Thermal control of coke furnace by data-driven approach

To achieve efficient and stable coke furnace operation, we developed an operation guidance system to reduce the variance of net coking time (NCT). A control algorithm that predicts the future NCT by locally weighted regression using the individual chamber database and adjusts the fuel gas flow rate was constructed. After a simulation validation using a newly developed one-dimensional transient model, the operation guidance system based on the developed control algorithm was implemented in a real plant. As a result, the root mean square (RMS) of the control error of NCT was reduced by 0.12 h, and the combustion chamber temperature was reduced by 28.8 degrees C. The developed operation guidance system contributed to the reduction of fuel consumption and CO2 emission of the coke furnace.