Ultra short-term forecasting for the propulsion energy consumption of all-electric ships based on TCFFA-GRU-parallel network

Improving energy efficiency has recently been a topic of great interest in the shipping industry. Advanced energy management systems are gaining attention due to their potential to reduce emissions and improve energy utilization. Predicting future energy demand is the basis for solving the energy management problem, so the accuracy of the ultra-short-term prediction model is essential. This paper studies ultra-short-term prediction models using real ship operation data. The potential impact of multiple electrical data on future energy demand is also considered. A novel attention mechanism is proposed to enhance the essential channels in sequences. Subsequently, a hybrid network architecture including two individual branches is proposed to overcome some shortcomings of state-of-the-art models. The TCFFA sub-network extracts high-dimensional coupling correlations, and the GRU sub-network captures temporal dependencies. Finally, the performance of the proposed model and numerous state-of-the-art models on ultra-short-term ship propulsion load forecasting is investigated, and potential problems of the existing state-of-the-art models are analyzed. The experimental results show that the relative accuracy of the proposed model improves significantly under a variety of load fluctuation scenarios, which are 11.63%, 0.74%, and 3.60%, respectively.