We present a new approach which identifies the number of customers wit
hin particular categories living in a chosen area of energy supply. Th
is identification is made from total energy consumption measured at di
stribution feeders by applying characteristic patterns of energy use a
nd nonlinear programming techniques. Customers were segmented into fou
r main categories with six sub-groups in each category. A total patter
n of energy use was obtained by summing for each sub-group the energy
consumption which is based on the characteristic patterns of energy us
e for that sub-group and the number of customers in that sub-group. Si
mulations were carried out to verify different optimization procedures
. Results show that a variable matrix optimization procedure can ident
ify customers most accurately by minimizing the assumed objective func
tion to a negligible value after 100 iterations. (C) 1997 Elsevier Sci
ence Ltd.