Design of an improved PID neural network controller based on particle swarm optimazation

A control algorithm of basic PID neural network (PIDNN) is introduced. A set of PID parameters is introduced into PIDNN to improve the processing ability of Proportional, Integral and Differential neurons. An additional momentum is also introduced into PIDNN to improve the learning efficiency of neural networks and to avoid trapping into local optimum. Also, an enhanced particle swarm optimization with adaptive variation operator and linear decreasing inertia weight is used to optimize the initial weights of PIDNN. A 3-in-3-out nonlinear coupling system is used in simulation to validate the proposed algorithm. The simulation result proves that this method has better dynamic and static characteristics than the original algorithm.