Application of Improved BP Neural Network for Wire Network Signal Prediction

It is difficult to measure wire net signal of high voltage power transmission lines accurately for power system in the process of monitoring and repairing. In order to solve this problem, the signal measured in remote substation or laboratory is employed to make multipoint prediction to gain the needed data, and then the predicted data is send to work-field via wireless network GPRS. The needed precise data may be computed based on current time and received data in work-field, which may provide reliably basis for fault diagnosis of air bracket high voltage power transmission lines. Here, BP neural network is employed to identify and predict the needed time signals. In order to overcome the shortcoming of general BP net that convergence speed is slow and plunge local extremum easily, the improved two stages learning and training method of BP network is designed to predict signal. Experiment results show that the designed BP network has good predicting ability. The designed system offers accurate data for the monitoring and fault diagnosis of high voltage power transmission lines.