Diploma_Project_Presentation_

1.

COMPUTER ENGINEERING
DIPLOMA PROJECT
Modelling a Growing Area and Developing
an IoT Microclimate Monitoring System to
Study and Predict Favourable Conditions
for Plant Growth Using AI
Students: Ksenia Semko, Yaroslav
Abrashkin, Islam Safarov, Yanis
Supervisor: Serik Kaparovich
Grigoryev
Computer Engineering
Almaty, 2026

2.

Research Aim and Objectives
AIM
1
3
5
To develop an autonomous IoT system for monitoring a growing area and predicting
favourable plant-growth conditions using AI.
Review existing solutions
Review IoT and machine-learning approaches.
Design the system
Design a multi-sensor monitoring architecture.
Build and evaluate
Implement and test an offline prediction model.
2
4
6
Analyse monitoring methods
Study methods for measuring environmental
conditions.
Model spatial variation
Estimate microclimate parameters across the area.
Compare approaches
Compare with conventional threshold-based control.

3.

Project Team and Responsibilities
01
02
03
04
TEAM MEMBER
TEAM MEMBER
TEAM MEMBER
TEAM MEMBER
• Ksenia Semko
Research coordination,
methodology,
documentation and
evaluation planning.
• Yaroslav Abrashkin
Responsibilities to be
confirmed with the team.
• Islam Safarov
Responsibilities to be
confirmed with the team.
• Yanis Grigoryev
Responsibilities to be
confirmed with the team.

4.

Research Relevance and Problem Statement
Problem Statement
Conventional irrigation often relies on fixed
thresholds. These rules may miss spatial
differences in soil moisture and local
environmental conditions. Sensor errors and
dependence on cloud services can further
limit reliability.
Why It Matters
1
Spatial variation
Conditions can differ across a growing area.
2
Sensor reliability
Low-cost sensors may need calibration and error
compensation.
3
Offline operation
Local processing can preserve core functions
without Internet access.

5.

Object and Subject of the Study
OBJECT
The broader process being studied.
Spatiotemporal monitoring and predictive
control of microclimate conditions in a
growing area.
SUBJECT
The specific methods and system components
under study.
Spatial interpolation, machine-learning
algorithms, and an autonomous IoT
architecture based on ESP32 and Raspberry
Pi.
Research question: How can these methods
improve prediction without Internet access?

6.

Theoretical Background and System Architecture
Local processing pipeline
Sensors → ESP32 nodes →
MQTT → Raspberry Pi →
Dashboard / Control
SENSORS
Soil moisture
Temperature · Air humidity
ESP32 NODES
Read and validate
measurements
RASPBERRY PI GATEWAY
Local database · Spatial
interpolation
Offline ML prediction
LOCAL MQTT BROKER → DATA PROCESSING → DASHBOARD / IRRIGATION
CONTROL
Design goal: essential monitoring and prediction should continue without external Internet access.

7.

Initial Results and Prototype Status
Reported prototype components
Evidence to include in the defence
• ESP32 microcontroller
• Capacitive soil-moisture sensor
• DHT22 temperature and humidity
sensor
• MQTT communication
• Local Mosquitto broker on Raspberry Pi
• Photograph of the prototype
• Example sensor readings
• MQTT message or data log
• Calibration code and test results
Numerical model performance has not yet been confirmed. Report only components and functions that have
been implemented and verified.

8.

Evaluation Plan and Evidence
Use measured data only; do not present unverified outcomes as achieved results.
DATA COLLECTION
1. Read sensor values
2. Validate and timestamp data
3. Calibrate sensors
4. Store measurements locally
Key point:
MODEL EVALUATION
• Train/test split
• Compare against threshold baseline
• MAE and RMSE; R² where appropriate
• Measure latency and memory use
The prototype and prediction model must be evaluated using logged measurements
and a predefined baseline before performance claims can be made.

9.

Proposed Solution and Practical Significance
01
02
03
04
Collect data
Calibrate sensors
Model conditions
Compare results
Compensate for
measurement errors and
check sensor
consistency.
Apply spatial
interpolation and test an
offline ML model.
Compare prediction, water
use, latency and offline
reliability with a threshold
baseline.
Read, validate and
timestamp sensor
measurements.
Evaluation metrics: MAE · RMSE · R² (where appropriate) · water consumption · latency · offline reliability. Any water-saving
target remains a hypothesis until tested.

10.

Conclusions and Future Work
1
The project proposes an integrated IoT architecture for monitoring microclimate conditions in a growing
area.
2
ESP32 sensor nodes and a Raspberry Pi gateway support local data collection and processing.
3
Spatial estimation and machine-learning prediction will be evaluated experimentally.
4
The main hypotheses must be tested against a predefined threshold-based baseline.
5
Future work: complete the prototype, collect data, evaluate performance and document the results.
Thank you for your attention! | I will be happy to answer your questions.