Cooperative Event-Triggered Fuzzy-Neural Multivariable Control With Multitask Learning for Municipal Solid Waste Incineration Process

Abstract

Municipal solid waste incineration (MSWI) is an industrial process with multiple mechanism reactions, which has strong coupling and time-varying dynamics. It is extremely difficult to design a feasible multivariable controller for MSWI process due to the complex composition of municipal solid waste and fluctuation of calorific value. To solve these problems, a cooperative event-triggered fuzzy-neural multivariable controller with multi-task learning (CETFNMC-MTL) is proposed to realize the adaptive multivariable control of MSWI process. First, a fuzzy-neural multivariable controller is established to control furnace temperature and oxygen content synchronously. Second, a dynamic self-organizing mechanism based on multi-task learning is designed, which splits and merges neurons by calculating the dynamic time warping distance and cumulative contribution of neurons in the continuous time. Third, a cooperative event-triggered mechanism is introduced to improve controller update efficiency while reducing mechanical wear and computational burden. Then, the stability of parameters learning and structure self-organizing process is analyzed to guarantee the successful application of CETFNMC-MTL. Finally, the effectiveness of the controller is tested with process data from an MSWI plant in Beijing, China. The results show that the proposed CETFNMC-MTL has adaptive learning ability, while reducing energy consumption and improving control accuracy.