基于电流积分电荷技术的神经网络ă模糊聚类电缆绝缘热老化状态评估模型

In order to realize the aging evaluation of full-size cable under complex factors and multi-parameters, which provides a basis for cable insulation diagnosis and evaluation application, this paper proposes a new evaluation model for thermal aging status of full-size cable insulation on the basis of neural network-fuzzy clustering from current integral charge measurement technique (DCIC–q(t)). Firstly, according to the DCIC–q(t) measurement system, the influences of applied voltage, time and temperature on the electric charge rate, dielectric constant, conductivity and other parameters of cable insulation were studied. Through parameter correlation analysis, it is found that the charge ratio, dielectric constant and conductivity are strongly correlated with temperature and voltage. Then, based on multi-layer parameter learning and adaptive BP neural network model, the mapping relationship between multi-parameter data input and thermal aging time of cables was realized. Finally, fuzzy C-means clustering (FCM) was used to classify the cable aging samples in the neural network model by membership degree and state group. An aging evaluation model of five-layer BP neural network combine with the FCM was established. The convergence speed and precision of BP neural network were improved by optimizing learning rate. The results indicated that the thermal aging status of cables can be divided into four categories: good, slight, moderate and severe. The accuracy of the evaluation results is 92.3%. A strong correlation between the charge ratio, conductivity and the thermal aging degree of cable insulation is observed.