考虑基于传统的介电常数法动态测量原油含水率时存在多变量交叉敏感性, 检测精度无法满足石油生产实时优化控制的需要, 研究了利用多传感技术对存在交叉耦合的多敏感参量进行测量, 提出了一种基于多维数据驱动的遗传优化小波神经网络逆模型及其辨识方法。该模型克服了传统神经网络初始参数随机选取的盲目性, 具有全局优化和复杂非线性自学习性能, 摒弃了多影响因素之间的交叉敏感性。仿真和实验结果表明了该模型的有效性, 其模型预测值与实验标定值之间的相关系数为0.999 6, 优于BP-NN模型。该方法具有较强的泛化能力和鲁棒性, 有效抑制了温度、矿化度等多参量交叉敏感性及传感器自身非线性对测量精度的影响, 改善了多传感器系统的非线性动态特性和检测精度。

As the traditional measuring method based on dielectric coefficients shows cross-sensitivity for multi-parameters in the measurement of oil/water two-phase flows, it can not meet the requirements of real-time optimization control for petroleum production. Therefore, this paper investigates a method to measure multi-parameters with cross-sensitivity by using multi-sensing technology.It presents an inverse model of wavelet neural network with genetic optimization and also researches its identification method. The model overcomes the blindness of initialization weight-value choice in traditional neural networks, provides the abilities of global optimization and nonlinear self-learning, and eliminates the cross-sensitivity of multi-factors. The simulation and experimental results demonstrate the validity and effectiveness of the proposed model and show that the correlation coefficient between the predicted values and calibration values is 0.999 6, which is better than that of BP-NN model. The method has strong generalized capability and robust convergence rate, and can effectively eliminate the influence of the cross-sensitivity of multi-factors and the nonlinearity of sensor self on the measuring precision, and improve the dynamic characteristics and measurement accuracy of sensor systems.