Long-term Wind Speed Forecasting: reality or wishful thinking?

Question

I have been studying/working with Wind Energy for a while and, one of my main responsabilities right now is to search for adequate long-term wind speed forecasting models.

However, 2 things strike me about this quest.

1. Wind speed forecasting is a task of great complexity for wind speed itself is highly stochastic. It depends on climate patterns, regional weather behavior, topography and atmospheric stability, or the lack of thereof, just to mention a few.

2. The very own idea of "long-term" varies from researcher to researcher. So far I've found authors who claim that 1 to 7 days can be considered long-term (Current status and future advances for wind speed and power forecasting,Jung & Broadwater, 2014); but also those who say one full year is the minimum threshold for long-term forecasting (The quality of weather information for forecasting of intermittent renewable generation, Sokolowska & Hossa, 2014). It's clear that forecasting horizon is highly linked to the aim of your research; therefore, when topics change, so will their forecasting horizons.

a. Isn't there a consensus about what long term means when it deals with wind speed forecasting?

b. Can it be assumed year(s)-long wind speed forecasting withsignificant accuracy is still some far-fetched, wishful thinking?

c. What are some efficient models out there yielding meaningful results?

I would like to hear your ideas/points as long as they can be backed by scientific reasoning. If you could point me out in the direction of research (proofs to your rationale) in long-term wind forecasting, I'd be downright grateful.

In case you have worked with this topic, then I'd enjoy hearing from your own experience.

Thank you so much in advance for all inputs.

Answer 1

Warning: "long-term wind-speed forecasting for generation" has (at least) two very different meanings. One refers to forecasting a distribution of wind speeds; the other refers to hour-by-hour (or half-hour by half-hour) forecasting of wind speeds.

Generally, when we talk about long-term forecasting of wind speeds for wind generation, we're talking about estimating the distribution of wind speeds over the next few years. That is to say, we're not talking about what the wind speed will be exactly two years from today, at 15.00 UTC. We're talking about what the shape of the distribution of the wind speeds will be - maybe in terms of the two parameters of a Weibull distribution, together with some indication of how the distribution varies by time of day, and by season of the year.

There is some skill in hour-by-hour wind speed forecasts up to about two weeks away. But for longer than that, there's very little skill in the models at all. In that regard, if we are talking about hour-by-hour wind speed forecasts, then short term is the next 0.5 - 4 hours, and long-term is 1-2 weeks.

In both cases, I'd look at recent publications from Reading University - David Brayshaw and others - and specifically on the accuracy of wind forecasts for wind generation, I'd also recommend work by Henrik Madsen and Pierre Pinson.

Answer 2

Wind can be an incredibly localized phenomena, both in the short term and the long term. Although the broad outlines of a long-term wind forecast shouldn't be too complex or controversial, getting the details right will generally require actual on-site measurement. Google wind measurement sodar for some cool tools to do such measurement.