Research On Elevator Group Control Algorithm Based On Adam Optimized Neural

Today,with the rapid development of intelligent technology,a large number of scholars have conducted in-depth research on the control algorithm of elevator group control system,considering the multi-objective comprehensive optimization problem,and proposed an elevator group control algorithm based on neural network.Aiming at the problems of poor convergence and low prediction accuracy of elevator group control system models based on traditional neural networks,an optimized neural network elevator group control algorithm is proposed to realize the rational allocation of elevator resources.The main research contents are as follows:(1)According to the structure diagram of the elevator group control system,analyze the composition and function of the modules in the elevator group control system,study different passenger flow modes and core module scheduling algorithm strategies,and summarize the advantages and disadvantages of different scheduling strategies in the group control system.Lay the foundation for subsequent research and analysis.(2)Study the importance of various performance indicators of the elevator group control system in multi-objective control,consider multiple factors,and design a comprehensive evaluation function of satisfaction.Then,build a neural network-based elevator group control model,combined with a comprehensive evaluation function,to train and test the model.The results show that the traditional neural network has unsatisfactory convergence speed,and it is difficult to deviate from the local extreme value,and the satisfaction evaluation value cannot be obtained quickly and accurately by the traditional neural network.(3)Aiming at the shortcomings of long training time and easy to fall into local extremes in traditional neural networks,an elevator group control algorithm based on Adam optimized neural network is proposed.Simulation experiments can show that the Adam algorithm improves the problems of long training time and low network prediction accuracy in the neural network model,and obtains faster convergence speed and better convergence performance.However,during the training process,there was an overfitting phenomenon that the effect of the training set was very good,but the effect of the test set was not good.Therefore,the Dropout method was introduced in the network training process to alleviate the occurrence of overfitting,thereby improving the neural network model to achieve the expected goal.(4)Build an elevator group control simulation system model in Py Charm to simulate and verify the algorithm proposed in this paper based on the Adam optimized neural network.The results show that compared with the traditional neural network model,the optimized model has reduced passenger boarding time and waiting time.Among them,the system energy consumption index represented by the number of elevator stops is the most obvious.The traditional neural network algorithm the average number of traffic stops under three modes about 294 times,and the optimized scheduling algorithm 230 times the average number of stops,compared to a 27.8% decrease.It can be concluded that the feasibility and superiority of Adam’s optimized neural network algorithm in the field of elevator group control indicate the significance of Adam’s algorithm for the application of elevator group control technology.