A Systematic Literature Review of Machine Learning Methods for Short-term Electricity Forecasting

Research in energy prediction is widely explored as it is used in long term planning like development investment and resource planning to estimating tariffs and analyzing and scheduling of distribution network. One of the methods applied in performing the forecasting is machine learning. There are many machine learning algorithms, dataset features, and evaluation metrics used. This paper offers to review articles on energy prediction published from 2016 until 2019. The review is made based on Systematic Literature Review method. A total of 119 articles were gathered from various sources such as IEEE, Science Direct and ResearchGate. The search made based on keywords such as machine learning, electricity, energy demand, forecast, and prediction. Based on the articles gathered, 31 articles were selected based on thorough examination on the title and abstracts. Six full materials are chosen for the final review. The review focused on i) standard dataset features chosen, ii) the machine learning algorithms applied and iii) the result based on evaluation metrics. Similarities found between the papers include the forecasting type, features selected, using various methods in performing the machine learning and applying multiple metric evaluations for a single dataset. The findings however show, the chosen machine learning algorithm and metric evaluation are different among the researchers and dataset size may influence the accuracy of the model generated.