Beyond perfect foresight: Sequential investment optimization for district heating capacity expansion

Abstract

District heating operators must commit to irreversible, capital-intensive investments while facing deep uncertainty about energy prices, and network evolution toward lower supply temperatures with associated demand trajectories. Traditional deterministic planning, which assumes perfect foresight, systematically undervalues flexibility and biases technology selection. This study develops a multistage stochastic optimization framework for district heating capacity expansion using Stochastic Dual Dynamic Programming (SDDP). The framework integrates sequential investment decisions with a hybrid uncertainty structure combining structural system temperature pathways and Markovian energy price dynamics, while vintage tracking captures the retirement dynamics of existing generation assets. Application to Stockholm’s district heating network over a 2025–2050 horizon reveals substantial value from stochastic optimization: the Value of the Stochastic Solution reaches 27 % under risk-neutral evaluation and 39 % under tail-risk measures. The analysis uncovers three critical strategic mechanisms. First, deterministic planning under-invests in foundational hedging assets, leading to a 330 MW capacity deficit and extreme cost penalties if the High-Temperature pathway materializes. Second, the stochastic policy leverages brownfield flexibility, utilizing legacy bio-oil boilers as a capacity bridge to delay capital-intensive commitments until long-term trajectories are resolved. Third, a decision hierarchy emerges: “no-regret” technologies (data center and wastewater heat pumps) form the strategic foundation, while biomass CHP deployment remains conditional, scaling significantly only if the High-Temperature pathway is realized. These findings demonstrate that multistage stochastic optimization transfers effectively to district heating, providing quantitative tools for managing the transition to electrified heat supply under uncertainty.