Dynamic modelling and control for multivariable output distribution systems

Following the recently developed strategies for the modelling and control for multi-input multi-output (MIMO) distribution control systems, this paper presents a dynamic model for the control of combined probability density function. A multilayer perceptron (MLP) neural network is used to approximate the probability density function (PDF) of the systems and a high-order dynamic nonlinear state-space model is obtained. Then nonlinear principal component analysis (NLPCA) is adopted to reduce the obtained model order and a lower-order time-varying system is achieved. The controller design for the lower-order time-varying system is given. The effectiveness of the result has been demonstrated by two examples in the end.