Investigating renewable energy systems using artifcial intelligence techniques

Niroomand, Kamran (2023) Investigating renewable energy systems using artifcial intelligence techniques. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (6MB)

Abstract

This research investigated applying Artificial Intelegence (AI) and Machine Learning (ML) to renewable energy through three studies. The first study characterized and mapped the recent research landscape in the field of AI applications for various renewable energy systems using Natural Language Prcoessing (NLP) and ML models. It considered published documetns at Scopus database in the period (2000-2021). The second study built hybrid Catboost-CNN-LSTM architecture pipeline to predict an industrial-scale biogas plant’s daily biogas production and investigate the feedstock components importance on it. The third study investigated prediciting biogas yield of various subtrates and the significance of each organic component (carbohydrates, proteins, fats/lipids, and legnin) in biogas production using hybrid VAE-XGboost model. The first study showed seven main metatopics and ascent of "deep learning (DL)" as a prominent methodology led to an increase in intricate subjects, including the optimization of power costs and the prediction of wind patterns. Also, a growing utilization of DL approaches for the analysis of renewable energy data, particularly in the context of wind and solar photovoltaic systems. The research themes and trends observed in the first study signify substantial recent investments in advanced AI learning techniques. The developed Catboost-CNN-LSTM pipeline achived a significant results and presented a superior approach when compared to previous relevant studies by eliminating the requirement for feature engineering, enabling direct prediction of biogas yield without the need for converting it into a classification task. The VAE-XGboost pipeline could ovcercome data limitation in the field and produced significant results. It has shown that the "fats" category is the most influential group on the methane production in biogas plants, however, “proteins” illustrated the lowest impact on biogas production.

Actions (login required)

View Item