This paper considers the Volterra-Hammerstein (V-H) system identification problem by means of a time-varying parameter expression. In order to make full use of the time-varying parameter data information, the multi-innovation identification theory is used to batch the data. The real-time identification of time-varying parameters is achieved during the processing of extensive information data sets. By incorporating the concept of weighting and forgetting factor, we derive a weighted multi-innovation forgetting factor gradient algorithm, which can enhance the ability of tracking the time-varying parameters and improve the flexibility of the algorithm. Numerical simulation example tests the effectiveness of the proposed algorithm.