Plantwide simulation and monitoring of offshore oil and gas production facility

Khaled, Mohammad Shaluddin (2020) Plantwide simulation and monitoring of offshore oil and gas production facility. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (2MB)

Abstract

Monitoring is one of the major concerns in offshore oil and gas production platform since the access to the offshore facilities is difficult. Also, it is quite challenging to extract oil and gas safely in such a harsh environment, and any abnormalities may lead to a catastrophic event. The process data, including all possible faulty scenarios, is required to build an appropriate monitoring system. Since the plant wide process data is not available in the literature, a dynamic model and simulation of an offshore oil and gas production platform is developed by using Aspen HYSYS. Modeling and simulations are handy tools for designing and predicting the accurate behavior of a production plant. The model was built based on the gas processing plant at the North Sea platform reported in Voldsund et al. (2013). Several common faults from different fault categories were simulated in the dynamic system, and their impacts on the overall hydrocarbon production were analyzed. The simulated data are then used to build a monitoring system for each of the faulty states. A new monitoring method has been proposed by combining Principal Component Analysis (PCA) and Dynamic PCA (DPCA) with Artificial Neural Network (ANN). The application of ANN to process systems is quite difficult as it involves a very large number of input neurons to model the system. Training of such large scale network is time-consuming and provides poor accuracy with a high error rate. In PCA-ANN and DPCA-ANN monitoring system, PCA and DPCA are used to reduce the dimension of the training data set and extract the main features of measured variables. Subsequently ANN uses this lower-dimensional score vectors to build a training model and classify the abnormalities. It is found that the proposed approach reduces the time to train ANN and successfully diagnose, detects and classifies the faults with a high accuracy rate.

Actions (login required)

View Item