Study Of Analytical Method Of Spectral Data Based On Grey Correlation Analysis

To optimize the performance of low dimensional complex renewable energy related devices,spectroscopic techniques become increasingly important.However,the traditional data analysis methods become less effective in revealing the key elements that constrain the performance of the devices.Based on the traditional grey relational analysis model,this thesis combines the information entropy theory to improve the calculation method of the resolution coefficient in the grey relational model,and explore the most suitable data transformation method for the spectral data.Furthermore,multiple linear regression,partial least squares algorithm and BP neural network is used to analyze the chemical composition of samples.Finally,the spectral data of photoelectrochemical cells electrode materials are taken as an example to test the validity and superiority of the model.The main content is as follows:First,the properties of traditional grey relational models are discussed in detail.The model does not satisfy the associated four axioms in grey theory.The modeling concept of grey relational model is analyzed,which show the limitation of using fixed resolution coefficient in the model.Secondly,this paper observes the curve before and after the smoothing of the X-ray emission spectrum,and uses the fitting accuracy and smoothing evaluation criteria to select the 5 point 1 polynomial smoothing algorithm as the first step of data preprocessing;then compares the different data.The gray correlation order and coefficient of variation under the transformation method are comprehensively evaluated to determine the optimal data transformation method suitable for spectral data analysis,namely the extreme value standard transformation method.Third,a dynamic calculation method for grey resolution coefficient based on information entropy is proposed.A new grey relational analysis model is established based on the individual difference of the data,and verified using simulation data,which confirm the new model complies with normative and proximity in grey relational axioms.Finally,based on the spectral data of renewable energy-related materials(such as silicon,silicon carbide,etc.),the simulation data with different component ratios are generated.Using the improved grey correlation analysis algorithm to calculate the correlation between simulation data and reference sequences,the inversion model of the chemical composition of the photoelectrochemical electrode material was established by solving the regression equation by partial least squares.In order to further verify the reliability and validity of the model,the regression accuracy of the grey correlation analysis model under noise interference is analyzed.The results show that the proposed grey correlation algorithm is better than the traditional gray correlation algorithm.