Recursive Neural Networks and its Application in Forecasting the State of Electric Power Equipment

This paper proposes a method based on the recursive neural network rather than the usual BP algorithm. A three-layer BP network structure with input layer, hidden layer, and output layer is used. The inputs are resistance leak current of continuous time sequence, and the values behind are outputs; after the training and learning according to the recursive neural network algorithm, the state forecast of MOA is realized. The result indicates that recursive network is more adapted to the state forecast of MOA.