基于ARIMA误差修正的混合预测方法

Hybrid Prediction Method Based on ARIMA Error Correction

由于水文时间序列中存在着大量的不确定性、灰色性、线性性以及非线性性等问题,传统单一的预测模型并不能完全提取水文时间序列中的信息,因此,为了提高径流量模拟的准确性,获得精度更高的径流量模拟数据。提出了一种基于求和自回归移动平均(ARIMA)修正反向传播(BP)神经网络的混合模型。实验结果发现,该混合模型可以提取单一BP神经网络未提取到的水文信息,并获得更高的模拟精度。

The traditional single prediction model cannot extract the complete information in the hydrological time series,for the reason that there are a lot of uncertainties, grey, linear and non-linear problems. Therefore, a hybrid model based on autoregressive moving average model (ARIMA) modified back propagation (BP) neural network is proposed, in order to improve the accuracy of runoff simulation and obtain more accurate runoff simulation data. The results show that the proposed hybrid model can extract the hydrological information which can not be extracted by a single BP neural network, and obtain higher simulation accuracy.