(710b) Optimal Design and Operation of Resilient Power Distribution Systems

Power systems are increasingly subject to uncertain stressors that threaten their ability to deliver electricity to end-users. Some uncertain stressors include natural disasters that can damage equipment and cause line faults, large fluctuations in renewable power generation, and uncertain power demands or loads. When power systems are not resilient to these stressors, there can be dire consequences; for example, the 2021 Texas winter storm caused long-term blackouts and ultimately cost at least 246 people their lives (Svitek, 2022). Hence, it is essential to enhance the resilience of power systems through measures such as equipment improvements, redundancy, and energy storage.

In this work, we develop a two-stage stochastic programming framework for the design and retrofit of power distribution systems to improve their resilience to a variety of uncertain stressors. While most existing works only focus on one particular type of stressor, we consider simultaneously uncertainty in line faults, renewable generation, locational marginal prices paid by the system operator, and power demand. We assume that possible resilience-enhancing investments include line hardening projects, mobile battery storage systems, and mobile renewable-ammonia-based energy storage systems (an emerging technology for chemical energy storage) (Riley et al., 2023). Once a specific realization of the uncertainty is observed, operational decisions, including line switching procedures, are adjusted in response to the conditions in that scenario.

Given the multitude of uncertain parameters considered in the problem, a large number of scenarios is required to accurately model the uncertainty. For a power distribution system of relevant size, the resulting two-stage stochastic mixed-integer linear program (MILP) can usually not be solved directly using an off-the-shelf MILP solver. Several measures are taken to improve the computational tractability of the problem: (i) We implement a scenario generation and reduction scheme that uses several sources of historical data to improve the accuracy of the estimated probability distribution while maintaining a manageable number of scenarios. (ii) We derive additional valid inequalities that result in a tighter MILP formulation. (iii) We apply a tailored decomposition method, namely a modified integer L-shaped method with additional problem-specific cuts. Large instances of the stochastic programming model, containing on the order of 100 million variables and constraints, are solved using the proposed solution algorithm in a case study on a modified IEEE 123-bus system (Hosseini et al., 2023). This case study is performed to investigate the optimal investments into and use of resilience-enhancing measures in a realistic test system that faces several uncertain stressors. Additionally, the results of the case study are used to assess the techno-economic feasibility of coordinating chemical energy storage systems with more mature technologies for resilience enhancement.

References

Hosseini, M. M., Rodriguez-Garcia, L., & Parvania, M. (2023). Hierarchical Combination of Deep Reinforcement Learning and Quadratic Programming for Distribution System Restoration. IEEE Transactions on Sustainable Energy, 14(2), 1088–1098.

Riley, B. P., Daoutidis, P., & Zhang, Q. (2023). Multi-scenario design of ammonia-based energy storage systems for use as non-wires alternatives. Journal of Energy Storage, 73(PA), 108795.

Svitek, P. (2022, January 2). Texas winter storm official death toll now put at 246. The Texas Tribune. https://www.texastribune.org/2022/01/02/texas-winter-storm-final-death-t...