Abstract: The convergence-confinement method is mostly used in the preliminary design stage of tunnels, but the formulas for the deformation of the surrounding rock in this method are implicit nonlinear functions, which makes it difficult to solve them by using traditional reliability calculation methods.In view of this, the stability reliability calculation of the surrounding rock of hydraulic tunnels was transformed into a constrained optimization problem based on the geometric meaning of the reliability index.Then, the Gaussian process regression was used to fit the real performance function, and the Genghis Khan Shark Optimizer was used to solve the optimization problem.As a result, a new method for calculating the stability reliability of the surrounding rock of hydraulic tunnels, based on the Gaussian process regression and the Genghis Khan Shark Optimizer, was proposed.The analysis of a calculation example showed that the proposed method has strong applicability, high computational efficiency, and accuracy.In the calculation, the reliability index and failure probability could be obtained by a few times of operating the numerical model, and at the same time, the design points were given, which could be used to further analyze the sensitivity of performance function to each factor.The example of a practical hydraulic tunnel showed that the probability of producing moderate or higher degree of extrusion deformation at the tunnel crown was 83.81% and the probability of producing rheological deformation was 14.46%.Compared with the deterministic analysis, the reliability analysis can give probabilistic significance to the deterministic analysis.With regard to the rheological deformation, the result of the reliability analysis was more conservative than that of the deterministic analysis.