【摘要】 针对大坝变形影响因素的复杂性以及监测数据的非线性、随机波动大和预测难度大等问题,提出一种改进自适应粒子群(particle swarm,PSO)算法的混合核函数最小二乘支持向量机(least squares support vector machine,LSSVM)模型,实现了大坝水平变形的时间序列预测方法。基于Mercer理论,将多项式核函数和高斯核函数进行线性组合,构建混合核函数,作为LSSVM模型的核函数,并以特征因子与大坝变形间的相互联系为基础,采用动态自适应惯性权重的PSO算法,对混合核函数的LSSVM模型进行参数寻优,以确保建立最佳LSSVM预测模型。将模型应用于丰满大坝,并与传统多项式核函数和传统高斯核函数的LSSVM模型进行对比仿真实验,对所提方法的有效性和准确性进行验证评估。结果表明,该模型在预测精度上有了明显提高,预测性能尤佳。可见改进自适应粒子群的混合核函数LSSVM模型对大坝变形的时间序列预测有良好的实用价值。更多还原

【Abstract】 Since the influencing factors are complex, the monitoring data are nonlinear, and the random fluctuations are large, it is very difficult to forecast the dam deformation. A hybrid kernel function least squares support vector machine(LSSVM) model with improved adaptive particle swarm optimization(PSO) algorithm was proposed to realize a time series forecast for horizontal deformation of dams. Based on the Mercer theory, the polynomial kernel function and the Gaussian kernel function were linearly combined to construct a hybrid kernel function as the kernel function of the least squares support vector machine model. Based on the correlation between the characteristic factor and the dam deformation, dynamics were adopted. The PSO algorithm with adaptive inertia weights optimized the parameters of the LSSVM model of the hybrid kernel function to ensure the optimal LSSVM forecast model. The model was applied to the plump dam, and simulation experiment with the LSSVM model of the traditional polynomial kernel function and the traditional Gaussian kernel function was carried out to verify and evaluate the validity and accuracy of the proposed method. The results show that the model significantly improves the accuracy of forecast, and its performance is excellent. It can be concluded that the improved kernel function LSSVM model with improved adaptive particle swarm is practically valued for time series forecast for horizontal deformation of dams.更多还原