Multi-model predictive function control based on neural network and its application to the coordinated control system of power plants

Steam power plant configuration, design, and control

“…GPC is then designed on the basis of the global model to achieve a wide range control of the power plant, as shown in Figure . Based on the Radial Basis Function (RBF) neural network and Adaptive Neuro‐Fuzzy Inference System (ANFIS) approaches, local linear MPCs developed at different loading points are combined for the coordinated control of a 500 MW plant . Although the multi‐model strategy seems more complicated than the direct nonlinear approaches, in fact, it is easier and more efficient to implement.…”

“…50 Because the advances in linear modeling and control theory can be directly taken into account, the multi-model techniques bring an alternative way to handle the nonlinearity, and its integration with the MPC approach has been shown to be effective to control the power plants. 49,[51][52][53][54][55][56][57][58][59][60][61][62][63][64][65][66] The earliest multi-model MPC used in power plant control is presented in Ref 51, where networks of dynamic local linear models are created after dividing the whole operating region into a number of zones and the global model is built by using the interpolation among these local models. GPC is then designed on the basis of the global model to achieve a wide range control of the power plant, as shown in Figure 15.…”

“…Based on the Radial Basis Function (RBF) neural network and Adaptive Neuro-Fuzzy Inference System (ANFIS) approaches, local linear MPCs developed at different loading points are combined for the coordinated control of a 500 MW plant. 52,53 Although the multi-model strategy seems more complicated than the direct nonlinear approaches, in fact, it is easier and more efficient to implement. To improve the wide-range operation of a boiler-turbine system, a fuzzy interpolated DMC is proposed in Ref 54, which can be viewed as an extension of the linear DMC in Ref 35. In Refs 49,55,56, GPCs are developed on the basis of neuro-fuzzy networks (NFNs), both the local linear model weighted and controller weighted approaches are given in these works.…”

“…The essential idea of the multi‐model technique is ‘divide and conquer’, which uses a combination of several linear models to approximate the nonlinear behavior of the plant . Because the advances in linear modeling and control theory can be directly taken into account, the multi‐model techniques bring an alternative way to handle the nonlinearity, and its integration with the MPC approach has been shown to be effective to control the power plants …”