Uncertainty is an increasingly important aspect of decision-making relating to the electricity systems of the future. Over the long-term time horizons required for investment decisions and government policy making, history indicates that forecasts tend to be varied and uncertain. Hence, long-term forecast uncertainty should be an important aspect of any electricity market modelling or planning exercise. However, surprisingly, we do not find this to be the case in the literature.This thesis contributes to the study of the incorporation of long-term uncertainty into the generation expansion planning class of decision-making models in a number of ways. Firstly, in order to represent a wider range of long-term uncertainties into the generation expansion planning model this thesis first investigates one of the more promising possibilities for reducing model complexity, the representation of time. A methodology for adjusting the weighting derived from common representative day clustering algorithms is proposed for use in generation expansion planning models that ensures the targeted level of net demand is captured in the model without altering the underlying net demand shapes that define ramping challenges. The results demonstrate the importance of carefully performing the clustering of representative days with the model selected expansion plans differing greatly in terms of both the total installed capacity and technology choice. The thesis then investigates the role of uncertainty in the important system planning question of quantifying decarbonization costs. I focus on the Ghanaian system to provide a benchmark for developing countries and provide insight into the relatively under-studied sub-Saharan region. To do so a generation expansion planning model is modified to incorporate the reality of fuel shortages and fuel switching typical of a developing country’s power system. From this modelling, a range of emission reduction costs are generated that provide important benchmarks and I identify drivers of these costs specific to developing countries. The results demonstrate that discount rates, representing Ghana’s access to capital, are a particularly important variable for developing countries. Lower discount rates can lead to more investment in capital intensive renewable energy in the long run but can also lock-in additional conventional generation investment in the short term. The thesis then turns to the investigation of the importance of representing a wide range of economic and physical sources of uncertainty into the modelling of the electricity system and focuses on the method with which uncertainty is incorporated, both for investment decision making and policy analysis. The results of a United Kingdom case study demonstrate the importance of combining uncertainty across different inputs, finding that the difference between a deterministic and stochastic solution increases non-linearly when uncertainty inputs are combined. Further, it is demonstrated that combining uncertainty sources by adding a limited number of scenarios to multiple sources of uncertainty outperforms adding additional scenarios to any individual source of uncertainty. Finally, the representation of uncertainty as individual scenarios is shown to underestimate the range of price outcomes and overestimate the range of potential CO2 emission outcomes, given uncertainty.The final study of the thesis compares six different policy options for reducing carbon emissions in the electricity system: a cap on CO2 emissions (as with a cap and trade scheme), a CO2 price, a renewable capacity target, a green certificates scheme, a renewable generation subsidy, and a renewable capital grant under different treatments of long-term uncertainty. In a case study of a small power system, the results show that using common modelling approaches that attempt to capture uncertainty as multiple different independent scenarios (such as scenario analysis or Monte-Carlo simulation) perform poorly at representing the reaction of a competitive electricity market as measured by a stochastic optimisation model. A policy maker using a scenario-based approach to make decisions could set a policy 55% more restrictive than required to meet their emission target. Further, a deterministic model that ignores uncertainty can underestimate carbon abatement costs by up to 85%. Incorporating uncertainty as individual scenarios only slightly improves this result and biases the estimated costs between price and quantity-based policy approaches to decarbonizing the system. Throughout this thesis, a continuous set of results are presented that make the case for long-term uncertainty being a critical consideration for the electricity system modeller.
A long term generation expansion planning model using system dynamics – Case study using data from the Portuguese/Spanish generation system
