Intelligent Multi-Model Optimization for Fracture Parameter Inversion in Multistage Fractured Horizontal Wells

Abstract An analytical-solution library that couples a finite-conductivity hydraulic fracture with either a homogeneous or dual-porosity matrix was constructed. History matching employs a hybrid inversion workflow: particle-swarm optimization (global search) is followed by a least-squares routine (local refinement), while a time-dependent, dual-objective loss function dynamically balances pressure and derivative misfits. Candidate solutions are ranked with the Akaike information criterion (AIC), and the best model is used to estimate fracture and matrix parameters for each stage. For Well N7, the automated procedure yields stage-average half-fracture lengths of 82.00–96.00 m, fracture-network volume ratios of 0.06–0.20, and inner-zone permeabilities of 1.1–1.7 mD. These single-stage results are subsequently integrated into a multi-stage, hydraulically fractured horizontal-well model to interpret the full pressure buildup test. Relative to the shut-in pressure-decline analysis of individual stages, the multi-stage interpretation indicates systematically lower inner-zone permeabilities and fracture-network volume ratios.