The forecast of hourly electricity demand and its price in the free power market is one of the most important tasks when managing modern power systems of energy scheduling. Improving the assertiveness of this forecast can support government agencies, generators and concessionaires of the energy industry to develop more optimized planning w.r.t. the cost of the electricity for immediate demands, in addition to establishing a safer energy consumption horizon for general customers. In this research, we are interested in investigating both hourly electricity demand and its corresponding price after direct consumption by customers. For this task, we apply modern Machine Learning (ML) models, Data Science machineries and exploratory data analysis tools so as to support the electricity industry in its attempt to formulate more optimized, safer and economical energy allocation plans. To properly conduct our research, three well-established approaches of Machine Learning will be considered: Random Forest, Support Vector Machines, and Gradient Boosting. All the ML-driven frameworks will be operated and validated from public databases made available by both foreign and Brazilian energy operators, therefore providing a more in-depth and realistic analysis of the power demand problem while still consolidating partnerships between university and the energy industry.
News published in Agência FAPESP Newsletter about the scholarship: