A robust method of dual adaptive prediction for ship fuel consumption based on polymorphic particle swarm algorithm driven

Fuel consumption prediction plays an irreplaceable role in the assessment of ship emissions and energy efficiency optimization. However, in practical and diverse situations with different ship types, routes, and operating conditions, single-algorithm-based fuel consumption prediction models fail to operate effectively. Additionally, conventional fusion models that lack construction and optimization mechanisms also lack adaptability. To tackle this issue, this paper proposes a dual-adaptive prediction method. Initially, the Particle Swarm Optimization (PSO) algorithm is utilized to initialize each base-model. Next, the Binary Particle Swarm Optimization (BPSO) algorithm is employed to construct an adaptive Blending architecture. Finally, the Phasor Particle Swarm Optimization (PPSO) algorithm is used for adaptive cascaded optimization of the fusion model. Based on the aforementioned approach, the BPSO-Blending-PPSO model, which consistently maintains optimal performance, is constructed and validated using operational data from two ships. The results demonstrate that the BPSO-Blending-PPSO model reduces the RMSE value by 5.2886–46.7492 % compared to single-algorithm models, effectively adapting to various sailing conditions of different ships. This method not only provides a new perspective to improve the predictive accuracy and robustness of ship fuel consumption models but also exhibits good scalability by incorporating various novel prediction algorithms into the base-model pool.