This invention discloses a low-carbon control method for the linkage of
waste heat recovery and
air conditioning system in data centers, relating to the field of energy-saving and low-carbon control technology for data centers. The specific steps of this method are as follows: S1, collecting real-time and historical operating data of
server racks, central cold and heat sources,
pipe networks, outdoor environment, and
power grid, standardizing and cleaning them to form a full-chain
thermal state dataset; S2, using the full-chain
thermal state dataset from step S1 to build a time-series neural network for prediction. The prediction model input layer receives the time-series sequence of
server rack computing
power load, the time-series sequence of
server rack
waste heat medium temperature and flow rate, and the time-series sequence of
outdoor temperature and
humidity as input features to predict future operating parameters. This invention constructs a complete
thermal state dataset through full-chain multi-dimensional data collection and standardized
processing, combined with a rolling time-series prediction model to predict the changing trends of
server room heat, heat exchange, and outdoor environment in advance, and dynamically divides the
waste heat value time period based on the waste heat value quantification model to form a global cold and heat resource
prediction system.