Investment decisions in the electricity sector under uncertainty - An agent-based modelling approach

Motivation The energy transition requires coherent and consistent investment in new technologies and infrastructure. However, investment decisions are neither pre-planned nor centrally organized. Instead, they are done by a large set of heterogenous companies, each with its own interests and conditions. How these companies react to policy interventions and what combined effect they have on the energy system is still poorly understood [1]. Traditional optimization methods of determining the capacity expansion for the energy system do not account for this [2], leaving policy makers without clear guidance [3]. Approach We present an agent-based model (ABM) of capacity investment decisions in the electricity sector to study the most important features of this complex system. The model assumes companies with different levels of information, capital resources, and risk tolerances. Given the imperfect foresight regarding future developments of the electricity market and policies, the companies rely on uncertain forecasts to inform their decision making. Importantly, the long-term electricity price forecasts differ between companies. Each forecast is based on the assumptions of a specific companyregarding parameters like carbon and fuel prices or regulations. Methodology The long-term electricity price forecasts are calculated with a combination of models parametrizedaccording to the company’s assumptions. A fundamental energy system model estimates the capacity expansion in the electricity sector over several decades. The resulting technology mix is used for an initial dispatch simulation with the agent-based market model AMIRIS [4]. Plants that represent possible investments are included in the model. The dispatch simulation yields the expected power price and the optimal operation profile of the potential plants. These values areused to calculate the net present value (NPV) for each investment, a central metric for theinvestment decision. Companies make their individual investment decisions in annual rounds. The sum of all investments, the capacity expansion, contributes to the new technology mix of the electricity system. The dispatch of the electricity system with the agent-based capacity expansion is modeled in a second AMIRIS run with hourly resolution (Fig. 1). This is repeated for each simulated year. Plants compete on a market whose technology mix changes over time. This has effects on price formation and profitability. We track and assess the actual profitability of each investment. Plants can be shut Figure 1: The investment model determines the capacity expansion, which AMIRIS uses to simulate the electricity market for one year. The revenues are returned to the investment model, and the cycle repeats.down if they underperform and generate losses, e.g. because the realized power price is different from the forecasted one. Given this setup, our model allows to study 1) the impact of uncertainassumptions 2) possible emergent behaviors, and 3) the long-term implications and transition pathways on the electricity market.To conclude, we show that agent-based modeling can offer a fresh perspective on the effect of uncertainty on transition pathways. With our open-source software, we strive to provide the tools to spur this line of research forward. References [1] A. Stechemesser et al., ‘Climate policies that achieved major emission reductions: Global evidence from two decades’, Science, vol. 385, no. 6711, pp. 884–892, Aug. 2024, doi: 10.1126/science.adl6547.[2] M. Haugen et al., ‘Power market models for the clean energy transition: State of the art and future research needs’, Appl. Energy, vol. 357, p. 122495, Mar. 2024, doi: 10.1016/j.apenergy.2023.122495.[3] P. Barbrook-Johnson et al., ‘Economic modelling fit for the demands of energy decision makers’, Nat. Energy, vol. 9, no. 3, pp. 229–231, Mar. 2024, doi: 10.1038/s41560-024-01452-7.[4] C. Schimeczek et al., ‘AMIRIS: Agent-based Market model for the Investigation of Renewable andIntegrated energy Systems’, J. Open Source Softw., vol. 8, no. 84, p. 5041, Apr. 2023, doi: 10.21105/joss.05041. Acknowledgements This research has received funding from the German Federal Ministry for Economic Affairs and Climate Action under reference numbers 03EI1068A and 03EI1054C.