Research On Traffic Configuring And Optimal Dispatching Of Elevator Vertical Traffic System

The characteristics of the fast rhyme of the modern life, the high efficiency requirements of modern business transactions and the diversities function of the building service make the vertical passenger traffic undergo a potential change. The classical methods of the traffic design, analysis, forecast and elevator group dispatching of elevator traffic system, are facing unprecedented challenges. So, the research on traffic configuring and optimal dispatching of elevator traffic system are of important theory significance with enormous potential economic benefit.In this dissertation, taking the elevator vertical traffic system in high rise buildings as the background, the relatively thorough and systematic study of the problems related with which including the problems of elevator traffic configuring, the elevator traffic demand forecasting, elevator traffic pattern recognition and the optimal dispatching of elevator group are conducted respectively in details. The main studies and obtained results in this dissertation are as follows:1. In view of the up peak traffic condition, the performance indices of elevator traffic system are analyzed and based on which the bi-level programming mathematical mode with multiple-objective of elevator banking in high rise buildings is established; A method employing genetic algorithm to solve a kind of multi-objective question is proposed, in which the objective function is optimized by globally optimizing the decision variables. The presented approach overcomes the drawback of the conventional method of determining weight value by preference in dealing with linear combination weighted method; The final optimal configuring scheme of elevator banking is achieved by analyzing the cost caused by the layout of elevator banks in the given high rise building for each obtained optimal solution, which provides the basic conditions for implementing the optimal dispatching of the elevator group control system.2. Aiming at the properties of time-variant and uncertain of elevator traffic demand in high rise buildings, the prediction model based on the LS-SVM (Least Squares Support Vector Machine) with iterative learning is established, the proposed model is independent of the distribution of traffic demand and it can dynamically track the varying regularity of traffic demand. Mean while, the continuous forecasting function of traffic demand is obtained by employing the LS-SVM; The inspection results show that the new model has a fairly ideal prediction effect which lay a foundation for establishing the efficient scheme of recognizing elevator traffic patterns.3. Based on the reliable prediction of elevator traffic demand, the method of establishing schemes for recognizing elevator traffic pattern by using the proposed concepts of critical points and the critical time ranges of elevator traffic pattern is presented. Two schemes for recognizing elevator traffic patterns are conducted in this dissertation, one is the filter function based method of pattern recognition with very important reference value in theory, and the other is the least squares support vector machine based method with better engineering significance in which the concept of critical points of elevator traffic pattern are extended to the critical time ranges which can be determined according to the dynamical property of dispatching strategy switching.4. In order to efficiently implement the optimal dispatching of elevator group, the Origin and Destination distribution ("O-D distribution" for short) of passengers traffic under different traffic patterns is needed to drive elevator group control system. In this dissertation, a prediction method of passengers' O-D matrix is proposed where the predicted information of O-D distribution is taken as the prior traffic demand information employed to conduct the optimal dispatching of elevator group. The novel method employs the merits of both grey theory and neural network by introducing the grey forecasting technique to RBF neural network to construct GM-RBFNN. By conducting the conversion of the accumulated generating operation (AGO) on the initial observed traffic data, the sample data which are exponentially distributed for modeling and training GM-RBFNN are obtained; Meanwhile, a method of modifying abnormal initial observed traffic data is presented to further reduce the randomness of observed traffic data. The proposed method not only avoid the theoretical error of grey model, but enhanced both the training speed greatly and the precision of prediction of the neural network. The predicted O-D distribution including the actual traffic information provides the important traffic data for the optimal dispatching of elevator group control system for a class of high-rise buildings.5. In order to realize the optimal dispatching of elevator group control system, in this dissertationseveral, multi-assessment indices are considered including average waiting time (AWT), average riding time (ART), average service time (AST), the average number of stops(ANS) and the average long waiting percent (LWP), etc, based on which the method of constructing multi-objective function is conducted. To solve the multi-objective optimal dispatching problem of elevator group , an improved particle swarm optimization algorithm(IPSO, Improved PSO) is proposed. The optimal dispatching of elevator group with multi-objective under the main traffic patterns is realized by finding the optimum Hamilton cycles based on minimizing the multi-objective function. So the system performance and the comprehensive service level are efficiently improved.