United Illuminating



2008 Off-Peak Water Heater Study

Final Report

Prepared for:

United Illuminating Company

December, 2008

RLW Analytics, Inc.

179 Main Street, 3rd Floor

Middletown, CT 06457

[pic]

Table of Contents:

1 Executive Summary 1

2 Introduction 1

2.1 Evaluation Objectives 1

3 Sample Design On-site Evaluation Activities 2

3.1 Sample Design 2

3.2 Site Scheduling 2

3.3 On-Site Data Collection 3

3.4 Test Population 4

4 Evaluation of Data 5

4.1 Initial Data Analysis 5

4.2 Initial Calculation of Demand Reduction Value (DRV) 6

4.3 Post-Stratification 7

4.3.1 Stratification by Tank Size 7

4.3.2 Stratification by Household Occupancy 9

4.4 Stratum Weighted Averages 11

4.5 Bias Assessment & Standby Heat Loss 11

4.6 Improperly Configured Time Clocks 12

4.7 Expected Annual Savings 13

5 Conclusion 13

Appendix A: Site characteristics used in evaluation of data A-1

Appendix B: On-site Data Collection Form A-3

Appendix C: Improperly configured time clocks sites RLW #5 & #47 A-6

Appendix D: Meter Compliance A-8

Appendix E: Calibration Certifications A-12

Table of Tables:

Table i-1: Demand and Savings Results per Controller……………………………………………………….……….2

Table 1: Analysis results for all sites during ISO-NE Summer On-Peak hours 6

Table 2: Stratum weighted mean calculation 11

Table 3: Statistical findings based on strata groups 14

Table 4: Change in demand reduction with adjustment of bad time clocks 15

Table 5: M & V metering requirements A-8

Table of Figures:

Figure i-1: Distribution of tank sizes within Population and the Sample……………………………………………...1

Figure 1: Progression of initial site visits from June 9th to July 9th 3

Figure 2: Average hourly demand reduction for all sites 6

Figure 3: Distribution of tank sizes within Population and the Sample 8

Figure 4: Average demand reduction of the sample by tank size 8

Figure 5: Distribution of the number of household occupants within the sample 9

Figure 8: Misconfigured time clock compared to an accurate time clock configuration 13

Figure 9: Site RLW #5, misconfigured time clock, operation during on-peak hours A-6

Figure 10: Site RLW #47, misconfigured time clock, operation during on-peak hours A-6

Executive Summary

This off-peak water heater study involved a three month period of data collection over the course of June, July, and August, and occurred at 66 residential sites within UI’s service area. Each of these sites consisted of participants in the water heater lease program, in which customers were given a time clock controlled electric water heater. The study examined the average energy demand reduction achieved at each location based on the use of a time clock controlled water heater.

The 66 residential sites were randomly selected to be representative of the 2,031 participants while reflecting the geographic distribution of all those involved in the water heater lease program. The sample size was selected to provide a relative precision of ±10% at the 80% confidence level assuming a coefficient of variation (CV) of 0.64. Figure i - 1 provides a comparison of the on-site sample and the population of participants based upon water heater tank size variable, which shows a bias in the sample for the larger 120 gallon water heaters. The population consisted of about 15% 120 gallon water heaters and the sample included 20% of the larger tanks so the on-site results were adjusted back to the population ratio so that the bias would not be included in the results.

|[pic] |[pic] |

Figure i - 1: Distribution of tank sizes within Population and the Sample

The primary data collection activity consisted of three on-site visits, during the first visit the power metering was installed and 50% of the controllers were left on and 50% were bypassed so that the heating elements would operate under baseline conditions. The second site visit was conducted after approximately six weeks and the state of the controllers was changed so that those that were bypassed initially were switched back on and those that were operating were no bypassed the initial data from the loggers was also collected. After four to six weeks the final visit was conducted and the metering equipment was removed and the final data was downloaded. Metering of each site involved one of three metering scenarios, each of which allowed for the collection of either time of use on/off operation of the elements or true RMS interval power data. All metering and spot wattage power measurements were performed using equipment that was in compliance with the ISO-NE M&V manual for FCM, which requires a true RMS power accuracy of ±2% or better.

The demand reduction value for the water heater time clock was calculated during the ISO-New England summer on-peak hours of 1:00 PM to 5:00 PM end of hour. During the data analysis it was determined that three of the sites had bad time clock controls that allowed the heating element to operate during control hours and locked the lower element out during non-control hours. These data points were included in the analysis to represent the natural occurrence of bad time clocks in the population. The overall Coefficient of Variation (CV) of the sample was about 0.81 and with the bad time clocks removed from the data the CV was 0.71. The sample size was selected based on a CV of 0.64 and therefore the target relative precision of ±10% was not achieved due to the increased variability in the metered data.

Table i - 1 provides a summary of the Summer On-Peak DRV, average daily savings and average annual savings per site for the controllers as observed, and if there were no malfunctioning time clocks. The summer on-peak DRV for the program was 0.23 kW/unit, with a relative precision of ±12.9% at the 80% confidence interval. The average annual savings was approximately 245 kWh per unit based upon the data collected during the summer months. If there were no malfunctioning time clocks, then the summer on-peak DRV would be 0.25 kW with a relative precision of ±12.0% and the average annual savings would be about 241 kWh per unit.

[pic]

Table i - 1: Demand and Savings Results per Controller

Introduction

The United Illuminating Company (UI) is a New Haven, Connecticut based electric distribution utility serving 17 towns and 320,000 customers in the greater New Haven and Bridgeport areas. UI offers a comprehensive set of Conservation and Load Management (C&LM) programs that are designed to benefit all of its residential, commercial, and industrial customers. Currently, residential UI customers are offered a C&LM program in which they can lease a high quality electric water heater, as part of the UI Water Heater Lease Program. Participating customers go onto UI’s Rate RT (a time of day rate), and UI takes full responsibility for service and replacement of the water heater. The leased water heaters are made by the Vaughn Manufacturing Co. and sold under the label Hubbell Heaters. The Hubbell water heaters are supplied to UI customers at various tank sizes of 40, 50, 80, or 120 gallons based upon factors such as house size and occupancy.

The UI rental water heater is equipped with an electronic time clock that regulates the operation of the lower heating element. The lower heating element is prevented from operating during On-Peak hours of 9:00 to 21:00, prevailing Eastern Time. The upper heating element of the water heater is allowed to operate as needed based on hot water demand. It was found that 35% of the sites had a time clock mounted on the water heater, giving the customer the option of pushing an override button. The override button initiates a 2 hr period where the lower heating element is not regulated. The remaining 65% of the sites consisted of a time clock located on the meter panel on the exterior of the customer’s house. The meter panel mounted time clocks do not give the customer the option to override.

Approximately 2,031[1] customers are involved in the UI Water Heater Lease Program. UI has estimated from a recently completed study that the ISO-NE summer On-Peak demand reduction is 0.24 kW.

1 Evaluation Objectives

The overall objectives of the C&LM program evaluations conducted by the Company is: a) to provide timely feedback of information in order to improve the design, implementation, cost-effectiveness, and evaluation of C&LM programs; b) to provide information concerning the attainment of the Company’s corporate C&LM objectives; and c) to support the needs of regulatory authorities.

The RLW evaluation will determine the following:

Average Hourly Load Reductions that is achieved over the ISO Summer On-Peak Hours of 1 p.m. to 5 p.m. non-holiday weekdays in June, July and August

Energy Saving associated with off-peak water heater

The evaluation of this program was designed and executed in a manner which would meet the high standards of UI and will be able to withstand external scrutiny from the Connecticut ECMB, performance filings and necessary internal review.

Sample Design On-site Evaluation Activities

This section will discuss the sample design for the evaluation that was utilized to meet the statistical requirements of the ISO-NE M&V Manual for FCM as well as the metering and primary data collection activities.

1 Sample Design

The design and selection of the on-site sample was the first phase of the evaluation. To estimate the sample size required to achieve ±10% precision at the 80% level of confidence the following formulas were used:

[pic] [pic]

Where,

n0 = the required sample size before adjusting for the size of the population,

z = a constant based on the desired level of confidence = 1.282 for 80%,

cv = coefficient of variation = standard deviation / mean = 0.64,

D = desired relative precision = 0.10,

n1 = the required sample size after adjusting for the size of the population

N = the population size (2,031), i.e., the number of program participants.

The data from a previous study yielded a coefficient of variation (CV) of 0.64, which leads to a sample of 65 for a finite population of 2,031 RLW was supplied with participating customer information table. The only data contained in the table was tank size and geographic location of the participants. It was determined that a random sampling would be representative of these variables. Had more information been available, such as number of occupants, a stratified sample could have been designed.

A random sample was drawn from the table; and it was verified that the sample was representative of the program’s geographic distribution. A secondary sample was randomly selected, which could be used to replace refusals and participants who could not be reached. Although the sample design did not differentiate between single-family and multifamily customers, RLW verified that the final selected sample adequately represented the program population proportions of each group.

The sample design was selected to comply with ISO-NE MV and requirements for a stand alone project, but since we anticipated that UI will include this measure as part of their energy efficiency portfolio it is not necessary to achieve 80/10. However, an additional sample point was included bringing the total sample to 66 sites.

2 Site Scheduling

Each customer in the primary and secondary samples was sent a letter on UI letterhead that introduced RLW as a contractor hired by UI to perform the study. The letter described the purpose of the study and a brief explanation of what the on-site visit would entail. The letter also informed the customer that they will receive an incentive for their participation.

After allowing a few days for delivery of the letters, an experienced RLW recruiter contacted each customer in the primary sample via phone. The recruiter referred to the letter and ensured that the customer fully understood their role in the study. The initial visit was then scheduled at the convenience of each customer. If a customer refused participation, they were replaced by a customer from the secondary sample. The first sites were scheduled for June 9, 2008 and the last sites were visited on July 9, 2008.

[pic]

Figure 1: Progression of initial site visits from June 9th to July 9th

3 On-Site Data Collection

Once onsite, an experienced RLW field representative performed a brief survey with each customer. The survey gathered demographic information about the household, including number of daytime occupants, number of nighttime occupants, etc. Information regarding the amount of water consuming appliance such as showerheads and dishwashers was collected, as well as the frequency at which each of these appliances are used. The onsite representative also collected data on the specifications of each water heater, including model number, installation date, rated wattage, location, and general condition. A copy of the on-site data collection survey is included in Appendix 1. This information allowed for an understanding of each customer’s baseline hot water consumption; while also supplying variables by which the sample could be post-stratified.

After collecting this information, the field representative was directed to the water heater where spot wattage measurements were performed and recorded. The power measurements were performed using a clip-on power meter. RLW only uses properly calibrated ISO-NE compliant handheld meters when performing spot power measurements for FCM project measurements, which must have a true RMS power accuracy of ±2% or better. The meter allowed for a one time true-RMS power measurement. Additional information regarding metering compliance, including calibration certifications, is available in Appendix 3 and 5.

Metering devices were then installed either in the heating element panels located on the water heater or inside the breaker panel. Metering was done with one of two different meter types, Dent Instruments Elite Pro Power Meters, and Dent Instrument Time-of-Use Current Transformer Loggers. Elite Loggers which record average power measurements across a programmed time period, i.e. 15 minutes, made up 18% of the sites. CT Loggers, which covered 82% of the sites, monitor Time-Of-Use (TOU) data. TOU data was then converted into power data based on the spot wattage measurements.

There existed three metering scenarios throughout the 60 sites. The most prominent metering scenario involved 2 CT Loggers, 53% of the sites were metered in this fashion. A CT Logger was placed on the supply feed to each individual heating element. This data showed the separate operating scheduled of the upper and lower heating elements. 28% of the sites were metered with one CT Logger, and the remaining 18% with Elite Loggers.

Upon installation of the loggers, the field representative disabled the electronic time clock on approximately half of the water heaters during the initial visit. The remaining time clocks were disabled during the second visit, and the time clocks previously disabled were reconnected. With the time clock disabled the upper and lower heating elements operate as needed based on hot water demand.

The water heaters were logged during phase 1, between initial and secondary site visit, for an average of 35 days. The following period, phase 2, between secondary and final site visits, lasted an average of 51 days. The period where the time clocks were disconnected established a baseline of water heater use; which could then be compared to the period when the water heater’s operation was time clock controlled. The comparison of the two phases yields load reduction per average participant water heater, through load control that occurs during the ISO critical times of 1pm-5pm, June through August.

4 Test Population

Based on the random sampling 66 sites were selected and scheduled. Three of these sites had to be dropped halfway through the study. One of these sites was dropped as the participant was moving and the water heater was to be removed. Another site was dropped because the water heater failed to operate after the time clock was bypassed. The time clock was then reconnected, and the water heater continued its normal operation. It was recommended that the customer contact UI for service of water heater and/or time clock. The third site was dropped as a language barrier existed, and the customer refused to allow for a second visit to bypass the time clock. Three additional sites had to be dropped due to logger issues.

The final test population included 60 sites, this still allowed for an achievement of ±10% precision at the 80% level of confidence. Appendix contains a table of the test population including number of occupants, tank size, and upper and lower heating element wattage, as these were the most prominent site characteristics used in the evaluation of the data.

The participants in the study were compensated for there cooperation in the study, and to cover any increase in water heater operating cost. The customers were initially offered $50, but in order to schedule the 66 sites it was necessary to offer $75 to some customers. The incentives were given out on the final visit; 68% of the participants received $50, and 32% of the participants received $75.

Evaluation of Data

The data evaluation included analysis of the raw metered data to determine the demand reduction by tank size and number of occupants. The sample data was post stratified and re-weighted to adjust for bias in the sample, which included a larger percentage of the 120 gallon tanks in the sample than in the population. As a result the raw sample data had to be adjusted and re-weighted to correct for the sampling bias in the sample.

The experimental design included a random selection of customers that had their controls overridden during the first on-site visit and those that had their controls overridden during the second on-site visit so that the split was 50/50 during each period. This was meant to control for any bias in overall water usage that may have occurred during the two time periods. Additionally an analysis of the expected savings due to standby losses was conducted and compared to the calculated savings for each customer to investigate whether there appeared to be systematic bias in the data. The results of the analysis indicated that there was no systematic bias in the data and there was no need to adjust the results.

Finally the issue of malfunctioning time clocks was addressed as part of the data analysis, because there were three sites that had distinctly different usage patterns than the other participants when the controllers were operating. The abnormal usage pattern for these three sites did not occur when the controllers were bypassed and were clearly due to problems with the time clock settings. The data for these sites were left in the analysis when calculating the final Demand Reduction Value (DRV) and energy savings because they represent the naturally occurring rate of bad time clocks in the population.

1 Initial Data Analysis

Following the final site visits all data was downloaded from the loggers. The data from each site consisted of power data for the water heater both with and without the time clock enable, pre and post period respectively. The interval power data was brought into RLW’s Visualize-It software for quality control inspection and the data was then converted into hourly interval data for analysis. Data from each of the 60 sites was output into hourly power consumption from hours 1 through 24. Each site consisted of two 24 hour data sets and was labeled as either pre or post period. The difference between the two data sets at each hour was calculated and referred to as the hourly demand reduction.

Figure 2 depicts the average hourly demand reduction or increase for all sites within the sample. The values were calculated as the difference between the pre and post period for each hour of the day for each individual site. The average was then taken for each hour. Demand reduction is illustrated in the chart as a negative value, highlighted in blue, and the demand increase is a positive value, highlighted in red. It is clear that savings or demand reduction is achieved between hours 6 to 22 or 6:00 AM to 9:59 PM. On-peak hours fall within the time frame where the demand reduction occurs. This verifies that the demand reduction is occurring during on-peak hours, as intended by the water heater lease program.

[pic]

Figure 2: Average hourly demand reduction for all sites

The demand increase occurs during off-peak hours. The time clock allows the lower heating element, which has a higher temperature setting than that of the upper heating element, to operate during this time frame. The demand appears as an increase during the late night hours of 23 to 5. When the time clock is not in operation there is little need for the water heater to operate during these late night hours. This is because the water heater functions strictly on a demand basis, meaning the heating elements kick on as a result of water being drained from the tank. However, when the time clock is functioning as intended the majority of water heating is performed during the midnight hours, as this time frame represents off-peak. Therefore, the pre period of the study, without the time clock in operation, results in a power consumption of nearly 0 kW during late night hours, while during the post period power consumption is at its highest the same hours.

2 Initial Calculation of Demand Reduction Value (DRV)

The data was analyzed to determine the Demand Reduction Value (DRV) for this measure during the ISO New England FCM summer On-Peak performance hours, which are between 1:00 PM to 5:00 PM during the months of June, July, and August. The initial calculation of the DRV utilized the un-weighted data for each site during the pre and post metering period and an average demand reduction was calculated for each site. The overall DRV was calculated as the simple average of the 60 sites with no adjustments for sample selection bias.

|Unadjusted Mean |0.24 |

|Standard Deviation |0.19 |

|Coefficient of variation |0.79 |

|Relative Precision |±13% |

Table 1: Analysis results for all sites during ISO-NE Summer On-Peak hours

Table 1 provides the unadjusted mean demand reduction value of 0.24 kW, which has not been corrected for bias. The coefficient of variation (CV) of 0.79 and the relative precision (RP) of ±13%, were higher than expected when compared to the CV of 0.64 and the RP of 10%, which were used in designing the sample. This is the result of increased variation from site to site within the sample. The average demand reduction of the entire sample ranged from -0.10 kW to 0.98 kW, this range is including sites with misconfigured time clocks, without the misconfigured time clocks the range is 0.02 kW to 0.98 kW. The presence of misconfigured time clocks within the sample will be discussed in detail in section 4.6. Certain factors specific to each site, for example number of occupants and tank capacity, can affect the expected demand reduction. The variation in site characteristics is the cause of the wide range in demand reduction. To verify that such trends existed within the sample, and to determine their impact on the demand reduction the sample was post-stratified using the information collected on site.

3 Post-Stratification

The primary reason for the post stratification of the data was to adjust for the sample selection bias and to investigate whether the overall relative precision of the DRV would improve when the data was stratified. Although there were a lot of variables that were evaluated for potential impact on the data including;

• Tank Size

• Number of Occupants

• Number of Showers per week

• Number of Showerheads

• Number of Faucets

Only the first two variables, tank size and number of occupants proved to have a good correlation with the savings results. Since tank size was the only variable that was also available in the tracking data, the distribution of number of occupants observed in the sample was assumed to be representative of the population.

1 Stratification by Tank Size

Stratifying the sample groups the sites with similar properties together. The 60 sites were first divided by the storage capacity of the water heater. All sites fit into one of two strata; 80 gallon tanks (including one 40 gallon tank exception) or a 120 gallon tank. According to Figure 3, it was found that at 80% the majority of sites were equipped with a water heater having a storage capacity up to 80 gallons, with one of these sites having a 40 gallon tank. The remaining 20% of the sample was equipped with a 120 gallon tank water heater, which is significantly higher than the 15% frequency of 120 gallon water heaters in the population data. This discrepancy in sampling rate can easily be accounted for analyzing the savings for the two groups independently and weighting the savings for each group to reflect their frequency in the population, i.e. 85% for 80 gallon tanks and 15% for 120 gallon tanks.

|[pic] |[pic] |

Figure 3: Distribution of tank sizes within Population and the Sample

The average demand reduction was then calculated based on tank size. When compared to the average demand reduction of all sites the 80 gallon tank sites had a lower average demand reduction, while the 120 gallon tank sites had a higher demand reduction. Figure 4 shows that the average demand reduction for all sites, 80 gallon tank sites, and 120 gallon tank sites where 0.24 kW, 0.22 kW, and 0.31 kW, respectively.

[pic]

Figure 4: Average demand reduction of the sample by tank size

This suggests that the time clock is more successful at reducing the energy consumption of the water heaters during on-peak hours if it has a storage capacity of 120 gallons. The figure includes the number of sites in each strata labeled as ‘n’, and as previously mentioned there were 12 sites with 120 gallon tanks. This is a relatively small sample size. There is a potential bias within the average demand reduction of 0.31 kW. Therefore, it cannot be definitively stated that the actual average demand reduction for all sites with 120 gallon tanks within the entire population is 0.31 kW.

2 Stratification by Household Occupancy

The sample was then stratified to determine how the number of household occupants affects the demand reduction associated with the operation of the time clock control. All sites were separated into two groups: occupancy of two or less and occupancy greater than two. The distribution of the sample into these two strata can be seen in Figure 5. As depicted in the graph 38 sites had either one or two household occupants, making up the majority of the sample at 63 %. The remaining 22 sites had a household occupancy of more than two residents, at 37 %.

[pic]

Figure 5: Distribution of the number of household occupants within the sample

Compared to the demand reduction for all sites, which achieved an average demand reduction of 0.24 kW, the households with two or less occupants had a lower average demand reduction at 0.21 kW. The households with greater than two residents achieved an average demand reduction of 0.29 kW. Therefore, the demand reduction which can be achieved by the time clock controlled water heater is greater for households with more than two occupants. The values for demand reduction of the two occupant based groups can be seen in Figure 6.

[pic]

In order to determine if the number of occupants effects the demand reduction based on tank size stratification, it was necessary to look at the household occupancies within each of the tank size groupings. Figure 7 illustrates that 65% of the sites equipped with an 80 gallon tank water heater have two or less occupants in the household, and the remaining 35% of the sites have two or more occupants. The sites equipped with 120 gallon tank water heaters have a closer divide with 58% having two or less occupants and 42% having more than two occupants. It is clear that sites with a tank size of ≤ 80 gallons have a larger majority of households with only two or less occupants when compared to the sites having a tank size of 120 gallons. This suggests the potential for bias based on number of occupants within the tank size groupings which as previously discussed the average demand reduction was lower for households having fewer occupants.

4 Stratum Weighted Averages

Prior to stratification of the sample the average demand reduction was calculated assuming equal weights for each site. However, the over sampling of the 120 gallon water heaters would clearly cause a sample bias in the results that would increase in the calculated demand reduction if the simple mean value of the sample were used. Therefore it was necessary to evaluate the demand reduction value using population weights that reflect the frequency of tanks in the population. Additionally the data was post stratified based upon the number of occupants (less than 2 or greater than 2) within each tank size category, resulting in a total of four strata. Population weights were calculated for each stratum which reflects the frequency that each type of water heater would occur in the population.

Table 2 provides the mean summer On-Peak DRV for each stratum, along with the stratum weighted DRV for the population which is 0.23 kW per water heater controller. The relative precision of this weighted DRV estimate is ± 12.9% at the 80% confidence interval. Note that the weighted DRV of 0.23 kW is slightly lower than the unadjusted mean DRV of 0.24 kW, which was calculated using the sample data. Although both of these values are relatively close we would recommend that UI use the weighted DRV of 0.23 kW per controller because it reflects the current distribution of water heaters in the population.

[pic]

Table 2: Stratum weighted mean calculation

5 Bias Assessment & Standby Heat Loss

During the course of the study there is potential for bias in the demand reduction if there existed a difference in hot water usage between the pre and post periods. The initial step in determining if such a systematic bias existed in the sample was to ask the customers if they were aware of any change in their water use. Such changes would have included vacation or having guests for extended periods of time. Based on the collected field information it appeared that this would be of little influence on the data.

To further analyze for the presence of such a bias an estimated demand reduction was calculated, which could be compared to the metered demand reduction. This calculation was done based on the principles of standby heat loss, as with all storage tank based water heaters energy is lost by the stored water to the surrounding environment. This energy loss results in reheating of the water and further energy consumption. The standby heat loss was calculated using the equation below, which dictates conductive heat transfer.

Q= U∙A∙∆T

Where,

Q = stand by heat loss (Btu/hr)

U = thermal conductance of water heater’s insulating material (Btu/(hr∙ft2∙°F))

A = surface area of the water heater (ft2)

∆T = temperature difference between upper and lower heating elements (°F)

The thermal conductance of the water heater is derived from the insulating material of the water heater, which has an R-16 thermal resistance rating. The inverse of the R-value rating yielded a thermal conductance of 0.625 Btu/(hr∙ft2∙°F) for each of the water heaters. The surface area was calculated using the water heater specs collected from each site. The calculated surface areas of the 120, 80, and 40 gallon tanks were approximately 50, 40, and 31 ft2, respectively.

The thermal properties and geometry of the tanks were for the most part constant; therefore ∆T was the driving force of the standby heat loss. The average temperature difference between the set points of the upper and lower heating elements was 8°F, but ranged from 0 °F to 19 °F. The most commonly occurring ∆T was 10 °F, with a lower temperature setting of 135 °F and an upper temperature setting of 125 °F. Standby heat loss is reduced during the time clock controlled period because of the lowered temperature setting of the upper heating element compared to that of the lower heating element. During the period when only the upper heating element is allowed to operate, heat lost is replenished to the point of the upper heating element temperature setting, which, as stated previously, is on average 8 °F less.

The standby heat loss, Q, having units of Btu/hr was converted into a demand reduction having units of kW, using the conversion of 3,412 Btu/hr per kW. The demand reduction occurred over a period of 13 hours per day, as this is the amount of time that the water heater is controlled by the time clock or then number of on-peak. The product of the demand reduction and the 13 hour time frame yielded an estimated daily demand reduction.

Comparing the estimated daily demand reduction to the metered average daily demand reduction showed that any differences in hot water usage between the pre and post periods was of little significance. The difference between the two demand reductions appears to be randomly distributed; meaning no pattern exists which would show the effects of flow change. Therefore, it was not necessary to adjust for any bias based on hot water usage.

6 Improperly Configured Time Clocks

The average demand reduction across the time period was determined for each of the individual sites. Three of the sites, RLW # 5, 12, and 47, had a negative demand reduction, showing no savings in the time period of interest. It appears that the cause of the negative demand reduction was due to improperly configured time clocks. A graph output by RLW’s Visualize-It software can be seen in Figure 8, it shows a comparison of a typical time clock operating schedule and that of a time clock which is not properly configured. Site RLW #12 represents the misconfigured time clock, and site RLW #18 is used to demonstrate the correct operating schedule of the time clock controlled water heaters. Site RLW #12, represented by the blue data line in the figure, shows that the time clock controlled heating element is not allowed to operate during off-peak hours when it should be, and is instead forced to perform the majority of its work during midday on-peak hours. The red data line in the figure, site RLW #18, represents a typical operating schedule one that regulates the water heater so that the majority of work occurs during off-peak hours. Sites with an improperly configured time clock have an occurrence of 5% within the sample. An analysis of the entire sample yielded a lowered average demand reduction because of the presence of these improperly configured time clocks. Appendix 2 includes a graph illustrating the time clock operation of sites RLW #5 and #47.

[pic]

Figure 8: Misconfigured time clock compared to an accurate time clock configuration

7 Expected Annual Savings

To develop an understanding of the success of the water heater lease program it is important to determine an expected kWh savings. Using the average hourly demand reduction an average daily kWh savings was calculated at 0.67 kWh per day per site. This daily savings value was converted into an annual estimated savings of 245 kWh per year per site based on the unadjusted summer data. As this is an average savings per site it would be expected that the annual savings would occur at each of the 2,031 sites within the total population of the water heater lease program.

Conclusion

Over the course of the study data was collected and analyzed for 60 sites. The data which yielded pre (without the time clock in operation), and post (with the time clock in operation) periods showed that a demand reduction did in fact occur during the period of interests between 1:00 PM and 5:00 PM end of hour. As shown in Table 3, the average demand reduction for all sites was 0.24 kW. This yielded a higher than expected coefficient of variation at 0.79 and relative precision of ±13 % as well as a relatively high standard deviation at 0.19. This is because of the large variation in site characteristics throughout the sample. Some of the factors influencing the achievable demand reduction are tank size, household occupancy, and presence of hot water consuming appliances, i.e. dishwashers, washing machines, and showers. Other factors exist which would impact the demand reduction, but are difficult to quantify. An example of this is changes in occupants’ daily schedules, and whether or not the occupants are conscious of their hot water use patterns, or if they actively attempt to focus their high hot water consuming activities during off-peak hours. The impact of the variation from site to site is evident in the range of average demand reductions specific to individual sites. The lowest average demand reduction was -0.10 kW and the highest was 0.98 kW.

Table 3 illustrates the average demand reductions found based on the various strata. It appears that for sites with high household occupancies achieve a larger demand reduction, but there is also a higher relative precision associated with this stratum as a result of the small sample size. The small sample size is the result of a randomly selected sample which was designed to represent the entire population, and sites with an occupancy of more than two people represent the minority. This is also the case with the 120 gallon tank stratum.

|  |Average Demand Reduction |Sample Size |Standard Deviation|Coefficient of |Relative |

| |(kW) | | |Variation |Precision |

|All Sites |0.24 |60 |0.19 |0.79 |±13% |

|Occupancy ≤ 2 |0.21 |38 |0.19 |0.94 |±20% |

|Occupancy > 2 |0.29 |22 |0.17 |0.58 |±16% |

|80 gallon tank |0.22 |48 |0.18 |0.84 |±15% |

|120 gallon tank |0.31 |12 |0.20 |0.64 |±24% |

Table 3: Statistical findings based on strata groups

As mentioned in the winter study conducted on the time clock controlled water heaters, the amount of savings achieved is higher with a larger household occupancy. This is due to the fact that in low occupancy households the hot water demand is minimal during on-peak hours, which is when savings is most commonly achieved. When the time clock is not in operation, pre period, and the tank volume is depleted during the day the lower heating element immediately starts reheating the replenished water to the higher temperature set point. During the post period the time clock allowed only the upper heating element to reheat the new water entering the tank during the on-peak hours. The temperature difference at which the replenished water is reheated during pre and post periods drives the demand reduction. Sites with only one or two occupants did not drain the tanks down frequently enough during midday to acquire such savings. Therefore, the time clock controlled water heaters achieve a higher average demand reduction when installed in houses with greater than two occupants. It is recommended that UI takes this into consideration when electing participants for the Water Heater Lease Program.

Sites equipped with an 80 gallon tank water heater consist of a large majority of households with two or less occupants. This suggests that the reason the sites with an 80 gallon tank water heater have a lower average demand reduction when compared to the average of all the sites is because a high percentage of these households have two or less occupants. The demand reduction is driven by the number of occupants in the household, and therefore, it would be a bias statement to claim that the 80 gallon tank water heaters do not achieve as much demand reduction when compared to the 120 gallon tanks. It is in fact, the number of occupants which drive the expected demand reduction associated with the use of the time clock controlled water heater.

Of the 60 sites included in the study three consisted of improperly configured time clocks. These sites showed a negative average demand reduction during the hours of interests. The presence of these sites in the sample lowered the average demand reduction, as they account for 5% of the sample. It can be concluded, based on the size of the sample, that the same percentage of sites have misconfigured time clocks in the total population. Table 4 shows the increase in average demand reduction achieved when the three sites containing the bad time clocks are removed from the sample. The average demand reduction increases by approximately 6%, changing from 0.23 kW to 0.25 kW when the bad time clocks are removed from the sample.

| |All Sites w/ Bad Time |All Sites w/o Bad Time |

| |Clocks |Clocks |

|Mean |0.23 kW |0.25 kW |

|SD |0.19 |0.18 |

|CV |0.81 |0.71 |

|RP |±13% |±12% |

Table 4: Change in demand reduction with adjustment of bad time clocks

Each of the sites with a time clock that was not functioning properly had an exterior ring time clock mounted on the meter panel; therefore, it is possible the time clocks faulted due to weathering. It is recommended that the exterior time clocks be checked for proper configuration, and installation of the water heater mounted time clocks is encouraged.

Appendix A: Site characteristics used in evaluation of data

[pic]

Appendix B: On-site Data Collection Form

On-Site Data Collection Form

United Illuminating

Water Heater Data Collection Form

Name:________________________________________________________

Address 1:___________________________________________________

Address 2:___________________________________________________

City:_____________________________ State:_______ Zip:____________

Telephone: (____) -_____-______

|Visit |Date |Time |Override |

|1 | | | |

|2 | | | |

|3 | | | |

Logger ID: Time Clock Location:

CT Logger:________________ 1 Water Heater

Elite:________________ 2 Meter Panel

Demographics

Residence Type:

1. Single-family detached home

2. Condominium

3. Apartment/Townhouse

4. Duplex

5. Mobile home

6. Resort cottage or cabin

7. Other (please describe:_____________________________________)

Own vs. Rent

1. Own

2. Rent

Home Age:

_____ Years

Square Feet (DO NOT INCLUDED UNFINISHED BASEMENT)

__________sq. ft

If customer is not sure -------( Do you have any idea if it is:

1. Under 1,000 square feet

2. 1,000 – 1,999 square feet

3. 2,000 – 2,999 square feet

4. 3,000 or more

5. Don’t know

Quantity of each in home:

_____ Dishwasher

_____ Clothes washer

_____ Showers

_____ HW Faucets other than bathtubs with showers

_____ Spas & Hot tubs

Weekly Usage of HW Appliances:

____ Dishwasher cycles per week ____________Time of Use

_____ Clothes washer cycles per week ____________Time of Use

_____ Showers per week ____________Time of Use

_____ Spas & Hot tubs ____________Time of Use

Number of people in household (by age group):

|Age Group |Number at home during day |Number at home during evening/nigh |

|Under 5 years old | | |

|5 – 18 years old | | |

|19 – 34 years old | | |

|35 – 44 years old | | |

|45 – 54 years old | | |

|55 – 64 years old | | |

|Over 64 years old | | |

Water Heater Inventory Data: Fill in matrix as appropriate

|Parameter |Data |

|Manufacturer | |

|Model Number | |

|Date of Installation | |

|Rated Wattage | |

|Location (cond. vs. uncond) | |

|Storage (gallons) | |

|Size (dia. and height) | |

|Power Measurement (Watts) | |

|Pipe Insulation present (%) | |

|General Condition | |

Spot Watt:

| |Volts |Amps |kW |Temp Setting |

|Upper Element | | | | |

|Lower Element | | | | |

Appendix C: Improperly configured time clocks sites RLW #5 & #47

[pic]

Figure 9: Site RLW #5, misconfigured time clock, operation during on-peak hours

[pic]

Figure 10: Site RLW #47, misconfigured time clock, operation during on-peak hours

Appendix D: Meter Compliance

Section 10.2 of the ISO-NE M&V manual contains a list of seventeen requirements for metering equipment that is used for measuring Demand Reduction Values (DRV) in the Forward Capacity Market (FCM). The requirements can be divided into four general categories, which are Standard Conformance, Technical Specifications, Usage & Method and Calibration & Maintenance, as presented in Table 5.

| |Application |Standard Conformance|Technical Specs |Usage/ Method |Calibration/ |

| | | | | |Maintenance |

|1 |All solid-state measurement equipment |X | | | |

|2 |Equipment directly measuring power or |X | | | |

| |demand | | | | |

|3 |Instruments measuring volts, amps, and |X | | | |

| |phase angle | | | | |

|4 |Data recorders that are recording pulses | | |X | |

|5 |Equipment exposed to significant harmonics|X | | | |

|6 |True RMS kW measurements and accuracy | |X | | |

|7 |Measuring imbalanced three-phase loads | | |X | |

|8 |Sampling rate on circuits with significant|X |X | | |

| |harmonics | | | | |

|9 |Accuracy of demand calculated with proxy | |X |X | |

| |variables | | | | |

|10 |Demand calculations to use power factor of| | |X | |

| |the end-use | | | | |

|11 |Data recorders must be synchronized in |X |X |X | |

| |time with NIST | | | | |

|12 |Equipment calibration to appropriate |X | | |X |

| |standards | | | | |

|13 |Equipment maintenance to appropriate |X | | |X |

| |standards | | | | |

|14 |Documentation of calibration and | | | |X |

| |maintenance activities | | | | |

|15 |Availability of calibration and | | | |X |

| |maintenance records | | | | |

|16 |Alternative accuracy, calibration, and | | | |X |

| |maintenance standards | | | | |

|17 |Interval data collection frequency | |X |X | |

Table 5: M & V metering requirements

Some of the requirements listed are clearly not applicable to the single phase meters used in this study, including # 4 for pulse meters, and # 7 for three phase loads. There are eight requirements (#1, #2, #3, #5, #8, #11, #12, and #13) that refer to some set of standards, be they ANSI, IEEE, relevant, equivalent, or industry/manufacturer in nature. Five of the requirements (#6, #8, #9, #11, and #17) indicate a technical specification that must be met. Six requirements (#4, #7, #9, #10, #11, and #17) involve the methods by which the equipment is installed or used. Finally, five requirements (#12, #13, #14, #15, and #16) specifically target calibration and maintenance of M&V equipment. Many of the ANSI and IEEE and other documents are lengthy and are primarily focused on revenue grade socket meters. The following sections will discuss the metering compliance with requirement numbers 1, 6, 8, 11, 13 and 17, which we have deemed to be relevant and reasonable to address.

The metered data used for this study was collected using Dent Instruments Elite Pro Power Meters and Time-of-Use Current Transformer (TOU CT) Loggers to measure the True RMS power of the room electric water heaters. The manufacturer lists the meter accuracy at ± 1.5% for true RMS power measurement and all of the meters were calibrated to that standard prior to shipping. The meters were also checked for accuracy prior to installation. The Dent Instruments Elite Pro Power Meters and Time-of-Use Current Transformer (TOU CT) Loggers meters are solid state electronic meters that require infrequent calibration and they should have remained within factory specifications for the duration of the data collection period for this project, (# 6 and #13).

Section 10.2 of the ISO-NE M&V manual specifies that measurement tools must be synchronized in time within an accuracy of ±2 minutes per month with the National Institute of Standards and Technology (“NIST”) clock. The Dent Instruments Elite Pro Power Meters and TOU CT Loggers meter contains a solid state circuit that exceeds the ±2 minutes per month standard for time drift. RLW standard operating procedure was to synchronize all of Dent Instruments Elite Pro Power Meters and TOU CT Loggers to a desktop computer clock that is linked to our network server and maintained in synch with the NIST clock (#11).

The Dent Instruments Elite Pro Power Meters and TOU CT Loggers are UL rated for safety for indoor use at a temperature range between 43ºF and 117 ºF. The meters are also rated to operate up to an elevation of 6,562 feet and at a relative humidity of up to 80% at 98ºF declining linearly to an RH of 50% at 117 ºF and we can reasonably assume that all of the units we operated at these UL rated conditions. UL ratings are equivalent to the American National Standard Institute (“ANSI”) standards (#1). Requirement 17 of the ISO-NE M&V Manual states that “Interval metering devices shall collect electricity usage data at a frequency of 15 minutes or less.” Although it is unclear if this requirement is ambiguous for failing to differentiate between sampled and integrated values. The likely intention was to double the Nyquist frequency thus mitigating data aliasing in equipment that duty cycles such as air conditioning and water heaters. Many interval metering devices (including the Elite Pro Power Meters and TOU CT Loggers) sample data continuously, integrating and storing data at a programmable interval. For data of this nature, rigorous computation of coincident peak demand impacts requires no more resolution than hourly data. In any event, the Elite Pro Power Meters and TOU CT Loggers were set up to record data at 15-minute intervals so even under the strictest interpretation of this requirement the metered data would be in compliance (#17).

Finally the Elite Pro Power Meters and TOU CT Loggers meter’s literature indicates that the meter measures voltage and current thousands of times per second and is capable of measuring non-sinusoidal wave form power accurately. The manufacturer has indicated that the sampling rate is in excess of the 2.6 kHz requirement for measuring harmonics listed in the ISO-NE M&V manual (#8). Based on the meter specification data provided by the manufacturer we conclude that the metered data used for this study is in compliance with the ISO-NE M&V requirements.

Appendix E: Calibration Certifications

[pic]

[pic][pic]

-----------------------

[1] RFP stated that there existed 1,725 participants in the water heater lease program; however, the customer info table consisted of 2,031 participants.

-----------------------

Figure 7: Distribution of occupancy based on tank size

Figure 6: Average demand reduction after post-stratification of the sample by occupancy

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download