In large scale condensing turbine unit

the exhaust status always lies in wet steam area.Due to the lack of effective measuring method

the exhaust dryness of the steam turbine is difficult to obtain directly

which has been the difficult problem in online economic analysis for thermal power units.By taking an N1000-25/600/600ultra-supercritical steam turbine as an example

the nonlinear mapping ability of BP neural network was used to establish a model which can reflect the relationship between exhaust dryness and unit load and exhaust pressure.After learning and training under some typical conditions

this model was used for exhaust dryness online calculation under full condition.The results show the final error of the training samples and verifying samples were controlled within-0.006 1and-0.001 0

which satisfies the accuracy requirement for engineering calculation

indicating the established BP neural network can be used in exhaust dryness prediction.