Scheduling Electricity Production Units to Mitigate Severe Weather Impact: An Efficient Computational Implementation

TL;DR

This paper presents a robust optimization framework for scheduling electricity production units to minimize load shedding during severe weather events, incorporating detailed physical laws and an efficient solution algorithm.

Contribution

It introduces a more precise physical model and a highly efficient algorithm for a complex tri-level mixed-integer nonlinear optimization problem.

Findings

The proposed model effectively minimizes load shedding under worst-case scenarios.

The solution algorithm outperforms standard solvers in computational efficiency.

Solutions are obtained within a reasonable time for moderately large instances.

Abstract

In the electric system, extreme weather events can cause trips or physical damage to transmission lines, leading to large-scale load shedding. To mitigate power shedding, we propose a framework that pre-positions the commitment of production units--particularly slow-start units--to cope with transmission topologies that may result from such events. Our goal is to minimize load shedding under the worst-case scenario. The novel contributions of this paper are twofold: (1) a more precise description of the physical laws than those used in previous works reported in the literature, and (2) a highly efficient solution algorithm compared to state-of-the-art, off-the-shelf solvers. We formulate this framework as a two-stage robust optimization model. In the first stage, generation units are scheduled, and in the second stage, power dispatch decisions are made to minimize load shedding under the worst-case scenario. Convexified AC power flow constraints are incorporated to ensure system reliability and security. The resulting formulation is a tri-level mixed-integer nonlinear optimization problem. To address the computational challenges, we propose a problem-specific outer approximation algorithm embedded within a column-and-constraint generation framework. Computational results demonstrate that our model and algorithm can produce solutions within a standard optimality gap in a reasonable time for moderately large instances.