MIMO modeling and multi-loop control based on neural network for municipal solid waste incineration

The Municipal solid wastes incineration (MSWI) process with complex mechanism and strong nonlinearity is an important part of resource cycle. It is a challenge to design its multivariable controlled object model and control strategy. To solve this problem, a multi-input multi-output (MIMO) data-driven model and a multi-loop PID controller are proposed in this paper. Firstly, a feature selection method based on Pearson correlation coefficient (PCC) and expert knowledge is used to analyze the relationship between the manipulated variable and the controlled variable of MSWI process. Secondly, a MIMO Takagi-Sugeno fuzzy neural network (TSFNN) based on multi-task learning (MTL) is designed to construct the multivariable controlled object model. Thirdly, a multi-loop PID controller based on quasi-diagonal recurrent neural network (QDRNN) is constructed, which has self-feedback channel and interconnection channel, and can adjust the control parameters automatically. Next, the stability of control strategy is proved by Lyapunov second method. Finally, the modeling effect and control performance are confirmed on the simulation experiments based on the real MSWI process data.