Z-FFR (Z-Pinch driven fusion fission hybrid reactor) is composed of a Z-Pinch driven device, a fusion physics target chamber and a sub-critical reactor, the physical designed of the driving device is a key and central step, which plays a leading role in the engineering of the entire device system. There will be millions of capacitors, inductors, resistors, etc. in the circuitry, and the model parameters will be obtained through machine learning, so that the final parameters can be optimally accurate on the training set with minimal loss, and this approach is geared towards the learning outcome. With this initialisation parameter, only a small number of learning samples are needed to rapidly converge in the model in the face of similar scenarios of non-linear non-constant physics problems, thus improving the generalisation capability with few samples and balancing accuracy and learning efficiency. driven fusion fission hybrid reactor modeling is of great significance.