Microgrid Used as a Testbed for Neural Network

Share Button

A research team at the US Department of Energy’s Argonne National Laboratory has tested a neural network for controlling power systems on a microgrid.

neural network

By Who is Danny/Shuttertock.com

The team set out to solve the challenge of modeling static and dynamic characteristics of a power system. A static characteristic is an action that does not change within a certain amount of time, such as turning a generator on or off. A dynamic feature on the other hand, such as electrical frequency, could fluctuate over time if there is a disruption or an operation.

“If you put dynamic and static formulations together in the same model, it’s essentially impossible to solve,” said Argonne computational scientist Feng Qiu, who co-authored the study.

Using artificial intelligence

Instead of calculating static and dynamic features separately, or trying to fit existing formulas together, the researchers used a neural network; an artificial intelligence tool that can create a map between a defined input and output.

“If I know the conditions we start with and those we end with, I can use neural networks to figure out how those conditions map to each other,” said Yichen Zhang, Argonne postdoctoral appointee and lead author of the study.

Optimized the static resources in the microgrid

Using a microgrid, the team tracked how a set of static conditions mapped to a set of dynamic conditions. The neural network was used to optimize the microgrid’s static resources to maintain a safe electrical frequency.

Simulation data, made up of static inputs and specified dynamic responses, was fed into the neural network. Using this, it “learned” to map estimated dynamic responses, such as safe frequencies, for set static conditions.

“The neural network transformed the complex dynamic equations that we typically cannot combine with static equations into a new form that we can solve together,” Qui said.

Looking ahead, this approach is a starting point that could help operators of power systems to control their assets in a more effective way.

Track news about microgrid research. Subscribe to the free Microgrid Knowledge newsletter.

Share Button

Sign up for our newsletter and get the latest microgrid news and analysis.

Comments

  1. “Simulation data, made up of static inputs and specified dynamic responses, was fed into the neural network. Using this, it “learned” to map estimated dynamic responses, such as safe frequencies, for set static conditions.”

    Back in the late 1970’s and early 1980’s when TI brought out their DSP chip, inverter technology made a quantum leap in capability. The concept of taking “feedback” from the load and generating these “magic” numbers allowed noise reductions and better load impedance matching on the fly than ever before.

    Basically the DSP modifies the output frequency based on the “best” feed back noise value detected. In one case a motor running at 60HZ may have a better impedance profile at an output of 60.045 Hertz than a strict 60.000 Hertz. This could also be a way to “form” a grid connection from an inverter that acts like a large rotating mass generation system does now.

Leave a Comment

*