模糊神经网络技术在电梯群调度中的应用
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
随着国内建筑智能大厦的兴起和人们对电梯的服务质量要求越来越高,电梯群的优化调度已成为急需解决的问题。在电梯群调度上已提出了模糊控制法、神经网络法等多种应用到实际中去的方法。因电梯群调度的非线性、随机性和模糊性等特点,这些方法总是存在不同的缺点。将模糊推理和神经网络相结合的模糊神经网络技术,在处理非线性、随机性和模糊性等问题上有很大的优势,所以将模糊神经网络技术应用到电梯群调度上,成为当前国内研究热门。但是,目前很少有文章介绍模糊神经网络技术在解决电梯群调度中的选梯问题上的研究。为此,主要就这方面进行一些初步的探索。
     要实现调度,必须先实现上位机与PLC之间的串行通信,这是因为控制电梯的PLC运算能力不强、交互性差等缺点,必须在上位机端实现复杂调度运算和电梯运行模拟,所以要通过串行通信采集PLC中各电梯运行数据、分配响应电梯等。为此,详细分析了串行通信在电梯群调度中的特点,如:通信参数设置、单方通信编程、通信的实时性和可靠性要求。并根据FPI型PLC的MEWTOCOL—COM通信协议,采用并行多通信模块的网络结构,用VC通信控件实现了多线程高效通信。
     文章分析并选择了电梯群调度通常采用的外呼——分配策略的调度算法,它是以平均等待时间(AWT)、长时间等待率(LWP)、能量消耗(RNC)为评价标准的评价函数:
     Si--Wi*AWTi+W2*LWPiV3*RNCi 这三个电梯调度评价标准值是函数的关键,也是模糊神经网络方法要解决的核心对象。
     根据模糊控制方法中所采用的求解三个评价标准的模糊规则知识和电梯调度中响应层站召唤信号延时小等实际要求,新构建了三个分别求解评价标准值的四层模糊神经网络,实现了AWT、LWP、RNC的规则到神经网络的映射,网络具有结构简单、计算量小、运算速度快的特点,其中的模糊节点采用高斯型隶属度函数,该函数的中心值和宽度值对网络的计算输出和性能有很大影响,必须学习修正。因此详细地推导了新建的模糊神经网络的BP学习算法,为了避免局部最小和加快平均误差函数收敛,采用了加动量项BP算法,为保证网络每层权值的物理意义和改善规则之间的重要性,选用了部分层学习算法,即仅对网络第三、四层间的权值w_i进行学习修正。这样,BP算法只针对中心值、宽度值和权值w_i进行学习,并将训练好的参数保存在数据文件中,供调度使用。
     以训练成功的模糊神经网络实现的调度算法为核心,研究了电梯运行的模拟问题。在保证电梯运行的实质前提下,详细地讲述了对电梯运行信号的数量、各网络输入变量预处理等进行简化的内容,成功地模拟电梯的运行。仿真实现了电梯群调度的整个过程:当有新的召唤信号时,先要通过训练好的网络计算出各电梯的AWT_i、LWP_i和RNC_i;从而可以计算出各电梯对应的S_i;此时选择对应最大S_i值的电梯作为响应本次召唤信号的电梯,相应的更新电梯动画并将该信息发送给控制该电梯的PLC,就完成了一次调度过程。
     在实验室提供的FPI—C40型PLC的基础上,用VC开发了电梯群调度的仿真软件。初步探索了电梯群调度的模糊神经网络方法,基本说明了该方法的有效性。
With more intelligent buildings ware constructed and better service was need, the improvement of Elevator Group Dispatch has become an emergent problem. Many methods such as Fuzzy Control Approach and Neural Network Approach are already used in elevator group dispatch. Because of non-linear, random and ambiguous features of the elevator group dispatch, there are various drawbacks in these methods. Combining Fuzzy Inference with Neural Neuvork. the fuzzy neural network technology has great superiority in processing non-linear, random and ambiguous problems. So the application of fuzzy neural network in elevator group dispatch becomes one of focus in the field in home. But there is few theses to introduce fuzzy neural network technology about selecting elevator problem in the elevator group dispatch. So the thesis makes a primitive stud}' in this field.
    To realize dispatch, the Serial Communication between Epigny Computer and PLC must be first achieved. Due to the poor calculation capability and mutuality of the PLC which controls elevators, the Epigny Computer must realize the complicated dispatch calculation and the simulation of the elevator moving, collect the running data of the elevators and allocate responding-elevator. Therefore, the feature of the serial communication in the elevator group dispatch are analyzed in the thesis, such as communication parameter setting, one-side communication programming and real-time ability and reliability of communication. According to the MEWTOCOL-COM communication protocol of FP1-PLC, we construct parallel multi-communication module network structure and use VC to realize multi-threading high effective communication.
    The Hall Call Assignment Strategy that is usually used in the elevator group dispatch is chosen and analyzed. It is an Evaluation Function that takes Average Waiting Time (AWT), Long Waiting Probabilities (LWP), Power Consumptions (RNC) as Evaluation Standards:
    The three values are the key to the function and the core problem of fuzzy neural network approach's targets.
    According to the Fuzzy Rules in the fuzzy control approach that calculate the three evaluation standard values, and the limited lasting-time of responding hall call in the elevator dispatch, we construct three four-leveled new fuzzy neural network calculating the evaluation criteria values. The network with character of simple structure, small calculation and rapid arithmetic speed, realizes the reflection from the rules ( AWT. LWP and RNC) to the neural network. Its fuzzy node adopts Gauss Model Subjection Function, whose center-value and width-value exerts great influence on the network's operation transmission and performance, and needs modification by studying. Then the BP Algorithm is inferred in detail and adding momentum item is used to avoid local least and speed up die convergence of Average Error Function. To guarantee the physical meaning of the area-value and improve the significance among rules, we choose part-layer algorithm to modify the area-value between the third level and the forth one. Thus the BP algorithm only study die center-value, width-value and w, values. The improved parameters are saved in the
    
    
    
    data file for dispatch.
    By taking the schedule operation realized by the successful fuzzy neural network as its core, study the problem of elevator moving simulation. On the premise guaranteeing the matter of elevator moving, state in details the simplicity of the numbers of elevator moving signals and the pre-processing of the network's inputting-variable. Successfully simulate and achieve the elevator dispatch procession: A, when an new hall call signal is found, calculating the AWT, , LWP, and RNC, through the network; B, getting the S, values of each elevator; then selecting the elevator with the biggest S, value to respond the new hall call signal; C, refreshing elevator cartoons and sending message to the PLC which controls the selected elevator.
    With FP1-C40 PLC in lab, develop software to simulate elevator group dispatch by using VC. In a word, the thesi
引文
[1] 刘剑、梁延懂、李玉珍等,《电梯运行的神经网络控制》,沈阳建筑工程学院学报,1999,15(3):289~292。
    [2] 万健如、蔡昱、李银惠,《群控电梯模糊与神经网络控制》,电气自动化,1999,4:21~24。
    [3] 蔡昱、万健如,《电梯模糊与神经网络控制》,中国电梯,1999,2:34~38。
    [4] 张苗苗、谢剑英,《电梯智能群控系统的面向对象分析和研制》,微电子学与计算机,2000,2:51~54。
    [5] 刘载文等著,《电梯控制系统》,电子工业出版社,1996年第一版。
    [6] 邱公伟主编,《可编程控制器网络通信及应用》,清华大学出版社,2000年3月第一版。
    [7] 常斗南主编,《可编程控制器原理、应用、实验》,机械工业出版社,1998年7月第一版。
    [8] 宋海生、单根立、任有志、陈继荣,《上位机和PLC串行通信的程序设计》,计算机应用,2000,20(1):67~68。
    [9] 吴晓滨,《电梯群控系统的通信技术》,中国电梯,2000,8:22~23、31
    [10] 苏玉北、王莉娜、黄天锡、种衍文,《计算机32位串行数据通信接口的设计与实现》,计算机工程与应用,1999,12:59~62。
    [11] 王霞,《Windows多线程应用程序设计技巧》,现代计算机,1999,4:24~27。
    [12] 李慧娟、胡伟华、彭超、谢剑英,《基于领域知识的电梯群控系统的智能滚动优化方法》,电子技术应用,1999,4:26~29。
    [13] 张学军、张苗苗、谢剑英,《面向对象的电梯控制系统建模和电梯群控策略》,系统仿真学报,1999,11(4):269~272。
    [14] 张苗苗、张学军、谢剑英,《基于模糊推理的电梯群控系统的研究与仿真实现》,测控技术,2000,19(3):56~59。
    [15] 梁春艳、谢剑英,《智能大厦中的电梯群控系统研究》,测控技术,2000,19(30):16~18。
    [16] Chang Bum Kim, Kyoung A. seong, etc, A Fuzzy approach to elevator group control system, IEEE Trans Syst., Man, Cybern., 1995; 25(b): 985~990
    [17] Seiji, etc, Supervisory control for elevator group by using fuzzy expert system, Proceeding of the IEEE International Conference on Industrial Technology. Guangzhou. 1994
    [18] Albert T. P. So, Janson K. L. YU, etc, Dynamic Zoning Based Supervisory Control for Elevator, Proceedings of the 1999 IEEE International Conference on Control Application, 1999; 8: 1591~1596
    [19] 张学军、张苗苗、谢剑英,《电梯控制系统的面向对象分析与实现》,测控技术,1999,18(12):29~31。
    [20] 闻新、宋屹、周露,《模糊系统和神经网络的融合技术》,系统工程与电子技术,1999,21(5):23~26。
    [21] 姜智峰、刘泽民,《一种新型模糊神经网络结构确定的研究》,电路与系统学报,1997,2(1):173~178。
    [22] 宗群、尚晓光、岳有军、雷小锋,《电梯群控系统虚拟仿真环境设计》,制造业自动化,1999,21(5):24~25、31。
    
    
    [23] 王维新、常本康,《群控电梯系统的算法设计》,电子工程师,1999,8:21~23。
    [24] 吴晓滨,《电梯群控系统呼叫预想到达时间的计算》,中国电梯,1999,5:24~26。
    [25] 吴晓滨,《电梯群控系统软件模块划分与功能分析》,中国电梯,1999,12:38~40。
    [26] 邹北骥,《计算机动画与物体真实运动之关系探索》,湖南大学学报(自然科学版),1999,26(1):98~31。
    [27] 朱晓华、章玉鉴,《用ActiveX控件构造虚拟仪器》,计算机应用,1998,18(12):21~23。
    [28] Richard D P. Simulation for Control System Design and Traffic Analysis, Barney G C. ed. Elevator Technology 9,IAEE, 1998:226~235
    [29] Ronald J.Norman, Object-oriented system analysis and design, Prentice-hall international, inc., 1998.6
    [30] Edward Yourdon&Carl Argila, Case studies in object-oriented analysis & design, Publishing house of electronics industry, 1998.6
    [31] 李兴兰、张友根、陶以政、潘振显,《基于RS—485/422网络的远程数据传输系统》,电子技术,2001,2:10~13。
    [32] 王士同编著,《神经模糊系统及其应用》,北京航空航天大学出版社,1998年6月第一版。
    [33] 张乃尧、阎平凡编著,《模糊神经网络与模糊控制》,清华大学出版社,1998年10月第一版。
    [34] 王士同编著,《模糊系统、模糊神经网络及应用程序设计》,上海科学技术文献出版社,1998年12月第一版。
    [35] 李士勇编著,《模糊控制、神经控制和智能控制论》,哈尔滨工业大学出版社,1998年9月第二版。
    [36] 张福恩等著,《电梯制造与安装安全规范应用手册》,机械工业出版社,1994年3月第一版。
    [37] (美)Michael J. Yong著(邱仲潘等译),《Visual C++6 从入门到精通》,电子工业出版社,1999年1月第一版。

© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号

地址:北京市海淀区学院路29号 邮编:100083

电话:办公室:(+86 10)66554848;文献借阅、咨询服务、科技查新:66554700