Deep learning based prediction system to minimize the effect of thermal effluent without domain knowledge

Abstract

Coal-fired power plants, the most common power plants, take more than half of the total power generation in Korea. However, it emits not only carbon dioxide which causes air pollution but also emits thermal effluents which degrades the water quality. In this research, our main goal is to limit the effect of thermal effluents (T5) to the sea (T1) by operating extra Cooling Water Pump (CWP). We can sim-ply formalize the problem as ”T5-T1 - 7”. It requires domain specific experts in traditional approach,however, we propose the prediction system that does not require domain specific experts. Our prediction system achieved 0.84 degree of RMSE with real plants data and we found rooms for optimization and improvements. The main contribution of this work is that we have shown the possibility of AI techniques, which could exploit the complex environment like marine environments, without domain specific experts.