Climate change is making the sustainable use and conservation of raw materials, reduction of pollutants and recycling ever more important. Modern waste incineration plants support and fulfill the conditions for waste treatment laid down in the Closed Substance Cycle Waste Management Act. According to the Federal Environment Ministry’s climate protection plans for decarbonizing the heat, electricity and fuel sectors, waste incineration plants can make an important contribution by producing bioenergy. This is because the thermal utilization of waste generates electricity and heat, reduces the pollutants contained in the waste and reduces the volume of waste. Despite modern control, automation and measurement technology, the efficient and optimal operation of a waste incineration plant in compliance with requirements such as incineration temperatures, retention times and emission limits in accordance with the 17th Ordinance on the Implementation of the Federal Immission Control Act (BImSchV) is still a process engineering challenge due to the inhomogeneity of the waste composition. In this research project, an AI model of a waste incineration plant is developed based on modern AI algorithms such as machine learning and neural networks, which learns the processes of a waste incineration plant and provides new correlations between the target variables of combustion, pollution, throughput, emissions and efficiency. The scientifically validated findings are then integrated into the conventional control process of the system using a standard procedure developed in-house in order to solve vector optimization. The AI model is developed using the Python programming language. Machine learning can be divided into four main areas based on the way the data is monitored:
► supervised learning,
► unsupervised learning,
► semi-supervised learning and
► reinforcement learning.

The types of learning mentioned above are examined in the context of the research objective. The strengths and weaknesses of individual algorithms are worked out according to specific quality criteria so that the validated AI model represents a combination of all AI algorithms.