The power grid that services our homes traditionally relies on centralized power sources. This picture has now changed. Advanced technology has diversified the grid by adding new sources of energy generation/storage and two-way power flows in distribution grids. Many American households use rooftop solar photovoltaic (PV) panels to generate their own electricity, which can then be stored using home battery systems . Smart household appliances and meters are allowing users to adjust consumption according to the energy price and to help balance the grid. Distributed Energy Resources (DER), refer to small-scale units of local generation, storage, and controllable load connected to the grid at distribution level, for example, behind-the-meter renewable and non-renewable generation, and electric vehicles. DER penetration is growing every year. The wide-spread deployment of DER and two-way power flows are transforming the distribution grid and presenting new operation challenges, such as fast and deep ramps and increasing uncertainties.
With increasing penetration of DERs, they are expected to be effectively monitored and controlled to achieve an autonomous distribution grid with two key features: (1). optimizing energy performance of DERs to address stochastic and dynamic challenges, (2). supporting grid services of frequency and voltage regulation at distribution levels [1,2] . To this end, the development and home-scale implementation [1,2] of the theoretical, centralized, optimized DER control algorithms for the secure and economic operation of distribution grids have been intensively studied, but two shortcomings remain: (1) cybersecurity issues are not being considered; and (2) accurate power grid topologies are not available in real time. To overcome these barriers, we propose to develop a novel paradigm, Fine Grid Mind, to explore AI-based DER control algorithms including deep reinforcement learning and distributed cyber-physical algorithms for autonomous distribution grid operation.