Talk:Load Composition

= WECC/LMTF San Diego Meeting (4 March 2009) Comments =


 * 1) High priority
 * 2) * Add A/C SEER values as a function of temperature (see A/C test results). [Done by Dchassin 04:29, 12 May 2009 (UTC), see ]
 * 3) * Air-conditioning penetration diversity factor needs to be separate. [Done by Dchassin 19:41, 12 May 2009 (UTC), see ]
 * 4) * Get new rules of association from Dmitry
 * 5) * Separate ZIP to be six elements Z, I, P w/ P and Q separated on the feeder tab. [Done by Dchassin 22:42, 12 May 2009 (UTC)]
 * 6) * Retail process load is too high [Done by Dchassin 22:07, 12 May 2009 (UTC), see ]
 * 7) * Run validation cases (especially Vancouver WA and somewhere in California)
 * 8) Low priority
 * 9) * Temperature units in the load composition table still show the TMY units in the labels. This is confusing.  Try hiding everything in [].
 * 10) * Latitude unit is N. It should be degrees (N is positive).
 * 11) * Check units of solar flux Wh/m^2 vs. Btu/sf.h. Probably ok, because of hourly conversion, but check anyway.
 * 12) * Summer Phoenix 65% Motor D is right per Richard Bravo.
 * 13) * How to use the ZIP fraction?
 * 14) * Confirm that composition can be scaled by energy by using the “scalar” cell on “composition” tab.
 * 15) * AMI data are available. Explain how that might be incorporated.
 * 16) * Rule of association: engineering judgment. Current data are based on experience. Need to be reviewed by MVWG. Emphasize importance of reviewing these in instructions.
 * 17) * Commercial buildings: sf is set to the same sf for one category.
 * 18) * Remarks on commercial tab need to be revised. Some are not limited.
 * 19) * PF is for ZIP only on the feeder tab. Change the label to be “ZIP PF”.
 * 20) * Design conditions are used as extreme conditions on sensitivity tab. Summer – 1 degree down; winter – 1 degree up.
 * 21) * The table layout is important as formulas rely on the layout. Changing the layout is dangerous.
 * 22) * Need to distribute to MVWG for review and test.
 * 23) * Step-by-step instructions to get the spreadsheet from the internet. Website is SourceForge project GridLab-D, search for "load composition".

Comments from Anatoliy

 * Note : This comment was receive by email and posted for reference. It refers to an Excel spreadsheet that cannot be uploaded.  Contact David Chassin to obtain a copy. --Dchassin 17:43, 23 March 2009 (UTC)

Please consider a practical method to accommodate user knowledge about load class proportions for a given feeder/substation/area. This method is compatible with your “from bottom up” approach. The needed additions are illustrated in the two attached spreadsheets which supposed to be added to your Excel files.

The idea is in considering typical customer numbers for load categories in the “composition” sheet (column B) as variables. To find a number of customers (or scaling multipliers to initially assumed numbers), we need to calculate monthly kWh’s for the initial mixture of typical customers and correct the number of customers to satisfy the proportions given by the user for the entire feeder/substation/area. The attached Sheet 5 includes two load shape tables I obtained using your tool for the same set of conditions and differed only by the day of a week (weekday and weekend). Sheet 5 calculates kWh’s for 24 hours of the given day by summarizing every hour kWs. Doing that for weekdays and weekends of the given month and multiplying totals by say 22 and 9 and dividing the sum by 31, we get the needed average monthly energy for the given mixture of typical customers.

Sheet 5 includes also fractions of average daily energy obtained for 6 transmission buses for the same month as the load shapes (in general for unlimited number of buses).

Multipliers for adjusting customer numbers are calculated to satisfy the following: $$E_{ti}/E_{t\Sigma} \cdot M_i = E_{bi}/E_{bS}\,\!$$ where $$E_{tS}$$ - daily average energy for original mixture of typical customers $$E_{tS}$$ - daily average energy for class $$i$$ of typical customers $$E_{bi}$$ - billing data based fraction of daily average energy for class $$i$$ bus load $$E_{bS} =1$$, $$M_i$$ - multiplier to reduce/increase number of all customer categories inside class $$i$$ loads. $$M_i$$ values are calculated in cells  using formula: $$M_i = E_{bi} \cdot E_{tS} / E_{ti} \,\!$$ Sheet 5 calculates multipliers for customer numbers assuming that the original set of numbers correctly reflects customer mixture inside load classes (we would not need to make this assumption having load shapes for each customer category and corresponding billing data based fractions for those categories). Multipliers from Sheet 5 are used in Sheet 4 to scale numbers of customers. Besides this scaling, Sheet 4 is a duplicate of the “composition” spreadsheet.

Cell  in Sheet 5 is used for selection of a bus to process. Having say 3 in  puts the corresponding multipliers (with “3” in column K) to buffer 1 for using in Sheet 4. The results from Sheet 4 are delivered to the output model composition table via buffer 2. All the results can be obtained in one Excel execution with automatic increase of numbers in  (not done).

Another comment from Anatoliy

 * Note : This comment was received by email and posted for reference. --Dchassin 17:43, 23 March 2009 (UTC)

One my comment in the previous e-mail was about defining component P fractions relatively to total real power. The formulas for fractions should also take into account the exact meaning of CMPLDW inputs for zip load components. The coefficients P1c, P2c, Q1c and Q2c are defined in CMPLDW as fractions of only ZIP real and reactive powers (not total powers). Btw, do we really want to use LCM for deriving zip real and reactive load compositions? How sure are we LCM provides this composition for nominal voltages? I would fill better using default values for P1c, P2c, Q1c, Q2c and PFs.

Do we consider electronic load 100% real? If not, where is PFel needed for CMPLDW?

It looks like I figured out how to count the bus-by-bus load composition for unlimited number of loads in one Excel execution. It takes 20-30 seconds to get adjusted load model compositions for 1,400 loads.

I also made some corrections in additional sheets to count the situations when assumed number of customers in one or more load classes is zero but it is not zero in the known composition. If such situation occurs, there is no multiplier to the customer numbers which could provide a desirable proportion. To exclude such situations, I corrected formulas for customer numbers by always adding 0.1 customer (could be 0.01 or even less) for each load category in the "composition" sheet. This provides non-zero load shapes and energies for all categories and always derives the finite value scaling multipliers. Unfortunately, I need to make the described formula corrections in the main "composition" sheet. I would prefer to do this in a copy if this sheet, but coping does not work for your "Update Loadshapes" button. Is there simple way to correct your macros and get this button in a copy sheet?

Comment from Richard Bravo

 * Note : This comment was received by email and posted for reference. --Dchassin 17:46, 23 March 2009 (UTC)

The fraction of Motor D is somewhat similar for all regions. We believe that the current spreadsheet assumes that every house in each area has an air-conditioner. This assumption is not true, particularly for coastal and northern areas. I agree that new or remodeled homes will have A/C but older might not.

For Portland, while newer residential construction is very likely to have an air-conditioner, most of older construction does not.
 * 1) There is a value of adding a parameter “residential AC penetration” to the calculations. I think the same thing happens here in SCE.
 * 2) How can we get information on residential AC penetration by region? We might have to do this by area instead by region. In our case we have low penetration of A/C near the coast (~20%) where the temperature does not increase above 85F. We have high penetration of A/C inland (40% to 65%) where the summer temperatures goes from 90 to 110F. The case of Portland might be on about 20% but if you go east this numbers might rise drastically.


 * A less important question: Is there a value of adding diversity in the thermostat settings? Not everyone will turn their AC on during a normal (80-85F) summer day. What about a coincidental factor? We know that some air conditioners are "shut down" but others will be cycling (momentarily off) by thermostats.


 * Fixed in V1.6.6 : See

= SEER temperature correction =

[--Dchassin 20:46, 11 May 2009 (UTC)]

The approach to including outdoor air temperature effects in the cooling load calculation is as follows.

The following inputs are required
 * $$SEER$$ is the (Seasonal Energy Efficiency Rating) value for the average equipment in use in Btu/W.h.
 * $$T_{design}$$ is the cooling design temperature used for sizing the equipment in &deg;F.
 * $$T_{set}$$ is the thermostat cooling setpoint in &deg;F.
 * $$T_{out}$$ is the outdoor temperature in &deg;F.

The maximum EER (energy efficiency ratio in Btu/W.h) is typically used to calculate the SEER value using the assumption $$SEER = 0.875 EER_{max}$$. The EER is determined as follows:

$$ EER_{max} = EER_{coeff} \times 3.412 \frac{T_{set}+459.7}{{T_{design}}-T_{set}} \qquad (1) $$

where the $$EER_{coeff}$$ is typically around $$0.1$$. This is the relatively invariant ratio of the Carnot efficiency to the actual efficiency and can be used to compute $$EER_{coeff}$$:

$$ EER_{coeff} = \frac{SEER}{0.875} \frac{T_{design}-T_{set}}{3.412 (T_{set}+459.7) } \qquad (2) $$

The theoretical EER, $$EER_{theoretical}$$ can thus be computed for any outdoor temperature

$$ EER_{theoretical} = EER_{coeff} \times 3.412 \frac{T_{set}+459.7}{{T_{out}}-T_{set}} \qquad (3) $$

The test results obtained do not show this large variation in EER and $$EER_{theoretical}$$ must be adjusted using the following formula to obtain the $$EER_{model}$$

$$ EER_{model} = EER_{theoretical} \left ( 1 - E_{factor}(T_{design}-T_{out}) \right ) \qquad (4) $$

where
 * $$E_{factor}$$ is estimated by minimizing the weighted RMS error between the observed power used in the tests and the model's prediction at various cooling design temperatures, using the weights for the equipment sizes, as shown in Table 1. The results for the model calibaration are shown in Table 2.


 * Note : If the outdoor temperature is below 83&deg;F then the SEER values are for 83 &deg;F are used.



A quadratic fit of this data yields the following relation

$$ E_{factor} = 4.6\times10^{-5} T_{design}^2 - 1.08\times10^{-2} T_{design} + 0.656 \qquad (5) $$

with $$R^2 = 0.9969$$.

Figure 1 shows the diversified AC load curves (in kWh/h) as a function of outdoor temperature (in &deg;F) for various cooling design conditions.


 * Caveat : This approach assumes that the thermostat setpoint is maintained. If the duty cycle is 100%, that assumption is not valid and the efficiency may be better than that predicted based on the change in outdoor temperature alone.

= Air-conditioner penetration factor =

[--Dchassin 19:21, 12 May 2009 (UTC)]

The air-conditioner penetration factor specifies what fraction of homes have operating air-conditioning units. This number is based on the cooling design temperature established for the model, as follows:

$$ r_{AC} = \begin{cases} 0.1 : T_{design} \le 80 \\ 0.1 + 0.9\frac{T_{design}-80}{20} : 80 < T_{design} < 100 \\ 1.0 : T_{design} \ge 100 \\ \end{cases} $$

The air-conditioning penetration factor is used to discount the diversified cooling load as follows:

$$ D_{cooling} = r_{AC}\ D_{cooling}\ P_{cooling}\ A/1000

$$

where
 * $$D_{cooling}$$ is the cooling duty cycle
 * $$P_{cooling}$$ is the cooling power density
 * $$A$$ is the floor area


 * Fixed : Done in version 1.6.6 [--Dchassin 15:59, 12 June 2009 (UTC)]

= Retail process load density is too high =

[--Dchassin 22:08, 12 May 2009 (UTC)]

This problem seems to stem from a very low square footage for building with process loads in the CEUS data. As a result, the normalization process give it too much weight. The solution is to make the square footage representative of reality, rather than the sampling of the CEUS data.


 * Fixed : At this time, I recommend using 50,000 sf instead of 1,000 square, but this number is just a guess. This will need to be researched more carefully before we commit to any value. Done in version 1.6.6 [--Dchassin 15:59, 12 June 2009 (UTC)].

= Comments from Dmitry =

These comments were received from Dmitry Kosterev (BPA) on Fri Jun 5, 2006.

Rules of Association
Please see [below] suggested "rules of association". Your guess may be as good (or better) than mine.

AC Load Calculation
There seem to be an issue with AC load calculation. In load shape calculation, the kW amount of Motor D (corresponding to AC load) is nearly constant for the day.


 * Fixed : The problem is caused by how the temperature is determined based on the study type. Peak load studies used the peak temperature regardless of the time of day.  This was fixed in version 1.6.7 by having the temperature (and all other conditions) be taken for the time of day given. [--Dchassin 15:57, 12 June 2009 (UTC)]

Enduse Electrification by City
Is it possible to have a support file (or a sheet) with default entries for each city:

The default data is to be used in "Feeder" tab calculations.

The reason is that Phoenix may have 100% houses with air-conditioners, while Portland say 50% and Seattle 25%.


 * Done : This was fixed in version 1.6.6 [--Dchassin 15:57, 12 June 2009 (UTC)]

= Rob Pratt comments =

Single family

 * 1) Building height default to 8 feet
 * 2) Roof insulation should be composition less than R60, suggest R40
 * 3) Window should be about R2.5 (double w/low-E)
 * 4) Window/wall ratio is high (probably less than 10%)
 * 5) Suggest adding door wall ratio with door Rvalue (2 doors minimum)
 * 6) ACH should not be below 0.5 and no usually above 1.0 on average
 * 7) heating and cooling capacity needs to include internal gains
 * 8) check solarexposure calcs for fully diversified orientation
 * 9) building UA should include doors
 * 10) need to modernize ELCAP gains
 * 11) add UA heat gain to latent load numerator
 * 12) in rules of association
 * 13) what about CFLs
 * 14) why ZIP in washing? isn't washing 100% motor-D?
 * 15) there should be 1.3 fridges per house and 0.3 freezers per house
 * 16) ELCAP shapes should be normalized before being includes; the magnitudes don't mean much anyway.
 * 17) daily energy use is missing drying and refrigeration/freezer (table 5)
 * 18) the load calc equation is not shown
 * 19) needs to explain heating temperature COP discount

Multi family

 * 1) can't tell how many apts exclude dryer/washer or second fridge/freezer
 * 2) electrification is generally higher in apts

Commercial

 * 1) shouldn't assume that 100% of interior lighting is not T8 (electronic ballast)