With the rapid development of the Industrial Internet of Things (IIoT), industrial networks are encountering increasingly frequent and complex network attacks, leading to significant economic losses to production environments. Traditional intrusion detection methods have shown significant limitations in dealing with new types of attacks or their variants. Although deep learning technology has brought new possibilities to network intrusion detection, it still cannot effectively capture structural information and potential patterns in network data (such as the camouflage of attack flow). To address these challenges effectively, this paper presents a novel approach for network intrusion detection by leveraging Graph Enhanced Graph Neural Networks (GEGNN). After thoroughly studying the complex network topology and interaction capabilities between individual IP addresses, we propose a concise and efficient graph structure model. Meanwhile, we design a structure enhancement model, which, through research on both local enhancement and global processing, enables the model to better cope with the instability in network data and improve its ability to recognize complex structures. We comprehensively evaluate the proposed Network Intrusion Detection System (NIDS) model on two latest intrusion detection datasets. Experimental results demonstrate that our model outperforms existing methods in key metrics such as detection accuracy and F1 score.