Research Of Identification Method For The Oil Spill Species Based On The Three Dimensional Fluorescence Spectra

With the rapid development of the industry in the world and people’s increasing requirements for the standard of living, oil plays a more important role in the economic development of various countries.Offshore oil exploration activities have become increasingly frequent along with the oil demand more and more, which makes oil transport by sea also increasingly active. Marine oil spill incidents occurred many times and polluted marine environment seriously. According to statistics, it has occured a total of 175 large oil spill accidents that each volume exceeds 5000 tons. About 65 oil spill accidents happened in china from 1973 to 2005, leading a total of 36795 tons oil spill. To 2013, there is about 17150 square kilometers of sea area which oil content over the first and second class sea water quality standards, particularly serious in Bohai area that accounting for about 48% of contaminated area. In addition to make the greatest efforts to eliminate the influence of oil spill on the environment, marine plants and animals when oil spill accident happens, the identification of the species and source of the oil spill is also very important. Simple, rapid and accurate identification method can not only indicate the direction of the oil spill identification work, but also improve the identification efficiency and save manpower. Becides having advantages of traditional flurescence, the three-dimensional fluorescence spectrum can measure the emission specrum and excitation spectrum at the same time. So its application in identification for oil spill species has got more and more attention and research.Crude oil and fuel oil are two common spill oil types. This paper chooses 6 kinds of crude oil and 3 kinds of fuel oil. They are put outdoor for natural weathering experiment for 45 days. The classification on crude oil and fuel oil is analysed for single oil samples, mixed oil sample and single oil part dissolved in water. Apply the better method of which identification is higher to the samples collected from Bohai to see the mail oil species in this area.The single oil samples are analysed by the following three methods. (1) The first method is using PARAFAC and NNLS together. The samples are scattered and normalized, then subsequently for parallel factor analysis and get each sample’s fluorescence characteristics made of 7 fluorescence components. Choose the weathered samples taken in 3rd,15th,45th days and the first parallel unweathered samples for the cluster analysis. Then a fluorescence standard libraries is constructed by 12 crude oil standard spectrum and 6 fuel oil standard spectrums. Use non-negative least squares linear regression to have linear fit for the rest of the samples. The recognition accuracy for crude oil and fuel oil using PARAFAC-NNLS is 87%, 100%, respectively. (2) The second method is WT-NNLS. The oil EEM after scattered is reduced dimensionality and then normalized. Subsequently it is decomposed by bior5.5 wavelet function. The 4th layer wavelet detail coefficient is selected as sample’s feature vector. It is by cluster analysis to construct fluorescence standard libraries. The rest of the samples are had linear fit using non-negative least squares linear regression. It’s 100% recognition accuracy both for the two oil types when using the method WT-NNLS. (3) The third method is WT-PLS. Choose the fourth layers wavelet detail coefficient decomposed by bior5.5 of weathered samples taken in 3rd,15th and 45th days and the first parallel unweathered samples as training set, while the other samples are used as validation set. Partial least squares is combined with the training set to establish the calibration model, and then use the calibration model to predict the validation set. It’s 100% recognition accuracy for the two oil types using WT-PLS.The main oil type and weathered law of mixed oil samples is analysed by method PARAFAC-NNLS and WT-NNLS. When the proportion of crude oil reach up to 70% during the whole weathered experiment, it can be identified as the main oil type absolutely by PARAFAC-NNLS in mixture of sea 2 station crude oil and 380CST fuel oil, but the crude oil is identified as the mail oil by WT-NNLS only when its proportion reach to 90%. In the mixture of the same crude oil and 5-7 fuel oil, PARAFAC-NNLS can recognize crude oil as the main oil type totally when its proportion reach up to 70%, while it totally can be identified as the main by WT-NNLS only when the proportion of crude oil reach to 90%. From the day of 15 to 45, however the amount of crude oil change, fuel oil is identified as the mail by the two methods. Generally speaking, the recognition accuracy using PARAFAC-NNLS is better than that using WT-NNLS.The results for the oil samples that dissolved in water are as follows. Crude oil, fuel oil recognition accuracy rates analyzed by PARAFAC-NNLS is 83%,22%, respectively. Analysis results for the two oil types using WT-NNLS and WT-PLS are the same that are 100%,0%, respectively. After oil, especially fuel oil, dissolved in water, the identification accuracy is reduced substantially. Distinguish analysis is tried again after oil component extracted by Hexane from water. The identification accuracy for crude oil extracted is 44%,89% and 89% when using PARAFAC-NNLS、WT-NNLS and WT-PLS, respectively. It is 78%,44% and 67%, respectively, when apply these three methods to extracted fuel oil.The methods PARAFAC-NNLS and WT-NNLS are applied to the petroleum hydrocarbon samples collected from Bohai Sea, mainly according to the result which get from PARAFAC-NNLS because of its higher identification efficiency. Crude oil is the main petroleum hydrocarbon contaminants in Bohai area. The stations that fuel oil is identified as the main oil type are most situated in the southern coastal area. This recognition results may related to the number of offshore drilling platforms and the nearshore marine vessels density.