The invention discloses a remote fault prediction method suitable for a phase modifier system. The method comprises the following steps of determining normal value ranges corresponding to all operation parameters; in combination with known faults, adopting an artificial intelligence learning technology, and creating corresponding intelligent monitoring models for different electrical devices according to device parameter characteristics; comparing acquired current operating parameters with the corresponding normal value ranges, judging that the phase modifier system is in an abnormal state atthe moment when any one operating parameter exceeds the normal value range, positioning abnormal electrical equipment, and generating an abnormal early warning signal; and importing the acquired current operating parameters of the abnormal electrical equipment into the corresponding intelligent monitoring model, performing fault analysis on the electrical equipment, and outputting fault types, fault mechanisms and troubleshooting suggestions. On the premise of discovering early-stage potential defects, state false alarm can be avoided as much as possible, and the automatic monitoring is realized through the comprehensive consideration of multiple parameters and the consideration of environment and external factors.