基于ABC-BP的土壤侵蚀量预报模型研究

摘要

Taking the prediction of soil erosion amount as a goal,a BP neural network model with a topological structure of 5-7-1 is established according to the data from the Bulletin of River Sediment in China.In the model,the soil type,topography,slope,vegetation and rainfall are taken as input data and the soil erosion amount is taken as output.In view of the shortages of BP neural network method,the weight value and threshold value of BP neural network are optimized by using Artificial Bee Colony algorithm,and then the simulation effect of ABC-BP model is analyzed.It shows that the correlation coefficient and average relative error between simulation value and measured value are 0.9942 and 4.13％ respectively.The ABC-BP neural network model has perfect consistency and higher prediction precision.%以预测土壤冲刷量为目标,根据《中国河流泥沙公报》数据资料,建立了以土壤类型、地形、坡度、植被、降雨为输入因子,土壤侵蚀量为输出因子,拓扑结构为5-7-1的BP神经网络预测模型.针对BP神经网络模型缺陷,采用了人工蜂群算法(ABC)对BP神经网络的权值和阈值进行优化,建立了ABC-BP模型,并对该模型的性能进行了验证.结果表明,所建立的ABC-BP土壤侵蚀量预报模型模拟值与实测值的相关系数、平均相对误差分别为0.994 2和4.13％,两者之间无显著的统计学差异,具有较好的一致性和较高的模拟精度.