A TECHNICAL REPORT FOR PROJECT PART I ON

OPTIMIZATION OF DUAL AXIS SOLAR TRACKING SYSTEM USING

MACHINE LEARNING

Submitted in partial fulfillment of the requirements for the award of the degree
of

BACHELOR OF ENGINEERING

in

ELECTRICAL AND ELECTRONICS ENGINEERING

By

S SIDDARTH CHANDRA (160120734054)

CH. MONITHA SAI (160120734010)

P. KUNDANA (160120734009)

Under the esteemed guidance of

SRI.P. HEMESHWAR CHARY
Assistant Professor
Department of EEE, CBIT (A), Hyderabad.

1
 Department of Electrical and Electronics Engineering
Chaitanya Bharathi Institute of Technology
(Autonomous) Gandipet, Hyderabad-500075.

SL.NO. CONTENTS PAGE NO.

1
1. Abstract

2. Introduction 2

5
3. Literature Review

7
4. Objectives

8
5. Methodologies Proposed

11
7. Machine Learning Integration

12
6. The level/status of the work

2
 1. ABSTRACT:
In the pursuit of optimizing solar energy harvesting, this project presents an advanced Dual-
Axis Solar Tracking System that integrates Machine Learning (ML), data visualization,
monitoring, error handling, and user reporting. The primary objective is to enhance the
efficiency and reliability of solar panel orientation, ensuring maximum energy generation
throughout the day.

Our DASTS incorporates dual axis tracking and ML algorithms to adjust panel angles based
on real-time solar conditions, acting as a redundant backup system while also ensuring
efficient energy capture. One key addition to our system is the integration of an obstacle
detection system using ML models, which enables the DASTS to detect and maneuver away
from obstacles, thus eliminating shading and further enhancing energy capture.

Moreover, we address Dual-Axis Solar Tracking System optimization and management,

implementing ML algorithms for calibration, fault detection, redundancy, and maintenance
scheduling. These techniques not only improve energy capture but also maintain system
efficiency by proactively identifying and mitigating issues.

To enhance data presentation and user accessibility, we utilize a cloud service provider and
a data visualization library. This setup allows users to access real-time and historical data
remotely, making informed decisions about system performance, and overall energy
optimization.

1
 2. INTRODUCTION:
Solar panels, the bedrock of sustainable energy generation, possess an inherent promise of
harnessing the sun's radiant energy. However, to unlock their full potential, they must be
meticulously oriented toward the sun's trajectory throughout the day.

This pivotal alignment directly influences the efficiency of energy conversion. Passive
systems, fixed in position, invariably fall short of realizing this aspiration. It is through the
application of cutting-edge technology and meticulous engineering that we embark on a
journey to elevate solar energy generation to new heights.

Our dual-axis solar tracking system exemplifies the pinnacle of solar panel optimization. By
actively and dynamically repositioning solar panels to perpetually face the sun, we not only
transcend the limitations of fixed installations but redefine the notion of energy efficiency.
Light Dependent Resistors (LDRs) in conjunction with precision motors orchestrate this
delicate solar ballet. Machine learning algorithms stand as vigilant sentinels, ready to
intervene should sensor anomalies arise.

We introduce a sophisticated fault detection and user alert system, steadfastly safeguarding
against operational irregularities. By closely monitoring critical parameters, such as
temperature, component health, and motor performance, we ensure the longevity and
reliability of the system. Stakeholders remain promptly apprised of system status through a
comprehensive alert system that spans email, SMS, and mobile application notifications.
Our cloud-based interface adds an extra layer of utility and convenience, offering insightful
data visualization. Comprehensive operational records and performance reports further
empower users to make informed decisions and facilitate expedited maintenance routines.

(1) The study delves into the implementation of a solar tracker system, focusing on the
incorporation of a Machine Learning model for obstacle detection and avoidance. This
innovative approach intertwines traditional engineering practices with advanced
technology, specifically Machine Learning, to enhance the performance and functionality of
renewable energy systems. The research emphasizes the crucial synergy between solar
tracking mechanisms and intelligent algorithms, showcasing the capacity for these systems
to adapt and overcome obstacles, ultimately bolstering the safety and efficiency of solar
energy capture.

(2) The study not only introduces a novel physical form but also integrates Machine Learning
to forecast future power output and temperature. This integration marks a compelling leap
in the application of predictive technology within the renewable energy sector, indicating a
potential for more precise and informed energy generation estimations. It illuminates the
advancements made in solar panel design by introducing a 3-dimensional structure that
promises increased energy generation and enhanced efficiency compared to traditional
rectangular panels. By employing Machine Learning for predictive analysis, the study
underlines the potential to accurately anticipate energy output and temperature, ultimately
contributing to more informed decision-making in the renewable energy landscape.

2
(3) The study delves into the development and execution of a dual-axis solar tracking
algorithm within a solar tracking system. It highlights the successful design and
implementation of this innovative system. However, the research identifies critical
challenges that hinder the optimal use of tracker systems in solar energy applications.
Notably, the study underscores concerns related to power consumption within the electrical
components of the tracker system, indicating the pressing need for more energy-efficient
designs to ensure sustainability and cost-effectiveness. Additionally, the paper emphasizes
the significance of refining the technology used for tracking to better estimate the most
effective path for solar panels.

(4) The paper centers on the creation of an innovative control design approach that
dynamically alters the orientation of solar panels, employing cost-effective soft actuators to
heighten their energy harvesting efficiency. The study introduces a novel concept of
leveraging low-cost soft actuators to adjust the positioning of solar panels, aiming to
optimize their exposure to sunlight. This groundbreaking approach signifies a departure
from traditional rigid mechanisms, offering adaptability and cost efficiency. Additionally, the
research incorporates machine learning techniques for calibrating and fine-tuning panel
movements. This integration of machine learning signifies a progressive step towards
enhancing the precision and effectiveness of solar panel adjustments, marking a
transformative shift in the calibration and optimization processes within the realm of
renewable energy.

(5) The paper presents a comprehensive overview of the significance of solar tracking
systems in augmenting the energy generation of solar panels. It emphasizes that solar
energy, when coupled with solar tracking mechanisms, holds the capacity to significantly
boost energy output by continuously adjusting the panels to align with the sun's position.
Despite the initial high setup costs associated with these systems, the paper highlights the
evolution of cost-effective alternatives proposed over time, showcasing the strategic
placement of solar panels to capture maximum sunlight. The discussion delves into the
comparison between different types of motors, emphasizing the efficiency of servo motors,
which maintain torque at high speeds and operate within an impressive 80-90% efficiency
range.

(6) The paper delves into the realm of photovoltaic technology and its applications,
specifically focusing on the critical factors of installing effective Maximum Power Point
Tracking (MPPT) and the calibration of an automatic MPPT algorithm. The efficient
operation of a photovoltaic (PV) system hinge upon the ability to generate the maximum
amount of power, necessitating a precise method for finding the global maximum power
point of a PV array. The study introduces an innovative approach utilizing an Artificial Neural
Network (ANN) to automatically determine the global maximum power point in a suggested
MPPT technique system. The paper meticulously considers various aspects, such as function
approximation in the MPPT problem for PV, optimal technique selection, and network
architecture, considering different training methods and network topologies to optimize
performance.

3
3.LITERATURE SURVEY:
 P. Gupta, V. Gupta, M. Sharma, R. K. Pachauri and J. Akhtar, "Design and
Performance Analysis of Three axis Solar Tracking System," 2022 7th Asia Conference
on Power and Electrical Engineering (ACPEE), Hangzhou, China, 2022, pp. 1876-1880,
doi: 10.1109/ACPEE53904.2022.9783762.

Summary: This study introduces the design and performance of a solar tracker system.
The primary objective is of evolving a Machine learning model for obstacle detection and
avoidance.

 C. Pavithra, R. Rithan, R. Rohith, M. Santhosh, M. Vijayadharshini and R. Lokith,

"Three Dimensional Solar Tracking Using Machine Learning," 2023 9th International
Conference on Advanced Computing and Communication Systems (ICACCS),
Coimbatore, India, 2023, pp. 369-372, doi: 10.1109/ICACCS57279.2023.10113091.

Summary: The primary target of this paper is to build a 3-dimensional solar panel that
generates more energy and is more efficient than a traditional shape of rectangular solar
panel. Machine learning is used to predict the future output power and temperature.

● F. A. Al-Muswe, I. K. A. Al-Amri, M. N. H. Al-Turfi and A. Alkhayyat, "Solar Tracking

System by Utilized Optimized Algorithm Based Deep Learning," 2022 5th
International Conference on Engineering Technology and its Applications (IICETA), Al-
Najaf, Iraq, 2022, pp. 37-44, doi: 10.1109/IICETA54559.2022.9888545.

Summary: In this study, a dual-axis solar tracking system-based solar tracking algorithm
has been designed and implemented. However, several issues still need to be overcome for
using a tracker system that includes power consumption related to the electrical part of the
tracker and the estimation of the best path based on tracking technology that is utilized.

● D. Barron, P. V. Chakravarthy, R. Nagarkar, Y. Cao and W. Gao, "IMU-based

Autonomous Solar Tracking System with Soft Actuators: A Machine Learning
Approach," 2021 IEEE Conference on Control Technology and Applications (CCTA),
San Diego, CA, USA, 2021, pp. 1172-1178, doi: 10.1109/CCTA48906.2021.9658869.

Summary: This paper focuses on developing a new databased control design approach

4
that adaptively adjusts the orientation of solar panels, built upon low-cost soft actuators, to
increase their energy harvesting efficiency, and uses machine learning calibration and tuning
panel movement.

● L. Nanda, A. Dasgupta and U. K. Rout, "Smart solar tracking system for optimal
power generation," 2017 3rd International Conference on Computational
Intelligence & Communication Technology (CICT), Ghaziabad, India, 2017, pp. 1-5,
doi: 10.1109/CIACT.2017.7977340.

Summary: Solar energy, when equipped with solar tracking systems, can generate a
greater amount of energy by continuously aligning with the sun's position. Despite the initial
high setup costs, various cost-effective alternatives have been proposed over time. The solar
panel is strategically placed to capture maximum sunlight. In comparison to other types of
motors, servo motors can maintain their torque at high speeds and exhibit impressive
efficiencies within the 80-90% range. In contrast, most panels operate below 40%, which
prompts people to either invest in multiple panels to meet their energy needs or opt for
single systems with high output. To enhance solar panel efficiency and reduce expenses,
solar tracking emerges as a viable solution. Solar tracking increases the panel's exposure to
sunlight, resulting in higher power output. Tracking systems can be either dual-axis or
single-axis trackers, with dual-axis trackers being more efficient as they follow the sun's
movement on both axes.

● B. R. Kumar, S. Srikanth, P. Sujeet, T. Swetha and P. Tanya, "Design and

Implementation of Solar based Maximum Power Point Tracking using Machine
Learning," 2023 5th International Conference on Smart Systems and Inventive
Technology (ICSSIT), Tirunelveli, India, 2023, pp. 247-253, doi:
10.1109/ICSSIT55814.2023.10060975.

Summary: This paper discusses the applications based on photovoltaic technology.

Effective MPPT installation and automatic MPPT algorithm calibration are necessary for a
photovoltaic (PV) system to generate the maximum amount of power. The global maximum
power point of a PV array is automatically found in our suggested maximum power point
tracker technique system by employing an Artificial Neural Network (ANN)-based approach.

5
This method is used on a battery-operated solar tracking system. AI-based MPPT techniques
are ideal for photovoltaic systems. The MPPT problem function approximation in PV, the
optimal technique, and the network architecture to select for the performance of many
training methods and network topologies are all considered.

• K. Ramaneti, P. Kakani and S. Prakash, "Improving Solar Panel Efficiency by Solar Tracking
and Tilt Angle Optimization with Deep Learning," 2021 5th International Conference on
Smart Grid and Smart Cities (ICSGSC), Tokyo, Japan, 2021, pp. 102-106, doi:
10.1109/ICSGSC52434.2021.9490485.

Summary: The increasing energy demand in recent decades has caused significant
depletion of natural energy sources, necessitating active research and development of
efficient renewable energy systems. Solar energy stands out as the most abundant
renewable energy source at our disposal, making the adoption of solar energy a promising
pathway towards a sustainable future. However, the varying position of the sun relative to
the earth throughout the day results in fluctuating sunlight intensity on the photovoltaic
cells of solar panels, leading to a notable drop in overall efficiency. Therefore, it is crucial to
focus on enhancing solar panel array efficiency to maximize solar power generation.

I. MACHINE LEARNING:
We have learnt how to develop a Machine learning model for obstacle detection and
avoidance for enhanced energy harvesting efficiency. And to implement Machine learning to
predict the future output power and temperature.
 P. Gupta, V. Gupta, M. Sharma, R. K. Pachauri and J. Akhtar, "Design and
Performance Analysis of Three axis Solar Tracking System," 2022 7th Asia Conference
on Power and Electrical Engineering (ACPEE), Hangzhou, China

 C. Pavithra, R. Rithan, R. Rohith, M. Santhosh, M. Vijayadharshini and R. Lokith,

"Three Dimensional Solar Tracking Using Machine Learning," 2023 9th International
Conference on Advanced Computing and Communication Systems (ICACCS)

 D. Barron, P. V. Chakravarthy, R. Nagarkar, Y. Cao and W. Gao, "IMU-based

Autonomous Solar Tracking System with Soft Actuators: A Machine Learning
Approach," 2021 IEEE Conference on Control Technology and Applications (CCTA)

6
 II. AUTOMATED TRACKING OF SOLAR PANELS:

In these papers we learnt and implemented the dual axis solar tracking system.

 S. Mohan, K. Rajkumar, P. Rajakumar, S M Hari Pradosh, “Automatic Solar Tracking

System”, 2022 International Conference on Power, Energy, Control and Transmission
Systems (ICPECTS)

4. Objectives:
• To improve dual axis tracking system with ML via redundant tracking and
obstacle detection and avoidance. Aims to improve system efficiency, robustness,
and safety, ensuring uninterrupted solar panel alignment with the sun while
mitigating the impact of obstacles on energy capture.

• Implement a comprehensive fault detection and alert system in the dual-axis

solar tracking system to proactively identify and report issues. This system will
prioritize critical problems, provide real-time alerts to the user/owner, and offer
solutions or recommendations for resolution, ultimately ensuring minimal system
downtime and increased reliability.

• Create an intuitive and user-friendly interface for data visualization and

parameter control within the dual-axis solar tracking system. This interface will
enable the user/owner to easily monitor system performance, access historical data,
adjust tracking parameters, and visualize energy production trends, thereby
empowering them to make informed decisions and optimize energy efficiency
effectively.

7
5.Methodologies:
• Implement accurate machine learning models within the dual-axis solar tracking system to
enhance tracking reliability through redundant tracking capabilities and to enable effective
obstacle detection and aims to improve the system's ability to adapt to various
environmental conditions and obstacles while maintaining efficient solar panel positioning
for optimal energy capture.
• Implement user notification and exception handling to enable efficient and secure
communication between the dual-axis solar tracking system and external email services.
This objective aims to ensure reliable notifications and alerts from the system, enhancing
its monitoring and maintenance capabilities.
• To develop and deploy GUI using APIs to enable effective user management and provide
intuitive data visualization capabilities within the dual-axis solar tracking system. This
objective aims to enhance user accessibility, simplify system control, and facilitate real-
time monitoring of solar panel performance, contributing to improved system efficiency
and user experience.

6.Machine Learning Integration:

The machine learning integration segment focuses on the development and implementation of
predictive models to anticipate the solar panel's positioning in the absence of reliable sensor data.
The following sub-sections detail the data collection, preprocessing, model selection, training, and
integration with the motor control system.

6.1 Data Collection and Preprocessing

In our collaborative project, the group unanimously decided to implement a linear regression model
to predict the solar panel's positioning. We extensively collected historical data related to solar
panel angles, environmental conditions, and contextual information. The collected data were
preprocessed to ensure uniformity and consistency, encompassing normalization techniques and the
robust handling of missing values to optimize the dataset for linear regression analysis.

8
In the code snippet above we read data and then preprocess it according to our needs and test_train
split the dataset and build and train a linear regression model on it.

6.2 Model Selection and Training

After rigorous deliberation, we selected the linear regression model due to its simplicity and
effectiveness in predicting continuous output variables. The chosen model underwent
comprehensive training using the preprocessed dataset, with a strong emphasis on parameter
tuning and performance optimization. Cross-validation techniques were integrated into the training
process to thoroughly evaluate the model's predictive accuracy and its ability to generalize to real-
world scenarios.

9
The diagram above illustrates the fundamental concept of a linear regression model, showcasing its
ability to establish a linear relationship between the input and output variables. The model's
predictive capability is based on fitting a line that best represents the trend observed in the historical
data, enabling precise forecasts of solar panel positioning even in the absence of real-time sensor
data.

6.3 Integration with Motor Control

To mitigate the impact of potential sensor failures, our team developed a localized instance of the
linear regression model within the Arduino system. This local implementation allows for real-time
prediction of solar panel positioning, independent of sensor data availability. By employing the live
output from the local linear regression model, the solar tracking system ensures continuous and
precise adjustments, maintaining optimal solar energy harnessing even in the absence of real-time
sensor feedback.

6.4 Obstacle Detection Using Machine Learning

Innovatively, our group engineered an obstacle detection methodology based on analyzing sudden
power output fluctuations of the solar panel. Employing machine learning techniques, the system
can accurately predict potential obstacles by interpreting irregular power output patterns.
Subsequently, the system dynamically adjusts the solar panel's positioning, enabling it to navigate
around potential obstructions with agility and ensuring consistent energy generation and
operational safety.

10
Block Diagram Overview

The block diagram provided shows the main components of a dual axis solar tracking system
with machine learning improvements. The system consists of the following components:
● Sensors: The sensors are used to track the position of the sun and the solar panels.
The most common sensors used in solar tracking systems are light dependent
resistors (LDRs) and pyranometers.
● Machine learning model: The machine learning model is used to predict the sun's
position and to optimize the tracking algorithm.
● Tracking controller: The tracking controller is responsible for adjusting the position of
the solar panels based on the output of the machine learning model.
● Fallback to ML Model in case of fault: This block is used to ensure that the system
continues to operate even if there is a fault in the main tracking controller.

11
 ● Obstacle Avoidance System: This block is used to detect and avoid obstacles in the
path of the solar panels.
● Exception Handling, if necessary, notify to user: This block is used to handle any
exceptions that may occur in the system. If an exception occurs, the system will
notify the user.
● Data visualizing and monitoring: This block is used to visualize and monitor the
performance of the solar tracking system.
The following shows how machine learning is used to improve the performance of dual axis
solar tracking systems:
● Predicting the sun's position: Machine learning models can be trained on historical
data of the sun's position and weather conditions to predict the sun's position more
accurately. This can be useful for optimizing the tracking algorithm and for
compensating for atmospheric refraction.
● Optimizing the tracking algorithm: Machine learning algorithms can be used to
develop tracking algorithms that are more efficient and accurate than traditional
tracking algorithms. For example, machine learning algorithms can be used to
develop tracking algorithms that can account for the dynamics of the solar panel
system and the effects of wind and other environmental factors.
● Detecting and avoiding obstacles: Machine learning algorithms can be used to
develop systems that can detect and avoid obstacles in the path of the solar panels.
This can be done by training a machine learning model on a dataset of images of
obstacles.
● Diagnosing and troubleshooting faults: Machine learning algorithms can be used to
develop systems that can diagnose and troubleshoot faults in the solar tracking
system. This can be done by training a machine learning model on a dataset of
historical data from the solar tracking system.

12
FLOWCHART OVERVIEW:

These sequences of actions outline a sophisticated system for solar energy management
and data processing. It involves a series of conditional steps to ensure optimal energy
harvesting and data analysis.

13
SEQUENCE:
1. The first step involves checking if the sun is currently shining. If the sun is not out,
the system is programmed to wait until it is. This is fundamental for solar energy
systems, as energy generation is directly dependent on sunlight.
2. Upon confirming that the sun is indeed out, the system proceeds to the next step,
which involves checking if IoT sensors are actively collecting data. These sensors
likely monitor various environmental and solar-related metrics. If they are not
collecting data, the system will notify the user about the necessity of transitioning to
a machine learning redundancy approach.
a. In the event that IoT sensors are not actively collecting data, the system then
utilizes a machine learning model for prediction. This model likely uses the
collected data to make informed forecasts or decisions, such as optimizing
energy generation and consumption.
3. The IoT sensors come into play again after prediction. They are used to measure and
collect additional data, potentially including real-time environmental conditions,
energy production, or other relevant parameters.
4. Once this data is gathered, it is sent to BLDC (Brushless DC) motors for control. BLDC
motors are efficient for controlling various mechanisms, and in this context, they
might be used to adjust solar panels or other components for optimal sun exposure.
5. To make the data accessible and comprehensible, a SaaS cloud service provider is
employed for data visualization. This step allows users to monitor and analyze the
system's performance through user-friendly interfaces, graphs, and charts.
6. The process then repeats to maintain a continuous and dynamic system that adapts
to changing sunlight and environmental conditions. This comprehensive approach
ensures efficient energy harvesting, data-driven decision-making, and user-friendly
data visualization in a solar energy management system.

14
6.The Status/level of work completed:
1. Literature Survey:
a. Reviewed all relevant papers in order to conduct a complete
literature survey on the project.
2. Learned some lessons regarding the current research environment.
3. Component Investigation:
a. Inspecting essential elements of the project.
b. Carefully scrutinized all components in terms of their functions
and operation principles.
4. Budget Analysis:
a. Collected specific details about the budget for the project.
5. Methodology Exploration:
a. Applied various approaches relevant to the task at hand.
b. Weighed the pros and cons of each approach.
6. Model Selection:
a. A desirable model has finally chosen for implementation following
serious deliberations.
7. Mentor Consultation:
a. Informed discussion about the project with tutors.
b. Receive advice from the experience of improving project
management.
8. Progress Documentation:
a. Provided a detailed summary on their progress to be included in a
final report.
9. The appendix included a summary of everything done, observations that
were realized as well as the decisions made.
10.Submitted abstract and report.
11.Final Project Explanation.

15