Acousto-optical Joint Detection System And Method For Partial Discharge Of Electric Equipment

For partial discharge fault state of power equipment,real-time dynamic monitoring is necessary.To deal with this problem in complex strong electromagnetic interference environment,it develops a low-cost,high-performance and portable partial discharge acoustic-optical joint detection system,which is based on the combination of two nonelectrical detection technology: ultrasonic and ultraviolet pulse.As different fault states of partial discharge reflect different development stages,the higher the fault state level is,the closer the breakdown stage of partial discharge is,and inspectors need to take timely and effective maintenance strategies.Considersing many interference factors in partial discharge detection of power equipment,such as low accuracy and low stability recognition caused by sample imbalance and misclassification in some situatuions,it designs a two-stage integrated learning model: Stacking-Bagging,which is superior to the traditional pattern recognition model in the application of high-voltage cables and transformers,including KNN,SVM,RF,XGB and LGB.The designed Stacking-Bagging model can realize high accuracy and high stability identification of partial discharge fault state.The specific work is as follows:(1)Firstly,the research status of partial discharge detection technology is investigated,and the characteristics of each detection method are analyzed.Combined with the practical application requirements of this project,and considering the advantages of joint detection,it proposes a system scheme of joint detection of ultrasonic and ultraviolet pulse.(2)Based on the partial discharge acousto-optic joint detection system technology scheme,the hardware part has completed the sensor selection,the design of acousto-optic signal driver and the selection of the main control module of the system.In the software part,the functions of acousto-optic signal acquisition,fusion processing,real-time storage,display and early warning are designed.And the system interactive display interface is designed.The system designed has the advantages of anti-interference,non-contact,low cost,and portable.Finally,the effectiveness of the system is verified by the field test.(3)Combined with the characteristics of acousto-optic signal,it builds the characteristic engineering of partial discharge signal.First,data preprocessing(outlier detection,denoising,and standardization)was carried out,and then the partial discharge signal features were extracted from the multi-analysis domain angle,and the feature distinguishability was verified through feature selection and multi-view visualization.(4)It analyzes the principle of integrated learning algorithm,and to deal with the problems and difficulties of partial discharge fault state recognition in multiple scenarios,combining the characteristics of partial discharge signals collected by the acousto-optic joint partital discharge detection system,it designs a two-stage integrated learning model: Stacking-Bagging.Then it deacribes how to implement this model,and analyzes the characteristics of the model.(5)Based on filed test,the PD signal database of high-voltage cable and transformer is built,and it compares the recognition result of the two-stage integrated learning model(Stacking-Bagging)with five traditional pattern recognition models,and the stability of the Stacking-Bagging model is analyzed.The experimental results show that the average recognition rate of designed integrated learning model in high voltage cable scene is over 98%,and the average recognition rate in transformer scene is over 99%,which is better than the other five single models,and the recognition stability is the highest,which verifies the high accuracy and high stability of the designed model.