Natural gas production process fault localization based on direct transfer entropy with adaptive lag

Abstract

Natural gas production equipment has complex structures, numerous monitoring parameters, and intricate correlation relationships, making it challenging to trace the source of faults for abnormal events. This paper proposes a fault localization method for natural gas production processes based on direct transfer entropy with adaptive lag (AL-DTE). Firstly, a variable set is screened based on the contribution plot. To enhance the adaptability to multiple operating conditions and reduce the reliance on experience, this paper integrates the adaptively to DTE and proposed AL-DTE. A causal analysis is then constructed and the causality graph is utilized to visualize the abnormal propagation path and identify the root cause of the fault. The proposed method can locate abnormal and faulty sub-equipment without relying on historical experience, enabling rapid localization of unknown faults. The effectiveness of the method is verified through simulation data and real faults of a triethylene glycol (TEG) dehydration equipment.