Integrating residential photovoltaic (PV) power generation and energy storage systems into the Smart Grid is an effective way of reducing fossil fuel consumptions. This has become a particularly interesting problem with the introduction of dynamic electricity energy pricing, since consumers can use their PV-based energy generation and controllable energy storage devices for peak shaving on their power demand profile, thereby minimizing their electricity bill. A realistic electricity pricing function is considered with billing period of a month, comprising both an energy price component and a demand price component. Due to the characteristics of electricity price function and energy storage capacity limitation, the residential storage control algorithm should 1) utilize PV power generation and load power consumption predictions and 2) account for various energy loss components during system operation, including energy loss components due to rate capacity effect in the storage system and power dissipation of the power conversion circuitry. A near-optimal storage control algorithm is proposed accounting for these aspects. The near-optimal algorithm, which controls the charging/discharging of the storage system, is effectively implemented by solving a convex optimization problem at the beginning of each day with polynomial time complexity. For further improvement, the reinforcement learning technique is adopted to adaptively determine the residual energy in the storage system at the end of each day in a billing period.
- Energy storage
- Optimal control
- Reinforcement learning
ASJC Scopus subject areas
- Renewable Energy, Sustainability and the Environment