Aiming at fault diagnosis of wind turbine pitch system affected by state delay, a new fault diagnosis method based on multiinnovation stochastic gradient algorithm was proposed. This complex system was modeled as a state space model, and the system identification model was established. The algorithm extended the innovation scalar into innovation vector to improve the accuracy. The changes of parameters could be caused by the fault of the system, and the algorithm was used to estimate the parameters of wind turbine pitch system with state delay. The fault diagnosis issue was transformed into identification problem. Simulation results showed that this method could achieve the purpose of diagnosing the fault of wind turbine pitch system with state delay.