数据驱动的车载空调设定温度预测研究

Data-driven Prediction of Vehicle Air Conditioner Set Temperature

为了对用户期望的车载空调温度进行实时预测,本文提出了一种习惯温度预测模型和时间序列温度预测模型双模型耦合的方法对车载空调设定温度进行实时预测。该方法以车内和外界的多维度信息作为输入,通过过滤式和随机森林对特征进行筛选,并根据实际应用场景集成模型来对用户期望的空调设定温度进行预测。最后使用该模型对测试数据进行验证。结果表明本文提出的双模型耦合的方法对用户空调设定温度的预测结果平均绝对百分比误差(MAPE)为0.049,能够精确地对车载空调温度进行预测,从而为智能化、个性化调控空调提供辅助决策。

In order to predict the real-time temperature of the vehicle air conditioner a user desires,this paper proposes a method that is coupled with the dual models of habit temperature prediction and time series temperature prediction to predict the real-time temperature of the vehicle air conditioner.The method takes the multi-dimensional information on the car and the information on the outside world as input,filters the features through filtering and random forest,and predicts the user′s desired air-conditioning set temperature according to the actual application scenario integrated in the model.Finally,the model is used to verify the test data.The results show that the dual model-coupled method predicts the mean absolute percentage error(MAPE)of the user′s vehicle air-conditioning set temperature to be as accurate as 0.049,thus providing auxiliary information on decision-making for intelligent and personalized air-conditioning control.