In order to find an algorithm which can solve the Stochastic Chance-Constrained Programming (SCCP) problem more effectively, a hybrid intelligence algorithm based on Clonal Selection Algorithm (CSA), random simulation technology and neural network was proposed. Random simulation was used to produce random variables sample matrix for training Back Propagation (BP) neural network to approximate the stochastic function. Fitness value was calculated and feasible solution was checked by the trained neural network in CSA until it could get the solution to the optimization problems. In order to make the searching rapid and effective, double cloning operators and double mutation operators were adopted in CSA. The simulation results show that satisfactory result has been achieved before 500 generation; moreover, the precision in the single objective optimization problem is improved by 2.2% and the precision in multi-objective optimization problems is increased by 65% compared with other existing algorithms. In addition, the algorithm was applied to solve the problem of optimal reservoir scheduling. The simulation results also show the correctness and effectiveness of the model and the algorithm.