Online Calibration of ERT Based on Impedance Analysis in Oil–Water Two-Phase Flow Measurement

Oil-water two-phase flow is widely exists in the process of oil extraction and transportation. The measurement of water-to-liquid ratio (WLR) of oil-water two-phase flow is crucial for oilfield extraction, which not only helps to manage the oilfield production process, but also calculates the single-phase flow rate when the total flow rate is known. Electrical resistance tomography (ERT) is suitable for WLR measurements in today’s fields at high water content, but requires calibration with measurements under full water conditions. Calibration values are often changed due to factors such as temperature and mineralization, and the originally obtained calibration values cannot be applied. In the past, manual recalibration was often required, but this recalibration process was cumbersome and costly. For this, a method based on impedance analysis is proposed to achieve the online calibration of ERT. First, the feature data for conductivity prediction is constructed by combining the measurement location information and frequency domain information of ERT; then, the conductivity prediction is achieved by using the modified residual network (ResNet)-18 deep neural network; finally, the mapping model from the conductivity value of water to the calibration value of ERT is constructed through approximately 30 h of experiments, which achieves the online calibration of ERT in the process of oil-water two-phase flow measurement. The proposed conductivity prediction model can reach the highest accuracy of 93.75% with no more than 5% relative error, and the accuracy of the mapping model from conductivity to ERT calibrated values is higher than 85% in all cases. Meanwhile, all errors are lower than 10%, proving the feasibility of the entire calibration process.