BENEFIT ANALYSIS OF COMBUSTION OPTIMIZATION WITH A NEURAL NETWORK FOR VATTENFALLS 6×500 MW JAENSCHWALDE PLANT

摘要

Vattenfall operates 6 units at Jaenschwalde, each unit running one turbine of 500 MWel fed by 2 boilers. The units burn lignite coal. The boilers were modernized in the mid 90ies in order to meet the then actual requirements on a low emission plant. Today besides the dominant task of low generation costs additional expectations exist regarding environmental friendliness. The combustion efficiency is of significant effect for the compliance with these requirements. Improvements in the combustion process enable lower lignite consumption and thus reduced CO{sub}2 emissions and fuel cost. In order to assess the potential of improvement of neural networks on the combustion efficiency, a test was carried out on one of the 815 t/h steam generators in the Jaenschwalde power plant over several weeks by installing a NMPC (non-linear model based predictive control). The boiler is equipped with new sensor equipment developed at the Technical University Zittau showing the furnace temperature distribution across the boiler located right underneath the over fire air nozzles. The relationship of regular process variables to this temperature distribution was examined. The analysis aimed for: centering and smoothing of the fire position; emissions reduction; lowering excess air; reduction of reheat spray flow. This paper shows the method of plant testing within the plant constraints and the benefit-analysis results. Further optimization benefits are shown in other references. Finally the presentation shows the regular procedure of a NMPC installation for a combustion optimization project.