Automatic Control
Zhang Guobin, Guo Ruijun, Du Ronghua, Xin Xiaogang, Zhang Qian, Niu Yuguang
The main steam temperature control system was proposed based on the prediction and compensation with long short-term memory (LSTM) in a thermal power unit, and the nonlinear relationship was established between the main steam temperature and following factors from the operation history database of the unit, such as the feed coal flow rate, the main steam pressure, the feed water flow rate, etc. The prediction model with LSTM was built for the main steam temperature control system, after which the main steam temperature control system based on prediction and compensation was constructed through the prediction value. Results show that, compared with the back propagation (BP) algorithm, the LSTM algorithm has a higher prediction accuracy, while in the actual application, the control quality of the steam temperature can be improved, which is caused by the advanced action of the control system based on the predictive compensator with the prediction value.