基于多元非线性回归算法的动车组变压器温度预测研究

Research on Temperature Prediction of EMU Transformer Based on Multiple Nonlinear Regression

对动车组状态参数进行预测是实现动车组数字化精准维修的重要前提之一。目前对动车组数据的利用方式主要以定量观测和定性分析为主,缺乏准确的状态预测量。为此,文章提出了一种基于多元非线性回归的预测算法,通过建立变压器温度预测模型,实现了变压器温度值的次日预测,并以此指导由温度导向的变压器日常运维项目。通过对比CRH5A型动车组车载数据中变压器温度原始数据和预测模型数据,结果显示采用该模型变压器温度预测精度达到95%以上,验证了所提方法的有效性,其对提高列车检修效率、节支降耗有着重要的作用。

The prediction of EMU state parameters is an important basis to realize digital precision repair.At present,the use of EMU data is mainly based on quantitative observation and qualitative analysis,lacking of accurate state prediction.Therefore,this paper proposes a prediction algorithm based on multiple nonlinear regression to establish a transformer temperature prediction model.The next day prediction of transformer temperature value is realized,which can guide the daily operation and maintenance project of transformer guided by temperature.By comparing the original data of transformer temperature in CRH5A EMU on-board data with the results of prediction model,the prediction accuracy of transformer temperature achieved by the model is more than 95%,which verify the validity of the proposed method that plays an important role in improving train maintenance efficiency,saving expenditure and reducing consumption.