You are on page 1of 7


Published by EnergyPulse December 9, 2002

Unlocking the Promise of Advanced Metering
Analyzing Interval Data
By Gregory Cmar Cofounder & CTO Interval Data Systems, Inc.

..................................................................................... Inc........5 Deregulation: Putting Value on Risks in Energy Supply Contracts.........................4 Using Regression to Build a Model .............................. All rights reserved...........................................................4 Minimizing Complexity..........................................................................................Table of Contents Advanced Metering ........... 4 Controlling the View ...................................................................................................................................................... 3 Information: The Value in Advanced Metering ....................................................................... –2– ...... 5 Conclusion.................................................................................................................................................................................................................................................. 6 © 2004 Interval Data Systems............................................................................................................................3 Analyzing Interval Data .........................

Substantial benefits for the end user—lower energy costs. Looking not the main line of business in the company. It’s the height (kW) of each hour’s consumption column.84 $ 1. the utility bills are up.000 7 6.64 $ 1. All of these can and will be resolved.51 to the total amount.68 the most significant development since the 10.67 $ 1.000 1.628 102 $ 2.001 to 200.53 normally yield significant Over 500. at least you didn’t have to actively manage consumption.19 $ 2. so they had the heating The chart shown above is a typical daily consumption pattern for system run longer and the time clock was never reset. The expectation is that consumers will apply modern management techniques.34 $ 1.25 $ 1.001 to 50. thermostat on that 0 really hot day and never 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 bother to turn it back.32 $ 1. Economies of scale drove costs down to the point where even if you metered.05 $ 1. Interval data.66 $ 1. facility Number of Floor Space mBTU managers and end users NorthMidBuildings (million per Building Floor Space have learned how they are (square feet) (thousands) square feet) Square Foot South West east west expected to respond.79 $ 1.000 708 11.31 $ 1. of course. Inc. kW –3– .774 123 $ 3.348 6. and extract new efficiencies for the industry – all so that the utilities can keep overall costs down while meeting their mandate for plentiful. Advanced Metering Advanced Metering is becoming the premier information tool of the electric power industry. change their habits. reliable power.271 105 $ 1.000 59 8.15 information.65 power was used. For whatever 24x7 Baseload is consumption reason. Capturing the savings potential will not be easy. The 24x7 Baseload proportional to how opportunity exists people use equipment because most people are 150 and lights when they are unaware of the in the facility.311 81 $ 1. All rights reserved. Somebody calls it “load creep”. does not 200.48 $ 1.12 Interval data presentation is 5.72 $ 1.001 to 10. Square foot consumption consumption is the ability and cost metrics have been adjusted to exclude warehouses. For the commercial building sector.851 107 $ 1. as supplied by the meter. Information: The Value in Advanced Metering In the 1950’s.54 $ 1. interface applications and operational strategies put in place in order for it to produce results.000 145 10. Advanced Metering clearly has the potential to provide the information necessary to tap into this potential.46 $ 1. Advanced Metering is more than just the installation of a meter. lead to different strategies. Firms that specialize in facility and energy management have found they can routinely save 15% of the total energy bill in any building that has not had an expert fine tune operation within the past year. better monitoring of facility operations. All Buildings Consumption Energy Cost per Square Foot symbiotic relationship after the business owners. Pretty a building. The more significant hurdle is that the market needs to be educated about Advanced Metering and how to achieve its benefit. accurate budgeting tools. Different times. The device needs a communications pathway. Time of Day There was late meeting.80 $ 1. etc. Variable Baseload but marbling in the 200 Variable Baseload is meat of operation. Just as clearly.238 79 $ 1. There is also a question of who should make the investment.48 $ 1.—are being promoted by the industry.16 $ 1. this implies a potential savings of $12 billion annually. it is possible to see that there is a block of consumption and the cause is forgotten. The ability to interpret the patterns of Data obtained from EIA’s 1999 Commercial Buildings Energy Consumption Survey (CBECS).Unlocking the Promise of Advanced Metering Analyzing Interval Data By Gregory Cmar.153 70 $ 1.000 257 9.24 $ 1. the industry hopes its users will achieve.35 building owner now has the capability of seeing when 50. © 2004 Interval Data Systems.05 $ 1. The promise of Advanced Metering is the ability to resolve consumption according to time of day while the utility varies the time-of-day price according to market conditions. of course.001 to 25.000 2.12 invention of metering.45 $ 2.001 to 5.001 to 100. 1. It promises to be the perfect.112 93 $ 1. the first of which is cost. in addition 100.17 $ 1. The 25. They occupancy or turn down the weather. they don’t that occurs bother to shut off 50 without regard to unnecessary lights. Cofounder & CTO.11 $ 1.000 square Daily Electric Consumption Weather Dependent Load feet of floor space. Weather Load (70°F) fat that can be trimmed. so the bill gets paid closely. Interval Data Systems. electric utilities had a vision of power so cheap there would be no reason to meter. compounded implications of their 100 actions. There are two substantial hurdles to overcome.110 8. only describes soon. that $12 billion benefit rounds out to merely $15-20 per month per 1. Raw data.001 to 500.000 23 6.50 $ 0. Inc. This 250 is proportional to outside potential savings is not temperature.

Analyzing Interval Data Analytics can be summarized as a process for identifying patterns within a given set of data. office machines. however.that occurs 24 hours per day (green). alludes to a pattern. a timestamped strip. etc. The final power block is weather dependent consumption. electric. around 6 AM. etc. these patterns are often hidden deep within the data. The next block of power identified in the chart is Variable Baseload (green stripe). As will be seen. And. Saturday and Sunday show the same saw tooth pattern. Once the bills are modeled—whether they are gas. Analyzing Interval Data will explain a process that reduces the complexity of analyzing utility data. This consumption is a result of equipment – lights. Past performance predicts future performance. The new level of information available as a result of Advanced Metering is both self-apparent and obscure in the facility management industry.400 1. Eventually. Looking at interval data in the manner that it is received. purchasing strategies can be maximized to obtain the best price. with 24x7 occupancy. Since we have already seen that the hourly consistency exists.000 800 600 400 200 0 Mon 2/5 1 PM Tue 2/6 1 PM Wed 2/7 1 PM Thu 2/8 1 PM Fri 2/9 1 PM Sat 2/10 1 PM Sun 2/11 1 PM 40 kWh per Day 30 20 Time of Day 10 Spreading the time interval out over a single day.400 1. This minimum. Any deviation becomes a flag for management's attention. 1. The benefits. Thousands 60 Weekday 50 Weekend kWh per Hour (kW) 1. unoccupied level of consumption normally occurs seven days per week. (A little background information is in order. The key is to find a process that will unlock that information in a cost-effective way.— they are reduced to that state which allows them to be scientifically managed. Tools to take advantage of its value still need to be developed. When considering utility metering data. weather load can be extracted through statistical analysis. A critical piece of information also recognized is that in a hospital – as well as most other business activities – process consumption varies in proportion to schedule (in this case the doctor’s 9-5 M-F). There are two viable methods for looking at the annual pattern of consumption from interval data. To see if this pattern holds together over an entire year. begins to tail at 5 PM and everyone is asleep by midnight. while consistency confirms good practice.200 Time of Day then look at the pattern that emerges. All rights reserved. consumption is far lower indicating that the doctors are responsible for driving consumption during their work hours. It is obvious that Weekend consumption trends lower than Weekday. Minimizing Complexity Weekdays versus Weekend consumption patterns were identified in the previous graphics. And as they are developed they need to be integrated into the daily operation of operating a building. it is easy to see that energy consumption is a function of process and schedule. computers. – that people use when at work. 1. this method will allow the load profile of the facility to be extracted. Rearranging the data will provide better insight to the pattern. the simplest step is to summarize the data by day. then overlaying each day upon the next. This colinearity of factors that drive consumption will make it simpler to define a model. Monday through Friday appears to have the same peaks. and facility managers. The sample data has been derived from a hospital. service companies. 0 3/29 4/29 5/29 6/29 7/29 8/29 9/29 10/29 11/29 12/29 1/29 2/29 Day of the Year –4– . As we have seen when each day of the week’s pattern (24-hour) was overlaid. Inc. there is a consistency for day type according to the hour of the day. because the bills predict future consumption on the most detailed levels. Pulling out the patterns requires an understanding of how buildings operate as well as the ability to perform complex statistical analyses. allows you to look into the pattern.200 kWh per Hour (kW) 1. Based upon this chart. Every decision regarding the construction and use of a building is locked within the bill. Therefore. levels off around 9 AM. are substantial.000 800 Mon 2/5 600 400 Tue 2/6 Wed 2/7 Thu 2/8 Fri 2/9 200 Sat 2/10 Sun 2/11 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Controlling the View Interval meter data provides a rich source of information for energy traders. but with considerably lower peaks. That pattern holds true.) There is a standard workday pattern apparent. the first method is to examine the annual pattern of each hour of the day. Occupancy begins to ramp up © 2004 Interval Data Systems. Even though the building is occupied on the Weekend. a simpler method is to sum all twenty-four hours into a single day and examine the daily pattern.

The nature of facility operations is such that this model is also an accurate predictor of future consumption. Deregulation: Putting Value on Risks in Energy Supply Contracts We have seen that historical energy consumption for a facility can be modeled. process and weather. each CDD is equal to the MDT minus 51° F. Weekday/Weekend & Holiday.6345 Price ($/mWh) $100 $80 $60 $40 $20 $9.000 Volume (mWh) © 2004 Interval Data Systems. 0 0° F 10° F 20° F 30° F 40° F 50° F 60° F 70° F 80° F 90° F 100° F Mean Daily Temperature The colinearity of schedule and process is an extremely important insight into energy consumption patterns.000 23. Mean Daily Temperature (MDT) was actually represented by calculating site-specific Cooling Degree Days (CDD). Mean Daily Temperature). In this case. For companies where the sharing of production information can be an issue. A typical pattern of occupancy. Price versus Volume (ISO-NE June 2002) $160 $140 Trendline Price vs Volume $120 y = 3E-07x2 .0. whenever the MDT is greater than 51° F. All rights reserved. one for Weekdays and the other for Weekends.000 19. Even though there are particular processes that drive energy consumption differently.000 17. This means it is often possible to describe consumption patterns by an artificial variable (day type) as opposed to having to have access to the actual production data. e. As the market becomes more educated it will demand products that accurately match its risk profile with cost.000 13. describes a majority of facilities. Inc. At this point there is enough information about the patterns of energy consumption and the factors that describe it to build a model: consumption is a function of schedule. this is an extremely attractive benefit of the modeling process. The implication for commercial building energy consumption is somewhat obvious. p. The daily kWh is reorganized again to highlight its relationship to weather (as characterized by the most common metric.000 15. Thousands 60 Q = f (s. Thousands 60 50 Weekday Weekday Model Weekend Weekend Model 50 Weekday Weekend 40 kWh per Day 40 30 kWh per Day 20 30 10 20 0 0° F 10° F 20° F 30° F 40° F 50° F 60° F 70° F 80° F 90° F 100° F 10 Mean Daily Temperature One benefit of regression analysis is that the standard deviation of the model versus actual use can be calculated for both baseload and weather dependent consumption (±3% and ±26%.000 21.251 R2 = 0. The accuracy of the energy model has implications in both energy conservation potential and the cost of power. Using Regression to Build a Model Regression is used to develop the equations that describe consumption. a single factor will describe both. there is often a consistent schedule for those processes that can be extracted from the interval patterns of energy consumption. –5– .000 11.But there is another pattern overlaid on consumption that creates an increase during the summer months while blurring the Weekday/Weekend relationship. w) Since process is known to be colinear with schedule. Industrial facilities often exhibit this same pattern of colinearity among process and schedule. Two curves are shown is this chart.g.0067x + 58. respectively).

The model on the right.000 mWh/Day End Use/Day Poly. There is a caveat – something that is often lost in the euphoria of advanced metering. Benefits from advanced metering comprise a mass market.000 60. By default this presumes that the risk associated the volatility of consumption is equal across all end users. the model of consumption for the facility identifies an additional component of volatility. volatility associated with end use consumption is a significant risk in providing power. has unequivocally shown that market system to be flawed. the price of providing power for this facility can be nearly 5% greater.000 $90ºF Temperature Weekday Daily Volume vs End Use 500. (End Use/Day) Poly. Which one will it be? These are exciting times.000 250.000 $10 200. In the chart Price versus Volume.000 mWh/Day Avg. This volatility is unique. which requires different tools than the industry is used to supplying. …the promised market will fail to materialize. The variation from the mean price at any volume is clearly identified in the chart.69/m Wh 400. its wholesale cost of power would be $31. Energy modeling shows that this is not true and that the process of deriving the variation is economic. Recent trials with Price Responsive Load Management (PRLM) indicate that dramatic savings in system costs and increases in system reliability are achieved when demand management is included as part of the supply of energy. In order to quantify the risk associated with supplying power to a facility or group of facilities. Energy suppliers will push for its implementation because it allows a proper assessment of the risks within their portfolio.000 50.000 Daily Volume (mWh) Clearly.000 $40 One way to look at electrical power and quantify its volatility is to compare the price of electricity at the Independent Systems Operator (ISO) with the volume of electricity being delivered. © 2004 Interval Data Systems.25/m Wh 300.000 90ºF Temperature it represent a tool for the mass market since the presentation is far too complex. More importantly. 350. If the facility consumption exactly tracked that at the ISO.000 $30 300. regulated public utilities have used a highly complex methodology to determine the cost of service and set the price for energy consumption.400. 40. The next chart.000 150. This paper has shown one method that can be used for identifying and accessing value within advanced metering. (mWh/Day) 30. By no means does 10. which split the price of energy from the price of the delivery system. or $33. the average user cannot. just as the price of power has a risk component associated with its volatility. (Avg.25/mWh.000 200.000 $60 450.69/mWh. However.000 50ºF 60ºF 70ºF 80ºF 20. (mWh/Day) 50ºF 60ºF 70ºF 80ºF $20 250. Weekday Daily Volume and Price. Weekday Daily Volume vs End Use. All rights reserved. exposes an additional source of risk. The fact is that facilities that are managed better have less risk associated consumption volatility. Therefore it is possible to model the data as if it were a single interval meter. they will demand better management of energy within their own organization. the correlation between volume and price remains. Deregulation.000 450. redisplays the previous chart as a daily summary with separate plots for both the volume and average price for the day. Weekday Daily Volume and Price 500. you can see that the hourly price of electricity varies with the amount (volume) being purchased. –6– Facility Use (kWh/Day) Average Daily Price ($/mWh) Daily Volume (mWh) Historically. As is shown. Consumption at the ISO is the same as an aggregation of facilities. Although industry wonks can identify tremendous value within the data. but quantifiable.000 ISO Cost: $31.000 350. it helps to view the variables in the same context. for each facility. (Typically. $/mWh) Poly. $/mWh Poly. traders hedge their position by balancing their portfolio against changes in the price of energy. They will also demand retail power products that recognize the value of their efforts.000 $50 .) As users become familiar with this tool they will appreciate its value to managing their facilities. The price of energy has significantly greater bias toward time of use and seasonality than has been captured under rate structures.000 User Cost: $33. The state-of-the-art with regards to Information Technology for advanced metering data is about to change or. Inc.000 Conclusion Advanced metering will continue to expand in the marketplace because significant value can be found within its data. In the projection shown.

Interval Data www. Inc. Inc. All rights reserved. MA 02472 (+1) 617-744-1091 (+1) 617-744-1092 © 2004 Interval Data Systems. 104 Summer Street Watertown. Inc.For more information: Mailing Address: Telephone: Facsimile: E-Mail: Web: Interval Data Systems and the IDS logo are trademarks of Interval Data Systems.intdatsys. –7– . All other trademarks and trade names are the property of their respective owners.