【摘要】 在对烃源岩品质和生烃潜力进行评价时，总有机碳含量(TOC)是一个重要参数。一般采用岩心样品进行测试，试验过程耗时、岩心样品数量有限、成本高昂；而采用常规的利用测井曲线的TOC预测方法以线性拟合为主，方法自身的缺陷会导致试验结果产生一定误差。为此，利用粒子群算法(PSO)改进的BP(Back propagation)神经网络模型对TOC进行预测。试验结果表明：自然伽马单因素法预测误差最大，BP模型次之，PSO-BP模型预测效果最好。其中PSO-BP模型具有强大的非线性拟合能力，可更准确地反映页岩气TOC与各测井曲线之间的关系。相较于传统BP模型，PSO-BP模型在预测页岩气TOC方面具有更高的准确度，研究结果为松辽盆地页岩储层TOC预测提供了新思路。更多还原

【Abstract】 When evaluating the quality and hydrocarbon generation potential of source rocks, total organic carbon(TOC) content is a crucial parameter. Generally, core samples are used for testing, which is time-consuming, limited in quantity, and high cost. Conventional TOC prediction methods based on well-logging curves mainly rely on linear fitting. Due to inherent flaws in the method, there may be certain errors in the experimental results. So in this paper, the BP neural network model improved by particle swarm optimization(PSO), abbreviated as the PSO-BP model, is used to predict TOC. The results indicate that the natural gamma single-factor method yields the highest prediction error, followed by the BP model, while the PSO-BP model has the best prediction performance. With robust nonlinear fitting capabilities, the PSO-BP model more accurately reflects the relationship between shale gas TOC and various well-logging curves. Compared with traditional BP models, PSO-BP model shows higher accuracy in predicting TOC in shale gas reservoirs.This method can provide a novel perspective for TOC prediction of shale reservoirs in the Songliao Basin.更多还原