Design methodology to determine the water quality monitoring strategy of a surface water treatment plant in the Netherlands

The primary goal of a drinking water company is to produce safe drinking water fulfilling the quality standards defined by national and international guidelines. To ensure the produced drinking water meets the quality standards, the sampling of the drinking water is carried out on a regular (almost daily) basis. It is a dilemma that the operator wishes to have a high probability of detecting a bias while minimizing their measuring effort. In this paper a seven-step design methodology is described which helps to determine a water quality (WQ) monitoring scheme. Besides using soft sensors as surrogate sensors for parameters currently not available online, they can possibly provide a cost-effective alternative when used to determine multiple parameters required through one single instrument.

Primary goal of a drinking water company is to produce safe drinking water fulfilling the quality standards defined 23 by national and international guidelines. To ensure the produced drinking water meets the quality standards, in the 24 Netherlands, sampling of the drinking water is carried out on a regular (almost daily) basis.

26
Common practice in the Netherlands is that (drinking) water treatment plants (WTPs) are designed in such a robust 27 way that the effluent quality can be guaranteed without direct control on the incoming water quality (Vanrolleghem surface water as a source, experience increased pollution in the form of organic micropollutants and increased taken at a regular interval to check the on-line measurements and that the produced drinking water meets the 1 quality standards set by national and international guidelines. However, besides the rapid tests performed at Site, 2 the time between sampling and laboratory results takes at least one day. This delay in results and interval between 3 measurements makes it difficult to only use the laboratory measurements for real-time control of a treatment plant 4 (van de Ven et al., 2010). 5 6 Retrieving reliable and robust on-line information is therefore important in order to be able to control a WTP. This 7 information can be retrieved from on-line sensors that measure a specific parameter directly, but also from generic 8 sensors that give indirect information. Roccaro et al. (2008), Rieger et al. (2004) and van den Broeke et al. (2008) 9 showed the ability of UV-Vis spectra measurements, measuring the absorbance of ultraviolet or visible light, to 10 estimate different parameters such as chlorine decay, nitrite and nitrate, ozone and assimilable organic carbon 11 (AOC) concentrations. These estimations were derived from algorithms developed, based on a change in UV-Vis 12 absorbance during a treatment step and laboratory measurements, using principal component analysis followed by 13 partial least squares regression. These types of generic sensors are so-called soft-sensors, sensors that require 14 software to give the required information. Juntunen et al. (2013) developed a soft-sensor to predict the turbidity in 15 treated water and to find the most significant variables affecting turbidity. 16 17 18 Optimized control can only be reached if there is a high probability of detecting a bias in the operation of the WTP.

19
At the same time, from an economical perspective, the data should be obtained with minimal measuring efforts 20 and costs. Understanding the requirements with respect to on-line monitoring and data reliability is a first step 21 towards direct control of the drinking water production based on the incoming water quality. Therefore, in this 22 paper a design methodology is described which helps to develop a water quality monitoring scheme. This will be 23 explained by means of a case study for the WTP Weesperkarspel in the Netherlands.     45 The treatment step objectives depend on the feed water quality and the type of treatment step considered. The 46 overall objective of a drinking water treatment plant is the production of safe drinking water fulfilling the quality 47 standards defined by national and international guidelines. The main objective of a treatment step for an existing 48 plant should be the focus on water quality and less on the chemical or energy consumption (van der Helm et al.,

49
2008b). Therefore it should be evaluated which parameters, present in the feed water quality, can be influenced per treatment step. In order to do so process knowledge on the different treatment steps is indispensable (Poch et 1 al., 2004). Van Schagen (2009) indicated that mathematical models are a powerful tool to evaluate the sensitivity 2 to process objectives and disturbances and help find the appropriate controlled variables.

4
Operational control options 5 Depending on the design of the treatment step certain operational control options are available to make changes to 6 the treatment process. Examples of operational control options are the change in chemical dosage, flow division 7 and backwash and regeneration frequency. The primary focus is on the operational changes that can be performed 8 within the existing plant lay-out.

10
Required water quality parameters 11 Based on the treatment step objectives and existing operational control options, the water quality parameters that 12 are influenced by the treatment step are determined. Ideally these water quality parameters should be monitored.

13
Besides the water quality parameters that are influenced by a treatment step, there are water quality parameters 14 that influence the efficiency of a treatment step. For example, the water temperature has an effect on the ozone 15 decay rate. The decay rate increases with increasing temperatures (Elovitz et al., 2000). This may result in a higher 16 required ozone dose in summer time, taking into consideration that the disinfection requirements are also different 17 with different temperatures.

19
Process characteristics 20 The required monitoring frequency and sensitivity of the selected water quality parameters may also vary

26
For example, ozone typically degrades quickly in water due to the reaction with organic compounds in the water.

27
This determines that the required measurement frequency should be high.

29
Evaluate available measurements for the identified water quality parameters 30 Based on the evaluation of the required water quality parameters and existing process characteristics the available 31 (on-line) measurements should be evaluated. A wide range of methodologies exist for determining water quality 32 parameters, from certified laboratory measurements to on-line measurements. Depending on the variability of the 33 process, the turnaround time of laboratory measurements is not always fast enough. To come to an optimal water 34 quality monitoring scheme also on-line water quality sensors should be considered. In this study the following 35 evaluation criteria for the available on-line sensors were assessed: 36 Easiness; is the sensor easy to use, is the measuring principle easy to understand; 37 Sensitivity; is the method sensitive enough; 38 Maintenance; does the sensor require much maintenance;

39
Costs for laboratory measurements as well as the purchasing and maintenance costs for on-line sensors were 40 indicated. Besides on-line sensors developed to measure one specific parameter, available surrogate sensors, used 41 to estimate a water quality parameter value, and soft-sensors were assessed.

44
The individual monitoring strategy defines which water quality parameters per treatment step should be monitored, 45 with a selected frequency and location. The evaluation, of available measurements for the identified water quality 46 parameters forms the basis for the monitoring strategy, subsequently ranked by the most critical parameters in the treatment plant. Criticality is determined by two factors, 1) parameters of which the measured concentrations are 48 close to the not to exceed limit and 2) parameters that can be potentially harmful to human health.

50
Determine integrated monitoring strategy of treatment plant 1 consideration the interaction between the different individual treatment processes. The evaluation, of available 2 measurements for the identified water quality parameters forms the basis for the monitoring strategy, again ranked 3 by the most critical parameters in the treatment plant. The monitoring strategy can be embedded into the process 4 control strategy to ensure optimized control based on the most critical parameters.

6
Case study: Ozonation and biological activated carbon filtration at Waternet 7 At the production location Weesperkarspel of Waternet, the water cycle company of Amsterdam and surroundings, 8 ozonation, pellet softening, biological activated carbon (BAC) filtration and slow sand filtration are the main steps 9 in the production of safe drinking water. The feed water is humics' rich seepage water from the Bethune polder, 10 sometimes mixed with Amsterdam-Rhine canal was, which is pre-treated by coagulation, sedimentation, 11 approximately 100 days retention in a lake reservoir followed by rapid sand filtration, before it is transported to 12 the Weesperkarspel treatment plant. At Weepserkarspel, the production of drinking water is roughly divided into The results of the evaluation of each step, to come to an optimised water quality monitoring scheme, are described 32 below, followed by a discussion on the outcomes of the assessment versus the previous and current monitoring

47
Water quality parameters that influence the efficiency of the ozonation step are temperature, pH and, for the total organic carbon (TOC) concentration. In order to assess the character of NOM, the specific UV absorbance  that can achieve disinfection, high frequency monitoring is required enabling direct control of the ozonation step. 41 42 In contrast to ozonation, BAC filtration is not a dosing process, but a separation/degradation process by means of

9
Evaluate available measurements for the identified water quality parameters 10 A summary of the required water quality parameters, as determined in the paragraphs describing the water quality 11 parameters, can be found in the first columns of Table 1 (ozonation) and Table 2    n.a.= not applicable, n.r. = not required.

Determination of individual monitoring strategy per treatment step
1 Figure 2 shows the individual monitoring strategy per treatment step determined by the water quality assessment 2 captured in Table 1 for ozonation and Table 2 for BAC filtration. The results are described in detail below.  therefore there is no need to continuously monitor this concentration in the influent. For Weesperkarspel, the 28 temperature of the water and pH will not change due to application of ozonation, hence there is no need to monitor 29 this in the influent of the BAC filters. 30 31 In Figure   1 on the integrated approach, only 4 differences are observed. In the influent of the ozone step only UV254 is 2 measured instead of UV254 and DOC and the turbidity is measured. In the effluent of the ozone step the i::scan TM 3 is installed measuring at a wavelength of 254 nm instead of the s::can TM able to measure the full spectrum allowing 4 for estimation of bromate and Ct value. However, the Ct value can also be calculated by the installed ozone 5 measurements and the UV254 can give a good indication of the achieved Ct as well (Westerhoff et al., 1999). No 6 differences are observed for the BAC filtration step, when considering the integrated approach.

39
Currently the installed sensors act as an early warning system to flag any deviations in water quality and operation.

40
The next step would be the direct control based on water quality. 41 42 43 CONCLUSIONS 44 The main objective of this paper was to develop a design methodology supporting the development of a water 45 quality monitoring strategy.. A seven step approach was defined, and each step was demonstrated for the treatment 46 processes ozone and BAC filtration. It was shown how the previous on-line water quality monitoring program of the treatment plant Weesperkarspel was adjusted based on a better understanding of the processes taking place.

49
Evaluation of available on-line sensors showed that the parameters temperature, pH and DO are commonly 1 available. Direct measurements of the more complex parameters such as AOC and bromate are not available on-2 line. The use of soft-sensors, able to estimate the bromate and AOC formation, help to gain continuous on-line 3 data. Besides using soft-sensors as surrogate sensors for parameters currently not available on-line, they can also 4 provide a cost effective alternative when used to determine multiple parameters required through one single