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

Area of Science:

• Computer Science
• Engineering
• Optimization

Background:

• Data centers are critical for modern information technology.
• Rising power consumption in data centers, particularly by cooling systems, is a global environmental and economic concern.
• Current operational planning methods for data center cooling are inefficient.

Purpose of the Study:

• To formulate and solve an optimization problem for the efficient operational planning of data center cooling systems.
• To develop a novel method that minimizes power consumption in data center cooling operations.

Main Methods:

• Formulation of a mixed-integer nonlinear optimization problem for cooling system operations.
• Decomposition of the problem into continuous and discrete subproblems.
• Development of a collaborative neurodynamic optimization framework using projection neural networks, Boltzmann machines, and particle swarm optimization.

Main Results:

• The developed neurodynamic optimization method was tested on data centers of varying sizes (800 to 8000 racks).
• Experimental results show that the proposed method achieves lower power consumption than mainstream approaches.
• The method effectively optimizes the operation of data center cooling systems.

Conclusions:

• The novel neurodynamic optimization framework offers a more efficient solution for data center cooling system planning.
• This approach contributes to reducing the significant energy footprint of data centers.
• The findings suggest a promising direction for sustainable data center operations.