The invention discloses a peak-valley electricity quantity prediction-based residential user distributed internet-surfing optimal selection method. The method solves the technical problem in the prior art that the precision of the power demand prediction is not high. The residential user distributed internet-surfing optimal selection method comprises the steps of S1, obtaining the peak electricity quantity and the valley electricity quantity of a target user during each quarter of the past n years, wherein n is more than or equal to 4 and is an integer; S2, according to the peak electricity quantity and the valley electricity quantity of the target user during each quarter and obtained in the step S1, predicting the peak electricity quantity and the valley electricity quantity of the target user during the fourth quarter of the m year within a prediction period, wherein is larger than or equal to 1 and is an integer; S3, according to the peak electricity quantity and the valley electricity quantity of the target user during the fourth quarter of the m year predicted in the step S2, determining the annual total electric charge and the annual total electric quantity of the target user in the m year within the prediction period, and figuring out the annually average electric charge of the target user in the m year; S4, comparing the average electric charge of the target user in the m year with a current on-grid price; selecting the allowance mode for the internet-surfing if the annually average electric charge of the target user is larger than the current on-grid price; and selecting the all internet surfing mode if the annually average electric charge of the target user is smaller than the current on-grid price.