PID neural networks in multivariable systems

In this paper, the authors advance some new opinions about neural networks (NNs). A new kind of NN, the PID neural network (PIDNN), is introduced. A PIDNN consists of three layers and its hidden layer consists of P neurons, I neurons and D neurons. The PIDNN´s weights are adjusted by back-propagation algorithms and a PIDNN can be used to control multivariable systems. The authors give the functions of proportional (P) neurons, integral (I) neurons and derivative (D) neurons and present simulation results of a multivariable control system based on the PIDNN.