Optimal control of multi-bus DC microgrids based on distributed dual-projection-layer recurrent neural network considering bus voltage regulation

Yuanyuan Zhu; Fan Yang; Guoyu Lin

doi:10.1016/j.isatra.2025.06.029

Optimal control of multi-bus DC microgrids based on distributed dual-projection-layer recurrent neural network considering bus voltage regulation

ISA Trans. 2025 Jun 26:S0019-0578(25)00327-1. doi: 10.1016/j.isatra.2025.06.029. Online ahead of print.

Authors

Yuanyuan Zhu¹, Fan Yang², Guoyu Lin³

Affiliations

¹ College of Mechanical & Electronic Engineering, Nanjing Forestry University, Nanjing, 210037, PR China. Electronic address: 3230300268@njfu.edu.cn.
² College of Mechanical & Electronic Engineering, Nanjing Forestry University, Nanjing, 210037, PR China. Electronic address: yfan0510@njfu.edu.cn.
³ Department of Intelligent Equipment, Changzhou College of Information Technology, Changzhou, 213164, PR China. Electronic address: linguoyu@ccit.js.cn.

PMID: 40579341
DOI: 10.1016/j.isatra.2025.06.029

Abstract

With the broad application of plug-and-play loads, it brings new challenges to conventional control issues in DC microgrids. This work addresses the joint optimization of generation costs and transmission line power losses considering bus voltage regulation. Specifically, the concept of virtual load nodes is first introduced so that loads can be implemented as plug-and-play. Through Kron Reduction, bus voltage of load nodes is indirectly controlled by DG nodes. Then, a distributed dual-projection-layer recurrent neural network (DRNN) is proposed for real-time optimal control. The coupled voltage and current are simultaneously maintained within safe bounds. By using Lyapunov synthesis, the convergence of the DRNN is demonstrated. The effectiveness of the proposed methods is evaluated by simulations in terms of plug-and-play test and comparative analysis.

Keywords: DC microgrids; Neurodynamic optimization; Real-time control; Recurrent neural network; Voltage regulation.