2024 IEEE Flagship International BIT Conference (BITCON-24)

A Practical Power Theft Detection System with

Simulation and Hardware Implementation
Bindu Vadlamudi Bora Rakesh Nakka Ajay Kumar
Electrical and Electronics Engineering Electrical and Electronics Engineering Electrical and Electronics Engineering
Velagapudi Ramakrishna Siddhartha Velagapudi Ramakrishna Siddhartha Velagapudi Ramakrishna Siddhartha
Engineering College Engineering College Engineering College
Vijayawada, India Vijayawada, India Vijayawada, India
[email protected] [email protected] [email protected]

Marasu Karthik Redrouthu Mohan Sai Shaik Sharuk Babu

Electrical and Electronics Engineering Electrical and Electronics Engineering Electrical and Electronics Engineering
Velagapudi Ramakrishna Siddhartha Velagapudi Ramakrishna Siddhartha Velagapudi Ramakrishna Siddhartha
Engineering College Engineering College Engineering College
Vijayawada, India Vijayawada, India Vijayawada, India
[email protected] [email protected] [email protected]

Abstract— Power theft remains a persistent challenge for created by unauthorized power consumption not only results
utility providers, leading to significant revenue losses and in financial losses but also jeopardizes grid stability. Utilities
inefficiencies across power distribution networks. In this paper, are increasingly looking for innovative methods to detect and
a practical and scalable power theft detection system is prevent electricity theft to reduce these losses and ensure the
proposed, utilizing real-time power comparison at multiple sustainability of power distribution systems.
points along a distribution line, as well as at individual customer
utility points. A Central Monitoring System (CMS) aggregates Electricity theft occurs through various techniques, which
the data from these locations, enabling the detection of are broadly categorized into consumer-level tampering and
unauthorized power usage and theft. distribution line interference. At the consumer level, common
methods include meter tampering, where the meter is
To validate the system's effectiveness, a MATLAB/Simulink modified to display lower readings, and stopping the rotating
simulation model was developed to simulate various power theft disk of mechanical meters using external mechanisms. At the
scenarios and monitor the system’s detection capabilities. distribution level, unauthorized connections are made directly
Additionally, a pilot hardware implementation was conducted to power lines, often referred to as direct tapping or hooking
using microcontroller-based sensors to test part of the system in systems, bypassing the meter entirely [3]. Another notable
a real-world environment. The combined results from the
form of tampering involves the physical destruction of meters
simulation and hardware tests demonstrate the system’s
to stop them from recording usage. These forms of electricity
accuracy and adaptability, providing a cost-effective and real-
time solution for detecting power theft across localized grids.
theft are prevalent in areas with limited enforcement, and
This dual approach of simulation and hardware validation detecting such activities requires robust monitoring and
highlights the system’s potential for widespread deployment in detection mechanisms across the power distribution system.
both small-scale grids and larger distributed energy networks. Various approaches have been proposed to address the
problem of electricity theft, ranging from manual inspection
Keywords— power theft detection, CMS, MATLAB/Simulink,
to more sophisticated technological solutions. Among the
microcontroller
technological solutions, IoT-based systems have gained
I. INTRODUCTION attention, enabling real-time monitoring and alerting utility
companies when discrepancies are detected. These systems
With the increasing demand for energy worldwide, rely on wireless communication networks to transmit data
electricity is a vital component for modern economies and from smart meters to a central system, providing real-time
societies. According to the International Energy Agency updates on power usage and highlighting anomalies that may
(IEA), global electricity demand is expected to increase by indicate theft [4]. While these systems offer notable
more than 3% annually, driven by the ongoing electrification improvements in theft detection, they are often expensive to
of various industries, the rise of electric vehicles, and the rapid deploy on a large scale, particularly in rural or underdeveloped
digitalization of economies [1]. As countries strive to meet areas where infrastructure is limited.
these growing energy demands, ensuring a reliable, and secure
supply of electricity becomes more critical. However, the One of the major challenges in electricity theft detection is
persistent issue of electricity theft undermines the stability and the variation in types of theft and the methods employed by
financial sustainability of power grids, resulting in significant offenders. Many IoT-based solutions focus on monitoring
revenue losses for utilities and globally. individual meters or consumer devices, but they often fail to
detect theft along the distribution line, where unauthorized
Electricity theft, a widespread issue in both developing tapping can occur. Traditional machine-learning approaches
and developed countries, causes substantial damage to that analyze consumption patterns have also shown promise in
national economies. A study on the global impact of electricity identifying theft, but they require large datasets and extensive
theft estimates that power theft accounts for over $96 billion computational resources, which may not be feasible in all
in losses annually [2]. The economic and operational strain settings.

979-8-3315-1839-4/24/$31.00 ©2024 IEEE

 In response to the limitations of existing solutions, this deploying such a system in underdeveloped or rural areas
paper presents a practical, scalable power theft detection remains a barrier to widespread adoption.
system that combines real-time power comparison and data
monitoring at customer utility points with a Central B. Machine Learning Approaches to Theft Detection
Monitoring System (CMS). The CMS, as illustrated in Fig. 1, Machine learning techniques have also been applied to
can serve as the central hub for multiple critical functions, detect electricity theft by analyzing consumption patterns over
including Power Theft Detection, Fault Detection, Real-time time. Reference [7] proposed a hybrid model combining deep
Monitoring, and Monthly Electricity Billing. The proposed learning and random forest algorithms to identify suspicious
system continuously compares power readings at multiple usage patterns from smart meter data. This model showed high
points along the distribution line and monitors individual accuracy in detecting theft but requires extensive
customer usage data. Any discrepancies in power computational resources and large datasets, making it
consumption are flagged as potential theft events. impractical for smaller grids or low-resource environments
where advanced infrastructure is unavailable. Moreover, the
system primarily relies on post-event data analysis rather than
real-time detection, limiting its immediate applicability.
C. GSM-Based Power Theft Detection
This approach of a power theft detection system uses
sensors placed at both the distributor box and individual
consumer meters, leveraging GSM/GPRS communication to
transmit data to a central server. This approach allowed for the
detection of theft through direct current comparison between
the distributor and consumer end, with alerts generated upon
the detection of discrepancies. However, the system's
dependence on GSM communication introduces limitations in
areas with poor signal coverage, and the use of GSM/GPRS
requires a SIM from a network provider, leading to additional
monthly or yearly maintenance costs for the SIM card. [8].
Fig. 1. Key Functions of Central Monitoring System D. Simulation Models for Power Theft Detection
Several studies have also explored the use of simulation-
The system is validated through MATLAB/Simulink
based models to identify discrepancies in power consumption.
simulation to model power flow and detect abnormal usage. A
For example, a MATLAB/Simulink-based simulation was
pilot hardware implementation using microcontroller-based
developed to detect theft by comparing expected and actual
sensors was also deployed to test real-world feasibility. This
power consumption across the grid [9]. While this simulation
integrated approach offers a cost-effective, real-time solution
offers an effective way to detect theft theoretically, it lacks
that can adapt to small-scale grids and scale for larger
hardware validation, limiting its effectiveness in real-world
networks. Bluetooth communication between sensors reduces
applications where noise, interference, and hardware-specific
latency and enhances detection speed. With the CMS at its
issues need to be accounted for.
core, the system enables utilities to detect and respond to
unauthorized power usage, making it a versatile and scalable Many existing power theft detection systems, such as IoT-
option for grid security. based and machine-learning approaches, struggle with key
challenges that hinder their broader implementation. Although
II. RELATED WORK these methods can be effective in certain contexts, they often
A. IoT-Based Solutions for Power Theft Detection fall short in practical deployments, particularly when it comes
to smaller, localized grids or scenarios requiring real-time
One of the more widely explored approaches is the use of detection and cost-effective solutions. The proposed system
IoT-based systems for monitoring power consumption and addresses these issues by employing real-time power
detecting theft. A power theft identification system using IoT comparison across multiple points on the distribution line
was developed to monitor voltage and current levels at the along with monitoring at individual customer utility points,
consumer’s end, with anomalies detected through preset offering a more comprehensive theft detection mechanism
thresholds. This system leverages Raspberry Pi for data compared to traditional methods focused solely on consumer-
processing and uses wireless communication to report theft level data. The inclusion of a Central Monitoring System
[5]. However, while the system offers a medium-cost solution, (CMS) further enhances the system by enabling the
it is limited by its reliance on static thresholds and does not aggregation of real-time data from multiple locations, thus
adapt to dynamic load conditions. Additionally, the system improving both scalability and detection speed.
focuses solely on consumer-level theft, leaving distribution
line theft undetected. Moreover, Bluetooth-based microcontroller sensors are
used as a cost-effective, low-maintenance, and decentralized
In another approach, researchers implemented a real-time alternative to heavier infrastructure solutions such as GSM.
theft monitoring system using wireless sensors placed along This makes the system particularly suitable for localized grids
the grid [6]. These sensors transmit data to a centralized with limited resources. By integrating both
monitoring system that compares consumption at multiple MATLAB/Simulink simulation and hardware validation, the
points. Although this system provides real-time updates, its system ensures theoretical soundness and practical viability,
reliance on centralized data processing introduces offering a solution that is adaptable to real-world scenarios
communication delays and potential system vulnerabilities, while maintaining cost efficiency.
especially when deployed on a larger scale. The cost of
 III. SYSTEM ARCHITECTURE B. Centralized Monitoring and Control System
The proposed electricity theft detection system is centered The system architecture depicted in Fig. 3, is built around
around a Central Monitoring System (CMS), integrating the CMS, which connects various distribution points and end
power sensors, data processing algorithms, and consumers, ensuring efficient electricity distribution and
communication protocols for real-time monitoring and theft monitoring. The architecture integrates both traditional and
detection. Designed to be scalable and adaptable to various renewable energy sources to provide a comprehensive
power sources and consumer setups, the system continuously solution for diverse energy networks:
collects and analyzes data across the distribution network to
• Central Monitoring System (CMS): The CMS collects
detect discrepancies efficiently.
data from multiple nodes across the grid. It processes
A. Process Flow for Electricity Theft Detection this data using built-in algorithms to detect anomalies,
The flowchart shown in Fig. 2, represents the overall such as power theft or grid imbalances. The CMS is
process of detecting theft within the power grid. The system also responsible for sending alerts and controlling grid
initiates by measuring the power consumption at two critical operations to prevent unauthorized power usage.
points: line points (at distribution transformers or feeders) and • End Consumers: The architecture caters to a variety of
consumer points (end-user meters). The sensors at these points electricity consumers, from smart homes and
transmit real-time data to the CMS. commercial buildings to factories and smart transport
At the heart of the CMS is an algorithm that compares the systems. These consumers are increasingly
power readings from the line and consumer points. Any interconnected with smart technologies, allowing for
significant discrepancy between the two, which suggests more precise energy consumption monitoring and
power loss beyond acceptable limits, triggers the theft control.
detection mechanism. If no discrepancy is found, the process By uniting these different entities through the CMS, the
completes without any alerts. However, if the system detects proposed system offers an adaptable solution for modern
a potential theft, an alert is generated, prompting immediate energy grids. It not only focuses on theft detection but also
action from grid operators or utility providers. This process promotes efficient energy usage and grid optimization.
ensures continuous and automated monitoring, minimizing
the need for manual checks while enhancing the system's C. Communication and Scalability
ability to detect theft in real time. By leveraging a decision- One of the strengths of this architecture is its reliance on
based flow as shown in Fig. 2, the system ensures prompt modern communication technologies, such as Bluetooth
responses to unauthorized power usage while maintaining a between sensors and the CMS. This reduces latency and
clear and streamlined approach to managing grid integrity. enhances the system's real-time detection capabilities. All
Bluetooth modules are interconnected, forming a continuous
linked communication network, which ensures seamless data
transmission across the system. Moreover, the architecture is
scalable to larger grid systems, enabling its application in
both small-scale and large-scale distributed networks.
IV. SIMULATION MODEL AND ANALYSIS
The effectiveness of the proposed system for electricity
theft detection was validated through a MATLAB/Simulink
simulation. This simulation was designed to mimic the
operation of a typical small-scale electricity distribution
network, ensuring realistic power flow and consumption
monitoring. Through the simulation, the theft detection
algorithm was tested in a controlled environment where both
normal and abnormal consumption patterns were introduced
to evaluate the system's responsiveness.
A. Simulation Setup
The simulation replicates the operation of a small-scale
electricity distribution network. It measures the power
consumption at the line points (transformers or feeders) and
the consumer points. The CMS receives data from these
measurement points and processes it using an algorithm to
detect discrepancies that could indicate power theft.
Fig. 4, represents the simulated grid divided into three
distinct distribution areas, each representing a section of the
overall power network. Each distribution area contained
several consumer points, representing the end-user meters that
consume electricity. For the purposes of this study:
Fig. 2. Theft Detection Process Flowchart
 Fig. 3. Central Monitoring System Architecture

1) Distribution Area 1 consisted of 3 consumer points, entering and leaving the area. By comparing the total power
each with varying power demands ranging from supplied with the aggregate consumption at the consumer
residential to small commercial usage. points, the system could detect discrepancies indicating theft
or abnormal behavior.
2) Distribution Area 2 included 3 consumer points, with
a mix of residential, industrial, and commercial loads B. Detection of Power Anomalies in Distribution Area 3
to simulate different types of consumers within a During the simulation as depicted in Fig. 5, Distribution
local grid. Area 3 experienced a deliberate power theft event. The load of
3) Distribution Area 3 had 4 consumer points, where the theft event in this area was set to a 20% deviation from the
power theft was simulated by introducing an normal consumption level at the specific consumer point. As
unauthorized load at one of the consumer points illustrated in Fig. 6, the power anomaly detection graph for
Distribution Area 3 highlights the deviation at the theft event.
Each consumer point in the distribution areas had a The power difference between the line point and consumer
predefined load profile, transmitting power consumption data points, which typically remains stable, displayed a sudden dip
to the Central Monitoring System (CMS). The CMS analyzed at the 50-second mark. This discrepancy triggered an alert in
the power flow from distribution transformers to individual the CMS, indicating potential theft.
meters. Line points were also monitored to track the power

Fig. 4. Simulation of the Theft Detection System

 Fig. 5. Distribution Area 3 with Theft load

This real-time detection of power anomalies is a crucial

feature of the proposed system, ensuring that utility operators
can respond quickly to unauthorized consumption. The
system’s sensitivity to even small deviations in power flow
demonstrates its effectiveness in identifying theft across
different distribution areas and load types.

Fig. 7. CMS Alert States During Theft Detection

Fig. 8, shows the power consumption patterns for each

distribution area. Before the introduction of theft, power
consumption was stable. However, after introducing theft
loads at the specified times, the consumption sharply
increased, which allowed the CMS to detect the anomalies.
Fig. 6. Power Anomaly Detection From Fig. 8, we can observe the moment each theft occurs,
with Distribution Area 1 experiencing a power spike at 40
C. System Theft Response Evaluation seconds, Distribution Area 2 at 60 seconds, and Distribution
The proposed system was tested through a simulation that Area 3 at 50 seconds. These sudden increases in power are
modeled three distribution areas with varying loads and theft directly tied to the theft loads introduced during the
events introduced at specific times. Table I highlights the simulation.
current load values for each distribution area, including both
the baseline and the introduced theft loads.
TABLE I. Load and Theft Events

Time of Total
Normal Theft
Distribution Theft Load with
Load Load
Area Introduction Theft
(kW) (kW)
(s) (kW)
DA_1 4.3 2 40 6.3
DA_2 5.4 2.5 60 7.9
DA_3 6.2 3 50 9.2

The loads were continuously monitored, and thefts were

detected and recorded by the Central Monitoring System
(CMS) at different time intervals. As illustrated in Fig. 7, Fig. 8. Power Consumption in All Distribution Areas with Theft Loads
which plots the alert state of the CMS, we observe how the
system responds to theft events across the three distribution V. HARDWARE IMPLEMENTATION
areas. When an alert is triggered, the corresponding CMS alert
The hardware implementation was developed to verify the
state changes from 0 to 1, indicating that theft has been
feasibility of detecting electricity theft using embedded
detected.
systems. The setup consists of microcontroller-based sensors
to monitor power flow at both distribution and consumer sensors and Bluetooth communication, the system monitors
points. Current and voltage sensors were used to measure power flow discrepancies between distribution lines and
electrical parameters, and these were connected to a central consumer points, enabling immediate anomaly detection.
unit via Bluetooth modules. Each consumer point has sensors MATLAB/Simulink simulations validated the approach,
that record power usage in real-time. These sensors send data while the hardware prototype demonstrated its scalability and
to a microcontroller (such as the ATmega328P), which then suitability for real-world deployment. The system's low cost,
transmits the information wirelessly to a Master Unit. The energy efficiency, and rapid detection capabilities make it a
Master Unit compares the consumer power usage with the valuable tool for improving grid security, reducing losses, and
total power supplied to detect any discrepancies, indicating enhancing the reliability of energy distribution networks.
potential theft.
REFERENCES
Fig. 9, shows the hardware setup, where sensors monitor [1] Eren Çam, Zoe Hungerford, Niklas Schoch, Francys Pinto Miranda,
power flow between the distribution transformer and and Carlos David Yáñez de León, “Electricity 2024 Analysis and
consumer points. In Fig. 10, the system is shown with loads Forecast to 2026,” INTERNATIONAL ENERGY AGENCY, Jan.
turned on, demonstrating how power consumption is 2024. Available: https://www.iea.org/reports/electricity-2024
recorded. [2] V. E. Arkorful, “Unravelling electricity theft whistleblowing
antecedents using the theory of planned behavior and norm activation
model,” Energy Policy, vol. 160, no. 0301-4215, p. 112680, Jan. 2022,
doi: https://doi.org/10.1016/j.enpol.2021.112680.
[3] M. B. Shahid, M. O. Shahid, H. Tariq and S. Saleem, "Design and
Development of An Efficient Power Theft Detection And Prevention
System through Consumer Load Profiling," 2019 International
Conference on Electrical, Communication, and Computer Engineering
(ICECCE), Swat, Pakistan, 2019, pp. 1-6, doi:
10.1109/ICECCE47252.2019.8940644.
[4] Potego Maboe Kgaphola, Senyeki Milton Marebane, and Robert Toyo
Hans, “Electricity Theft Detection and Prevention Using Technology-
Based Models: A Systematic Literature Review,” Electricity, vol. 5,
no. 2, pp. 334–350, Jun. 2024, doi:
https://doi.org/10.3390/electricity5020017.
[5] P. Leninpugalhanthi, R. Janani, S. Nidheesh, R. V. Mamtha, I.
Keerthana and R. S. Kumar, "Power Theft Identification System Using
Fig. 9. Pilot Hardware Setup Iot," 2019 5th International Conference on Advanced Computing &
Communication Systems (ICACCS), Coimbatore, India, 2019, pp.
825-830, doi: 10.1109/ICACCS.2019.8728361.
[6] Celimpilo Lindani Zulu and O. Dzobo, “Real-time power theft
monitoring and detection system with double connected data capture
system,” Electrical Engineering, vol. 105, no. 5, pp. 3065–3083, May
2023, doi: https://doi.org/10.1007/s00202-023-01825-3.
[7] S. Li, Y. Han, X. Yao, S. Yingchen, J. Wang, and Q. Zhao, “Electricity
Theft Detection in Power Grids with Deep Learning and Random
Forests,” Journal of Electrical and Computer Engineering, vol. 2019,
pp. 1–12, Oct. 2019, doi: https://doi.org/10.1155/2019/4136874.
[8] N. K. Mucheli et al., “Smart Power Theft Detection System,” 2019
Devices for Integrated Circuit (DevIC), pp. 302–305, Mar. 2019, doi:
https://doi.org/10.1109/devic.2019.8783395.
[9] S. Shah, T. Singh, Y. Rai, V. Pandey, and L. Gahalod, “Power Theft
Detection Using IoT,” Journal of Emerging Technologies and
Innovative Research (JETIR), vol. 11, no. 5, pp. k170–k173, May 2024.
[10] Dr.S.Thangalakshmi, G.Sangeetha bharath, and S.Muthu, “POWER
THEFT PREVENTION IN DISTRIBUTION SYSTEM USING
SMART DEVICES,” International Journal of Applied Engineering
Research, ISSN 0973-4562, vol. 10, no. 42, pp. 30841–30845, 2015.
Fig. 10. Hardware Setup with Loads Turned On [11] L. Zhou, "Power System Monitoring and Management Based on Smart
Sensors," 2023 3rd International Conference on Intelligent Power and
The system uses low-latency Bluetooth modules (e.g., Systems (ICIPS), Shenzhen, China, 2023, pp. 122-127, doi:
HC-05) for short-range communication, ensuring efficient 10.1109/ICIPS59254.2023.10404486.
data transmission. Standard household lines powered the [12] Y. A. Badamasi, "The working principle of an Arduino," 2014 11th
setup, and sensors were calibrated for accurate measurements. International Conference on Electronics, Computer and Computation
(ICECCO), Abuja, Nigeria, 2014, pp. 1-4, doi:
The hardware demonstrated effective theft detection by 10.1109/ICECCO.2014.6997578.
monitoring power usage in real-time and can be scaled for [13] K. V. S. S. S. S. Sairam, N. Gunasekaran and S. R. Redd, "Bluetooth
larger power grids. The system's microcontroller design is in wireless communication," in IEEE Communications Magazine, vol.
easily integrated into existing grids, and its low power 40, no. 6, pp. 90-96, June 2002, doi: 10.1109/MCOM.2002.1007414.
consumption ensures long-term operation with minimal [14] Ankit Negi, “HC-05 pinout, specifications, datasheet and HC05
maintenance. Arduino connection,” www.etechnophiles.com, Oct. 06, 2020.
https://www.etechnophiles.com/hc-05-pinout-specifications-
VI. CONCLUSION datasheet/
[15] RiftligerMy, “Arduino Bluetooth Master, and Slave Using Any HC-05
This paper presents a real-time system for detecting Modules,” Instructables. https://www.instructables.com/Arduino-
electricity theft, combining simulations and practical Bluetooth-Master-and-Slave-Using-Any-HC-05/
hardware implementation. Using microcontroller-based