data analytics, Data to Insight, Process Modeling

Using Data to Optimize Processes: Part IV—Minimizing Total THM Formation in a Distribution System

Here is the latest post in the series “Using Data to Optimize Processes,” by Ruby Daamen and John Cook from ADMI.

As has been demonstrated from previous articles in this series, data can be used to build process models for both natural systems such as rivers, estuaries and groundwater movement over expansive areas, but can also be used to model man-made systems such as water treatment plants (WTP). The last article featured optimizing THM removal in a water treatment plant.  This article will address optimizing total THM (TTHM) formation in the distribution system. (The same approach can be used for HAA5 or other DBPs.)

Model for TTHM Formation on Distribution, THMD1

Based upon the most critical sites in which TTHM levels have been the highest, the TTHM data was obtained from ten locations.  Where multiple samples for TTHM were obtained from a single site using a split sampling technique, the samples were averaged.  A stacked dataset was then created in which the time-series of data is repeated for a given static parameter, in this case, the water age at each sample site.   

The resulting model predicted distribution TTHMs as a function of:


in which:

CL2 = free chlorine concentration leaving the clearwell

PTTHMACW = modeled TTHM formation leaving the clearwell

WATERAGE = water age at any particular site on the distribution system

TMPDEC = temperature, de-correlated

The resulting model, shown in Figure 1, yielded an R2 = 0.76.  Figure 2 shows the response surface with TMPDEC at its minimum value (left plot) and maximum value (right plot) and water age at its mean.  The effects of water age on the distribution system are shown in Figure 3.

Figure 1. Predicted distribution system TTHM (red) vs. actual THM levels (blue) vs. water-age (green)

Figure 2. Predicted distribution system TTHM levels given modeled clearwell THM and measured Cl2 levels with hidden mean water age and maximum/ minimum de-correlated temperature

Figure 3. Predicted TTHM levels on distribution system given water age, from 27 to 425 hours; Hidden variables include mean water temperature and mean Cl2.

Distribution TTHM Mathematical Experiment No. 1

The design goal for the example WTP is that the 365 day running average of the TTHM value at any critical sampling location on the distribution system not exceed 64 µg/L, as opposed to the USEPA standard of 80 µg/L.  Figure 4 shows the measured TTHM, predicted TTHM and predicted TTHM averaged over 365 days at YD027, the site with the longest estimated water age, and YD030, one of the sites with the shortest estimated water age.  Historically YD030 has always met the goal, whereas YD027 has never met the goal.  A mathematical experiment was run to check the feasibility of meeting the goal under modified operations.

In part 1 of the experiment, the TTHMDIST model was run with historical water age, CL2CW, and TMPDEC.  Pred_TTHMACWUSER (a constant user set TTHMACW) was set to the historical minimum and then increased incrementally to the historical mean.  Results suggest that a clearwell TTHM of 30 µg/L or below will allow site YD027 to always meet the stated goal, with a goal of 35 µg/L meeting the goal for all intents and purposes.

In part 2 of the experiment, the TTHMCW2 model was run with historical RAIN (rainfall) and TMPDEC data.  The remaining inputs were user set as follows:

  • CL2DOSE at historical minimum of 1.79 mg/L
  • CL2CW-CL2FIN_R at historical maximum of 0.92
  • TURBREM-FLT-SET at historical maximum removal of 97%
  • ALKESET to its 95thpercentile value of approx. 4 and mode of approx. 3 mg/L
    • [The historic minimum of 0.57 was not thought to be a reasonable goal]

The results show that theoretically the goal could be reached if all assumptions are met.  However, the treatment process would have to be run at optimal process control.

Figure 4. Measured TTHM, predicted TTHM, predicted TTHM 365 day MWA and water age for sample sites YD027 and YD030.

Figure 5. YD027 Predicted TTHM – 365 day MWA for using historical data, and a user predicted clearwell TTHM of 23, 30 35, 40 and 53 µg/L.

Figure 6. Two important clearwell models yield THM results around 30 µg/L. Predicted clearwell THM results with alkalinity = 4 mg/L (pink); and predicted clearwell THM results for alkalinity = 3 mg/L (green), plus other variables.

Figure 7 shows the measured distribution TTHM and measured and predicted 365 day MWA distribution TTHMs at each of the 10 locations.  Only 3 of the 10 sites have historically met the goal of 64 µg/L (namely, sites YD015, 30 and 60).

Figure 7. Measured TTHM (blue), measured and predicted TTHM 365 day MWA (red and pink) and water age (light gray) for all ten sample sites. Straight red line shows TTHM goal of 64 µg/L.

Figure 8 shows the historical chlorine residual at the clearwell vs. the historical chlorine dose.  Chlorine residual is also shown as a 30 day moving window average to remove high frequency variability and clearly show the trend.  Clearwell chlorine residual has maintained an average of 1 mg/L over the entire historical period.

Figure 8. Chlorine residual (blue), 30 day MWA of chlorine residual at clearwell (red) and chlorine dose (green). While the chlorine dose has been reduced, chlorine residuals have averaged 1 mg/L over the entire historical period.

Distribution TTHM Mathematical Experiment No. 2 (Site YD054)

The design goal for the example WTP is a 365 day running average not to exceed 64 µg/L TTHM at any critical sampling location on the distribution system.  However the USEPA standard is 80 µg/L and historically the most critical sampling site is YD054.  A mathematical experiment was conducted to check the feasibility of meeting the standard of 80 µg/L for site YD054 under current operations.  For this experiment the TTHM formation at Site YD054 was modeled independently and therefore did not include waterage.    In Figure 9, it is shown that site YD054 will remain in compliance with the standard with a clearwell TTHM value somewhere in the range of 35 to 40 µg/L.

Figure 9. TTHM model for Sampling Site YD054. In general very accurate, model does not capture extreme TTHM event in 2009. However, it is suspected that this extreme event may be due to analytical error.


About noahmorgenstern

Entrepreneurial Warlock, mCouponing evangelist, NFC Rabbi, Innovation and Business Intelligence Imam, Secular World Shaker, and General All Around Good Guy


No comments yet.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: