HETEROGENEOUS ENSEMBLE LEARNING FOR VIRTUAL FLOW METERING APPLICATIONS

With the increase of marginal fields and deepwater offshore fields in the petroleum
industry, the demand for continuous and cost-effective multiphase flow monitoring
becomes increasingly significant for reservoir management, operational diagnosis, and
well-level production optimization. Virtual flow metering (VFM) is a software-based
computational model that represents an attractive solution to meet these rising demands and
accomplish fully integrated operations. VFM also plays a significant role in augmenting
and backing up physical multiphase flow meters. However, mode-driven VFMs are
difficult to deploy and very expensive to maintain, while current data-driven VFM studies
are limited, suffer inherent limitations, and do not deliver performance analysis over the
complete operating envelope.