User generated content intelligent analysis for urban natural gas with transformer-based cyber-physical social systems

Intelligent analysis of user generated content (UGC) plays an important role in ensuring urban natural gas safety and controlling process risks. However, most existing analysis methods are single-task driven and ignore the spatio-temporal information of gas sensing data. To address these problems, we propose a Transformer-based cyber-physical social security system (CPSS) for UGC analysis. Specifically, this unified system integrates multiple tasks, i.e. quality assessment and control of gas data, security factors of user consumption, and spatio-temporal abnormal gas signal detection. In the developed Transformer-based model, a time-space cross-attention module is embedded for combining the long-range spatio-temporal dependences of gas data. Moreover, a feature memory block module is introduced for abnormal feature enhancement and high-level representation of gas quality. Experimental results on related gas datasets demonstrate that this Transformer-based method achieves state-of-the-art performance, and the security system significantly improves the safety factor of natural gas use in smart cities, providing a robust framework for risk management and safety enhancement.