Operation Parameter Optimization Of Cooling Systems In Data Centers Based On NN-HS

The rapid development of Internet services such as big data and cloud computing has led to the increasing number and scale of data centers.High energy consumption has become a prominent problem for data centers.Cooling system accounts for a considerable proportion in the total energy consumption of data center.Optimizing its operating parameters and reducing its energy consumption is one of the key technologies to build a green data center.This paper attempts to use neural network modeling combined with harmony search optimization,namely NN-HS method to solve the operation parameter optimization problem.Firstly,the operating parameters of the cooling system are taken as the input,and the corresponding Power Usage Efficiency(PUE)of the data center is taken as the output.After comparing with the relevant learning algorithms,the BP neural network is constructed,and the PUE function is trained by using the historical operation data of a data center of Tencent in North China and the data center of Beijing CMBC,Based on the sensitivity analysis,the key operating parameters are selected,and the objective is to minimize the PUE function which is obtained by using the key parameters to approach again.Considering other performance constraints,the mathematical model for optimizing the operating parameters of the cooling system in the data center is established.Secondly,the algorithm of the model is studied.based on the fact that the quality of initial population of harmony search(HS)algorithm needs to be improved,the blindness and randomness of search direction lead to low search efficiency and slow convergence speed,and the fixed parameter setting leads to low convergence accuracy in the later stage of the algorithm,the corresponding improvement measures are given,and then an improved harmony search algorithm(IHS)is proposed.The improved algorithm adopts the random opposition-based learning strategy to improve the initial population quality of harmony memory,and uses the search direction of "seeking advantages and avoiding disadvantages" to replace the search direction of the basic HS algorithm,which makes the algorithm quickly approach to the optimal solution and stay away from the worst solution in the search process,and effectively improves the convergence speed of the algorithm;At the same time,the parameters of the algorithm are adaptively adjusted to make it decrease nonlinearly in the whole iteration process of the algorithm,which can effectively increase the diversity of harmony population at the beginning of the iteration,and further accelerate the convergence of the algorithm at the later stage,so as to improve the overall solution performance of the algorithm.Based on the classical function optimization problem,the proposed algorithm is calculated and tested,and compared with the results of the related algorithms in the literature,which verifies the feasibility and effectiveness of the improved algorithm.Finally,the proposed algorithm is applied to the optimization of the operation parameters of the cooling system in the data center.The related parameter optimization process is designed,and the above two different types and different scales of engineering instances are selected for simulation and analysis.Based on each instance,different working conditions are selected for optimization simulation,and compared with the operation parameter adjustment method based on expert experience.The relevant results show that the proposed NN-HS method can better reduce the power consumption of the cooling system,and can effectively reduce the PUE value for different data center instances,and the optimization effect is good.The work of this paper provides exploration and support for the improvement of related algorithms,the construction of green data center and the intellectualization of its operation and maintenance,which has certain theoretical significance and application value.