Conference programme

Back to the programme printer.gif Print




Delegates are invited to meet and discuss with the poster presenters during the poster presentation sessions between 10:30-11:30 and 16:00-17:00 on Thursday, 19 November 2015.

Lead Session Chair:
Stephan Barth, ForWind - Center for Wind Energy Research, Germany
Di Li WindSim AS, Norway
Co-authors:
Di Li (1) F Mana Matteo (1) Catherine Meissner (1) Aurelie Bencharel (2) Thomas Galopin (2)
(1) WindSim AS, Tønsberg, Norway (2) Eurocape New Energy, Monaco, Monaco

Share this poster on:

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2015 and are published here unedited

Di Li wants to leave the world better than he found it. This is why he works timely with passion and innovation, to reduce the impact on climate change.

As Wind Service Manager he has always been working on making uncertainty certain. With knowledge, reliability and experience gained from hundreds of wind farms in more than 30 countries, and from prestigious Royal Institute of Technology in Stockholm, his clients witness unbelievable detail to build and maintain their wind farms on an unbelievable scale.


Poster

Poster Download poster (6.97 MB)

Abstract

Benefits of using CFD for Wind Power Forecasting in complex terrain

Introduction

Nowadays, wind power forecasting is mostly done using statistical tools converting wind speeds calculated by weather prediction models to power production. Statistical tools find connections between historical mesoscale model output and observed power and can apply them to forecasted mesoscale model output to predict the power production.
One type of statistical methods are Artificial Neural Network (ANN) which connect mesoscale model as input to wind farm power production as output. ANN provides in average a good forecast but it is almost like a black box where it is not easy to understand what is physically happening and ANN has sometimes limits which can lead to a wrong forecasts.
To provide a different approach Computational Fluid Dynamics (CFD) is used which describes the local wind field around the wind farm. It is a deterministic tool which dynamically downscales the wind speed from the mesoscale model onto the wind farm level. It has a deeper physical basis than the ANN forecast.
We present the case of a wind farm where the forecast using CFD performs better than the standard ANN and we explain the reason for that and the shortcomings in the ANN approach.


Approach

The wind power forecast is calculated and compared using two approaches:
1) One single ANN from the mesoscale model output to the power production of the whole wind farm (ANN wind-power).
2) An ANN connects the wind of the mesoscale model to observed wind conditions on site and this corrected wind is then used by the CFD to transfer it from the wind measurement position to the turbine positions (ANN wind-wind + CFD) as in [1].


Main body of abstract

The wind energy forecasts are calculated for a wind farm with 6 turbines. To do so a CFD model is calculated on an area of 30x25 kilometres around the wind farm. The model takes into account roughness of some forested areas around the wind farm and the presence of some mountains about 9 kilometre on the east side of the farm.
The presence of these mountains is quite important as the main wind directions on the site are NNW and SEE/SSE with most energy coming from SEE/SSE. The mountainous area is quite high with around 1000 metre a.s.l. while the wind farm is around 200 metre a.s.l.
The forecast with the two approaches are run on this site using as training data set data from the year 2014 and the validation is performed over four months from February to May 2015.
The CFD shows a better performance than the single ANN approach. Analysing the periods where the CFD performs better we can recognize that those are mainly cases of wind coming from sectors E and SEE. The CFD is able to recognize the local recirculation of the wind caused by the mountains upstream of the wind farm and describes the under performance of the wind farm due to that.
This under performance happens only in those directions and the decrease in power production affects only some of the wind turbines. A small wind direction shift makes a large change in the final power production that is why the simple ANN forecast is not able to recognize it.


Conclusion

The studied wind farm shows a specific behaviour depending on the direction of the wind. This is due to the presence of high mountains in the east side of it which influence the local wind field.
The CFD simulation is able to recognize this and to describe this behaviour. Using the approach “ANN wind-wind + CFD” to do power forecasting improves the forecast performance compared to forecasts with only ANN.





Learning objectives
The audience will learn about the difference between statistical and deterministic tools in wind power forecasting.
The audience will learn how to perform wind power forecasting based on CFD.
The audience will learn when it is better to use deterministic tools than statistical tools.