According to uncertainty of hot metal silicon content and the blast furnace temperature univariate time series model contains less input information, it is difficult to reveal the relationship between variables and the change rule of characteristics. Taking blast furnace hot metal temperature as the research object, BP neural network multivariate time series model and T-S fuzzy neural network model of multivariate time series were built. It applies a steel blast furnace data simulation test.The results show that the T-S fuzzy neural network multivariate time series model can achieve better shooting and prediction accuracy.