Automated scheduling of deferrable PEV/PHEV load in the smartgrid

We consider the scheduling of deferrable charging demand for plug-in electric or hybrid-electric vehicles (PEVs/PHEVs) in the smart grid such that the grid is operating within the safety charging threshold, and as many as consumers/users are satisfied by the end of a finite horizon (e.g., 8pm6am). Given that the charging profiles of PEVs/PHEVs are not constant and have a (truncated) prescribed triangle shape, the grid has to take into account such unevenness in the scheduling of their charging process. We assume that the consumers’ charging profiles are known at the beginning of the finite horizon, and their charging can be interruptible. We develop an automated discrete-time scheduling algorithm in which the grid dynamically tracks the unevenness of each consumer’s charging profile, and gives priority to consumers with highest unevenness that it can tolerate in each time period. For each consumer, the unevenness is measured by the cumulative difference of the charging profile and the average residual demand of his remaining charging profile in each time period until charging completion. This schedule is dynamic because it measures the average residual demand of each consumer in each time period. By comparing with an alternative scheduling algorithm that only measures the unevenness by the average demand of each consumer from their charging profile at the beginning, we show that the dynamic algorithm can better avoid bustiness in the charging process of the grid and can also satisfy much more consumers by the end of the finite horizon. We also compare with another algorithm, the SRPT algorithm, that gives priorities to consumers with shortest demand period, without taking into account unevenness. We conduct simulations to show the effectiveness of our scheduling algorithm in two scenarios: PEV/PHEV consumers, a mixture of PEV/PHEV consumers and constant-charging-profile consumers. With renewable energy supply, we allow the safety charging threshold to be random in each period and also conduct simulations to compare the performance of the algorithms in the scenario with mixed consumers.