Improved neural network with least square support vector machine for wastewater treatment process

Abstract

This research offers a unique interval by using the predicting approach for discharge indicators of water quality data such as biochemical oxygen demand (BOD) and ammonia nitrogen (NH3-N). This is considered one of the significant quality metrics in wastewater treatment plants for water quality management as well as surveillance. To begin, the effluent information for BOD/NH3-N and their supplementary parameters are gathered. Hence BOD and NH3 are considered major feature sources for estimating water pollutants. BOD is high then oxygen level is very low in the water due to pollutants or algae. Ammonia nitrogen is an organic waste component in water from sewage. The significant characteristics with good correlation levels of BOD and NH3-N are examined and identified using a grey correlation analysis method after certain basic data pre-processing procedures. The BOD/NH3-N effluent information of a water treatment plant is predicted using an upgraded feed-forward neural network with the least square support vector machine (FFNN-LSSVM) method. An optimization approach for an enhanced feed-forward neural network (IFFNN) is built by Machine Learning Algorithms. The IFFNN used regular influent water quality, influent rate of flow, and Wastewater performance monitoring and operational conditions as input parameters. For future prediction, input variables were previous different wastewater quality measurements. Lastly, the analysis shows that, when compared to other current algorithms, the proposed methodology can forecast wastewater quality of water with high accuracy in predicting BOD and NH3 levels, limited computation duration, mean error less than 10% and R² is 90% proves better than existing techniques.

Keywords: Accuracy; Intervals; Machine learning; Neural networks; Optimization; Prediction; Wastewater management; Water quality.