Novel Approach to Hydraulic Support Pressure Prediction by EMD-Transformer

Hydraulic support pillar pressure prediction has been a pivotal basis for decision-making during the mining process. It has been one of the fundamental pieces of information for ensuring the stability of the surrounding rock. However, the pressure of hydraulic support pillars follows specific patterns; it cannot be predicted using simple mathematical models. Additionally, during the mining process, issues such as the support attaching and detaching the roof, roof fragmentation, and sensor detection errors introduce a significant amount of random perturbance, turning the pressure data into a non-stationary time series. Non-stationarity significantly complicates the pressure prediction. Based on the Transformer model, this paper proposes an EMD-Transformer model. Empirical mode decomposition aims to decompose non-linear non-stationary data pressure data into intrinsic mode functions, and the transformer model establishes detailed correlations of all components and projects predicted components. The aggregate sum of predicted components gives pressure prediction. Comparative experiments on a real coal mine support pillar dataset showed that the EMD-Transformer model proposed in this paper achieved notably improved performance on all prediction lengths compared to LSTM and vanilla Transformer model.