About Aggregations
Aggregations are measurements that represent a summarization of other measurements from a set of devices, items, and/or measuring components. For example, an aggregation may be based on the sum of the electric consumption of all residential customers in a particular postal code within the utility's service territory. The group of related measuring components for which measurements are summarized are referred to as constituent measuring components.
Every type of aggregation has one or more dimensions. For example, an aggregation might be defined with three dimensions as follows:
• Substation
• Feeder
• Transformer
Aggregations are defined by aggregator measuring components. A separate aggregator measuring component should be defined for every distinct combination of dimensions for a given type of aggregation. For example, the transformer aggregation described above would require an aggregator measuring component for every distinct combination of substation / feeder / transformer found among the electric Service Points.
Aggregations can also be aggregated. For example, the aggregator measuring components for each substation / feeder / transformer combination can in turn be used to aggregate consumption at each substation / feeder combination. In this case, a distinct aggregator measuring component would be defined for each unique substation / feeder combination. Likewise, the aggregator measuring components for each substation / feeder combination could be used to aggregate consumption at each substation, in which case, distinct aggregator measuring components would be defined for every unique substation. Finally, the aggregator measuring components for each substation can be used to aggregate total consumption for all substations, with an aggregator measuring component defined to represent the entire service area.
Aggregation Calculations
The system periodically aggregates consumption via batch process, using a deferred monitor on the aggregator measuring components. In addition, users can re-aggregate data in real-time if they don't wish to wait for the batch process, or if the original aggregation needs to be re-calculated due to incorrect data. Users can also create ad hoc aggregations "on the fly" and these will persist in the database in the same manner as other aggregations.
Understanding Aggregation Periods
The start and end dates and times for aggregation calculations are based on the following:
• Aggregation Horizon
• Aggregation Lag
• Aggregation Cut Off Time
Whenever aggregation is performed for an aggregator measuring component, consumption is aggregated for every day in its "Aggregation Horizon." The "Aggregation Horizon" is the number of days during which there's a potential change in measurement data for one or more of the measuring components associated an aggregator measuring component.
Aggregation calculations typically lag behind the current date by a few days to give the system time to upload and perform validations and create final measurements. The amount of lag time is referred to as the "Aggregation Lag" and is the number of days between the date on which aggregation calculations are performed and the end date of the aggregation period. This defines the time period between the aggregation calculation date and the end of the aggregation horizon that serves to allow all measurements to arrive. This together with the Aggregation Horizon is used to determine the start and end dates of an aggregation period. For example, with an Aggregation Horizon of 5 and an Aggregation Lag of 2, aggregation calculations performed on January 9 would be for an aggregation period of January 3 through January 7. The next day (January 10), the aggregation period would shift to January 4 through January 8.
![Understanding Aggregation Periods Understanding Aggregation Periods](../MDM_25000/images/D1_BG_image10.jpg)
![Understanding Aggregation Periods Understanding Aggregation Periods](../MDM_25000/images/D1_BG_image11.jpg)
Aggregation is always performed through a given "through time" (such as 12:00 AM) rather than through the actual time of the aggregation calculation. This time is referred to as the "Aggregation Cut Off Time." For example, the stop time for aggregation calculations with an Aggregation Lag of 2, and an Aggregation Cut Off Time of 10:00 PM will always be 10:00 PM 2 days prior to the date on which the calculations are performed. In the above examples, aggregations performed on January 9 would have an end date/time of 10:00 PM on January 7, and aggregations performed on January 10 would have an end date/time of 10:00 PM on January 8.
Aggregation and Re-Aggregation
The use of the aggregation horizon means that aggregated totals for some days will be re- aggregated until those days are no longer covered by the aggregation horizon. For example, with an aggregation horizon of 5 days and an aggregation lag of 2 days, on the night of January 9 the aggregation period would be January 3 through January 7 (as in the above example).
On the night of January 10, the horizon will shift 1 day (to January 4 through January 8, again as above). This means the following:
• The totals for January 3 calculated on January 9 will be untouched (January 3 now falls outside the aggregation horizon)
• The totals for January 4 through January 7 will be re-derived because corrections may have occurred (and they still fall within the aggregation horizon)
• The totals for January 8 will be calculated for the first time (because it now is within the aggregation horizon).
Manually Read Meters and Aggregation Lag
Aggregations that include manually read meters often have much longer aggregation lags than those for automatically read meters. This allows more time for manual meter readings to be imported into the system for use in aggregations.
For example, suppose a situation where manual meter reads arrive approximately 1 month after the date of the reading. In this case, it wouldn't be until a meter read upload on February 7 that the last manual reads including consumption from January 6 will exist. Since we don't want to perform aggregations for January 6 until there's a decent chance that all consumption for that date exists, an aggregation lag of 32 days ensures that the data for January 6 is in the system when the aggregation is performed on February 7.
When An Aggregator Measuring Component Contains Both Manual and AMI Measuring Components
The previous section suggests that if an aggregator measuring component contains both manually read and AMI measuring components, the aggregator measuring component should have a long lag time rather than the short time shown in the earlier example.
While this is one possible approach, an alternative approach could allow users to see timely aggregations of the AMI channels and only lag the manual channels. To do this they could configure the system as follows:
• Create an aggregator measuring component that only aggregates manual measuring component types (horizon: 5 days, lag: 31 days)
• Create a separate aggregator measuring component that only aggregates interval measuring component type: 5 days types (horizon: 5 days, lag: 2 days)
• Create a third aggregator measuring component that aggregates the above aggregator measuring component types (horizon: 5 days, lag 31 days)
Expanding The Aggregation Horizon Periodically
Even with a 30-ish day aggregation lag, it's possible for an account with several months of estimated consumption to have its consumption "invalidated" when real readings arrive whose dial readings are less than the estimated readings (or when major retroactive data change takes place). Another similar situation is when a user manually corrects consumption after the aggregation horizon has moved forward.
To address situations like these, the use of ad-hoc aggregations with a user-specified aggregation horizon supports the notion of periodically (such as once a month or once a week) aggregating with a long aggregation horizon (up to 90 days) to catch as many retroactive changes as possible.
A Note About Using Non-Effective Dated Dimensions
Be aware that whenever aggregation takes place, the system uses master data that exists as of that date. For example, if an aggregation is performed using non effective-dated master data attributes (for example, aggregating by measuring component type) and a measuring component's type is changed, all days in the aggregation horizon will reflect the change regardless of when the user makes the change, because the measuring component type relationship is not effective-dated.
The impact of this depends on the nature of the change:
• If the master data was wrong, this provides a way to re-aggregate the data to accurately reflect the master data.
• If the change occurred effective as of a given date (but there's no effective dated attribute), this can result in incorrect aggregation values.
Because of this, it is recommended that where possible dimensions used in aggregations be effective-dated.
For information on configuration, reference the related section for
Aggregations in the Administration Guide.
Parent topic