Long-term Energy-system Optimization Models - Capturing the Challenges of Integrating Intermittent Renewable Energy Sources and Assessing the Suitability for Descriptive Scenario Analyses

This dissertation focuses on energy system optimization models (ESOMs). These models are used to generate possible transition pathways of the entire energy system in a single or multiple countries over a time horizon of multiple decades. Experimenting with different transition pathways allows gaining insights into the complexity of the energy system transition and can help in forming a long-term vision of this transition. In addition, these transition pathways can be used to evaluate the adequacy of the current policy framework to achieve a desired transition. As such, these models form valuable tools for policy makers. 

Due to the large scope of ESOMs, solving these models quickly becomes computationally demanding. To limit the computational cost, ESOMs have historically used a low level of temporal and technical detail to represent the operation of the power system, i.e., intra-annual variations in demand and renewable generation are typically represented by 4-48 so-called time slices and the technical constraints faced by thermal power plants when changing their power output, starting up or shutting down are neglected. However, in the context of an increasing penetration of strongly fluctuating and limitedly predictable renewable energy sources such as wind turbines and solar PV panels, this low level of temporal and technical detail might not be sufficient to grasp the challenges related to integrating these intermittent renewable energy sources (IRES). 

In this regard, a first objective of this dissertation is to assess the impact of this low level of temporal and technical detail on the results provided by ESOMs. The presented research indicates that both the low level of temporal and technical detail lead to an overestimation of the uptake of IRES, an overestimation of the electricity that can be generated by baseload technologies and an underestimation of the system costs. As the penetration of IRES increases, particularly the low level of temporal detail starts to have a significant impact on the obtained results. This is shown to result from the fact that traditional time-slicing methods lead to smoothing of the variability of IRES.

A resulting second objective is to develop improved time-slicing methods. In this regard, two time-slicing methods are proposed which are shown to better capture the variability of IRES without necessitating an increase in the number of time slices. This dissertation focuses in depth on one time-slicing method which relies on representing the different conditions occurring throughout a year via a small number of representative historical periods (e.g., days). The selection of the representative set of historical periods is key for the accuracy of this method. In this regard, a novel, optimization-based, approach to select a representative set of historical periods is developed and benchmarked to state-of-the-art approaches available in the literature. This developed approach is shown to achieve better results than the approaches available in the literature. The significance is that, given that a limited number of time slices can be used, a better selection of representative periods allows improving the results provided by ESOMs.

A third objective is to develop methods to tractably account for technical constraints in ESOMs. To this end, reduced formulations of the technical constraints faced by power plants are formulated. The results of a planning model integrating these reduced formulations are compared to the results of a planning model which integrates detailed clustered unit commitment (CUC) constraints for a variety of scenarios and cases. This analysis shows that the developed reduced formulations are sufficiently accurate for long-term planning purposes while reducing computation time by a factor of 5-600 with respect to the model with integrated CUC constraints. However, the research presented in this dissertation also highlights that, due to assumptions which need to be made regarding the cycling capabilities of thermal power plants and the requirements for operating reserves, there is a risk that the incorporated technical constraints are overly and unrealistically restrictive, which can lead to strong overestimations of the system costs and suboptimally low penetration levels of IRES. Recommendations to avoid this potential pitfall are presented. 

The final part of this dissertation relates to the fact that since the liberalization of the electricity markets, investment decisions in generation capacity are made by private, profit-maximizing, actors. The decisions made by these actors can be influenced by the market design and the policy framework. In this regard, the last objective of this dissertation is to determine to what extent ESOMs can account for specific market designs, policy interventions and behavioral characteristics. An analysis is presented which shows that a number of inherent assumptions are made in optimization models which prevent from representing certain market designs, policy interventions as well as behavioral characteristics.