Study of soft sensor modeling based on deep learning

Soft sensor are widely used to estimate process variables which are difficult to measure online in industrial process control. This paper proposes a new soft sensor modeling method based on a deep learning method, which integrates denoising auto-encoders (DAE) with support vector regression (SVR) method. The denoising auto-encoders are designed to capture robust high-level feature representation of import data and the SVR model is employed to precisely estimate output data based on the feature representation obtained from DAE. In case study, the method combining denoising auto-encoders with support vector regression (DAE-SVR) is applied to the estimation of oxygen-content in flue gasses in ultra-supercritical units. The results show DAE-SVR is a promising modeling method for soft sensors.