data is proposed in this paper to estimate the actual harmonic distortions of a solar PV inverter. A simple power system is modelled and simulated for different cases to train the ANN system and improve its prediction performance. The method is validated in the IEEE 34-bus test feeder with established harmonic sources, and it has estimated the individual harmonic components with a maximum error of less than 10% and a maximum median of 5.4%.

Keywords

• Artificial neural networks
• Harmonic distortions
• Harmonics estimation
• Power harmonics
• solar PV inverter