Sensitivity of MPC Performance to Component Scaling in a Battery–Hydrogen Storage System

Abstract

With the increasing share of volatile renewable energies, there is a growing need for flexible storage systems to balance fluctuations between generation and demand. Multi-energy systems, featuring battery and hydrogen storage systems, provide an efficient and scalable solution for this purpose. While rule-based control is primarily used in industry, model predictive control (MPC) is considered the most promising strategy for cost-optimized and safe operation. The performance of such controllers depends heavily on the capacities, power rates, and degradation behavior of the storage systems; for hydrogen systems, it can also depend on minimum switch-on/off times and ramping rates. This work aims to use sensitivity analysis to quantify the influence of variations in these component parameters on control quality. To this end, a lab-scale battery-hydrogen storage system (TU Wien) is modeled and operated using MPC to minimize grid exchange and thus increase self-sufficiency. Experimental data from electrolysers and fuel cells were used to calibrate the model. Actual energy production and consumption data from one year are condensed into five representative weeks, serving as test cases. In a systematic study, the capacities and power rates of the components are varied. For the given lab-scale system, the maximum power of the electrolyzer has the greatest influence on control performance, followed by the battery and hydrogen storage capacity. The power of the fuel cell and battery has no limiting effect. It can thus be concluded that the electrolyzer is undersized, and the battery should have greater storage capacity to improve performance.