Assessment and Development of Heavy Oil Viscosity Correlations

Abstract

Abstract Heavy oil always poses a great challenge to production and transportationsystems due to its high viscosity. This paper evaluates fourteen dead, eight saturated and nine under saturated live oil viscosity correlationsagainst a databank consisting of heavy oil data from various parts of the worldwith wide ranges of temperature, pressure and fluid compositions. Theexisting empirical viscosity correlations were mostly developed forsignificantly lighter oils. Most of them cannot reasonably predict theheavy oil viscosity at low temperatures. Three new empirical correlationsfor dead, saturated and under saturated oils are developed that are applicablefor the heavy oils with API gravity ranging from 10 to 22.3. When comparedwith the databanks, the new dead oil, saturated and under saturated oilviscosity correlations showed 3 to 50%, 3 to 13% and 22 to 27% improvement overthe existing best correlations, respectively. Saturated, Aromatics, Resinand Asphaltenes (SARA) data of some heavy oils were also analyzed to understandthe role of the asphaltene and resins in heavy oil viscosity. Introductory Remarks There have been a number of empirical correlations developed for medium andlight crude oils. However their applicability is limited to specific oilsdue to the complex formulation of the crude oils. Moreover, theirapplicability to heavy oils is very much in question. De Ghetto et al.[1]first defined the heavy oil in terms of API gravity. Later, he divided theheavy oil into two groups: heavy oil (10<oAPI<22.3) and extra heavy oil(oAPI<10). His definitions are adopted in this paper. The correlations are generally presented as functions of the measuredparameters such as API gravity, temperature, pressure and solution gas ratio(Rs). Lohrenz and Bray[2] also used the crude oil chemical composition todevelop an empirical correlation for oil viscosity. Egbogah and Ng[3] usedthe pour point as an additional input parameter for dead oil viscositycorrelations. Argillier et al.[4] reported that asphaltene and resin havea great effect on heavy oil viscosity. In this study, the existing correlations are thoroughly evaluated using alarge databank, and a new set of correlations for heavy oils based on simplelaboratory measured parameters at low temperatures are proposed. Literature Review Literature search and review have identified a number of correlations forpredicting crude oil viscosity. These correlations can be divided intothree categories: dead oil viscosity (µod), saturated or bubble point viscosity(µob) and under saturated oil viscosity (µo). The correlating variables andtheir ranges used by the authors to develop their correlations are listed inTables A-1 - A-3 in Appendix A. In 1946, Beal5 presented dead oil viscosity correlations as a function ofAPI gravity and temperature in the range of 100 - 229 oF. A total of 655data points were collected from 492 oil fields in the United States attemperature 100 oF.98 data points were collected above 100 oF. Bealalso developed a correlation for estimating the viscosity of under saturatedoil based on 52 data points from 26 crude samples. Beal's viscosity correlationcompares with the original data with an average error of 24.2%. In 1959, Chew and Connally[6] proposed a correlation to predict thesaturated oil or bubble point viscosity based on 457 data points collected fromCanada, the United States and South America. Later on the Chew and Connallyequation was simplified by Aziz[7]. In 1975, Beggs and Robinson[8] published two correlations for dead andsaturated oils. The dead oil correlation was developed using 460 datapoints from 93 different oil samples. An average error of -0.64% withstandard deviation of 13.53% was reported when compared against the data usedfor its development. Saturated oil viscosity correlation was developedfrom 2073 live oil data, and an average error of -1.83% with a 27.25% standarddeviation was reported. In 1978, Vazquez and Beggs[9] presented a correlation based on 3593 datapoints for under saturated oil viscosity by using most commonly required PVTparameters for crude oils. An average percentage error of -7.54% and astandard deviation of 8.7% were reported when the correlation was testedagainst the data from which it was developed.