Research On Prediction Of Water Cut Of Crude Oil Based On BP Neural Network

Water cut of crude oil is an improtant parameter during the crude oil production and processing, along with researching oil field development, the accurate detection of water cut of crude oil is technical support of crude oil extraction, dehydration processing, gathering and transportation measures, sales and petroleum refining processes. Water cut of crude oil has great significance for the determination of oil-water boundary layer position, grasp the dynamic oilfield, estimated flow and predict life of development, so the accurate measure of water cut of crude oil is very important.The measure of water cut of crude oil can be divided into manual sampling measurement and online real-time measurement, manual sampling measurement has high precision, but it can not be achieved on-line measurement and the sampling period is longer, difficult to meet the need of field production auto-management. On-line measuring device's stability, accuracy, timeliness, reliability and cost are difficult to adapt to the requirements of the actual oil production because the majority of our oil has entered the high-water extraction period, water cut in well head production vary by a big margin. Carrying out the on-line research of water cut of crude oil is a key premise that improve the detection precision of water cut of crude oil and the level of automation control.so there is an urgent need to develop a high-precision, wide range measurement instrument of water cut of crude oil, or the introduction of new data processing methods to improve the detection accuracy of existing instruments to meet the need of the on-line accurate measurement of water cut of crude oil. The measure of water cut of crude oil by microwave method is influenced by many kinds of factors, and it has complex nonlinear relationship with its influencing factor. In this paper, the experimental data of water cut of crude oil by microwave method is processed by BP neural networks method, and the prediction model of water content of crude oil based on BP neutral networks is established. According to the disadvantages of BP neural networks algorithm itself, such as slow convergence, fall into local minimum point, the method that variable step size combined with momentum is used, the compensation for influencing factors of microwave method measuring water cut of crude oil is realized, so the precision of measurement is improved.