Modelling approaches for optimization of integrated regional energy systems

This thesis deals with the subject of energy system optimization models. Energy systems in this context are the connected networks of energy sources, energy conversion, transport and storage technologies, and various energy end users in different sectors of society. Energy in different forms such as heat, electricity or fuels is at the core of many functions in human societies, enabling activities such as cooking, manufacturing and global logistics. At the same time, detrimental environmental impacts of such energy use have been recognized, with the burning of fossil fuels accounting for a large part of anthropogenic climate change. Reducing impacts of energy systems while at the same time providing energy services to a growing global population means that energy systems must undergo transformation to more sustainable technologies and structures.

Energy system optimization models can be used to plan both operation of and investments in energy systems. Optimization of operation takes place in short-term dispatch optimization of conversion units in existing systems as well as in investment planning models to evaluate how potential system structures would operate to satisfy estimated long-term energy demands. Investment optimization models can help identify favorable routes of development for energy systems with short- or long-term time horizons, on regional, national or even larger scales. In the research work behind this thesis, mixed-integer linear programming has been applied to optimize two types of energy systemmodels: regional heating system models including district heating network topologies and regional integrated energy system models with a focus on incorporating intermittent renewable technologies. District heating network optimization entailed a high spatial resolution to capture the large amount of relevant pipeline options to consider, which in turn required reduced temporal resolution to ensure manageable computational times for solving the model. Despite this, the modelling approach was found practically viable for evaluating investments in district heating and decentralized options such as heat pumps, boilers and thermal storages. Additionally, a decomposition method was developed for solving complex model instances which the original model formulation could not solve.

The other studied modelling approach sought for answers to questionsof how intermittent renewable technologies could in practice interact with each other and with other more or less flexible technologies to satisfy varying energy demands in an integrated energy system, and how these modelled interactions would affect investment optimization solutions. This required high temporal resolutions, as wind speeds,solar irradiation and energy demands can all change rapidly. An hourly resolution over a one-year time horizon was thus selected, which in turn required very simple technical detail levels for the conversion technologies to reduce model complexity. Example cases showed how wind turbines, heat pumps and gas engines could interact to provide electricity and heat in a regional energy system. Since the relevant question regarding investment decisions is how to plan long-term investments starting from a present situation, the model was extended to span a selectable time-horizon divided into several periods. With this extension, estimates of near-future progression of energy demands, costs and efficiencies, as well as potential dismantling of existing powerplants could be taken into account in the optimization. Example cases again showed how sector integration can provide flexibility in regional energy systems and how long-term investment optimization can be formulated as mixed-integer linear programs, but also highlighted the necessity of dealing with uncertainties in model parameters and estimates of future circumstances.

Research on energy system optimization models is ongoing and besides just developing the models themselves, it is necessary to develop frameworks and common practice for how to use the models in a decision-making process. Uncertainties are inherent in both the models and in the input data, and getting relevant information from the models requires careful formulation of optimization scenarios and analysis of the optimization output. Since this may involve large amounts of optimization runs, it is all the more desirable to reduce required computational efforts to solve the models. There is thus a call for general modelling formulations that are malleable to problems of different scales and with different technological options as energy system planning problems exist in diverse circumstances with differentgeographic, climatic and social conditions.