Complex Engineering Problem — Artificial Intelligence

Open Book Assessment

Total Marks: 100
Title:

AI-Based Smart Energy Analytics System for University Campus

Problem Statement
A university campus aims to develop an intelligent AI-based energy analytics system capable of:
1. Predicting hourly energy consumption of academic buildings
2. Classifying buildings into operational categories
3. Detecting abnormal energy usage patterns using deep learning
The university has installed IoT-enabled smart meters across multiple buildings. The collected dataset
contains environmental, operational, and electrical parameters.
Students are required to design, implement, analyze, and evaluate an integrated AI solution using:
 Regression techniques
 Classification techniques
 Multilayer Perceptron (MLP) Neural Networks
The project should demonstrate engineering decision-making, model comparison, analytical
interpretation, and performance evaluation.

Dataset Description
The dataset contains the following features:

Feature Description

Temperature Ambient temperature (°C)

Humidity Relative humidity (%)

Occupancy Number of occupants

HVAC_Load HVAC power consumption (kW)

Lighting_Load Lighting power consumption (kW)

Equipment_Load Equipment consumption (kW)

Solar_Generation Solar PV generation (kW)

Time_Index Hour of day

Total_Consumption Total building consumption (kWh)

Feature Description

Building_Type Academic / Laboratory / Hostel

Fault_Status Normal / Faulty

Students may generate a synthetic dataset or use publicly available smart energy datasets.

Tasks

Part A — Data Analysis & Preprocessing (15 Marks)

Students are required to:
1. Perform exploratory data analysis (EDA)
2. Handle missing values and outliers
3. Normalize or standardize features
4. Visualize feature relationships
5. Justify selected preprocessing techniques
Deliverables
 Correlation matrix
 Histograms / scatter plots
 Feature importance discussion

Part B — Regression Model (25 Marks)

Develop a regression model to predict:
[
\text{Total Energy Consumption}
]
using:
 Linear Regression
 Multiple Linear Regression
Requirements
1. Develop mathematical formulation
2. Train and test the model
3. Compute:
o MAE

o MSE
 o RMSE

o (R^2) score

4. Compare regression performance

5. Interpret coefficients and engineering significance
Use:

Part C — Classification Model (25 Marks)

Develop a classification system to predict:
Building Type or Fault Status

using:
 K-Nearest Neighbors (KNN)
 Logistic Regression
 Decision Tree (optional bonus)
Requirements
1. Compute classification accuracy
2. Generate confusion matrix
3. Calculate:
o Precision

o Recall

o F1-score

4. Analyze impact of feature scaling

5. Study impact of varying K in KNN
Use:

Part D — Multilayer Perceptron (MLP) Neural Network (25 Marks)

Develop an MLP model for:
 Energy prediction OR
 Fault classification
Requirements
1. Design network architecture
o Input layer
 o At least two hidden layers

o Output layer

2. Select suitable activation functions

3. Perform forward propagation analysis
4. Train the network using backpropagation
5. Analyze:
o Loss convergence

o Overfitting/underfitting

o Learning rate effects

6. Compare MLP performance with conventional ML models

Use sigmoid/ReLU where appropriate.
Sigmoid activation:

Part E — Engineering Analysis & Discussion (10 Marks)

Students must critically discuss:
 Why certain models performed better
 Computational complexity
 Real-time deployment feasibility
 AI reliability in smart grids
 Ethical and sustainability considerations

Report Requirements (IEEE Format)

Students must prepare the report in standard IEEE conference format including:
1. Abstract
2. Introduction
3. Literature Review
4. Methodology
5. Dataset Description
6. Experimental Results
7. Discussion
8. Conclusion
9. References
Submission Requirements
Students must submit:
 IEEE-format report (PDF)
[Link]
[Link]

 Source code
 Dataset
 Result graphs and screenshots
 Trained model files (optional)

Marking Distribution

Component Marks

Data Analysis & Preprocessing 15

Regression Modeling 25

Classification Modeling 25

MLP Neural Network 25

Discussion & Presentation and References 10

Expected Learning Outcomes

Upon completion, students should be able to:
 Develop AI pipelines for engineering systems
 Apply regression and classification techniques
 Design and evaluate neural networks
 Analyze real-world engineering datasets
 Present technical findings professionally in IEEE style
 Assessment Rubric

Sr.
Mar
No Component Excellent Good Satisfactory Poor
ks
.

Clear
Problem Good
objectives, Basic
Understandi understanding Unclear problem
1 strong explanation 10
ng & with minor definition
engineering only
Introduction gaps
justification

Proper
Data preprocessin
Minor Limited
Analysis & g, Poor or missing
2 preprocessing analysis 15
Preprocessi visualization, preprocessing
issues performed
ng and feature
analysis

Correct
implementati Minor Basic Incorrect or
Regression
3 on with calculation/mo implementati incomplete 20
Modeling
detailed del issues on only model
evaluation

Accurate
Minor Limited Incorrect/
Classificatio classification
4 evaluation metric incomplete 20
n Modeling with proper
issues analysis classification
metrics

Well-
Good
designed Basic MLP Poor or
MLP Neural implementatio
5 network with implementati incomplete 20
Network n with small
strong on neural network
issues
analysis

Excellent
Good
interpretation
Results & discussion Limited Weak or missing
6 and 10
Discussion with minor interpretation discussion
comparison
gaps
of models

IEEE Fully
Minor
Formatting professional Acceptable Poor presentation
7 formatting 5
& IEEE formatting and formatting
issues
Presentation formatting

Total 100