基于径向基函数网络的垂柳茎体含水量日变化模型的研究

Diurnal Variation Model of Water Content in Stem of Weeping Willow Based on Radial Basis Function Network

研究了利用径向基函数网络建立垂柳茎体含水量的日变化模型。以SWR乔木茎体水分传感器测量垂柳茎体含水量的日变化数据为时间序列。通过调整系数sce、g,改变径向基函数的输入向量维数,以达到模型的精确化。设置输入向量的维数分别为2、4、5,对输入向量按照不同的维数进行分组后,输入到神经网络中进行训练。为了得到模型的具体参数,选取另一组样本数据进行验证。结果表明,利用径向基函数网络建立的垂柳茎体水分日变化模型具有可行性,预测数据与观察数据具有很好的跟随性。同时试验结果也表明,输入向量的维数并不是越大越好,这次试验中,2维的要明显好于4维、5维的。

The establishment of diurnal variation model of water content in stem of weeping willow based on radial basis function network（RBF） was researched.The diurnal variation data of water content in stem of weeping willow was measured by standing-wave ratio（SWR） sensor,which was taken as time series.Through adjusting coefficients of sc and eg and changing the dimensions of input vectors of RBF to obtain more precise model.The dimensions of input vectors were set up as 2,4 and 5 respectively.After the input vectors were grouped according to the different dimensions,they were imported into the neural network to training.In order to obtain the specific parameters of the model,another group of sample data was selected to verify the results.The results showed that the diurnal variation model of water content in stem of weeping willow based on RBF network has feasibility for the simulation results agreed well with the measurement value.Meanwhile,the results also indicated that the dimension of input vectors is not the bigger,the better.In the experiment,the result of two dimensions of input vectors was obviously better than that of four dimensions and five dimensions of input vectors.