Analysis of Experimental Data of ESP Performance Under Two-Phase Flow Conditions

Abstract

Abstract Electrical Submersible Pumps (ESP's) are well known to have a good predictable performance for single phase, low viscosity liquids. In the oil industry, oil production is associated with the production of gas. The presence of gas deteriorates the performance of the pump. The performance deterioration varies with the amount of gas and the intake pressure at which the pump is operated. So far no good predictive method is available to predict the performance of the ESP under two-phase conditions. Presently the industry is using the homogeneous model, correlations and other models to predict this performance. In the homogeneous model the two-phase flow head is assumed to be the same as single-phase head (provided by the manufacturer) at a total mixture flow rate. The available correlations are specific to the tested type of pump and to the tested number of stages. Finally to predict performance on model, no general model has been developed to predict two-phase performance of the ESP due to complexity of two-phase flow, pump geometry and speed at which it is operated. Researchers, Lea and Bearden (1982), Cirilo (1998), Romero (1999) and Pessoa (2000) have concluded through experimental results that the application of the homogeneous model gives good prediction only at low gas fractions (in the order of 2 to 5 %) at the intake. At higher gas fraction the experimental results shows that performance is well away from the homogeneous model predicted performance. The University of Tulsa Artificial Lift Projects - TUALP is currently conducting experimental and theoretical research on the two-phase behavior of electrical submersible centrifugal pumps, using a 22-stage Mixed flow type, series 513 pump with best efficiency flow rate of 6100 BPD to gather data on stage wise performance under two-phase flow conditions. Air and water were used as working fluids. This is a unique facility that has pressure gauges fitted across each stage. It not only helps to study the stage wise behavior but also the effect of number of stages on cumulative performance of pump under two-phase flow conditions. This paper focuses on analysis of data collected at TUALP facility. Introduction Centrifugal pumps are dynamic single or multistage devices that use kinetic energy to increase liquid pressure. To handle low viscosity, single-phase incompressible fluids, existing impeller and diffuser designs are very successful, but are severely impacted by free gas, highly compressible or viscous fluids. The relationship between the head developed by the pump and the flow rate through the pump for a certain rotational speed is determined by a specific single-phase performance curve, which is experimentally determined using water as the working fluid. The head performance curves are valid then for any other low viscosity single-phase liquid, independent of its density. Brake horsepower and efficiency curves are usually presented on the same chart. The performance of multistage pumps handling incompressible fluids is presented on average per stage. An example of these performance curves is shown in Figure 1. Two vertical lines define the lower and upper limit where it is recommended to operate the equipment. The best efficiency point is the point of maximum efficiency within the operating range, and is usually abbreviated as BEP. For low viscosity oil with no free gas or very low volumetric free gas fractions (< 2%) at pump intake condition, the sizing of a multi-stage ESP has shown good agreement based on the water performance curves supplied by the manufacturer, corrected by the homogeneous model approach. While handling higher contents of free gas, the centrifugal pump suffers head degradation. Performance prediction based on single-phase water performance curves corrected by the homogeneous model cannot be used. In addition to performance degradation while handling free gas, submersible pumps also require prediction of surging and gas lock conditions. The homogeneous model is incapable of correctly addressing those problems.