Development Of The Status Monitoring Device Of Power Communication Equipment Based On Image Recognition

Power dispatching data network is the basis to ensure the safe and stable operation of China’s power system.Abnormal operation of power communication equipment will cause communication business interruption,which will bring great economic losses to plants and users.At present,there are three problems in the operation status monitoring of vertical encryption authentication gateway,router and switch devices:(1)The superior dispatch can only locate abnormal sites and devices,and the plant station will send personnel to the site to check after receiving the feedback.(2)The new energy plant is remote and unattended,and the daily patrol inspection still depends on manual completion.(3)Some plant and station personnel have little knowledge about the running state of communication equipment and cannot make timely and accurate judgments.Based on the real-time requirements of the power dispatching data network,in order to assist the staff to monitor the running state of the equipment and shorten the time for troubleshooting,this topic applies the image recognition technology to the running state monitoring of the power communication equipment,and realizes the image acquisition,processing and analysis on the hardware device.And through the way of building a server to upload the monitoring results for the staff in the remote computer real-time view.In view of the operation status identification and monitoring device development of these three kinds of equipment,the main content of this research is as follows:(1)The equipment images obtained from the site were enhanced to solve the problem of insufficient data sets.Based on the same number of samples,the accuracy of the three methods is compared,and the conclusion is drawn that the three methods can not meet the requirements of positioning devices.Then,the template matching method is used to identify and detect the three kinds of equipment respectively,in which the switch has a very low recognition accuracy due to the occlusion of network cables.Finally,the YOLO v3 model based on convolutional neural network training data set samples were selected to learn a large number of sample features,and then the model training parameters were adjusted according to the change of loss function,and the detected target position information was output with the recognition accuracy of98.89%,98.01% and 97.32%.(2)According to the fixed position of the indicator light on the equipment panel,the indicator light area is intercepted to reduce the size of the image to be processed and reduce the amount of calculation.After switching to HSV color space,set the value range of red,green and yellow indicator light colors,select specific colors and find contour for counting.An array is defined for each of the three devices to store the counting results.The length of the array is determined by the flashing period of the indicator light.By referring to the device manual,the corresponding relationship between the indicator light and the device state is clarified.Then,according to the changes in the stored values in the array,the different states of the device are matched and the results are output..(3)After the embedded operating system is installed on the hardware platform,the device identification and indicator light detection program is transplanted to the device,and the relevant dependencies(OpenCV and Flask)required by the program are configured in the python3 environment.Flask framework is used to build a Web server,and HTML files are used to design the title and form of the browser page.Finally,the URL is set to upload the identification results into the form.After starting the server program,access the set IP port with the browser to open the monitoring page,and the device can run stably and accurately identify the running state of the device.