FvM is right, if you are switching PF capacitors, you'd better use more rugged devices such as mechanical contacts or thyristors. Maybe thyristors would be advantageous thanks to their self-turnoff at zero current crossing. You don't need the fast response of IGBTs. An added bonus of thyristors is they are less prone to destruction from high voltage spikes: though this is NOT the way to turn them on, if you exceed the breakdown voltage they will just conduct; an IGBT, MOS or bipolar could die.
Anyway, my post was about how to simulate a bidirectional device with a unidirectional one. Think of the "AC" bridge's terminals as the "main terminals" of a TRIAC, and connect a bipolar, MOS, IGBT or SCR to the "DC" terminals: the device will always see the same polarity, but will be effectively switching AC to your load.
OR, do as FvM says: two transistors, in anti-parallel connection, each with a series diode, and excite each gate-emitter with separate isolated drivers.