The autumn identifies the highest mean intensity of tropical cyclones over western North Pacific and accumulated cyclone energy (ACE) serves as the characteristic index for the mean intensity of tropical cyclones. Based on the tropical cyclone best-track dataset from Japan Meteorological Agency, the sea ice data from National Snow and Ice Data Center, and the SST data from Met Office Hadley Centre from 1979 to 2015, the ACE index in autumn over western North Pacific is analyzed and predicted using regression analysis and multiple stepwise regression. The results show that the ACE index in autumn over western North Pacific is of obvious interannual variations and is closely related with El Niño-Southern Oscillation (ENSO), and the maximum (minimum) value appeared in 1991 (1999) when El Niño (La Niña) occurred. ACE is generally stronger in autumn of the El Niño developing years and the tropical cyclones are weaker in autumn of the La Niña declining years. ACE anomalies in autumn over western North Pacific are influenced by the forcing of the mid-latitude anomalous wave trains resulting from the change of the Arctic sea ice and by the SST pattern in El Niño. Due to the abnormal increase of sea ice in the Beaufort Sea, anomalous wave trains (similar with Circum-Global Teleconnection in the Northern Hemisphere) appear at the upper troposphere in summer, whose positive pressure makes the subtropical high over western North Pacific move eastward and northward in summer and autumn. The variations of the subtropical high and the abnormal distribution of SST over the Pacific Ocean affected the Walker Circulation. To the east of Philippine Sea, the eastward vertical wind shear is weaker, which is favorable for developing tropical cyclones. At the place where tropical cyclones are generated, the positive vorticity anomalies contribute to strengthen their intensities. Finally, the forecast skill score of the model is 0.69. If SST or the sea ice was used as the only factor, the forecast skill score would be greatly reduced.