ANN-based UPFC for Power Flow Control

Author: Subigya Khanal

Institution: Institute of Engineering, Tribhuvan University, Nepal

Program: Bachelor of Engineering in Electrical Engineering

Year: 2025

This project presents the design and simulation of an Artificial Neural

Network (ANN) based Unified Power Flow Controller (UPFC) for enhancement

of power flow control and improvement of voltage profile in transmission

systems. The ANN-based controller was developed to replace conventional

PI controllers and demonstrated superior performance in terms of response

time, accuracy, and stability. The system was implemented and tested using

MATLAB/Simulink on a standard IEEE test bus system.

In this project, I modeled and simulated a transmission system in MATLAB/Simulink and analyzed its performance without any controller. I then implemented a conventional PI-based UPFC to observe its impact on system behavior.

After that, I designed and trained an Artificial Neural Network (ANN) using system data and replaced the PI controller with the trained ANN controller. I tested the system under different operating conditions such as load changes and disturbances.

Finally, I analyzed system performance by comparing key parameters such as power flow, voltage profile, and system stability, and concluded that the ANN-based controller provided better overall performance.

Modern power systems face increasing challenges due to the rapid integration

of renewable energy sources such as solar photovoltaics and wind energy.

These sources introduce variability and uncertainty into the grid, making

it harder to maintain stable voltage profiles and balanced power flow across

transmission lines.

A Unified Power Flow Controller (UPFC) is one of the most powerful FACTS

(Flexible AC Transmission Systems) devices available to address these

challenges. However, conventional controllers used in UPFC systems are

often slow to respond to dynamic changes and require precise system

modeling.

This project explores the use of Artificial Neural Networks (ANN) as an

intelligent alternative controller for UPFC systems, enabling faster and

more accurate power flow regulation without requiring an exact mathematical

model of the system.

This work is directly relevant to emerging challenges in integrating

solar PV systems and electric vehicle charging loads into transmission

networks, areas that are central to the mission of the

Free Appropriate Sustainability Technology research group.

To design an ANN-based controller for a Unified Power Flow Controller

To improve active and reactive power flow control in transmission systems

To improve voltage profile stability under varying load conditions

To compare ANN controller performance with conventional PI controllers

To simulate and validate the system using MATLAB/Simulink

A Unified Power Flow Controller (UPFC) is a power electronics device that

can simultaneously control active power, reactive power, and voltage

magnitude in a transmission line. It combines the functions of a STATCOM

and a SSSC connected through a common DC link.

An Artificial Neural Network (ANN) is a computational model inspired by

the human brain. In this project, a feedforward ANN was trained using

system input-output data to learn the optimal control signals for the UPFC

under different operating conditions. Once trained, the ANN can produce

fast and accurate control outputs without needing to solve complex

mathematical equations in real time.

Modeled a standard IEEE transmission test system in MATLAB/Simulink

Designed a conventional UPFC with PI controllers as a baseline

Collected training data by running the system under various load and

generation conditions

Designed and trained a feedforward ANN using the collected data

Replaced the conventional PI controller with the trained ANN controller

Ran simulations under multiple test scenarios including normal operation,

sudden load changes, and fault conditions

Compared results between the conventional and ANN-based controllers

The ANN-based UPFC controller showed faster response time compared to

the conventional PI controller

Voltage profile improved significantly under varying load conditions

Active and reactive power flow was maintained closer to reference values

The ANN controller remained stable under sudden disturbances where the

PI controller showed oscillations

Overall system performance improved in terms of settling time and

steady-state error

MATLAB / Simulink

The integration of intelligent controllers like ANN into power systems has

direct relevance to sustainable energy goals:

As more solar PV systems connect to the grid, transmission networks need

smarter and faster control to manage the variability

ANN-based controllers can be trained on real grid data, making them

adaptable to different network configurations without expensive remodeling

The same neural network approach used in this project has been applied in

recent research on machine learning models for electric vehicle charging

and national-level solar photovoltaic planning

Open-source implementation of such controllers could help developing

regions like Nepal manage their growing renewable energy integration

challenges at low cost

Extend the simulation to larger IEEE bus systems

Test the ANN controller under renewable energy variability scenarios

including solar PV fluctuations

Explore real-time hardware implementation using open-source platforms

such as Arduino or Raspberry Pi

Investigate application of the controller for EV charging station

integration into distribution networks

Subigya Khanal

B.E. Electrical Engineering

Institute of Engineering, Tribhuvan University, Nepal

Email: subigyakhanal345@gmail.com

Appropedia: User:Subigya Khanal