You are on page 1of 1

MAP A INSTITUTE OF TECHNOLOGY

School of Electrical, Electronics and Computer Engineering


CHRISTIAN JAY L. BANAAG | KRIZIA C. BERNARDO | CLAUDINE L. GUTIERREZ| OLIVER JAMES V. LANDICHO

Thesis Adviser: Engr. Jesus Martinez and Dr. Felicito Caluyo

Electricity Theft Monitoring System Using Artificial


Neural Network
ABSTRACT
System Losses can be seen in our monthly electric bills, dissected into different factors by the power distributors. These system losses are comprised of
technical, non-technical, and administrative losses. The focus of this study is the non-technical loss, mainly from illegal connections. Electricity Theft is one of the
major contributors as to why both utilities and consumers shoulder extra charges. The problem is, utilities are only able to measure system losses, and have their
consumers pay 8.5% of it, and the rest will be shouldered by them. The purpose of this study was to train a system that will be able to detect whether electricity theft
exists in a certain area. MATLAB Simulation was applied for the Artificial Neural Network (ANN), whose main role was to continuously learn the behavior of the
system and provided the researchers the algorithm needed for the objectives of the study. The result of the study was remarkable, because the success of it proved
that finding a way to solve this problem is not a lost cause. It showed that designing such system can help eradicate electricity pilferage altogether, with the help of
human intervention. The government, utilities, and consumers just have to be more fervent in addressing whats at hand.
INTRODUCTION
According to the site Index Mundi, a data portal that gathers facts and statistics from multiple
sources and turns them into easy use visuals, the estimated annual electric power consumption in the
Philippines is 54.4 billion kWh (2009). This figure consists of the total power generated in the Philippines
plus imports and minus exports. The discrepancy between the amount of electric power consumed and
electric power generated and imported will be accounted as System Loss. System loss can be in the form of:
illegal connections, meter tampering, system design errors, or unpaid billing charges. And in the
Philippines, the bulk of the system loss comes from illegal electrical connections. Since the consumers are
only allowed to pay 8.5% of the total system loss, the rest will be shouldered by the power companies.
Leading not only to an unfair billings for the consumers, but also costs the power companies more. With an
increasing demand in electricity, it is in utmost importance to pay more closely in minimizing system loss
due to theft.
At present, measurement of system losses is one of the ways utilities do to monitor their power
distribution. Designs for electricity theft monitoring system are well supported by theory, but its realization
in the real world has yet to be implemented. In a recent study, the system operates by measuring the meter
data of consumed and delivered power.
Utilities are only able to measure system losses. That is why the researchers will design a system
that will be able to recognize the existence of electricity theft. And since billing irregularities and unpaid
bills are also considered as system losses, modifications will be done so that the system will only focus on
illegal connections.
The main objective of this study is to design a system that will be able to detect whether there is
an electricity theft occurring. Specifically, it aims to (1) design a small scale environment to be used for
testing; (2) train an algorithm using MATLAB that will determine whether there is theft or not; and (3)
validate and display the output at a central system.

RESULTS AND DISCUSSION

Table 1
Table 1 shows a sample scenario, Scenario 5 was when all the loads from Meter 1, Meter 2, Meter 3,
Theft Line and Other Line were turned on. At 8AM, most of the consumers were already awake, starting their
day by cooking food for breakfast, watching television for morning news, etc. It was expected that at this
scenario, power consumption is at its maximum.

METHODOLOGY

Fig. 1
Figure3.1 (b) shows how the process for the detection of electricity theft. First was gathering all the
data measured by the power meters from different scenarios set by the researchers that simulated the events
of theft in the system.

(
(

Table 2

)(

) (
)(

)(

)
)

X2 = 320.0245
The researchers are to reject the null hypothesis. Therefore, the detection of theft by
the neural network is associated by the actual presence of theft in the system.

CONCLUSION

Fig. 2
Legend: MCU Microcontroller Unit (Arduino AtMega2560) ; ML Main Line Meter ; M1
Meter 1 or Power Analyzer 1 ; M2 Meter 2 or Power Analyzer 2 ; M3 Meter 3 or Power Analyzer 3 ;
T Theft Line ; and L Other Line (Unmetered Legal Consumer) Figure 2 (a) shows the Small Scale
Environment that was designed to replicate the actual connections of power lines. All loads (L1, L2, L3, T,
and L) were directly connected to the Main Power Line, whose power came from the Main Power Source.
AC Power Analyzers were used as sensors, connected in Main Line, Meter 1, Meter 2, and Meter 3.

As the world continues to evolve through time, its people also strive to improve its lifestyle and
adapt to the changes that are inevitable. Gone are the days that a household uses bonfire for cooking, torches
for lighting, burning of coals for operating machineries, and the likes. These are the primitive ways of living,
which were buried decades ago as people discover how to utilize electricity in all of the aspects of life.
Unmistakably, electricity has played a great role in the evolution of our civilization. In this study, the
researchers were able to design a small scale environment that was used for testing. It was strategically
designed so that it can represent a main power line and the consumers that are legally connected on it. Also, a
node for theft is installed on the board, which is significant because it shows how illegal connections are
infringed, and how it affects the overall power consumption of the system. Through multiple and continuous
testing, the researchers were able to train an algorithm that can detect the existence of theft, based on the
behavior set by the researchers. This is possible because of an intelligent system which is the Artificial Neural
Network, whose main role is to learn the behavior of the designed small scale environment. Also, the trained
system is capable of displaying an output of Normal Operation or Theft Detected, in a real time basis. As
a conclusion, it is safe to say that it is actually possible to address electricity theft with the help of the
Artificial Neural Network (ANN). The learning process of ANN depends on the behavior of the system, and
should be trained continuously up until a desired output is obtained. The government and power distributors
just need to put more extra effort in implementing solutions to this problem, and ardently monitor the
situation from time to time. The effectiveness of this study is hoped by the researchers and for it to materialize
in the society and help lessen the expenses of both consumers and distributors by eradicating electricity theft.

You might also like