The Prediction And Analysis Of Huili’s Meteorological Factors To "Hongda" Tobacco’s Nicotine Content Based On BP Neural Network

Nicotine content determines the quality and taste of tobacco products, thereby affecting the economic benefits of the production of tobacco. Meanwhile, it is closely related to the ecological environment especially the meteorological factors. The apparent three-dimensional climate characteristics of Huili country, which is an important part of liangshan tobacco area, has a great influence on "Hongda" planting and the nicotine content of tobacco leaf.This study aims to clear characteristics of the county meteorological factors through statistical analysis, and forecast the content of nicotine of "Hongda" leaf based on the regression analysis of the meteorological factors and the BP neural network. According to the comparison analysis, the optimal forecasting method is selected in the prediction of the nicotine content. The main results were as follows:(1) In the meteorological factors of different time periods, transplanting period humidity, transplanting the reach root period of rainfall, and humidity, prosperous long-term rainfall, mature period humidity have a significant negative correlation with nicotine content; meanwhile, root period of sunshine percentage, prosperous long-term sunshine percentage, prosperous long sunshine time, and nicotine content was significant positive correlation;(2) the whole fields of meteorological factor in the phase of the sunshine percentage in open field, rainfall, humidity, and "big red" tobacco nicotine content has obvious relevance, the stage in which field rainfall, field humidity has a significant negative correlation with nicotine content, the field phase of the sunshine percentage and nicotine content was significant positive correlation.(3) In the regression analysis, compare forecast results, the overall regression prediction accuracy is better than that of forced regression; Not to withdraw the principal component factor conditions, based on the period of time meteorological factors stepwise regression prediction precision is highest; After extracting principal component factor, based on the principal component factor F4 (humidity, temperature), F5 (sunshine time, radiation), stepwise regression prediction precision is highest; Before and after contrast to extract the principal component factor regression analysis, regression analysis from the field to stage principal component factors F4 (humidity, temperature), F5, sunshine time, radiation) as the independent variable stepwise regression prediction precision is highest, the most suitable predict nicotine content.(4) in the neural network models to predict, extract the principal component factor conditions, meteorological factor neural network prediction model based on phase field measurement accuracy is the highest; After extracting principal component factor, based on the period of time (humidity, temperature), principal component factor F1, F2, sunshine time, radiation), F3 (percentage of sunshine, rainfall), the neural network prediction precision is highest; Before and after contrast to extract the principal component factor prediction accuracy of neural network is principal component factors as input layer of neural network is higher than that in pure prediction. accuracy of meteorological factors as the input layer. Overall, neural network analysis of which period of time in the principal component factor F1 (humidity, temperature) and F2 (sunshine time, radiation), F3 (percentage of sunshine, rainfall) as the input layer of neural network model to predict the most suitable for nicotine content.(5) compared with neural network model and regression analysis shows that in a period of time (humidity, temperature), principal component factor F1, F2, sunshine time, radiation), F3 (percentage of sunshine, rainfall) as the input layer of neural network model prediction accuracy is the highest, the most appropriate in view of the "big red" tobacco nicotine content is predicted.