Hybrid ship unit commitment with demand prediction and model predictive control

Janne Huotari*, Antti Ritari, Jari Vepsäläinen, Kari Tammi

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

11 Citations (Scopus)
96 Downloads (Pure)


We present a novel methodology for the control of power unit commitment in complex ship energy systems. The usage of this method is demonstrated with a case study, where measured data was used from a cruise ship operating in the Caribbean and the Mediterranean. The ship’s energy system is conceptualized to feature a fuel cell and a battery along standard diesel generating sets for the purpose of reducing local emissions near coasts. The developed method is formulated as a model predictive control (MPC) problem, where a novel 2-stage predictive model is used to predict power demand, and a mixed-integer linear programming (MILP) model is used to solve unit commitment according to the prediction. The performance of the methodology is compared to fully optimal control, which was simulated by optimizing unit commitment for entire measured power demand profiles of trips. As a result, it can be stated that the developed methodology achieves close to optimal unit commitment control for the conceptualized energy system. Furthermore, the predictive model is formulated so that it returns probability estimates of future power demand rather than point estimates. This opens up the possibility for using stochastic or robust optimization methods for unit commitment optimization in future studies.

Original languageEnglish
Article number4748
Number of pages21
Issue number18
Publication statusPublished - 11 Sept 2020
MoE publication typeA1 Journal article-refereed


  • Gaussian Process
  • Maritime
  • Mixed-integer linear programming
  • Model predictive control
  • Optimization
  • Predictive model


Dive into the research topics of 'Hybrid ship unit commitment with demand prediction and model predictive control'. Together they form a unique fingerprint.

Cite this