Uncertainty Modelling in Renewable Energy-Based Electricity Systems Stochastic Optimization

Large contributions to the increasing Global greenhouse gas (GHG) have been attributed to unsustainable energy use, consumption, and production patterns. However, current energy and electricity modelling tools developed to give insight into the complex interactions of the components of the energy and electricity system and handle its challenges are themselves challenged with the requirements to meet up with the fast-moving energy transition. Hence, there is a need to demonstrate the importance of uncertainty modelling methods of representing variability transparently and incorporating forecast uncertainties in renewable-based electricity system optimization.
This thesis achieves this in three important research aspects. The first aspect is aimed at developing a novel methodology that integrate a state-of-the-art spatially resolved energy system co-optimization framework with uncertainty and sensitivity methods. The second aspect is aimed at applying a deep distributional Deep Reinforcement Learning (DRL) algorithm based on a reward probability distribution approach to the intelligent operational scheduling of energy and electricity management in a real-world microgrid considering uncertainties. The third and last aspect introduce a novel methodological framework that combines probabilistic forecasting system with distributional DRL algorithm to achieve an optimal least-cost operation of the microgrid. The proposed framework is tested on a real-world UK microgrid energy system.
The results for example show for the first aspect, that in comparison on average of the three scenarios considered, a total annual system cost reduction of about 48.3% is observed. In terms of the average marginal price of electricity, a reduction of about 44.2% is observed. Through the above case investigations, this thesis demonstrates the feasibility of incorporating uncertainty modelling and forecasting into energy and electricity system expansion and operation planning modelling problems to achieve cheaper operation cost. These results will assist policy-makers and researchers in providing reliable, transparent, efficient, and risk-informed decision-making on energy studies.