Introduction
The global shift towards sustainable energy has placed wind power at the forefront of renewable energy generation. Efficient wind farm operation is crucial for maximizing output and economic viability, with accurate power prediction playing a key role. Supervisory Control and Data Acquisition (SCADA) systems in wind farms collect high-resolution data on turbine performance and environmental conditions, offering opportunities for advanced analysis. Machine Learning (ML) and Deep Learning (DL) models are effective in capturing the complex relationships in this data. This thesis explores the use of Machine Learning (ML) and Deep Learning (DL) to model
the complex relationships in SCADA data and improve power predictions. By evaluating various ML and DL models, the research aims to pinpoint the most effective strategies for reliable and accurate power forecasting.
Motivation
Accurate wind power forecasting is crucial for stakeholders like wind farm operators, grid managers, and energy traders. Traditional methods often struggle to address the complex patterns in SCADA data. Although ML and DL models are increasingly used, there is a lack of comprehensive studies comparing different models consistently. Understanding model performance is key for choosing the best forecasting tools. Moreover, as SCADA data grows
in volume and complexity, evaluating the scalability and efficiency of these models becomes essential. This research aims to address these gaps, offering insights to improve efficiency and decision-making in wind farm management.
Goal
This MSc thesis aims to evaluate the effectiveness of various ML and DL models in predicting wind farm power production from SCADA data. The study covers several key areas including data preprocessing and feature selection, where SCADA data is processed to enhance quality and impactful features for power production areselected. It also involves model development, where models are created using various ML techniques and DL models. A comparative analysis is conducted to determine the models’ accuracy, computational efficiency, scalability, and ease of implementation. Finally, the thesis offers recommendations on the best models for wind
power forecasting.
Expected Outcomes
Upon completion of this study, the expected outcomes are:
• Enhanced Understanding of SCADA Data: A detailed analysis of SCADA data, leading to insights into the most influential factors affecting wind farm power production.
• Developed Predictive Models: A suite of ML and DL models developed and optimized for wind power prediction.
• Performance Benchmarking: A comprehensive evaluation of each model’s performance, highlighting strengths and weaknesses.
Special entry requirements
Advanced proficiency in Python programming, Proficient in ML and DL concepts, Good proficiency in Linux.
Supervisor
Hamidreza Abedi, Associate Professor, Div. of Safety and Transport, Dept. of Electrification and Reliability, Unit Renewable Energy Systems, RISE Research Institutes of Sweden (hamidreza.abedi@ri.se)
Educational area
Computer engineering, Mechanical and Industrial design engineering, Physics, Mathematics and Environment
Location
Gothenburg
Credits
30-60 hp (1-2 students). The compensation will be 30 000 SEK upon completion of a high-quality thesis.
Welcome with your application
Candidates are encouraged to send in their application as soon as possible. Suitable applicants will be interviewed as applications are received. Last day of application is November 12, 2024.
