Computer scientists at University of Southampton in the U.K. have developed mathematical techniques that encourage accurate predictions of distributed electric power contributions to regional power grids. The work of Valentin Robu and colleagues from Southampton’s Agents, Interaction and Complexity Research Group was presented this week at the Twenty-Sixth Conference on Artificial Intelligence in Toronto.
As more distributed producers of power are installed by homes and businesses, including renewable sources like wind and solar, power grids need a way of estimating with confidence the contributions of these producers. The more accurate the estimates of power from these distributed sources, the better the grid operators can estimate and generate power from conventional power plants to meet customers’ power needs.
Because most of these distributed power sources are small operators, however, tracking and predicting their contributions can be difficult. In addition, Robu notes that “current feed in tariffs, that simply reward production are expensive and ineffective.”
The research by Robu and colleagues aims to provide a reliable way for distributed power producers to more accurately estimate their contributions to the grid, with a payment mechanism that rewards producers for the accuracy of those estimates. That mechanism consists of scoring rules with incentives for accuracy in predicting power production under conditions of uncertainty.
Robu’s team tested their methods with 16 commercial wind farms in the U.K, where they collected data every half-hour from the wind farms over a 10-week period. Their findings show their reward mechanism outperformed current payment mechanisms in effect for distributed power producers.
The ability to accurately predict contributions of renewable, but intermittent, power sources on the grid can make possible creation of cooperative virtual power plants (CVPPs), collections of distributed energy sources with the reliability and robustness of conventional electrical power plants. “CVPPs that together have a higher total production and, crucially, can average out prediction errors is a promising solution,” says Robu, “which does not require expensive additional infrastructure, just intelligent incentives.”
Read more:
- Simulation Software Helps Optimize Utility Transport Grids
- Ignore the Sticker Shock, Electric Trucks Can Pay You Back
- Engineers Calculate, Write Software for Solar Fluctuations
- New Math Tools Developed to Monitor Power Grid
* * *
You must be logged in to post a comment.