Fault detection for NOx emission process in thermal power plants using SIP-PCA.

Area of Science:

• Environmental Engineering
• Data Science
• Chemical Engineering

Background:

• Big data necessitates advanced data-driven models for real-time decision support in pollution emission management.
• Conventional principal component analysis (PCA) has limitations in capturing complex interactions and non-Gaussian distributions in industrial processes.
• Monitoring nitrogen oxides (NOx) emissions from coal-fired boilers is crucial for environmental compliance.

Purpose of the Study:

• To evaluate the usability of a novel data-driven model for monitoring NOx emissions from coal-fired boilers.
• To address the limitations of traditional PCA in handling complex, non-Gaussian process variables.
• To develop an effective system for early fault detection and prevention of exceeding emission standards.

Main Methods:

• Development of a novel survival information potential-based principal component analysis (SIP-PCA) model.
• Utilizing the SIP performance index to improve PCA's information extraction capabilities for non-Gaussian data.
• Employing kernel density estimation for determining control limits in fault detection.
• Application of the proposed algorithm to a real-world NOx emission process.

Main Results:

• The SIP-PCA model successfully extracts more information from non-Gaussian process variables compared to conventional PCA.
• The developed fault detection system effectively identifies potential failures in the NOx emission process.
• The algorithm demonstrated its capability in a real NOx emission monitoring scenario.

Conclusions:

• The proposed SIP-PCA model is a viable and effective tool for monitoring NOx emissions in complex industrial processes.
• Early detection of process failures can be achieved, enabling timely intervention.
• The method aids in preventing NOx emissions from exceeding regulatory standards, contributing to better environmental management.