The invention discloses a method and system for realizing adaptive management of an intelligent gateway through AI edge calculation. The method comprises the steps: constructing a human body activityscene portrait model, carrying out discrimination of content layering and grading, and obtaining the original data of each layering and grading of a motion state in real time; obtaining scene data after edge calculation, constructing and self-adaptively perfecting an identification model, and optimizing a scene portrait model sample of the motion state of the smart home product; finely distinguishing part of portrait features by using an optimized SVM algorithm and combining different levels of data; and through a deep learning algorithm, carrying out self-learning on the plurality of model samples, and optimizing and perfecting a scene portrait model of the motion state of the smart home product. According to the invention, motion conditions of smart home equipment, a human body and the like are collected in real time through the terminal nodes, the activity scene portrait model is established, regular downloading of the user Internet portrait feature model is realized through edge calculation, model verification is carried out through a supervised learning mode of an SVM, and the calculation amount is reduced.