BP Neural Network Based On Sub-humid Plains Irrigation Area Of Soil Moisture Forecast

Huantai, as a representative area of plain well irrigation district in Semi-humid region, its agriculture was on great dependence of irrigation.The Irrigation level directly determined the agricultural production, effected the GDP. Higher requirements of irrigation management must be proposed to achieve the sustainable development of agriculture. To achieve higher yield, improve water use efficiency, reasonable irrigation schemes should be made, also more accurately soil moisture conditions of future should be forecast. Therefore, making real-time crop irrigation amount forecast, and irrigating crops as optimization of the forecast, to make the unit water produce the best possible results, now is an important problem to solve.In the situation that water resources are seriously shorted, high active irrigation water management is important part of water saving and improving water productivity in irrigated agriculture. The core of water saving irrigation is putting planned water use into practice.Based on consulted losts of achievements about soil moisture forecast and irrigation experiments,this thesis analysed the method of soil moisture forecast.This thesis used MATLAB language to set up a BP neural network prediction model of soil moisture, studied the area based on actual situation, considered temperature, air humidity, average wind speed, light intensity, total radiation, irrigation, irrigation dates, rainfall, crop growth conditions as the input items, soil water storage as output items.In Water-saving agricultural demonstration base of National 863 plans projects. This thesis analyzed water using situation of the test area and carried out experimental studying on non-sufficient irrigation of winterwheat. TRIME-TDR was used in the experiment to measure soil moisture on the field for a long-term observation,for it's of high precision, good stability, without destroying soil structure, easy to carry. Application showed that: the model was of high precision, the maximum absolute error was 0.15mm, it was beneficial for scientific and precise irrigation,also was benefical for improving irrigation management level.