New generalized correlations for oil rate predictions through wellhead chokes for high GOR reservoirs

Accurate oil production rate is crucial for well performance evaluation. Testing separators and multiphase flow meters are associated with many technical and operational issues. The available correlations are based on the Gilbert method with low accuracy at high gas production wells. In this study, the artificial neural networks (ANN) technique was applied to estimate the oil rate (Q o ) as a function of the wellhead choke parameters. A dataset with 1640 wells was collected from oil and gas fields. Q o was estimated as a function of gas-oil ratios (GOR), water-cut (WC), up-and down-stream (P wh & P sep ) pressures, and choke diameter (D 64 ). The wells were categorized to flow under critical (CF) and subcritical (subCF) conditions, where ANN was applied to build different models to estimate the production rates for each category. New correlations were established using the optimized weight and biases from the ANN models. In the case of subCF conditions, the oil rate was highly related to the pressure differential across the choke, while at CF conditions, P sep does not affect the oil rates. ANN models were constructed for CF conditions and another for subCF conditions. The average absolute percent error (AAPE) in the case of subCF was 1%, while in the case of CF, the AAPE was 1.13%. The developed models perform better than the published correlations, where the AAPE values for published correlations under comparison were generally greater than 25%. The findings from the study will highly help petroleum engineers to estimate the oil rates using readily available data from wellhead chokes with less additional costs or site involvement while maintaining quality.