Machine learning methods in power forecasting: a systematic review

Abstract

Forecasting power consumption is a crucial aspect of managing cities and regions. Accurate forecasts ensure smooth and uninterrupted operation of consumer and industrial units. While the traditional methods of forecasting have been useful, the advent of big data has enabled the use machine learning techniques. In this paper, we discuss the applications of machine learning to power forecasting. We describe the technical details of the existing methods and highlight their strengths and weaknesses. We find that there is no single method that fits all scenarios. The optimal method depends on various factors including the characteristics of the data, the size of the data, and the forecasting horizon.