Application of additional momentum in PID neural network

Abstract

PID neural network (PIDNN) is a new type of feed-forward neural network which has been found by Huai-lin Shu, in which neural network is integrated with PID control law. The priori knowledge of PID control can be used to choose the initial weights to make sure that PIDNN control systems are initial stable. Without priori knowledge and using the random initial weights, initial stable of the PIDNN system may be uncertain. The paper proposes an improved algorithm which has momentum factor to overcome the local minimum of the PIDNN. The distinct improvement of the PIDNN control system is proved by simulation results.