Power grid operators require fast balancing actions to regulate grid frequency; in a Smart Grid scenario this service may be provided by flexible loads. We consider a load aggregator that manages a large quantity of loads, and must align their overall consumption with an operator reference. Starting from a stochastic queueing model for the set of loads, we derive a macroscopic differential equation model, excited by noise, that represents the aggregate, and includes a scalar deferral action variable. An ℋ2-optimal control method is then used to optimize the tracking error in power consumption while keeping the control variable within its hard limits. The controller is tested with real-world frequency regulation signals, and achieves high performance using the relevant industry performance metric. An implementation is given which achieves these benefits with mild communication requirements.
展开▼