铁道学报
鐵道學報
철도학보
2014年
3期
9-13
,共5页
物联网%城市轨道交通%多维度%客流密集度指数%实时预测
物聯網%城市軌道交通%多維度%客流密集度指數%實時預測
물련망%성시궤도교통%다유도%객류밀집도지수%실시예측
internet of things%urban rail transit%multi-dimension%passenger crowd index%real-time forecast
实时准确的客流预测是城市轨道交通客流预警和疏导的基础。本文针对城市轨道交通车站站台的实时客流密集度指数预测问题,根据实时客流的“间歇性”特点,依据30 s为周期的真实检测数据,分别构建低、中、高3个时间维度在线实时预测模型。根据应用需要,对30 s 低纬度预测采用自回归与移动平均整合模型(ARI-MA),对3 min左右中维度提出多因素logistic预测模型,15 min构建一种灰色与移动平均整合模型,并分别对预测参数进行估计。通过对10余个车站早晚高峰及平峰不同数据的大量在线实验验证模型的准确性,以北京地铁动物园站为例进行介绍,3个维度精度分别达到97%、95%、99%。结果表明：采用本文提出的模型较其他时间序列模型进行城市轨道交通车站设施的实时客流预测,具有更好的预测性能。本文所提模型已经用于北京市轨道交通安全防范物联网示范工程中,初步取得较好的实践效果。
實時準確的客流預測是城市軌道交通客流預警和疏導的基礎。本文針對城市軌道交通車站站檯的實時客流密集度指數預測問題,根據實時客流的“間歇性”特點,依據30 s為週期的真實檢測數據,分彆構建低、中、高3箇時間維度在線實時預測模型。根據應用需要,對30 s 低緯度預測採用自迴歸與移動平均整閤模型(ARI-MA),對3 min左右中維度提齣多因素logistic預測模型,15 min構建一種灰色與移動平均整閤模型,併分彆對預測參數進行估計。通過對10餘箇車站早晚高峰及平峰不同數據的大量在線實驗驗證模型的準確性,以北京地鐵動物園站為例進行介紹,3箇維度精度分彆達到97%、95%、99%。結果錶明：採用本文提齣的模型較其他時間序列模型進行城市軌道交通車站設施的實時客流預測,具有更好的預測性能。本文所提模型已經用于北京市軌道交通安全防範物聯網示範工程中,初步取得較好的實踐效果。
실시준학적객류예측시성시궤도교통객류예경화소도적기출。본문침대성시궤도교통차참참태적실시객류밀집도지수예측문제,근거실시객류적“간헐성”특점,의거30 s위주기적진실검측수거,분별구건저、중、고3개시간유도재선실시예측모형。근거응용수요,대30 s 저위도예측채용자회귀여이동평균정합모형(ARI-MA),대3 min좌우중유도제출다인소logistic예측모형,15 min구건일충회색여이동평균정합모형,병분별대예측삼수진행고계。통과대10여개차참조만고봉급평봉불동수거적대량재선실험험증모형적준학성,이북경지철동물완참위례진행개소,3개유도정도분별체도97%、95%、99%。결과표명：채용본문제출적모형교기타시간서렬모형진행성시궤도교통차참설시적실시객류예측,구유경호적예측성능。본문소제모형이경용우북경시궤도교통안전방범물련망시범공정중,초보취득교호적실천효과。
Accurate real-time forecast of passenger flow is the basis of passenger flow early warning and evacua-tion in urban rail transit.This article focused on the real-time station platform passenger crowd index on fore-cast problem present with urban rail transit.According to the intermittent feature of real-time passenger flow, the on-line real-time forecast models for 30 s,3 min and 15 min dimensions were built on the basis of 30 s cyc-ling of real test data.In view of practical needs,the ARIMA model for 30 s,the multi factor logistic model for 3 min and the integrated gray and move average model for 1 5 min were built to estimate forecast parameters re-spectively.On-line tests during morning and evening peak time and common time at more than 10 stations proved the correctness of the structrual models.Taking the Beij ing Metro Zoo Station for case study,the de-grees of accuracy corresponding to the three above-mentioned time dimensions were obtained as 9 7%,9 5% and 9 9%.The research results show that the proposed models are of good effects on predicting real-time passenger flow in urban rail transit stations.The models have been used in the demonstrative project of the Beijing Metro internet of things for safety and initial results have been achieved.
