I’m performing a set of simulations were some basic parameters are modified (such as wind conditions).
I have to analyze the output data (such as generated power) and I’d like to know what is a reasonable simulation time. My doubt arises because I’ve read posts were in a 400 sec simulation the first 100 sec must be ignored to obtain reasanoble data. I have to perform a significat number of simulations so I’d like to know what is an optimal time of simulation.
This depends on what outputs you are looking at. You mention generator power, often only which the mean value is critical. Mean values tend to converge faster than standard deviations, and standard deviations tend to converge faster than extremes. The IEC typically requires that you run six seeds of 10-minute simulations for each condition, or equivalently, one seed of one hour. These times are after the start-up transients die out. Start up transients last longer when proper initial conditions are not set, but are typically 30-100 s in length when good initial conditions are set (shorter for land-based wind turbines and longer for floating wind turbines).
Thanks for your answer. I’ve been working on my project to try to rephrase the question in a more clear way.
The goal is to compare the generated power of a 5MW onshore wind turbine (5MW_Land_DLL_WTurb) with different wind profiles (it is mainly focused on steady wind conditions). A 3D graphic will be created so that it is easier to understand. It will involve the wind speed varying from 3 to 25 m/s (the cut-in and cut-off wind speeds), the power-law wind-profile exponent (PLexp) varying from 0.05 to 0.5 and the generated power.
My doubt arises because I don’t know what the quickest way is to perform all these simulations. I’ve came up with two possible solutions. The first one is to perform a 3600 s simulation (according to the IEC conditions you mentioned in the last message) for every PLexp at each wind speed. It will look something like this:
PLexp = 0.05 → WSpeed = 3 m/s
PLexp = 0.05 → WSpeed = 4 m/s
PLexp = 0.05 → WSpeed = 5 m/s
PLexp = 0.10 → WSpeed = 3 m/s
PLexp = 0.10 → WSpeed = 4 m/s
PLexp = 0.10 → WSpeed = 5 m/s
The second solution is faster, but I don’t know if it is very rigorous. I will create a single “.hh” wind profile file where the wind speed increases a unit every 400 s (for example) for each PLexp. It will look something like this:
I’ve chosen 400 s because it looks like it’s a time enough to eliminate the start up transients like in this picture happens:
Finally, what I obtain when simulating the last “.hh” wind profile is this:
Hope I’ve been more clear this time. I’ll be pleased to know your opinion about this issue I’m facing.
The latter option you mention is likely sufficient if you are not interested in considering the effects of turbulence. Of course, if you have a nice computer with many cores, it would probably be faster to run separate simulations for each wind speed/shear condition in parallel (in batch mode).