Energy System for a Data Center with 100% Hourly Renewable Utilization

Abstract

The Paris Agreement aims to limit global warming to 2 degrees Celsius by reducing carbon dioxide emissions. Data centers are essential to the digital world and contribute approximately 1% to global electricity consumption. Major tech companies like Amazon, Microsoft, Google, and Meta are large contributors to this consumption. In countries like the Netherlands, this contribution is higher due to the Netherlands being a digital hub, accounting for 3% of the total national electricity consumption. Considering their significant energy usage, it is interesting to research how data centers can transition to using only renewable energy sources on an hourly basis so they can decrease their footprint. This thesis aims to identify a feasible and cost-effective configuration for a 100 MW data center that operates 100% on renewable energy on an hourly basis. The research will assess the environmental impacts, focusing on CO2 emissions (CO2e) and land usage, with the main goal of achieving the lowest Levelized Cost of Electricity (LCOE). First, an overview of potential energy storage technologies is provided to find suitable solutions. Based on different criteria, Lithium-ion (Li-ion), lead-acid, vanadium redox flow batteries (VRFB), zinc-bromide flow batteries (ZBFB), alkaline and PEM (proton exchange membrane) electrolyzers, storage tanks, and PEM fuel cells are chosen. With these technologies, eight different configurations are composed, combining various battery types and electrolyzer types with the storage tank and PEM fuel cell: Li-ion with alkaline (1), lead-acid with alkaline (2), Li-ion with PEM (3), lead-acid with PEM (4), VRFB with alkaline (5), ZBFB with alkaline (6), VRFB with PEM (7), and ZBFB with PEM (8). An energy model is developed to identify the most feasible and cost-effective solution, using solar, wind, and load profiles as input. The assets are modeled based on their technical parameters and limitations. A 10-year simulation of the hourly energy flows is executed. In this simulation, the constraints, degradation, and replacement of the assets are taken into account. The costs are determined based on Capital Expenditures (CAPEX), Operating Expenses (OPEX), replacement costs, and residual costs of the asset after 10-year. The connection costs are also considered. The configuration dimensions are optimized based on the LCOE using the particle swarm optimization method. The two configurations with the lowest LCOE are evaluated for environmental impacts, including CO2e and land usage. The study shows that combining a Li-ion battery or VRFB with an alkaline electrolyzer is the most cost-effective and feasible configuration for a data center. The LCOE is 0.243 and 0.241 €/kWh, respectively. Additionally, this reduces kg CO2e/kWh by 7-8 times compared to the current grid emissions in the Netherlands, which are approximately 0.037 and 0.032 kg CO2e/kWh. However, the required land usage is approximately 20 km², and the LCOE of the configuration is about 2.5 times higher than the current power purchase agreement (PPA) price for a data center, which is 0.10 €/kWh. Therefore, it is recommended to initially start with an 80% renewable scenario using only a Li-ion battery and renewable energy sources. This approach reduces the LCOE to 0.12 €/kWh, which is approximately the same as the current PPA price if an additional 0.01 €/kWh for the connection to the grid is included. Furthermore, kg CO2e/kWh is three times lower compared to the current situation, which is 0.08 kg CO2e/kWh. This strategy also mitigates the potential risks associated with the infancy of VRFB and the combination of intermittent renewable sources and an electrolyzer, as these are excluded in this scenario.