Operational planning of data-center cooling systems via collaborative neurodynamic optimization

Data centers are essential facilities for information processing, storage, and transmission. The rapidly rising power consumption of data centers has become a global concern. In data centers, cooling systems consume a significant portion of power consumption. Existing planning methods are inefficient for the operation of data-center cooling systems. In this paper, a mixed-integer nonlinear optimization problem is formulated for the operational planning of data-center cooling systems. The problem is then decomposed into two subproblems involving continuous and discrete variables, respectively. Within a collaborative neurodynamic optimization framework, a method is developed to solve the subproblems. The method employs neurodynamic-model pairs consisting of projection neural networks and Boltzmann machines, and uses a particle swarm optimization rule. Experiments are conducted on data centers with 800, 2700, 4000, and 8000 racks. The experimental results demonstrate that the developed method achieves the lowest power consumption compared with mainstream methods.