Abstract: The support state perception and data processing of hydraulic support are the key factors to realize adaptive support of hydraulic support and advance prediction and early warning of roof disasters.However, the existing technology mainly performs statistical analysis of hydraulic support initial support force and end-of-loop resistance.There are problems such as insufficient perception information, in-depth data mining and inaccurate prediction and early warning.The relationship between mining stress, roof fracture and hydraulic support load change is analyzed, and the 'five-stage' viewpoint of the destabilization process of roof rock fracture and the 'double-factor' control method of determining the reasonable working resistance of hydraulic support are explained.The characteristic parameters of the support state of the hydraulic support are given, and a technology architecture for the comprehensive perception of the support state of the hydraulic support based on the coupling relationship between the hydraulic support and the surrounding rock is proposed.Moreover, it is pointed out that the non-contact sensor will be the key to solve the problem of inadequate sensing of the support state of the group hydraulic support.In view of the characteristics of low dimension, small number of samples, and correlation of multiple characteristic parameters of support state perception data of hydraulic support, the analysis and prediction method of hydraulic support support state perception data based on template curve library is proposed.Based on the mapping relationship between mining stress and hydraulic support support state, the technology architecture of roof disaster intelligent prediction platform is proposed, which can realize the advance prediction and early warning of abnormal hydraulic support support condition and roof disaster.