An advanced system of thermal diagnosis with the application of neural modeling of the boilers

Generation of the information for the process of proper operational decision making represents the task for the control systems of the power unit operation. The effectiveness of the decision making is in a large extent dependent of the quality of the information accessible for decision-maker. The evaluation of existing state of operational condition is not sufficient for optimal decision making. Besides the knowledge of indices characterizing the operational conditions, the information about the influence of the operational parameters on the deviation of these indices from optimal point is also very important. Generation of such information requires the application of simulative mathematical model. There are two main groups of mathematical models. The first one represents analytical models based on physical laws. The second one are empirical models based on experimental data. Processes proceeding in energy boilers are very complicated and for this reason in this case the creation of analytical models based only on physical laws are extremely complicated and time-consuming. The development in measurements and computer techniques caused that mathematical modeling based on registered measurements data are more often applied. The paper presents the application of neural modeling method for creation of the empirical model of the boiler. The analytical-empirical models have been analyzed. In the analyzed models the dependence between input and output parameters has been approximated by means of the neural models. The possibilities of application of obtained models for exploitation control system of the boiler as well as for evaluation of the deviation of the index of specific consumption of fuel from reference level has been also demonstrated.