Fractured tight sandstone reservoirs are promising targets for underground energy storage, but their heterogeneous nature and often-incomplete historical test data pose significant challenges for accurate deliverability prediction and reservoir evaluation. To address this, a novel hybrid methodology is proposed. For wells with complete historical data, deliverability is calculated using a binomial inflow performance relationship (IPR) model. For wells with incomplete data, a weighted fusion model integrating a Random Forest algorithm and least squares regression is developed to predict natural blowout capacity, a key proxy for energy storage injectivity/productivity. The fusion model achieved superior performance with a mean absolute error (MAE) of 7.19 × 104 m3/day and a Mean Relative Error (MRE) of 8.5%, outperforming standalone methods. Based on the predicted deliverability, reservoirs in the Bozi–North block (Kuche Depression, Tarim Basin) were classified into three potential grades (I, II, III). The study provides a data-adaptive framework for deliverability prediction and offers tailored reformation process recommendations (e.g., sand fracturing for Grade I reservoirs), thereby providing a more reliable and practical decision support tool for the efficient development of tight sandstone energy storage reservoirs.
Ren et al. (2026) studied this question.