Image processing based combustion control in coal combustors

The issue of improving thermal efficiency in boilers is getting more and more important for engineers because of rising fuel prices and environmental concerns. In this context, common point of the boilers working efficiently and cleanly is to have a closed loop control system controlling fuel/air ratio on its ideal value. From this point of view, recently, monitoring of combustion chamber by using the cameras and the efficiency analyze of combustion with image processing techniques are trendy topics in this field. This study mainly consists of three consecutive stages. In the first stage, the combustion chamber of a coal-fired burner with a capacity of 85000 kcal / h was monitored by cameras from two different angles. The coal loading to the burner was carried out in a constant and loading-waiting manner to obtain flame images for different combustion conditions. At the same time, excess air coefficient and emission values of combustion were obtained via a flue gas analyzer. Different image processing techniques were applied to the obtained flame images and significant information was gathered from image matrices. Then, the prediction of excess air factor was carried out in two different ways. The first ANN model was formed by taking image information and the flue gas temperature as input, excess air factor as output. The second ANN model was formed by taking only the image information as input and the excess air as output. In the second step, the relationship between the air supply to the combustion chamber and the excess air coefficient was defined by the system identification process with a high accuracy transfer function. Finally, three different controllers (Weighted geometric center-based PI-PD controller, H∞ controller and Model Predictive Controller) which are from three different families were designed and applied to the transfer function. To provide virtual reality, estimated lambda values were included to the system as disturbance.