Gas Concentration Prediction Based on IWOA-LSTM-CEEMDAN Residual Correction Model

Area of Science:

• Computational intelligence
• Artificial intelligence
• Safety engineering

Background:

• Standard whale optimization algorithm (WOA) suffers from local optima, slow convergence, and low accuracy in coal mine gas concentration prediction.
• Single-factor long short-term memory (LSTM) neural network residual correction models exhibit limitations in accurately predicting gas concentrations.
• Accurate prediction of coal mine gas concentration is crucial for preventing accidents and enhancing safety management.

Purpose of the Study:

• To address the limitations of existing WOA and LSTM models for coal mine gas concentration prediction.
• To develop a novel hybrid model for improved prediction accuracy and enhanced coal mine safety.
• To improve the global search capability and convergence speed of the whale optimization algorithm.

Main Methods:

• Developed an improved whale optimization algorithm (IWOA) by enhancing population diversity and global search capabilities.
• Integrated IWOA with a long short-term memory (LSTM) neural network and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN).
• Utilized CEEMDAN for signal decomposition and analyzed intrinsic mode function (IMF) prediction errors for optimal weight combination.

Main Results:

• The proposed IWOA-LSTM-CEEMDAN model demonstrated superior prediction accuracy compared to BP, GRU, LSTM, WOA-LSTM, and IWOA-LSTM models.
• Achieved prediction accuracy improvements of 47.48% over BP, 36.48% over GRU, 30.71% over LSTM, 27.38% over WOA-LSTM, and 12.96% over IWOA-LSTM.
• The model exhibited the highest accuracy in multi-step prediction tasks for coal mine gas concentration.

Conclusions:

• The IWOA-LSTM-CEEMDAN model effectively overcomes the limitations of traditional methods for coal mine gas concentration prediction.
• The hybrid approach significantly enhances prediction accuracy, contributing to improved coal mine safety management.
• The model offers a robust solution for real-time monitoring and early warning systems in hazardous mining environments.