Intelligent prediction of coal mine water inrush based on optimized SAPSO-ELM model under the influence of multiple factors

Abstract

Mine water inrush is affected by many factors such as geological structure and fracture zone. However, there may be overlap among these factors, leading to uncertainty, fuzzy similarity and nonlinear relationship among most of them. Therefore, the traditional mathematical model is not ideal to predict water inrush. This paper proposes an intelligent model for predicting water inrush from coal floor based on simulated annealing particle swarm optimization-extreme learning machine (SAPSO-ELM). Based on 144 groups of learning data and 36 groups of predictive validation data, the proposed model extracted common factors from 14 geological factors that might be related to water inrush in a mining area, so as to reduce information interaction among discriminant indexes. In this paper, simulated annealing particle swarm optimization (SAPSO) is innovatively used to optimize the model parameters and compared with other intelligent models (SVM, BPNN, PSO-ELM and ELM) for the learning prediction of the same data. The results show that the common factors extracted from the original variables contain most of the comprehensive information and can reduce information redundancy. Compared with traditional intelligent models (SVM, BPNN, PSO-ELM and ELM), the proposed model improves the computational efficiency of convergence, and the prediction accuracy is higher. It is proved that SAPSO-ELM intelligent algorithm is indeed scientific and has broad application prospect in result prediction induced by complex multi-factors.