Dear Allstaters
We have the following data structure
Day Hour X1 X2 ... Xn Sum
1 0 200 42 ... 360 6060
...
1 23 232 48 ... 432 5460
and so on for several months, where
Xi is the energy consumes by cell i, measured by its own counter which
may involve error component.
Sum is the energy consume of all the cells in the cell group measured by
a group counter which is exact. We can have up to 50 cells in a cell group,
n=50.
Ideally, when all the counters are exact, we have X1+X2+:+Xn=Sum.
The problem we've faced is to measure the impact each cell counter has to
the common error, i.e. to the difference Sum - (X1+X2+:+Xn). In fact we have
to decide which counter must be adjusted first, second and so on.
The error component of a counter is dependent not necessarily in a linear
mode on the measured volume of energy, for the simplest model we, probably,
can assume the linearity: more the energy volume more the error
Obviously the energy consume is considerably influenced by the hour and the
day of week, so the simple regression model Sum=a1*X1+a2*X2+:+an*Xn is
inappropriate.
Does there exist any well-known model to settle this problem?
Any hints will be greatly appreciated.
Alexandru Hlibiciuc
Department of Mathematical Cybernetics,
Moldova State University,
Chisinau, str. Livezilor 60, Moldova, MD-2009.
e-mail: [log in to unmask]
[log in to unmask] (current address)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|