中国整体能源格局呈“多煤、贫油、少气”态势，富油煤作为一种特殊煤炭资源，其开发对改善国家油气资源匮乏现状意义重大。焦油产率作为富油煤判识的唯一指标，目前主要通过格金干馏试验获得，而利用地球物理手段进行焦油产率评价的研究相对薄弱。为了建立基于测井参数的煤焦油产率预测技术方法，以新疆三塘湖煤田条湖矿区为例，通过26次工业分析、元素分析、格金干馏等实验与13次自然电位、自然伽马、补偿密度等测井参数数据收集，建立了煤焦油产率-煤质参数-测井参数的数据库，利用统计方法开展研究区煤焦油产率敏感因素分析，厘定与煤焦油产率相关地质参数，利用多元回归分析方法和随机森林回归方法首次开展了双预测模型联用的煤焦油产率测井预测方法探究。结果表明：①研究区内煤焦油产率敏感的地质因素主要有煤质灰分、H、C质量比（H/C）以及镜惰比等，对煤焦油的敏感测井参数主要包括自然电位、自然伽马、补偿密度等；②建立了测井参数预测煤质参数的多元回归分析、煤质参数预测煤焦油产率的随机森林回归分析，首次开展了双模型组合的煤焦油产率预测评价模型，且预测模型精度达到89%。研究成果为预测富油煤焦油产率提供了新思路，提高了煤焦油产率评价模型的精度，对煤焦油地质储量评价具有重要的技术支撑作用。

China's energy structure exhibits a pattern dominated by coal, with comparatively lower reliance on oil and natural gas. As a unique coal resource, oil-rich coal plays a crucial role in alleviating the nation's oil and gas shortages. Coal tar yield serves as the primary criterion for identifying oil-rich coal and is typically determined using the Grignard distillation test. However, research on evaluating coal tar yield through geophysical methods remains limited. To develop a predictive approach based on logging parameters, taking the Santanghu Coalfield Tiaohu Mining Area in Xinjiang as an example, a comprehensive database integrating coal tar yield, coal quality parameters, and logging data was compiled. The dataset was constructed through 26 industrial and elemental analyses, Grignard distillation tests, and 13 logging measurements, including spontaneous potential, natural gamma, and compensated density. Statistical techniques were applied to identify key factors influencing coal tar yield in the study area, determine the geological parameters associated with coal tar production, and develop a predictive model integrating multiple regression analysis and random forest regression. The findings indicate that the primary geological factors affecting coal tar yield include coal ash content, the H/C, and the mirror inert ratio, while the most influential logging parameters are spontaneous potential, natural gamma, and compensated density. Moreover, a multiple regression model was developed to estimate coal quality parameters based on logging data, while a random forest regression model was applied to predict coal tar yield using coal quality parameters. The research innovatively integrates these two models for coal tar yield prediction, achieving an accuracy of 89%. The findings offer a novel approach for forecasting oil-rich coal tar yield, enhance the precision of coal tar yield evaluation models, and provide valuable technical support for assessing geological reserves of coal tar.