Finding optimal control solutions for systems of high complexity cannot be achieved by applying straightforward optimization algorithms. The results of our recent research revealed that using distributed building automation systems together with an accurate temporal thermal model of server room, substantial amount of energy can be saved. Although we have created a predictive thermal model for server rooms, we realized that current building automation systems utilized in data centers are rigid and they do not exhibit enough intelligence and adaptability. To address these inadequacies, in this research novel methods based on cognitive algorithms, machine learning, and bio-inspired heuristics will be explored. The results will be incorporated into an integrated solution and will be evaluated via simulations.