建立土壤硫释放过程的人工神经网络模型

以温度、土壤含水率、胱氨酸添加量和土壤pH值作为土壤释放挥发性含硫化合物的主要影响因素,采用正交实验方法分析这些因素与土壤硫释速率的关系,利用BP神经网络算法对实验结果建模,并用模型对不同影响因素下的土壤硫释放情况进行预测。结果表明,网络模型对学习过的样本有较高预测精度,预测结果相对误差在2％以下,对未学生过的样本,误差为10％左右,表明人工神经网络方法建立的模型适用于土壤硫释放预测。

Establishment on Emission Model of Sulfur Gas from Soil Using Artificial Neural Network

Temperature, water content, adding amount of cystine and pH of soil are important factors which effect on emission of volatile sulfur compound from soil treated with cys-tine. A mathematieal model of the relation between influence factors and volatile sulfur compound emission rate has been established based on promoted BP neural network algorithm. With this model, the emission rate under different conditions were predicted. The result showed that the prediction capability for the studied samples was better, and its relative error was below 2%;and for the samples without being studied, the relative error was about 10%.Therefore, the artificial neural network has good ability for predicting emission of volatile sulfur compound from soil.