关键词: SG-KNN, 光谱感知, 返出钻井液, 地层流体, 定性预测

Abstract: During the drilling of oil and gas wells, downhole reservoir and formation information is brought to the surface in the form of return drilling fluid. Therefore, through the detection and analysis of the returned drilling fluid, the information of the oil and gas reservoirs and the judgement of formation fluid composition can be obtained, timely understand the formation fluid status. This paper proposes a new method based on SG-KNN for spectral sensing analysis of of fluid composition of formation while drilling. Diesel oil, methane and water were mixed in a certain proportion to obtain a gas-liquid mixture to simulate the actual main hydrocarbon composition of the returned drilling fluid. A large number of experimental data were obtained by compose sensing spectroscopy to detect the mixed fluid samples, and the data were preprocessed by S-G smoothing and normalization method. Finally, the KNN algorithm model was used to detect the returned drilling fluid in real-time, and the oil, gas and water in the returned drilling fluid were qualitatively analyzed and predicted. Compared with other models, the method described in this paper has stronger data processing ability, and higher qualitative analysis accuracy, which can meet the technical requirements of component analysis and prediction of returned drilling fluid in wellsite.

Key words: spectral perception, SG - KNN, return drilling fluid, qualitative prediction