Air cooling data center energy saving method based on deep Q learning conclusion network

Abstract

The invention discloses an air cooling data center energy-saving method based on a deep Q learning decision network. The method comprises the following steps: S1, acquiring environment states of a sampling data center and a monitoring data center; and S2, inquiring environment states of the sampling data center and the monitoring data center periodically, performing training of a deep Q learning decision network, and generating the temperature of the energy-saving air conditioner set value. The method has the beneficial effects that self-learning of the controller is carried out by adopting a reinforcement learning mode, the controller can adapt to environmental changes by setting rewards and feedback, exploration and learning mechanisms, and the environment of the data center is ensured to be in a safe operation interval while learning and optimizing the air conditioner setting value of the air cooling unit; therefore, empirical data with a poor estimation effect can be trained in future learning; and a new evaluation item is added to the environment state by applying the decision network, and the sensitivity of the controller to the environment change is improved.

Description

technical field [0001] The invention relates to the field of air-conditioning control and energy saving, in particular to an air-cooled data center energy-saving method based on a deep Q-learning duel network. Background technique [0002] Energy saving is to reduce energy consumption as much as possible to produce products with the same quantity and quality as before; It is to apply methods that are technically realistic and reliable, economically feasible and reasonable, and environmentally and socially acceptable to effectively use energy and improve the energy utilization efficiency of energy-consuming equipment or processes. [0003] Existing energy-saving control schemes based on energy consumption prediction need to collect historical sensor data, air-conditioning power consumption, and air-conditioning setting data to construct an initial training data set, and use sensor data and air-conditioning settings as input, and air-conditioning power consumption and ambient ...

AI Technical Summary

Problems solved by technology

When the control is issued, the prediction model predicts the power consumption of different air-conditioning settings based on the sensor data collected in the current environment, and the temperature prediction selects the air-conditioning setting with the minimum power consumption under the temperature limit to achieve energy-saving air-conditioning. However, the existing technology The medium energy consumption prediction method has high requirements for training data, but in actual scenarios, there is often a relatively large bottleneck in the collection of rich data sets. First, the temperature state throughout the year is complex and changeable, and the historical data in stages is of great importance to the future. Second, the setting values ​​of air-conditioning and refrigeration equipment are often not too much intervention, and historical data is often a limited combination of air-conditioning settings; third, once the equipment is replaced in the data center, the energy consumption prediction control scheme will be difficult to adapt New physical environment, which leads to failure of predictive methods

Images