Research On Soft Sensor Method Of Nitrogen Oxide Content In The Entrance Of SCR System Of Coal-fired Unit

As China’s policy for pollutant emission become increasingly strict,the precise control of nitrogen oxide emissions from coal-fired units becomes very important.Obtaining accurate and real-time value of the pollution concentration is fundamental to effective and energy-saving operation for pollution control.However,the existing hardware devices cannot measure in a long-term and timely manner,and soft sensors without proper consideration of input variables selection,system delay,and update frequency have problems of poor accuracy and long computational time.Firstly,I propose using-effect relationship as an evaluation of the input set selection for the time series soft sensor model.A method based on transfer entropy is proposed to test the causal-effect relationship between variables and detect the delay between variables,and the sequences of variables are also adjusted to be aligned.The method is verified using a simple nonlinear model.Secondly,I propose an iteration based soft sensor model of online adaptive soft sensor method based on the least squares support vector machine.The iteration is used to reduce the number of retraining procedure of the online model to reduce the calculation time and improve the computational speed of the online model.The advancement of the soft sensor model is verified using a simple nonlinear model.Finally,the proposed soft sensor method is applied to measure NO_x content in the entrance of SCR system of coal-fired unit,in order obtain the accuracy value in short time.The actual data of a coal-fired unit is processed by the proposed method.I first detect the causal-effect relationships and delay between variables.Then a set of variables that best cover features without introducing redundant information to the soft sensor model are selected.The sequences of variables are also adjusted according to the delay.After the variables are processed,the concentration of NO_x from the entry of the SCR system is obtained by the proposed iteration based soft sensor model.The simulation results show that the proposed soft sensor method can meet the requirements of field application regarding calculation time and measurement accuracy.