针对综采放顶煤工作面在智能放煤理论、智能感知与识别关键技术、智能放煤综合决策技术和远程放顶煤智能控制技术方面存在的问题，依托“十三五”国家重点研发计划—— 千万吨级特厚煤层智能化综放开采关键技术及示范，开展了千万吨级综放工作面智能化放煤理论及关键技术研究，取得如下成果：①开展了顶板顶煤组合体压缩试验、煤岩组合体破碎块度分布试验和不同顶板−顶煤条件下综放开采放煤相似模拟试验，阐明了顶板顶煤的破碎−运移的相互作用过程；开发了采空区三维激光空间探测技术，证明了群组放煤过程中顶煤“面接触块体成拱现象；开展以采放协调、高回收率、低含矸率为约束条件的特厚煤层多口群组智能放煤数值模拟，确定了群组放煤口数量；综上，为智能放煤工艺优化提供了可靠理论基础。②对工作面煤矸地质信息及物性特征、顶煤放落过程全周期感知要素进行探索研究，形成了包含“顶煤厚度在线探测−煤矸精准识别−煤流动态测量”的综合感知技术体系，为智能化放煤决策技术提供了充分的数据信息支撑。③建立了综放面“人−机−环”多源信息数据库，构建了基于采放时间协调、采放空间协调、采放运能协调的特厚煤层综放面采放协调决策模型，开发了基于Q-learning算法的智能放煤决策软件，形成了基于煤矸识别、顶煤厚度探测和过煤量监测感知，结合煤矸运移时序特征和经验数据的人工智能决策技术。④研发了智能综放面三机位姿高精度惯导监测与控制技术，实现了采煤机、液压支架、刮板输送机的实时定位、姿态监测及动作控制；开发了智能化矿山融合通信调度系统，构建了智能综放远程综合控制平台，成功实现了“远程一键启动”模式的智能放煤。⑤在塔山矿8222工作面，开展了基于探地雷达的顶煤厚度在线探测技术、融合振动−音频−高光谱的煤矸精准识别技术、基于激光三维扫描的放煤量实时监测技术和综放工作面智能化放煤决策软件的应用，现场顶煤厚度探测、混矸率和放煤量的误差分别控制在10.71%、9.32%和7.8%以内，平均每个放煤循环节省时间约30 min，实现了年产1500万t的综放工作面智能高效放煤。

The research focused on addressing various challenges in intelligent top caving theory, intelligent perception and recognition key technology, intelligent caving comprehensive decision-making technology, and remote caving intelligent control technology of fully mechanized top coal caving face. This is being done under the “13th Five-Year” national key research and development plan, specifically designed for the key technology and demonstration of intelligent fully mechanized top coal caving mining method with annual production of 10 million of tons in extra-thick coal seam. The research has resulted in the following outcomes: ① Comprehensive experiments were conducted to understand the interaction process of crushing and migrating of roof and top coal combination (RTCC) and the fragmentation distribution of RTCC under different roof conditions. A three-dimensional laser goaf space detection technology has been developed, and the arching phenomenon of top coal blocks on contact, at coal discharge process with multiple coal discharge ports, has been validated. In addition, the numerical simulation of multi-port intelligent coal caving in extra-thick coal seam is carried out with the constraints of mining and caving coordination, high recovery rate, and low gangue ratio, and the number of coal caving ports is determined. those provide a reliable theoretical basis for optimizing intelligent coal-caving processes. ②The research has explored the geological information and physical characteristics of coal and gangue in the working face, along with full-cycle sensing elements of the top coal caving process. This has led to the development of a comprehensive sensing technology system, including real time detection of top coal thickness, accurate identification of coal and gangue, and dynamic measurement of coal flow, providing crucial data information support for decision making of intelligent coal top caving technology. ③A multi-source information database has been established for the man-machine-environment interface, and a decision-making model has been developed for fully mechanized top caving in extra-thick coal seams. An intelligent coal top caving decision-making software based on the Q-learning algorithm has been created, utilizing artificial intelligence for coal and gangue identification, top coal thickness detection, and coal quantity monitoring. ④A high-precision inertial navigation monitoring and control technology for intelligent fully mechanized caving faces has been developed, enabling real-time positioning, attitude monitoring, and action control for the shearer, hydraulic support, and scraper conveyor. An intelligent mine-integrated communication scheduling system and a remote-integrated control platform for fully mechanized caving have also been established. Those allow for the successful implementation of intelligent coal caving in remote one-button start mode. ⑤Advanced technologies such as ground-penetrating radar for top coal thickness detection, vibration-audio-hyperspectral for coal-gangue identification, and laser three-dimensional scanning for real-time coal caving monitoring are utilized in the 8222 Working Face of Tashan Mine. Intelligent coal caving decision software is applied in fully mechanized caving operations, leading to a control of errors within 10.71% for top coal thickness detection, 9.32% for mixed gangue rate, and 7.8% for coal caving amount. On average, each coal caving cycle now saves about 30 minutes, leading to intelligent and efficient coal caving operations with an annual output of 15 million tons.