Off-grid PV systems for hydrogen production: From prospection analysis to system control

Hydrogen is not only the most common element in the universe, but also a versatile molecule which is an option to decarbonize sectors that are not easily electrified or are energy intensive, such as long-distance transport, chemical industry, and metallurgy among others. Hydrogen gas is typically obtained from fossil fuels, and only a small fraction is produced nowadays from electrolysis, or the process of using water and electricity to produce gaseous hydrogen.
If the electricity for powering the electrolysis process comes from renewable sources, the produced gas will have no associated greenhouse emissions. This is the so-called green hydrogen, which is the base for decarbonization of carbon intensive industries. This work investigates the potential of stand-alone green hydrogen production from solar energy, covering the whole design process, from an allocation and feasibility analysis, to system control. To do so, this thesis is separated in two parts. The first part focuses on the preliminary assessment phase of photovoltaic (PV) systems and the solar resource, while the second covers the integration of PV and electrolysis systems finalizing with a control strategy for these systems.
Chapter 2 presents a methodology for analyzing potential sites for PV deployment, including information on the degradation of the site. This provides the designer with additional information beyond the purely technical and economical layers that are typically considered in this type of study. The more degraded a site is, the more suitable it is for deploying new PV projects, avoiding pristine natural areas. This, combined with mitigation measures can minimize the environmental impact of new PV projects.
An analysis of the efficiency loss of PV systems is discussed in Chapter 3. In particular, the efficiency loss caused exclusively by quick variations in irradiance, as a consequence of passing clouds. These abrupt and quick changes affect not only the solar modules, but components downstream, such as the maximum power point tracker. The implemented algorithm might be sensitive to these changes and move the operating point of the PV module away from its maximum power point, leading to energy loss.
Predicting quick changes of irradiance is a topic covered in Chapter 4. Using sky images and artificial intelligence, it is possible to predict ultra-short-term irradiance. The proposed method is an ensemble of models, each trained on a particular sky condition. Because each model is highly specialized, once the sky condition is determined, the model that performs best on each sky type is employed, leading to lower prediction errors, more precise predictions and lower training data needed. Yet, an accurate prediction is a topic for further research.
The integration of PV with hydrogen systems is introduced in Chapter 5, which presents a literature review on integration methods for PV and electrolyzers as well as the main challenges for operating these systems in a variable manner.
Moving to the design phase, Chapter 6 proposes a sizing procedure, based on Particle Swarm Optimization to minimize the energy that cannot be used by the hydrogen equipment (electrolyzer and compressor), aiming at the maximum energy utilization in the system. Horizontally-placed PV modules provide a good compromise between efficiency, hydrogen production and cost.
Once the system has been designed, Chapter 7 puts together all the topics covered in this dissertation proposing a control strategy for an optimally-sized stand-alone PV electrolyzer systems, without electrical storage. The control is based on prediction of irradaince changes using sky-images. From Chapter 4 it was clear that the prediction using sky images is far from perfect, yet this is needed for control. To solve this problem, the strategy proposed in Chapter 7 relies on information on the uncertainty of the prediction and uses fuzzy logic to account for imperfect predictions. This strategy can effectively smooth power changes without the need of additional storage components.