The invention belongs to the technical field of power frequency communication channel state analysis algorithms, and particularly relates to a method of evaluating the state of a power frequency communication channel. The method comprises steps: through acquiring data influencing the quality of the power frequency communication channel such as lengths of different power lines, the power line branch number, a branch length, load impedance characteristics, power line types and power line noise, a power frequency communication channel quality evaluation data storage and sharing platform is built; on the basis, the data are mined and calculated; and an Adaboost ensemble learning algorithm is used, through an expert analysis system and a neural network algorithm based on deep learning, the state of the power frequency communication channel under multiple parameters is evaluated comprehensively. Compared with the traditional simplex and fixed channel state evaluation method, the method of the invention can continuously learn and optimize the evaluation model according to changes of influence factors, and a more accurate channel state is obtained.