AI model boosts microgrid optimisation
Researchers at Gwangju Institute of Science and Technology in Korea have developed an AI-based optimization model to boost the performance of electrical microgrid systems.
The model incorporates possible variations in future power outputs to arrive upon an optimal scheduling decision and reduce operational costs and load shedding.
Microgrids are smaller, localized electricity grids, often with renewable energy sources such as solar panels or wind turbines. These can be connected to the main grid of the region, but also can also be disconnected or “islanded” if needed. Models that guide the operation of microgrids, such as scheduling load shedding, are key to their efficient functioning.
So far though most microgrid models have either neglected the uncertainty and variations in renewable energy or assumed the worst-case scenario, which can lead to an increase in energy not supplied (ENS) and operating costs.
- Sony looks to open source cloud microgrid
- Lidl builds Finland’s largest power microgrid
- Siemens teams for African solar microgrid projects
The team at GIST developed a new two-stage stochastic optimization model to minimize operating costs and load shedding.
“One of the problems with microgrids is that they sometimes cannot supply enough electricity for the load, causing load shedding, and at other times they produce too much electricity,” said Dr. Yun-Su Kim, who led the study. “We created an operation algorithm that can reduce operation costs and load shedding.”
Key to the new optimization model was an artificial neural network (ANN) to create a prediction model for the power output of renewable energy sources. This power output is obtained in the form of a probability density function, providing the likelihood that a given power output will be obtained at any given point of time.
This accounts for variations and uncertainty in the renewable energy supply. This probability density function is then fed into a stochastic optimization model that makes operating decisions, such as scheduling.
The researchers validated their model using data from a microgrid designed by the Natural Energy Laboratory of Hawaii Authority. They found that the ANN predicted power output with a low error of 9.7%. The stochastic optimization model also offered an approximately 20% reduction in average ENS, as well as around 19% lower operating costs.
“Reforming the power grid using microgrids can help renewable energy integration. Thus, improving the efficiency and integration of microgrids will bring us one step closer to energy security and stability,” said Kim.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
