The Pressure Is Still On: Deep Well Injection Performance for RO Concentrate Disposal. Abstract

Transcription

1 The Pressure Is Still On: Deep Well Injection Performance for RO Concentrate Disposal Christopher J. Stillwell, PE CDM Smith th Street, Suite 1100 Denver, Colorado Scott Neibur Clint Carter Chris Douglas Tim Rynders Stephanie Mericka Abstract In the summer of 2012 the East Cherry Creek Valley Water and Sanitation District (ECCV) completed construction of a 10 million-gallon-per-day (MGD) brackish water reverse osmosis (RO) water treatment facility. The plant provides a renewable water source to areas of southeast metro Denver which are currently served exclusively by non-renewable groundwater. In addition to the primary RO and UV treatment systems, ECCV utilizes a multi-tier concentrate management approach which includes high recovery RO for brine minimization and an EPA approved deep well injection system for brine disposal. The result is a state of the art facility boasting an overall recovery of percent with no off-site process water discharge, I.e. Zero Discharge Desalination (ZDD). Operation of the facility is dependent upon reliable and continuous operation of the deep well injection system, which delivers gallons per minute of brine or off-specification water to 10,000 foot deep subsurface sedimentary layers for disposal. The system currently operates at pressures ranging from 1,100 1,800 psi, with the capacity for future expansion allowing injection pressures of up to 3,200 psi. The design and startup processes had to carefully consider challenging factors such as uncertainty in long term well behavior and the potential for well fouling. The potential for occurrence of precipitate forming chemical reactions in the injection zone was theoretically modeled, but remained a primary concern during facility design and initial operations. During the initial 3+ years of plant operations, the injection well has been closely monitored for flow, pressure and seismic activity. Data analysis performed bi-annually continues to provide insight into long term well behavior. The resulting conclusions were used by the District operators and engineers to develop standard operational and best management practices for the injection well system. As the first Colorado municipality to utilize this 1

2 approach, operation of the ECCV facility provides a wealth of information regarding the feasibility of utilizing high pressure deep well injection as the primary long term method for drinking water RO concentrate disposal. This paper and the corresponding presentation focus on the first four years of injection well system operation and lessons learned from the gathered information. Summary During the first four years of operation, the pressure and flow data for the ECCV injection well has been collected and evaluated in an attempt to better understand well performance and operation. The approach to data evaluation focuses on three primary properties of the well: 1. System Curve. When designing a pumping system, the system curve or flow vs. pressure relationship is calculated to determine the required pumping head. In the case of a high pressure disposal well, the system curve is defined by the pressures required to inject flows into subsurface geologic formations. When a new disposal well is drilled, a step test is performed. That test forces flow down the well at a constant rate while recording the corresponding injection pressure. This is repeated for 5-6 different flow rates. During design the results of the step test define the system curve for the injection system. However, those results are for a new, clean, unused well and typically are misrepresentative of actual long term well performance. It is therefore useful to replot the flow vs. pressure data for actual operations and evaluate any changes that occur. 2. Historical trends. Simply looking at the life time flow and pressure trends for the well can provide valuable insight into well operations. Long term trends can highlight gradual changes in injection pressure, variation in operations, and disposal well reactions to changes in operation. 3. Normalized Pressure. For a municipal injection system, flows can be variable depending on season, demand, weather, facility operation, and many other factors. This variability in operations can cause data to be erratic and difficult to analyze, or draw conclusions from. In order to deal with this variability, it is useful to normalize the data and evaluate trends that occur in the normalized dataset. For the disposal well, data is normalized as psi/gpm, representing the pressure it takes to force a unit of flow into the well. Analyzing trends in normalized data allows projection of well trends into the future and has provided fairly accurate indication of future well operational conditions. In the following sections, these three properties are evaluated based upon actual data collected over four years of operation for a deep well injection system disposing of RO concentrate. In each case, the conclusions and lessons learned are valuable to engineers and operations staff when developing the designs for new systems or when planning for the future maintenance and operation of existing systems. 2

3 System Curve Generating the actual system curve, or pressure vs. flow plot for the injection well system provides a good indication of the progression of well injection pressure requirements. Evaluating the system curves of the injection system over the life of the well allows the progression of pressures to be illustrated. Figure 1 below combines the annual system curves for 2012 through September 2015 and the initial well step test for pressure and flow comparison. The most notable features of the trends include: A sustained increase in the operating pressure required to inject similar flows when compared to previous years of data. Similar data distribution patterns, annually. New historical high pressures seen each year at varying flows and changing operational scenarios. Figure 1 Deep Well Injection System Curve 3

4 In order to better understand the development of this data, it is useful to look at the data by time period and consider how the pressures changed over shorter periods of time. This approach helps correlate well operations with plant operations. In the paragraphs below, a series of plots show the system curve progression from late 2014 through summer Each plot represents a period of similar plant operation. All data gathered prior to December 19, 2014 (for each time period) is included as light grey on the plots and data in the current monitoring period is shown as a gradation of red. Deep Well Performance Background: The deep well has been in-service since May of 2012 and has seen increased annual use as the Northern Water Treatment Plant increases annual production. As of Sept 2015, the deep well has disposed of 220 million gallons of brine. As shown above in Figure 1, the pressure required to inject brine into the formation has gradually been increasing both due to increased annual volumes and short and long term pressure mounding phenomenon. This paper focuses on the short term evaluation of the deep well performance from Dec 2014 to September 2015 and compares these results with the actual and projected trends from May 2012 through December This evaluation will help determine how closely the deep well is following projections and responding to changes in actual operation. 2015: From late December 2014 through April 2015 (see Figure 2), the plant was intermittently operated at half plant capacity, and less than 260 gpm was sent to the deep well for disposal. Injection pressures were similar to those previously recorded for flows in the gpm range, and well within pump capacity. During this period, injection pressures tended to be similar to the 2013 and spring 2014 data (if compared to Figure 1). This is likely due to periods of 3-4 days in between plant operation when no flow was injected and the effects of short term pressure mounding are minor. 4

5 Deep Well Injection System Curve (Dec 2014-April 2015) Injection Pressure (psi) Injection Pressure (Ft H2O) December April 2015 Half Plant, Intermittent Operation Historical Data Before December Deep Well Injection Flow (gpm) Figure 2 Deep Well Injection System Curve (Dec 2014 April 2015) Plant production from May 2015 through June 2015 (see Figure 3) experienced similar flow rates, but operation of the RO facility and deep well system became continuous. During this time period, injection pressures increased to pressures similar to those previously seen in prior years for this time period. During the period of mid-june 2015 through September 2015 (see Figure 4), the RO facility was operated at full capacity and injection pressures climbed considerably. As seen in previous years, operation of the deep well injection system at higher flow rates ( gpm) resulted in an increase of the required injection pressure. This increase of the injection pressure is extremely prevalent at flow rates in the gpm range where pressure can be seen increasing above all previous data. Also, similar to previous trends, once plant operation was scaled back to half plant, the injection pressures decreased with flow, but remained higher than previously recorded pressures for similar flow rates. 5

6 Deep Well Injection System Curve (May-Mid Jun 2015) Injection Pressure (psi) Injection Pressure (Ft H2O) May - Mid June 2015 Half Plant, Continuous Operation December April 2015 Half Plant, Intermittent Operation Historical Data Before December Deep Well Injection Flow (gpm) Figure 3 Deep Well Injection System Curve (May 2015 June 2015) 1800 Deep Well Injection System Curve (Mid Jun-Sept 2015) Injection Pressure (psi) Mid June - September 2015 Full Plant Operation May - Mid June 2015 Half Plant, Continuous Operation December April 2015 Half Plant, Intermittent Operation Injection Pressure (Ft H2O) Historical Data Before December Deep Well Injection Flow (gpm) Figure 4 Deep Well Injection System Curve (mid-june 2015 Sept 2015) 6

7 Historical Trends The second method used to evaluate well history is to look at the historical trends over the life of the well. Due to the nature of reverse osmosis system operations, the Northern Water Supply Project Membrane Treatment Facility operates at several common production flow rates. These common treatment plant operational flows result in common injection well flow rates, depending on the plant operational scenario. Plotting the pressure and flow data sequentially allows trends to be highlighted illustrating the correlation between pressure and flow at common operational conditions over the life of the well. Figure 5 (below) shows pressure and flow data for the well plotted by date. Data from 2012, 2013, 2014 and 2015 are included. Figure 5 Overall Pressure and Flow Trends by Date There are a few key features that should be highlighted when analyzing the overall flow and pressure trends for the well. The RO facility commonly operates at flow rates of 200 and 400 gpm. This is demonstrated by the grouping of red data points along the 200 and 400 gpm axis. This is the result of operation of the RO facility at full plant operation (10 MGD treated water production with overall facility recovery of 95-97%, 2 primary RO systems and 2 7

8 secondary brine minimization RO systems in operation) and half plant operation (5 MGD treated water production with overall facility recovery of 95-97%, 1 primary RO system and 1 secondary brine minimization RO system in operation). The other flow rates are the result of other intermittent disposal demands such as off-spec storage pond flows during plant operation or when the RO system is not in operation. During the summer months, injection flows are commonly near 400 gpm and injection pressures are high. During winter months, injection flows are commonly near 200 gpm and injection pressures are lower. There is a noticeable increase in yearly maximum injection pressures each year when compared with the previous year. There is a noticeable increase in maximum winter season injection pressures each year when compared with previous years. Evaluation of this data can provide insight into the overall performance of the well. However, due to the very large distribution of pressure data, it can be difficult to identify rates of injection pressure increase. Variations in injection flow and volumes over short durations also make it difficult to trend pressures and illustrate constant change in injection well operations. Therefore, the results illustrated in Figure 5 have been further analyzed to better understand the well trends, and to allow the analysis to yield predictions for future operations. Normalized Pressure Trends Normalization of the injection pressure demonstrates the pressure required to inject a common unit of flow (psi per gpm). Figure 6 includes data for the entire life of the well. This plot provides the most insight as it begins to compensate for variations in short term plant operations such as low flow in the winter, a few weeks of high demands, or occasional shutdowns. These various operational regimes will balance out overall and the normalized trend line is likely more indicative of the current conditions in the well injection zone. The trend line indicates that on average the unit pressure required to pump each unit (1) gpm has risen from 4.1 psi/gpm in 2012 when the well was new to 5.0 psi/gpm in 2015, resulting in ~360 psi greater average pressure to inject a brine flow rate of 400 gpm. 8

9 Items to note include: Figure 6 Normalized Pressure over the Life of the Well There is an increasing normalized pressure trend over the life of the well. Figure 5 best demonstrates the increasing trend of an annual ~5-7% increase in normalized pressure each year. There is an inverse relationship between flow rate and normalized pressure. At higher flow rates, the pressure required per gpm decreases. At lower flow, the pressure required per gpm increases. In either case, the normalized pressure tends to gradually change and approaches a relatively stable value during sustained operation at a common flow rate. The inverse relationship between normalized pressure and flow is potentially largely due to the residual pressure in the injection zone. After periods of high flow the well pressure is already high and therefore higher pressure is necessary to inject low flows until the well pressure falls and stabilizes. Inversely, after periods of low flow the well is at a low pressure and less pressure is initially required to inject high flows until the well pressure increases and stabilizes. As Shown in Figure 6, a linear trendline has been fitted to the normalized pressure data. The normalized pressure tends to approach a relatively stable value near the trendline after changes in flow. Fitting a trendline to the normalized pressure data allows the normalized pressure to be projected into future operational periods. This method has 9

10 been used to estimate when injection well capacity will be limited. Detailed normalized pressure trend results are summarized in Table 1, below. Table 1 - Specific Pressure Trend Summary Date Normalized Pressure Average Normalized Yearly % Increase Yearly % Increase (Trend) Pressure (Actual) (Trend) (Actual) 13-Jul N/A Jul N/A N/A N/A 13-Jul % Jul % 3.13% % 13-Jul % Jul % 12.59% % 13-Jul % Jul % 1.86% % Jul N/A 5.42% N/A Jul N/A 5.71% N/A Jul N/A 4.50% N/A When assessing the new data for 2015, a few things stand out. Primarily, the actual increase in normalized pressure between the summer of 2014 and the summer of 2015 was slightly less than predicted by the trend, and slightly less than previous years. This can likely be attributed to following plant operations: In 2016, Colorado had a cool and wet spring. As a result, the ECCV RO facility did not begin operating at full plant flows until the middle of June. One month later, in the middle of July, the deep well was likely still accepting flow at a lower normalized pressure. During 2015, deep well injection flows never exceeded 400 gpm. The ability of the facility to maintain constant RO recovery reduced the load on the deep well. Off-spec pond flows were minimized during Efficient management of off-spec water which can include startup and shutdown flush water; raw water, concentrate and brine from blow-off events; and plant process water also reduced the load on the deep well. 10

11 However, overall results from the summer of 2015 as well as an overall data analysis for the life of the well indicate an increase in deep well injection pressures, and the normalized pressure trends still indicate the likelihood for a steady increase into the future. Conclusions The data evaluated confirms that injection pressure requirements are increasing yearly at a relatively predictable rate based on the normalized pressure trends. Also, the injection pressure requirements decrease after periods of rest or low flow, but quickly return to previous high values when flow rate is increased. The rise in normalized pressure may be indicative of well fouling, or increased pressure mounding in the injection zone but the fall off of the injection pressure after relaxation indicates most of the pressure increase is likely due to pressure mounding in the injection zone. Pressure mounding as described above is caused by the volume of water injected into the well formations increasing, so there is an increasing resistance to accepting more water. The sub-surface formations become saturated and pressurized and therefore a larger force is required to move the existing volume of water further away from the injection well to make room for new flows. This effect is the lessened after periods of rest, and is exacerbated after periods of high flow. Injection Well Maintenance Since the startup of the injection well, the injected brine fluid has been super saturated with calcite, barite, and ferrous hydroxides. The brine is also saturated amorphous silica and gypsum. Pilot testing with 3 mg/l of scale inhibitor in the feed water, as well as additional grab sample testing during actual plant operations showed that the brine is stable (does not precipitate) for several months when it is not exposed to air or seed crystals that can initiate the precipitation process. Precipitation of salts has occurred in brine samples which were exposed to air. It is possible that injected brine may react with the injection well formations and begin to precipitate, especially during periods of downtime when the well is unused. The other factor that may contribute to well fouling is biological activity. Extended storage of water in the off-spec storage pond results in algae and bacteria formation. The algae growth is significant enough to cause fouling/blinding of the screen filter upstream of the injection well system and may have a similar effect in the injection well formation for the small particles that can pass through the screen filter. 11

12 In general, minimizing the potential for particulate fouling by installing screens on the feed supply to the deep well pumps and by minimizing the potential for precipitate formation is good practice and may increase the life of the well. Good management of disposal flows and a minimization of the introduction of biological constituents is also a best practice. Lessons Learned Design: When establishing the design points for a high pressure deep well injection system, ample thought should be given to the potential for pressure increases over time. The initial well step test data are likely only indicative of initial system operations and significant increases in operational pressure over short term well operations should be expected and designed for. The results of this study indicate that pressure increases of 40 50% above initial well test results are possible within 4-5 years, depending on flow, formations and operational philosophy. Operations: Well capacity is a valuable resource. The capacity of a high pressure deep well injection system may be significantly impacted by heavy loading of the well. Periods of high flow can result in injection pressure increases that are un-recoverable without long periods of well rest. Even with rest the well may not return to like new condition. 12