劉嬌 史國友 朱凱歌 張加偉 李爽 陳作桓 王偉



摘要:為提高潮汐預測精度,解決單一調和分析預測精度不高的問題,提出一種基于調和分析和自回歸綜合移動平均-支持向量回歸機(autoregressive integrated moving average support vector machine for regression,ARIMA-SVR)的組合潮汐預測模型。潮汐分析中,潮汐可認為是由受引潮力影響的天文潮位和受環境因素影響的非線性水位的疊加。采用小波分析對潮汐樣本數據進行去噪處理,使用調和分析法計算天文潮位,以調和分析法計算產生的殘差作為非線性水位樣本數據,并使用ARIMA-SVR模型進行潮高計算,最后將兩部分的計算結果進行線性求和得到最終的潮汐預測值。利用美國舊金山港口實測潮汐數據進行預測仿真,結果表明,該組合模型解決了調和分析忽略非線性影響的問題,提高了潮汐預測準確率,可行且高效。
關鍵詞:潮汐預測; 組合模型; 調和分析法; 支持向量回歸機(SVR); 自回歸綜合移動平均(ARIMA)模型
中圖分類號: U675.12
文獻標志碼: A
Abstract:To improve the accuracy of tide prediction and solve the problem of low accuracy of single harmonic analysis, a combined tide prediction model based on the harmonic analysis and the autoregressive integrated moving average-support vector machine for regression (ARIMA-SVR) is proposed. In tide analysis, tide can be considered as the superposition of astronomical tide level affected by tide-generating force and non-linear water level affected by environmental factors. The wavelet analysis is used to denoise the tide sample data. The harmonic analysis method is used to calculate the astronomical tide level. The residual sequence generated by the harmonic analysis method is used as the sample data of non-linear water level, and ARIMA-SVR model is used to calculate the tide height. The tide prediction value is obtained by linear summation of the calculated results of the two parts. The simulation of prediction is carried out using measured tide data of San Francisco Port of the United States. The results show that: the combined model solves the problem of ignoring nonlinear effects in the traditional harmonic analysis, and the accuracy of tide prediction is improved; the combined model is feasible and efficient.
0 引 言
潮汐是海平面周期性的升降運動,它的漲落與人們的生產生活有著密切的聯系。實時精準的潮位信息在船舶航行安全、海洋資源開發利用、海洋災害的減輕與預防中發揮著舉足輕重的作用。因此,研究一種簡單、高效的潮汐預測方法成為亟待解決的問題?,F有的潮汐預測方法主要分為傳統預測模型和智能化預測模型。
傳統潮汐預測模型主要是用調和分析法進行預測的。Thomson于1866年首次將調和分析法應用于潮汐預測中,Darwin對該方法做了進一步完善,創立了典型的平衡潮理論?!?br>