A Study on the Application of Reinforcement Learning to the Operation of Dust Feeders of Stoker-type Refuse Incinerators

Abstract

Stable fuel supply in a stoker type refuse incineration plant is necessary for stable power generation, but it is difficult to achieve stable fuel supply by control based on a single rule, because refuse is affected by climate and other factors in addition to its uneven properties. In this study, we aimed at the automatic establishment of action rules using reinforcement learning for fuel feeder, and evaluated the difference in the optimization of the action of the feeder by the difference in the reward design using a simple simulator simulating the feeder.