Prediction Of Light Absorption Properties Of Organic Dyes In Dye-sensitized Solar Cells Using Machine Learning Technology

In dye-sensitized solar cells(DSSC),the development of highly efficient organic dyes for light absorption throughout the visible region remains one of the most important scientific challenges.However,there are still some problems in the current two mainstream analysis methods.For example,the comprehensive process of the experiment is complicated and the cost is high,and the simulation calculation takes too long.In order to solve the above problems,we introduce machine learning technology into the performance calculation of organic dye molecules.The specific work is as follows:1.We studied the ground state structure,energy levels and optical properties of a range of organic dyes,with a focus on predicting the absorption spectrum of dye molecules by machine learning techniques.By screening a series of machine learning algorithms and constructing a database of dye molecules,we first carried out the feasibility of self-learning of dye molecules,further verified our models and databases,and successfully applied machine learning techniques to the absorption spectrum prediction of organic dye molecules.2.At present,we have implemented the following four algorithms to predict the absorption spectra of dye molecules: support vector machine(SVM)algorithm,extreme gradient enhancement(XGBOOST)algorithm,random forest(RF)algorithm,and extremely random forest regression(ET)algorithm.We found that the predicted absorption spectra of dye molecules using the SVM algorithm and the XGBOOST algorithm have good matching with experimental data.Most of these dyes have absorption peak deviations in the range of 20 nm,especially DS-1 with a deviation of only 3.66 nm.In this work,we successfully predicted the light absorption properties of a series of dye molecules by establishing a database of dye molecules and screening an effective machine learning model.Our machine learning model saves time compared to traditional DFT calculations;our machine learning model saves more cost than experimental synthesis3.In order to quickly predict the absorption spectrum of the new dye,and to facilitate the further expansion of our database and enhance the practicality of the model,we have further developed the model calculation implementation software,integrated the built-up complete dye molecular database and algorithm,which can be realized by simple operation.The absorption spectra of the dye molecules are quickly predicted and the results are derived,making it easier for everyone to use the new machine learning model for research.We have constructed a machine learning model that can be used to predict the absorption spectrum of dye molecules.We believe that our model can be used as a reference for the design of novel organic dye molecules,and we believe that our machine learning model has very good performance in the calculation of organic dye molecules.Big prospects.