摘要
研究根据室内尿素水解试验资料,建立了以温度、水分、时间为输入因子,尿素态氮含量为输出因子,拓扑结构为3-2-1的BP神经网络预测模型,以及Verhulst灰色预测模型和零级动力学模型,并分析比较了三种模型的预测效果。结果表明:3种预测模型均能满足模拟精度要求,所建立BP神经网络模型模拟值与实测值的平均相对误差、相关系数和决定系数分别为2.39%、0.992 4和0.984 5,具有较高的预测精度和良好的稳定性,并且模拟效果明显优于Verhulst灰色预测模型和零级动力学模型,可以较好地描述尿素水解动态变化过程,为尿素水解定量研究提供了精确的科学依据。
According to laboratory test data of urea hydrolysis, it established BP neural network model, Verhulst grey forecasting model, zero-order kinetics model and compared simulation effect of the three models. In the BP neural network model, temperature, water content and time were taken as input data, Urea content was taken as output, topological structure was 3-2-1. It shows： the three models could meet the requirements of simulation accuracy, the average relative error, correlation coefficient and determination coefficient between simulation value and measured value, which was calculated by BP neural network model, was 2. 39%.0. 992 4 and 0. 984 5, respectively. The BP neural network model has higher prediction precision, better stability and better simulation effect than BP neural network model and Verhulst grey forecasting model. The BP neural network model well described the dynamic process of Urea hydrolysis and provided accurate scientific basis.
出处
《节水灌溉》
北大核心
2016年第9期57-61,共5页
Water Saving Irrigation
基金
国家自然科学基金资助项目(51579168
51249002)
教育部博士点基金(20121402110009)
山西省科技攻关项目(20140311016-6)
