Evolutionary self-organising modelling of a municipal wastewater treatment plant

Application of Selected Methods of Artificial Intelligence to Activated Sludge Settleability Predictions

“…Conversely, the genetic programming method shows the worst predicting abilities for the inputs mentioned above. In GP, settleability is expressed as follows: (10) and the values of mean absolute and relative errors are equal to: MAE = 43.21 cm 3 /dm 3 and MAPE = 23.18%, respectively. ...where: Q(t-1, t-2), T(t-1, t-2), SE(t-1) -flow to the treatment plant, wastewater temperature, and settleability measured in a time step Δt from the predicted value of the technological parameter, Q(t), T(t) -inflow and temperature specified at the same time as SE(t).…”

“…Consequently, to model the processes going on in the individual parts of the treatment plant, i.e. degree of pollutant load reduction (suspended solids, biogenic compounds), to optimise the operation of the treatment plant parts (determination of oxygen amount in activated sludge tanks, the electrical energy used in pumping stations), and also to produce biogas, artificial neural networks [6][7][8][9], genetic programming [10], Support Vector Machines (SVM) [11][12][13], autoregressive models [14], regression trees [15][16], and others are used.…”

Influential Parameters on Biomethane Generation in Anaerobic Wastewater Treatment Plants

“…Using this technique, the system could be adapted and operated in a variety of conditions showing a more flexible performance (Zeng et al, 2003). Hong et al, 2003). KSOFM differ from feedforward networks in a way that they provide a clustering methodology which leads to data reduction (Kasabov, 1998) and also project the data nonlinearity onto a lower-dimensional display.…”

“…Regarding the maximum coefficient of correlation (R) and the minimum mean absolute percentage errors (MAPE) of the predictions, the developed model showed a satisfactory performance in comparison with the pure ANN models. Therefore, the model developed could be recommended in order to optimize design considerations of the treatment process (Pai et al, 2009 (Zhu et al, 1998), optimal control of a wastewater treatment process integrated with PCA (Choi and Park, 2001), Kohonen Self-Organizing Feature Maps (KSOFM) to analyze the process data of municipal wastewater treatment plant (Timothy Hong et al, 2003), Unsupervised networks for modeling the wastewater treatment process (Garcia and Gonzalez, 2004;Hong and Bhamidimarri, 2003;Cinar, 2005), Grey Model ANN (GM-ANN) to predict suspended solids (SS) and COD of hospital wastewater treatment reactor effluents (Pai, 2007), on-line monitoring of a reactor (Luccarini, 2010 (Chen, 2003), control and supervise the submerged biofilm wastewater treatment reactor , modeling the nonlinear relationships between the removal rate of pollutants and their chemical dosages in a paper mill wastewater treatment plant Table 5 summarizes the aforementioned models together with the advantages and drawbacks which might be considered for selection in applied projects and utilization in industrial scales. As observed the most extensively used model is ANN.…”

Selección de tecnologías para el tratamiento de aguas residuales municipales

“…Estas plantas son utilizadas en la remoción de contaminantes presentes en el agua residual cruda, y deben responder a una alta variabilidad temporal del flujo o caudal de entrada, así como a la variabilidad de las concentraciones o componentes propios de estas aguas residuales. Esto requiere una interacción en los mecanismos biológicos, físicos y químicos entre los procesos unitarios, los fenómenos hidrodiná-micos y la adaptabilidad del consorcio microbiano ante las condiciones ambientales cambiantes (Hong, 2003;Bdour A, 2009).…”