Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350878724

Electric Energy Meter System Integrated with Machine Learning and


Conducted by Artificial Intelligence of Things - AioT

Conference Paper · April 2021


DOI: 10.1109/ElConRus51938.2021.9396280

CITATION READS

1 2,926

3 authors:

Nimai Chandra Das Md Ziaul Haque Zim


Daffodil International University Petersburg State Electrotechnical University
6 PUBLICATIONS 1 CITATION 7 PUBLICATIONS 3 CITATIONS

SEE PROFILE SEE PROFILE

Md sazzad Sarker
Daffodil International University
3 PUBLICATIONS 1 CITATION

SEE PROFILE

All content following this page was uploaded by Md Ziaul Haque Zim on 15 April 2021.

The user has requested enhancement of the downloaded file.


Electric Energy Meter System Integrated with
Machine Learning and Conducted by Artificial
2021 IEEE Conference of Russian Young Researchers in Electrical and Electronic Engineering (ElConRus) | 978-1-6654-0476-1/20/$31.00 ©2021 IEEE | DOI: 10.1109/ElConRus51938.2021.9396280

Intelligence of Things – AioT


Nimai Chandra Das Md Ziaul Haque Zim Md Sazzad Sarkar
1 Saint
Department of Automatic Control Systems, Petersburg Electrotechnical Department of Electrical and Electronic
1 Saint Petersburg Electrotechnical University “LETI” Engineering
University “LETI” Saint Petersburg, Russia Daffodil International University
2 Department of Computer Science and
Saint Petersburg, Russia Dhaka, Bangladesh
2 Department of Electrical and Electronic Engineering,
Engineering, Daffodil International University
Daffodil International University Dhaka, Bangladesh
Dhaka, Bangladesh ziaul15-1133@diu.edu.bd
nimaidey67@gmail.com

Abstract—the field of electric power distribution is still in its Power outages or load shedding is a common problem for
infancy, and state of the art solutions from modern technology is many countries. It will be reduced to a certain amount by
not easily adapted. IoT and AI could also be bringing a wind of properly observance of power consumption and avoiding
change, but hitherto in large amounts of electrical power users wastage of electric energy. The most important point at issue
and distributors hooked in to humans is to see the vitality meter that continues inspection or observance of energy cannot be
and abandoning the bills to the claimant of that home monthly. done expeditiously as a result of the user can't conscious of
Most defects of this infrastructure are that human reliance to their power consumption. The power distribution authority
filter the meter of every house and abandoning the bills. So ever, issued bills with area units and only they'll get a plan
further bill amounts or notice from power distributors after
concerning their consumption. These bills are distributed solely
paying bills are basic blunders. To defeat this kind of inaccuracy
this paper considers the creation of a smart energy meter device.
once in an exceedingly month and clients always stayed in the
Considering power consumption, cost efficiency, speed, and dark concerning their energy usage throughout this era of your
importance on the performance we used ESP8266 based time. This method has got to be continuously varied times to
nodeMCU and AVR microcontroller based Arduino development expeditiously management energy usage in an exceedingly
board. The task of nodeMCU gives notification and shows month. Sometimes a customer can’t understand why they
current pursuing with costs through the online page. In this received a big amount of bills and they refuse to pay it out.
exploration, we complete the accuracy of meter perusing with the Therefore power distribution authority cuts the power supply
help of an LCD that shows which prone to KWh, voltage, current due to payments and that particular customer will suffer many
and power factor perusing. The main convenience of this system months and sometimes it takes a long time to resolve the issue.
doesn't lose any data caused by power interrupt, where others
AVR based system fails to save the delivery data before turning Power related problems are one of the major concerns all
off the power. over the world. It is also recognized as a crucial issue because
of the industry and technological dependency. During this
Keywords— Artificial Intelligence of Things (AIoT); ESP8266; world, there are differing types of power supply, distribution,
nodeMCU; AVR microcontroller; Arduino and energy consumption related problems. In this research, we
work on “Smart Energy Meter Conducting with Artificial
I. INTRODUCTION Intelligent (AI) and Internet of Things (IoT)” to real-time
Electric energy is currently the backbone of modern and observance on energy utilization in the portable or versatile
technological society. And electric power distribution is largely computer as opposed to checking energy meter. Analyzing the
pragmatic, but evaluating with the modern tendencies regular energy usage data with an AI program user will notify
nowadays the power distribution infrastructure is still ancient. maximum and regular energy uses. Customers can easily
The major issue of recent infrastructure is maintenance, monitor their daily and monthly energy uses by using portable
protection, and inspection. Although every power distribution internet and Wi-Fi active devices like android smartphones,
authority has its own specialized team for maintenance, computers and they can also easily understand their uses with
protection, and inspection. But the problem is they are not well the help of graphs and charts.
equipped with modern technologies. So if infrastructure caused In addition to this, we see that now-a-days technical world
damage from sudden incidents like a natural disaster or any such a lot focused on Artificial Intelligence (AI) and the
synthetic accident, it’s very hard to point out the problem and Internet of Things (IoT) based systems. Users expect
give a particular solution. Where we can easily point out this everything that automated and advanced are going to be
kind of problem with the help of Artificial Intelligent (AI) and recommended to them by the system. To form this type of
the Internet of Things (IoT). system to be recommendation capable must have the power to

826 978-1-6654-0476-1/21/$31.00 ©2021 IEEE

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
form decisions by itself. To require decisions by itself must got III. LITERATURE SURVEY
to have classified data that system does it already by gathering In past studies, thousands of researchers from all over the
a day and store it within the memory of the system. world are working on the different types of energy meter
These above reasons make us inspired to try and systems and trying to improve technical methods and their
experiment this type of research-based tasks. Our task is firmly accuracy. In recent decades researchers are thinking about new
associated with the IoT, has some AI programs inside IoT and newer approaches and technologies for getting better and
devices and Electrical procedures too. automated energy consumption output from users. But
effective works and resources for Smart Energy Meter are not
II. BACKGROUND STUDY going well parallel to modern time.
The previous time this related developments that were In 2003, B. S. Koay, S. S. Cheah, Y. H. Sng, P. H. J.
already done by many researchers during this field. Except, Chong, P. Shum, Y. C. Tong, X.Y. Wang, Y.X. Zuo and H. W.
giving a crystal clear explanation of this, this chapter will show Kuek did a “Design and Implementation of Bluetooth Energy
what the restrictions, lackings, and ineffectiveness of those Meter” [6] where they present and implement a Bluetooth-
works were and lastly, this section also recites the scope of this Enabled Energy Meter that is able to show energy consumption
research also because of the challenges of it. data and give readings wirelessly based on Bluetooth.
It's sad that only a few developmental works during this In November 2006, Marek Huczala, Tomas Lukl, and Jiri
field are accomplished by this point though within the current Misurec talked about the automatic energy measurement
period of time, working during this field is increasing day by system and captured energy meter data over a secured power
day. There are a lot of scopes and resources worldwide having line [7]. They work on Broadband Power Line (BPL), Power
tons of power distribution companies. Line Communication (PLC) and Home-Plug that gives its own
Some related developmental tasks associated with our or upper-level communication TCP/IP. They also implement
research are given below with a brief description. the AES cryptographic algorithm in hardware.

In the paper, “Smart Energy Meter”[3], introduce B. Sivaneasan, E. Gunawan, and P. L. So, they present
automated system that provides remote access to utility automatic meter reading (AMR) systems using power-line-
companies for fault diagnosis and communicating with clients communication (PLC)-access networks and also present the
and clients can also recharge and control loads remotely. The analysis of two different topological structures and the
highlights they considered in their proposed energy meter performances of these two schemes analytical models they
system (EMS) transmit data like energy consumed in kWh and studied in July 2010[8].
generates a bill over a GSM mobile network and advantages In the year of 2011, Carmine Landi, Pietro Merola, and
include that the system provides domestic power consumption Giacomo Ianniello proposed an ARM microcontroller based
accuracy, safely and with a relatively fast update rate. power management system integrated with a Web Server that
In this paper, we found that using GSM technology is not allows collecting the statistical data of power consumption, to
cost-efficient because GSM technology requires a SIM card the reduction of energy consumption, costs and power quality it
that has network subscription charges as well as for reading is able to interface devices for load-displacements [2].
meters. This system sends SMS and it's not that much of an Asma Garrab, Adel Bouallegue, and Faten Ben Abdallah
excellent idea because SMS have charges. And also like the proposed an automatic meter reading (AMR) system for a
situation in a natural disaster or network failure this project smart grid with power efficient microcontroller
doesn’t work because of using GSM technology. MSP430FE423A that includes metering module ESP430CEl
A paper from “M. M. Mohamed Mufassirin” shows IoT has and the Power Line Communication standards in March
a better arrangement exactness [4]. A distinctive study of using 2012[9].
the Internet of Things (IoT) for measuring electricity In the year of 2014 and 2015, papers from M V Aleyas and
consumption and generates bills in an automated way by the Dr O. Chandra Sekhar respectively talked about PLC
help of IoT and telemetric communication techniques are Communication [10] and also automatic meter reading for
talked about. Various types of unique technical methods that smart metering by using QPSK modem with PLC Channel and
have been utilized are: Detecting and controlling energy theft, GSM Modem [11] in their own paper.
disconnect and reconnect the electrical power supply, and
capability to continuously connect with providers and S.V.Anushree, T. Shanthi asserts a framework for IoT
consumers. Infrared sensors have been used in this system to based intelligent energy meter surveillance and theft detection
detect any illegal alteration happens in the metering system. systems [1] in November 2016.
An assessment report has been presented as titled “Impact The work by Dakka. Obulesu, K. S. V. Phani Kumar and V.
Assessment of Smart Meter Grouping on the Accuracy of Mahesh deal with an energy-monitoring system aspect of IoT,
Forecasting Algorithms” where concentrate to uncover the to overcome human errors and reducing dependency on manual
importance of characteristic choice in Smart Energy Meter is labor, besides reducing costs in energy consumption at
depicted.[5]. In this, they talked about the prediction of future consumer end [12].
energy consumption demand, bills forecasting errors,
Advanced Metering Infrastructure (AMI), Distribution System
Operator (DSO) with its grid balancing efforts.

827

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
IV. RESEARCH METHODOLOGY
This chapter deals with the theoretical part of our research
task. This section will provide a crystal clear grasp of the
concept of the task. Making it more clear, the very first, we try
to explain the Subject of Research and Instrumentation very
shortly. Then we’re going to discuss the implementation of the
Internet of Things (IoT) and Artificial Intelligent (AI) process
in the system.
Fig.2. ACS712 IC & MCU schematic diagram
A. Subject of Research and Instrumentation
We utilize a step-down transformer that takes AC 220volts
We want to explain by subject of research is Internet of as input and gives DC 9v as output, also additionally associated
Things (IoT), AVR microcontroller and Artificial Intelligent with the two diode circuits that fill in as a full-wave rectifier
(AI) that is being researched and studied for clear circuit and it’s currently separated by the capacitor. Capacitor
understandings and being implemented as an Energy Meter changes over the throbbing DC into smooth DC with the
Systems. The subject of research is deeply responsible for the assistance of charging and releasing impact. The capacitor is
analysis, giving the right knowledge of different kinds or sorts presently controlled by the IC 7805 controller. IC 7805 gives a
of research parameters, and also for a clear understanding. As 5-volt guideline to the circuit and gives managed 5-volt control
opposed to, Instrumentation refers to the required tools or kit supply. And the controller is presently again sifted by the
like ESP8266 based nodeMCU, AVR microcontroller-based capacitor. In the yield of the capacitor, we utilize one resistor
Arduino development board and LCD, etc that are being used and one drove in arrangement to give a visual sign to the
by the researchers for implementation. circuit.
B. Flowchart of our proposed system ESP8266 nodeMCU is working as the heart of this project
and complete its predefined task which is already defined in the
code. According to code, the sensor recognizes the data of
voltage, current, watts, and amp, balance and sends it to the
web server we define and continue working without any
interruption.
D. Schematic Diagram

Fig.1. Flowchart of our system

C. Implementation Procedure
The proposed system consists of four major modules
namely, ESP8266 based nodeMCU, Arduino AVR
microcontroller based development board, ACS712 IC and
LCD. Where ESP8266 based nodeMCU is mainly the
controlling of the system and Arduino Uno (AVR Fig.3. Schematic Diagram of our system
microcontroller based development board) controller of the
system. In this system, we have two types of current supply E. Data Collection Procedure
alternating current (AC) and direct current (DC). DC used for Machine learning a subfield of Artificial Intelligence (AI)
operating and controlling the internal components of energy where data is the fast and leading thing. Actually, this data is
meters and alternating current (AC) used for loads in this taken into most consideration, because data is the heart of the
system [3]. ML process. In this section, there was no alternative to data and
it became the most challenging task of our research.
The current sensor is hall effect based linear ACS712 IC
utilizes the live current abuse and this module will understand TABLE I. DATA STORED IN TWO DIMENSIONAL ARRAY
live current DC or AC current beginning from -5A to +5A, -
20A to +20A and -30A to +30A. It has 8 pins. The VCC is Date Cost (tk) Watts
associated with +5V to control the module and in this way, the 04 January, 2020 0.75 78
GND is associated with the base of the microcontroller. And 08 January, 2020 0.87 84
the VIOUT is connected with the microcontroller Analog 12 January, 2020 1.95 1.92
Input/Output pin. 16 January, 2020 0.50 0.55
20 January, 2020 1.12 1.15
24 January, 2020 0.28 0.26
31 January, 2020 0.18 0.15

828

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
As we said before, our proposed system doesn't lose any In Figure 6, demonstrates the cost of power consumption on
data caused by power interruptions, where other AVR based a daily basis where dates are successively localized from up to
systems fail to save the delivery data before turning off the bottom on the left side of the chart. It is seen the cost of power
power. Because we store real time data in a web server and consumption rapidly increased due to demand.
save it in a two dimensional array. This data is highly pre-
processed. We know that the accuracy of outcome depends on
how efficiently data are going to be pre-processed.
We know that computers only work with numerical value,
so, we already convert all features categorical to numbers by
the label encoding [13]. As an example, here we label the
“Dates” as 1, 2, 3, 30 and 31 respectively.
F. Bar Chart and Graph of Data
Here, bar charts illustrate the cost of power consumption on
a weekly basis. It is generally obvious that the cost is in its
maximum around 3.5$ due to maximum demand and the cost is
in its minimum less than 0.2 $ due to minimum demand.
Fig.7. Watt vs Time (Daily)

In Figure 7, this bar chart represents the use of electricity


on a daily basis where dates are successively localized from up
to bottom on the left side of the chart and on the horizontal axis
at the bottom of the chart the magnitude of watts are shown.
G. Implementation of Machine Learning Algorithm
We optimize a simple “Decision Tree” algorithm for
comparing the uses of electricity and its help users for reducing
their energy consumption.
Fig.4. Cost vs Time (Weekly) Working procedure of this algorithm is very simple and in
general this algorithm is widely used for decision analysis [14].
In Figure 5, this bar chart represents the use of electricity in
This algorithm can be linearized by depending on decision
weekly basis. It is noticeable that the maximum demand was
rule[15], where its result is the contents of the leaf node, and
around 4.5 watt last week and the minimum demand was less
therefore after a significant determination along the path form a
than 0.2 watt on the first week.
conjunction within the clause if . An example of rule form,
If con1 and con2 and con3 then generate results.
Here, if we consider ‘con1’ variable is mean value of
previous day energy usage and ‘con2’ is mean value of today’s
usage then we can build condition like
If con2 greater than con1 then generate notification.
Before users forget about reduction of energy consumption,
this “Decision Tree” predicts the uses of energy and notify the
user. And this is how our system enables Artificial Intelligence
to reduce future energy consumption.
Fig.5. Figure 5: Watts’s vs Time (Weekly)
V. ANALYSIS AND SIMULATION
In this chapter we are going to discuss some sorts of our
developed systems. We also decide to talk about some
components, show our simulation process and how it’s
implemented in our systems and how the sensor data, power
converter data can be taken for our system. It helps us in our
final development of the system.
A. Current Sensor
A linear Hall sensor integrated circuit is ACS712 IC, which
is a copper conduction path located near the surface of the die
and this IC is able to measure voltage. The integrated hall IC
has capabilities of sensing magnetic fields that are generated if
Fig.6. Figure 6: Cost vs Time (Daily) current is applied flowing through this copper conduction path

829

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
and after that it also converted into a proportional voltage. We B. Output Waveform of Buck-Converter
decide to choose the incentive through oscilloscope by entering
220 voltage AC source with some opposition esteem as
indicated by proteus ohm's law its ascertain and show the
present proportion of a sine wave.

Fig.11. Output waveform of buck converter


Fig.8. Simulation of current sensor C. Output Waveform of Relay
We also simulate current data through current sensor by
using Arduino Uno development board and Arduino IDE.

Fig.12. Output waveform of buck converter

VI. EXPERIMENT WITH IMPLEMENTED SYSTEM


In this chapter, we test our Electric Energy Meter System
and before completing the development process of our full
system, we test it by integration testing system. In every test
Fig.9. Output of current sensor in Arduino IDE
section we found some problems and we solved those problems
as soon as possible. Researchers from all over the world are
trying to upgrade this energy meter system.
Below in Figure 13, we connect different types of load with
our system at the same time we connect the same load with an
analog meter for comparison.

Fig.10. Test of current sensor graph

Fig.13. Meter Reading Load

830

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
The whole system is run on board power supply if the Initially it also checks the connectivity with WiFi. If it got
supply is breakdown or increase the system will not work. For connected with the internet then it will check the connection
securing user data, every time the system updates automatically for Web Server. After establishing a connection with a web
and save this data in a Web server. server it will show all its data on the web page.
By using this system the users can check the status of the VIII. FINAL RESULTS
appliances at whatever time of the day. So anyone can access
anytime in this meter. To prevent this problem we use a simple Firstly, we'd switch on the main topic. The ACS712 sensor
encryption and decryption. sense the electric power that is utilized by the load and this
current sensors output is analog form. This sensor output is
A. Result of Experiment connected to the analog input in Arduino Uno board and inbuilt
After the scheme and development of energy meters using ADC converts analog input as digital output and showed in
the IoT system is proposed and when the power is on, the relay LCD.
switches the first relay on and all are rest. Overall system was The ESP8266 based nodeMCU allows connectivity with
designed and tested by us. We are introducing the home the internet to this system. A web page displayed the power
appliances controlling and regarding security issues. Our utilized by the loads and continuously exchange data with a
project has completed a successful implementation & testing. web server through this nodeMCU.
So we said that our system was effective fast response, most
secure and easy to construct. A. Hardware System

VII. WORKING PROCEDURE OF THIS SYSTEM


To work with this system we need to power up this firstly.
Then connect this system with loads. We receive a command
from the user. It has the ability to respond through ESP8266
based nodeMCU.

Fig.15. Developed Hardware Model

B. Cloud and Web Server

Fig.14. Working principle of developed system

Figure 14 shows the flowchart of our system which starts


and establishes its connection with sensors, load, controller and
display. Then the sensor processes its data with a
microcontroller unit. Then the system sends command to
ESP8266 nodeMCU. Fig.16. Developed Web Server

831

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
Server work as remotely observation of our system data explanation of the issues additional bill from power distribution
through the internet. We can see our system reading data stored authority and can deliver the simplest service to the end-user.
in the cloud through this server from anywhere in the world by
using server address which is called as IP address from our REFERENCES
energy meter. [1] S.V.Anushree, T. Shanthi. “IoT Based Smart Energy Meter Monitoring
and Theft Detection Using ATMEGA”, International Journal of
IX. FUTURE WORK Innovative Research in Computer and Communication Engineering.
[Vol. 4, Issue 11, pp. 19801- 19805, November 2016].
Our main target is to create a proficient and upgraded [2] Patrick Mapulane, Tshepang Letshwiti, Mompati Molibe, Oagile
electric energy meter system and the prime future scope of this Gaogane, Keaboka M. Sethebe, N.M.J. Ditshego. “Smart Energy
project is to make a fully automated system with maximum Meter,” BIUST Research and Innovation Symposium 2019 (RDAIS
accuracy and minimal human supervision. In this work, using a 2019) ISSN: 2521-2292 [Page no 30-36]
small machine learning (ML) algorithm enables the future [3] M. M. Mohamed Mufassirin, A. L. Hanees. “DEVELOPMENT OF IOT
scopes of Artificial Intelligence and IoT based energy meters. BASED SMART ENERGY METER READING AND MONITORING
SYSTEM”, 8th International Symposium - 2018, SEUSL
After implementing this, we can add more division or class of
[4] B. Sivaneasan, E. Gunawan, and P. L. So. “Modeling and performance
things regarded as having particular shared characteristics in analysis of automatic meter-reading systems using PLC under impulsive
this task, which can make this more efficient. noise interference,” IEEE TRANSACTIONS ON POWER DELIVERY,
[vol. 25, no. 3, pp. 1465–1475], Jul. 2010.
Features like, when real time energy usage is greater than
[5] Garrab, A.; Bouallegue, A.; Ben Abdallah, “A new AMR approach for
previous days energy meter can generate an automated energy saving in Smart Grids using Smart Meter and partial Power Line
notification and if user doesn’t respond within 5 minutes, the Communication”, IEEE First International Conference on Renewable
system will automatically disconnect the power supply. By Energies and Vehicular Technology (REVET), pp. 263 – 269, March
using machine learning (ML) we can train this system about 2012
when users need energy supply and when time to go on [6] M V Aleyas, Nishin Antony, Sandeep T, Sudheesh Kumar M, Vishnu
sleeping mode, disconnect/continue the power supply. The Balakrishnan. “Automatic Meter Reading and Load Management Using
Power Line Carrier Communication”, International Journal of Advanced
consumption data can also be used, not just for billing but also Research in Electrical, Electronics and Instrumentation Engineering.
by the power distributors for forecasting the consumption and [Vol. 3, Issue 5, pp 9425-9432] May 2014.
plan their resources accordingly using data mining techniques. [7] Dr O. Chandra Sekhar, Mr. B V Rajanna, and Mr. M .Kiran Kumar.
It'll be resulting in good to the forthcoming researchers to do “Automatic Meter Reading For Smart Metering By Using QPSK
such sorts of research on Artificial intelligent electric energy Modem With PLC Channel and GSM Modem.”, Journal of Electrical
meters. Engineering, January 2015
[8] Dakka. Obulesu, K. S. V. Phani Kumar and V. Mahesh. “Design and
X. CONCLUSIONS Implementation of IoT Energy Meter to Monitor Energy Flow at the
Consumer End”, Asian Journal of Electrical Sciences. ISSN: 2249-
An elite system to specialise in real-time observation with 6297, [Vol. 8, No. 2, 2019], pp. 1-4
efficient use of cheapest, smart and artificial intelligent electric [9] “Smarter Ways to Encode Categorical Data for Machine
energy meter system that permits the longer term reduction of Learning”.https://towardsdatascience.com/smarter-ways-to-encode-
electrical power consumption. Through this technique, it'd be categorical-data-for-machine-learning-part-1-of-3-6dca2f71b, (Accessed
determined that there are often a big number of improvements on 08/01/2020)
in society, industry, and agricultural sectors with the [10] Kamiński, B.; Jakubczyk, M.; Szufel, P. . "A framework for sensitivity
analysis of decision trees". Central European Journal of Operations
dissipation of IoT. Thus, the system would be a feasible Research. 26 (1): pp 135–159 (2017).

832

Authorized licensed use limited to: Md Ziaul Haque Zim. Downloaded on April 15,2021 at 09:07:48 UTC from IEEE Xplore. Restrictions apply.
View publication stats

You might also like