Scholarlink

<p>High gas concentration in coal mine is easy to lead to explosion and poisoning accidents, and timely and accurate prediction can effectively prevent such accidents. A new prediction method is proposed to improve the accuracy of prediction coal mine gas concentration, which combines Variational Mode Decomposition (VMD) with Long Short-Term Memory (LSTM) to mitigate the impact of the complexity on prediction accuracy. Firstly, the penalty factor and the modes number of VMD are optimized using the Particle Swarm Optimization. Subsequently, the lengths of one-step prediction windows for different Intrinsic Mode Functions (IMFs) after VMD decomposition are determined through autocorrelation and partial autocorrelation analysis. LSTM one-step prediction models are built for different IMFs to obtain prediction results. Finally, by reorganizing the dataset to construct an error sequence, a VMD-LSTM model is established on the error sequence to perform a secondary decomposition and predict the errors at future time steps. The prediction results and the error prediction results are combined to obtain the final prediction results for the test set. Experimental results show that the proposed method achieves the highest prediction accuracy compared to the other four methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Coal Mine Methane Gas Concentration Prediction Based on PSO-VMD-LSTM and Error Quadratic Decomposition

  • 期刊论文

  • 发表日期：

    2025年1月31日

查看全文
  • Qingsong Hu,
  • Yuanxun Cheng,
  • Shuo Zheng,
  • Die Zhao,
  • Shiyin Li,
  • Yanjing Sun,
  • Yuansheng Zhang

摘要

High gas concentration in coal mine is easy to lead to explosion and poisoning accidents, and timely and accurate prediction can effectively prevent such accidents. A new prediction method is proposed to improve the accuracy of prediction coal mine gas concentration, which combines Variational Mode Decomposition (VMD) with Long Short-Term Memory (LSTM) to mitigate the impact of the complexity on prediction accuracy. Firstly, the penalty factor and the modes number of VMD are optimized using the Particle Swarm Optimization. Subsequently, the lengths of one-step prediction windows for different Intrinsic Mode Functions (IMFs) after VMD decomposition are determined through autocorrelation and partial autocorrelation analysis. LSTM one-step prediction models are built for different IMFs to obtain prediction results. Finally, by reorganizing the dataset to construct an error sequence, a VMD-LSTM model is established on the error sequence to perform a secondary decomposition and predict the errors at future time steps. The prediction results and the error prediction results are combined to obtain the final prediction results for the test set. Experimental results show that the proposed method achieves the highest prediction accuracy compared to the other four methods.