Research On The Detection Method Of Electric Vehicle In-home Charging Based On Edge Computing

As an important part of building smart buildings,fire safety has become a research hotspot in the field of building intelligence.With the increasing popularity of electric vehicles,limited charging places can no longer meet the increasing demand of people.Many people choose to bring electric vehicles into their homes for charging.Because electric vehicles are prone to explosion when charging,and the main parts of electric vehicles are easily accessible and combustible materials.Once a fire occurs,it will pose a huge threat to the lives and property of itself and other residents in the community.This paper takes the deep learning target detection technology as the core,combined with the fast-developing edge computing technology.This paper proposes an edge intelligent electric vehicle in-home charging detection method,which can quickly and accurately detect the illegal behavior of electric vehicle in-home charging.It is of great significance to ensure the safety of buildings.The main research work of this paper:(1)An improved Mobile Net-SSD algorithm is proposed to identify electric vehicle targets in elevators.Aiming at the problem of excessive calculation of Mobile Net-SSD,hyperparameters are added to reduce the amount of model parameters,making the model more suitable for embedded devices.In order to prevent the gradient from disappearing,LRe LU is used as the loss function of the model.Aiming at the problem of insufficient data set,it is proposed to use Cycle GAN data enhancement methods to expand the data set to improve the generalization ability of the model.During the experiment,the BOHB(Bayesian Optimization and Hyperband)method is used to optimize the hyperparameters adaptively according to the self-made elevator data set in this paper.Experimental results show that the improved Mobile Net-SSD algorithm can increase the m AP of the traditional SSD algorithm by 6% and reduce the memory usage by 30%,only 41% as the amount of original calculation of Mobile Net-SSD.(2)Aiming at the problem that some residents choose to disassemble the battery and take it home for charging,a non-intrusive battery charging detection method based on TinyYOLOv4 is proposed.This method collects electrical power data through non-intrusive load detection equipment,and transmits the data to the Cambrian development board via Ethernet.The development board calls the Tiny-YOLOv4 algorithm to identify different electrical waveforms and realizes battery charging event detection.If the battery charging behavior is detected,the power meter will be powered off through the development board control relay.(3)Design and develop an electric vehicle detection system model in an elevator based on edge computing.The hardware development environment of this system model is based on the Cambrian 1H8 development board,and the software framework adopts the edge intelligence platform.The system captures images in the elevator through the camera,and calls the electric vehicle detection algorithm in the elevator deployed on the Cambrian 1H8 development board,aiming to realize real-time detection of the electric vehicle target in the elevator.The administrator monitors the real-time detection results of the video stream in the elevator through the VLC player.Once the electric vehicle target is detected,the system controls the closing of the elevator door through the relay module,and calls the horn module to play alarm audio to prevent the electric vehicle from entering the elevator.Figure[54] table[8] reference[72]...