In order to guarantee the normal operation and maintenance of the transmission line under the icing and galloping conditions, a novel early warning method of transmission line galloping based on BP neural network is proposed through treating it as a problem of classification and prediction under supervised machine learning. Through analyzing the external meteorological factors that influencing galloping, a BP neural network learning algorithm is established by taking wind, inducted angle of wind direction and line, relative humidity, and ambient temperature as input vectors. The galloping probability is predicted by judging whether the prone-galloping weather conditions are satisfied utilizing the proposed method, and its prediction performance is assessed through several test indexes with the purpose of improvement. A case study is presented by adopting historical galloping data of Henan power grid, and the result shows that the proposed method is effective and practical, which can provide support for power system operation staffs to make reasonable decisions as well as ensure the power grid securely tiding over the peak-load during winter. This work is supported by Major and Basic Project of State Grid Foresight Science and Technology (No. SG20141187).