Integration of Machine Learning in IoT

based Home Automation System

Team: weautomate
Bijaya Sapkota
Dipesh Poudel
Sandesh Thakali
Swagat Gautam
Project overview:
Ø With the growing energy demand and the
residental electricity consumption in Nepal
being 43% of total consumption, we aim to
develop cost effective home automation
system having capabilities of internet
connectivity, integrated with machine learning
for situational awareness and autonomous
decision making.
Ø Overall the implementation of home
automation system can significantly enhance
energy efficiency in residental sectors,
contributing to both environmental
sustainability and energy saving.
2
Problems in Traditional systems :

Ø Lack of interconnectivity between devices.

Ø Disabled and old aged persons can’t access switches manually.
Ø Inefficient energy usuage .
Ø Monitoring the real time consumption rate is not possible.
Ø Peak hour Identification and load balancing can’t be done.

3
Developed system was divided into two parts:

1)Voice commanded IOT based home automation system

2)Integration of AI in developed model.

4
 Part A:Voice commanded IoT based
home automation system
ØThe system we have developed is capable
of real time monitoring and control of all
non smart home appliances through a
smartphone via internet.
ØVoice command on google assistant, alexa
can remotely control home appliances from
anywhere.
ØThis system introduces inter device
connectivity and hence is capable of
switching many devices at a time.
5
Why home automation?
• Old aged/disabled persons who cannot
access switches manually can control it
via mobile app or voice commands .
• Data can be monitored in real time via
web/App interface .
• Each device is interconnected with
each other .
• Non smart appliances can be controlled
from anywhere if connected to internet.

6
Hardware Specifications:

Ø ESP-32 wifi module

Ø 4 channel relay module
Ø Bulbs and socket
Ø Switch board

Software Specifications:
Ø Cadio app
Ø PLX-DAQ
Ø Adafruit library
Ø IFTTT
7
Methodology:
Integration of voice commanding
Google assistant/ Alexa

System architecture consists of internet User interfaces Mobile app and

web app
connectivity (ESP-32) which is attached to
the main supply unit of the home system
which can be accessed through the internet .
Web server

This brings control of every switch of your

home within your smartphone. So, the main ESP-32

objective of our system is to deliver low cost

and energy efficient product that optimizes
Relay Module Manual switches

costumer comfort.
Automation of non smart
Home appliances
8
Need for further AI Integration
ØPeak hour identification and adjustment of energy intensive
activities.
ØNatural language processing can make voice commanding even
easier .
ØReal time Energy monitoring and optimization.
ØSituational awareness and autonomous decision making.
ØDynamic pricing adaptation.

9
 Part B:
Integration of AI in
developed model
Methodology:

Ø Data from temperature sensor

(DHT22),motion sensor (PIR) to be
collected in sampling rate of every 15 secs,
allowing us to capture precise and real-time
environmental changes.
Ø The data was saved in Ms Excel file using
PLX-DAQ
Ø After pre processing the data, it is fed into
Temp data acquisition using PLX-DAQ
machine learning algorithm.
Ø We have used the Logistic Regression
under supervised machine learning to
process the data from sensors. 11
 Parameters used in Sigmoid Activation function
Ø Train data to be fed into algorithm was taken from open source database
like kaggle.
Ø The probability that a device has to be turned on or off is,
1
�(ON or OFF) =
1 + �− (b1.temp + b2.occupancy + b3.time + b4.weekend/weekday + const)

Ø ML algorithm decides the state of home

Appliances either to be turned on ON or
OFF (0 or 1) on the basis of result
obtained from activation function.
12
 Problems tackled
i. Electrical devices in “standby mode” add up to 40% of the total energy used.
Ø With a remotely accessible home automation system, devices can be powered off directly instead of
being kept in standby mode.

ii. Brings manual control to zero.

Ø With NLP assisted voice commands ,easy accessibility for old aged and disabled and remote access
and ML capabilities ,this system gets more efficient and accurate over time.

iii. Since AI integrated home automation system observes and predicts temporal and occupancy
variation adjustments in usage patterns, they can dynamically optimize energy consumption.

Winter season High heater usages,more artificial lighting due to shorter days

Summer season Low heater usage,potential use of fans for cooling

13
 Conclusion

Seeing the significant increase in electricity consumption rate in Nepal over past few years,

And 43% of Electricity usage of Nepal referring to residential consumption.

Home automation and maximizing the efficiency using AI is the most crucial
Step forward in the energy management sector.

With use of our simple, cost effective and one time installation home automation system
equipped with machine learning capabilities,domestic consumption can be made
significantly efficient with reduced electricity bills and financial benefit for homeowners. 14
 References
• https://www.researchgate.net/publication/371607691_SMART_HOME_AUTOMATION_SYSTEM_
BASED_ON_IoT
• Pavithra.D, RanjithBalakrishnan, “IoT based Monitoring and Control System for Home Automation”,
23 April 2015.
• Sudha Kousalya, G Reddi, Priya Vasanthi, B Venkatesh, IOT Based Smart Security and Smart Home
Automation presented at International Journal of Engineering Research & Technology 04, April-2018.
• Smart Home Monitoring System Using ESP32 Microcontrollers by Marek Babiuch and Jiri Postulka.
• Research Article of Internet of Things-Based Intelligent Smart Home Control System Olutosin Taiwo
and Absalom E. Ezugwu
• Smart Home Automation and Analysis Using Machine Learning by Aman Shukla, Masoom
Shaikh ,Jugal Paswan ,Siddhyant Mishra ,Baban U. Rindhe
• Dey S, Roy A, Das S. 2016 Home automation using internet of thing.

15
 Thankyou!!
Any Queries?

16