Voltage Control In Smart Distribution Network With Limited Reactive Power Capacity Of Smart Inverter Using Machine Learning - Master’s thesis in Electrical Engineering

Aug 5, 2026

Upload by Eden

Maintaining consumer voltages within specified limits is essential, yet grid operators may struggle when renewable generation increases operational uncertainty. This thesis addresses voltage quality challenges by using inverter-based control instead of relying solely on traditional grid voltage regulation. It analyzes how inverter reactive power capability depends on loading conditions and inverter size, including coordinated control under network modeling constraints. Using grid and inverter data, the work builds machine learning models, with reinforcement learning to adjust reactive power to correct voltages at adjacent buses, keeping even far-end bus voltages within limits under overload. Reactive power control is shown to be more robust and aggressive than a conventional OPF approach.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages189

Reading Time7 h 56 min

AccessFree to download

1 Introduction

1.1 Background

1.2 Motivation

1.3 Problem Statement

1.4 Objective

1.5 Scope of Thesis

1.6 Limitations and Assumptions

2 Literature Review

2.1 Traditional Voltage Regulation

2.2 New Approaches for Voltage Regulation

Faculty of Electrical Engineering Voltage Control In Smart Distribution Network With Limited Reactive Power Capacity Of Smart Inverter Using Machine Learning Syed Muhammad Aizaz Haider Master’s thesis in Electrical Engineering ELE-3900 May 2023 “The person who taught me one single word, made me his/her servant for rest of my life”. IMAM ALI (SALAWAT) Dedicated to BABUL ILAM, MAULA ALI (SALAWAT), Preface This master's thesis was the capstone project for the electrical engineering program at UiT, The Arctic University of Norway, in Narvik, Norway. My supervisors Professor Pawan Sharma and his Ph.D. colleague Raju Wagle have been an excellent supervisor and advisee in this project. Specially Mr. Raju, who helped me even in his holidays and without his kind help and constant guidance, this thesis would have not been possible to finish. It was mr. Raju Wagle, who suggested this thesis topic and motivated me to work in the domain if machine learning and power systems. I'd also want to express my gratitude to my friend Hammad Tariq, who helped me get started with PandaPower and taught me the how to work in that. Finally, I appreciate everyone in my family specially my parents and siblings for always being there for me. I want to specially thank my son, Syed Muhammad Aoun Akbar, whose smile is a constant source of motivation for me. I'd want to express my gratitude to UiT for providing me with an enriching educational experience, and to my seniors for keeping me motivated and allowing me to enjoy my time with them and learn from them. I also want to express my gratitude to Ventum Dynamics AS for encouraging me to complete my Master’s degree. Syed Muhammad Aizaz Haider 19th May, 2023 Abstract It is very important for consumers to have the operating voltages within the specified limit. Sometimes, this is not possible to provide by the grid operators although grid operators invest heavily in voltage regulating devices. With the integration of variable renewable energy systems in the grid, it has become more challenging for the grid companies to regulate them to provide stable voltages and good quality power. However, the control of inverters in renewable energy sources can be utilized for overcoming such challenges. Another aspect that should be considered is analyzing the capacity of the inverter to provide reactive power support to mitigate the voltage violation in the power system network. The amount of reactive power support from inverters is dependent on different loading conditions and the size of the inverter. Hence, the analysis of such scenarios is of the great significance of renewable energy sources and load, modeling of the network for implementing coordinated control from such inverters is another challenge. Hence, with available data from the grid and inverters, it is possible to model the system using machine learning (ML) algorithms. From the model developed using ML, reactive power support requirement from inverters, analysis of the impact of reactive power and active power from inverters and change of loading in the power system network can be analyzed. Reinforced machine learning is implemented in this work where the inverter learns to correct its action to adjust the voltages at the adjacent bus by controlling the flow of reactive power into the distribution network based on data available for the next stage. The voltages obtained on the far most bus were also within the specified limit even for the overloading condition. And the reactive power control was more robust and aggressive to keep the voltages within the specified limit as compared to the conventional Optimal Power Flow OPF tool. Contents Abstract ...................................................................................................................................... 4 List of Figures ............................................................................................................................ 8

FAQs

Why is inverter control needed for voltage regulation in smart distribution networks?

Grid operators may not always maintain consumer voltages within limits, especially after integrating variable renewable energy. Inverter control helps overcome these challenges by adjusting power flows to stabilize voltages.

How does the thesis treat inverter reactive power capacity?

It considers that reactive power support depends on different loading conditions and the inverter size. This makes scenario analysis crucial for effective coordinated control.

What machine learning approach is used and what outcome is reported?

The work uses reinforced machine learning so the inverter learns to correct its actions for voltage adjustment via reactive power control. Voltages at the farthest buses stay within limits even under overloading, with more robust and aggressive reactive power control than conventional OPF.

Aug 5, 2026

Upload by Eden