Importance subsampling for power system planning under multi-year demand and weather uncertainty

TL;DR

This paper presents importance subsampling, a novel method for reducing time series data in power system planning models, enabling accurate multi-year analysis under climate variability with significantly lower computational costs.

Contribution

The paper introduces importance subsampling, a new systematic approach to preserve extreme events in reduced datasets for long-term power system planning models.

Findings

Importance subsampling outperforms traditional clustering methods.

It enables multi-decadal analysis with reduced computational effort.

Open-source models and data support reproducibility.

Abstract

This paper introduces a generalised version of importance subsampling for time series reduction/aggregation in optimisation-based power system planning models. Recent studies indicate that reliably determining optimal electricity (investment) strategy under climate variability requires the consideration of multiple years of demand and weather data. However, solving planning models over long simulation lengths is typically computationally unfeasible, and established time series reduction approaches induce significant errors. The importance subsampling method reliably estimates long-term planning model outputs at greatly reduced computational cost, allowing the consideration of multi-decadal samples. The key innovation is a systematic identification and preservation of relevant extreme events in modeling subsamples. Simulation studies on generation and transmission expansion planning models illustrate the method's enhanced performance over established "representative days" clustering approaches. The models, data and sample code are made available as open-source software.