Improved Modelling of Wastewater Treatment Primary Clarifier Using Hybrid Anns

Abstract

Abstract-This paper presents the results of modelling to predict the effluent biological oxygen demand (BOD 5 ) concentration for primary clarifiers using a hybridisation of unsupervised and supervised artificial neural networks. The hybrid model is based on the unsupervised self-organising map (SOM) whose features were then used to train a multi-layered perceptron, feedforward back propagation artificial neural networks (MLP-ANN). In parallel with this, another MLP-ANN was trained but using the raw data. A comparison of the outputs from the two MLP-ANNs showed that the hybrid approach was far superior to the raw data approach. The study clearly demonstrates the usefulness of the clustering power of the SOM in helping to reduce noise in observed data to achieve better modelling and prediction of environmental systems behaviour.

Review of Latest Advances in Nature-Inspired Algorithms for Optimization of Activated Sludge Processes

“…Results showed the hybrid model outperformed the ANFIS model in predicting the necessary values [8]. KSOM was used to extract features from noisy data and fill in missing values [26][27][28][29]. Three years of data were taken from a WWTP in Edinburgh, UK, and two models were tested, one with ANFIS alone and another with the hybrid KSOM-ANFIS [27].…”

Anaerobic digestion process modeling using Kohonen self-organising maps

“…This is due to the aspects of KSOM in dealing with such uncertainty in a similar way of human thinking and its power of dealing with missing values. KSOM has been successfully used to model activated sludge wastewater treatment plant (Asadi et al., 2017; Begum et al., 2016; Liukkonen et al., 2013; Machón-González et al., 2017; Rustum and Adeloye, 2007; Rustum, 2009; Rustum and Adeloye, 2013a, Rustum and Adeloye, 2013b; Rustum et al., 2008; Rustum and Forrest, 2017; Szelag et al., 2017). Thus, this inspired the authors to try model the anaerobic system using the same approach.…”

Application of Selected Methods of Black Box for Modelling the Settleability Process in Wastewater Treatment Plant

“…Considering the fact that the methods of classification are used in conjunction with models of regression and improve the ability of predictive mathematical models, which is why they are being increasingly used in the discussion of practical documented through publications [11,12,17]. The review of the literature shows [10,18], that these analyzes are made by hybrid models which are usually a combination of self-organizing artificial neural network with another type of neural network (probabilistic, multi-layer perceptron, recurrent).…”

“…In these models the first stage of calculating training set, using a suitable algorithm discriminatory type of method k -medium [8], c -medium [9], self-organizing neural networks [10], is divided into appropriate classes, within which they are made predictions based on independent models. In many cases [11,12], this approach can improve the ability of predictive mathematical model, but in a situation of extremely diverse number of data in each class or insufficient input data there may be problems at the stage of learning and testing statistical models.…”