Selective Catalytic Reduction System Ammonia Injection Control Based on Deep Deterministic Policy Reinforcement Learning

The control of flue gas emission in thermal power plants has been a topic concern. Selective catalytic reduction technology widely used as an effective treatment technology. However, precisely controlling the amount ammonia injected remains challenge. Too much not only causes secondary pollution but also corrodes reactor equipment, while too little does effectively reduce NOx content. In recent years, deep reinforcement learning achieved better results than traditional methods decision making and control, which provides new for selective systems. purpose this research is to design intelligent controller using technology, can accurately injection, achieve higher denitrification effect less pollution. To train controller, high-precision virtual denitration environment first constructed. order make more realistic, was designed special structure with two decoders unique approach fitting environment. A deterministic policy agent ammonia. stable, actor-critic framework experience pool were adopted. show that emissions nitrogen oxides at outlet after training