Sizing renewable energy microgrids for supercomputing centers using particle swarm optimization

Abstract

High-performance computing (HPC) centers face growing challenges in managing energy consumption, operational costs, and environmental impact due to their continuous, high-density loads. This study presents a data-driven optimization framework for sizing hybrid renewable energy microgrids tailored to the specific demands of HPC facilities. Using one-minute resolution data collected over a full year from the Data Analysis and Supercomputing Center (CADS) at the Universidad de Guadalajara, we applied Particle Swarm Optimization (PSO) to determine the optimal number of photovoltaic (PV) panels and wind turbines.

The optimization simultaneously minimizes total system cost, energy deficit, and operational penalties under strict reliability constraints. Results show that a PV-only system comprising 960 panels (no wind turbines) yields the optimal configuration, covering 75% of the annual energy demand. The solution achieves a Levelized Cost of Energy (LCOE) of 1.42 MXN/kWh and a payback period of 7.8 years. Due to low average wind speeds (2.1 m/s), wind turbines were excluded. A comparative validation using a Genetic Algorithm (GA) confirmed the robustness of the PSO solution, with a cost deviation below 1.3%.

This study highlights the importance of high-resolution real-world data and site-specific modeling in optimizing microgrids for mission-critical environments. The proposed framework is generalizable and supports informed decision-making for renewable integration in data-intensive infrastructure.