Research On Scheduling Optimization Of Elevator Group Control System For Large Public Buildings

With the continuous advancement of urbanization,high-rise buildings are emerging in cities,and elevators are playing an increasingly critical role in modern architecture.A single elevator or multiple elevators working independently in a building can no longer meet the travel needs of people,and the Elevator Group Control System(EGCS)using advanced technology for unified scheduling of multiple elevators can be applied to the complex elevator operating environment in various large public buildings to better meet the commuting needs of passengers.How to improve passenger transport efficiency while improving passenger ride experience,saving energy and reducing loss are the series of problems urgently needed to be solved for EGCS.This thesis takes the elevator group consisting of multiple elevators installed in a typical large public building in the city as the research object,takes the situation of 20 floors and 4 elevators as an example,and focuses on improving passengers’ use experience,shortening waiting time and riding time,and improving energy waste.The research is carried out from the following three parts.Firstly,aiming at the problem that the elevator traffic flow of large public buildings keeps periodic change in the main period,but the performance of traditional machine learning methods in elevator traffic flow prediction is poor,this thesis proposes an elevator traffic flow prediction method based on long short-term memory network.The method utilizes the long short-term memory neural network and Dropout principle to build a prediction model,and selectively retains the historical information of elevator traffic flow time series,avoiding the problems of over-fitting and gradient disappearance in traditional neural networks.Simulation results show that this method can effectively improve the accuracy of elevator traffic flow prediction.Secondly,the accuracy of elevator traffic pattern recognition is related to whether EGCS can reasonably dispatch ladders.In this problem,Support Vector Machine(SVM)is widely used and the recognition effect is good,but there are still some problems such as unreasonable setting of key parameters and low classification accuracy.Aiming at this situation,this thesis proposes an elevator traffic flow pattern recognition method combining Genetic Algorithm(GA)and SVM.This method uses genetic algorithm to encode the relevant parameters and search the optimal parameter pairs in the space,and then constructs the elevator traffic pattern classification model based on GA-SVM,which effectively improves the identification accuracy of elevator traffic pattern.Thirdly,the elevator scheduling strategy is the core part of EGCS.In this thesis,the PSO-SA hybrid optimization algorithm is proposed based on the introduction of Simulated Annealing(SA)operator on the basis of the original Particle Swarm Optimization(PSO)algorithm,so as to obtain the jump ability to avoid the generation of local optimal solution,and it is applied to the EGCS in large public buildings.Then,the rules of the elevator dispatching system are determined,the system state matrix is designed,and the mathematical models of the average time of passengers taking the elevator,the time of waiting for the elevator and the average energy consumption of the system are summarized.By combining these optimization objectives with different weights,the multi-objective comprehensive evaluation function of the EGCS is constructed.Finally,the performance of several algorithms is compared and verified on the simulation platform of EGCS.Simulation results show the effectiveness of the proposed method in elevator scheduling.