An Exact Algorithm for Non-preemptive Peak Demand Job Scheduling

Peak demand scheduling aims to schedule jobs so as to minimize the peak load in the schedule. An important application of this problem comes from scheduling power jobs in the smart grid. Currently, peaks in power demand are due to the aggregation of many jobs being scheduled in an on-demand fashion. Often these have some flexibility in their starting times which can be leveraged to lower the peak demand of a schedule. While the general version of the problem is known to be NP-hard (we observe it is even NP-hard to approximate), we provide an optimal algorithm based on dynamic programming that is fixed-parameter tractable (FPT). Simulation results using household power usage data show that peak power demand can be significantly reduced by allowing some flexibility in job execution times and applying scheduling.