Trend Classes and Trend Areas Should be Defined Appropriately
In addition to defining a large enough number of reads on the trend class, you want to make sure that the system can collect the number of reads in a relatively short period of time or the benefit of trends is diluted. For example, if the system collects the number of reads over a period of two days as opposed to two months, the seasonal trend information is better represented in your estimates.
Consider the following example. Assume the number of reads required in the trend sample is 1000 and your system collects the following trend information for the area/class:
Month | Number of Reads | Average |
March | 400 | 200 kWh |
April | 400 | 300 kWh |
May | 400 | 500 kWh |
The trended average in May is 333.33 kWh ((80000 + 120000 + 200000) / 1200).
However, if the number of reads required in the trend sample is 1000 and your system collects the following trend information for the area/class:
Month | Number of Reads | Average |
March | 1100 | 200 kWh |
April | 1100 | 300 kWh |
May | 1100 | 500 kWh |
The trended average in May is 500 kWh (550000/1100), which is a more accurate representation of the trend in May.
Note:
Example Values. The examples above are meant to illustrate the importance of ensuring that the system collects enough reads in a short enough time span to accurately capture trends. The values are summarized by month and do not represent actual records in a trend profile.
The collection of an appropriate number of reads is a function of the number of reads defined for a trend class and the number of customers who are in each trend class/trend area combination. Make sure that the number of reads is not too large for the number of customers who are in each trend class/trend area combination and that the trend class/trend area combinations do not create groups of customers that are too small to calculate accurate trends.
Parent topic