【作者】 李俊芳；

【导师】 宗群；

【作者基本信息】 天津大学 ， 控制理论与控制工程， 2013， 博士

【摘要】 在高层和中高层建筑中，为了提高电梯运行的效率，降低运行能耗，电梯群控系统的调度问题已经引起极大重视。提高电梯运行效率的关键因素之一是事先获取详细的电梯交通流量信息。电梯交通流作为电梯群控系统的调度对象，对群控系统的节能降耗起决定性的影响，电梯交通流研究的主要内容是特性分析和流量预测。本文全面、系统地对三种流量特性进行了详细的定性分析和定量描述，同时从三种流量特性出发分别建立了相应特性的流量预测模型，为电梯群控调度提供了理论依据。主要研究内容及创新点是：（1）首先对电梯交通流的典型特性即周期性和随机性，进行了定性分析和定量描述。其次，重点研究电梯交通流中存在混沌性。通过重构电梯交通流量序列的吸引子进行定性分析；计算数据序列的关联维数和最大李亚谱诺夫指数进行定量分析。结果表明：上、下高峰模式下多种交通流时间序列都呈现出了相似的低维混沌行为特征，即存在混沌现象。本结论将有助于电梯群控系统根据流量混沌特性进行流量预测或调整群控调度策略，从而提高各项性能指标。（2）对办公大楼、教学楼及住宅楼等进行电梯交通流量分析表明，该类大楼内的交通流量具有明显的周期性，并且交通流量时间序列中存在线性和非线性特征数据。提出采用ARIMA模型与GP模型相结合的混合预测方法实现了电梯交通流量的混合预测。仿真结果表明，文中所提混合预测模型较其它单一预测模型都具有较好的预测精度。（3）对医院就诊大楼、机场大楼以及宾馆等建筑物的电梯交通流量分析表明，这类大楼的交通流量具有明显的随机性。提出了采用模糊加权马尔可夫链模型对电梯交通流量交通状态进行预测。预测结果与实际电梯交通进行比较，结果表明所提方法的有效性和可行性。（4）对于不具有周期性且随机性又不明显的不规则流量，已分析其具有的混沌特性。针对具有混沌特性的交通流量，提出了基于支持向量机的电梯交通流量混沌时间序列预测方法，并采用PSO寻优的方法确定了模型的最优参数。分析结果表明，基于SVM的预测模型，相比于其他两种单一的线性模型及非线性模型，更适合于电梯交通流这样的小样本预测，预测精度较高。更多还原

【Abstract】 Group elevator scheduling is important to transportation efficiency for mid-riseand high-rise buildings, and how to improve the service efficiency of elevators hasreceived considerable attention. One important trend to improve elevator systems is tocollect advance traffic information. In this paper, the three characteristics have beenanalyzed and described systemly, and at the same time the three different predictmodels have been founded. It can provide the thoeriotical basis for the elvator groupcontrol system.(1) At first, we have discussed the periodic, stochastic characteristics carefully.Second, The phase space is reconstructed to get the chaotic attractor of the flow. Wethen calculate the correlation dimension and the Largest Lyapunov Exponent aimingat the above data. Based on these analysis and calculation, these results indicate thatlow dimensional chaotic characteristics obviously exist in the up-peak and down-peaktraffic flow data. The result will help to adjust the group control scheduling methodsaccording to the chaotic behaviour of the peak flow so as to save the building energyconsuming.(2) It is shown that there are periodic flow in the offce-building, dormitory, hoteland so on. So, we proposed the autoregressive integrated moving average (ARIMA)model has been one of the most widely used linear models in time series forecasting.Then combining genetic programming (GP), a hybrid forecasting model will be usedfor elevator traffic flow time series which can improve the accuracy. At last,simulations are adopted to demonstrate the advantages of the proposed ARIMA-GPforecasting model than any of the GP and the ARIMA forecasting models separately.(3) There are the randomness in the traffic flow of the hospital building, airportbuilding and hotel. So a new model is proposed to predict the elevator traffic flowstate called fuzzy-weighed Markov model. And the final prediction model canreazlized. The simulation shows the effective of the proposed method.(4) The traffic flow has the chaotic characteristic based on irregular feactures,which has not obvious periodic characteristic or randomness. Aiming at this chaotictraffic flow, the predict model by support vector machine is presented. We havediscussed the influence of the parameters to the predict result, then the parameters canbe optimized by PSO method. Finally, simulations results have shown theeffectiveness and advantages of the proposed SVM forecasting model to predict thesmall samples.更多还原