Anomaly detection of Municipal Wastewater Treatment Plant operation using Support Vector Machine

It is difficult to run a wastewater treatment process (WWTP) stably in the long term. In this work, to monitor the operation state of the treatment process, Support Vector Machine is applied for anomaly detection in Municipal Wastewater Treatment Plant based on operational data. Considering the characteristics of the water quality parameters and relevant regulations, we select the detection vector and choose C-SVM and Radial Basis Function (RBF).Then this paper analysis the parameters optimization of SVM, using the Grid search and Particle Swarm Optimization for model calibration. By comparing the accuracy with 10-Cross Validation of three models, we determine the final classification model. Validation demonstrates the model is able to gain high classification accuracy.