The impact of renewable energy intermittency on the operational characteristics of a stand-alone hydrogen generation system with on-site water production
Faculty of Health, Engineering and Science
School of Engineering
In renewably powered remote hydrogen generation systems, on-site water production is essential so as to service electrolysis in hydrogen systems which may not have recourse to shipments of de-ionised water. Whilst the inclusion of small Reverse Osmosis (RO) units may function as a (useful) dump load, it also directly impacts the power management of remote hydrogen generation systems affecting operational characteristics. This research investigates the impact on the hydrogen generation system when simulations utilise different methods to account for solar power needed to drive the system as well as varying scales of (short-term) battery capacity. The simulations, in MATLAB/Simulink, utilise two specific methods of irradiance prediction (ASHRAE clear sky model and Neural Networks) and are benchmarked against measured irradiance data for Geraldton (Western Australia). This imposes different levels of accuracy and intermittency. Laboratory testing and device-level models help simulate the operational characteristics of a hydrogen generation system including on-site water production. Operational characteristics studied include: total energy available for PEM electrolysis, ontime of the electrolyser in steady-state, the number of start/stop cycles for the electrolyser and its duty factor (litres of hydrogen generated over the total number of start/stops). Results show that increasing systems' battery capacity only affects the operational characteristics of a PEM electrolyser up to a threshold capacity (Ah). Increasing battery capacity generally allows for more renewable energy penetration. Additionally, the hydrogen generation system behaviour is (generally) more accurately predicted, across all battery capacities, when Neural Networks are used to predict the availability of solar irradiance. The ability to predict accurate levels of solar irradiance becomes more important in winter when irradiance is at its lowest. The results also highlight the need to use highly resolved (temporal) simulations which are able to better capture device-level operational characteristics.