Abur Receives DOE Solar Energy Technology Office 2020 Award
Professor Ali Abur, electrical and computer engineering, has been studying and researching how electricity is generated, transmitted, and distributed within power systems for 30 years. He recently received a $750K Department of Energy (DOE) grant from the Solar Energies Technologies Office in the category of Systems Integration to fund his most recent project, titled “Graph-Learning-Assisted State and Event Tracking for Solar-Penetrated Power Grids with Heterogeneous Data Sources.”
AI and Machine Learning to Optimize the Solar Power Grid
The project uses artificial intelligence and machine learning techniques to integrate a diverse set of measurements and use them to calculate the state of the power grid. The resulting tool will be able to detect topology changes and faults in the grid and update grid models accordingly, which will improve the situational awareness of power grids that contain large amounts of solar energy sources. This will be accomplished by exploiting a large volume of data and measurements available from a highly diverse set of measurement devices. The project will also provide tools to detect and identify unexpected disturbances or switching events by exploiting the recently developed sparse estimation methods in the data analytics area.
According to Abur, the DOE has “increased its efforts in supporting projects which facilitate integration of renewable energy sources into power grid.” As the world turns toward clean energy solutions, the DOE has been investing in solar renewable energy sources as the advances in solar energy technologies significantly accelerated in the recent years. “I hope this project will contribute in a minor way to the integration of these renewables, which will hopefully help to decrease clean energy costs,” says Abur.
To understand Abur’s contribution to the clean energy initiative, one must first understand how power grids are monitored. A power grid contains many substations which are interconnected through power lines. These substations can either generate and supply power to the grid or connect to distribution feeders that withdraw power from the power grid. All substations have multiple types of sensors which collect various measurements such as power flows and voltages. This information is automatically processed and passed along to system operators who are constantly monitoring the power system to avoid power outages or violation of safety limits for equipment by taking appropriate corrective actions.
Abur explains, “My interest and work is mainly having the monitoring done in a reliable and robust fashion.” Using the information from sensors, Abur’s research aims to enhance the processing of these measurements and determine the best estimate of the operating state of the system used by the operators and many other network applications.
Over the last 10-15 years, the connection of renewable energy sources in distribution systems has complicated the monitoring process of the power grids. Historically, distribution systems have withdrawn power from the transmission system; however, with the increasing numbers of solar energy sources, power flows became bi-directional between the two systems.
“You have renewable energy sources and sometimes the direction of power flow may change. It may actually reverse direction and inject power to the transmission grid. As a result, now you have to start monitoring what’s happening inside these distribution systems so that we can exploit this added capacity to improve the overall system dispatch, reducing the production cost of electricity and increasing system reliability,” says Abur.
From Theoretical to Implementation
In the first part of this research project, Abur and his collaborators from UMass-Lowell and Brandeis University will be using machine learning tools to reconcile measurements obtained at different rates and resolution from different sets of measuring devices. Various measurements from different types of sensors will be processed using machine learning tools in order to create a usable set of measurements, which will be compatible with the proposed state estimator.
“The differentiating feature of this project is that we take the output of the state estimator, and we feed it back into this machine learning tool. This feedback loop allows continuous improvement of both applications, helping the machine learning tool to receive less noisy measurements and the state estimator to receive better approximations to missing measurements.”
The second goal of this project is to heighten and strengthen the situational awareness in the power system by creating monitoring techniques to detect, identify, and locate unexpected disturbances such as a line outage or a short circuit fault in the grid. This will allow system operators to dispatch a team to the exact location of the disturbance and fix the problem in a timely fashion.
After the theoretical development is complete, Abur and his collaborators will develop the necessary software and use simulated measurements to test and validate their approach. In the final year of the grant, they will partner with Commonwealth Edison, a utility company in Chicago, to use actual measurements from their microgrid network to prove that the developed state estimator will work in a real utility system.
Abur expects that every few years there will be changes in the composition of the sensors that will require variations to the formulation. “Technology is also driving the research,” he says. Thus, as developments in renewable energy change the landscape of the power grid, Abur’s research will continue to follow the changes to facilitate seamless integration of these renewable energy sources into the system.