Energy Management Control of Hydrogen Fuel Cell Powered Ships

Abstract

An Energy Management System for electric vessels is described, based on a Model Predictive Control (MPC) with Anticipative Action. The electric ship has a power system composed of a hydrogen fuel cell generator, a battery storage system, a propulsion system, an auxiliary load module, and a command system. The controller defines the power allocated among the vessel’s power system components. The MPC design uses a Linear Parameter-Varying (LPV) model to approximate the nonlinear dynamics of the vessel’s power system and components. To improve the performance of the LPV-MPC, an additional predictor is included, based on data-driven Machine Learning. This is included in the LPV-MPC so the future predicted trajectory of the reference signals can be estimated to improve the allocation of power. The reference trajectory is generated using a Neural Network trained to estimate the future power demand determined by representative ship manoeuvres. A simple baseline Rule-based (RB) strategy was compared with the basic LPV-MPC and with the data-driven LPV-MPC that includes the prediction generated by the Neural Network.