Abstract：BP neural network and genetic algorithm are used in identification of in-situ stresses for the problems that are stress measuring points few,measurement discreteness and calculation nonlinear. BP neural network is used to map the nonlinear relationship between borehole pressures and stress parameters,and is trained by using samples from practical engineering to ensure accuracy of predicted results. At last,applying genetic algorithm,a Wanfu engineering example is conducted to obtain the in-situ stresses. By comparing theoretical calculations with identified results,it validates that the proposed method is feasible,reliable and convenient in determination of in-situ stresses,and can used effectively in similar engineering.

Keyword：hydraulic fracturing;in-situ stress measurement;neural network;genetic algorithm;parameter identification;pressure-based back analysis;