I describe efforts at LANL in modeling and controlling aggregated ensembles of stochastic loads. The modeling part of the talk will focus on accurate analysis of the steady state and transient behavior of the probability distribution of an ensemble of loads (thermal, pools, electric cars or alike) as observed by an aggregator. Then in the control part of the talk I describe novel algorithmically tractable MDP approaches to balancing tracking of the exogenous system operator signal with the customers welfare. This presentation is based on the joint work with V. Chernyak (Wayne State U) and M. Goldshtein (Skoltech).