Algorithm Research Of Boiler Furnace Temperature Field In Power Station

Acoustic temperature measurement of boiler furnace temperature field is arapid developing temperature testing technology which have not realized in ourcountry. This paper based on the real-time on-line monitoring technology in boilerfurnace temperature field of power station, some related technologies in acoustictemperature measurement is researched in theory. This paper has mainly completedthe following works:Put forward a scheme of equipping the acoustic sensor unevenly in furnacetemperature field to measuring temperature. In analysis and understanding theleast-square algorithm, the paper put forward steepest descent method, the conjugategradient method and the radial basis function neural network algorithm. Usingmathematical software Matlab, simulation experiments have been conductedrespectively on the single-peak symmetrical temperature model, the single-peakskewed temperature model and the twin peaks temperature model. The RMS error ofsingle-pink symmetric temperature model using these four reconstruction algorithmsare1.99%,1.72%,1.36%and1.04%; the RMS error of single-pink skewedtemperature model using these four reconstruction algorithms are2.62%、2.97%、2.17%、3.53%; the RMS error of twin pinks temperature model using these fourreconstruction algorithms are4.34%、3.92%、2.38%、0.97%. The simulation datashows that the radial basis function neural network reconstruction algorithm hashigher precision than other algorithms, this algorithm of image quality is better thanother reconstruction algorithm.In order to study the influence of interference environment on temperature fieldreconstruction results, joining the random error with normal distribution, temperaturefield models are reconstructed respectively under different level of SNR such as40dB,30dB and24dB. The RMS error of reconstruction results are1.41%,3.61%and1.13%;1.62%,4.37%and1.39%;1.81%,4.86%,1.52%. The results show that the reconstruction image is conformity with the original temperature field, and the radialbasis function neural network algorithm has good anti-noise ability.