Enhanced Monitoring of Water Quality in Crude Oil Desalting/Dehydration Plant (DDP) using Soft Sensing Techniques

The present study introduces a novel soft sensor based on State Dependent Parameter (SDP) models utilizing the Local Instrumental Variables (LIV) method for monitoring a crude oil Desalting and Dehydration Plant (DDP) system. A key advantage of the LIV modeling method is its ability to interpolate directly without necessitating extensive model parameterization. Additionally, the inherent complexity and non-linearity of the process are effectively addressed by LIV-based soft sensors, which require fewer process variables, thereby reducing training time and computational complexity. Two distinct soft sensors were developed to assess the salinity efficiency and water cut efficiency of the DDP system. The efficacy of these soft sensors was evaluated using a dedicated testing dataset, revealing a robust correlation between salinity efficiency, water cut efficiency, and five secondary parameters. Comparisons between SDP-LIV model predictions and real observations of the DDP process show strong agreement. By leveraging these developed soft sensors, continuous evaluation of product properties is possible with minimal delay compared to traditional laboratory analyses. This capability is crucial for pollution control and environmental monitoring, as it allows for real-time detection and mitigation of contaminants in crude oil processing. Lastly, the performance of the proposed soft sensor is benchmarked against other models, such as Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN), demonstrating superior predictive capabilities. This study underscores the potential of SDP-LIV-based soft sensors in enhancing environmental protection and operational efficiency in crude oil processing.