Practical Hydrocarbon Allocation – A Machine Learning Approach

Abstract

Abstract The current conventional method of hydrocarbon production allocation of reservoir fluids’ contribution to the separate producing strings from commingled production in the oil and gas industry considers some rigid assumptions that make the allocated volume mostly unreasonable. This uncertainty is usually due to the fixed decline rate assumption normally adopted for a specified period until a new well test of the contributing strings is available to generate a new allocation factor anytime production allocation is required. This paper presents an artificial intelligence approach in the determination of real-time allocation factors for determining contributing production flow performance from commingled production using a machine learning algorithm with the fixed rate assumption using well flow parameters such as the flowing tubing head pressure, flowline pressure and other well parameters to generate transient rates for the producing strings, to create new allocation factor when required. Data from marginal fields in the Niger Delta were used as case studies and the results generated from this exercise after proper data pre-processing depict reasonable output with precision of high confidence level. Results from this approach can also be used in the absence of a reliable well test.