A Data-Driven Approach for Analysing the Operational Behaviour and Performance of an Industrial Flue Gas Desulphurisation Process

摘要

This paper investigates the use of data-driven modelling techniques, in particular artificial neural networks, to analyse the operational behaviour and performance of environmental control systems at the example of a flue gas desulphurisation (FGD) process. Using real data from a long-term campaign at an industrial plant, different stationary and dynamic models are developed and assessed, and issues related to sensor accuracy, non-stationary process phenomena and the effect of signal processing techniques are discussed. The work illustrates that flexible data-driven methods can represent the state and performance of such a process with good accuracy and provide useful indications about the relation between the process operation and its efficiency.