Development of the Photovoltaic (PV) Analysis and Response Support (PARS) platform

Many countries have set a goal to achieve 100% clean energy by 2035. In the United States, a preliminary goal is to increase solar energy generation from today’s 3% to approximately 40%. This will require PV systems at all levels to provide grid services for maintaining system frequency and voltage - not only under normal operating conditions, but also during emergency restoration. To develop PV-based grid support functions and test them on a wide variety of possible operating conditions, we have developed the Photovoltaic (PV) Analysis and Response Support (PARS) platform for the GridWrx lab at North Carolina State University. PARS is designed to be a network of power system digital twins for developing and testing microgrid energy and power management systems, providing real-time situational awareness, and selecting optimal response plans for restoring grid services during a prolonged outage. This seminar will introduce three key technologies behind the development of the PARS platform: generative adversarial network (GAN) based synthetic feeder topology and data generation, meta-learning based load forecasting methods, and dynamic modeling of microgrids via co-simulation. The GAN based synthetic data generator allows us to produce realistic simulation scenarios from actual utility networks and customer data. Meta-learning based load forecasting provides a flexible tool for power and energy management applications to obtain load forecasts at all levels with different forecasting requirements. PV and other DER resources operating within a power grid are modeled using co-simulation techniques. This co-simulation utilizes a combination of electromagnetic transient (EMT) and phasor domain analysis to improve the simulation computational speed while still capturing the fast transient responses of inverter-based resources.