Neural modelling of steam boilers

The paper presents a method and example results of calculations of the neural modelling of steam boilers. Empirical models can be worked out based on the results of specially organised measurements or continuous measurements recorded in the computer system storing the operational performance. The introduction of operational measurement data for material and energy balances required the separation of stationary sub-periods of boiler operations. For each separated sub-period of stationary operation thermal calculations based on DIN 1942 have been carried out. The results of calculations are utilized to estimate the neural model of a steam boiler. This model describes the dependence of the main operational parameters of the boiler upon the flue gas losses and losses due to unburned combustibles. The parameters of the neural model have been estimated by means of the back-propagation method. (C) 2007 Elsevier Ltd. All rights reserved.