Prediction of fold-of-increase in productivity index post limited entry fracturing using artificial neural network

Unified Fracture Design (UFD) bridges the gap between practices and theory in the hydraulic fracturing industry. It represents a technique to design hydraulic fracturing treatment with a particular amount of proppant. This design could provide the maximum fold-of-increase (FOI) in productivity-index (PI) after hydraulic fracturing treatment. The UFD optimization tool is very effective, but it has assumptions like any other model. One assumption of UFD optimization technique; is a single-layer assumption. This assumption does not align with the limited entry fracturing design concept. In limited entry fracturing, the frictional pressure is employed to offset the stress differences between multi-layers reservoirs to attain fluid injection through these layers, intended to deliver an optimal fracture conductivity in all layers. The drawback of this assumption is the underestimation of the actual value of FOI in PI. This paper aims to recast the original unified fracture design approach to extend the optimal UFD to a multilayer reservoir to predict the FOI in PI after limited entry fracturing treatment. The recasting tool for this problem to find the optimum solution is Artificial Neural Networks (ANN). The architected ANN model is based on actual historical data of limited entry fracturing treatments. A statistical comparison between the proposed ANN model and classical UFD technique demonstrates that ANN model solution has a more reliable estimation of FOI in PI with the actual historical data.