DeepCCHP: Intelligent Comprehensive Optimization of Energy Consumption and Carbon Emission for Data Center CCHP Systems

The existing data center energy saving and carbon emission reduction optimization methods do not comprehensively consider the carbon footprint involved in the coupling of energy input, production energy consumption, and waste utilization, making it difficult to achieve systematic energy saving and carbon emission reduction. To this end, a deep reinforcement learning (DRL) based optimization algorithm, DeepCCHP, is proposed for data center Combined Cooling Heating and Power (CCHP) systems to jointly control the power supply system and the cooling subsystem thereby optimizing the costs of energy, Carbon Dioxide Emissions (CDE) and Power Usage Effectiveness (PUE). DeepCCHP combines the Long and Short-term Time-series Network-Attention (LSTNet-Attn)) and a DRL approach to solving the joint optimization problem to achieve the joint control of the power generation unit and chiller equipment proactively. It is trained and validated with Alibaba data center cluster data in a simulation environment based on Trnsys software. Experimental results show that the DeepCCHP algorithm can save up to 40% of the cost and 28% of the carbon emission compared to the baseline algorithm, and can achieve a better compromise and balance between energy cost, carbon emission, and energy efficiency.