Measurement of the flow rate of oil and water using microwave and Venturi sensors with end-to-end dual convolutional neural network

Abstract

Considering that the one-dimensional convolutional layer (conv1d) is specifically designed for time series, this paper proposes a method for measuring the water-to-liquid ratio (WLR) in oil-gas-water multiphase flow, using a convolutional neural network based on a conv1d, microwave amplitude time series and microwave phase time series. In addition, it also proposes the use of convolutional neural network based on conv1d and Venturi double differential pressure time series to measure the mixed liquid flow rate in the oil-gas-water multiphase flow. However, the calculation of the oil or the water flow rate through a combination of the Venturi and the microwave calculation results can result in the problem of transmission error. To solve this problem, this paper proposes a method using microwave and Venturi sensors with end-to-end dual convolutional neural network to measure oil flow rate and water flow rate in oil-gas-water multiphase flow. The experimental results show that this method can effectively reduce the transmission error and improve the accuracy of measurement.