论文部分内容阅读
改变应用最小二乘法求解大坝统计预警模型的传统方式,利用粒子群算法随机搜索的优化能力确定统计模型的回归系数。针对粒子群算法收敛速度较慢等问题,提出一种新的自适应策略,能够依据粒子个体和种群的优化信息,调整学习因子,并将该策略与遗传算法的交叉、变异算子相结合。通过工程算例表明,该方法具备较好的搜索多样解能力,自适应地调整粒子飞行的步长,提高了粒子群算法的收敛速度;基于该方法的大坝预警评价模型与最小二乘法、基本粒子群算法相比,数据挖掘能力强,预警评价结果与大坝的实际运行状态更加吻合,有效地提高了统计模型的预测精度。
The traditional method of applying least square method to solve the dam early-warning model is changed, and the regression coefficient of the statistical model is determined by using the optimization ability of the PSO algorithm. Aiming at the problem of slow convergence speed of particle swarm optimization, a new adaptive strategy is proposed, which can adjust the learning factor according to the optimization information of particle individuals and populations, and combines the strategy with the crossover and mutation operators of genetic algorithm. The engineering example shows that this method has better searching ability of multiple solutions and adaptively adjusts the step of particle flight and improves the convergence speed of particle swarm optimization algorithm. Based on the method of early warning evaluation model and least square method, Compared with the basic particle swarm optimization algorithm, the data mining ability is strong and the results of the early warning evaluation are in good agreement with the dam actual operating conditions, which effectively improves the prediction accuracy of the statistical model.