In elevator group control system, to choose the weight values of multi-objective optimization problem is always subjected to the designers’ understanding and personal preference, so sometimes the weight values can’t directly reflect the real change of elevator’s performance with the weight coefficients. If we put the Pareto solution set into the research of choosing the weight values, and use the random weight method, after one time, we can get several groups of different weight values, then seek the optimal Pareto solution according to the elevator dispatching strategies generated by different weight values. Because this algorithm can provide us several groups of different and optimal elevator dispatching strategies, at the same time, provide the researchers more useful data to refer. According to the advantages and disadvantages of each algorithm, we join gradient decent algorithm with the clonal selection algorithm, and speed up the convergence rate in later period, define random balanced traffic flow between layers as passengers flow, put hybrid algorithm and clonal selection algorithm into elevator group control system to find the saving energy policy, to some extent, hybrid algorithm has much more superiority than clonal selection algorithm.