Application of machine learning to predict the time evolution of condensate to gas ratio for planning and management of gas-condensate fields

Abstract

One of the most important parameters for the evaluation, forecast, and management of gas-condensate fields is the evolution of the condensate to gas ratio (CGR) over time. This parameter tends to decrease as reservoir pressure declines. In the conventional approach, gas and condensate samples are collected at the beginning of production and periodically later to conduct laboratory experiments on composition, CGR, and fluid properties. However, sample collection, transportation, and analysis require a lot of time and effort and could be very expensive. Likewise, dynamic models are also frequently used to predict CGR over time. However, these models could include many uncertainties due to ambiguous input data, including reservoir structures, fluid phase interaction, and reservoir property distribution. Therefore, the application of machine learning to predict the time evolution of CGR in this research could be a new and effective approach to supplement conventional methods.