Prediction of sump oil growth rate towards gravitational settling of produced water in oilfield based on machine learning

Abstract

Gravitational settling is crucial for treating produced water, and its efficiency affects the separation in multi-stage treatment processes. Due to complex water properties and thickening upper oil layer during the settling process, sump oil recovery in appropriate time is essential to avoid secondary pollution of separated water and recover oil resources. This paper constructs and analyzes the dataset of sump oil recovery for produced water separation. It is determined that the factors affecting the sump oil growth are oil content, daily processing capacity, polymer concentration, and suspended solids content. The produced water of water-flooding, polymer-flooding, and ASP-flooding, with average oil contents of 0.03893, 0.20072, and 0.05387, respectively, is carried out to train the prediction model for the sump oil growth rate. It is indicated that for actual settling tanks with cross-sectional areas ranging from 100 m2 to 900 m2 and daily processing capacity range from 5000 m3 to 30000 m3, the sump oil growth rate for produced water of water-flooding, polymer-flooding, and ASP-flooding are distributed in the ranges of 0.1094 m/day to 2.5206 m/day, 0.0309 m/day to 0.7524 m/day, and 0.0505 m/day to 1.8764 m/day, respectively. The predicted results are consistent with actual operation.