Federated learning-based recurrent dynamic network for frequency control of networked microgrid system

An essential goal within power system operations is to ensure reliability, safe and high standard of power supply. A pivotal measure for achieving this goal is through the implementation of load frequency control (LFC). Essentially, LFC is an automated process designed to regulate the system frequency and net tie line power between a control area and its neighboring areas to align with their predetermined values of ±0.2 Hz as per the IEEE Standard 1547. Nowadays, the power systems are becoming more complex due to the increasing penetration of renewable energy sources (RES) and the transition of the conventional unilateral system to digitization through communication systems leading to the development of Microgrids and Networked Microgrids system. This transition leads to frequency instability due to the intermittent nature of RES. Any fluctuations in the load frequency that occur without a clear direction can cause damage to distributed generators, auxiliary engines, and other hardware equipment. In severe cases, such fluctuations can even pose a threat to the safety of the power grid itself. The primary focus of this thesis centers on addressing the issue of frequency deviation in a highly RE dominated power network and proposes a suitable control scheme with an appropriate control technique. Initially, the study is made considering a fractional-order proportional-integral-derivative (FOPID) controller designed in a centralized fashion, whose parameters are tuned using a proposed chaotic atom search optimization (CASO) for a RE integrated multi-area power system. However, this approach is deemed unsuitable for ensuring the necessary flexibility for a stable network operation in the event of an increase in Distributed Energy Resources (DERs) penetration. Furthermore, this scheme is susceptible to a single-point failure. To address the limitation of centralized control, a decentralized control strategy is employed using an improved version of FOPID controller called nonlinear FOPID (NLFOPID) controller. During this case, a hybrid atom search–particle swarm optimization (AS-PSO) approach is proposed to tune the parameters of the NLFOPID controller. However, a coordinated operation is impossible for a decentralized control scheme. This problem is addressed using distributed control strategy (DCS) based on a leader-follower framework for the coordination of multiple distributed energy resources (DERs) in NMGs. A chaotic hybrid AS-PSO optimized federated average learning of recurrent zeroing neural dynamics designed self-adaptive FOPID (FAL-ZND FOPID) controller is proposed for frequency regulation of NMG using DCS. Furthermore, the LFC is studied for prosumer and consumer-based P2P energy markets considering the proposed adaptive control technique. Finally, the real power system validation is carried out by utilizing a New England IEEE-39 test bus system. The results obtained proves that the proposed distributed adaptive controller improves the frequency response of the system by 65 %, 54 %, 45 % and 42 % in terms of settling time, peak overshoot and undershoot and also controller effort, respectively compared to decentralized and centralized control schemes.

Text FK 2023 25 - Declaration Form.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (1MB) Text FK 2023 25 - Full Text.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (9MB) Text FK 2023 25.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (9MB) Official URL or Download Paper: http://ethesis.upm.edu.my/id/eprint/18941

Actions (login required)

View Item