Control System for Electrical Power Grids with Renewables using Artificial Intelligence Methods

Ibrahim Salem Jahan

Doctoral thesis

Abstract:

Modern electrical and electronic devices are very sensitive to the power supply and require steady and stable electric power. Factories may also need electric power within a specific standard range of voltage, frequency, and current to avoid defects in the production. For these reasons electric power utilities must produce an electric power of a specific standard of power quality parameters [EN50160]. Nowadays, renewable energy sources, such as wind energy and solar energy are used to generate electric power as free and clean power sources as well to reduce fuel consumption and environmental pollution as much as possible. Renewable energy, e.g. wind speed or solar irradiance, are not stable or not constant energies over the time. Therefore smart control systems (SCSs) are needed for operate the power system in optimal way which help for producing a power with good quality from renewable sources. The forecasting and prediction models play a main role in these issues and contribute as the important part of the smart control system (SCS). The main task of the SCS is to keep the generated power equal to the consumed power as well as to consider standard levels of power quality parameters as much as possible. Some of previous studies have focused on forecasting power quality parameters, power load, wind speed and solar irradiance using machine learning models as neural networks, support vector machines, fuzzy sets, and neuro fuzzy. This thesis proposes designing forecasting systems using machine learning techniques in order to be use in control and operate an electrical power system. In this study, design and tested forecasting systems related to the power and renewable energies. These systems include wind speed forecasting, power load forecasting and power quality parameters forecasting. The main part of this thesis is focus in power quality parameters forecasting in short-term, these parameters are: power frequency, magnitude of the supply voltage, total harmonic distortion of voltage (THDu), total harmonic distortion of current (THDi), and short term flicker severity (Pst) according to the definition in [EN50160]. The output of the forecasting models of power quality parameters can be used in shifting the load to run in switch time which will help for correct and optimize the quality of the power.

Abstract:

Modern electrical and electronic devices are very sensitive to the power supply and require steady and stable electric power. Factories may also need electric power within a specific standard range of voltage, frequency, and current to avoid defects in the production. For these reasons electric power utilities must produce an electric power of a specific standard of power quality parameters [EN50160]. Nowadays, renewable energy sources, such as wind energy and solar energy are used to generate electric power as free and clean power sources as well to reduce fuel consumption and environmental pollution as much as possible. Renewable energy, e.g. wind speed or solar irradiance, are not stable or not constant energies over the time. Therefore smart control systems (SCSs) are needed for operate the power system in optimal way which help for producing a power with good quality from renewable sources. The forecasting and prediction models play a main role in these issues and contribute as the important part of the smart control system (SCS). The main task of the SCS is to keep the generated power equal to the consumed power as well as to consider standard levels of power quality parameters as much as possible. Some of previous studies have focused on forecasting power quality parameters, power load, wind speed and solar irradiance using machine learning models as neural networks, support vector machines, fuzzy sets, and neuro fuzzy. This thesis proposes designing forecasting systems using machine learning techniques in order to be use in control and operate an electrical power system. In this study, design and tested forecasting systems related to the power and renewable energies. These systems include wind speed forecasting, power load forecasting and power quality parameters forecasting. The main part of this thesis is focus in power quality parameters forecasting in short-term, these parameters are: power frequency, magnitude of the supply voltage, total harmonic distortion of voltage (THDu), total harmonic distortion of current (THDi), and short term flicker severity (Pst) according to the definition in [EN50160]. The output of the forecasting models of power quality parameters can be used in shifting the load to run in switch time which will help for correct and optimize the quality of the power.

Keywords

Smart power grid, power quality parameters, power load, renewable energies, weather data, decision tree, neural networks, linear regression, support vector machine, forecasting system

Keywords

Smart power grid, power quality parameters, power load, renewable energies, weather data, decision tree, neural networks, linear regression, support vector machine, forecasting system

Language used: English

Date on which the thesis was submitted / produced: 27. 4. 2021

Thesis defence

• Date of defence: 16. 6. 2021
• Supervisor: Stanislav Mišák
• Reader: Břetislav Stacho, Peter Braciník, Lukáš Prokop

VSB – Technical University of Ostrava

Fakulta elektrotechniky a informatiky

Doctoral programme / field:
Elektrotechnika / Elektroenergetika