Research Into Prediction Model Of Water Content In Crude Oil Based On Intelligent Information Processing Technique

The measurement of water content in crude oil is an important and widely encountered practice in all aspects of oil industry operations, crude oil production, processing, transportation and refining. The study and implementation on the online detecting technique for water content in crude oil is regarded as a more significant thing at home and abroad. Aiming at dealing with some defects in the present measuring instrument for meeting the practical requirement of oilfield production, an intelligent information processing technique is used to develop a novel measuring instrument for water content, which takes the advantages of high-precision, extensive measuring coverage and higher performance-cost ratio. This subject is of vital significance in both theory and practice to promote the development of on-line measuring technology of water content in crude oil.The measurement of water content in crude oil is affected by various factors, including flow states of oil/water, temperature, salinity content, non-linear character, etc. A measuring system based on multi-sensor is designed in this dissertation for doing experiments about oil/water two-phase flow, the mathematic relationships and influencing rules among temperature, salinity content, water content, dielectric constant of oil-water mixed fluid are studied. The measured temperature, output signal of water measuring sensor and salinity content are selected as main parameters of oil/water two-phase flow, the prediction model for water content in crude oil is researched thoroughly by using a set of methods of multiple regression analysis based on nonlinear transformation of sample matrix, artificial neural network, combined algorithm of multiple genetic-neural networks, intelligent compound method of pattern recognition-genetic neural networks. The experimental results indicate that the intelligent compound model for subsections of water content evidently improves the measuring precision for water content under influence of phase-transformation,viscosity of emulsion, temperature and salinity content, gains a good performance of pattern recognition and generalization ability, overcomes the short-coming of present measuring methods, and is proved to be a novel intelligent measuring method with higher precision.Except above, the advantages/disadvantages and practicalities of four different prediction models adopted in this dissertation are analyzed deeply, which provides a base for realizing high-precision online measurement of water content. The problem about transformation from laboratory achievement to practical application is also discussed preliminary.