Abstract
As renewable energy resources are more environmentally friendly than the conventional energy sources, the problem of effectively utilizing renewable energy sources has received a lot of attention lately. In this thesis, we consider a stand-alone power system with renewable energy sources and a perfect battery. We assume that the loads to the system are deferrable so that they can be scheduled to balance the supply and the demand. The main objective of a scheduling policy in such a system is to effectively utilize renewable energy sources so that the loads can be served as quickly as possible and the amount of wasted energy can be minimized. For such a purpose, we consider three classes of scheduling policies: (i) no lookahead policies, (ii) 1-period lookahead policies, and (iii) 1-period prediction policies. Our findings are (i) a maximal 1-period lookahead policy is always better than a maximal no lookahead policy, and this can be formally proved by using a sample path argument, (ii) increasing the battery capacity has little effect on the performance of a maximal 1-period lookahead policy, and this can also be proved by using a sample path argument, (iii) increasing the battery capacity is very effective for improving the performance of a maximal no lookahead policy, and (iv) prediction can be very effective in improving the system performance without the need of increasing battery capacity.