Abstract: Compared with the forecast of typhoon track, numerical models lack the ability to forecast typhoon intensity. To further reduce errors in typhoon intensity forecast and improve the mitigation ability of typhoon, an equation of the typhoon intensity in the Northwest Pacific that consists of typhoon initial intensity, initial field error term, change of ensemble mean term and dispersion term has been established using the ECMWF ensemble forecast data from 2018 to 2019. Partial correlation analysis and collinearity test are applied. The 2020 data are used to compare and test the forecast effects. Conclusions are as follows. The prediction error of typhoon intensity calculated by the prediction equation is always less than that of various statistics in the ensemble prediction in each prediction time. Among them, the RMSR of 24 hour prediction is lower than the maximum, the ensemble mean prediction and the deterministic model forecast by 34.36%, 14.58% and 20.38%, respectively. Also, the RMSE of 72 hour prediction is lower than the maximum, the ensemble mean prediction and the deterministic model forecast by 25.68%, 12.91% and 11.13%, respectively. In the comparative case analysis of typhoon, the prediction effect of the revised forecast is also better than the ECMWF ensemble forecast and deterministic model forecast, which is the closest to the reality, and can perform better in the forecast of rapidly enhancing typhoon. The prediction equation of typhoon intensity based on ensemble forecast can quantitatively extract key information from ensemble forecast and generate more accurate prediction, which provides a reference for typhoon forecast in the Northwest Pacific.