Investigation of Adapted Data Center Load Balancing Algorithms for Battery Charge Distribution in Direct Current Nanogrid Systems
Résumé
This research explores the design and performance evaluation of a direct current (DC) nanogrid system powered exclusively by renewable energy sources, with a particular focus on solar energy. The study is conducted in a highly remote rural location, specifically a cirque on Réunion Island, where conventional power infrastructure is limited or unavailable. The nanogrid architecture consists of three main components: a photovoltaic (PV) energy source, a battery storage system, and a consumer load. The system operates entirely in DC, ensuring energy efficiency and minimizing conversion losses. The evaluation is based solely on simulations, without hardware implementation, and considers a network composed of four nanogrids operating under worst-case conditions of variable solar generation and consumption. Several load balancing algorithms, adapted from server management techniques such as Round Robin and Least Connection, are applied to the energy distribution process. A new algorithm, Weighted SOC Round Robin (WSRR), was developed to account for the batteries’ state of charge. Results show that WSRR reduces power outages by up to 45% and lowers energy underload by 25% compared to baseline approaches. The algorithm ensures more stable and efficient energy delivery, highlighting the relevance of simple, adaptive strategies for managing decentralized renewable energy systems in isolated environments. This study contributes to the development of robust control methods for renewable-powered nanogrids, addressing both energy autonomy and system resilience in remote areas.