Research On The Signal Filtering Technology Of Complex Thermal System

Complex thermal system signal filtering technology is widely used in thermal process modeling and simulation,power system condition monitoring and optimization control.Conventional filtering methods such as averaging filtering,inertial filtering,and wavelet filtering are based on significant differences between the signal and the noise in the time domain,frequency domain or time-frequency domain.However,affected by the increaing installed capacity of new energy,the proportion of time occupied by thermal power unit operating in variable and low-load conditions is greatly increased,and the dynamic characteristics of the thermal signals are more complex and the signal to noise ratio is lower than those under normal conditions.It may cause or aggravate the problem that the useful components and the useless components do not have significant characteristic differences in the complex thermal system signals.In the complex thermal system,there are many kinds of signals.While ensuring the safe and stable operation of the system,it also provides the possibility of using the multi-information data fusion technology to achieve signal filtering.The research idea of the subject is that the direct or indirect measurement signals that can reflect a certain thermal state are fused in the time domain,frequency domain or time-frequency domain according to objective conditions and filtering purposes to realize signals filtering under non-significant feature difference conditions.The following attempts were made:(1)The independent fluctuation component and its corresponding source signal are obtained by independent component analysis and multi-scale correlation analysis.Based on this,a multi-channel least mean square(LMS)adaptive filter was designed.The result of fusing the output signals and the source signals in the time domain was used as the filter output.This method was applied to filter out the unwanted and non-significant feature differences fluctuation components of the furnace pressure signal caused by uncontrolled disturbances;(2)In order to further reduce the requirements on the input signals type and extend application range of the multi-channel LMS adaptive filter,one of the plurality measurement signals was selected as the desired signal of the multi-channel LMS filter,and the remaining measurement signals were used as the fused signals.This method was applied to solve the problem of conventional filters' poor filtering effect caused by unknown noise characteristics of sub-critical boiler drum water level signals;(3)According to the direct measurement signals or the indirect measurement signals have complementary advantages in the dynamic and static characteristics,after using the soft measurement technique to eliminate the dimensional difference between different measurement signals,the dynamic and static complementary fusion filter was constructed in the time domain,the frequency domain and the time-frequency domain respectively.This method was applied to solve the problem of co-channel interference in the wind signals of double-inlet and double-out steel ball mill.According to engineering experience and comparative analysis of operation data and simulation data,The filtering method based on multi-information source data fusion can effectively solve the problem that the useful components and the unwanted components do not have significant feature differences in some thermal signals.The subject has made a meanful exploration in signals filtering of complex thermal systems.