Estimation of gas/liquid and oil/water interface levels in an oil/water/gas separator based on pressure measurements and regression modelling

Gravity separators are widely used for separation of gas/oil/water/sand from both offshore and onshore oil production facilities. Estimation of the gas/liquid and oil/water interface levels in gravity separators have been a concern since these parameters are important for reliable operation. Most of the instruments on the market today do not provide reliable measurements of both gas/liquid and oil/water interface levels. The few instruments that do provide reliable measurements are however based on radioactive principles. Nevertheless these radioactive instruments possess a strong health, safety and environmental risk. An alternative inexpensive, environmentally friendly, accurate and cost effective way for gas/liquid and oil/water interface level estimation based on pressure measurement is presented. The root mean squared error of prediction (RMSEP) for gas/liquid and oil/water interface level estimation from traversing a pressure sensor based on partial least square regression (PLS-R) were 14.5 mm and 17.7 mm respectively. A comparison of results from models based on PLS-R and ordinary least square regression (OLS-R) techniques proved that the RMSEP from the PLS-R technique was better in estimating the oil/water interface level but in the case of gas/liquid interface level estimation the OLS-R technique was slightly better. It was concluded that the PLS-R technique provided a better overall result and is recommended when pressure measurements in combination with multivariate data analysis is applied for gas/liquid and oil/water interface level estimations in gravity separators.

► Two independent experiments were conducted in a pilot scale gravity separator.
► A single pressure sensor was traversed from the top to the bottom of the separator.
► The operation of the gravity separator was simulated in a MATLAB application.
► Pressure measurements coupled with regression modelling was adapted in this study.
► The gas/liquid and oil/water interface levels were satisfactorily estimated.