Temperature Prediction Model for Power Data Centers Based on GRU

With the rapid expansion of data center infrastructure, the energy consumption has become a major concern. Reducing data center power consumption and improving energy efficiency are crucial. This paper proposes a power-efficient data center temperature prediction model based on Gated Recurrent Unit ( GRU ) integrated with Convolutional Neural Network ( CNN ) and Attention Mechanism. The model leverages historical temperature data from a data center in Inner Mongolia as input features. High- dimensional features are extracted using a one-dimensional CNN, and the extracted feature vectors are fed into the GRU for analysis and feature extraction. Additionally, the Attention Mechanism is employed to compute attention scores for the hidden layers of GRU and allocate attention weights, highlighting the influence of highly correlated features. The model ultimately achieves temperature prediction. Experimental evaluations are conducted using a temperature dataset provided by the Inner Mongolia data center, and comparisons with other neural network models are made. The results demonstrate that the proposed model exhibits better convergence and higher accuracy.