Integrated Optimization of Photovoltaic Systems and Waste Heat Recovery in Data Centers

Sammanfattning

Data centers (DC) consume large amounts of electricity, of which virtually all is ultimately dissipated as heat. This study investigates the integrated potential of combining DC waste heat recovery (HR) with conventional photovoltaic (PV) and hybrid photovoltaic-thermal (PVT) solar installations at the TOFAS automotive manufacturing complex in Bursa, Turkey, a large industrial facility with substantial electricity and thermal demands and an extensive available rooftop area. The analysis is conducted using a MATLAB-based computational tool developed as part of this work, which combines energy balance modelling, techno-economic evaluation, and multi-objective Pareto optimization that finds a balance between the economic and the environmental perspectives, applied to real monthly operational data provided by the facility. The HR assessment shows that, despite having a system efficiency (defined as the fraction of data centre electricity consumption ultimately delivered as useful heat after accounting for all recovery chain losses) above 60%, the recovered heat covers less than 1% of the complex's total thermal demand. The DC is small relative to the industrial complex it serves, and the economic case for HR in isolation is weak at this scale. Its value lies primarily in the consistent, low-effort CO₂ savings it generates once installed. The solar optimization results show that conventional PV strongly dominates under economic criteria, driven by the higher value of displaced grid electricity relative to displaced natural gas, while PVT receives only a marginal allocation even under environmental optimization due to the seasonal mismatch between peak solar thermal output and minimum heating demand. The multi-objective optimization reveals that moderate expansions beyond the economic optimum can achieve substantially greater emissions reductions at relatively small additional cost. The Pareto front provides decision-makers with a transparent view of the trade-off between net present value and avoided CO₂ emissions across the full range of feasible configurations, rather than prescribing a single solution. The computational tool is designed for reuse across other industrial facilities through adjustable input parameters and an interactive graphical interface.