Robust energy management for a smart grid with frequency regulation service

Abstract

With the development of a modernized power grid, a smart grid, functioning as the next-generation power grid, is integrated with various renewable energy sources (RESs). In such smart grid, energy management plays a critical role on improving the utilization of RESs. However, due to the uncertainty and intermittency of RESs, the stability issue of the smart grid occurs. To handle this issue, a robust energy management with the frequency regulation service is designed for providing reliable and stable operations in a smart grid. Such energy management technologies can be divided into two aspects, namely V2G scheduling and regulation reserve. Specifically, vehicle-to-grid (V2G) technology aims to utilize electric vehicles (EVs) to perform active power balancing with the frequency regulation service. Furthermore, day-ahead regulation reserve can be procured for operating the active power compensation in the system. In this thesis, we focus on the design of the robust energy management approaches with the frequency regulation service for the smart grid system.

First, we propose a hierarchical system framework for joint optimal power flow (OPF) routing and V2G scheduling with the regulation service. At the grid level, the OPF routing problem is formulated through the semidefinite programming (SDP) relaxation. Then, at the EV level, both decentralized forecast-based and online scheduling algorithms are devised to control EVs schedules by providing the V2G regulation service. The numerical results show that both the devised V2G scheduling algorithms can help flatten the power fluctuations at the buses attached with EVs so as to stabilize the power grid effectively. In addition, for different systems, the use of PFRs both diminishes the system power loss greatly and provides voltage regulation.

In addition, we propose a novel smart cross-system framework to leverage on the interactions and dependencies between the smart transportation system and city smart grid system through information sharing. More specifically, we formulate a joint problem to solve allocation and V2G scheduling for electric buses (EBs) in a city. We utilize the deep learning approach to account for the short-term traffic conditions in the city. After that, the joint problem is formulated as a mixed-integer quadratic programming (MIQP) problem, and it can be decoupled into each EB subproblem through the Lagrangian dual decomposition. The distributed algorithm is then devised to solve each EB subproblem in a scalable manner. The numerical results present the efficacy of the proposed joint framework such that the quality of the V2G regulation service can be further improved.

Finally, we propose a deep learning approach to estimate the frequency regulation reserve in a general distribution system. First, we consider multiple features in the distribution system, such as RESs, loads, and city weather information. Second, we obtain the net active power imbalance and the power matrix of a general distribution system through the power flow model. Third, after data-preprocessing stage, we deploy the long short-term memory (LSTM) to estimate the frequency regulation reserve. The numerical results show that the accurate estimation on the frequency regulation reserve can be achieved through our proposed deep learning model.