Determination of optimal power management strategies under operating uncertainty of a photovoltaic – battery – heat pump system with thermal storage

Details

Supervisors

Contino, Francesco

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil électromécanicien, à finalité spécialisée: énergie

Abstract

Balancing intermittent energy such as solar energy can be achieved by batteries, heat pumps and thermal storage tanks. This enables the coupling of the electricity and heating sectors. When the performances of such systems are evaluated, the parameters are often assumed as fixed and deterministic. Nevertheless, parameters such as grid electricity price, loads and investment costs are uncertain in reality. Moreover, power management strategies for combined operation are still an active research topic. To address these limitations, Robust Design Optimization (RDO) is performed on different power management strategies where uncertainties have been propagated. A photovoltaic - battery - heat pump - thermal storage tank system is studied for a household located in Belgium. First, the model undergoes the RDO using the NSGA-II algorithm and the Polynomial Chaos Expansion (PCE) method. Then, a comparison is made based on the performance of the different strategies. Finally, an analysis based on the stochastic performance of the different power management strategies is performed. The robust design optimization finds a trade-off between minimizing the Levelized Cost Of eXergy (LCOX) standard deviation and optimizing the levelized cost of exergy mean. This MSc thesis provides the optimized designs and the advantage of different power management strategies based on the financial flexibility of the household owner. The results demonstrate the possibility to reduce the impacts of the uncertainties on the system by having an improved strategy. Being able to sell its electricity surplus to the grid (i.e. grid selling strategy) allows obtaining a lower LCOX mean and LCOX standard deviation than the reference case (up to 20% and 13% respectively). Operating the system differently during the peak and off-peak hours (i.e. peak shaving strategy) or by doing a day ahead forecast (i.e. forecast strategy) also gives improved results regarding the reference case (up to 1.2% and 4.5% respectively for the LCOX standard deviation). However, a method based on buying electricity from the grid at low prices (i.e. grid buying strategy) shows weaker results regarding the LCOX standard deviation and the LCOX mean (up to 14% and 9% respectively). Subsequently, the uncertainty quantification shows that the uncertainties on the future grid electricity price and electricity load are dominant for the system (mean effect of 46% and 28% respectively on the robust designs). Finally, if the system owner predefines an upper limit of 550€/MWh, selling its electricity surplus achieves a higher probability than the reference case (i.e. 35%) of being lower than this limit. Having a different operation during peak and off-peak hours or doing a day ahead forecast improves the results regarding the reference case (i.e. both of 8%).