For fun and learning.

[mordred]

http://www.webmet.com/met_monitoring/621.html

this link is one I sent to Peter at the beginnning of this problem its the meteorologists way of handling the measurements. I've also seen this problem on another control forum they however could not come up with a solution that satisfied everyone. You will notice that one of the things they mentioned is that wind direction is defines as between 1 and 360 degrees. 6.2.1 to 6.2.3 covers how to handle the standard deviation. You can see that they mentioned the statistical mean values covering section 6.2.4. as a side not they have some exceptions where incorrect data can occur for this part. So the go further. If you look at 6.2.10 they mention the effects of meander wind low wind that for long periods of time overwrite statistical data so it jumps you to another page 6.4.4.

Your correct in the definitions of the problem if for example I had to design a system that requires highly accurate wind direction and speed the vextor base method simply would not cover all the possibilities. Or for weather forecasting for that matter. After reading this and thinking about what I would need to program this I start to prefer Peters alternatives in post 25.

I would be curious if a combination of Peters idea and adding statistical equations to the mix could compensate for some of the problems mentioed

 

[mordred]

Lets take this a step further with this question what if my coustomer wants the daily average pick whatever sample rate you care to choose.
Vector based averaging cannot handle that. Or at least not by itself

 

[Peter Nachtwey]

Lets take this a step further with this question what if my coustomer wants the daily average pick whatever sample rate you care to choose.
Vector based averaging cannot handle that. Or at least not by itself

Why not? The vector based average is truly an average. It is just that in your example you don't want the average in a true sense but in a practical sense.

For a practical solution I would average the wind speeds to get the average wind speed. I would not care about direction just the magnitude. Then use the incremental direction technique as in the link. If the wind isn't blowing I would not average in the direction for that minute. I think this is the simplest and most straight forward solution. I don't see any obvious flaws other than the one listed in the link where if the wind rotates more than 180 degrees in one sample time.

 

[mordred]

That makes more practical sense with what you stated above and you are correct I was worried about the practical sense as I ran into this problem with that SQL system so I honestly wanted to know if there was a better way to handle a problem of this nature in case I come across it again down the road. Also to ensure that someone does not simply try the Vector method without considering the variables and problems we mentioned in this thread.

 

[PcRider]

Friday night is a good night because there is little activity and there is the whole weekend to work on the problem.
There weren't many players. I can think of all sorts of problems but most are too difficult.

Here is a good one though and not too difficult, calculate the average wind direction when sampling every minute. Most of the time it will be easy because the wind will not come from the north but if it does it will come from 355 degrees some times and 05 degrees other times and the average is 180 which isn't right.

Should the direciton be weighted by the wind speed? How would you do that?

You guys are WAY over my head so forgive me if this sounds to simplistic. When you talk about averaging the direction, aren't you really talking about an average of the amount of time spent at any given direction within a given time period?
I don't know calculus so maybe you said that and I didn't get it.
Seems like you would have to use many intervals in order to get any real data no matter how you get the average, so the question of using only two intervals would be pointless in any circumstance.

humbly PcRider

 

[Peter Nachtwey]

You guys are WAY over my head so forgive me if this sounds to simplistic. When you talk about averaging the direction, aren't you really talking about an average of the amount of time spent at any given direction within a given time period?

Yes, the real question is should the direction be weighted by the wind speed?

I don't know calculus so maybe you said that and I didn't get it.

No calculus is required. Just vector math and trig. I had this in high school.

Seems like you would have to use many intervals in order to get any real data no matter how you get the average, so the question of using only two intervals would be pointless in any circumstance.

True, some sort of averaging technique is required whether it be batch, rolling average or exponential averaging.

 

[Pego]

Sorry, what was the question...?

I think someone should post a PLC programming challenge each week. Must post via Acrobat since we all use different software.

 

[SMOKE]

There use to be a stump me test here or another place. It was way math oriented. Simply an average is the result of the test data divided by number of tests. I do not think what Peter wanted to ask is the average. Simple e.g. 15 results are wind North and 15 results wind South and you use the 360 thing then the average is average wind blew East. The wind never blew east since we know it blew. Other wise the wind never blew. The question is not valid. I sorta think Peter did this on purpose.

 

[Bobbias]

Well, yeah, the purpose was to give us something that looks easy at first, but the traditional way of solving wouldn't give us very satisfactory results.

 

[mordred]

I'm not sure that was Peters original purpose the flaws came up as we looked closer at the way most people solve this with the vector averaging method using Arctan posted earlier.

 

[Bobbias]

lol, well, intended or not, it's a great example of exactly that sort of situation. It's pretty easy to see that if the wind moved 180 degrees between samples, that would pose a problem, and that averaging it normally, would possibly end up with it saying the wind came from somewhere it never actually did.

 

[Peter Nachtwey]

I saw the thread at control.com

http://www.control.com/thread/1026210133
They were having fun with it.
Sometimes there are two or more imperfect solutions.
I think that adding the vectors is just as valid as doing the incremental changes. It all depends on the intended purpose.

 

[Bobbias]

Well the "right" answer is always the one that satisfies the need, not the most "mathematically perfect" solution.