Various uncertainties self-correction method for the supervisory control of a hybrid cooling system in data centers

Abstract

The uncertainties of the crucial parameters have a considerable influence on the supervisory control performance of the water side free cooling system in data centers, especially for the model-based optimal control. To reduce the uncertainty of the key parameters in an operational data center hybrid cooling system, this study proposed an uncertainties self-correction method based on Bayesian Inference and Markov Chain Monte Carlo (BI-MCMC) theory. Four novel enhanced self-correction strategies, including the benchmarks extended, adding sensitivity coefficient, local correction and prior distribution updated, were developed to handle the negative impacts caused by the various uncertain working conditions and complex relationships between parameters. The performance of the proposed method was fully investigated under the cases with single/multiple uncertain parameters with different degrees of uncertainty and deviation. For the uncertainties problem with single parameter error scenario, the basic BI-MCMC method reduced the error by at least 92.5%. For the multiple parameter uncertain scenarios, the designed enhanced strategies can significantly reduce the degree of the uncertainty, i.e. the correction accuracy is up to 98% and the absolute correction deviations are not more than 0.02. Therefore, the proposed method could provide a solution to the current challenges in handling the uncertainties and establishing a reliable model for the supervisory control.