The mixture fluids of oil–gas–water metamorphose through the change of phase, properties and structure. This poses a challenge to decipher their dynamic entanglement for phase velocity measurement. It is evident that progressively grasp of process states provides valuable prior information for further flow velocity measurement. As a significant flow characteristic, multiscale gives unique inspiration for flow state monitoring and velocity measurement from a global–local view. In this work, empirical wavelet transform-based Hilbert-Huang transform (EWT-based HHT) and multiscale robust empirical permutation entropy (Ms-rePE) are employed for the analysis of CWUD and conductance signals respectively, achieving the joint characterization of oil–gas–water flow structure by acoustic-electrical dual-mode information. On this basis, a data-mechanism dual-driven model for flow velocity measurement is established under a unified multiscale framework. And a novel approach of hypergraph-based global–local structure analysis (HGLSA) is proposed to simultaneously extract global–local information embodied in the complex flow process. The proposed scheme of global–local state monitoring-based flow velocity measurement is verified by the dynamic experiment of oil–gas–water three phase flow.