Abstract:

Highly imbalanced data for fault diagnosis in wastewater treatment process seriously affects fault diagnosis performance, especially in identification of faulty classes. Reduced recognition accuracies of faulty classes may lead to occurrence of other issues, such as failure to reach quality standard of effluent water, high operation cost and secondary pollution. An ensemble weighted extreme learning machine method (WELM) for imbalanced learning was proposed for fault diagnosis modeling in wastewater treatment process. AdaBoost ensemble classification algorithm based on WELM base classifiers was integrated into assessment index G-mean of imbalanced classification. New updating rules for initial weight matrix in the base classifiers and ensemble weight formula were defined for iterative learning of the base classifiers. Simulation results show that this fault diagnosis model of wastewater treatment process can improve classification performance, such as G-mean value, overall classification precision, and recognition accuracy of faulty classes. The proposed method is effective in imbalanced fault diagnosis of wastewater treatment process.

Key words: weighted extreme learning machine, AdaBoost ensemble algorithm, imbalanced learning, wastewater treatment, fault diagnosis, modeling

摘要：

污水处理过程的故障诊断数据具有高度不平衡性，影响了故障诊断效果，尤其是降低故障类别的识别正确率，导致出水水质不达标、运行费用增高和环境二次污染等问题出现。据此提出一种基于加权极限学习机集成算法的污水处理故障诊断建模方法。该方法将不平衡分类评价指标G-mean引入以加权极限学习机为基分类器的AdaBoost集成分类模型，定义新的基分类器初始权值矩阵更新规则和集成权重计算公式，用于基分类器的迭代学习。由仿真实验结果可知，基于加权极限学习机集成算法的污水处理故障诊断模型，可有效提高分类性能G-mean值和整体分类精度，特别提高了故障类的识别正确率，验证了基于加权极限学习机的集成算法在不平衡性污水处理故障诊断问题上的有效性。

关键词: 加权极限学习机, AdaBoost集成算法, 不平衡学习, 污水处理, 故障诊断, 模型