碩士 / 朝陽科技大學 / 環境工程與管理系碩士班 / 91 / Activated sludge processes have been the most widely used procedures for biological wastewater treatment processes. So far, many studies have estimated the effluent quality in procedures. However, they mainly focus on the study of using single model to estimate effects, and few of them use various models to compare the estimated effects of single system. The primary goals of the study are: (1) to probe into the effects that Activated Sludge Model No. 2d and backpropagation neural network respectively estimate and predict the effluent quality in the TNCU procedure (2) to probe into the effects that backpropagation neural network and grey prediction respectively estimate the effluent quality of various types of wastewater treatment plants (3) to compare the effects that the three different modeling ways mentioned-above respectively predict the effluent quality in biological wastewater treatment processes and the differences among them.
The results of this study indicate that in TNCU plant ASM2d has excellent effects to simulate the effluent quality in the TNCU procedure. The simulated value for all ingredients in the procedure is identical to the experimental value, with the correlation coefficient above 0.96.In addition to simulating the dissoluble ingredients, it also can be utilized to simulate the granular ingredients in the procedure. Neural network has better estimated effects on the effluent of SS and SNO3 in the TNCU procedure. The correlation coefficient between estimated value and experimental value is respectively 0.95 and 0.98, and Mean Absolute Percentage Error (MAPE) is respectively 19.79% and 6.84%. However, it has poor prediction effects on the effluent of SNH4 and SPO4. The correlation coefficient is respectively 0.62 and 0.97, and MAPE is respectively 138.26% and 95.34%.
With regard to the three wastewater treatment plants, neural network has the best prediction effects on Plant B, and Plant C is the second while Plant A is the worst. Besides, according to the results of sensitivity analysis, the pH value and temperature of the influent has a significant influence on the effluent quality in these three plants, and consequently, becomes the important reference for the operation of the wastewater treatment plants. Four types of Grey Model respectively estimate the effluent quality in the three plants, and the GM(1,1) has the best estimated results.
ASM2d can simulate the dissoluble ingredients and the granular ingredients, but the modeling process requires experiments to calibrate and validate, rather than large amounts of data. Neural network requires large amounts of data to establish the input-output relation, but it fails to describe clearly the mass flow and transformation mechanisms in the ingredients in the procedure. Grey Theory only requires more than four output system data to operate the grey prediction, but it also fails to describe clearly the inner various transformation procedures or mechanisms in that system.
| Identifer | oai:union.ndltd.org:TW/091CYIT5087014 |
| Date | January 2003 |
| Creators | Chia-Ho Tsai, 蔡嘉和 |
| Contributors | Tzu-Yi Pai, 白子易 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 130 |
Page generated in 0.0018 seconds