Multi-condition modeling and optimal control of boiler based on improved FCM clustering algorithm

• Main Authors: Xia, Heping, Zhang, Xinyu, Han, Zhongyang, Zhao, Jun, Wang, Wei
• Format: Conference Proceeding
• Language: English
• Subjects: Analytical models; Boilers; Clustering algorithms; Data models; Feature extraction; Fuzzy clustering; Modeling; Operating condition identification; Prediction algorithms; Predictive control; Production
• ISSN: 1948-9447
• DOI: 10.1109/CCDC62350.2024.10587809

Boiler is the main equipment of coal-fired power plant, which has complex and variable characteristics of multiple operating conditions due to different loads in daily production. The overall operational efficiency of the plant is directly affected by its control level, especially for the accommodation of the different operating conditions. In this paper, the improved fuzzy C-mean clustering algorithm is proposed to realize the classification of boiler conditions by extracting the features of boiler combustion history data and modeling each sub-condition using the prediction error method at first. Then a multi-model predictive control method based on T-S fuzzy theory is designed to control key variables such as active power in the boiler. In experimental study, real data is used to comparing the proposed with other commonly used methods, such as the conventional Fuzzy C-means clustering algorithm. The results show that the method proposed in this paper has obvious advantages in boiler modeling accuracy and control performance, which verifies the reasonableness for boiler condition identification as well as the practicability for real-world application.