Vous êtes sur la page 1sur 33

DR.T.

THIMMAIAH INSTITUTE OF TECHNOLOGY,


Oorgaum, Kolar Gold Fields-563 120

2017-2018
DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING
1

.F, Dept of CSE


IMPLMENTATION OF MACHINE LEARNING
ALGORITHM FOR PREDICTING USER
BEHAVIOUR AND SMART ENERGY
MANAGEMENT

TEAM MEMBERS

Under the guidance of


Ms.REVATHI.S RONISHA BASKER(1GV14CS047)
Asst.Prof ,Dept.of CSE SANGEETHA .S(1GV14CS051)
SHILPA J.P(1GV14CS053)
M SHRUTHI
2 NIGADE(1GV14CS054)
Dr TTIT K.G.F, Dept of CSE
CONTENTS

 INTRODUCTION
 EXISTING SYSTEM
 PROPOSED SYSTEM
 IMPLEMENTATION
 ARCHITECTURE
 MODULES
 TEST CASES
 SNAP SHOTS
 CONCLUSION

Dr TTIT K.G.F, Dept of CSE


1.INTRODUCTION
• A great need arises to reduce consumption of electricity as energy is costlier and fossil fuels are not
good for the environment.

• This project targets domestic usage which has a more direct approach in changing the way we
consume energy. In this project we take up House Hold Loads as the application.

• Smart energy metering and normalized energy data on load usage are one of the major goal setters
for the future smart grid and improved energy efficiency in smart homes.

• Load Monitoring (LM) is essential for energy management and cost fixing. Appliance-specific
energy consumption statistics is obtained and is used to formulate load scheduling strategies for
optimal energy utilization, disaggregation of Load is essential.

• Non-Intrusive Load Monitoring (NILM) is an alternative and best method for Load Disaggregation,
as it can distinguish devices from the aggregated data measured at only a centralized location.

Dr TTIT K.G.F, Dept of CSE


2.EXISTING SYSTEM

 Smart Meter is the existing solution


 Smart meter measures the electricity consumption on unit level. User can threshold on
usage in smart meter and control the usage.

Dr TTIT K.G.F, Dept of CSE


3.PROPOSED SYSTEM

 NILM technology is used by implementing sub-metering system for each load


and feeding back the energy usage data of each appliance back to the controller
which is being interfaced with all the household appliances through Machine
Learning Algorithm and is used to predict energy usage.

 A GUI is developed for making the user informed about his electrical usage
giving various options like Budgeting, Maintenance needs, etc.

 The user can view energy usage predictions in both units , set budgets and
receive warnings if average daily usage is more than set budget.

Dr TTIT K.G.F, Dept of CSE


4. IMPLEMENTATION

 Load Prediction-ARIMA
 Failure prediction –THRESHOLDING
 PEAK /NON PEAK

Dr TTIT K.G.F, Dept of CSE


5.ARCHITECHURE
Power

Submeterin
Devices DB
g

Power Event
Calculation Analysis
Usage alert

Unit Failure
day/ consumed prediction GUI
month

Operation 8
Load
Dr TTIT K.G.F, Dept of CSE
failure
6.MODULES

 SUBMETERING
 UNIT CONSUMPTION FORECAST
 FAILURE FORECAST
 USAGE ALERTS

Dr TTIT K.G.F, Dept of CSE


Flow of Data

Unit Measurement/ Fail


Send to Server
Measurement

Unit Prediction/Failure
Unit Display Prediction
OVERVIEW OF ALL MODULES

Unit consumption
forecast

Failure
forecast

11

Dr TTIT K.G.F, Dept of CSE


6.1 SUBMETERING
Bulb
Switch

5V

Arduino
Microcontroller

Serial
Interface

Data Controller
DB

12

Dr TTIT K.G.F, Dept of CSE


Device Side Connection
6.2 UNIT CONSUMPTION FORECAST

Calculate unit
consumption

Unit consumption
forecast

14

Dr TTIT K.G.F, Dept of CSE


UNIT CONSUMPTION CLASS DIAGRAM

Dr TTIT K.G.F, Dept of CSE 15


6.3 FAILURE FORECAST

Failure forecast days

16

Dr TTIT K.G.F, Dept of CSE


FAILURE FORECAST CLASS DIAGRAM

Failure forecaster

Dr TTIT K.G.F, Dept of CSE 17


6.4 USAGE ALERTS

18

Dr TTIT K.G.F, Dept of CSE


USAGE ALERT CLASS DIAGRAM

Dr TTIT K.G.F, Dept of CSE 19


TEST CASES

UNIT TESTING Classes integrated Tests done Remarks

Class: Admin Interface Class tested burning Success


program to Smart meter
and handling the
consumption and recharge

Class: Electricity User Class tested to check web Success


user profile is managed

Class: Smart Meter Class tested to check Success


whether monitoring
consumption and control is
implemented
Class: Predictor Class tested to check Success
whether prediction is
working

Dr TTIT K.G.F, Dept of CSE 20


VALIDATION TESTING

Functionality to be Input Tests done Remarks


tested
Working of Front-End User interaction with help Appropriate forms open when Success
of a mouse and keyboard buttons are clicked

Working of Monitoring Device is switched on Consumption is monitored and Success


reported to server

Working of Prediction Unit is collected for many Based on past days unit Success
days consumption, future
consumption and failure is
predicted

Dr TTIT K.G.F, Dept of CSE 21


WEB PORTAL(Registration)

Dr TTIT K.G.F, Dept of CSE 22


Web Portal(login)
Web portal(Account details)
Backend database
unit consumed

Dr TTIT K.G.F, Dept of CSE 25


Failure data

Dr TTIT K.G.F, Dept of CSE 26


User profile

Dr TTIT K.G.F, Dept of CSE 27


UNIT CONSUMPTION FORECAST

Dr TTIT K.G.F, Dept of CSE 28


FAILURE FORECAST

Dr TTIT K.G.F, Dept of CSE 29


User alert

Dr TTIT K.G.F, Dept of CSE 30


7.CONCLUSION

 User can manage electricity in a conserved manner without wastage.


 User can use electricity according to his budget
 Failures are notified in advance, so user can make backup plan to manage the
failure.

31

Dr TTIT K.G.F, Dept of CSE


Advantage

 User can manage electricity in a conserved manner without wastage.


 User can use electricity according to his budget
 Failures are notified in advance, so user can make backup plan to manage the
failure.
Dr TTIT K.G.F, Dept of CSE 33