Emission-aware and data-driven many-objective voyage and energy management optimization of solar-integrated all-electric ship

All-electric ships (AES) with energy storage systems (ESS) and solar photovoltaic (PV) are gaining popularity due to their capability to provide clean energy and improve operational efficiency. However, it introduces additional complexity to voyage and energy management schedules. Hence, this research proposes a data-driven many-objective voyage and energy management scheduling of AES with ESS and PV while considering non-linearity in constraints and objectives formulation. It minimizes fuel consumption, emission, ESS cumulative damage, and the operating cost of auxiliary units (PV, ESS, and shore-power supply). It investigates the performance of non-dominated sorting algorithms (NSGA-II and NSGA-III) on the proposed AES mathematical model. Furthermore, a long-short term memory (LSTM) based PV forecast is carried out based on the environmental factors, PV characteristics, and operation time to improve the voyage and energy dispatch accuracy. In addition, the proposed framework identifies the conflicting and associating objectives based on the Pearson correlation coefficients to establish the relationship between the objective functions. The result illustrates that the proposed non-linear AES model can converge to the solutions within 300 iterations. Hence, allowing the vessel operator to identify the desired operating points that meet the objectives and constraints quickly and efficiently. Furthermore, the correlation coefficients can identify respective associations and conflicting objective terms in the proposed coordinated AES voyage and energy dispatch model.

Automated summary

A data-driven many-objective voyage and energy management scheduling of all-electric ships (AES) with energy storage systems (ESS) and solar photovoltaic (PV) is proposed to improve operational efficiency. The study demonstrates that non-dominated sorting algorithms and an LSTM-based PV forecast model can efficiently identify optimal operating points that meet all constraints and objectives within 300 iterations.