A probabilistic gap based condition prediction approach for desulfurization slurry circulating pump

The Desulfurization Slurry Circulation (DSC) Pump is crucial for controlling gas emissions in limestone-gypsum wet flue gas desulfurization (WFGD) power plants. However, predicting the condition of DSC pumps is hindered by a lack of samples, obstructing field development. This paper proposes a solution using probabilistic gap positive-learning (PGPU) and biased SVM algorithms. Key contributions include: (1) a comprehensive feature model based on expert experience and vibration signal extraction for condition classification, (2) a PGPU and bias-SVM method that updates the model by leveraging probability gaps between true and known samples, and (3) cross-comparisons with other classifiers like SVM and neural networks. Results show the proposed method meets ISO 10816-3 standards and improves accuracy over other methods that don’t account for biased datasets, effectively addressing the state prediction issue for primary WFGD equipment.