摘要
为准确预测用能负荷,通过调研高速公路典型构造物用能负荷历史数据,分析了交通量、天气状况、星期、月份以及节假日/工作日等多重因素对用能负荷的影响,采用主成分分析(PCA)方法对这些影响因素进行降维处理,消除原始序列的冗余性;分析了节假日/工作日属性和天气状况等因素对用能负荷特征的影响,提出了高速公路用能负荷预测模型候选数据集(CD)构建策略,在此基础上,采用长短时记忆(LSTM)网络模型对多变量用能负荷预测进行了动态时间建模,并依托广西桂柳高速公路实测数据对提出的预测模型进行了验证。分析结果表明:交通量、天气状况、星期、月份、节假日/工作日这5个主成分累计贡献率为85.54%,是高速公路用能负荷主要影响因素;隧道、收费站和服务区用能负荷的高峰时段分布各异,隧道用能负荷高峰时间波动较小,收费站用能负荷高峰集中在10:00~21:00,服务区用能负荷高峰时间段最短,仅集中出现在11:00~12:00;在不同季节测试日中,采用候选数据集构建策略及PCA处理后,提升了训练集数据的针对性,预测结果的精度也得到了提高,平均绝对百分比误差不超过12.33%,均方根误差不超过3.86,提出的预测模型在不同典型场景的负荷预测中均具有良好的适用性,可为高速公路自洽能源系统设计提供理论支撑。
To accurately predict the energy load,historical energy load data of typical highway structures were investigated,and the influences of multiple factors on energy load,including traffic volume,weather condition,week,month,and holiday/workday,were analyzed.The principal component analysis(PCA)method was applied to reduce the dimensionality of these influencing factors,thereby eliminating the redundancy in original sequences.The effects of holiday/workday attributes and weather condition on energy load characteristics were examined,and a strategy for constructing a candidate dataset(CD)for energy load forecasting models of highways was proposed.On this basis,the dynamic time modeling for multivariate energy load forecasting was performed by using a long short-term memory(LSTM)network model.The proposed forecasting model was validated by using empirical data from the Guilin-Liuzhou Highway in Guangxi.Analysis results show that the cumulative contribution rate of the five principal components of traffic volume,weather condition,week,month,and holiday/workday,is 85.54%,and they are the primary influencing factors of energy load on highways.The peak energy load periods for tunnels,toll stations,and service areas vary,with tunnel energy load peaks showing minimal fluctuations,toll station peaks concentrated between 10:00-21:00,and service area peaks occurring within the shortest period,only concentrated between 11:00-12:00.Across different seasonal test days,the application of CD construction strategy and PCA processing improves the specificity of the training data,resulting in enhanced prediction accuracy,with the mean absolute percentage error not exceeding 12.33% and the root mean square error not exceeding 3.86.The proposed prediction model demonstrates strong applicability for load prediction in various typical scenarios and provides theoretical support for the design of self-consistent energy systems for highways.
作者
张宇飞
蒋玮
张硕
王腾
肖晶晶
袁东东
ZHANG Yu-fei;JIANG Wei;ZHANG Shuo;WANG Teng;XIAO Jing-jing;YUAN Dong-dong(Key Laboratory for Special Area Highway Engineering of Ministry of Education,Chang'an University,Xi'an 710064,Shaanxi,China;School of Highway,Chang'an University,Xi'an 710064,Shaanxi,China;School of Civil Engineering,Chang'an University,Xi'an 710064,Shaanxi,China)
出处
《交通运输工程学报》
EI
CSCD
北大核心
2024年第5期40-53,共14页
Journal of Traffic and Transportation Engineering
基金
国家重点研发计划(2021YFB1600200)
国家自然科学基金项目(52122809)
陕西省重点研发计划(2024SF-YBXM-577)
陕西省秦创原“科学家+工程师”队伍建设项目(2024QCY-KXJ-020)。
关键词
交通与能源融合
预测模型
主成分分析
用能负荷
能源自洽
用能特征
integration of transportation and energy
forecasting model
principal component analysis
energy load
energy self-consistency
energy consumption characteristics
