| Along with the social economy rapid development, the shortage of water resources and increasing labour costs and other issues emerging. Processing tomato as the Xinjiang characteristic industry in the country has a play a decisive role position, its planting area occupies90%above of countrywide, crop occupies countrywide80%, significant economic benefits. Processing Tomato under film drip irrigation as an advanced irrigation technology, inevitable toward automation, intelligent direction. To achieve such technology, on the basis of work is essential.In view of Processing Tomato under drip irrigation automation system of soil moisture forecast problem, on2010,2011in Xinjiang production and Construction Corps Agricultural Division eight national agricultural science and Technology Park, to carry out the forecast of soil moisture was studied in field experiment.In the processing of Tomato in the field soil moisture content based on the analysis of the features, using observed in2010,0～20cm,20～40cm,40～60cm,60～100cm of moisture ratio values and the corresponding mean daily temperature data as input factors, and then to two days and a week after the20～40cm moisture content values for output factors, establish a respectively to two days a week, for a time interval based on BP artificial neural network model for soil moisture forecast model, and then the input, output for training, using2011data to verify the trained network, error analysis and test of model accuracy.The paper obtained the following results:1、In the science and Technology Park soil conditions, processing tomato with drip irrigation under membranes spatial soil moisture content had moderate or weak variation;0～20cm soil moisture spatial structure is poor, and20～60cm soil moisture spatial structure is obvious, surface soil and field climate environment are closely linked.20～40cm soil water content is relatively stable, but on behalf of processing tomato soil moisture condition, can be used as a selection of forecast factors based on.2、Processing tomato with drip irrigation under membranes in120m x90m scale range, average block omnidirectional range value of23.81m, if the codomain is emplaced within the monitoring point to repeat cloth drop, reasonable layout of monitoring points number is20.3、Drip irrigation under plastic film processing tomato soil vertical direction is divided into three layers:0～20cm is the surface layer of the soil moisture in blast crisis, the external environment change greatly;60cm～100cm is relatively stable layer, soil moisture changes very little. After irrigation, in0-60cm soil layer, soil moisture content is before irrigation is increased obviously, while in the60～100cm little changed;Surface soil moisture change in water before, after vary greatly.4、In this paper, the prediction model of the input factor is five, the output factor A, so the input node number is5, the output node number is1,therefore the network hidden layer neuron number is4toll. The model uses9, good results.5、Set up two kinds of three layer structure based on BP network for drip irrigation under plastic film on soil moisture forecast model of processing tomato.Two days later the forecasting model BP (5,9,1) of the maximum relative error is14%, the average relative error is3.7%, the maximum absolute error is2.26%, the average absolute error of1.13%;A week after the forecasting model BP (5,9,1) of the maximum relative error is11%, the average relative error is3%, the maximum absolute error is2%, the average absolute error of1%.The two models can satisfy the actual production requirements. The model prediction accuracy is above85%, is in line with the forecast requirement. |
PDF Full Text Request