The fusion of heterogeneous data from multiple sensors is a key issue in the field of smart grids. With the rapid development of computer technology and multi-source sensing, the scale and structure of distributed IoT has become increasingly complex. It generates massive, multi-source, heterogeneous and sparse diverse data sets. The types of this data are constantly changing and the form of access is shifting from static to dynamic. This network heterogeneity poses a significant challenge in handling large volumes of heterogeneous data and maintaining the operational status of sensors in real time. In this paper, a two-channel CNN-based data fusion scheme for multiple sources of information is proposed to address the management of heterogeneous networks and data fusion communication issues. Among them, an improved data collection platform achieves more comprehensive and accurate monitoring and estimation of the network operation status. Finally, the scheme proposed in this paper is subjected to simulation experiments, which prove that the data fusion algorithm not only ensures real-time efficiency but also guarantees high accuracy.