For low order process with large time delay, a kind of optimal PI controller tuning method is proposed based on generalized Hermite-Biehler theorem and Genetic Algorithm. Firstly, the stable region of PI controller is obtained by using the generalized Hermite-Biehler theorem. Then the optimum parameters are selected from this region based on ITAE criterion and genetic algorithm. A tuning formula is obtained by nonlinear fitting of optimization result, which has the capability to cover the variety of normalized time delays up to 100. Simulation of Monte-Carlo stochastic experiment indicates that the proposed method has good performance robustness when parameter uncertainty occurs, compared with other four PI tuning methods.