Spectral Characterization Of Oil Content In Oilfield Reinjection Water Based On Neural Network

With the wide application of tertiary oil recovery technology in China’s oilfields,the oilfield wastewater produced in the process of oilfield production is also increasing.In order to reduce the environmental pollution caused by oilfield wastewater,oilfield wastewater needs to be treated and injected back into the layer.However,there are strict requirements for the quality of reinjection water in the oilfield.The reinjection of oilfield wastewater with substandard water quality will pollute the reservoir and affect the physical properties of the reservoir.Oil content is an important water quality index of reinjection water.Timely monitoring the oil content of reinjection water is conducive to the normal operation of oilfield production.Traditional detection methods can not accurately monitor the oil content of reinjection water on-line.Therefore,it is very important to establish a method that can quickly and accurately determine the oil content of oilfield reinjection water.This paper summarized the traditional oil content detection methods,and established the quantitative analysis model of oil content of reinjection water combined with spectroscopy and neural network model.The quantitative analysis model of oil content was optimized by analyzing turbidity,p H value and temperature.It provided a theoretical basis for on-line monitoring of oil content of reinjection water in oilfield.The details are as follows:1.The measurement principle and the current states of spectroscopy combined with stoichiometry stoichiometric method were consulted and understood.The theoretical basis of oil content quantitative analysis model was established based on the detection principle of spectral method,spectral preprocessing method,optical constant inversion method,model evaluation parameters and chemometric modeling method.2.The UV spectra of oilfield reinjection water under different temperature and concentration conditions were measured.Based on the double-thickness inversion method,the inversion model of optical constants for oilfield reinjection water was established,and the inversion model of optical constants was verified by the optical constants of distilled water.The optical constant of oilfield reinjection water was solved based on the inversion model of oilfield reinjection water optical constant.3.The convolution neural network model and residual neural network were used to establish the neural network quantitative model of oil content in reinjection water based on the ultraviolet spectrum of oilfield reinjection water.Different modeling data,model structure and spectral preprocessing methods were optimized,and the best oil content quantitative analysis model based on convolution neural network model and residual neural network was determined respectively.The best model of residual neural network,the best model of convolution neural network and traditional modeling methods were compared and analyzed.4.The effects of turbidity,p H value and temperature on the quantitative analysis model of oil content were studied,and the quantitative analysis model of oil content was optimized.Through the above research,this paper established a fast detection method of oil content by oilfield reinjection water based on neural network,which provided theoretical basis and data support for the development of real-time monitoring technology of oil content in oilfield reinjection water...