Diagonal Recurrent Neural Networks with Application to Multivariable Temperature Control

A multivariable control strategy for steam temperature control system in a power plant is presented with the aid of diagonal recurrent neural networks (DRNN). There exists strong couple between reheating process and desuperheating process in this system. The unknown controlled process dynamics are identified by two diagonal recurrent neuro-identifiers (DRNI), which provide the sensitivity information of the objects to two diagonal recurrent neuro-controllers (DRNC) respectively. The convergence of the proposed control algorithm is analyzed. The control strategy has been used in the simulation to control reheating steam temperature and superheated steam temperature of a 200 MW coal fueled power plant, the simulation verifies the effectiveness of the proposed method