Waste-to-energy plant evaluation of combustion stability using an automatic crane system with AI-based waste identification

Waste-to-energy facilities play a major role in realizing volume reduction and stabilization of organic waste, as well as energy recovery and utilization. In the operational management of waste-to-energy plant, stable combustion is required for preventing emission of hazardous substances and for realizing efficient energy recovery. It is known that proper mixing and homogenization of the wastes in the bunker by waste crane is crucial for achieving stable combustion, and this makes limitation on rate of automatic crane operation due to unavoidable visual inspection of waste mixing status and necessary manual intervention by crane operators, which was dependent on the experience and judgement of the operators. To increase the rate of automatic crane operation, the authors developed an innovative automatic crane operation system with an artificial intelligence (AI) that uses deep learning to identify waste types inside a waste bunker, as well as an advanced control device that recognizes optimal crane operation based on the inference results made by the AI. The AI is developed by applying semantic segmentation algorithm to waste bunker images as training data, for which labels of each waste types are added. The F1 scores of the trained AI were high enough and close to 1 for each type of waste. The AI and the system were implemented at a municipal waste-to-energy facility in Japan and a 6-day demonstration experiment was carried out. An automatic operation rate of about 90% was achieved during the demonstration experiment period, without adversely affecting waste combustion or requiring constant monitoring of the bunker by an operator. Thereafter, the experimental plant continued stable operation using the AI based system.