论文部分内容阅读
为识别并去除棉花滩水电站大坝监测数据中的噪音,采用软硬折衷阈值小波去噪法对坝顶水平位移进行去噪处理,利用MATLAB中自带的支持向量机工具箱对去噪后的坝顶水平位移进行拟合、预测,并与传统多元回归方法的结果进行对比。结果表明,该方法在建模时段和预测时段的误差平方和分别为3.49、5.49,均优于传统多元回归方法,软硬折衷小波去噪方法与支持向量机结合具有较好的拟合、预测效果。
In order to identify and remove the noise in the dam monitoring data of Mianhuatan Hydropower Station, the software and hardware trade-off threshold wavelet denoising method is used to denoise the horizontal displacement at the dam crest. Using the support vector machine toolbox in MATLAB, The horizontal displacement at the top of the dam is fitted and predicted, and compared with the results of the traditional multiple regression method. The results show that the square error sum of this method is 3.49 and 5.49 respectively in the modeling period and the forecasting period, which are better than the traditional multivariate regression method. The hardware and software eclectic wavelet denoising method and the support vector machine have good fitting and prediction effect.