Day-ahead Electricity Demand Forecasting in an Electrified Seaport using Crane Scheduling

Abstract In the coming years, seaports will undergo significant electrification process, moving away from fossil fuels. In such new reality, obtaining accurate electricity load forecasting is critical for reducing costs, planning infrastructure improvements, and ensuring a stable energy supply. However, studies specifically addressing this need in ports are scarce. This paper presents several novel Long Short-Term memory (LSTM) models for forecasting the electricity demand of a highly electrified port, using the Port of Sines as a case study. These models incorporate operational data, such as vessel arrival schedules and quay crane usage, to enhance forecasting accuracy. Our results show that including these variables significantly improves forecast accuracy, reducing the Mean Absolute Percentage Error (MAPE) from 10.55% to 3.59% compared to models relying solely on historical data. This research provides a robust framework for ports to improve energy management and supports the broader goals of energy efficiency and sustainability in the maritime industry.