Abstract: In view of the problems in intelligent system construction of Menkeqing Coal Mine, such as inconsistent interface standards of multiple subsystems, difficulty in data sharing resulting in isolated island, low efficiency in cross-system adaptation and invocation of algorithm models, and relatively slow scene linkage disposal response in areas like gas exceeding limit, ventilation abnormality, drainage malfunction and power supply outage and so on, etc, a four layer integrated platform of "perception-infrastructure-middle platform-application" is constructed based on hyper-fusion and data middlem platform. Taking distributed block storage, resource pooling, and "three-stage type data fusion" (mode registration-stream computing-feature warehouse) as the core, achieving multi-source high-frequency data alignment within 500 ms and hot swap invocation, and implementing company-mining end two-level deployment, uniform interface, and linkage contingency plan. The operation shows that over 30 sets of business and production subsystems achieve 100% interconnection and interoperability, and the interface compatibility rate is improved from 40% to 100%; The daily collection exceeds 300 million records, the cumulative historical data scale exceeds 200 TB. The daily operation and maintenance personnel is decreased from 5 people/day to 1 person/day, the average response time of complex failures is shortened from 90 minutes to 54 minutes, the dispatching efficiency is improved by about 25%, and the closed-loop of major accident hidden dangers is ≤24 hours. The research results show that this platform has good engineering applicability and replicability in terms of system integration, data governance, and linkage management and control.