Wind Learn Mode

[pechan]

Good Morning boB,

Any updates on Wind Learn mode for the Classic?

[ChrisOlson]

IMO, a Wind Learn Mode might be hard to do properly.  In my experience running lots of wind turbines with Classics I've found that programming a power curve for a particular turbine is not always programmed to the peak power output of the machine.  This is done on-purpose, especially at high outputs, to keep the machine quiet and under control by running the rotor at less than ideal tip speed ratio.  Every turbine is different, so trying to come up with an automatic learning algorithm that covers all might not be the best thing.

If a thing like Perturb and Observe were used to "learn" a wind turbine curve, I could see where you would purposely program a curve to operate a turbine at slower tip speed, then have the "learning" algorithm find out that the turbine could put out more power at a specific wind speed and end up with a screamin' demon on the tower that wear blades out faster than you can bolt 'em on. 

[Halfcrazy]

Chris
Thats a good point. It is very true that a lot of times a curve is "De Tuned" for particular reasons and learn mode could cause issues there. Definitely something to think about

[boB]

Well, the idea of wind learn is to find the max power point curve which is usually the optimum TSR of the whole turbine.

Not taking into account noise or anything like that.  De-tuning can be done after the max curve is found.

Perturb and Observe cannot be done to find the optimum curve except with a wind tunnel, unless you mean
P&O over a long time, like, hours or days or weeks.

It should be a lot of fun if and/or when we can get it to work though.

boB

[dgd]

Well, the idea of wind learn is to find the max power point curve which is usually the optimum TSR of the whole turbine.
...
Perturb and Observe cannot be done to find the optimum curve except with a wind tunnel, unless you mean P&O over a long time, like, hours or days or weeks.

When my Clipper arrived in March last year I noticed the connector marked RPM and I sort of imagined that in the (near) future this would become an AUX input for the Classic's turbine management and perhaps wind learning  curve processing. At least providing an upper limit detection to prevent self destruction in high wind.

However,  as is, the Clipper should prevent runaways as long as its resistors are good values.

So what would wind curve learning achieve?  Just fine tuning an existing curve would be nice but probably not worth the effort for possibly small returns in power.

Where I thought it would be really useful is to allow a turbine to be connected to a Clipper/Classic combo then starting with a generic wind curve (or more likely a linear line for inV/outA) have the learn mode P&O the curve voltage setpoints to get the near max A out at each in Volts.

Is actual wind curve learning more complicated than this?

dgd

[boB]

Where I thought it would be really useful is to allow a turbine to be connected to a Clipper/Classic combo then starting with a generic wind curve (or more likely a linear line for inV/outA) have the learn mode P&O the curve voltage setpoints to get the near max A out at each in Volts.

Is actual wind curve learning more complicated than this?

dgd

Starting with a linear curve (or something like that) is exactly how it would start out.  i.e.  not knowing anything about
the power curve.

It is more complicated that P&O because, for one,  wind turbine inertia, and 2, the wind never sits still long enough
to do a test on it except for a wind tunnel where the wind can be made constant long enough to find one point
on the curve.  Then, you can crank up the wind (or crank it down) and add another point.

Yes, the RPM output could possibly help but would not really be necessary to help with wind learning AFAIK.

It should be learnable by taking enough time to get the wind power curve betterized.
boB

[dgd]

It is more complicated that P&O because, for one,  wind turbine inertia, and 2, the wind never sits still long enough to do a test on it

Ok, the wind tunnel method is never going to happen.
P&O on a single voltage input point is also not practical so the method would probably need to use the rising voltage between voltage setpoint mid values and on detecting this note if the averaged Aout over the voltage rise time matched the Aout for the voltage setpoint.
If it did then adjust up the Aout for the setpoint. Keep doing this until a max value Aout is found for the voltage setpoint. May take several dozen/hundred iterations to focus into the true value for Aout.

Although the current wind graphs use 16 voltage set points there could be many more so the graph becomes better shaped.
Every time its windy and the blades accelerate through set point value there could be many averaged Aout values obtained for tuning the overall curve. Probably just one or two windy days and the curve would be pretty good.

Falling voltage over the same range would not be used since the inertia of the turbine as it slows will not be wind power via loaded blades.

I suppose the real question is whether there is enough resource in the Classic's cpu to implement a wind curve learning routine with mutiple point updating.
Maybe the Clipper is where this should be happening with a learn module 

dgd

[dgd]

Is there a set of modbus registers that hold the current wind curve voltage and matching current values?  If so are these directly writeable so that changes take effect immediately?

dgd