Proving the Business Case for the Internet of Things

North Carolina researchers develop energy scheduling technique to advance smart grid concept

Steve Rogerson
June 30, 2015
Researchers from North Carolina State University have developed a technique for scheduling energy in electric grids that moves away from centralised management by tapping into the distributed computing power of energy devices. The approach advances the smart grid concept by coordinating the energy being produced and stored by both conventional and renewable sources.
Currently, power infrastructure uses a centralised scheduling approach to forecast and coordinate the energy produced at the thousands of large power plants around the USA. But as renewable energy systems – such as rooftop solar panels – proliferate, and are incorporated into the power grid, the infrastructure will need more advanced systems for tracking and coordinating exponentially more energy sources. Addressing this issue is essential to the idea of a smart grid that can make efficient use of widespread renewable energy resources.
“A key challenge for renewable energy generated on-site – by home solar panels, for example – is determining how much of that energy needs to be stored on-site and how much can be shared with the larger grid,” said Mo-Yuen Chow, a professor of electrical and computer engineering at NC State and senior author of a paper describing the power scheduling technique.
The existing approaches to scheduling are highly centralised, with power plants sending data to a control centre that crunches the numbers and then tells plants how much they’ll be expected to contribute to the grid.
“This approach doesn’t scale up well, which is a problem when you consider the rapid growth of on-site renewable energy sources,” said Navid Rahbari-Asr, a PhD student at NC State and lead author of the paper. “The rise of on-site energy storage technologies presents an additional challenge, since that means energy can be stored for use at any time – making power scheduling calculations significantly more complex. In addition, the centralised approach is vulnerable. If the control centre fails, the whole system falls apart.”
To address these problems, the researchers developed technology that takes advantage of distributed computing power to replace the traditional control centre with a decentralised approach.
“Our approach taps into the computational resources of each energy device,” Rahbari-Asr said. By having each device communicate with its immediate neighbours, the device can calculate and schedule how much energy it will need to store, how much to contribute to the network, and how much to draw from the network.
“Collectively, this distributed technique can determine the optimal schedule for the entire grid,” said Rahbari-Asr.
The distributed technique would also help protect the privacy of home owners and other power generators, because they wouldn’t be sharing their energy production, storage and consumption data with a control centre.
The technology has been validated in simulations, and the researchers are in the process of implementing it in an experimental smart grid system at the National Science Foundation Freedm Systems Center on NC State’s campus.
“We hope to have experimental results to report in 2016,” Chow said.
The paper – “Cooperative Distributed Scheduling for Storage Devices in Microgrids using Dynamic KKT Multipliers and Consensus Networks”  – will be presented at the IEEE Power & Energy Society General Meeting, being held July 26 to 30 in Denver. The paper was co-authored by Yuan Zhang, a PhD student at NC State. The work was supported by the NSF Freedm Systems Center.
The paper’s abstract says that scheduling of storage devices in microgrids with multiple renewable energy resources is crucial for their optimal and reliable operation. With proper scheduling, the storage devices can capture the energy when the renewable generation is high and utility energy price is low, and release it when the demand is high or utility energy price is expensive.
This scheduling is a multi-step optimisation problem where different time-steps are dependent on each other. Conventionally, this problem is solved centrally. The central controller should have access to the real-time states of the system as well as the predicted load and renewable generation information. It should also have the capability to send dispatch commands to each storage device.
However, as the number of devices increases, the centralised approach would not be scalable and would be vulnerable to a single point of failure. Combining the idea of dynamic KKT multipliers with consensus networks, this paper introduces a novel algorithm that can optimally schedule the storage devices in a microgrid solely through peer-to-peer coordination of devices with their neighbours without using a central controller.