Steady Data Acquiring and Robust Neutral Network Training for Boiler Combustion Optimization

Abstract:

Assuming the measuring parameters obey the Gaussion distribution, this paper proposes a Chi-square distribution based steady data judgment algorithm, which includes two parts, i.e. gross error detection and steady data judgment criterion. In order to overcome the noise’s influence on the result of neutral network training, the paper introduces the RANSAC algorithm into the neutral network training and put forward a RANSAC-BP neutral network training algorithm, which culls noisy data during neutral network training and then retrain the neutral network with noise free data, thus robust to data noises. This algorithm has been validated by a simulation experiment.

• Periodical: Applied Mechanics and Materials (Volumes 157-158)
• Pages: 1527-1532
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.157-158.1527
• Online since: February 2012
• Authors: Jie Su, Xu Guang Wang
• Keywords: Chi-Square Distribution, Combustion Optimization, RANSAC, Steady Data

References

[1] Wang Peihong, Li Leilei, Chen Qiang et al. Optimal algorithm research on high efficiency and low emission combustion of coal-fired boiler [J]. Power Engineering, 2004, 24(4)_4.

[2] Zhou Jiangxin, Fan Zhengbing, Si Fengqi et al. A review about the development of boiler combustion optimization technology [J]. Boiler Technology, 2008, 39 (5)_5.

[3] Li Zhihong. Visual method of flame's development and application [D]. Institute of Engineering Thermophysics, Cas, (2006).

[4] S.C. Hill1, L. Douglas Smoot. Modeling of nitrogen oxides formation and destruction in combustion systems [J], Progress in Energy and Combustion Science, 26(2000).

DOI: 10.1016/s0360-1285(00)00011-3

[5] Gao Daqi, Wushou Yi. An optimization method for the topological structures of feed-forward multi-layer neural networks [J], Pattern Recognition, 1998(31): 1337-1342.

DOI: 10.1016/s0031-3203(97)00160-x

[6] Wang Peihong, Li Leilei, Chen Qiang et al. Research on applications of artificial intelligence to combustion optimization in a coal-fired boiler [J]. Proceedings of Chinese Society for Electrical Engineering, 2004, 24 (4)_5.

[7] U. schnell, M. kaess, H. Brodbek. Experimental and numerical investigation of NOx formation and its basic interdependencies on pulverized coal flame characteristics[J], Combustion science and technology, (1993).

DOI: 10.1080/00102209308935284

[8] J.O. Lwendt. Mechanisms governing the formation and destruction of NOx and other Nitrogenous species in Low NOx coal combustion systems[J], Combustion science and technology, (1995).

DOI: 10.1016/s0140-6701(97)80348-7

[9] Fisher M.A. and Bolles R.C. Random sample consensus: a paradigm for model fitting with application to image analysis and automated cartography. Communications of the ACM, 1981, 24(6): 381-395.

DOI: 10.1145/358669.358692

[10] I. Rivals, L. Personnaz. Construction of confidence intervals for neural networks based on least squares estimation [J]. Neural Networks, 2000, 13: 463-484.

DOI: 10.1016/s0893-6080(99)00080-5