The invention discloses a vehicle-mounted flywheel battery fault diagnosis system based on probability distribution. The structure of a shaftless flywheel battery is taken as a research object; the method comprises the following steps: performing dynamic analysis on three single faults of shafting misalignment, rotor cracks and dynamic and static part rub-impact, establishing a motion differentialequation, establishing three single fault rotor models of shafting misalignment, rotor cracks and dynamic and static part rub-impact according to the dynamic equation, and establishing a dual-parameter system dynamic model for other single fault rotors. Residual errors are obtained through the four models and input into a probability algorithm module, the conditional probability of each Kalman filter model is calculated, the weighted average value of the conditional probabilities of all the states is obtained, and therefore mixed state estimation of the flywheel battery is formed. Interference of human factors can be avoided, the rotor fault diagnosis precision is improved, real-time monitoring can be achieved, development of a flywheel battery control system based on a model is promoted,and application of a vehicle-mounted flywheel battery is promoted.