Error Correction Method of Crude Oil Moisture Content Detection Based on BP Neural Network

Abstract

Microwave phase-shifting sensor is one of the effective means to realize online detection of water content in high water-cut crude oil, but its detection accuracy is easily affected by salinity. Aiming at the mineralization components (NaCl and CaCl2) existing in water-bearing crude oil, the influence of different proportion and content of dual-component mineralization on the accuracy of microwave phase-shifting crude oil water content detection sensor was studied experimentally, and the influence rule of dual-component mineralization (NaCl and CaCl2) on the accuracy of crude oil water content detection was obtained. It is difficult to establish an accurate error compensation model because the relationship between the composition and content of salinity and the measured water content is affected by many factors. Therefore, a BP neural network model for error correction is established, which reduces the detection error of microwave phase-shifting crude oil moisture sensor from 13.912% to 1.821%, and improves the detection accuracy. BP neural network prediction model is superior to multiple linear regression prediction model.