Research And Design On Elevator Group Control Policy With Destination Registration

In high-rise buildings, elevator is an indispensable vertical transportation equipment, and quality of elevator service is directly related to the manufacture and daily life. With the rapid development of the elevator control technology, many advanced and efficient ideas have been introduced into the elevator group control technology in the past, new requirements are also proposed about more intelligent, networked, energy saving. Security stability and efficiency of the elevator group control technology is greatly improved. But there are still some problems unsolved.This thesis optimized four kinds of typical passenger flow model of traffic mode based on dynamic characteristic of the group control system with destination, the type of elevator traffic model, performance evaluation indexes and part of the general theory of intelligent algorithm. The BP Neural Network was used on the traffic pattern recognition. The data that generated from the simulation of passenger flow model is the data source of traffic pattern recognition and cluster-car dispatching validation. Multi-object dispatching strategy of the elevator group control with destination registration was studied. A multi-objective decision model was established, which has taken the passenger, elevator service and energy consumption into consideration. A cluster-car dispatching method of medium passenger flow was designed.Firstly, A multi-objective policy of group control was overall designed, the function of the group control system with destination is divided into three modules:data management module, the traffic pattern recognition module, decision module of group control dispatching. And each module functions were analyzed in detail. Secondly, the passenger flow model was optimized, the data which generated from the simulation of passenger flow model was used as training samples, while a five-input-four-output of BP neural network was established,which was used for the traffic pattern recognition. The initial value of learning rate using gradient descent with momentum and self-adaption, the number of training and training error were obtained after analyzing the data. It can train the network and identify traffic patterns well. Finally, this thesis describes the advantages of passengers clustering in groups as a unit to boarding the car. And a multi-objective decision model was established based on the fastest return of the elevator. This model can be transformed into 0-1 programming problems, thereby optimal clustering group was selected. The model proposed that the waiting time was divided into a fixed waiting time and dynamic waiting time. And it was treated as indicators. Integrated call effect time and the total running time of one-way were the objective function. Traffic flow of simulation obtained perfect results, considering the model of passenger satisfaction index, elevator service indicators and energy consumption indicators.