When a process engineer is asked to design transportation and treatment facilities or to recommend operating procedures, he needs to run steady state and transient multiphase flow computer programs. The predictions regarding steady state two-phase flows of the mechanistic models OLGA, TACITE and TUFFP have been assessed in a previous SPE paper (ref. 1). As a logical continuation, the performances of these models regarding transient two-phase flows are compared in this paper first on transient loop data and then on field data. The main features of each model are presented as well. From the analysis of the features and the performances of the three codes, some guidelines to improve each model are indicated.
Formerly, people used to run steady state empirical correlations for the design of multiphase production system including well, pipeline and downstream facilities. But it appeared very quickly that such correlations could be used only in the case of wells or pipelines similar to the ones the correlations were derived from (same diameter, same profile and same fluids).
Since the early 1980's, with the coming of a number of mechanistic models onto the market (PEPITE, WELLSIM, OLGA, TUFFP), people know that it is less risky to use steady state mechanistic models to design the multiphase production systems. Given the inlet and the outlet pressures as well as the gas and liquid flow rates, these steady state models are reliable enough to figure out the right diameter. But when it comes to designing facilities downstream of a pipeline in which flow rates can change fast, these models often prove inadequate.
They are incapable of even predicting the flow rates at the pipeline outlet when the inlet flow rates are low enough to give rise to such well-known phenomena as terrain slugging or severe slugging. The only way to deal with these highly varying outlet flow rates and consequently to design properly the facilities downstream of the pipeline is to use transient flow simulation models. Among such models we have tested OLGA, TUFFP and TACITE.
The three programs were compared first against loop data and next against field data. The advantage of data measured on a loop is that the fluid properties and the flow rates are known precisely. If discrepancies exist between the measures and the predictions, only the hydrodynamic model can be called into question. For the code developers, this is a first step needed to establish confidence before pursuing with the comparison against field data. Such a comparison is required by any potential user of a model to get him to trust the program in his turn.
Data from the loop set-up within the frame of the Tulsa University Fluid Flow Project were used for the comparison. This loop will be briefly described here-after.
Specific measurement campaigns were conducted on a number of real pipelines operated by Elf Exploration Production and Total to acquire relevant transient flow data necessary to test any transient flow simulation program. Among them, we have selected one pipeline with very severe changes in the inlet flow rates.
As expected, none of the three programs can predict perfectly all the transient flow conditions. From the analysis of the main features of the models on the one hand and from the analysis of the conditions where the discrepancies are the largest on the other hand, the authors will try to make clear some areas where improvements are needed.
Loop data. The data have been acquired on the test loop set up in the frame of the Tulsa University Fluid Flow Project (TUFFP). The test facility (fig. 1) is described by Minami (ref. 2). It is a 420 m long, 77.9 mm diameter horizontal steel pipeline that uses air and kerosene as working fluids. It is operated at atmospheric pressure.
P. 631^
