Smart investigation of artificial intelligence in renewable energy system technologies by natural language processing: Insightful pattern for decision-makers

Abstract

This study aims to provide a framework which enables decision-makers and researchers to identify AI technology patterns in renewable energy systems from a massive data set of textual data. However, the study was challenged by the Scopus database limitation that allows users to retrieve only 2000 documents per query. Therefore, we developed a search engine based on the Scopus Application Programming Interface (API) that enables us to download an unlimited number of documents per query based on our desirable settings. We extracted 5661 renewable energy systems-related publications from Scopus database and leveraged Natural Language Processing (NLP) and unsupervised algorithms to identify the most frequent computational science models and dense meta-topics and investigate their evolution throughout the period 2000-2021. Our findings showed 7 meta-topics based on the class-based Term Frequency-Inverse Document Frequency (c-TD-IDF) score and term score decline graph. Emerging advanced algorithms, such as different deep learning architectures, directly impacted growing meta-topics involving problems with uncertainty and dynamic conditions.