Abstract:

Tahe Oilfield, known for its substantial crude oil reserves, features fracture-vuggy carbonate reservoirs with diverse and heterogeneous characteristics shaped by structural and karstic influences. Each reservoir type within this field exhibits distinct development traits, making the precise identification of these reservoir types crucial for devising effective production strategies and optimizing oil reservoir development. However, the identification of reservoirs through drilling and geophysical data is challenging and costly, hence, this paper focuses on the dynamic identification of the vuggy, fractured-vuggy, and fractured reservoirs in the buried hill carbonate reservoirs in the Yakela block of Tahe Oilfield. The research initially involved analyzing the dynamic data of the production wells in this area and dividing the development stages of each well. Subsequently, the discriminant indicators, such as the initial oil production in the elastic stage, elastic time, cumulative oil production, and production decline rate were extracted. These indicators are generally available in each well and have less human interference. They form the basis of a dynamic quantitative characterization method for determiningreservoir types Through the utilization of mathematical statistics and artificial neural network technology, an automatic identification system for carbonate reservoir types based on dynamic data was established. Remarkably, the results obtained from this method align with over 80 % of the reservoir types determined through drilling logging and geophysical data. This automated identification method proves to be highly operable and complements geological data effectively, enabling more precise reservoir determination, especially in areas where geological information is scarce. Its applicability extends to carbonate reservoir research in regions with limited data, offering reliable reservoir-type results that are essential for informed development planning.

Key words: Tahe Oilfield, Yakela buried hill, dynamic identification of reservoir type, production stage division, neural network, fractured-vuggy reservoir