Ship energy system design optimisation
Optimisation of the design and operation of ship energy systems in terms of energy efficiency and lifecycle costs
The optimisation software was developed to aid ship designers in simulation-based optimisation of energy system design. The optimisation can consider e.g. installation of waste heat recovery systems and battery systems. Energy system configurations are simulated in operational scenarios that describe the demand of propulsion and electrical power over long periods based on historical data. The optimisation is guided by a machine learning based surrogate model. Multi-objective optimisation is employed to explore the trade-offs between objectives such as investment cost, operational cost, energy efficiency and emissions.
The distributed optimisation can run hundreds of energy system simulations in parallel in a high-performance computing (HPC) environment. The dynamic simulation model of ship energy systems is based on the VTT proprietary energy system design & operation platform (NDOO), which has been implemented in the Matlab/Simscape environment. The NDOO platform includes a library of energy system component models, from which the potential energy system configurations are assembled in design optimisation.
Both genetic algorithms and distributed Bayesian optimisation are available as alternative methods of multi-objective optimisation. The Bayesian optimisation algorithm uses a machine learning surrogate model of energy system behaviour to quickly evaluate potential designs for their key performance indicators such as lifecycle costs and energy efficiency, and the most promising designs are selected for dynamic simulation.
The system is under development with a test case based on the refit of a container ship. The main focus is on the effect of installing different waste heat recovery systems and battery systems. Besides installation of additional components, the goal is to optimise also their operational settings in terms of e.g. setpoints and thresholds.
The VesselAI project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 957237.