Integrating Machine Learning Clustering into PVT Based Reservoir Fluid Characterization Workflows

The characterization of reservoir fluids is pivotal to multiple facets in the life cycle of a hydrocarbon bearing asset. The process of reservoir fluid characterization starts with the collection of representative samples which are then subjected to laboratory testing for the measurement of Pressure-Volume-Temperature (PVT) properties as well as other properties such as the fluid and isotopic composition. Subsequently, an equation of state may be developed from the measured data. Ultimately, the equation of state provides the capability to predict other fluid properties that may not have been physically measured in the laboratory. The process of sampling and conducting laboratory analysis are time consuming and have an associated cost. This paper presents a novel process by which the properties of a characterized ‘parent’ sample may be utilized to infer the properties of a new ‘child’ sample based on compositional data and saturation pressure measurements from the ‘child’ sample.