Level estimation in oil/water separators based on multiple pressure sensors and multivariate calibration

Abstract

Separation of oil, water and gas is an important process stage in oil and gas production. Such mixed fluids with different densities are often separated using a gravity separator. An unwanted emulsion will develop in the layer between oil and water and should not be a part of the oil output flow from the separator. The level and thickness of the emulsion layer together with oil and water content is therefore one of the important properties when controlling the oil output flow rate. The water output flow can be used to adjust the position of the interface/emulsion layer which should be below the oil output. Most of the level estimators are based on radioactive level measurements where the radiation is influenced by the density of the liquids. The radioactive principle is however expensive, and the resolution depends on the number of receivers and the robustness of the software. The main disadvantage is that process operators are exposed to radiation. This paper shows how a set of inexpensive pressure sensors and radar is used together with multivariate calibration for the estimation of the levels and the positions of the different layers in a vertical oil/water/gas separator. The method is calibrated and tested on a pilot scale vertical separator based on the same design as full-scale separators used in the oil/gas industry. The signals from the radar and pressure sensors are calibrated against known reference levels using partial least squares regression (PLSR). Models based on both competitive and complementary sensor inputs are evaluated. Copyright © 2010 John Wiley & Sons, Ltd.