Development of Multivariate Statistical-Based Tools for Monitoring of Sour Water Unit

摘要

High production combined with high product quality and low environmental damage and energy consumption represents one of the main challenges of any chemical plant. In petrochemical plants, sour water treatment unit plays a key role regarding environmental issues, due to the necessity of ammonia and sulphur-based compounds removal. Although fundamental-based models can be successfully used for monitoring tasks, faster models based on multivariate statistics theory have been widely introduced in industrial sites. They present attractive features such as the capacity of handling a high amount of data, contain the process history and can be used for on-line and realtime process monitoring. In this study, multivariate statistical based methods were used to develop an alternative tool for an actual operating sour water treatment unit monitoring. Input variables collected from the process history were used to not only adequately predict the behavior of a given response variable, but also another potential use of this is to gain understanding of how and to which extent different process variables affect a specific response variable. Some issues like the length of predictive horizon, data structure, data selection and treatment during the development of the tool are also discussed. The obtained results indicate that multivariate techniques were able to monitor the unit variables behavior, representing a key alternative tool for sour water unit monitoring.