Energy & Power Systems

Upcoming energy technologies such as roof-top solar panels, batteries, and electric vehicles are interfaced to our AC electric power grids through AC/DC converters. These devices come with advanced sensing, communication, and control functionalities. We have shown through our grid probing project that if an electric utility can control these devices, we can determine non-metered consumption and the connectivity of the grid, which are often unknown.

With funding from the U.S. Department of Energy and the Pacific Northwest National Laboratory, we are exploring the resiliency of distribution systems with respect to extreme events. 

The main focus is on restoring critical load/services after a catastrophic outage where the utility system is severely damaged and not available. In this situation, microgrids and distributed energy sources would be tapped to serve the critical load. We have proposed a new metric for resiliency based on the total MW-hour that the system is able to provide to critical services during system restoration. We also have developed new operation and control methodologies that enable a weak system to maintain a stable operating condition using only distributed energy and control resources.

The models that have aided power engineers over decades were not designed to accommodate today’s massive data from new sensing devices or the random dynamics created by the growing numbers of intermittent, renewable energy sources. With funding from the NSF and the Pacific Northwest National Laboratory, we are developing new dynamic state estimation and control methods for today’s grid. We have developed new dynamic estimators supported by a strong mathematical foundation that quantify the uncertainties involved with renewables and integrate them in a unified framework. We also have proposed a new voltage control scheme that minimizes the communication bandwidth requirement while exhibiting a system-wide situational awareness. 

Cyber intrusions have caused major power outages by disrupting the grid’s operation and control systems. An ECE team has developed a cyberphysical system testbed to demonstrate how falsified control commands can be detected and stopped. In this project, anomalies are specified and identified for substation automation and supervisory control systems. Collaborative attacks are identified using relation-based models and algorithms. Ongoing work is to defend the supervisory control systems from attacks by falsified measurements through the communication systems of the power grid. 

We also have used tools from machine learning and big data analytics to train AC/DC inverters to behave harmoniously with each other, and improve the stability, reliability, and efficiency of our residential electric grids. Localized solutions, where each inverter operates in silo, have been ineffective. On the other hand, network-level coordination necessitates formidable cyber overhead—both in computation and communication. Leveraging advances in machine learning, we are hitting the sweet spot by using data to learn inverter control rules in an offline fashion, and later apply them in real-time with tunable cyber overhead.

Solar and wind energy are highly variable. Sometimes there is a surplus of renewable energy, and we have to waste it. Other times, there is a sudden power deficit, and some electric loads have to shut down. As a possible solution, can we charge batteries during periods of surplus, and discharge batteries at deficit times? How can that be done in a smart way without knowing the upcoming surplus/deficit? What size should the battery be? Alternatively, during times of surplus, can we run pumps and push the water circulating in our city networks up to water towers in lieu of a battery? We are exploring these and other questions regarding energy variability in the grid.

ECE researchers are continuing to hone the open-source software platform for building energy management systems to improve small and medium-sized building efficiency and help implement demand response.

Decreasing natural gas prices and the capability of gas-fired electric power plants to rapidly respond to fluctuations from wind generation, both have resulted in a strong coupling between gas and electric power systems. To better capture this coupling and optimize the two energy systems in tandem, we are developing algorithmic solutions for improved modeling, optimization, and monitoring of natural gas networks.

Modeling the power grid—such a complicated system that it’s sometimes referred to as “the world’s largest machine”—is a task the size of the grid itself. To expand existing models, researchers are using synchrophasor data to find the impulse response between any pair of locations by cross-correlating their angle and power flow data streams. This measurement can occur during normal grid operation, making it a useful tool for understanding the state of the power system.

The increasing number of renewable energy sources is complicating the already huge task of monitoring the power grid. ECE researchers are leveraging domain-specific machine learning tools to model, monitor, and optimize power system dynamics. Using synchrophasor data, Gaussian processes, and established grid stability metrics, these new methods can improve grid efficiency and security while increasing the contributions of renewable energy.