基于多模型离散覆盖的宿州初霜期动态灰预测

初霜冻害是影响宿州市秋收作物收成的重要因素。根据季节事件灰预测和多模型离散覆盖机理，创建不同零起点的灾变子序列并建立多个GM(1,1)模型，按照灰关联度择优选模，利用新息或灰数依次递补，对初霜期10月23日出现年份进行预测。提出按照频次对多模型离散覆盖进行白化以确定最有可能出现的年份，经预报检验效果较好，对初霜期的长期预测有一定的参考价值。

Dynamic Prediction of the First-Frost Period in Suzhou City-Based on Multi-model Discrete Cover

First-Frost disaster is a key factor of affecting the outcome of the kharif crop in Suzhou city.This paper, according to the mechanism of seasonal event grey prediction and multi-model discrete cover, establishes different zero-start-point catastrophic subsequences and multi-GM(1,1) models and by optimizing the choice of model by grey correlation degree and by means of successive modeling of innovation and gray data, conducts a dynamic prediction of grey target interval of the First-Frost period, thus to ascertain the most possible year of disaster by whitening multi-model discrete cover according to frequency of disaster happening, which is conducive to long-term prediction of the First-Frost period after experiments with significant effects.