Fluorescence-Enhanced Treatment Process Optimization

Size: px
Start display at page:

Download "Fluorescence-Enhanced Treatment Process Optimization"

Transcription

1 Fluorescence-Enhanced Treatment Process Optimization Christopher Miller, PE Associate Professor Civil Engineering AWWA Conference October 29, 2015

2 Acknowledgements

3 2014 Headlines

4 2015 Headlines

5 2015 Headlines (Water21, 2015)

6 2015 Headlines

7 Scope and Perspective Efforts are driven by expressed concerns at WTPs regarding: 1. Taste-Odor Issues and Algal Toxins 2. Evaluating and Implementing New Technology and Options (i.e. chemical selection, AOP s) 3. Data-Driven Management (a.k.a. Decision Support) 4. Intermittent elevated THM and HAA levels and more stringent compliance requirements 5. Emerging Contaminants and Unregulated DBPs

8 Background 2015 AWWA Guidance for Water Utilities 2015 WRF Report: if the removal of DOC, UV, and color was optimized, this resulted in optimized cell removal as well, in most cases. 8

9 Fluorescence as a surrogate for DOC DOC is a complex mixture of compounds impacting multiple treatment processes Fluorescence offers a rapid (< 10 minutes per sample) approach to characterizing the nature of DOC and monitoring DOC removal Fluorescence can also be used to monitor algal pigments and algal cells Coagulation Performance Pre-oxidant Demand of DOC PAC Effectiveness Algal Activity Risk

10 AWS Fluorescence Monitoring 2 sample locations (raw and settled water) Sampled once per day or as needed faster than a jar test (make adjustment and check!)

11 AWS Fluorescence Monitoring Fluorescence monitoring has been ongoing since 2010 Three unique fluorescence components have been identified Fluorescence monitoring clearly identifies differential component removal PAC generally increases component removal PAC added Fraction Fmax Removed Humic Like Component Fulvic Like Component Protein Like Component

12 AWS Fluorescence Monitoring Algal pigments fluoresce at higher excitation/emission wavelengths Fluorescence can also be used to monitor raw water algal activity and settled water cell removal (cell lysing and pigment analysis) Raw Water

13 Project Objectives 1. Develop operational data-based models for the coagulation and oxidation processes 2. Identify optimal solutions (i.e. more efficient allocation of chemicals) 3. Evaluate optimal solutions under three scenarios: Varying raw water DOC Varying raw water alkalinity Simulated harmful algal bloom conditions (High raw water DOC and oxidant restrictions) 4. Identify specific solutions to: Reach treatment targets (TO1) Maintain current quality, but reduce cost (LC2) Maintain current cost, but improve quality (BQ3) 5. Assess optimum specific solutions using one year of AWS operational data

14 Modeling DOC and Turbidity Removal Paper accepted for publication: Coagulation modeling using artificial neural networks to predict both turbidity and DOM-PARAFAC component removal Settled Turbidity Model Raw water turbidity, ph, alkalinity, hardness, temperature, chemical dose 1,300 + data points over 4 years Fluorescence Component Models Raw water component scores, alkalinity, ph, chemical doses 215 data points over 10 months Models were assessed using goodness of fit, parametric analysis, and external validation criteria (comparison with line of ideal fit) R = 0.99 R = 0.95 R = 0.93

15 Multiobjective Optimization Simultaneously optimize conflicting objectives i.e. Higher cost = improved water quality (lower turbidity, lower DOC) There is not one unique solution to optimize all objectives; the optimal solutions represent a tradeoff between objectives Goal: Identify the set of solutions that are not dominated by any other solution

16 Identifying Optimal Treatment Solutions There are many possible solutions (chemical dose combinations) The solutions can be evaluated dynamically as incoming water quality changes Chemical Dose Ranges Chlorine Dioxide Dose Interval Number of Possible Combinations Coagulant PAC Perm mg/l 0-50 mg/l 0-3 mg/l 0-3 mg/l , mg/l 0-50 mg/l 0-3 mg/l 0-3 mg/l ,128, mg/l 0-50 mg/l 0-3 mg/l 0-3 mg/l ,211, mg/l 0-50 mg/l 0-3 mg/l 0-3 mg/l ,728,445,721 We define optimal solutions as Pareto efficient solutions Pareto efficient solutions are non-dominated: A solution (x*) dominates another solution if 1. f i (x*) f i (x) for all i 2. f j (x*)<f j (x) for at least 1 j Narrowing solution space to only those solutions that are efficient

17 Optimal Solutions for Different Scenarios Three different scenarios were evaluated and the resulting solution sets were assessed Scenarios: Scenario FA Score HA Score Alkalinity (mg/l) Varying Raw Water DOC Low Moderate High Varying Raw Water Alkalinity Low Moderate High HAB HAB Non-HAB All other variables were held constant at their median value Constraints Scenario Alum (mg/l) PAC (mg/l) Perm (mg/l) Chlorine Dioxide (mg/l) Changing DOC Changing Alkalinity HAB

18 Solutions for 15 Variable DOC and Alkalinity 8 Results show that the Pareto efficient set of solutions capture expected trends As DOC increases: Treatment cost increases (approximately Low Mod 38% High from low to high DOC) DOC Recommended PAC and coagulant doses increase with DOC Alum Dose (mg/l) Alum Dose (mg/l) Perm Dose (mg/l) Low Mod High DOC Low Mod High DOC Settled HA Score Settled HA Score Settled Turb (NTU) As Alkalinity increases: PAC Dose (mg/l) Alum doses remain similar PAC doses increase 2 Treatment costs increase Low Mod High Alkalinity Settled Turb (NTU) Chlorine Dioxide Low Low Mod Mod High High DOC DOC 3 PAC Dose (mg/l) Low Mod High 2 Alkalinity Low Mod High 40 DOC 20 0 Low Low Mod Mod High High Alkalinity Alkalinity Settled FA Score Cost ($/d) Settled FA Score Cost ($/d) Low Mod High DOC x Low Mod High DOC Low Mod High Alkalinity x 10 4 Low Mod High Alkalinity

19 Solutions for HAB Scenario The HAB Scenario was run for High DOC, High Alkalinity Water with restricted oxidant doses Under the HAB Scenario Settled HA Score Pareto efficient solutions had higher PAC and coagulant doses Treatment cost was similar DOC removal was similar These results suggest that under 0 HAB scenario, cost effective DOC removal can still HAB Non-HAB HAB Non-HAB be achieved by limiting oxidants and increasing PAC dose Alkalinity Settled FA Score Alum Dose (mg/l) HAB Non-HAB Settled Turb (NTU) 1.5 PAC Dose (mg/l) HAB HAB Non-HAB Non-HAB Cost ($/d) x HAB Non-HAB How 3 do we select just 1.5 one solution to implement? (mg/l) 2 ioxide 1

20 Selecting Solutions for Implementation For implementation in real time, the set of solutions must be further constrained Approach: identify three solutions from the optimal set of solutions for use by a Decision Maker that meet the following criteria: 1. Reach targets for objectives as specified by Decision Maker (TO1) 2. Lower cost while maintaining same water quality (LC2) 3. Improve water quality for similar cost (BQ3) This approach was tested using 268 daily observations of water quality and chemical dose from Akron WTP from August 2014 through August 2015 BQ3 Current position TO1 LC2

21 Reach Target Objectives (TO1) BQ3 TO1 LC2 Current Targets were set as follows (but could be set at any value): HA score = 4; FA score = 5; Turbidity = 1NTU The algorithm routinely returned solutions with objective function values at or near these targets The ability to achieve a target on any given day is dependent on raw water quality Input from a decision maker would be needed to set these targets as they change over time and to decide if the cost of the recommended solution is worth it

22 Lower Cost for Similar Quality (LC2) BQ3 Current TO1 LC2 Estimated annual savings of $350,000 Recommended turbidities generally within 0.4 NTU of actual turbidities Recommended FA scores generally with 2 units of actual scores Recommended HA scores generally within 1 unit of actual HA scores

23 Improved Quality at Similar Cost (BQ3) BQ3 Current TO1 LC2 In most cases, recommended solutions show cost differences of less than 1% Improved quality was defined as an improvement in all objective function values At these similar costs, calculation of reallocation of chemicals yielded: FA reductions as high as 4 HA reductions as high as 10 Settled turbidity reductions as high as 1.4

24 Fluorescence Processing Dashboard Fluorescence analysis of raw and settled samples is conducted as needed The resulting files are uploaded to the cloud and immediate feedback on coagulation efficiency is provided

25 Decision Support Current water quality and chemical doses are input Recommended solutions are provided

26 Conclusions The removal of fluorescence components and turbidity during the coagulation/oxidation process was successfully modeled These models can be used to identify optimal chemical doses given current water quality conditions General solution evaluation of various scenarios, including HAB, yielded reasonable results and trends (i.e. as source water DOC increases, treatment costs increase) Fluorescence can be used to monitor and optimize DOC removal, which can help a water treatment plant optimize algal toxin removal (TO1) Optimized solutions for regular operations yield: Cost savings (as much as 5-10% annually) while maintaining water quality (LC2) Improved water quality while maintaining similar cost (BQ3) if the removal of DOC, UV, and color was optimized, this resulted in optimized cell removal as well, in most cases.

27 Questions?