Study On The BP Prediction Model Of Kostiakov-Lewis Infiltration Model Parameters

In view of the serious shortage of water resources in our country, the agriculture water consumption accounts for over 60% of the total water consumption in China and the reality of serious water waste, soil water infiltration parameters in this paper, we study the BP prediction model, to improve the effect of agricultural irrigation water use efficiency and provide the basic theory and technical support, has important practical significance to the national water resources sustainable development.Based on the regional field of frozen soil, frozen soil, salinization of soil infiltration experiments, to soil water infiltration parameters of Kostiakov- Lewis infiltration model as the main research object, analyses the field soil water infiltration parameters are the main influence factors and frozen soil, characteristic factors of saline-alkali soil water infiltration model to establish the corresponding parameters of BP prediction model, realized through basic soil physical and chemical index of soil water infiltration model parameter prediction. The results show that:Field(not freeze or not) saline-alkali soil water infiltration characteristics of main influence factors mainly include: the soil moisture content, structure, organic matter content, quality of a material, among them, the soil structure indicators need to be considered were for the first time the irrigation occurred in the process of structural deformation. For these main factors influencing soil water infiltration parameters input variable of BP prediction model, can obtain satisfactory prediction results, the average error is less than 6%. Among them: infiltration coefficient k of the average error less than 6%, the infiltration index, the average error less than 5%, a steady infiltration rate f0, the average error less than 5%. With the same linear regression forecast model, the comparison of nonlinear regression forecast model and prediction accuracy of BP prediction model is better than nonlinear regression prediction model(the average error of 10% smaller) and linear regression forecast model(the average error of 12%.).Frozen soil infiltration parameters of BP prediction model has good feasibility. In a freezing the basic physical and chemical parameters of saline-alkali soil and the characteristic factors affecting soil ground temperature(5 cm) as the input variable of BP prediction model of frozen soil infiltration model parameter prediction accuracy is higher, the average forecasting error is less than 9%, has good feasibility. Infiltration coefficient k of the BP prediction model to predict the average error of 8.319%; Infiltration index a BP prediction model model to predict the average error of 6.773%; Steady infiltration rate f0 BP prediction model to predict the average error of 4.1739%.Saline-alkali soil infiltration model parameters of the BP prediction model has good feasibility. Saline-alkali soil water infiltration parameters established by the BP prediction model of the average prediction error is less than 7%, among them: BP prediction model of rainfall infiltration coefficient k average error is 6.5583%; Infiltration index alpha BP prediction model of average error is 6.2243%; Steady infiltration rate f0 BP prediction model of average error is 5.8405%.In this paper to establish the freeze BP prediction model of saline infiltration parameters, especially in the frozen soil, saline-alkali soil infiltration parameters of BP prediction model, feel inadequate, perfect sample data, which can consider the influence factors of limitations, in the later study, strive to extend the sample set, improve the network generalization ability, thus improve the prediction accuracy of the network.