As Adam said, it's really difficult to trace the source of the problem without having access to the model. This is fine, though -- ultimately, you should try out the troubleshooting :)
It looks like your model runs fine, so the problem is either in the solver settings (is the model running accurately?) or the design itself (are the RLC parameters + Inverter controller a good design?)
For the solver settings, I see that you are using a discrete simulation. If you switch this over to Continuous, it will likely be slowers -- but do the results look the same? If they do not, then you may need to switch a smaller discrete sample time that can adequately capture enough resolution. For these types of systems, the limiting factor is the PWM frequency. I like to choose a sample time that is at least 50-100 times the PWM frequency.
For your design, is it the passive rectifier + RLC filter values, or is it the active inverter control in your "Regulator" subsystem? I would try the following:
- Instead of using the AC Link on the left, try use a constant DC Voltage source. Does your inverter create a nice sinusoidal signal if the DC link is a constant voltage source? If not, then the rectifier on the left isn't doing its job.
- Instead of the voltage regulator on the right, try using a constant PWM generation. That is, use a PWM Generator (2-Level) block to generate a constant duty cycle, frequency, and phases (0, -2pi/3, +2pi/3). Does this open-loop inverter strategy give you a sinusoidal output? If so, then your inverter regulator may need improvements.
Since you say that the model works fine without the DC Link, could it be that those resistors in series with the AC links are eating up most of the power, and there is therefore no more power to overcome the semiconductor forward voltages in the rectifiers/inverters? Not sure...