Ansteel Group

The transmission of manual coal blending experience is challenging and heavily relies on individual expertise. In the plant, the manual experience of coal blending engineers cannot be effectively passed down and can only be handed over by senior workers training new employees. New coal blending engineers are slow to get started independently and are increasingly difficult to recruit. Additionally, the cost of raw coal is high, traditional coal blending values are conservative, and quality indicators are often excessive, increasing coking costs. The quality is unstable, and manual experience can only provide rough estimates of coal blending and coking results, making fine calculations impossible. The results of coal blending and coking are highly dependent on personal experience, making it difficult to accurately predict coke quality.

The coal blending process takes too long; due to the complexity of coal quality information indicators, coal blending engineers spend a lot of time meticulously planning to adapt to new raw coal.

The Pangu Large model, based on the estimated ratio of raw coal, accurately predicts coke quality indicators while ensuring coke quality. By optimizing coal blending parameters, it reduces costs, lowering the average coal blending cost by over 5 RMB. Combining artificial intelligence with expert experience helps coal blending engineers improve their coal blending skills, expand new ideas, and ensure effective transmission of experience. It helps improve the output ratio efficiency, reducing the manual time from 1-2 days to 1-2 minutes. The accuracy of coke quality predictions has increased from 90% to over 95%.