Development of big data intelligent early warning platform for coal mine water hazard monitoring based on cloud service

A microseism-electric coupling system to alert the coal disaster of water inrush has been proposed,aimed at North-China-Type coalfields and based on the theory of Down Three Zones. Firstly,a migration subsystem to collect o- riginal target data by using Flume has been designed,enabling the effectiveness and instantaneity of data preprocess- ing. It also aggregates and transmits these data from raw signals to a database system. This database as a storage center employs Spark and HDFS to store this large-scale,real-time,multi-source,heterogeneous,and spatial-temporal data, and to possess the TB-level capacity. This storage center has a unified multi-source sequential storage architecture to support saving big data by using MapReduce to process the data in parallel and YARN to dispatch and manage the re- source efficiently. It also leverages Spark Streaming to design real-time and high-efficiency data processing with an in- telligent warning module and a remote service interface to serve appli-cations. By performing the migration subsystem and storage center to obtain processed coal-water inrush data,a Long Short-Term Memory ( LSTM) model is used to predict the disaster of water inrush using the theory of Down Three Zones corporately. In the semi-supervised training of this model,the microseism-electric data are inputted,and then the model’s output is the alert level that divides the coal disaster of water inrush into four ranks. The LSTM model can warn the level of water inrush automatically and show the visualization of alert information. In practice,the system can provide timely and effective warning capabilities for east well in Gequan coal mine,Jizhong Energy Group,China.