Multi-step Predictions for Generalized Heave Motion of Wave Compensating Platform Based on ELMAN Neural Network

BP algorithm is the typical supervised learning algorithm, so neural network cannot be trained on-line by it. For this reason, a new algorithm(TD-DBP), which was composed of temporal difference (TD) method and dynamic BP algorithm(DBP), was proposed to overcome the restriction. TDDBP algorithm can make Elman network train on-line incrementally. The gradient descent momentum and adaptive learning rate TD-DBP algorithm can improve the training speed and stability of Elman network effectively. Using the collected real time data,the modified TD-DBP algorithm was able to realize direct multi-step predictions for generalized heave motion of wave compensating platform.