Modelling and analysis of multi-timescale uncertainty in energy system planning

TL;DR

This paper models and analyzes the impact of short-term and long-term uncertainties on energy system planning, demonstrating that multi-timescale uncertainty significantly influences planning decisions and solution stability, with a new parallel Benders decomposition method improving computational efficiency.

Contribution

It introduces a comprehensive analysis of multi-timescale uncertainty in energy planning and proposes a parallel stabilised Benders decomposition method for improved solution speed.

Findings

Including multi-timescale uncertainty increases the value of stochastic solutions.

Long-term uncertainty impacts solution structure more than cost coefficient uncertainty.

Parallel stabilised Benders decomposition is up to 7.5 times faster than serial methods.

Abstract

Recent developments in decomposition methods for multi-stage stochastic programming with block separable recourse enable the solution to large-scale stochastic programs with multi-timescale uncertainty. Multi-timescale uncertainty is important in energy system planning problems. Therefore, the proposed algorithms were applied to energy system planning problems to demonstrate their performance. However, the impact of multi-timescale uncertainty on energy system planning is not sufficiently analysed. In this paper, we address this research gap by comprehensively modelling and analysing short-term and long-term uncertainty in energy system planning. We use the REORIENT model to conduct the analysis. We also propose a parallel stabilised Benders decomposition as an alternative solution method to existing methods. We analyse the multi-timescale uncertainty regarding stability, the value of the stochastic solution, the rolling horizon value of the stochastic solutions and the planning decisions. The results show that (1) including multi-timescale uncertainty yields an increase in the value of the stochastic solutions, (2) long-term uncertainty in the right-hand side parameters affects the solution structure more than cost coefficient uncertainty, (3) parallel stabilised Benders decomposition is up to 7.5 times faster than the serial version.