Forecasting-aided state estimation based on deep learning for hybrid AC/DC distribution systems. (15th January 2022)

• Record Type: Journal Article
• Authors: Huang, Manyun; Wei, Zhinong; Lin, Yuzhang
• Is Part Of: Applied energy. Volume 306:Part B(2022)
• Publication Date: 2022-01-15
• Subjects: Forecasting-aided state estimation -- Deep learning ensemble Kalman filter -- Hybrid AC/DC distribution systems; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2021.118119
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: The forecasting-aided state estimation (FASE) approach is presented for hybrid AC/DC distribution systems in the first time. A distributed FASE framework based on the constrained EnKF is designed to enhance computational efficiency and reduce the communication burden. A physics aware deep learning-based state forecasting model can capture the complex and nonlinear relationship of system states with the aid of historical data. Features of the proposed method have been illustrated on a sample system and its scalability is demonstrated on a larger hybrid system. Abstract: To accommodate a higher penetration of distributed energy resources, distribution systems are moving toward hybrid AC/DC configurations for secure and economic operation. In this regard, this paper proposes a forecasting-aided state estimator (FASE) for hybrid AC/DC distribution systems to obtain accurate estimates for online security monitoring and control. The proposed FASE is designed in a distributed framework, with decomposition into several subproblems and solution by a constrained ensemble Kalman filter algorithm. In the proposed methodology, a deep neural network-based state forecasting model is developed to imitate the complex temporal and spatial relationship between system states, avoiding the state transition model built by unfounded explicit formulations. Furthermore, smart meter data is integrated by deep regression learning to obtain power injections of consumers and address the system observability issue. Extensive comparisons with two alternatives are carried out on a sample 33-node hybrid AC/DC distribution system to show the effectiveness and benefits of the proposed FASE, and on a larger 106-node hybrid AC/DC distribution system to demonstrate scalability.