MoSiCS: Modeling, simulation and optimization of complex systems–A case study on energy efficient datacenters

Abstract

This paper proposes a methodology for modeling and simulation the operation of Complex Systems considering the energy perspective which may feature both discrete and continuous as well as linear and nonlinear sub-systems. The methodology integrates in a unified view, sub-systems with behavioral or phenomenological models, each of them featuring a set of inputs, outputs and control variables with their associated actions. A proactive optimization process is defined in terms of system level flexibility as well as operation self-adaption, and is leveraging on model simulations to calculate over a time window the system outputs for a set of given inputs and actions. We aim to determine the optimal plan of actions over a set of predicted inputs that will minimize the distance between the system outputs and a given target. To prove its effectiveness, we model and simulate the thermal-electrical processes inside a Data Center, aiming to optimize its operation by exploiting the thermal energy flexibility to re-use the otherwise wasted heat in nearby neighborhoods.