Model-based control on populations of air conditioners: Shaping aggregated power for demand side management

A population of thermostatically controlled devices such as air conditioners can be managed as a whole to reduce their load during a critical period or to participate in the electricity market. We present a model based on first principles for the power response of a population of air conditioners controlled by a common temperature set point change. We then develop a model-based controller for the aggregated power of the population. Our controller requires fine temperature set point resolution. For coarse resolution, as in ACs installed nowadays, we propose to divide the population into logical clusters and send different coarse-resolution signals to each cluster. Clustering achieves the same performance as using one global, high resolution, signal.