Symbolic Regression for Modelling Decarbonisation Pathways in the Global Energy-Economy-Climate System

The ETSAP-TIAM Model is a process-based optimisation model of the global energy system, integrated with a climate and economy model that explicitly represents detailed energy technology processes including generation, transmission, and end-use across sectors. It provides a tool to inform policy by examining the effects of policy changes on outcomes. However, the model is complex, large, and unwieldy. In this paper, we develop a suite of Genetic Programming Symbolic Regression (SR) surrogate models. The benefits are both interpretability and instant response of the new model to parameter changes. Considering both regression performance and model complexity, we compare a SR system with strong baselines, exploring different points on the performance—interpretability curve. Finally, we read and interpret several of our models.