Prediction of conductivity of propped fractures in shale reservoir based on neural network

The propped fracture inflow capacity in unconventional reservoirs' fracture networks is vital for assessing fracturing effects. It's influenced by factors like fracture width (sand spreading concentration), proppant type, and closure pressure. Based on extensive experimental data, a BP neural network prediction model based on genetic algorithm was established to realize its efficient and accurate prediction. Experiments were conducted using an API-standard flow-conducting chamber to test the flow-conducting ability of ceramic and quartz sand proppants under different sand placement concentrations, based on unconventional reservoir slickwater sand-carrying proppant placement experiments. Through model learning and training, the optimal values of parameters such as learning rate and the number of hidden layer nodes were determined, and then flow capacity prediction was carried out. The experimental results show that the quartz sand proppant under high closure pressure conditions has an abrupt change in the change trend of the flow-conducting capacity of the propped cracks due to crushing and other reasons. And the prediction model can achieve good prediction results for both ceramic and quartz sand. Comparing the predicted test set value and the actual inflow capacity value, its MAE is 0.582 D-cm and R2 is 0.993.