Based on the gold cyanidation leaching process, a dynamic mechanism model is established according to material balance equations. The kinetic reaction rates are estimated by the Tikhonov regularization method, which reduces the propagation of measurement noise effectively. And then the estimated results are used to identify the unknown parameters. To reduce the influence of model uncertainty on real-time optimization (RTO) results, the constraint adaptation method is applied to RTO of leaching process. The simulation results show that this method can track the optimal set point of the actual process under model uncertainty. Moreover, the measurement gradients are unnecessary, which is less influenced by measurement noise and more available for practical applications and has laid an important foundation for the successful implementation of the optimization and control for hydrometallurgy process.