Big data analysis and prediction model in intelligent dispatching system of coal yard

This paper discusses the big data analysis and prediction model in the intelligent coal yard dispatching system, aiming at solving the problems of low efficiency and unreasonable resource allocation caused by traditional coal yard dispatching relying on manual experience. By building an overall architecture based on Internet of Things, cloud computing and artificial intelligence, the system realizes real-time collection, storage and processing of multi-source data. The research focus is to use big data analysis technology to mine the value information behind the coal yard operation data, and combine the time series prediction model and machine learning model to accurately predict key indicators such as coal demand, inventory and quality. The experimental results show that the Prophet-LSTM hybrid model excels in demand forecasting, with a Mean Absolute Error (MAE) of 152.9 tons, a Mean Absolute Percentage Error (MAPE) of 7.3%, and a Root Mean Square Error (RMSE) of 189.4 tons, significantly outperforming traditional ARIMA models as well as standalone LSTM and Prophet models. Meanwhile, the inventory optimization strategy based on linear programming effectively reduced inventory fluctuations, with a standard deviation decrease of approximately 43.6%, and maintained better stability and safety during extreme weather conditions. The improved Isolation Forest algorithm demonstrated higher accuracy (94.7%), lower false alarm rates (3.6%), and shorter response times (1.9 seconds) in safety monitoring, significantly enhancing the system's real-time monitoring and response capabilities.