Professional Documents
Culture Documents
Teep Program
_____________________________________________
This research addresses the need for continuous monitoring of solar panel systems in
factories to prevent electricity loss and financial implications. It aims to develop a machine
learning-based methodology for detecting outliers in solar panel data, providing an
automated and efficient solution. The significance of this research lies in its potential to
enhance the efficiency and reliability of solar panel systems, contribute to the prevention
of electricity loss, and promote renewable energy technologies. The outcomes can inspire
advancements in anomaly detection techniques for other industrial systems and support the
global shift towards sustainable energy sources. This research has practical implications
for solar panel factories, renewable energy technologies, and the broader goal of
sustainable energy.
LITERATURE REVIEW
Data monitoring involves collecting and analyzing real-time data from sensors within the
system, storing it in a centralized database or cloud-based platform for easy access and
analysis. This monitoring serves several purposes. First, it performs maintenance under
performing components, enabling prompt or replacement to maximize energy generation.
Second, it detects and diagnoses anomalies such as shading or faulty connections that can
decrease system efficiency. Additionally, long-term data analysis helps assess system
performance trends, identify degradation, and plan for upgrades or improvements.
In summary, data monitoring is essential for optimizing solar power system performance,
detecting anomalies, and extending their lifespan. Effective monitoring techniques and
tools are crucial for reliable operation, whether in residential settings or large-scale solar
farms. By harnessing the power of data, solar power systems can achieve higher efficiency
and contribute to a sustainable energy future.
METHODOLOGY
We will call data from csv file by Pandas library, like as:
Shall we check if solar data is missing? If it is missing, we can use methods to deal with it.
But here I will remove the missing data because my data is enough.
As we see, the data is missing, namely I_AC, I_DC, U_AC, U_DC missing 2533 values.
Irradiance and T_MODULE are missing 2298 values. So we will delete by most missing
value we will use the df.dropna() function.
The correlation coefficient is a measure of some kind of correlation, that is, a statistical
relationship between two variables.
● When this correlation is less than 0, it means that the relationship is inverse, when
one variable increases, the other decreases and vice versa
● When this correlation is greater than 0, it means that the relationship is positive,
when one variable increases, the other increases and when this variable decreases,
the other decreases.
● When this correlation is 0, 2 variables are not related to each other.
Show correlation matrix. This matrix displays all the correlation values between the data
columns. In practice, this display is often to identify pairs of strongly correlated features
from which to make a decision not to use either feature to build the model.
We can see outliers outside of linear regression as shown below:
Data collection and preprocessing lay the foundation for accurate and reliable anomaly
detection in solar panel data. These steps ensure that the dataset is clean, normalized, and
appropriately prepared for training machine learning algorithms. By meticulously
collecting and preprocessing the data, the research can build a robust anomaly detection
system capable of accurately identifying outliers and anomalies in the solar panel system.
3.2 Selection and Implementation of Machine Learning Algorithm
The OneClassSVM algorithm was chosen for anomaly detection in solar panel data due to
its effectiveness in unsupervised outlier detection. By finding a hyperplane to separate
normal data from outliers, OneClassSVM maximizes the margin around normal data
points. The algorithm utilizes kernel functions to map the data into a higher-dimensional
space and constructs a hyperplane defined by support vectors. OneClassSVM is capable of
handling non-linear data distributions and is robust to noise. Implementing the algorithm
allowed for the identification of anomalies in solar panel data by detecting deviations from
normal patterns. Overall, OneClassSVM provides a powerful tool for unattended anomaly
detection, contributing to the development of a reliable monitoring system for solar panels.
This is my result:
It creates a boundary around the normal instances in a dataset, considering them as the
only class during training. Instances that fall outside the boundary are considered outliers.
How can the boundary be evaluated?
Scatter Plot: A scatter plot can be created to visualize the distribution of the solar panel
data. Normal data points can be plotted in one color, while the outliers detected by the
OneClassSVM model can be highlighted in a different color. This allows for a visual
inspection of how well the model identifies and separates anomalies from the normal data
instances.
This is result:
By using research data visualization techniques it is possible to better understand the
effectiveness of the OneClassSVM model in accurately detecting anomalies in the solar
panel data. These assessments aid in assessing the model's performance and its potential
for real-world deployment in monitoring solar panel systems.
3.4 Design and deploy model machine learning of the anomaly detection on the
ưebsite
To make the anomaly detection model accessible and user-friendly, it can be designed and
deployed on a website using Docker and Flask API. This allows users to interact with the
model and obtain anomaly predictions conveniently. Here's an outline of the design and
deployment process:
Web Interface Design: Create an intuitive and user-friendly web interface where users can
input their solar panel data for anomaly detection. Design the interface to accept relevant
parameters such as irradiance, voltage, and current values.
Dockerization: Dockerize the Flask application to ensure easy deployment and portability.
Create a Dockerfile that specifies the application's dependencies, configurations, and
runtime environment. Containerization and Deployment: Build a Docker container using
the Dockerfile and deploy it on a web server or cloud platform of your choice. This allows
the web application, along with the anomaly detection model, to be easily deployed and
scaled.
API Development: Implement a Flask API to expose the anomaly detection functionality.
This enables developers or other applications to interact with the model programmatically,
allowing for integration with external systems.
● Step 2: Wait 5 minutes - Until container log show something like this
● Step 3: Go to http://127.0.0.1:8050/
This is my result:
By following these steps, the anomaly detection model can be seamlessly integrated into a
user-friendly web application. Users can access the website, input their solar panel data,
and receive real-time anomaly predictions. The use of Docker ensures easy deployment
and scalability, while the Flask API enables seamless integration with other systems. This
design and deployment approach the accessibility and usability of the anomaly detection
model in monitoring solar panel systems.
CONCLUSION
Furthermore, the TEEP program facilitates the formation of meaningful connections and
friendships with fellow interns from different parts of the world. Collaborating with them
on projects and sharing our diverse perspectives was truly enriching. The program fostered
a supportive and collaborative environment, promoting personal growth and professional
development.
In addition, the TEEP program equipped me with practical skills and knowledge through
the hands-on experience gained during my internship. The opportunity to work on the solar
panel anomaly detection project enhanced my understanding of machine learning
algorithms and their application in real-world scenarios. The skills acquired will
undoubtedly contribute to my future academic and professional pursuits.
Overall, my participation in the TEEP program has been a transformative journey filled
with valuable experiences, cultural immersion, and lifelong friendships. I am immensely
grateful for the support and guidance provided by the program organizers and mentors.
The TEEP program has undoubtedly shaped my personal and professional growth, leaving
a lasting impact on my life.
I am extremely grateful to Professor Kemo and English Teacher Cindy for their invaluable
guidance, support, and mentorship throughout my TEEP journey. Professor Kemo's
expertise in solar panel systems and anomaly detection has been instrumental in
overcoming challenges and deepening my understanding of the subject. I am truly thankful
for his patience and encouragement. English Teacher Cindy's unwavering support and
assistance in improving my language skills have greatly contributed to my overall
development. Her feedback and motivation have played a vital role in my growth as a
student and researcher. I feel privileged to have had the opportunity to work with such
exceptional teachers who have inspired me to strive for excellence. Their mentorship has
left a lasting impact on my personal and academic growth. I extend my sincere
appreciation for their contributions to my TEEP experience.