Low-regret Strategies for Energy Systems Planning in a Highly Uncertain Future

TL;DR

The paper addresses substantial uncertainty in energy system planning by introducing a regret-based decision-support framework that automatically identifies, evaluates, and visualizes low-regret strategies. It applies the framework to Switzerland's biomass allocation under a 2050 net-zero target, considering eight biomass pathways and comparing outcomes across a large ensemble of futures. The analysis identifies five robust strategies, with Fuel&Chemicals typically minimizing regret across criteria, while continued use of biomass for low-temperature heat positions as high-regret; the framework's decision trees, cumulative regret curves, and decision maps offer actionable, uncertainty-aware guidance. The approach is generalizable to other energy-system decisions, enabling policymakers to weigh trade-offs under uncertainty and design more resilient transition pathways.

Abstract

Large uncertainties in the energy transition urge decision-makers to develop low-regret strategies, i.e., strategies that perform well regardless of how the future unfolds. To address this challenge, we introduce a decision-support framework that identifies low-regret strategies in energy system planning under uncertainty. Our framework (i) automatically identifies strategies, (ii) evaluates their performance in terms of regret, (iii) assesses the key drivers of regret, and (iv) supports the decision process with intuitive decision trees, regret curves and decision maps. We apply the framework to evaluate the optimal use of biomass in the transition to net-zero energy systems, considering all major biomass utilization options: biofuels, biomethane, chemicals, hydrogen, biochar, electricity, and heat. Producing fuels and chemicals from biomass performs best across various decision-making criteria. In contrast, the current use of biomass, mainly for low-temperature heat supply, results in high regret, making it a must-avoid in the energy transition.

Figures (5)

• Figure 1: Scatterplot showing the distribution of biomass flow across all 1000 scenarios into different usage options (i.e., outputs of interest) of the energy system. The scatterpoints of each output of interest are randomly spread along the x-axis to enhance visualization. The outputs of interest are ordered from left to right based on their range of observed values across the scenarios.
• Figure 2: Decision tree applied to the space of the outputs of interests. Decisions are indicated by the symbols in the tilted squares, with the corresponding split given below each box. The axes of the radar plots show the biomass allocation to the specific sectors. Each axis is normalized to the maximum observed value of the respective output of interest. The grey-shaded area in each node indicates the range between the minimum and maximum observed value of each output of interest. Each $N$ represents the number of scenarios contained in a leaf node. The leaf nodes of the decision tree return five strategies for the use of biomass. The solid colored line indicates the average value of the outputs of interests in the leaf node. We name the strategies by the biomass allocation that distinguishes them from the other leaves. Hence, from now on, we refer to the five strategies as Chemicals, Hydrogen, Biomethane, Biofuel, and Fuel&Chemicals strategy.
• Figure 3: The cumulative regret curves display the regret of the biomass strategies across all 1000 scenarios. Leftward arrows highlight the y-intercept, indicating the number of scenarios in which a strategy is the optimal choice. Downward arrows mark the minimal regret for strategies that are never optimal, while upward arrows show the maximum regret for each strategy. Crosses represent the average regret across all scenarios. The value-at-risk VaR$_{\alpha}$ for a strategy is determined by selecting a confidence level $\alpha\in[0,100]$ on the y-axis and identifying the corresponding regret value on the x-axis. This regret value represents the level that will be exceeded in (100-$\alpha$)% of scenarios. Solid lines represent the cumulative regret of strategies derived from the decision tree (Fig. \ref{['fig: strategy_tree']}), while dashed lines show the regret curves for the additionally evaluated Business-as-Usual (BAU) and No Biomass strategies.
• Figure 4: The Pearson correlation coefficients indicate the sensitivity of different strategies' regret to variations in input parameter values. Positive correlations (red) suggest that higher input parameter values increase regret, whereas negative correlations (blue) imply that higher values reduce regret. The five input parameters with the highest correlations (ordered by the average of absolute values across all strategies) are displayed for the five strategies derived from the decision tree.
• Figure 5: The figure illustrates the dependence of regret for the biomass strategies on the five most influential parameters. The diagonal panels display the dependence on individual parameters, while the upper triangular panels present decision maps for pairwise parameter combinations. These decision maps highlight the conditions under which one strategy outperforms the others, with the color of a pixel representing the strategy with the lowest mean regret under the given conditions. Each parameter range is divided into $n=7$ equally sized bins, resulting in an average of $N/n = 1000/7 \approx 143$ scenarios per bin and $N/(n\times n) = 1000/49 \approx 20.4$ scenarios per pixel. In the diagonal panels, the mean regret (solid line) and the standard error of the mean (shaded area) are shown for each bin.