부산대학교 도서관

The temperature of the incinerator plays a critical role in ensuring the efficiency and safety of thermal power generation units. Accurate temperature prediction models are essential for controlling furnace combustion efficiency and detecting abnormal combustion states. This study aims to develop an advanced data-driven model that addresses the challenges associated with long-term temperature forecasting. The proposed model adopts an encoder-decoder architecture that integrates a multi-scale temporal vector and a partial attention vector, enabling the model to learn correlations effectively. Combining a Temporal Convolutional Network with a Gate Recurrent Unit as the encoder, the model can adaptively capture the underlying relevance and long-term dependencies among multiple variables in a furnace combustion system. To evaluate the performance of the proposed models, a real-world dataset from a waste-to-energy plant was utilized. The results demonstrate remarkable performance, with a root mean squared error of 3.74 and a mean absolute error of 2.38 in a 30-step prediction. These findings underscore the superiority of our model as an optimal solution for temperature prediction modeling.