Application of Principal Components Analysis and Artifical Neural Networks on Quality Predictionof Effluent in Wastewater Treatment Plant

• Main Authors: Shiu-ping Lin, 林修平
• Other Authors: none
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/31259099742037576676

碩士 === 立德管理學院 === 資源環境研究所 === 94 === The variety of factories and their products in the industrial park has contributed to the dramatic change of influent water quality and influent flow rate of wastewater treatment plant in the industrial park as well as increase the uncertainty of operations management. On the basis of the influent of wastewater treatment plant and the operating conditions of each processing element, the effluent water quality can be calculated accurately and punctually. By this way, operating conditions of wastewater treatment plant can be adjusted or modified in time which is very beneficial to the operations management of wastewater treatment plant. Previous studies in relation with the calculation of effluent water quality of wastewater treatment plant have only applied neural networks and simulation. However, little concern has been paid to the engineering practices. There are multiple and interrelated input parameters pertaining to characteristics of effluent water quality. To consider both perspectives of numerical analysis and engineering practice and the rationality, the present study aims to apply principal component analysis which is widely used to complete the screening of input parameters in the case of the most complicated activated sludge system of wastewater treatment plant. Due to the reason that each component contains all input parameters, such method can present from selecting the main parameters inconsistent with that of real practice, lessen the unnecessary input parameters efficiently and preserve the basic information from original data structure so as to promote the process of back-propagation neural networks. Data were collected and based on the monitoring parameter and operating parameter from wastewater treatment plant of industrial park located in southern Taiwan from August, 2000 to September, 2002. First, principal component analysis was used to gain the new parameters. Then, the hydraulic retention time of each processing element is used to determine the influence order. Through back-propagation neural networks, the calculation of final effluent water quality of activated sludge system of wastewater treatment plant of industrial part is proved to be more effective than the calculation from conventional neural networks. Finally, after improved tests, the historical value of calculation items and the new parameters derived from principal component analysis can be employed to calculate COD of effluent water quality and the concentration of SS. After calculation, the results show that the test correlation coefficients are 0.774 and 0.706 respectively.