Operation state calculation (OSC) provides safe operating boundaries for power systems. The operators rely on the software-aid OSC results to dispatch the generators for grid control. Currently, the OSC workload has increased dramatically, as the power grid structure expands rapidly to mitigate renewable source integration. However, the OSC is processed with a lot of manual interventions in most dispatching centers, which makes the OSC error-prone and personnel-experience oriented. Therefore, it is crucial to upgrade the current OSC in an automatic mode for efficiency and quality improvements. An essential process in the OSC is the tie-line power (TP) adjustment. In this paper, a new TP adjustment method is proposed using an adaptive mapping strategy and a Markov Decision Process (MDP) formulation. Then, a model-free deep reinforcement learning (DRL) algorithm is proposed to solve the formulated MDP and learn an optimal adjustment strategy. The improvement techniques of ''stepwise training'' and ''prioritized target replay'' are included to decompose the large-scale complex problems and improve the training efficiency. Finally, five experiments are conducted on the IEEE 39-bus system and an actual 2725-bus power grid of China for the effectiveness demonstration. INDEX TERMS Operation state calculation, tie-line power adjustment, deep reinforcement learning, stepwise training, prioritized target replay.
The tie-line power adjustment is an essential part of the power system operation state calculation. Various existing algorithms to solve the tie-line power adjustment problem are mainly implemented by introducing tie-line power equation constraints into conventional power flow calculations. Such methods have low calculation efficiency, not enough automation, and are prone to non-convergence in the power flow calculation. In this paper, the tie-line power adjustment problem is formulated as a Markov decision process, and the proximal policy optimization algorithm is introduced to optimize the decision policy. In order to enhance the effectiveness of the proposed method, a new deep neural network structure suitable for the proximal policy optimization algorithm is designed. The proposed method is verified with the IEEE 39-bus system.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.