Research On Prediction Method Of Polypropylene Cable Insulation Life Based On Accelerated Test And LMBP Neural Network

The new environmentally friendly polypropylene(PP)main insulated cable is gradually developing into an important part of power cables,and its service life is affected by various aging factors.At present,the research method for cable insulation life is mainly electrothermal aging test,selecting an appropriate end of life and taking the physical and chemical performance degradation reflected in the aging process of the material as the basis for evaluating the life.However,the conventional aging test has the problems of long time,cumbersome process,and large dispersion of the obtained test data.In recent years,the research on intelligent algorithms by scholars in related fields has confirmed that artificial neural network(ANN)can well fit the life test data with large dispersion.The unique radial approximation global optimization ability of ANN is used for the life prediction of cable insulation.provides a new directionIn this thesis,two types of PP cable materials are used as raw materials for testing and comparative analysis.Differential scanning calorimeter is used for thermo-oxidative aging test to study the change characteristics of PP’s oxidation induction period under high thermal stress conditions.The electrical aging characteristics of PP were studied in the step-by-step boost test;Python programming was used to process the data of the electrical aging test;based on the data of the aging test,Matlab was used to establish the LMBP neural network life prediction model,and Bayesian regularization was used The rule further optimizes its weight correction process,and obtains a more practical PP insulation life prediction model,which provides a reference for the life evaluation of polypropylene insulated cables.The corresponding oxidation induction period of PP at different temperatures was obtained through the thermo-oxidative aging test,and the temperature-life model of the material was extrapolated based on the Arrhenius formula.The reliability of polypropylene insulation is obtained through constant voltage,step-by-step,step-by-step acceleration test to obtain polypropylene aging data,and the life index n value of PP6 polypropylene insulating material is 12.5 and PPS is 12.9.The least square method is used to match the aging data.The reliability of the n values obtained by the three accelerated electrical aging tests was verified,and the electrical aging life model of polypropylene was determined and the corresponding life prediction curve was obtained.Lower than PP6,it is inferred that the two PP cable materials have a threshold field strength and PP6 type polypropylene is more suitable for higher working field strength;based on the BP neural network,the parameters affecting the insulation aging of polypropylene are extracted as feature quantities,and LM is used.The algorithm improves the network accuracy and establishes a three-layer neural network,and determines the appropriate modeling parameters.After optimizing the weight correction process of LMBP neural network by Bayesian regularization,the network convergence speed is improved,the output error is reduced,the generalization ability is stronger,and the prediction value has higher accuracy.The results show that the network is more suitable for prediction under lower working field strength,and the fitting correlation coefficient of the prediction results reaches 0.98.Comparing the predicted life of the network with the accelerated test results,it is found that the predicted value of the network is greater than the life evaluated by the accelerated electrical aging test.It is believed that because the neural network takes the insulation thickness as a feature quantity to participate in the life evaluation,the impurities,defects and pores in the insulation materials with different thicknesses are different.Realistic.