Photovoltaic and Behind-the-Meter Battery Storage: Advanced Smart Inverter Controls and Field Demonstration
Electric utilities have little visibility of the electrical distribution system, and consequently, limited diagnostic capabilities. The distribution grid was designed for a unidirectional power flow, where energy is supplied by large centralized power plants; however, this is changing to meet California’s aggressive de-carbonization goals. The large-scale deployment of distributed renewable generation, such as photovoltaics (PV), can have negative effects on an unprepared grid. As illustrated in the California “duck curve,” PV generation modifies load profiles during the day, which causes steep ramping in the evening. This effect indicates the need to revise the electrical grid’s design and operation. This project sought to (a) create a centralized resource in California to test and validate distribution technology controls with industry standard PV, storage, and high-fidelity sensors, (b) support strategic and operational decisions for new grid architectures by providing simulation models, (c) promote new ways to control clusters of PV smart inverters, in accordance with California Rule 21, and (d) innovate, develop, and field test a predictive controller to maximize profit for the customer while supporting the grid. The controller was built using the state-of-the-art model predictive control methodology to optimally control behind-the-meter PV and battery storage. In consideration of the duck curve, the controller optimally controls the battery by charging during excess generation periods and discharging during the critical afternoon demand hours. The controller was evaluated in annual simulations and revealed the potential cost-effectiveness of behind-the-meter battery storage. The simulations showed as much as 35 percent of an annual electricity bill could be saved, with a payback of the investment in battery storage in about 6 years – significantly shorter than the manufacturer’s 10-year warranty. All developed simulation models, the grid event library, and the model predictive controller are open-source and available online.