1 of 16

DESIGN AND IMPLEMENTATION OF A SMART IRRIGATION SYSTEM POWERED BY DEEP LEARNING ALGORITHMS.

AKOLA MBEY DENIS 5946016

LOUIS MARIE AYARIGA ATOLUKO 5948516

1

2 of 16

Presentation Outline

  • Introduction
  • Overview of project
  • Aims and objectives
  • System Design and Implementation
  • Challenges
  • Conclusion
  • Recommendation for future work

2

3 of 16

AIMS AND OBJECTIVES

  • Predict rainfall for the next two weeks using Deep Learning Algorithms.

​

​

  • Apply the rainfall information to build an efficient and smart irrigation system.

​

​

  • Build a mobile application to present weather prediction to users.

3

4 of 16

System Design and Implementation

4

5 of 16

System Architecture

5

6 of 16

Our proposed system architecture block diagram

6

7 of 16

MOBILE APPLICATION

  • The mobile application was built to send real-time information from the farm to the farmer.
  • The some features of the mobile app are:
    • It registers and authenticate a new farmer.
    • It has a dashboard that displays the real-time farm measurements to the farmer.
    • The features of the dashboard includes the amount of rain, the status of the water pump.

7

8 of 16

How the mobile app works

8

The activity diagram describes how the mobile app works

9 of 16

Mobile application

  • It was built using react-native and firebase.

​

9

10 of 16

Deep learning Model

  • The dataset we used to build the model was obtained from Raspisaniye Pogodi Ltd website.
  • The weather data for Kumasi was available for the period from January 2010 to May, 2020.
  • Long Short Term Memory network was used to train the model.

10

11 of 16

Deep Learning Model

  • The performance metrics for the model were accuracy and the root mean squared error.
  • After several training experiments the root mean squared error of 7.346 was obtained.

11

12 of 16

Deep learning model

12

Training and validation loss plot

Model’s Prediction plot

13 of 16

Hardware Design and Implemention

13

14 of 16

Challenges

  • The dataset we had to train our deep learning model had a lot of null-valued fields which does not give a clear and real-life reflection of the daily weather for Kumasi.
  • The cost of some of the sensors the sensors we needed were so exorbitant and we could not afford them
  • Due to the Covid-19 pandemic, we were not able to integrate the various aspects of the system.

​

14

15 of 16

Recommendation

  • We recommend that more research should be done in this area to help the agriculture industry converse water as water is becoming a scarce resource.

​

15

16 of 16

​

​

​

Thank you

16