论文部分内容阅读
基于最大熵谱原理和云模型理论,对巢湖流域11个市县1955-2005年汛期的历史降水数据进行分割,并逐一建立对应的降水历史云和趋势云,将二者根据权重叠加形成预测云;通过预测云的正向发生器产生并加权平均得到对应的降水预测值,与神经网络预测的降水值以及2006年实际降水值进行了比较,预测结果明显好于神经网络,更加接近于实际降水值。研究表明该方法对于挖掘短序列时间降水特征,模拟降水随机性与模糊性具有一定的优势,较好地预测了研究区域的降水。
Based on the principle of maximum entropy spectrum and cloud model, the historical precipitation data of 11 cities and counties in Chaohu Lake basin during the flood season from 1955 to 2005 are segmented, and the corresponding precipitation history cloud and trend cloud are established one by one, and the two are combined according to the weight to form a forecast cloud ; By predicting the cloud forward generator generates and weighted average of the corresponding precipitation forecast value, compared with the predictive value of neural network precipitation and the actual precipitation in 2006 were compared, the prediction results were significantly better than the neural network, closer to the actual precipitation value. The research shows that this method has some advantages for mining short time series precipitation features, simulating precipitation randomness and fuzziness, and predicting precipitation in the study area well.