Roughness & Energy Production

[jacksonlord]

I have been playing with surface roughness change lines and roses in my layout and getting some very curious AEP results. There are two main issues happening, (1) the production results from roughness change lines in a terrain which is presumably run on the entire layout are substantially different than the production results from roughness roses with the same roughness value placed on each turbine of the site, and (2) roughness roses have opposite effects on wind atlases than I would expect--when the roughness value is increased in the rose child to the met station, the predicted wind speeds and power densities go up for each R-class and height.

(1)- When using a roughness change line as part of the vector map which is placed as a child to the project, it has a very slight effect on the calculated expected production value of the entire site (for a particular SWT 2.3-101 turbine, a change line which effectively set the entire area with a Zo=0.3, the AEP=10.494 MWh; with the change line removed, giving the default Zo=0.03, AEP=10.554 MWh). This is a very small difference, and doesn't seem accurate to me. Vegetation with a Zo=0.3m, like a short forest, would very negatively impact the flow in a site as compared with farmland. This difference does seem to be recognized when a roughness rose is placed as child to the individual turbine, bringing the AEP=7.973 (seems like almost a little too much, actually). Why such a large discrepancy between change lines and roses?

(2)- When a rose is place child to a met station, higher values of uniform surface roughness cause expected average wind speeds and power to go up in the recalculated wind atlas. This seems counter-intuitive: with taller vegetation, there would be a decrease in expected wind speeds at each height, right?

Perhaps I don't fully understand the model, but not a lot is written about how exactly the roughness values effect the results in the help doc.

Also, is there a way to make the roughness rose a child of a wind farm layer, rather than having to repeat loading the rose as a child for each individual turbine in a layout?

[Duncan]

Hi JacksonLord,

I'd like to try to answer part two of your question.

As the aerodynamic roughness length around a site increases, the wind you'd expect to measure there decreases.

The wind atlas calculation is taking observed wind data and cleaning them for site effects, so the 'cleaned' wind for a rough site will be faster than for a smooth site.

For the same measured wind speed, the site-cleaned wind will be higher as the roughness length increases. For example, if you measure 10m/s just above a forest canopy, then the general conditions must be very windy to produce that measurement. The data in the wind atlas are trying to capture or express these 'general conditions'.

Does that help? I hope that I understood your point correctly.

[jacksonlord]

Duncan,

Thanks for clarifying. I understand your answer and it's what I suspected--that the resulting values are those that would be sufficient to "overcome" the roughness/vegetation to result in the observed wind climate. I'm still unclear as to how it is meant to interact with the rest of the items in the project, however. It seems that by adding roughness to the observed data and increasing the atlas values, this in turn increases the wind farm production results--the logic being that the wind is stronger at this location thereby producing more power.

To my humble opinion, this seems counter-intuitive and perhaps contradictory. When you measure wind at a mast, you are finding the wind speeds empirically. To say that they WOULD be higher if not for the given landscape is moot: the given landscape exists and is inextricable. It should be "baked-in" to the values both of the average wind speed and the resulting true production values.

Then there is the question of assigning roughness values on the turbine sites or wind farm. Taking the model as it is, by assuming that the met mast values have a "baked-in" roughness which is homogeneous throughout the site, assigning roughness to the turbines would then be double-counting the effect of the surface on the wind. Therefore, it seems it would be best to assign the roughness to both the met station AND the turbine sites, to counteract the increased wind atlas predictions with the decreased turbine spot flow values, thereby settling the production values to their correct predictions. This is especially true in a site whose surface roughness values are spatially heterogeneous, since different spots will produce different predictions. With spatially homogeneous roughness, assigning a roughness rose to the met station and the turbine sites should cancel each other out, since one is counting up and the other down.

I tried this out in WAsP and found it not necessarily to be true. With the Zo=0.3 uniform rose in the atlas, a given turbine AEP=13.2 GWh. With the (same) rose in both the atlas and the turbine, AEP=11.0 GWh. With no roses at all, AEP=10.554. So it seems that a rose on the met station will have a greater impact in magnitude that the rose at the turbine site.

Which all comes back to my original question (1). With a gradient of surface roughness defined entirely in the vector map using roughness change lines, none of this should be a question to the user (although presumably the processes still go on behind the scenes), since the vector map is applied to all items in the hierarchy at once if placed properly. Unfortunately, it isn't consistent with the roses, which was my main issue raised in the first post.

Sorry to be long-winded here (no pun intended)!

-jackson

[Duncan]

Before replying, can I just check whether the heights of your met. station and turbines are the same, and how far away are the roughness change lines?

[jacksonlord]

The height of my met station is 58 m, and the turbines are at 80 m hub height. I figured there would be some generic sheer exponent applied?

There is a single roughness change line which encircles the entire project area and assigns Zo=0.3 both inside and out. It is about 10-20 km from the site.

[Duncan]

Hello again,

I did a few simple tests and (as far as I can see) roughness roses are working correctly as perfect substitutes to the equivalent map data. Note that if a roughness rose is used for a site, this overrides completely any associated roughness data from the map: it's an alternative not a supplement.

You might be interested to have a look at the reference site hierarchy member. This lets you examine the roughness rose which WAsP derives from the map: click on the "Roughness survey" tab to see it. It might be useful when you want to check how WAsP is interpreting your map data. Another place to get some insight is the "Site effects" tab of the met.station, turbine site and reference site windows.

Your experiment which gave AEP predictions of 13.2, 11.0 and 10.6 sounds OK to me. The reason for these differences is likely the height difference between your met station and turbine site. If you set these to be the same, then you're essentially performing a self-prediction and you should see the 11.0 and 10.6 values converge. The model might introduce a slight discrepancy in the self-prediction as the roughness changes between 0.3 and 0.03, so they may never match precisely.

The surface roughness has a greater effect at lower heights. So if your met. station is at 58m and your turbine site is 80, then assigning a longer roughness length to the area will increase the predicted AEP.

You write that...

"When you measure wind at a mast, you are finding the wind speeds empirically. To say that they WOULD be higher if not for the given landscape is moot: the given landscape exists and is inextricable."

... which I think reveals some confusion about the principle of modelling in WAsP. (I'm assuming you are using "moot" in the US-English sense to mean practically irrelevant.)

The WAsP approach is to estimate the site-specific effects and then remove them from or apply them to a wind (or more precisely a wind PDF). So I'm afraid that WAsP's calculation of a wind atlas from measured wind data is doing exactly what you regard as counter-intuitive, and that's pretty much the basic idea of the model.

I hope that helps to explain the program behaviour.

[jacksonlord]

Duncan- thanks for the response.

I see now the logic of this method. I was beginning to realize that this weekend as I examined the atlas again and put two-and-two together. My key misunderstanding, I think, was the nature of the effects on sheer from surface roughness, rather than as a retardation of the wind flow on the entire vertical profile.

Also, I see now that the roughness rose has to be placed both as a child to the met tower and the turbine sites (for a homogeneous site, that is). Only placing a rose under the met tower, for example, would imply that the flow at the turbine site is less inhibited, thus more productive given the same wind resource. That also explains the AEP numbers I reported, as you said.

I'm now getting the rose and change line derived values to converge as well.

Thanks for the tips,
jackson

[brunofunk]

Duncan, I have a 60m lattice tower in the middle on a complex site with 15m trees all over - Avg 6m/s. Applying a displacement height of 10m, I reach a 0.26 shear factor.
In order to match the WAsP selfprediction at 95m-10m with estimated average wind speeds at 95m - 6.8-6.9 m/s - I need to use a roughness class of 4.But it is known that forested regions should have a roughness class of 3.1-3.2. What should I use 3.2 or 4? Using a roughness 4 the estimated AEP is much higher, as WAsP needs to clean much more the wind from the forest effects.

[Duncan]

Hi Bruno, I'm a bit confused by what you mean when you say a "roughness class of 3.1-3.2" can you explain?

[brunofunk]

Hi Duncan,

Let me rephrase:

I have a 60m lattice tower in the middle on a complex site with 15m trees all over - Avg 6m/s. Applying a displacement height of 10m, I reach a 0.26 shear factor.
In order to match the WAsP self-prediction at 95m-10m with estimated average wind speeds at 95m - 6.8-6.9 m/s - I need to use a roughness length of 1.6 .But it is known that forested regions should have a roughness length of 0.5m. What should I use 1.6m or 0.5m? Using a roughness length of 1.6m the estimated AEP is much higher, as WAsP needs to "clean the wind much more" from the forest effects when compared with the 0.5m scenario.

[Duncan]

Hi.

I'm not really qualified to answer this, so maybe someone else will contribute.

I guess you've read the WAsP in the Forest guidelines (http://www.wasp.dk/Support/DownloadFiles/How%20to%20handle%20forest%20in%20WAsP.pdf)

For what it's worth, I'd suggest that it's quite possible for a forest to have a roughness length of 1m or more. So I wouldn't get too tied to this forests=0.5m figure. Depending on the canopy structure, the displacement height could be something other than 2/3 stand height. Both z0 and dzero depend on wind speed too.

What do you know about the forest? Is it slow-growing, close-packed coniferous forest (easier to model)? Or is it mixed-height, mixed-species, varying-density deciduous woodland (harder to model)?

[brunofunk]

Duncan, thanks for the answer. The project is over a homogeneous forest. I estimated it with 15m tree height. 15*0.1=1.5 which is close to the 1.6m value I estimated before. I feel more confident about using other than 0.5m roughness length for forests!