Integration of machine learning with complex industrial mining systems for reduced energy consumption

Area of Science:

• Mining Engineering
• Artificial Intelligence
• Industrial Control Systems

Background:

• Deep-level mining faces declining profitability due to rising costs and reduced output.
• Implementing Internet of Things (IoT) technologies is challenging in underground mines owing to complex, integrated systems and lack of dynamic control.
• Refrigeration and ventilation systems are critical but often controlled independently, hindering energy efficiency and production optimization.

Purpose of the Study:

• To develop and compare machine learning (ML) prediction techniques for the integrated behavior of a key component in the refrigeration-ventilation system.
• To address the challenge of lacking continuous measurements for critical components.
• To evaluate ML models based on accuracy, prediction time, and data requirements.

Main Methods:

• Development of industrial machine learning models to predict system behavior without direct sensor data.
• Comparison of various ML techniques, including Support Vector Machines (SVM) and Artificial Neural Networks (ANN).
• Evaluation of model performance based on prediction accuracy, speed, and data volume.

Main Results:

• The Support Vector Machines (SVM) method achieved the lowest average error at 1.97%.
• The Artificial Neural Network (ANN) method demonstrated greater robustness with a maximum percentage error of 12.90%.
• Potential annual energy savings of 215 kW (2.9%) for the ventilation and refrigeration system, equating to R1.33 million ($82,900), are achievable with SVM.

Conclusions:

• Machine learning models can effectively predict the integrated behavior of critical mining components despite data limitations.
• The SVM approach shows significant promise for optimizing energy consumption in underground mining's refrigeration-ventilation systems.
• AI-driven predictive modeling offers a viable strategy to enhance efficiency and profitability in the deep-level mining industry.