Machine Learning Basics and Potential Applications in Power Systems

Power system studies have relied on physical model-driven methods for decades. However, uncertainties arising from integrating renewable energies, nonlinearities introduced by power electronic devices, increased dependence on cyber-physical systems, and the need for fast and accurate big data analysis challenge traditional power system methodologies. In recent years, machine learning (ML) has revolutionized scientific research, making it possible to address constantly changing and nonlinear questions without the need for pre-determined models. This paper first introduces the basics of ML and typical algorithms to new researchers and readers. Then typical examples of applying ML to power systems are proposed but not limited to electricity customer clustering, load and electricity price forecasting, power system dynamics prediction, impedance model identification, power system security, optimal load flow, load management control and inverter-based resources (IBR) control. In future studies, it is encouraged to embrace this emerging technology and utilize a combination of data-driven and model-driven methods.