| Summary: | Electricity has become an important aspect in daily life. In Indonesia, PT.
PLN has a role as the biggest electricity supplier. To collect and supervise the data
usage of customer, PT. PLN use human labour as the data collector. The
Indonesian kilo Watt hour (kWh) meter reading used mobile phone to store the
amount of usage, mobile phone camera to capture images of kWh usage, and a set
of papers to record in physical form. These processes are repetitive and time
consuming. Moreover, some human operators absence to collect data and record
wrong electricity usage deliberately which increasing the customer awareness. All
of the fraudness possibilities demand PT.PLN to continuosly improve its
performance, spesifically in operational area.
The purpose of this paper is to build software in order to improve existing
data collection processes that integrated with billing calculation and transaction
processes. The automated system presents the artificial neural network (ANN)
method, Backpropagation and Kohonen-SOM algorithm, to recognize the number
of electricity usage. The object of the research are image, computation time, and
billing calculation. On this research, the comparison of processing time from
existing and automated system calculated to choose better system with higher
productivity rate.
The Kohonen-SOM algorithm spent an average computation time with
repetitions as 8,7 seconds and Backpropagation spent as 4,32 seconds. The system
spent 60.90 sec/house that equal with 59 houses/hour. The automated system
presents a new methodology of artificial neural network, for avoiding the high
construction and maintenance costs in the existing meter reading technology.
Result of the automated design shows the data read-out integrated with billing
transaction system has 69,02 seconds at the average process time that equal with
33 houses/hour. From the comparison, it shows that there is a difference as much
as 2 houses between the methods of analysis. The automated system performs
higher productivity.
|
|---|