Why does the use of "svpwm generator 2-level"simulink module have a step change in motor speed in the closed-loop control of the motor

After building the following simulation model, the sampling method is set as discrete sampling, the sampling time is 1e-6s, and the ode3 solution method, the motor speed results show step phenomenon. Among them, the SVPWM module configured to alpha beta - components, Pattern1, 100000 hz, 1 e - 6, the Universal bridge module configuration for the IGBT, other parameters not mentioned as the default parameters.

Finally, what sampling and decampling method is suitable for the "SVPWM Generator(2-Level)" module? Is it better to be discrete or continuous? Which solution should be chosen？

Answer 1

[Joel Van Sickel]

For the 2nd motor, you have a single simulink line for dq0, except, the 0 value (3rd element) is the top motor's speed. I don't know if that motor is spinning or not, but that would mess up any of your clark/park transforms if you are taking zero sequence into account) You shoud set the value explicitly to 0 for this model since you are explicitly defining your dq0 values. However, if your top motor isn't spinning, you will accidentally be sending the correct value of 0. Can you share the model if this doesn't fix things?

[xue li]

I have set the third input of the coordinate transformation to 0, but it still does not solve the simulation result step problem. In this process, it is found that the above problems can be reduced or even eliminated by reducing the PWM frequency, but the actual IGBT switching tube frequency of 5k~10k is very normal. Therefore, it is not reasonable to reduce the PWM frequency. Is there any other suggestion to eliminate the step?

In addition, the upper part of the model is self-written m file to implement the seven-segment SVPWM. Compared with "SVPWM Generator(2-Level)" in the library file, high switching frequency and discrete solution Settings will have a step phenomenon, but self-written m files will not have a step in continuous solution.

Finally, could you please give advice on what solution is suitable for the SVPWM algorithm in the motor control process?

Simulation model and main. m has been placed in the attachment.Thank you for your help！

[Joel Van Sickel]

It looks like for this pwm generator, to get better results, you need to change the PWM resolution of the PWM generator. I got rid of most of the steps by switching to 0.5 us. It gets much better at 0.25 us. However, this will slow down your simulation. Unloaded motors typically do require smaller time steps to simulate accurately. Sometimes you can also put a snubber resistor on their terminals to allow for accurate simulation with larger time steps (but it adds innacuracies to the simulation). You can also set the PWM generator to a sample time of 0 to make it continuous time and get a very clean response. I didn't have your SVPWM file so I can't comment on the version that you implemented working where this one needed a smaller time step.

I would use the solver ode23tb for a specialized power system model like this.

[xue li]

Due to some work issues, please forgive me for not replying in time. Thank you for your help.