On the implementation of Prabhu’s exact solution of the stochastic reservoir equation

After 50 years of Prabhu’s paper on the exact solution of the stochastic reservoir equation for the important class of gamma inflow distributions with an integral shape parameter, a detailed implementation of the exact solution is still lacking, despite its potential usefulness from both theoretical and practical points of view. This paper explores some properties of Prabhu’s exact solution and investigates the numerical difficulties associated with its implementation. The solution is also extended to derive the distributions of deficit, spillage, yield, and actual release from the reservoir. Explicit analytical solutions for three relatively simple cases are given in detail as examples and comparisons with approximate numerical solutions are made, which reveal some shortcomings of approximate methods. The implementation of the solution in the general case reveals some numerical problems associated with large values of the shape parameter of the inflow distribution and large ratios of reservoir size to draft, mainly due to accumulation of round-off errors. A Matlab program has been developed to calculate emptying and filling probabilities over a wide range of reservoir parameters using extended precision. Comparison of Prabhu’s solution with the numerical solution of the reservoir integral equation highlights possible problems with the numerical solution, which may produce inaccurate or even invalid results for large reservoirs, large drift, and large skewness of the inflow distribution. A comparison between gamma and lognormal distributions as models of skew revealed that as the reservoir size, drift, and skewness increase, the probability of emptying of the reservoir becomes smaller for the case of gamma inflow than in the case of lognormal flow having the same skewness coefficient.