基于BP神经网络的输电线路覆冰增长模型研究

分析了现有输电线路覆冰增长模型在预测中的不足以及神经网络对非线性映射变量表达的优越性,提出了一种基于Levenberg-Marquardt学习算法的BP神经网络的覆冰增长预测模型。通过实验获取的覆冰增长数据样本训练BP网络,利用收敛的网络进行输电线路覆冰增长的预测,仿真实验误差1mm以下的有7组数据,远高于对比模型makkonoe模型的3组,验证了模型有效性,对输电线路的覆冰研究和预防有重要意义。

Study on Transmission Line Ice Accretion Mode Based on BP Neural Network

After analyzing the deficiency of existing prediction accuracy of ice accretion on transmission lines and the superiority of neural network for nonlinear variable mapping,a new method based on BP neural network which taked the Levenberg-Marquardt learning algorithm was proposed.This new prediction model was practiced by the ice growth data of experiment.Using the convergent prediction model,a successful ice growth prediction experiment was set up.The simulation result shows that there are 7 groups of prediction error less than 1mm,which is much better than the 3 groups of Makkonoe model.The prediction simulation verified that this new prediction model is a effective model.This new model plays a significant role in the prediction and prevention research.