Abstract With the rise of cruise services, energy consumption and emission of the maritime area are increasing. Due to the negative effect of greenhouse gases, many policies have been issued in the world to save energy and reduce emission. Adhering to the principle of energy conservation and emission reduction, an artificial neural network model with strong nonlinear fitting ability is introduced to explore the dynamic sailing data, and predict the fuel consumption for cruise ships based on automatic identification system data. Considering the constraints of station arrival time and the uncertainty of sailing speed and load during sailing, which can obtain the change rule from the historical voyage data, the objective function is to minimize the fuel consumption of a voyage. The established artificial neural network model is embedded into these four improved particle swarm optimization algorithms (GPSO, LPSO, MCPSO and SIPSO) with global search capability to optimize the sailing speed between stations, achieving the economic and environmental protection of a voyage. This method is applied to a real case study of Norwegian waters. By comparing the optimization results of these four algorithms, the total fuel consumption is potential to reduce from 97.4 t to 86.6 t of a voyage with the help of multi-swarm cooperative particle swarm optimization algorithm when its inertia weight is 0.7. It demonstrates that the method can be used as a tool to plan the sailing speed of cruise ships in advance.