UC Riverside

The power grids are continuously evolving over the past decades, where new challenges and opportunities are embraced at the same time. On one hand, the penetration of renewable generations and other distributed energy resources (DER) is growing rapidly, whose different generation and control patterns could significantly impact the daily operation. On the other hand, the new communication, monitoring and regulating devices are gradually installed, which enable more control abilities of the generations, demands, and grids, and the feasibility to deploy more sophisticated control schemes.

To leverage the new technique and overcome the new challenges in the smart girds, different optimization and control problems need to be solved for different roles including the system operator, demand, and financial traders. For the system operators, it is critical to maximizing the total social welfare while satisfying the operational constraints. To better coordinate the DER and improve the efficiency of distribution systems, the three-phase optimal power flow (OPF) problem algorithms are developed including the DCOPF algorithm for robustness and the ACOPF algorithm for optimality. Moreover, the deep reinforcement learning-based Volt-VAR control schemes are proposed to better maintain the voltage stability and electricity service quality.

For demands resources, minimizing their energy bills will satisfy the energy needs is always their goal. Providing ancillary services by proactively adjusting their total demand is one of the potential choices. Through the provision of the services, the demands can not only receiving incentives from the system operators but also help to improve the reliability and stability of power grids. We develop control schemes specifically for the data centers to provide the phase balancing service in the distribution system and the frequency regulation service in the transmission system.

The financial traders, it is desired to maximize their total profits. A better trading strategy with a more accurate forecast model can help increase the traders' gain and further improve the price convergence of the electricity market. Our machine learning based trading framework outperforms the existing approach and lays the foundation for market efficiency evaluation across the markets.