A Virtual Measurement Method of Industrial Power Consumption Scenario Based on Self-Supervised Contrast Learning

With the acceleration of industrialization and the increasing demand for energy, energy consumption and carbon emissions in the industrial sector have become a global concern. Effective monitoring and management of industrial electricity consumption, along with precise decomposition and evaluation of carbon emissions, are critical for optimizing energy efficiency and reducing greenhouse gas emissions. However, traditional electrical measurement methods face limitations due to hardware costs, data acquisition challenges, and real-time computational requirements, particularly in high-dimensional, complex, and dynamic industrial environments. This paper proposes a virtual measurement method based on self-supervised contrastive learning (SimCLR) to meet these challenges. By leveraging unlabeled data through contrastive learning, the method extracts discriminative features from industrial electricity consumption data, enabling accurate load decomposition and virtual measurement under small-sample conditions. Experimental results on a real industrial dataset demonstrate the method's superior performance in terms of F1-score and root mean square error (RMSE), highlighting its potential for intelligent power monitoring systems.