The series on developing process models continues with this article from John Cook and Ruby Daaman of ADMI on optimizing the removal of total organic carbon TOC through a conventional sedimentation process. This case study is based on a 24 MGD conventional WTP with a reservoir source of supply and highly variable TOC.
TOC Historical Data
Provided below in Figure 1 are the TOC data over the historical period of interest, designated in the Figure as TOCRAW for the raw water TOC concentration; TOCESET as the TOC from the east-side sedimentation basin; TOCWSET from the west-side sedimentation basin; and lastly TOCREM, the TOC removed through the sedimentation process presented as a percent removal. It should be pointed out that over the period of study, the historical high in TOC removal was 63 percent (%) and the mean was 37% removal.
Model 1 for TOC Removal, TOC1
Using artificial neural networks and spectral analysis, the TOC1 model includes additional water quality parameters in an attempt to further explain TOC removal variability. TOC1 includes alum, coagulation pH (PHSET), raw water alkalinity (ALKRAW), color, hardness (HARD) and a de-correlated water temperature signal (TMPDEC), R2 was boosted to 0.62 (Figure 2). The plot of the surface response diagrams (Figure 3) demonstrate the sensitivity to coagulation pH of TOC removal whether the hidden parameters are held at their minimum or maximum historical values.
Model 2 for TOC Removal, TOC2
Model TOC2 includes source water (Q) and rainfall inputs. Rainfall and Q are correlated to one or more of the water quality parameters and ALUM. For this reason a “super” model was created made up of 2 cascading models (Figure 5). Rainfall inputs include moving window averages of various sizes, decorrelated from each other by computing differences between successive window sizes. Figure 6 shows the modeling results for TOC2 as percent TOC removal.
Summary of TOC Removal Model Performance
A number of additional TOC removal models were developed and their results plotted below (Figure 7). It can be observed that all models are generally very good at capturing TOC removal outside of the low extremes. Capturing the low extremes proved to be more difficult but this may be related to measurement error given the limited number of instances in which the models were not able to capture the actual behaviors.
TOC Removal and pH Sensitivity—a Mathematical Experiment
The full historical data set shows that the minimum, mean and maximum coagulation pH to be 5.0, 5.8 and 6.9, respectively. To evaluate the sensitivity of TOC removal with respect to pH, three plots are made using the best TOC model and setting the pH at the model’s data set minimum, mean and maximum of 5.25, 5.84 and 6.6, respectively. These values were selected as reasonable minimum, mean and maximum pH values for coagulation. Optimal TOC removal occurred at a coagulation pH = 5.3.
Each graph (Figure 8) shows the approximate alum dose at which 45 percent and 50 percent TOC removal is achieved. In Figure 9, 50% is never achieved (the TOCREM goal of 38.5% for the alkalinity using 6.6 mg/L would be achieved). The other inputs are set to their mean value as indicated in Figure 9. The conclusion demonstrates that the lowest pH requires less alum to reach the desired TOC removal percentages.
If you have any questions you would like to address to John Cook, you may email him personally at firstname.lastname@example.org