In view of the accuracy and low time efficiency of the traditional load curve classification method based on statistics, the non-invasive load monitoring and decomposition technology is extended and applied to the bus load curve decomposition of substation in this paper. A method for identifying the daily net load curve of the bus based on SVM and SCADA big data is proposed by considering the output of new energy. Firstly, the change process of the load active power curve of typical industries is analyzed, and the active power mutation time for load pre-screening is extracted. Secondly, Fourier series are utilized to fit the active power waveform. The industry load feature tags are obtained, waveform features are extracted. In addition, for the purpose of identification of the industry load characteristics, the support vectors machine is employed for the waveform characteristics classification and recognization for the daily net load curve of the substation bus. In the end, a 330 kV substation in Gansu Power Grid is simulated by SCADA for verification. It is shown that this method can effectively classify the bus load, thereby improving the efficiency of load modeling.