摘要
采用小波理论和Elman神经网络相结合的方法对风速进行短期预测,建立小波Elman神经网络预测模型。为了避免梯度下降法存在收敛速度慢、易震荡、陷入局部极小值等缺点,在神经网络学习过程中采用LM(Levenberg-Marquardt)算法和附加动量项法。通过实例分析,与小波BP神经网络模型进行比较,表明该模型具有较强的逼近和容错能力、较快的收敛速度、较好的预测效果。
This paper made the short-term prediction of wind speed by means of wavelet theory and Elman neural network. The wavelet Elman neural network model was established. The Levenberg-Marquardt algorithm and momentum in neural network learning process were applied in order to avoid the shortcomings of the gradient decent method, such as slow convergence rate, vulnerability of producing vibrations and falling into local minimum value. As the case study shows, the proposed model has stronger approximation and fault tolerance, faster convergence rate, and better forecasting effect, when compared with wavelet BP network model.
出处
《华东电力》
北大核心
2013年第4期798-801,共4页
East China Electric Power
基金
国家重点基础研究发展计划项目(973计划)(2012CB215203)
