基于改进 BP 神经网络的管外测量原油含水率研究

摘要

In order to solve the many factors influence to the measurement of water content of crude oil with the electromag-netic conductance method, the relationship between temperature, mineralization degree, the sensor output voltage and the water content was processed with improved BP algorithm.On the basis of collected data with the electromagnetic conductance method, the water content of crude oil was predicted by using modified heuristic method and numerical optimization BP algorithm.The two modified BP algorithms can improve the prediction accuracy of water content of crude oil and algorithm convergence speed, and the Fletcher-Reeves correction algorithm has better effect than the adaptive learning rate momentum gradient algorithm, which provides the theoretical and experimental basis for the design of the intelligent instrument for measuring the water content of crude oil outside pipeline.%为解决多种因素对电磁电导法原油含水率测量准确性的影响，利用改进的BP神经网络处理温度、矿化度、传感器输出电压与含水率之间的关系。在电磁电导法采集数据的基础上，分别用启发式改进方法和数值优化BP算法预测了原油含水率，两种改进的BP算法均提高了预测原油含水率的精度和算法的收敛速度，而Fletcher－Reeves修正算法比自适应学习率动量梯度算法具有更好的效果，为设计智能化管外在线测量原油含水率仪表提供了理论和实验依据。