The invention discloses a transformer top oil temperature anomaly monitoring method based on multi-dimensional information, which comprises the following steps: A, initializing a CS algorithm, establishing a fitness evaluation model of a bird nest position, and initializing a fitness value; B, calculating a global optimal solution and a local optimal solution of each bird nest position; C, respectively constructing a Susa thermal circuit calculation model and an SVR regression compensation model according to the identified parameters; D, based on the Susa thermal circuit calculation model, obtaining a preliminary top oil temperature prediction value; E, based on the SVR regression compensation model, obtaining a final prediction value of the top oil temperature; and F, carrying out short-term trend prediction on the residual error after moving average by adopting an ARIMA method. According to the scheme, the environmental conditions of the transformer are fully considered, and the keyparameters in the temperature rise model are identified and corrected, and the residual error of the top-layer oil temperature is subjected to trend prediction, and the monitoring precision of the top-layer oil temperature of the transformer is improved.