Data center server energy consumption prediction method based on non-redundant feature selection

The invention discloses a data center server energy consumption prediction method based on non-redundant feature selection. The method comprises the following steps: (1) selecting strong correlation features in a server energy consumption original feature set; (2) selecting a non-redundant feature set from the strong correlation features by using a redundant feature determination algorithm; and (3) mining the incidence relation between the non-redundant characteristics and the energy consumption by using the gating circulation unit neural network, and constructing a prediction analysis model of the energy consumption of the server. According to the data center energy consumption management system based on deep learning, operation and maintenance personnel of the data center can visually and accurately master key factors influencing the energy consumption of the server and better analyze and predict the load change and the energy consumption trend during the operation of the server; based on the central server