Study on Slagging Characteristics of Boiler Pre-combustion Chambers Based on Deep Learning

Area of Science:

• Thermal Engineering
• Artificial Intelligence
• Combustion Science

Background:

• Pre-combustion chambers (PCC) in coal-fired boilers face slagging issues due to high heat loads, impacting safe operation.
• Existing AI models for boiler slagging prediction use limited parameters, neglecting complex flow and combustion dynamics.

Purpose of the Study:

• To develop a fast and accurate prediction model for boiler wall slagging degrees.
• To improve the precision of slagging prediction by incorporating complex operational and structural parameters.

Main Methods:

• Simulated boiler combustion processes under diverse conditions to generate a comprehensive dataset.
• Validated numerical simulations with experimental data for typical operating conditions.
• Employed a deep parallel residual convolution neural network (DPRCNN) for slagging degree identification.

Main Results:

• The DPRCNN model achieved 100% accuracy, precision, and AUC in identifying three types of boiler wall slagging.
• The study demonstrated the effectiveness of deep learning in predicting boiler slagging.

Conclusions:

• The proposed DPRCNN model offers a highly accurate and efficient solution for boiler slagging prediction.
• Deep learning technologies are applicable and effective for enhancing boiler operational safety and stability.