Abstract: The construction of geological models that accurately characterize the strong heterogeneity of different types of reservoir groups is the geological basis for efficient development deep paleokarst reservoirs. Currently, the geological model constructed by the paleokarst classification modeling method is mainly based on static data, which is not fully consistent with the actual production dynamics. There are problems such as the local inconsistencies between the model connectivity characteristics and the measured inter-well connectivity, the uncertainties in physical properties parameters of unfilled caves, and the great differences between model-calculated well control reserves and dynamic reserves of single wells. To improve the coincidence degree between the geological model and the production dynamic information of deep paleokarst reservoirs, a model optimization method and workflow based on annealing simulation method are constructed with the constraints of dynamic connectivity and dynamic reserves. In this paper, based on the connectivity characteristics of tracer and production dynamic discrimination, the mathematical model of fracture tracer conduction is established, and the objective function that characterizes the relationship between tracer conduction data and fracture attributes is constructed. The location of part fractures is optimized locally, so that the model connectivity is quantitatively consistent with the dynamic connectivity data. Taking the dynamic reserves of single well or well group as the target, the porosity, storage volume or co-optimization of the reservoir within its control range are carried out to improve the porosity accuracy of unfilled karst caves and reduce the difference between well controlled reserves and dynamic reserves of the model. The results show that the agreement between tracer simulation and measured curves after optimization of the geological model of a typical block reaches 86.3%, and the combined compliance rate between single well control reserves and dynamic reserves in the model is increased to 88.5%. The purpose of optimizing the model based on dynamic information is realized, the uncertainty of the model is reduced, and the effect is remarkable.

Key words: dynamic connectivity, dynamic reserves, annealing simulation, dynamic optimization, paleokarst reservoir, deep carbonate reservoir