As the second most widely used artificial lift method in petroleum production (and first in produced amount), electrical submersible pumps (ESPs) maintain or increase flow rates by converting kinetic energy to hydraulic pressure of hydrocarbon fluids. ESPs are generally characterized under water condition and the water performance curve is provided by manufacturers. However, properties of hydrocarbon fluids are very different from water and significantly alter the pump performance. Most of existing methods to estimate ESP pump boosting pressure under viscous fluid flow involve strong empirical nature by correlating the experimental data with correction factors. A universally valid mechanistic model is not available yet. In this paper, a new mechanistic model accounting for viscosity effects of working fluids on ESP pump hydraulic heads is proposed, which is validated with a wide range database collected from different pump types.
The new model starts from Euler equations for centrifugal pump, and introduces a conceptual best-match flowrate QBM, at which the outlet flow direction of impeller matches the designed flow direction. The mismatch of velocity triangles, resulted from various liquid flow rates, is used to derive recirculation losses. Other head losses due to flow direction change, friction, and leakage flow etc. are incorporated in the new model. QBM is obtained by matching the predicted performance curve with the catalog curve with water. With the determined QBM, the ESP hydraulic head under viscous fluid flow conditions can be calculated. The model predictions of the boosting pressure among various ESPs are compared with an experimental database, which is composed of more than 170,000 data points.
The specific speeds (NS) of ESP pumps in this study are ranged from 1600 to 3448, among which one ESP pump is radial type, and the others are mixed types. As baseline, the model-predicted ESP pump water performance well matches to the catalog curves. With viscous fluid medium, the boosting pressures predicted by the model agree well with the experimental data. For most calculation results with medium to high flow rates, the overall prediction error is less than 15%. Unlike the empirical correction factors that use experimental data points as inputs, the mechanistic model proposed in this study does not require any experimental data input, it could predict ESP pump performance with an acceptable accuracy.