Development of speed-power performance models for ship voyage optimization
Abstract: Various measures, such as voyage optimization, performance monitoring and ship cleaning schedules, have been developed to help increase the energy efficiency of shipping operations. One of the most important elements needed for these measures is a reliable ship speed-power model. Many research efforts have been devoted to developing such models to describe a ship’s energy performance for head-to-beam seas, which are important for ship design purposes. For measures to increase the energy efficiency of a ship’s operations, speed-power performance models for other heading angles are of equal importance but are rarely investigated. Therefore, the overall objective of this thesis is to develop speed-power models for arbitrary wave headings that are especially applicable for ship voyage optimization. First, a semi-empirical model is proposed based on experimental tests. Then, a machine learning model (black box) is developed based on a large amount of full-scale measurement data. For the semi-empirical model, formulas to estimate a ship’s added resistance in head waves are developed to effectively describe a ship’s hull forms and other main characteristics. The formulas are then extended to estimate the impacts of wave headings from different angles, and these are verified by experimental model tests. A significant wave height-based correction factor is proposed to consider the nonlinear effect on a ship’s resistance and power increase due to irregular waves. For the machine learning-based model, the XGBoost algorithm is used to establish the model based on full-scale measurements of a PCTC. The input features include parameters related to ship operation profiles, metocean conditions, and motion responses. For the three case study ships, the discrepancy between power predictions and the actual values is reduced from more than 40% using today’s well-recognized methods to approximately 5% using the semi-empirical model proposed in this thesis. The machine learning model can further reduce the discrepancy to less than 1%. It is also demonstrated that the improved models can help to effectively optimize a ship’s voyage planning to reduce fuel consumption.
CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)