位移时序预测的APSO-WLSSVM模型及应用研究

Study and application of displacement time series forecast based on APSO-WLSSVM

引入改进的粒子群算法对小波核函数最小二乘支持向量机进行优化,提出了位移时间序列预测的改进粒子群优化小波最小二乘支持向量机预测模型(APSO-WLSSVM)。该模型具有小波变换的良好时、频域分辨能力和支持向量机的非线性学习能力;同时利用粒子群算法优化小波最小二乘支持向量机的参数,避免了人为选择参数的盲目性,从而提高了模型的预测精度。为证明该模型的优越性,将该模型与传统的高斯核函数支持向量机模型的预测结果作了对比,结果表明该模型较传统方法预测精度有了明显提高。最后将该模型用于锦屏一级水电站左岸边坡和导流洞进行变形预测,预测结果表明该方法科学可靠,在岩土体位移时序预测中具有良好的实际应用价值。

The model of wavelet least squares support vector machines （WLSSVM） is optimized by adaptive particle swarm optimization （APSO）, and a new model named as wavelet least squares support vector machine based on adaptive particle swarm optimization （APSO-WLSSVM） is put forward. The model combines good time domain, frequency domain resolving ability of wavelet transformation and nonlinear learning ability of SVM. The adaptive particle swarm optimization is used to optimize the parameters of SVM so as to avoid artificial arbitrariness and enhance the forecast accuracy. For comparison, the model of APSO-WLSSVM and the traditional SVM （Gauss kernel function） are used to forecast the same displacement time series. The result shows that the former is better than the latter in forecast accuracy. The model is used to forecast the left back slope and diversion tunnel of Jinping First-stage Hydropower Station. The forecast values are in good agreement with the measured ones, indicating that the APSO-WLSSVM is feasible and precise and can be well applied to the forecast of displacement time series.