Subtractive Register Where UOM Does Not Measure Peak Quantity
Assume the following read history exists for a subtractive register (i.e., one where you have to subtract the previous read from the current read in order to derive consumption) where the register's unit of measure doesn't measure a peak quantity.
Read Date
Reading
Derived Consumption
15-Jan- 1999
1000
N/A (first read)
15-Feb-1999
3000
2000
15-Mar-1999
4500
1500
Next, assume the meter's trend profile looks as follows:
Read Date
Total Qty (Assume kWh)
No of Units (Total Days)
Number of Reads
Average Consumption (Per Day)
13-Mar-1999
6,000,000
135,000
4,500
44.444444
14-Mar-1999
900,000
15,000
500
60
15-Mar-1999
5,000,000
137,750
4,750
36.297641
...
13-Apr-1999
4,000,000
135,000
4,500
29.629630
14-Apr-1999
4,650,000
155,000
5,000
30
Note: 
The system keeps a separate trend for every combination of trend area, trend class, unit of measure and time of use code. Trend area comes from the premise at which a meter is installed. Trend class comes from the SP type of the service point at which a meter is installed. Unit of measure and time of use come from the register read's register.
Next, assume a new register read is recorded on 15-Apr-1999 with a value of 5000.
In order to calculate estimated consumption for this register read, we have to calculate average daily consumption using the following formula:
The following points describe exactly how the system calculates each variable in this formula:
• Determine Average Customer's Usage In Current Period:
◦ Extract the number of reads from the meter's service point's SP type's trend class. We'll assume this is set to 7,500 reads.
◦ Read trend records in reverse chronological order from the read date (assuming there is no trend record on 15-Apr-1999, we'll start amassing trends from 14-Apr-1999).
◦ Read enough trend records until the number of reads is greater than or equal to the trend class' number of reads. We will have to read 2 trends records to amass this value (the ones on 14-Apr-1999 and 13-Apr-1999). We remember how many reads we extracted from the trend table for use a little later. In this case, we used 9,500 reads.
◦ Divide Total Qty by No of Units. Using our example, we'd divide 8,650,000 kWh by 290,000 days to get 29.827586 kWh per day.
• Determine This Customer's Usage In Previous, Non-Estimated Period:
• Find the consumption associated with the previous, non-estimated read for the register's unit of measure / time of use codes at the service point. In this case, we'd find the read on 15-Mar-1999. Because this register is subtractive, we'd also need to find the previous, previous non-estimated read in order to calculate consumption. In this case, we'd find the read on 15-Feb-1999.
Note: 
Minimum Days Between Readings. The algorithm parameter Minimum Days Between Readings controls the minimum number of days needed between the previous non-estimated read and the "previous, previous" non-estimated read.
• Divide the total consumption by the number of days. Using our example, we'd divide 1,500 kWh by 28 days to get 53.571429 kWh per day.
• Note: if the customer doesn't have consumption in the previous, non-estimated period, e.g., if it's a meter at a new premise, the system assumes the customer uses the same as the average customer's usage in the previous period (see next point).
• Determine Average Customer's Usage In Previous Period:
• Read trend records in reverse chronological order from the read date of the previous reading (15-Mar-1999).
• Read enough trend records until the number of reads is greater than or equal to the number of reads amassed when determining the Average Customer's Usage In The Current Period (the first point). Using our example, we'd have to read 3 trends records (the ones on 15-Mar-1999, 14-Mar-1999 and 13-Mar-1999).
• Divide Total Qty by No of Units. Using our example, we'd divide 11,900,000 kWh by 287,750 days to get 41.355343 kWh per day.
• Next, determine the number of days of estimated consumption. This will be equal the number of days between the estimation date and the prior reading (note, the prior reading could have been estimated). Using our example, we'd have 31 days (the number of days between 15-Apr-1999 and 15-Mar-1999).
• At this point, we have everything we need to estimate consumption. This will equal ((53.571429 / 41.355343) * 29.827586) * 31 days. This is equal to 1,198 kWh.
• If we need to calculate high and low boundaries, we multiply 1,198 kWh by the high and low values defined for the register's unit of measure, the service point's service type and the read's read type.