基于灰色投影优化随机森林算法的输电线路舞动预警方法

Early warning method of transmission line galloping based on random forest optimized by grey relation projection

为实现对输电线路舞动的准确预警,多种预警模型被建立。其中对于预警模型的研究点多集中于参数的最优选取,而尚未开展针对训练样本的聚类改进,特别是地形参量存在难以客观衡量的问题,容易导致预测精度不佳抑或模型存在适应性问题。为此,提出了一种基于灰色投影优化随机森林算法的舞动预警模型。通过加权灰色关联投影法来优化选取与待预测样本关联度较高的样本作为训练样本集,以此来改进随机森林模型对低关联度数据的拟合能力。此外,模型还通过多维度参量即导线分裂数、直径、档距等内部参量以及风速、风向角、湿度等外部参量作为输入特征量来提高预警的准确性。利用历史舞动数据对模型进行检验,结果显示该模型的预警结果相较于传统随机森林算法和BP神经网络具有明显优势。通过该方法,可为输电线路舞动预警提供一种新的解决路线。

In order to achieve accurate early warning of galloping,many kinds of early warning models have been established.The research points of early warning models are mostly focused on the selection of optimal parameters.However,the clustering improvement of training samples has not been carried out,especially the terrain parameters,which are difficult to measure objectively.It is easy to lead to poor prediction accuracy and the adaptability of models.Therefore,a galloping early warning model based on random forest optimized by grey correlation projection was proposed in this paper.The method of weighted grey correlation projection was used to select optimal training samples which were tightly correlated with the predicted samples,so as to improve the fitting ability of random forest model when inputting low correlation data.In addition,multi-dimensional parameters were considered as input factors in order to improve the accuracy of early warning,including internal factors(conductor splitting number,diameter,spacing,etc.)and external factors(wind speed,wind direction angle,humidity,etc.).The historical data of transmission line galloping was used to verify the effectiveness of the proposed model.The results show that the early warning results of this model have obvious advantages in accuracy when comparing with traditional random forest algorithm and BP neural network.This method could provide a new solution for the early warning of transmission lines galloping.