With the oilfield development, we are no longer limited to single-phase flow but must measure multiphase flow and mixed flow. The technical difficulty of measuring multiphase flows is greater than the accurate measurement of single-phase fluids. We need to know the density and viscosity of the single-phase fluid and the geometry of the measuring device to quantify the single-phase flow.
It would be convenient to measure multiphase flows using the above physical quantities for each flow phase. Unfortunately, the characteristics of multiphase fluids are much more complex than those of single-phase fluids, such as the inability of groups to mix uniformly, anomalies in the mixed liquid, flow pattern changes, relative velocities, fluid properties, pipe structure, flow direction, and other factors that can cause changes in the response characteristics of turbine flow sensors.
1. Accuracy measurement of turbine flowmeter
The separation of three-phase flow in oil wells requires split-phase metering accuracy, which is mainly affected by the degree of gas-liquid separation and the accuracy of the water content meter. Therefore, we propose using the coaxial line phase method; then, we use the annular air-water finder and sanitary turbine flow meter to control the gas-liquid separation.
The control is based on neural network self-correcting control. The water content measurement is carried out using a high accuracy coaxial line phase method – the water content meter meets the requirements for accurate measurement of three-phase flow in oil wells.
In oilfield production, when measuring parameters (e.g., temperature, pressure, etc.), it isn’t easy to measure these two parameters due to the complexity of measuring flow rates and phase holding ratios, but this has attracted the interest of engineers.
2. Turbine flow meters for single-phase flow
In single-phase flow conditions, the turbine speed, and the volume flow are a single linear function. However, in oil-water two-phase flow, the turbine response and volume flow are also linear functions as long as the flow rate exceeds the initial flow rate, within an allowable margin of error.
3. Turbine flow meters for multiphase flow
In multiphase flow, however, changes in the density of the fluid mixture, even when the total flow rate remains constant, can cause significant changes in turbine speed.
The uncertainty of the measurement is mainly influenced by the degree of gas-liquid separation, as the mixture is separated by gas-liquid and then measured using proven single-phase flow measurement techniques and phase fraction measurement techniques. Based on this, we propose a hybrid control method for the gas-liquid separator using the response function of a coaxial line moisture content meter and a hygienic turbine flow meter.
4. The principle of the turbine flowmeter
The sanitary turbine flowmeter is a velocity flowmeter, which reflects the flow rate by measuring the turbine’s rotational speed in the fluid.
4.1 The liquid working principle of the turbine flowmeter
A turbine is placed in the center of the pipe; when the fluid passes through the line, the medium impacts the turbine blades, generating a driving torque on the turbine. The turbine overcomes the resistance torque and produces rotation.
In a specific flow rate and viscosity range of fluid media, the angular velocity of the turbine rotation is proportional to the flow rate of the fluid. Therefore, the flow velocity of the fluid can be obtained through the rotational angular momentum of the turbine, which in turn can be converted to the flow rate of the fluid through the pipe.
4.2 The liquid and gas turbine flowmeter principle of operation
In the case of gas-liquid two-phase flow, the measurement results are increased due to the high velocity of the liquid phase compared to the liquid phase. The coaxial phase method measures the water content by measuring the phase difference between the electromagnetic wave propagating through the oil and water mixture.
When there is gas, the increase in the dielectric constant of the gas-liquid mixture results in a much lower response value for the Coaxial Phase Circumferential Water Finder than for the liquid. When the gas-liquid separation is complete, its response value rises again to the value of the pure juice, which is judged to be done.
5. Turbine flow meters reduce the measurement error.
Gas in the gas-liquid separation, we use sanitary turbine flowmeter and coaxial line phase method ring space water finder response function, so that you can control the gas-liquid separator will be good to reduce the error of measurement.
The self-correcting control of neural networks, without a reference model, relies on online recursive identification (parameter estimation) to estimate the unknown parameters of the system, which is used to control the design algorithm online for real-time feedback control.
Let the single input single output linear system be
YK-1 is the object output; the UK is the controller output.
f (-) and g (-) are unknown, and the BP neural network can approximate these functions and recalibrate the control law by learning the algorithm, for simplicity, let the controlled object by a first-order system, i.e.
Using the model through neural networks to obtain
to approximate the object model, where W=W[W0, W1, ……, W2P], V=[V0, V1, ……, V2q], and have
The corresponding control law is
Substituting equation (4) into equation (3), we get
It is very small, and thus there is: minor, and thus there is:
is not known, but its sign is known, and sgn[g (YK) ] can be used instead of g (YK), which gives the learning rule for adjusting W (k) and V (k) as
where:ηk, μk is the learning efficiencies, respectively.
6. Application examples
In the water-gas simulation loop, clean water is used to simulate the excellent output fluid, and the air is used to simulate the separation gas from the well. First, a standard flow rate is obtained by adjusting a traditional flow turbine. Then, the average flow rate of water is mixed with a certain amount of gas to bring a mixture.
The mixture is then passed through the separation unit and separated to obtain the flow rate of the separated water. In the test, when the response value of the water content meter for the total water value, the separation is complete, and the experimental water content meter is selected for the entire water value of 4500Hz.
The neural network structure was chosen as 4-5-3, with a learning rate of ηk = 0.78, μk = 0.64, and the initial values of the weighting coefficients were random numbers in the interval [-0.5, 0., 5]. The input command signal is a sampled coaxial line phase method water rate meter signal, and the output signal is taken from a hygienic turbine flow meter.
7. Conclusion
The gas-liquid separation can be controlled by the response function of the coaxial phase method water content meter and the hygienic turbine flow meter, which can be monitored and adjusted in real-time to achieve complete separation.
The turbine flowmeter can effectively achieve real-time control of the gas-liquid separation based on neural network self-calibration control. This method is effective and has been applied with excellent prospects.