【摘要】 为提高汛期降雨量中长期预报的精度,采用芒种日分析充分提取有用信息,基于BP神经网络模型,构建了芒种日分析的BP神经网络耦合模型,并将其应用于北京市中长期汛期降雨量的预测。结果表明:相比于常规BP模型,耦合BP模型能够有效提高预报的精度,验证期耦合BP模型模拟值与实测值相关系数为0.78,明显优于常规BP模型的0.42;耦合BP模型较常规BP模型的预报合格率提高了40%。芒种日分析能够充分发掘隐藏在原始数据中的有用信息,降低极端值等噪声数据对预报结果的影响,有效提高了模型的预报精度。将传统节气与人工智能预报技术相结合,为中长期汛期降雨量预报提供了一种新思路。更多还原

【Abstract】 In order to improve the accuracy of medium and long-term wet-season rainfall forecasting,a Hybrid models based on the Grain in Ear and BP neural network is established in this study,and applied to the forcasting of rainfall in mid-and long-term wet-seasons in Beijing.The results show that the hybrid model can effectively improve the accuracy of the rainfall forecasting,in comparison with the traditional BP model.The correlation coefficient between the simulated and the measured rainfall is 0.78,which is much better than the traditional BP model of 0.42.The hybrid model also has a 40%improvement in terms of the forecasting pass rate over the traditional BP model.The Grain in Ear can fully explore the useful information hidden in the original data,reduce interfere of noise data(e.g.,extreme values),and effectively improve the forecasting accuracy.This study combines the traditional 24-solar-terms with artificial intelligence forecasting technology and provides a new idea for medium and long-term wet-season rainfall forecasting.更多还原