Prediction Model Of Soil Heavy Metal Content Based On Contour Area Method And IGWO-DNN

In recent years,in the field of environmental pollution monitoring,the detection methods of heavy metals in soil have been widely studied.Among them,X-Ray Fluorescence(XRF)spectrometry is widely used in soil heavy metal detection because of its advantages of rapidity and accuracy.In the ideal case,the net fluorescence intensity of the element calculated by XRF spectrum processing should be in direct proportion to the content of the element.In fact,the net fluorescence intensity of the experimental sample will be affected by its water content,so the heavy metal content can not be calculated by the positive proportion function.In this paper,the preparation method of experimental soil samples under the influence of water content was studied.The XRF spectra were obtained and pretreated,and then the water content was taken as the environmental interference factor to establish the prediction model of heavy metal content in soil.The main research contents include:Firstly,the experimental soil samples with different water content and different heavy metal content were prepared.Based on the basic principle of XRF method,the XRF spectral data acquisition of experimental samples under the influence of water content was completed.Through soil collection,soil pretreatment,soil drying,preparation of experimental reagents,preparation of experimental samples,collection of experimental spectral data and other operations,XRF spectrum acquisition is completed.The principle and types of errors that may appear in each experimental link are analyzed,and the experimental steps are optimized.Secondly,the calculation method of XRF element net fluorescence intensity based on Contour Area Method is proposed to process the spectral data.Compared with the traditional Net Peak Area Method,this method can avoid the problem of poor fitting effect of Gaussian Function on characteristic peaks,thus reducing the calculation deviation of fluorescence intensity value.The results show that the prediction accuracy of the heavy metal content prediction model with Contour Area Method as input is obviously better than that of the traditional Net Peak Area Method.Thirdly,the Deep Neural Network(DNN)model was used to predict the soil heavy metal content under the influence of water content.Grey Wolf Optimization(GWO)algorithm is used to optimize the number of hidden layer neurons in DNN model,so as to establish the optimal prediction model.Compared with other models,the accuracy of DNN prediction results was evaluated.Finally,the internal iterative optimization principle of GWO algorithm is analyzed,and an Improved Grey Wolf Optimization(IGWO)algorithm is proposed from the direction of improving the internal location update strategy of the algorithm.By finding the optimal value of the test function,the optimization effect of GWO algorithm before and after the improvement is compared.Combining IGWO algorithm with DNN,the prediction model of soil heavy metal content was established.The results showed that the R~2 of IGWO-DNN reached 0.9841,which was 0.0188 higher than that of DNN.Therefore,IGWO-DNN is a widely used prediction method.