Energy storage management and load scheduling with renewable integration

Abstract

In this dissertation, the energy storage management and load scheduling problems are studied. The main objective is to design real-time cost-effective control policies at a residential site with integrated renewable generation. Stochastic nature of system dynamic for renewable generation, user load, and electricity pricing has been formulated in problems. Furthermore, battery degradation costs due to battery operation have been incorporated into the system cost. Both infinite and finite time horizon approaches have been designed in this dissertation. Lyapunov optimization technique has been applied to design the real-time control algorithms that rely only on the current system dynamics. Close-form solutions have been obtained for simple implementation. The proposed algorithms are shown to have bounded performance gap to the optimal control policies. The first problem is to minimize the long-term time-averaged system cost with i.i.d system inputs, where battery operation cost is considered. In the second problem, a finite time horizon approach is provided to minimize the system cost over a fixed time period. Non-stationary stochastic nature of system dynamics is considered in formulating the problem. Furthermore, the detailed battery operation costs is incorporated into the system cost. A special technique to tackle the technical challenges in problem solving is developed. In the third problem, a joint energy storage management and load scheduling problem is proposed. The problem is to optimize the load scheduling and energy storage control simultaneously in order to minimize the overall system cost over a finite time horizon. In this real-time optimization design, the joint scheduling and energy storage control is separated and sequentially determined. Both scheduling and energy control decisions have close-form solutions for simple implementation. Through analysis, it is shown that the proposed real-time algorithm has a bounded performance guarantee from the optimal T-slot look-ahead solution and is asymptotically equivalent to it as the battery capacity and time period go to infinite.