Portfolio optimization with irreversible long-term investments in renewable energy under policy risk: A mixed-integer multistage stochastic model and a moving-horizon approach

• Published in: European journal of operational research Vol. 290; no. 2; pp. 734 - 748
• Main Authors: Gatzert, Nadine, Martin, Alexander, Schmidt, Martin, Seith, Benjamin, Vogl, Nikolai
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 16.04.2021
• Subjects: Irreversible investments; Mixed-integer optimization; Multistage stochastic optimization; Policy risk; Portfolio optimization; Portfolio optimization; 90Bxx; Multistage stochastic optimization; Irreversible investments; 90C11; Policy risk; 90C06; 90C90; 90C15; Mixed-integer optimization
• ISSN: 0377-2217; 1872-6860
• DOI: 10.1016/j.ejor.2020.08.033

•We present a mixed-integer multistage stochastic model that includes investment opportunities in illiquid and long-term infrastructure projects in the context of renewable energies, which are also subject to policy risk.•We present a tailored moving-horizon approach together with suitable approximations and simplifications of the model.•We evaluate these approximations and simplifications in a computational sensitivity analysis.•We derive a final version of the model that can be tackled on a realistic instance by our moving-horizon approach. Portfolio optimization is an ongoing hot topic of mathematical optimization and management science. Due to the current financial market environment with low interest rates and volatile stock markets, it is getting more and more important to extend portfolio optimization models by other types of investments than classical assets. In this paper, we present a mixed-integer multistage stochastic model that includes investment opportunities in irreversible and long-term infrastructure projects in the context of renewable energies, which are also subject to policy risk. On realistic time scales for investment problems of this type, the resulting instances are by far too large to be solved with today’s most evolved optimization software. Thus, we present a tailored moving-horizon approach together with suitable approximations and simplifications of the model. We evaluate these approximations and simplifications in a computational sensitivity analysis and derive a final model that can be tackled on a realistic instance by our moving-horizon approach.