The invention discloses a thunder and lightning transient signal identification method and system based on spectral density attenuation. In an existing thunder and lightning transient signal identification scheme, a traveling wave head is difficult to identify, a sampling rate requirement is high, and setting lacks a theoretical basis and so on. The method of the invention is characterized by extracting the transient signal of a direct current power transmission system and carrying out wavelet transform, calculating a wavelet singular entropy, determining the ratio of a spectrum density, and identifying a fault signal, a lightning shielding failure interference signal, and a counter interference signal; and determining according to a recorded data energy distribution coefficient, and identifying a common short circuit fault signal and a lightning fault signal. In the invention, a thunder and lightning transient signal can be well identified, and the method and the system are less affected by factors such as a fault distance, a transition resistance and the like.