基于LSTM-LightGBM模型的车站环境温度预测

客运火车站环境温度易受其他环境特征变量如湿度、PM2.5、二氧化碳等影响,传统的单变量预测算法并未考虑其他环境特征变量的影响因素;为进一步准确预测车站环境温度值,提出了结合长短期记忆神经网络LSTM与梯度提升算法LightGBM的组合模型,对客运站环境温度值进行预测;首先将预处理数据输入LSTM模型,对环境特征变量湿度、二氧化碳、PM2.5、PM10进行单变量预测;再将环境特征变量的LSTM输出预测值输入LightGBM模型得出环境温度预测值;根据波形图与均方根误差RMSE对比分析,基于LSTM-LightGBM的组合模型预测方法可以保留LSTM模型对单变量预测的周期性特点,且可表现出环境特征变量输入LightGBM模型后对温度预测的非平稳变化;结果表明基于LSTM-LightGBM的组合模型方法比单纯使用LSTM方法更接近原始波形,具有更低的RMSE。

The Station Environment Temperature Prediction Based On LSTM-LightGBM Model

The environment temperature of passenger railway station is easily affected by other environmental characteristic variables such as humidity, PM2.5, carbon dioxide, etc. The traditional univariate prediction algorithm does not consider the influence factors of other environmental characteristic variables.In order to further accurately predict the environmental temperature of the passenger station, a combined model combining LSTM neural network and LightGBM gradient lifting algorithm is proposed to predict the environmental temperature of the passenger station. Firstly, the pre-processed data were input into the LSTM model, and environmental characteristic variables such as humidity, carbon dioxide, PM2.5 and PM10 were predicted. Then input the predicted value of environmental characteristic variables output by LSTM into LightGBM model to get the predicted value of temperature. According to the comparison and analysis of the waveforms and RMSE, the combined model prediction based on LSTM-LightGBM can retain the periodicity of the univariate prediction used by LSTM model, and can show the non-stationary changes of the temperature prediction after the environmental characteristic variables input into LightGBM model. The results show that the combined model method based on LSTM-LightGBM is closer to the original waveform and has lower RMSE than the method using LSTM alone.