Input-delay-based LSTM-GRU dynamic modeling method for steam flow in steel industry

The combustion process of gas-fired boilers in hybrid heat and power systems of steel companies is complex and varied, and the time scale of output steam generated by the boilers is different from input gases, leading to the relevant time delay. In order to accurately model the association between multiple input gases and output steam, a dynamic modeling method based on input-delay long and short-term memory network combined with the gated cycle unit (LSTM-GRU) is proposed in this study. To exactly identify the time delay parameters of the influencing factors of the steam, a data-driven method is presented based on the dynamic time warping and cause-and-effect probability, according to which the training samples are constructed. LSTM-GRU is subsequently employed so as to build a dynamic model of steam flow. Actual data coming from a typical steel industry are used to verify the validity of the presented method, the results indicate that the model established offers more precise characterizations for dynamic processes of the steam flow from the gas boiler.