基于支持向量机的钱塘江潮时预报方法

Prediction on the Occurrence-time of Qiantang River's Tidal Bores Based on SVM

针对在钱塘江潮时预报中经验模型和传统神经网络的可靠性不足的问题,提出一种基于支持向量机的钱塘江涌潮到达时间预报方法。通过历史数据了解并分析钱塘江涌潮的周期性以及各涌潮周期间的相关性,取预报日期前后一个月的数据作为一个预报模型,以预测日期前一个月以及近5年内同一月份的隔日时间差数据作为训练样本,利用支持向量机预测未来涌潮到达时间。最后,通过对钱塘江沿岸多个水文站2015年农历八月初一至八月二十一的隔日时间差实例预测,验证了方法的有效性。

Due to the lack of reliability of empirical model and traditional neural network model commonly used in the prediction of occurrence-time of Qiantang River＇s tidal bores, a new method based on Support Vector Machine is introduced to forecast the tidal bore occurrence-time. Based on historical data, the periodicity of tidal bores and the correlation between each tidal cycle are firstly analyzed, and then according to forecast date, the data at one month interval are selected to build the forecasting model, and the every-other day difference data of one-month before the forecast date and the corresponding same month of recent five years are used as training sample. Finally, the occurrence-time of tidal bore is predicted by Support Vector Machine. An experimental example on the tidal bore occurrence-time prediction at four tide observation stations on Qiantang River is presented to demonstrate the effectiveness of proposed method.