The invention discloses a transmission line comprehensive fault location method for intelligent substations. When a transmission line fails, a method of highest ranging accuracy is selected from the existing single-end power-frequency ranging methods to give accurate fault position information. The existing power-frequency ranging methods have different system error sources, and the ranging accuracy is affected by the power source, the impedance of an opposite-end system, the transition resistance and other parameters. Thus, under different fault conditions, the methods differ in ranging accuracy, and the ranging results are different. According to the method, a large number of training samples during failure of a transmission line are acquired first, the attribute of the training samples is reduced based on the rough set theory, the internal relation between the ranging accuracy and fault conditions is found out, and less influencing parameters are deleted to speed up the solving process. When a system fails, a method achieving a most accurate ranging result is found from a variety of power-frequency ranging methods using a KNN algorithm. Therefore, the ranging accuracy is improved effectively.