基于BP神经网络的江苏省秸秆资源量预测  

A Study of Straw Resources Prediction Based on the BP Neural Network: A Case Study of Jiangsu Province

作　　者：丁美[1] 籍春蕾[1] 邹碧莹[1] 赵言文[1] 

机构地区：[1]南京农业大学资源与环境科学学院,南京210095

出　　处：《资源科学》2011年第11期2197-2203,共7页Resources Science

基　　金：中国可再生能源规模化发展项目:"江苏省生物质能资源调查;评估和开发利用规划研究"(编号:A2-B10-CS-2009-001)

摘　　要：农作物秸秆是地球上第一大可再生资源,为能更好的合理开发利用农作物秸秆资源,缓解日益突出的资源短缺、环境污染与经济发展的矛盾,对其进行预测研究是非常之必要的。本文系统分析了江苏省秸秆资源现状及其资源量变化趋势影响因素,并以1990年-2008年历史数据和2009年农作物秸秆资源普查数据为基础,选取理论资源量、人均资源量和单位播种面积资源量为预测评价指标,基于BP神经网络(BP-ANN)对江苏省农作物秸秆资源的评价指标发展趋势进行预测。结果表明:建立的BP神经网络预测模型的相对误差基本在5%的范围内,平均相对误差在2%左右,预测结果与实际有较高的拟合度,且对数据具有较好的适应能力。在未来5年内,江苏省秸秆理论资源量呈平稳发展趋势;而人均资源量和单位播种面积资源量呈下降趋势,前者较后者下降幅度大。预测结果与当地发展规划趋势相一致,该方法具有很强的实际应用价值。本文最后针对江苏省实际,提出了农作物秸秆资源开发利用相关建议。Straw is the largest renewable resource on the Earth. There are about 60-80 million ton of straw resources produced each year in China. It is necessary to predict the amount of straw resources in order to reasonably develop and use straw resources, which would be helpful in easing increasingly serious contradictions among resources shortage, environmental pollution, and economic development. Jiangsu Province, rich in straw resources, is one of the biggest agricultural provinces. The Back-Propagation （BP） neural network method was used to identify and predict a non-linear procedure for resources assessment. First of all, the author systematically analyzed comprehensive utilization and trend factors of straw resources in Jiangsu. Currently, straw resources are primarily utilized as fertilizer, energy, industrial raw materials, animal feed, and base materials. The trend factors of straw resources are the economic output of crops, the ratio, and the collection coefficient of straw. This study examined straw resources in Jiangsu Province using data from statistical yearbooks during the period 1990-2008 and data collected in 2009 combined with the BP neural network method, predicting trends in straw resources with indices of theoretical amount, per capita amount, and per planting area amount. Results show that the relative errors of the established BP neural network model were basically -5% and the average relative errors were 2%. The prediction of the BP network model showed a higher fitting degree with reference to reality, which is indicative of its good ability to adapt to the data. Using the established BP neural network model, the author predicted the amount of straw resources in the next five years, showing that the theoretical amount would vary between 3.9-4.1 million tons, the per capita amount would vary between 650-900 kg, and per planting area amount would vary between 4.7-5.5 tons per hm2. The development trend of the theoretical estimates would be generally stable in Jiangsu Province while the trends

关 键 词：BP神经网络 秸秆资源 预测 开发建议 江苏省 

分 类 号：S216.2[农业科学—农业机械化工程] TP183[农业科学—农业工程]