【摘要】 航站楼等公共交通建筑中,客流特征是影响实际负荷需求的重要因素。本文以某航站楼的历史运行数据为基础,建立了基于随机森林的负荷预测模型,对时间、室外气象参数、客流统计数据进一步拆解得到24 h特征、星期特征、逗留人数、室外空气焓值,结合历史负荷预测空调负荷,并对不同特征组合下的模型性能进行对比。结果表明:引入逗留人数、前1 h内的历史负荷、室外空气焓值作为输入特征对于负荷预测模型的性能有显著提升。在选择恰当的输入特征前提下,随着输入特征维度增加,随机森林模型的训练效率更高,对空调负荷的预测性能更好。更多还原

【Abstract】 In public transport buildings such as terminal buildings, passenger flow characteristics are an important factor affecting the actual load demand. Based on the historical data and random forest algorithm, a load forecasting model of a terminal is established. The time, outdoor meteorological parameters and passenger flow statistics are further disassembled to obtain 24 h characteristics, weekly characteristics, number of stay and outdoor air enthalpy. The model performance under different feature combinations is compared. The results show that the performance of the load forecasting model is significantly improved by introducing the number of stay, historical load in the first hour and outdoor air enthalpy as input characteristics. Under the premise of selecting appropriate characteristics, with the increase of input feature dimensions, the training efficiency of model is higher and the prediction performance of air conditioning load is better.更多还原