State Key Laboratory of Coal Resources and Safe Mining,China University of Mining Technology (Beijing),Beijing　 100083,China

Hydrocarbon source rocks have large physical property differences because of their complicated sedimentary environments. According to the correlation analysis between the TOC and the logging parameters,it has been found that the source rocks in coal measures has the characteristics of marked correlation difference between the TOC and the logging parameters,and the logging parameters exist a cross correlation. To begin with,the logging parameters which selected by using the mean impact value method will participate in the final BP neural network modeling. Thus,the predication error of the model and the adding modeling time caused by non-interdependence between logging parame- ters are effectively avoided. In addition,the prediction models which called Δlog R model,BP neural network model and improved BP neural network with genetic algorithm model are established based on the determined TOC data for calculating the organic carbon content of source rocks in coal measures. Finally,the modeling computation and error a- nalysis for the three kinds of models are conducted. Results show that the improved BP neural network with genetic al- gorithm model has the best prediction results with strong stability and high precision,and it is hardly affected by the heterogeneity of source rocks. Therefore,it can perfectly reflect the subtle changes of TOC content on hydrocarbon source rocks in coal measures.

国家重大科学仪器设备开发专项资助项目(2012YQ030126)；国家自然基金煤炭联合基金资助项目(U1261203)；中国地质调查局基金资助项目(12120115102101)；

Wang Pan,Peng Suping,Du Wenfeng,et al. Prediction model of total organic carbon content on hydrocarbon source rocks in coal measures based on geophysical well logging[J]. Journal of China Coal Society,2017,42(5):1266-1276.