Abstract
This paper presents a climate-aware stochastic capacity expansion planning (CEP) framework that addresses the growing need for robust policy and investment decisions towards high renewable penetration, giving due consideration to climate variability. The developed model incorporates demand, renewable generation, and hydro inflows for 35 climate years, preserving intra-annual variability, correlation of resources, and variance of climate conditions. To address the computational burden of fully stochastic models, several clustering techniques are assessed, namely k-means, k-medoids, hierarchical, and a novel cost-driven approach. Unlike traditional clustering methods, the cost-driven approach clusters climate years according to their impact on system cost, delivering near-optimal results with more than a 90% reduction in computation time. The study also explores whether deterministic model runs, i.e. solving for each climate year individually, can approximate the stochastic optimum. A percentile-based analysis shows that energy mixes derived from intermediate-to-upper percentiles of the capacity envelope can approximate optimal cost within about 3-5%. Overall, these findings highlight the importance of considering multiple climate years and selecting appropriate scenario reduction techniques to ensure robust and cost-effective system planning.

This publication has 60 references indexed in Scilit: