Title: DDQN-based data laboratory energy consumption control model

Authors: Hui Cao; Xin Xu; Chenggang Li; Hongda Dong; Xiangyu Lv; Qi Jin

Addresses: Jilin Provincial Electric Power Science Research Institute Co., Ltd., Changchun 130000, China ' Jilin Provincial Electric Power Science Research Institute Co., Ltd., Changchun 130000, China ' Jilin Provincial Electric Power Science Research Institute Co., Ltd., Changchun 130000, China ' Jilin Provincial Electric Power Science Research Institute Co., Ltd., Changchun 130000, China ' Jilin Provincial Electric Power Science Research Institute Co., Ltd., Changchun 130000, China ' Computer Engineering College, Northeast Electric Power University, Jilin 132012, China

Abstract: With the rapid development of information technology, data laboratory plays an essential supporting role in various fields. A lot of energy consumption also accompanies its operation, and the intuitively controllable part is the air conditioning energy consumption expenditure. Our work is based on the sensor data of data centre infrastructure management (DCIM). We collect data such as server temperature and humidity, users, and data transmission rate, but we need more specific data on air conditioning energy consumption. To this end, we used the computational fluid dynamics (CFDs) model to simulate the time series data of air conditioning energy consumption, used the improved regression analysis method to align it with the sensor data, and constructed a complete time series data set of air conditioning energy consumption. Finally, based on double deep Q-network (DDQN), an improved neural network, we constructed an energy consumption control model for the data laboratory to improve energy efficiency to the greatest extent. The experimental results thoroughly verify the effectiveness of our proposed model, which achieves remarkable results in energy consumption control and provides a new concept and method for energy conservation.

Keywords: deep reinforcement learning; energy consumption control; computational fluid dynamics; CFDs; data feature engineering.

DOI: 10.1504/IJSNET.2024.137338

International Journal of Sensor Networks, 2024 Vol.44 No.3, pp.157 - 168

Received: 08 Oct 2023
Accepted: 12 Oct 2023
Published online: 12 Mar 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article