基于人工神经网络原油中气油体积比预测模型研究

针对国外原油气油体积比预测模型在国内一些油田并不适用,在分析BP神经网络基本原理的基础上,提出了原油气油体积比新的预测模型。该模型结构为4-10-1的三层BP网络模型,它考虑了压力、温度、地面原油重度和气体的相对密度对气油体积比的影响。利用该模型对大庆油田实测值进行了训练与测试。测试结果表明:利用人工神经网络方法建立的气油体积比预测模型比国外模型精度高,基本合理可靠。

RESEARCH ON THE FORECASTING MODEL OF GAS-OIL RATIO BASED ON THE NEURAL NETWORK

A new forecasting model of gas-oil ratio, which structure is 4-10-1 three-layer BP network, is put forward on the basis of analyzing the basic principle of BP neural network because foreign forecasting models are not appropriate for Chinese oilfields. In the model, the effect of pressure, temperature, weight of crude oil on surface and relative density of gas on gas-oil ratio is taken into account. The model has been trained and tested by the actual measured values in Daqing oilfield. The result shows that this new forecasting model is more accurate than foreign ones. It is practical and reliable.