Electricity consumption forecasting for sustainable smart cities using machine learning methods
Darius Petelează, Alexandru Matei, Radu Sorostinean, Árpád Gellért, Ugo Fiore, Bala-Constantin Zamfirescu, Francesco Palmieri
Abstract
Integrating smart grids in smart cities is pivotal for enhancing urban sustainability and efficiency. Smart grids enable bidirectional communication between consumers and utilities, enabling real-time monitoring and management of electricity flows. This integration yields benefits such as improved energy efficiency, incorporation of renewable sources, and informed decision-making for city planners. At the city scale, forecasting electricity consumption is crucial for effective resource planning and infrastructure development. This study proposes using a time-series dense encoder model for short-term and long-term forecasting at the city level, showing its superior performance compared to traditional approaches like recurrent neural networks and statistical methods. Hyperparameters are optimized using the non-dominated sorting genetic algorithm. The model’s efficacy is demonstrated on a six-year dataset, highlighting its potential to significantly improve electricity consumption forecasting and enhance urban energy system efficiency.