The invention discloses a characteristic quantity selection method and a characteristic quantity selection system applied to transformer hot spot inversion. The characteristic quantity selection method comprises the steps of: S1, designing a multi-factor and multi-level training sample set by adopting an orthogonal method according to the actual environment of a transformer; S2, performing temperature field simulation on the transformer under conditions corresponding to training samples, so as to obtain a temperature field distribution diagram reflecting hot spot distribution and oil flow temperature distribution inside the transformer; S3, acquiring a streamline diagram of the transformer according to the temperature field distribution diagram, and selecting a plurality of temperature characteristic points on an outer shell and radiating fins of the transformer according to the streamline diagram, wherein the temperature characteristic points and environmental factors constitute basiccharacteristic quantities; S4, and optimizing the basic characteristic quantities by utilizing an artificial intelligence algorithm. The characteristic quantity selection method and the characteristic quantity selection system can select the characteristic quantities having universality, and can be better applied to the engineering practice; and by applying the characteristic quantity selection method to transformer hot spot inversion, the problems of high difficulty, low precision, slow calculation speed and the like of an existing transformer hot spot measurement method can be solved.