CLOSED-CIRCUIT FRACTURING MODELS AND OPERATIONS

Claims

Some implementations include a method comprising: generating temporally variable engineering features from fracturing stage data; determining a plurality of machine learning models based on the engineering features; training the machine learning models by regression analysis with the engineering features as inputs and response product reactions as targets of the machine learning models; grouping some of the fracturing stages into clusters that have similar response product reactions for both the far field and near field; performing, using a physics-based fracturing model, fracturing modeling on each cluster to determine the uncertainty of the response product reactions; and determining improvements to the physics-based fracturing model from the clusters of fracturing stages.Generate a database of grouped temporal characteristics and remediation strategies; and generate, from the database, a digital twin to control fracturing operations in a well.