【作者】 陈斌；

【导师】 丁世飞；

【作者基本信息】 山东农业大学 ， 农业昆虫与害虫防治， 2007， 硕士

【摘要】 玉米螟[(Ostrinia palustralis （Hübner）],俗名玉米钻心虫,属鳞翅目螟蛾科。玉米螟在我国主要有欧洲玉米螟[Ostrinia nubilalis （Hübner）]和亚洲玉米螟[Ostrinia furnacalis （Guenée）],其中亚洲玉米螟在我国是优势种,发生严重,在我国各地均有分布。玉米螟是公认的世界性大害虫,一直是农业害虫预报和防治的重点。玉米是我国主要的粮食作物之一,同时又是主要的饲料作物。因此,玉米螟在我国为害严重,不但给玉米种植造成很大的经济损失,同时给养殖业带来很大损失。本文就气象因子对二代玉米螟种群的作用;用于预测的气象因子的筛选;基于逐步提高预测精度的统计预测;模糊数学应用于玉米螟预测预报进行等方面进行了研究,主要结论如下:1、本文利用山东省宁阳县1989-2003年玉米螟的发生资料以及气象资料,运用了相关分析、逐步回归分析、通径分析以及灰色关联分析对影响二代玉米螟种群动态的气象因子进行分析。结果显示气象因子对二代玉米螟种群消长的作用和二代玉米螟的生物学特性是相符合的。结合几种分析方法得出:影响我省影响二代玉米螟消长的主要气象因子是5月⁸月份的气温和降水。2、筛选合适的预报因子是预测预报成败的关键。本文利用多元逐步回归法筛选出了用于预测二代玉米螟虫量的气象因子,分别是6月份平均气温（ x₁ ）;6月份平均降水（ x₂）;7月份平均降水（ x₄）;5月份平均降水（ x₆）;8月上旬平均气温（ x₇）;8月上旬平均降水（ x₈）。3、利用筛选出的预测因子分别运用多元线性回归预测法、多项式逐步回归预测法建立了二代玉米螟发生程度的预测模型,并着重探讨了BP（Back Propagation）人工神经网络预测法在二代玉米螟预测中的应用,结果显示,该方法具有很高的预测精度,预测精度平均值达到95.53675%,有93.33%的年份预测精度超过了90%,100%的年份预测精度超过85%。本文为了进一步提高预测精度,提出了基于信息熵确定权系数的组合预测模型,并对玉米螟的发生进行了预测,取得满意的效果。预测精度平均值达到95.63094%,有93.33%的年份预测精度超过了92%,100%的年份预测精度超过86.9%。4、影响农业害虫种群发生发展的因素具有很大的不确定性和模糊性,因此本文尝试将模糊数学方法应用于玉米螟的预测。应用模糊模式识别和基于信息熵模糊物元综合评判模型对二代玉米螟虫量进行预测取得了很好的效果,经过验证两个模糊预测模型的准确率均达到100%。模糊数学理论充分考虑了农业害虫种群发生系统的复杂性和模糊性,经过验证模糊数学方法在农业害虫预测预报中是切实有效的。更多还原

【Abstract】 The corn borer, known as a famous pest, belongs to Lepidoptera Pyralidae. In our country the corn borer is comprised of the Asian corn borer and the Europe corn borer. The Asian corn borer is the preponderant species in our country and distributes all over our country with serious occurrence. The corn borer is the focus of the pest forecasting and the pest manage. The corn is the primary grain crop and fodder, so the corn borer brings high loss to the corn planting and the livestock breeding.This paper studies the effect of the meteorological factors to the second-generation corn borer, and then the feasibly meteorological factors for forecasting are filtrated, and the usage of the statistical forecasting methods in order to increase the precision of forecasting, and the usage of the Fuzzy math method to forecasting the second corn bore.1. Using the occurred data of the corn borer and the synchronous meteorological data of NingYang, ShanDong province in 1989-2003 we analyze the effect of main meteorological factors to the population dynamics of the second corn borer with the correlation analysis; the multinomial step-wise regression analysis; path analysis and the grey correlation analysis. The result shows that the effect of the meteorological factors accord with the biology characteristic of the second-generation corn borer.Integrated with the above ways the conclusion is gained that the main effect factor to the amount of the second-generation corn borer is the precipitation of 5-8 month.2. Filtrating the appropriate forecasting factors is the bottle-neck of forecasting. In this paper the meteorological factors for forecastinging the amount of the second-generation corn borer are gained by the multinomial step-wise regression analysis. These are the average daily temperature in June （ x₁ ）; the average precipitation in June （ x₂）; the average precipitation in July （ x₄）; the average precipitation in May （ x₆）; the average daily temperature in August （ x₇）; the average precipitation in August （ x₈）.3. After using the poly-linear regression model and the multinomial step-wise regression analysis, the BP （Back Propagation） artificial neural network （ANN） model is studied in forecastinging the population dynamics of the second-generation corn borer with the meteorological factor filtered by using poly-linear step-wise regression. It has very high forecasting precision by applied. Using the BP ANN the average forecasting precision reach 95.53567%, and have 93.33% samples with the 90% veracity at least. All samples have 85% veracity at least.In order to elevate the forecasting precision more, the combination forecasting model based on using entropy to ascertain the weighting coefficients is used to forecasting the corn borer and a satisfactory effect is acquired. Using this method the average forecasting precision reach 95.63094%, and have 93.33% sample with the 92% veracity at least. All samples have 86.9% veracity at least.4. The influence on the occurrence and the development of the pest is uncertain and fuzzy, so the Fuzzy theory is tried to forecasting the corn borer. The Fuzzy pattern recognition and the Fuzzy matter element decision model are used to forecasting the corn borer and gain satisfied results. After put into test the two methods have 100% veracity. Fuzzy theory sufficiently takes the fuzzy and the complexity of the pest developmental system into account, so it has feasible usage in the pest forecasting.更多还原