Emulation system on smart grid Fast Automated Demand Response of widely-distributed stochastically-operating building facilities

The Fast Automated Demand Response (FastADR) for ancillary services might require precise control on power consumption of widely-distributed target loads. Office building air-conditioning facilities are expected to be the FastADR target loads. However, the dynamic behaviors of widely-distributed inverter-driven air-conditioners are complicated and stochastic. In order to reproduce and analyze the behavior, we have developed a realistic emulation system for the FastADR for these facilities. A neural network model is built from actual facilities' time-series data. The emulation system consists of 120 models with different stochastic disturbances, an aggregator's one-min-sampling feedback controller, and actual FastADR communication networks with the dummy Internet. We found that some step responses of FastADR sometimes become oscillative depending on the parameters of aggregator's wide-area feedback control.