关键词: 随机建模, 地质统计学, 硬数据, 序贯高斯模拟, 碎屑岩油藏, 中国东部

Abstract:

When geostatistical tools were used to simulate reservoirs, the proportion of uncertain factors in the realization resulted from the classical Sequence Gaussian Simulation (SGS) algorithm correlated with both the data and the algorithm used. During reservoir simulations, a complete evaluation of realizations would not be obtained unless the data and the algorithm used are analyzed in details. The results of realizations may vary greatly, sometimes even showing a reverse trend. SGS has statistical significance, thus, the data processing will then exert an influence on the algorithm itself. The analysis and processing combined with algorithm and data may eliminate or reduce this influence and optimize the result through improving the SGS method on condition that hard data are fully used. Based on the distribution of hard data, this paper attempted to make the random quantity be less than the quantity of the estimated value during the course of reservoir simulations by means of conditionally controlling the proportion of the random quantity, consequently, no random quantity would be apt to shield the geological change trend of hard data.At the same time, the grid points that could search more hard data would be simulated first by measurably controlling the simulation sequence during the course of simulation,which could make full use of hard data and reduce the disturbance of random noise.The SGS approach presented in this paper can reflect more geology rules implied by the whole hard data than the classic SGS method previously used in reservoir simulations.

Key words: stochastic simulation, geostatistics, hard data, Sequence Gaussian Simulation (SGS), clastic reservoir, eastern China