Machine Learning Algorithms for Load Forecasting Based on Big Data

Big Data's importance in the context of modern Power Systems' transition is rapidly increasing. Thus, the research interest towards techniques to explore and extract the most relevant information is in a continuous expansion. In this framework, Machine Learning algorithms are the most suitable approach towards retrieving valuable knowledge in a heterogeneous and ever more complex data environment. Considering the high rate and velocity of data production in smart Power Systems, Machine Learning implementation is mandatory. Load forecasting is a hot topic due to its high fluctuation and economic impact considering the increasing and variable Renewable Energy penetration. This paper presents a comparison between the statistical load forecast and two Machine Learning algorithms based on multiple types of errors. It highlighted the necessity of refining the algorithms and enhancing the available data using Feature Engineering to improve the performance. Nevertheless, Machine Learning algorithms are the only approach feasible for the ever- increasing data pool with heterogeneous characteristics.