Artificial neural network model for a biomass-fueled boiler

In order to operate plants fueled with biomass in an optimum manner, it is important to create thermodynamic models of the same. However, these kind of plants are hard to model by "traditional" methods such as heat and mass balance programs. Some difficulties are the large inertia of some subsystems, as well as the fact that many important parameters are not constant nor unequivocally determined. For this reason, Artificial Neural Networks (ANNs), a technique within the field of Artificial Intelligence (AI), have been chosen as the main candidates to build an adequate model of these kind of plants. Data from an existing plant is used to train, validate and test the ANNs. More specifically, an ANN-based model of the biomass-fired boiler of the plant is implemented which is able to catch the non-linear behavior of the system at different operational conditions with a satisfying accuracy. A conclusion of this work is that ANNs can be considered as a useful tool to model the biomass-fueled boiler. Several sensitivity analyses and pruning of unnecessary inputs were carried out. For instance, some input parameters revealed themselves to not have significant influence on the accuracy of the ANN-model, while in physical modeling they are to be considered as essentials. One possible outcome of ANN modeling is to gain insight about which sensors could be excluded from the existing sensor configuration without lowering the reliability of the plant. A good plant model will supply the personnel in the control room with information necessary to make reliable predictions and arrive at correct decisions. This can lead to a considerable reduction of operational and maintenance costs and improved performance of the plant.