In opening the discussion the following questions were submitted by Prof. G. L. CHIERICI (Italy):
Did the authors of Paper No. I apply their method to cases of water-drive gas reservoirs, where one is usually confronted with lack of information about the aquifier?
Did the authors investigate how errors in reservoir pressure measurements, as observed in practice, affect the range of uncertainty in backed-out reservoir parameters values?
Dr D. W. GREEN answered that water-drive gas reservoirs had not been studied but that the influence of errors was investigated.
On Prof. CHIERICI'S comment that the reliability of the predictions still rests on the authors "educated' feelings, Dr GREEN answered that in their work they had tried to eliminate a great deal of art and intuition by prediction of the behaviour of oil-reservoirs and improving the reliability of the prediction.
Professor CHIERICI'S questions to the authors of Paper No. 3 were:
Was the method used in the example case to solve the linearised system an iterative one? If so, what was the number of iteration cycles to cope with non-linearity of the original system? What computing time was required for each time-step both on the digital computer and on ADCS?
Is the simultaneous-solution technique combined with the evaluation of pressure- and saturation-dependent coefficients at the new time level allowed in the ADCS?
Dr GESHELIN explained that the relaxation method was used and the number of iteration cycles on the digital computer BESM-3M was between 5 and 12 for each time-step. The acceleration of solution on ADCS is essential because of frequent changes of well production rates. He agreed that the implicit methods were preferable and added that the method proposed and the ADCS allowed their application.
In connection with numerical methods applied in Papers 2, 3 and 4 Prof. G. W. THOMAS (U.S.A.) made a contribution on the use of Galerkin's method in reservoir simulation. By finite difference techniques only discrete solutions in time and space are obtained for the fluid flow equations. This is a disadvantage when one would like to predict precisely the behaviour of bottom-hole well pressures, locate flood fronts and treat those processes that involve rapid changes in pressure, saturation or concentration. Furthermore, error bounds on the computed results from sophisticated finite difference simulators are difficult if not impossible to achieve. He said they have enjoyed some success with an alternative approach based on a method proposed by the Soviet scientist B. G.
Galerkin. This technique is capable of providing continuous solutions in space much like an analytical solution. This is illustrated in F