Using AI to predict the supply of renewable energy days in advance

Discussion in 'Intelligence & Machines' started by Plazma Inferno!, Jul 15, 2016.

  1. Plazma Inferno! Ding Ding Ding Ding Administrator

    Renewable energy like solar and wind power are changing the way we generate electricity. Our energy production is becoming cleaner and cheaper, and in many countries renewables are starting to overtake fossil fuels as the primary power source. But one of the biggest problems with renewables has yet to be solved: what happens if it's cloudy?
    More specifically, the problem is that renewable energy sources can never provide a constant source of power. No matter how many solar panels you build, they all provide zero power when the sun goes down. And if you build a large number of solar panels, you risk generating too much power when it's sunny, which can be a problem too. Ultimately, most countries are relying on fossil fuels for so-called baseline power, throttling up or down the output of coal and gas plants to ensure energy demands are met without being exceeded.
    But this isn't a perfect solution either. It takes time to make these adjustments, called "re-dispatches," and the government has to compensate utility companies for lower outputs, which costs taxpayers millions of dollars a year. To combat this, researchers in Germany are developing software that uses machine learning to predict the amount of energy generated by renewables over the next few days. This early warning system, called EWeLiNE, which takes realtime data from solar panels and wind turbines around Germany and feeds it into an algorithm that calculates the renewable energy output for the next 48 hours. This algorithm uses machine learning, and the researchers compare real data with EWeLiNE predictions to refine the algorithm and improve its accuracy.

Share This Page