石油含水率测量动态补偿模型研究

Research on Measuring Moisture in Petroleum Based on Dynamic Compensation Method

针对传统微波透射法测量石油含水率存在的测量误差大等问题,提出了一种基于神经网络的动态补偿方法,确定衰减和相移两个参量作为动态补偿模型的输入;针对传统BP算法具有收敛速度慢、容易陷入局部极小值等缺点,采用微粒群训练算法对神经网络动态补偿模型进行训练,可使微波透射石油含水率测量结果的补偿过程具有寻优全局性和精确性。实验结果表明,利用该技术对石油含水率测量结果进行校正是一种有效的方法,具有一定的应用价值。

To against the problem of big error in measuring moisture in petroleum by traditional microwave transmission method, the dynamic compensation technique based on neural netwnrk is proposed. Two of the parameters, i.e. microwave attenuation and phase shift are taken as the inputs of the dynamic compensation model. Considering the shortcomings of conventional BP algorithm, e.g. converging slowly and easily trapping a local minimum value, a learning algorithm using particle swarm optimization （PSO） is adopted to train the neural network dynamic compensation model. This will enable the compensation process optimal and accurate overall. Experiments show that the use of the technology in calibrating the measurement result of moisture in petroleum is effective and has certain applicable value.