Data-driven ship berthing forecasting for cold ironing in maritime transportation

Cold ironing (CI) is an electrification alternative in the maritime sector used to reduce shipborne emissions by switching from fuel to electricity when a ship docks at a port. During the ship's berthing mode of operation, accurately estimating the berthing duration could assist the port operator to manage the berth allocation and energy scheduling optimally. However, the involvement of multiple input parameters with a large dataset requires a suitable handling method. Thus, this paper proposed a data-driven approach for ship berthing forecasting of cold ironing with various models such as artificial neural networks, multiple linear regression, random forest, decision tree, and extreme gradient boosting. Meanwhile, RMSE and MAE are two main indicators applied to assess forecasting accuracy. The simulation-based result shows that the artificial neural network outperforms all other models with the lowest error performance of RMSE (3.1343) and MAE (0.2548), suggesting its capability to handle nonlinearities in complex forecasting problems of port activity. The high accuracy of forecasting output in this study, which is berthing duration contributes to close estimation of two info: 1) CI power consumption and 2) departure time of the ship. This information is vital to the port operator to be used in the energy management system (EMS) as well as in the berth allocation problem (BAP).