Operating Environment And Status Awareness System Of Secondary Equipment In Power Distribution Room

The environment in the power distribution room will be critical to the normal and stable operation of the power distribution equipment.Therefore,accurate and timely grasp of the environment and equipment operating status information in the power distribution room is of great significance to the prediction of the safe operation of the power distribution network.The existing monitoring methods of the power distribution room are mostly traditional methods,the application of artificial intelligence technology is less,and the intelligence of the power distribution room is low.In order to adapt to the development of smart grid,based on deep neural network technology,the paper developed a set of monitoring system for power distribution room that integrates collection,calculation and intelligent identification,and successfully realized the intelligent identification of analog quantity and image fault.The main work of the thesis is as follows:The article first designed the hardware circuit for collecting the environment,current,voltage and other analog quantities of the power distribution room,as well as the state image of the pressure plate and the smart meter,and studied the related filtering processing algorithm to complete the preprocessing of the collected signal.In order to improve the recognition rate of extracting equipment fault types from analog signals such as current and voltage,the paper designs a hybrid neural network.The hybrid neural network is composed of two parts: a fully connected neural network and a convolutional neural network.The convolutional neural network is composed of a Res Net network embedded with an attention mechanism.The network uses the analog signal features extracted by the fully connected neural network and the features extracted by the convolutional neural network from the analog signal time-frequency diagram to fuse,and finally realizes the equipment fault classification.The recognition rate of the hybrid neural network for equipment faults is increased by 10% compared to only using the fully connected neural network to extract the features of analog signals,and the recognition rate of only using the convolutional neural network to extract the features of the time-frequency image is increased by 5%.Aiming at the detection of the pressure plate and the recognition of the text displayed by the electric meter,the feature extraction network in the original SSD network structure is replaced with the Res Net network embedded with the attention mechanism,and the method of transfer learning and data enhancement is used to solve the insufficient data set.The problem;in addition,in order to improve the accuracy and accurately identify the state where the press plate is not in place,the bilinear finegrained recognition link is introduced into the improved SSD network.After two recognition and improvement processes,the improved deep neural network controls the press plate.The recognition accuracy rate reached 98.5%,and the improved network structure accuracy rate,specificity,recall rate and other model indicators performed well.Finally,in view of the difference between the displayed text of the electricity meter and the flat and long text characters,before recognizing the text characters,Laplacian gradient change processing is performed on the electricity meter picture.According to the text characteristics of the electricity meter,the default aspect ratio of the default frame in the SSD network is modified.It realizes the accurate recognition of the characters of the electric meter fault.At the end of the thesis,based on the Lab VIEW development platform,a set of power distribution room management system that can identify the data,equipment status,and text in the power distribution room is developed,and the functional test of the system is completed.The test results show that the various functions of the system It meets the design requirements and has good engineering practical value.