The invention discloses an equivalent salt deposit density (ESDD) prediction and early warning system for a power insulator surface contaminant, which comprises four function units of a genetic algorithm training unit, a BP (Back Propagation) neural network training unit, a genetic BP neural network predication unit and an early warning unit. The genetic algorithm is good at global search, the BP neural network is good at local search, and thus, when the genetic algorithm and the BP algorithm are combined, the prediction accuracy can be improved, and the algorithm convergence rate is improved. The genetic algorithm firstly optimizes the initial weight of the neural network, good search space is positioned, and the BP algorithm is then adopted to search the optimal value in the small space. The application conditions are wide, the system can be applied to insulators of any types and is not influenced by different contaminant accommodation features due to different insulator types, and the prediction method and the early warning system can realize insulator ESDD prediction, and give contaminant flashover and early warning signals.