A new Feature Selection Algorithm using Modified PSO for an Anaerobic Wastewater Treatment System

Wastewater treatment is necessary to preserve the environment and living organisms. COD (Chemical Oxygen Demand) measures the amount of oxygen consumed by the water in the decomposition of organic matter and oxidation of inorganic matter and chemicals. Predicting effluents in the water is a time consuming process. The proposed method uses anaerobic treatment process to utilize anaerobic bacteria (biomass) to convert organic pollutants or COD into biogas in an oxygen free environment. Upflow Anaerobic Filter(UAF) is used as an anaerobic reactor and Cheese whey is used as a wastewater in the UAF. Artificial Neural Networks (ANN) are used as an estimation tool for predicting the performance of filtering treatment process in wastewater treatment plant. Hence the proposed method uses an cost effective techniques such as Dynamic Score Normalization technique with Mahalanobis distance(DSN-M) as a preprocessing step , Modified Particle Swarm Optimization for feature selection process and Adaptive Neuro-Fuzzy Inference System (ANFIS) as a learning algorithm for prediction. The experiment was carried out by comparing dataset without normalization, without feature selection and with ANFIS(Method 1) and dataset with the proposed normalization(DSN-M), feature selection(Modified PSO) and ANFIS (Method 2). The results proved that the proposed method achieves better performance in predicting the COD effluents present in wastewater.