Evaluation And Optimization Of Surface Irrigation Performance And Its Parameters Based On BP Artificial Neural Networks

In North China, crop growth depends closely on agricultural irrigation because of less rainfall. However, agricultural water source is losing seriously due to unsuitable irrigation measures. Until now, surface irrigation is still widely used in Northern China by constraint of economic status, agricultural management and farmland use form. In such cases, to optimize surface irrigation system as well as its management that can improve irrigation efficiency relieve water resources crisis and protect food security.In this study, existing research results are firstly summarized on surface irrigation performance evaluation and optimization. Hereafter, main factors and their level values, that impact surface irrigation performance, are selected as focus. Then a suitable numerical model is used to simulate amount of irrigation processes, wherein, surface microtopography model is coupled. As a result,13440 groups of data for irrigation performance indices are obtained. By means of Matlab software, BP artificial neural network is used to analyze these data and the following results are achieved:(1) An approach, that combines numerical model of surface irrigation with microtopography model which includes spatial variability, is proposed to overcome limitation of only considering surface slope of existing methods.(2) 13440 groups of data about surface irrigation parameters are calculated out by simulations to obtain the mapping relationship between surface irrigation parameters and performance indices. As a result, mapping database is established between surface irrigation parameters and performance indices. This work provides a basis for intensive explorion of surface irrigation system optimization.(3) The mapping database shows that irrigation efficiency is significantly influenced by surface slope, land-leveling precision (measured by standard derivative of surface microtopography, that is Sd) and basin length. Concretely, the condition, Zmin>0, is suitable to long basin rather than short basin because low storage efficiency will appear in the latter case and longer irrigation should be performed.(4) BP neural network is constructed to evaluate surface irrigation system under any given conditions. Meanwhile, mapping database introduces variability of surface microtopography. As a result, land level precision (Sd) rather than surface microtopography is only given in the evaluation of surface irrigation performance.(5) The optimization model for irrigation time (toc) and unit discharge (q) is constructed. In the model, the inputs needed to be adjusted according to the outputs, thus the three performance indicators can be considered together.