Coal properties on-line analysis with synergy adaptive moving window support vector regression based on immune clone algorithm
- 1 June 2018
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 5913-5918
Abstract
It is of great significance to obtain moisture, volatile matter, ash and heating value in real time. Online type coal analyzer used in principle is generally used in X ray fluorescence or neutron induced gamma ray, but these methods are radioactive, bring security risks to the field staff. With the development of modern computer and chemometrics, NIR on-line detection of coal quality gradually get people's attention. However, when using near infrared spectroscopy for coal quality analysis, it is often used to select the optimal pretreatment method under the full spectrum, and the selection process usually depends on the expert experience or trial method. It is very difficult to get the optimal model solution. To solve this problem, a novel spectrum analysis tool named synergy adaptive moving window support vector regression based on immune clone algorithm (SA-MWSVR-ICA) is proposed, based on the working principle of the human immune system. Based on the working principle of the human immune system, the performance of the quantitative model is regarded as antigen, and a special vector corresponding to the above mentioned antigen is regarded as antibody. The antibody contains a pre-processing method optimization region which is created by 11 decimal digits, a spectrum variable optimization region which is formed by some moving windows with changeable width and position, and a model parameter optimization region. Through the on-line near infrared coal properties analysis device installed in the actual power station, the quantitative analysis of field pulverized coal shows that the model established by the immune clonal algorithm is better than the traditional methods. Additionally, the selected pre-processing methods and spectrum variables are easily explained.Keywords
This publication has 22 references indexed in Scilit:
- An improved changeable size moving window partial least square applied for molecular spectroscopyChemometrics and Intelligent Laboratory Systems, 2016
- Recent Advances in Artificial Immune Systems: Models and ApplicationsApplied Soft Computing, 2011
- A review of clonal selection algorithm and its applicationsArtificial Intelligence Review, 2011
- Variables selection methods in near-infrared spectroscopyAnalytica Chimica Acta, 2010
- Application of near infrared reflectance spectroscopy to predict meat and meat products quality: A reviewMeat Science, 2009
- Nondestructive measurement of fruit and vegetable quality by means of NIR spectroscopy: A reviewPostharvest Biology and Technology, 2007
- On-line coal analysis using fast neutron-induced gamma-raysApplied Radiation and Isotopes, 2005
- Analysis of coal by diffuse reflectance near-infrared spectroscopyAnalytica Chimica Acta, 2005
- The influence of data preprocessing on the robustness and parsimony of multivariate calibration modelsChemometrics and Intelligent Laboratory Systems, 1994
- Coal analysis by application of the partial least squares method to infrared spectraAnalytical and Bioanalytical Chemistry, 1992