Cost Avoidance Weather Adjustments in EnergyCAP

WATCH VIDEO ON COST AVOIDANCE--BASIC VIDEO 1

WATCH VIDEO ON COST AVOIDANCE--ADVANCED VIDEO 2

This topic deals with the theoretical basis for EnergyCAP weather adjustments in determination of Cost Avoidance. For related software interface and operation questions not discussed in this topic, see the Cost Avoidance Operations topic group.

Before making weather adjustments, EnergyCAP determines if a meter is weather sensitive. The important question is: Does the weather have a significant impact on how much energy this meter uses?"

Meters can be weather sensitive in the summer only, winter only, both, or neither. Usage patterns can be very different in each season. For this reason, EnergyCAP handles summer and winter graphs, weather factors, and adjustments separately and independently. If a meter shows no weather sensitivity, there is no rationale for weather adjustments to the baseline for determination of Cost Avoidance.

Understanding Degree Days

Weather analysis is based on a very specific type of weather data—the daily mean temperature. This historical data is widely available from a variety of sources. In EnergyCAP, there is even an option for automatic download of weather information for each weather station associated with the organization. EnergyCAP also provides an interface for manual data entry of known weather data.

Most facilities use energy for heating and/or cooling. The extent of use varies with the extremes of temperature. In EnergyCAP, these extremes are related to a theoretical non-use norm—the building Balance Point Temperature (BPT). The Balance Point Temperature is the point at which the building does not require energy for heating and/or cooling. Heating Degree Days refer to days where the daily mean temperature at the facility was below the building Balance Point Temperature. Cooling Degree Days refer to days where the daily mean temperature at the facility was above the building Balance Point Temperature. In EnergyCAP, the default building BPT is 55°F. However, this value can be changed globally or on an individual basis for each building (place).

EnergyCAP compares the individual building BPT and the known daily mean temperature to determine the number of heating or cooling degrees in the day.  More significantly, EnergyCAP sums the daily heating degree days and cooling degree days to determine the total number of heating and the total number of cooling degree days for the known utility billing period.

EXAMPLE:

On April 24, 2007, the mean daily temperature for the area around Fire Station Alpha was 51°F. The building BPT is 55°F. Therefore, there were four Heating Degree Days on April 24. A general cooling trend in April brought the total of Heating Degree Days for the utility billing period to 150.

As can be seen from the discussion following, the Degree Day concept can be very useful in comparing weather severity from year to year. Heating or Cooling Degree Days can be related to utility billing information to assess the weather sensitivity of individual meters, and to provide for a more accurate assessment of Cost Avoidance from year to year.

EXAMPLE:

In June of 2007, there were 140 Cooling Degree Days associated with Fire Station Beta. In June of 2008, there were 96 Cooling Degree Days. Per this information, June 2007 would probably required more air conditioning than June 2008. Therefore, if the meter serving Fire Station Beta is weather sensitive for cooling, the facility would have less weather-related consumption (and less cost, barring any significant utility rate changes) in June 2008 versus June 2007. To answer the question, “How much more/less would I have spent in 2007 if I had experienced 2008 weather?” we need to account for any variation in utility rates, and we also need to determine the degree of weather sensitivity for the meters associated with the building. EnergyCAP uses a simple linear regression analysis (see next section) based on several months or years of data to find how much meter consumption is due to weather and how much is not (the “base load”). The base load consumption for June 2007 is added to the consumption that would have been used, if the meter had experienced June 2008 weather. The adjusted June 2007 consumption can be compared to the June 2008 consumption to determine how much more/less consumption occurred. In addition, multiplying the adjusted June 2007 consumption by the Average Unit Cost of the June 2008 bill gives how much the adjusted June 2007 consumption would have cost us in June 2008. Comparing the adjusted June 2007 cost to the June 2008 costs answers the question, “How much more/less would I have spent in 2007 if I had experienced 2008 weather?”

The EnergyCAP approach is to plot the baseline bills, month-by-month, on a grid with consumption on the vertical axis and weather (as expressed using the “degree days” concept) on the horizontal axis. EnergyCAP does this internally via statistical linear regression analysis, but it is possible to obtain the same results by plotting the monthly consumption data on a graph or in a spreadsheet program.

Each point represents an electric bill received in a baseline winter month. The "best fit" line is found by statistical single linear regression analysis, and is called the "regression line."

The regression line in this case would be nearly horizontal. The meter uses about the same amount of electricity every month. Is it weather sensitive in winter? No, because the cold weather has no effect on usage. Usage stays the same.

The slope of the line is zero, which means that the weather factor, expressed as KWH per degree day, is zero.

Since the meter is not weather sensitive, there's no reason to adjust the baseline for degree days.

NOTE: A meter may be weather sensitive even though it is not statistically apparent from the baseline data.  The typical case is a building that was so poorly controlled (e.g. open windows are used to control excessive heat in the winter) that there was no statistical correlation between baseline usage and baseline weather.

Consider this data from a gas meter in winter. This graph is very different. The usage increases as the weather gets colder, and decreases as the weather gets milder. Clearly, this meter is weather sensitive.

At the point of zero heating degrees (mild weather), the usage is zero. This means that the meter has no base load. There is no non-weather sensitive usage. All of the energy used is related to the weather.

This winter utility data from an electric meter shows a consistent slope which rises to the right, so the meter is weather sensitive. In the cold months such as December, January, and February, the weather load is greater, while in the mild months the weather load is smaller.

Note, however, that unlike the previous example, the Y-intercept (the point at which the line hits the vertical axis) has a value greater than zero —in this case, about 5,000 KWH. The Y intercept reveals the base load—the non-weather-related portion of the total usage. This meter uses 5,000 KWH even if there is mild weather (zero heating need degrees) plus some additional KWH depending on the weather. Once the weather load for the meter has been established, it is possible to more accurately adjust the baseline consumption for Cost Avoidance calculations.

Baseline Adjustments and Cost Avoidance

Baseline adjustment for non-weather sensitive meters

In EnergyCAP, a non-weather sensitive meter must meet one of the following criteria:

• The user has unchecked the “Attempt summer/winter weather adjustments to baseline?” option in the Meter Properties – CAP tab (see image below).

OR

• The baseline length, as defined in the Meter Properties window, is greater than 12 months.

Setting the Baseline Length to any value greater than 12 months disables the weather analysis, even if the Usage Adjustments checkboxes have been previously enabled.

OR

• EnergyCAP performed a usage vs. weather analysis for this meter and did not find an acceptable correlation between usage and degree days. Correlation data is displayed in the Cost Avoidance – Savings and Normalizations – Use vs Weather tab; a “red” background color for the statistical results indicates a lack of correlation in the statistics, while green means that a weather correlation exists.

If a meter is found to be non-weather sensitive, no weather adjustments will be made to the baseline consumption.

Cost Avoidance calculation for non-weather sensitive meters

The EnergyCAP processor calculates Cost Avoidance for non-weather adjusted meters using these steps:

1. Obtain current bill data. A "current" bill is a bill in the post-retrofit period; i.e., a bill that has a billing period End Date later than the Savings Start Date.

Current Billing Period:        7/15/06 to 8/14/06 (30 days)

Current Usage:                    42,000 KWH

Cost:                                       $3,750.00

1. Using the 365-day baseline period for the meter, extract the daily baseline usage for the matching period of 7/15 to 8/13* and total the daily baseline usage values for each day in the 7/15 to 8/13 period. This total usage is called the "without CAP" because it is the raw baseline that has been "adjusted" to match the billing period of the current bill, and reflects what would have been used in the absence of a Cost Avoidance/energy management program.

*NOTE: The current bill is 7/15 to 8/14 but we use the baseline period of 7/15 to 8/13. Why? For convenience, EnergyCAP assumes that bills are read at noon, so the current billing period of 7/15 to 8/14 includes 29 full days (7/16 to 8/13) and two half days (7/15 and 8/14). To keep it simple, EnergyCAP uses all of the start day (7/15) and none of the end day (8/14) when processing the baseline. So, the last day is always dropped. This gives us the correct number of 24-hour days in the billing months and year.

1. Next, adjust the baseline usage value for any floor area adjustments (proportional) or other special adjustments. For this example, let’s assume the sum of daily baseline usage without CAP for that billing period after all adjustments is 45,500 KWH.
2. Next, calculate the cost of the 45,500 KWH used in the baseline year. Using today's average unit cost as the costing method, we first calculate the unit cost of today's energy.

For example: $3,750 ÷ 42,000 KWH = $0.0843 per KWH

Then, we apply that unit cost to the adjusted baseline usage of 45,500 KWH:

45,500 KWH @ $0.0893 = $4,063

The “without CAP” cost is $4,063. This means that we would have spent $4,063 in the period of 7/15 to 8/13 if our consumption level had stayed the same as the base year and we paid today's rates.

1. Finally, calculate the Cost Avoidance.

Cost Avoidance is simply the difference between what we would have spent and what we did spend. Thanks to energy management, our actual spending was lower than expected, so we have a savings.

Cost Avoidance = Without CAP Cost - Actual Current Cost

Adjusting the baseline for weather-sensitive meters

The baseline creation process for a weather sensitive meter is very different from that of a non-weather sensitive meter.

1. Determine that the meter is weather sensitive. To do so, enable the meter for weather regression analysis using the Attempt Weather Adjustments checkbox option in the Meter Properties - CAP tab. Once enabled, the statistical analysis will reveal if the meter is weather sensitive. If a meter is weather sensitive, the Weather Factor and R-squared values are displayed on the Use vs Weather tab against a green background, as illustrated below.
2. Calculate the degree of sensitivity using a statistical technique. A meter that consumes 1 KWH per cooling degree is not highly weather-sensitive (the weather does not cause high energy usage), while a meter that consumes 100 KWH per cooling degree is much more sensitive. The regression analysis calculates a "weather factor" in units of usage per degree day of weather. The slope of the statistical regression line is the indicator of weather sensitivity. TO view this line, click the Use vs Weather tab at the bottom of the Cost Avoidance Manager window. There are separate slopes for summer and winter. Click on the Display Summer/Winter radio buttons to select the desired analysis. A meter may be weather sensitive in one season but not in the other, as in this example.

Only the summer shows a positive (green) correlation for this meter. Winter calculations are highlighted in red, indicating a poor correlation or no correlation.

1. "Disaggregate" each baseline bill into daily weather and non-weather components.

In the chart above, the daily “base load” is indicated in dark blue and the weather load is indicated in light blue.

1. Store the daily weather and non-weather components in the baseline file.

It is possible to change the weather factor by deleting any abnormal "outlier" bills in the baseline year. An "outlier" point is a month that has an abnormal usage vs. weather value.

A good example of an outlier is the July (Jul) bill (see graph, above) for a school that is shut down from June 20-August 1. The July weather is hot, so the "predicted" usage is high, but the actual usage is low because A/C is turned off.

Clearly, July is an abnormal outlier. The outlier point affects the weather adjustment by lowering the R-square value, often below the minimum. This causes EnergyCAP to ignore weather adjustments in all months, even though other spring and fall months may be good candidates for adjustments.

The EnergyCAP Use Vs Weather tab enables the user to remove an outlier month from the weather calculation: simply move the mouse cursor to the outlier point and click on the point.

NOTES: EnergyCAP may have already deleted some outliers. Any point more than two standard deviations away from the regression line will be removed from the analysis automatically, and the point indicated by a red diamond rather than a green circle. Any points outside of the lines (which indicate one standard deviation from the regression line) are possible candidates for manual deletion, but may be valid. Do not remove points outside of the lines unless you are confident that they are abnormal and not representative.  When a point is removed from the analysis, it will change from a green circle to a red diamond. You can include it by clicking on it again.

It is not possible to include far outliers that have been automatically removed by the EnergyCAP software.

When outlier points are removed from the regression analysis, the entire date range of that bill is removed from weather analysis. For example, if you or EnergyCAP removes a 7/03 – 8/05 bill, the energy usage associated with that date range will never be weather adjusted — the non-weather sensitive rules will apply.

The minimum R-squared value changes when outliers are removed because the minimum is based upon the number of points. Fewer points require a higher R-squared value for the same 95% confidence level.

Adjusting the building Balance Point Temperature (BPT)

It is possible to adjust the "heating needed below" temperature (for winter analysis) and the “cooling needed above” temperature (for summer analysis) to a setting that may be more appropriate. The “heating needed below” temperature must be less than or equal to the “cooling needed above” temperature. EnergyCAP allows for a “dead-band”, a range of temperatures where the building requires neither heating, nor cooling, but it does not allow an overlap range of temperatures where the building is being heated and cooled at the same time. These set points are for the building “balance point temperature” (BPT) used to calculate the number of heating or cooling degree days for each day in the billing period. Based on experience, a building manager may be able to assess the proper BPT for a specific facility with great accuracy.

As the BPT for summer or winter is changed, the R-square value will change. There may be many temperature settings that all have R-square values over the minimum required for a weather correlation. Which one is best?

Here's where some judgment and understanding of the building is important. The very highest R-square value is not always the best.  For example, the best R-square for an elementary school may be at 37º, but the building manager knows that the boiler controls turn on the heat at a 55º outside temperature. Use 55º instead of 37º.

Since the regression line must be a single straight line, EnergyCAP plots the line as shown above.

However, the figure below shows a more appropriate plot for these points — the combination of a flat and sloped line. This reveals that the BPT is probably set too high (if heating) or too low (if cooling). The leftmost months are essentially all the same usage — weather has not yet affected them.

Experiment with editing the BPT so less sensitive weather months "fall off" the graph to the left, providing a better estimation of heating needed below temperature, as shown below.

Knowledge of a building can be used to make better assumptions. Is September a heating month? If so, it should appear on the heating graph. If not, lower the heating temperature so that September falls off the graph because it isn't a heating month. Use the same approach for cooling months.

As the BPT changes, several changes may be observed:

• The minimum R-square value changes as more or less points (months) appear on the graph. The minimum is based upon the number of points; fewer points require a higher minimum for the same 95% confidence level.
• The months will shift to the right or left. The vertical height of the point for each month is the usage, so that won't change. The horizontal position of each month is the degree days. Because a change in temperature changes the degrees of heating need, the month will move left or right. For example:

With more extensive changes to BPT, more months will disappear from the graph. As the temperature is lowered, some months will no longer be considered "heating" months because they will no longer have heating need. They will disappear from the graph.

If you don't use the Use vs Weather screen to adjust the weather factor, the processor will automatically create one for you using 55ºF as the default temperature, automatically deleting any outlier bills that are more than two standard deviations outside of the regression line.

EnergyCAP uses the weather factor to disaggregate each baseline bill. The technique is really quite simple. EnergyCAP starts with the first day in the billing period. Since the weather factor is the consumption per degree day, the weather factor times degree days equals the total weather-related consumption.

EXAMPLE:

100 KWH per degree of heating need x 10 degrees of heating need = 1,000 KWH

By multiplying the factor times the heating degrees on this day, it is possible to closely estimate how much energy was used on this day in the base year for weather related loads (heating or cooling).

EnergyCAP steps through each day in each baseline bill and calculates the weather component in this manner. Next, EnergyCAP totals up the weather usage for each bill and subtracts it from the total consumption. The "left-over" usage must be the non-weather (lights, equipment) component because the weather component has already been accounted for. This non-weather component is then evenly assigned to each day of the billing period.

The result is that each baseline bill is disaggregated into a daily weather component and a daily non-weather component. The sum of the daily components always exactly equals the actual bill.

NOTE: The CAP06 Baseline Report shows the result of the disaggregation and can be used to verify the process.

Calculating Cost Avoidance for a weather-sensitive meter

EnergyCAP calculates Cost Avoidance for weather sensitive meters on a daily basis following this procedure:

1. Retrieve the current year bill.
2. Retrieve the bill start and end dates. Each day is processed to include the start date but exclude the end date. (The end date is processed as the start date on next month's bill.)
3. Retrieve the usage and weather for day one.
4. Retrieve the baseline data.

NOTE: You can use the CAP06 Baseline Report to print the baseline data for each day. The baseline will have two components: non-weather and weather.

1. (see *NOTE below) The non-weather component stays the same throughout the month, except for the “fringe” dates around the first or last day of the billing period. They may be different since there is a possibility that the monthly utility bills will not match up precisely from year to year.
   
   The weather component for the day is determined by multiplying the number of degree days by the slope of the corresponding best-fit straight line of Use vs. Weather from the baseline. For example, if the current day has 10 CDD and the meter has a baseline Use vs. CDD slope of 57.0, then the meter would have consumed 570 units to maintain a comfortable indoor building temperature.

1. Add the adjusted weather sensitive component to the unadjusted non-weather component. The result is the "without CAP” usage (adjusted baseline) — the baseline after it has been adjusted for today's weather condition.
2. If any special adjustments of floor area adjustments exist, apply them to the baseline.
3. Calculate the average unit cost of today's bill by dividing total cost by total usage. This is the current average unit cost.
4. Multiply the average unit cost times the “without CAP” usage. The result is the “without CAP” cost for this day.
5. Repeat steps 3 through 9 for each day of the current billing period (from step #2).
   NOTE: From the Savings tab, you can open a new window with the day-by-day results for the month by clicking the relevant Cost Avoidance data table entry or graph bar. If you click on the graph outside of the display bars, the last-viewed month will display.
6. Sum all the daily “without CAP” baseline costs.
7. Subtract actual current cost from the “without CAP” adjusted baseline to determine Cost Avoidance:

Cost Avoidance = Without CAP Cost - Actual Current Cost

*NOTE: A special situation that may be encountered at step 5 is that a particular day may have been a heating day in the baseline year but a cooling day in the current year. For example, in the 2000 baseline year, April 3rd was a heating day because the average temperature for the day was 45ºF. This year, April 3rd was a cooling day because the temperature was 66ºF.

The procedure in step #5 does not work in this case of a "swing" day. Instead, a slightly different approach is used. Since April 3rd of this year was a cooling day, and the baseline must be adjusted to today's conditions, it is necessary to calculate what would have been used in the baseline if the temperature had been 66ºF. First, the cooling weather factor (shown on the Use vs Weather tab) is examined.

• If the factor and R-square are zero, this means that the baseline year was not weather sensitive in the summer. The “without CAP” usage will be equal to the non-weather component with no weather component.
• If the meter is summer weather sensitive, then the cooling weather factor will be determined and added to the non-weather component. The result is the total “without CAP” usage for the “swing” day.

Peak Demand NOTES:

Peak demand is not adjusted for weather in EnergyCAP. Adjusting peak demand using peak degree days (the most severe day in the billing period) is unreliable for two reasons:

• There is a two-in-seven chance (30%) that the worst weather day will occur on a weekend or holiday when mechanical systems are in setback modes.
• Extrapolating peak demand using peak weather can easily result in a calculated value that exceeds the installed load of the building.

Because peak demand weather adjustments are unreliable, they have been removed from all versions of EnergyCAP and FASER since 1992.