An improved two-stage optimization for network and load recovery during power system restoration

• Published in: Applied energy Vol. 249; pp. 265 - 275
• Main Authors: Liao, Shiwu, Yao, Wei, Han, Xingning, Fang, Jiakun, Ai, Xiaomeng, Wen, Jinyu, He, Haibo
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.09.2019
• Subjects: Load restoration; Network restoration; Power system restoration; Two-stage optimization; Network restoration; Two-stage optimization; Load restoration; Power system restoration
• ISSN: 0306-2619; 1872-9118
• DOI: 10.1016/j.apenergy.2019.04.176

•Proposed an improved two-stage optimization method for network and load recovery.•Decouples integer and continuous decision variables and solves them separately.•Proposed method significantly can improve the computational efficiency.•Verified the advantage in computational efficiency and plan optimality.•Demonstrated potential application through simulation in a practical system. Network and load recovery (NLR) during power system restoration is a multi-step mixed-integer nonlinear programming (MINP) problem. The NLR is difficult to be solved as it is NP-hard. Thus, NLR is commonly solved step by step, which is short-sighted and will result in longer restoration time. To obtain a far-sighted NLR plan, this paper proposes an improved two-stage optimization method for NLR. The first stage adopts a mixed-integer linear programming (MILP) model to obtain optimal solutions of integer variables in NLR, namely the load pick-up schedules and transmission line charging schedules. Then in the second stage, a continuous non-linear optimization method based on AC power flow with frequency constraints and load model is established to minimize the restoration duration of the plan generated in the first stage step by step. Case studies are undertaken on a 10-machine 39-bus system and Southeast Hubei Provincial power system of China. Simulation results indicate that the restoration plan obtained from the improved two-stage optimization method is highly effective, while the computational efficient meets the intensive need for restoration scheduling after blackouts.