基于实时占用感知的电梯群控调度仿真研究

doi:10.16182/j.issn1004731x.joss.25-0810

摘要/Abstract

摘要：

面向群控电梯调度中轿厢空间资源配置与高峰客流响应效率之间的矛盾，提出一种基于实时占用感知的PPO多目标调度方法。构建考虑轿厢容量约束的仿真环境，设计以乘客平均候梯时间、系统能耗与轿厢拥挤度为优化目标的奖惩机制，据此定义状态空间与动作空间，形成基于PPO的调度框架；开发集交通流可视化、策略调度与性能评估于一体的仿真平台。仿真结果表明：该方法在多种交通模式下均能显著降低长候梯率、能耗与拥挤度，表现出良好的适应性与收敛性。

关键词: 电梯群控调度, 强化学习, PPO算法, 实时占用感知, 多目标优化

Abstract:

To address the conflict between car space allocation and peak passenger flow response efficiency in elevator group control scheduling, amulti-objective scheduling method based on proximal policy optimization (PPO) with real-time occupancy perception was proposed. A simulation environment considering car capacity constraints was constructed, and a reward-penalty mechanism with average passenger waiting time, system energy consumption, and car congestion as optimization objectives was designed.Basedonthis,state and action spaces were defined to form a PPO-based scheduling framework; a simulation platform integrating traffic flow visualization, policy scheduling, and performance evaluation was developed. Simulation results show that this method significantly reduces long waiting rates, energy consumption, and congestion across various traffic modes, demonstrating good adaptability and convergence.

Key words: elevator group control scheduling, reinforcement learning, PPO algorithm, real-time occupancy perception, multi-objective optimization

中图分类号:

TP273

韩义勇,王诗雨,叶杨等 . 基于实时占用感知的电梯群控调度仿真研究[J]. 系统仿真学报, 2026, 38(7): 1993-2006.

Han Yiyong,Wang Shiyu,Ye Yang,et al . Simulation Study of Elevator Group Control Scheduling Based on Real-time Occupancy Perception[J]. Journal of System Simulation, 2026, 38(7): 1993-2006.

图/表 22

图1

表1

电梯环境参数

参数名称	参数取值
时间间隔 $Δ t$ /s	0.5
楼层数目 $H$	20
楼层高度 $h$ /m	3
电梯数量 $M$	4
电梯最大加速度 $a m a x$ /(m/s²)	1
电梯最大运行速度 $v m a x$ /(m/s)	2
乘客转移时间 $t p$ /s	3
乘客并行进出允许人数 $N p a r$	2
轿厢最大乘客数 $N c, m a x$	12
门动作时间 $t d$ /s	3
电梯门开关能耗 $E d$ /J	350
电梯待机平均功耗 $E s$ /W	100
电梯额定载重量 $m r$ /kg	800
电梯净重量 $m e$ /kg	300
曳引轮半径 $r t$ /m	0.27
曳引电机效率 $η m$	0.8
齿轮传动比 $i g$	0.9
轿顶反绳轮效率 $η s g$	0.9
轿顶反绳轮半径 $r s g$ /m	0.9

表1

图2

表2

电梯状态信息

环境	符号	参数说明
楼宇环境 $S$	$E t i$	$t$ 时刻电梯 $i$ 状态
	$R t u$	$t$ 时刻存在上行请求的楼层
	$R t d$	$t$ 时刻存在下行请求的楼层
电梯状态 $E$	$f i$	电梯 $i$ 当前所在楼层
	$v i$	电梯 $i$ 当前速度
	$D i$	电梯 $i$ 当前调度方向
	$f c u r, i$	电梯 $i$ 当前目的楼层
	$f r e s, i$	电梯 $i$ 预约目的楼层
	$L i$	电梯 $i$ 当前载荷率
	$F i$	电梯 $i$ 当前占用率
	$O i$	电梯 $i$ 轿厢拥挤度，若满载则不响应外呼信号
	$b d o o r, i$	1/0：电梯 $i$ 门当前打开/关闭

表2

表3

动作空间说明

参数符号

参数说明

f t a r, i

1~H：将电梯 $i$ 调配到相应目标楼层

-1：电梯 $i$ 不改变之前的目标楼层

0：电梯 $i$ 停梯，无调度任务

d t a r, i

1：电梯 $i$ 抵达目标楼层后向上运行

-1：电梯 $i$ 抵达目标楼层后向下运行

0：电梯 $i$ 抵达目标楼层后无运行方向

表3

表4

不同交通模式的权重分配

交通流模式	$ω 1$	$ω 2$	$ω 3$
上行高峰	0.6	0.3	0.1
下行高峰	0.5	0.3	0.2
层间交通	0.2	0.4	0.4
空闲交通	0.3	0.6	0.1

表4

图3

表5

信息编码规则及含义

符号	维度	编码值域/阈值	编码规则	图4对应
$D i$	3	{上,停,下}	对应方向置1	$D i$ 行
$f i$	20	{1,2,···,20}	对应楼层置1	紫色块 $f i$
$f c u r, i$	20	{1,2,···,20}		紫色块 $f c u r, i$
$f r e s, i$	20	{1,2,···,20}		紫色块 $f r e s, i$
$R t u d$	20	{1,2,···,20}	对应楼层有请求置1	$R t u d$ 行
$R t u$	20	{1,2,···,20}		$R t u$ 行
$R t d$	20	{1,2,···,20}		$R t d$ 行
$F i$	5	{0-0.2,···,0.8-1.0}	对应区间范围置1	$F i$ 行
$L i$	5	{0-0.2,···,0.8-1.0}	对应区间范围置1	$L i$ 行

表5

图4

图5

图6

图7

表6

PPO算法超参数设置

超参数	数值
裁剪系数	0.2
网络学习率	3×10^-4
最小批次尺寸	64
训练次数	50 000
GAE参数 $λ$	0.95
采样步数	2 048
更新次数间隔/s	10

表6

表7

图8

图9

图10

图11

表8

图12

表9

图13

参考文献 19

[1]	Li Huayu, Chu E T H. SmartRide: Intelligent Reservation and Scheduling for Elevators[J]. Journal of Ambient Intelligence and Smart Environments, 2025, 17(1): 75-100.
[2]	Al-Kodmany K. Smart Elevator Systems[J]. Journal of Mechanical Materials and Mechanics Research, 2023, 6(1): 41-53.
[3]	Yang Zhenshan, Yue Wenjiao. Elevator Traffic Pattern Recognition Based on Fuzzy BP Neural Network with SOM Algorithm[J]. Advances in Modelling and Analysis B, 2017, 60(4): 630-645.
[4]	Li Zewen, He Kemei. A Review of Research on the Development of Elevator Group Control Technology[J]. Engineering Advances, 2023, 3(5): 438-442.
[5]	田海, 赵小强. 群控电梯调度算法研究[J]. 计算机与网络, 2018, 44(18): 61-63.
	Tian Hai, Zhao Xiaoqiang. Research on Group Control Elevator Scheduling Algorithm[J]. Computer & Network, 2018, 44(18): 61-63.
[6]	Galarraga Joritz, Aitor Arrieta Marcos, Ali Shaukat, et al. Genetic Algorithm-based Testing of Industrial Elevators Under Passenger Uncertainty[C]//2021 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW). Piscataway: IEEE, 2021: 353-358.
[7]	高慧生, 许清风, 房志明, 等. 基于遗传算法的多目标电梯紧急疏散问题[J]. 计算机系统应用, 2021, 30(1): 168-173.
	Gao Huisheng, Xu Qingfeng, Fang Zhiming, et al. Multi-objective Elevator Evacuation Problem Based on Genetic Algorithm[J]. Computer Systems & Applications, 2021, 30(1): 168-173.
[8]	Zhang Yuting, Cui Wei. A Multi-objective Optimization Approach for Elevator Group Control Systems Based on Particle Swarm Algorithm[J]. Journal of Circuits Systems and Computers, 2024, 33(8): 2450138.
[9]	Shuoyan Chou, Aditya Budhi D, Dewabharata A, et al. Improving Elevator Dynamic Control Policies Based on Energy and Demand Visibility[C]//2018 3rd International Conference on Intelligent Green Building and Smart Grid (IGBSG). Piscataway: IEEE, 2018: 1-4.
[10]	张家瑞, 李海鹰, 苗建瑞, 等. 电梯群控系统调度模型及其改进ADMM分解算法[J]. 控制与决策, 2023, 38(1): 39-48.
	Zhang Jiarui, Li Haiying, Miao Jianrui, et al. A Dispatching Model and Improved ADMM Decomposition Algorithm for Elevator Group Control System[J]. Control and Decision, 2023, 38(1): 39-48.
[11]	Bapin Yerzhigit, Zarikas Vasilios. Smart Building's Elevator with Intelligent Control Algorithm Based on Bayesian Networks[J]. International Journal of Advanced Computer Science and Applications, 2019, 10(2): 16-24.
[12]	Hanif Mohammad, Mohammad Nur. Metaheuristic Algorithms for Elevator Group Control System: A Holistic Review[J]. Soft Computing, 2023, 27(21): 15905-15936.
[13]	顾玲丽, 董佳琦, 许洪华. 基于强化蚁群算法的高层住宅电梯群调度研究[J]. 计算机仿真, 2022, 39(1): 412-417.
	Gu Lingli, Dong Jiaqi, Xu Honghua. Research on Elevator Group Control and Scheduling of High-rise Residential Buildings Based on Reinforcement Learning and Ant Colony Algorithm[J]. Computer Simulation, 2022, 39(1): 412-417.
[14]	Wei Qinglai, Wang Lingxiao, Liu Yu, et al. Optimal Elevator Group Control via Deep Asynchronous Actor-critic Learning[J]. IEEE Transactions on Neural Networks and Learning Systems, 2020, 31(12): 5245-5256.
[15]	刘宇, 张聪, 李涛. 强化学习A3C算法在电梯调度中的建模及应用[J]. 计算机工程与设计, 2022, 43(1): 196-202.
	Liu Yu, Zhang Cong, Li Tao. Modeling and Application of Reinforcement Learning A3C in Elevator Scheduling Algorithm[J]. Computer Engineering and Design, 2022, 43(1): 196-202.
[16]	魏东, 张靖媛, 方硕. 电梯群控仿真模型构建及群控调度策略研究[J]. 控制工程, 2025, 32(10): 1895-1905.
	Wei Dong, Zhang Jingyuan, Fang Shuo. Construction of Elevator Group Control Simulation Model and Research on Group Control Scheduling Strategy[J]. Control Engineering of China, 2025, 32(10): 1895-1905.
[17]	Wang Shu, Gong Xuejian, Song Mulang, et al. Smart Dispatching and Optimal Elevator Group Control Through Real-time Occupancy-aware Deep Learning of Usage Patterns[J]. Advanced Engineering Informatics, 2021, 48: 101286.
[18]	Othman Saba A, A K Mohammed Jamal, Farag Mahel Mohammed. Variable Speed Drives in Electric Elevator Systems: A Review[J]. Journal of Physics: Conference Series, 2021, 1973(1): 012028.
[19]	Ciflikli Cebrail, Emre Oner Tartan. A Model for the Visualization and Analysis of Elevator Traffic[J]. Transportation Planning and Technology, 2019, 42(8): 868-880.

交通模式	满载识别	电梯能耗/J	平均候梯时间/s
上行高峰	有	9 730 886	48.96
上行高峰	无	13 980 575	55.41
下行高峰	有	10 947 640	49.38
下行高峰	无	12 192 587	52.11
层间交通	有	3 281 459	27.56
层间交通	无	3 530 393	28.73
空闲模式	有	1 794 582	14.90
空闲模式	无	1 872 118	15.17

数据序号	上行客流	下行客流	层间客流	乘客总数
1	40	9	16	65
2	58	4	12	74
…
24	22	86	27	135

数据序号	上行客流	下行客流	层间客流	乘客总数
1	18	9	39	66
2	8	7	37	52
…
24	6	8	55	69