Abstract:
Severe slug flow is undesirable in offshore oil production systems, particularly for
late-life fields. Active control through choking is one of the effective approaches
to mitigating/controlling severe slug flow in oil production pipeline-riser systems.
However, existing active slug control systems may limit oil production due to
overchoking. Another problem in most active control systems is their dependency
on information obtained from subsea measurements such as riser base pressure
for active slug flow control.
Both of these control challenges have been satisfactorily solved through the
introduction of new multiphase flow topside measurements that are reliable and
efficient in providing flow information for active slug control systems. By using
Venturi multiphase flow topside measurements and Doppler ultrasonic
measurements, an active slug flow control system is proposed to suppress
severe slug flows without limiting oil production. Experimental and simulated
results demonstrate that under active slug control, the proposed system is able
not only to suppress slug flow but also to increase oil production compared to
manual choking.
Another objective of this research was to assess the applicability of
continuous-wave Doppler ultrasonic (CWDU) techniques for accurate
identification of gas-liquid flow regimes in pipeline-riser systems.
Firstly, flow regime classification using the kernel multi-class support-vector
machine (SVM) approach from machine learning (ML) was investigated. For a
successful industrial application of this approach, the feasibility of conducting
principal component analysis (PCA) for visualising the information from intrinsic
flow regime features in two-dimensional space was also investigated. The
classifier attained 84.6% accuracy on test samples and 85.7% accuracy on
training samples. This approach showed the success of the CWDU, PCA-SVM,
and virtual flow regime maps for objective two-phase flow regime classification
on pipeline-riser systems, which would be possible for industrial application.
Secondly, an approach that classifies the flow regime by means of a neural
network operating on extracted features from the flow’s ultrasonic signals using
either discrete wavelet transform (DWT) or power spectral density (PSD) was
proposed. Using the PSD features, the neural network classifier misclassified 3
out of 31 test datasets and gave 90.3% accuracy, while only one dataset was
misclassified with the DWT features, yielding an accuracy of 95.8%, thereby
showing the superiority of the DWT in feature extraction of flow regime
classification. This approach demonstrates the employment of a neural network
and DWT for flow regime identification in industrial applications, using CWDU.
The scheme has significant advantages over other techniques in that it uses a
non-radioactive and non-intrusive sensor.
The two investigated methods for gas-liquid two-phase flow regime
identification appear to be the first known successful attempts to objectively
identify gas-liquid flow regimes in an S-shape riser using CWDU. The CWDU
approaches for flow regime classification on pipeline-riser systems were
successful and proved possible in industrial applications.