Research On Soft Measurement Method For Three-phase Flow Based On Hybrid Particle Swarm Optimization Algorithm

The principle of the three-phase flow system is very complex. And it is difficult tomeasure the parameters of this system. However, it is important to research its operatingcharacteristics and measure its parameters accurately for the industrial production and theresource exploitation. In recent years, the soft sensor modeling method based on the leastsquares support vector machine has been used widely in the pattern identification and theflowrate measurement of the three-phase flow. And the further research for this soft sensormodeling method has been an important direction in the field of multi-phase flowdetection. Therefore, a soft measurement model for the three-phase flow based on the leastsquares support vector machine algorithm which is optimized by a hybrid particle swarmalgorithm has been studied in this paper.Firstly, aiming at the uncertainty of parameters setting of the least squares supportvector machine and slow calculation speed and poor robustness of grid search method, aleast squares support vector machine optimization algorithm based on a hybrid particleswarm algorithm is studied. And the regularization parameter and kernel functionparameter of the least squares support vector machine are optimized by this algorithm.Secondly, a flow pattern identification model of the three-phase based on the leastsquares support vector machine which is optimized by the hybrid particle swarmalgorithm is established. Dynamic experiments were carried out on a simulation testdevice of the three-phase flow, and bubble flow, transition flow and slug flow patterns areidentified in these experiments. Mean, standard deviation, skewness, kurtosis, waveletpacket energy features, hurst exponent, correlation dimension, kolmogorov entrogy, EMDenergy entropy, power spectrum entropy, approximate entropy and Lempel-Zivcomplexity are extracted from the conductivity fluctuation signals of the three-phase flow,and they are used as the secondary variables of the identification model. Besides, themulti-classifier based on the binary tree method is constructed in this model.Finally, a soft measurement model for the split-phase flowrate of the three-phaseflow based on the adaptive iterative least squares support vector regression machine which is optimized by the hybrid particle swarm algorithm is established. In bubble flow pattern,transition flow pattern and slug flow pattern, the oil flowrate and the gas flowrate of theoil-gas-water three-phase flow are measured respectively by this soft measurement model.The total flow, the moisture content, the liquid velocity and the gas velocity of theoil-gas-water three-phase flow are used as the secondary variables of the model.