A Two-Stage Approach Combining Constraint-Based Algorithms and Gaussian Process Regression for Estimation of Cruise Ship Hotel Loads

This paper presents a hybrid approach based on two stages for estimating the hotel load power demand on cruise ships. In the initial stage, a constraint-based algorithm was used that integrated criteria related to engine continuous rating, propulsion type, and power. This formed the basis for determining maximum auxiliary engine power, crucial for onshore power system planning. In the second stage, Squared Exponential Gaussian Process Regression (GPR) was introduced as a machine learning tool, exhibiting outstanding performance in capturing intricate data relationships. The model, trained on specific features, underwent thorough evaluation, including cross-validation and sensitivity analysis. Key performance metrics for the GPR, such as a Root Mean Squared Error (RMSE) of 0.1421 MW and 0.9991 MW for training and testing, respectively, highlight the model’s accuracy and generalization capabilities. Non-linear relationships observed in certain features underscored their influence on predicted hotel loads. The two combined stages provide a robust framework for accurately estimating cruise ship hotel loads.