小波分解与变换法预测地下水位动态

通过小波分解方法将地下水位动态的非平稳时间序列分解为多个细节信号序列和逼近信号序列，然后运用时间序列自回归模型及人工神经元网络模型对各信号序列分别进行模拟预测，模拟结果比单纯用自回归法或人工神经网络模型更接近实测值，说明通过小波分解方法进行地下水位动态模拟和预测是适合的；同时用小波变换方法对地下水位动态进行了宏观分析，使隐藏的规律性显现出来，揭示出地下水位动态变化中除了具有一个水文年内的周期性变化规律外，还存在2～3年间隔的波幅强弱变化，可以推断未来短期内地下水位动态发展仍将延续当前总体下降的趋势，与小波分解方法得到的预 测结果相吻合。

Application of wavelet decomposition and wavelet transform method to forecasting of groundwater regime

The non-stationary time series of groundwater regime in we stern suburban of Beijing is decomposed into several detailed stationary time series and a smoothed non-stationary time series according to the principle of wavelet decomposition. The stationary time series is simulated by using auto-regressive method and the non-stationary series is simulated by using artificial neural network model. The comparison shows that the error of the simulation adopting this method is smaller than that by using auto-regressive method and artificial neural network model directly. Furthermore, the wavelet transform method is applied to the macro analysis of groundwater regime. It is revealed that several cycles of groundwater table exist, which are hidden in the long-term groundwater regime. Consequently, it may infer that the groundwater table will maintain the downward tendency in the near future. This conclusion is in accordance with the forecasting by using wavelet decomposition method.