Simulink Model of ROSCO TSR Tracking Mode

[abdulhaseeb1195-web]

Dear,

I am using OpenFAST ROSCO in Simulink, and I am trying to make the control similar to TSR tracking in ROSCO. for which I am using the given parameters from the given DISCON file. I am facing the issue in region 2 from 7 to 10 m/s wind speeds, where my torque and speed response is not the same as ROSCO.

In the Simulink control, I am getting the generator speed from the TSR formula, then it is compared with the generator speed (rotor speed to generator speed) from the wind turbine, which produces an error that passes through the PI control and results in torque. I have used a similar strategy as the Simulink model of ROSCO available online.

For ROSCO, In DISCON file, I am using one second low pass filter option (WE_Mode=0), and so as I have provided in my model as a transfer function (1/(1+s) that is a low pass filter for wind speed, I have also tried with low pass filter with corner frequency 0.20944 as mentiond in Discon file for F_WECornerFreq but results remain almost same.

Can you please guide me if you are doing something in region 2 that I am missing, or in any way, if you can guide me to improve the control?

Below, I have attached the discon file, wind profile, and generator speed and generator torque response for my control (red) vs ROSCO (blue).

Best regards,
Haseeb

[dzalkind]

I think you need to filter the wind speed signal at a lower bandwidth. Otherwise, when the wind speed increases, the rotor speed set point will increase, and that will cause the torque to decrease initially.

[abdulhaseeb1195-web]

Dear, Thank you for prompt response. I agree with you and it may resolve the problem. However, I want to understand since I have selected WE mode (0) that asks for one second low pass filter, does not it mean a filter with one second time constant (1/1+s) or I am making some mistake in understanding ? Best regards, Haseeb

…

On Wed, 8 Oct 2025 at 21:20, dzalkind ***@***.***> wrote: *dzalkind* left a comment (NREL/ROSCO#477) <#477 (comment)> I think you need to filter the wind speed signal at a lower bandwidth. Otherwise, when the wind speed increases, the rotor speed set point will increase, and that will cause the torque to decrease initially. — Reply to this email directly, view it on GitHub <#477 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BYPJIG7KEQ25IM7XG2GQ4M33WVPYFAVCNFSM6AAAAACIU64KKWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGOBSHA4TGMRSHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[dzalkind]

That's my understanding, but I would check the Simulink model to be sure.

[abdulhaseeb1195-web]

Dear,

I have obtained results for region 2 at different corner frequencies in the Simulink model and compared them with ROSCO WE_MODE = 0. No doubt the results for a 0.1 rad/sec corner frequency are better than those of others; however, they are still not the same as those of ROSCO.

I am attaching below my Simulink model as well as the results for you to review.

Please let me know if something is unclear.

Best regards,
Haseeb

[dzalkind]

I'm not sure what question you're asking here, but it's outside of the scope of this repo to comment on controllers other than ROSCO.

Since this is not an issue with the ROSCO controller, I'm going to move this to a discussion.

[abdulhaseeb1195-web]

Dear,

In the previous discussion, I mentioned that I modeled ROSCO in Simulink based on the ROSCO Simulink example available on GitHub. I am using NREL-5MW wind turbine, and I’m currently facing an issue related to the wind estimator mode.

In ROSCO, I’m using WE_mode = 0, which applies a one-second low-pass filter. I implemented the same one-second low-pass filter in my Simulink model (using 1/(1+s)). However, I noticed that in Region 2 (7–10 m/s), the transient responses for speed and torque differ between my model and ROSCO.

I’m unsure whether this difference is due to an error on my end or if it’s an inherent behavior of ROSCO when simulated in Simulink.

I’ve attached screenshots of my Simulink model and have also followed the suggestion to reduce the frequency bandwidth and run simulations with different values. The results are compared with ROSCO’s output.

In the attached plots, you can see that the simulation with a 1 rad/s (1-second time constant) filter (red line) behaves quite differently from ROSCO (black line). I’m trying to understand whether ROSCO is indeed using a one-second filter or if I might be misunderstanding how it’s implemented.

Please let me know if you need any additional details.

Best regards,
Haseeb