遗传算法与小波神经网络在ET_0预测中的应用

Application of genetic algorithm and wavelet neural network in ET_0 prediction

ET_0是计算作物需水信息的重要依据,对于发展精准灌溉、制定中长期的农业用水策略具有重要的指导意义。选择将P-M公式的计算值为ET_0标准值,利用SPSS软件对多种气象数据进行相关性分析并以最高气温、最低气温、日照时数和相对湿度作为输入因子,采取小波神经网络将归一化的历史ET_0标准序列分别进行高低频带的预测,建立结合遗传算法优化小波神经网络的ET_0预测模型。通过仿真对比,基于遗传算法优化的小波神经网络与未采用遗传算法优化的小波神经网络相比具有更高的精度和更快的收敛速度,其中的小波算法能有效处理ET_0序列在预测过程中的周期非平稳特性。经实验样机平台验证遗传算法优化小波神经网络在作物需水预测中具有良好的性能,模型能够为精准灌溉、农业中长期需水决策提供指导作用。

ET0 is an important basis for the calculation of crop water requirement information,which has vital guiding significance for the precision irrigation development and the formulation of agricultural water strategy for medium and long term.The values of P-M formulation are selected as ET 0 standard values.SPSS is utilized to analyze the relevance between multiple meteorological factors and ET 0 values.Then the highest temperature,lowest temperature,sunshine hours and relative humidity factors are chosen as input factors.Wavelet neural network is adopted to predict the different frequency band of the normalized history ET 0 standard value sequence respectively.The prediction model of wavelet neural network was established based on genetic algorithm to predict the ET 0.By comparison,the wavelet neural network optimized with genetic algorithm has higher prediction accuracy and faster convergence speed than wavelet neural network model which is not optimized by genetic algorithm,the problem of the periodic non-stationary characteristics of ET 0 sequence in the prediction process is solved effectively.Tested in the experimental prototype platform,the wavelet neural network model optimized with genetic algorithm has good performance in water demand prediction,which will provide the guidance in precision irrigation and agricultural water strategy of medium and long term.