Session: 01-01: AI for Energy Sustainability I

Paper Number: 124369

124369 - Efficiency-Driven Supervised Learning Regressors in Power Modeling and Optimization of Vertical Axis Wind Turbines 

Abstract: 

Utilizing renewable energy presents a viable option for fulfilling the increasing global need for energy. Of all the clean and sustainable energy sources, wind power stands as the most widely available and employed form. The evident utilization of wind energy across various sectors highlights its extensive application. This has led to a rise in the popularity of smaller wind turbine (WT) installations, including micro-scale ones, especially in distant or urban locations, in addition to the prominent onshore and offshore large-scale WT initiatives.

According to the rotational axis orientation, wind-power utilization equipment can be divided into Horizontal Axis Wind turbines (HAWT) and Vertical Axis Wind Turbines (VAWT).  In HAWTs the wind turbine’s axis of rotation is parallel to the wind direction; while in VAWT the axis of rotation is perpendicular to the wind direction. Besides the conventional HAWT, VAWTs are attracting significant interest, particularly for minor wind farms. As for structural and aerodynamic characteristics, change in wind direction does not affect VAWTs’ performance. Furthermore, due to their simpler mechanical structure, they produce less noise and necessitate lower maintenance. Consequently, they prove to be well-suited for urban settings.

One of the key aspects in studying wind turbines is how the turbine is modeled. The Modeling of VAWTs has developed Analytical models such as blade element, streamtube models which are based on momentum (actuator disk) theory, vortex models, panel models, and lifting line mode. Inaccuracies in these models made Computational Fluid Dynamics (CFD) models more preferred over analytical models for studying VAWTs enabling them to simulate complex fluid flows and more complex blade profiles. Therefore, existing and widely used VAWTs’ blade and structural designs and performance optimization methods include CFD simulation-based strategies and/or wind tunnel experiments, both of which have limitations and challenges. The computational method is costly and time-consuming in and of itself, and wind tunnel facilities are frequently unavailable and expensive. Over the past few years, several studies have highlighted the efficacy of utilizing Machine Learning (ML) algorithms and Artificial Intelligence (AI) in the system design and optimization. The optimization technique based on ML and AI only necessitates a dataset of turbine information. This dataset can be employed to train an ML or neural network model, enabling the estimation of potential output from the VAWTs.

In this study, the primary purpose is to investigate ML algorithms that can be employed to determine and optimize a 12-KW lift base VAWT performance in terms of power coefficient under different combinations of design parameters and operating conditions and draw comparisons between them. In fact, in the current study, a 3-blade lift-based VAWT is modeled using CFD simulation and validated using the available experimental data; then it will be considered as the source of obtaining the data set. Considering the nature of the problem and the dataset, robust nonlinear regression models will be used to model the power coefficient with respect to design parameters. Then the results and accuracies will be compared. The surpassed model, then, will be used for design optimization, as it has been observed that AI, specifically ML algorithms, yields favorable outcomes compared to conventional simulation-based optimization methods. In fact, this step is a major step towards a Digital Twining which is a dataset-based virtual representation of a physical entity that accurately mimics the object as well as reducing the computational costs while maintaining good accuracy.

Presenting Author: Ehsan Dorosti Simon Fraser University (SFU)

Presenting Author Biography: I am a Ph.D. student in Mechatronic System Engineering (MSE) at Simon Fraser University (SFU). I am working on Sustainable Energy systems including Vertical Axis Wind Turbines, their modeling, and optimization.

Authors:

Ehsan Dorosti Simon Fraser University (SFU)
Amir Shabani Simon Fraser University (SFU)
Krishna Vijayaraghavan Simon Fraser University (SFU)