**2. Method**

The study is designed to compile the relevant contributions from previous publications and to analyze their results in relation to multi-agent modeling design for the energy domain. This study firstly conducts a literature search of ontologies and multi-agent systems for the energy domain. The literature search was performed during the first quarter of 2019. To retrieve the relevant articles for this literature study, four online databases are selected that are relevant in the fields of energy, and MAS and ontologies: ACM digital library, IEEE Xplore, Web of Science, ScienceDirect. The review covers books, conference proceedings, academic journal articles, research articles, and review articles. Other forms of publications, such as newspapers, posters, etc., were not considered since their publication forms are not for scientific research purposes. There was no limitation on the publication years for the literature search.

The data collection was divided into three rounds with relevant keywords. The keyword search was only applied to titles due to a large number of the literature in the fields and the concerns of the relevance in the selected domains. The first round focused on the multi-agent systems in the energy domain. To avoid excluding any relevant study, the search strings were:

('multi-agent' OR 'multiagent') AND ('energy' OR 'electricity' OR 'heating' OR 'grid' OR 'electric' OR 'power' OR 'wind')

The strings, in the first round, resulted in 1433 publications. The result from each database is shown in Table 1. All these 1433 publications were imported to the reference management software-Endnote (https://endnote.com/).


**Table 1.** Results in the first round search.

To dismiss the duplicated publications, i.e., articles which were obtained through multiple databases or strings, 856 articles were removed by this criterion. The remaining 577 articles were selected for further analysis. This study searched the remaining articles with 'ontology' OR 'ontologies' in titles, abstracts, and keywords, and resulted in 24 articles with full-text.

Based on the text mining in the analysis software NVivo (https://www.qsrinternational.com/nvivo/ home) and careful review, the 24 articles were separated into six sub-domains (shown in Table 2). Majority of the selected articles only address one sub-domain, and one article [8] addresses three sub-domains (energy management, microgrid, and buildings), and another article [9] addresses two sub-domains (power system and microgrid). The publications show that the application of ontologies in the field of MAS for the energy domain was mainly conducted after the year 2004, with focus on the sub-domain of grid control between 2004 to 2014, and expanded into the sub-domain of electricity market since 2014. A list of the 24 articles in the Appendix A shows the focused aspects in the energy domain, ontology, and MAS design.

**Table 2.** Six addressed sub-domains by the selected articles.


## **3. Results**

This study reviews and analyses the selected 24 articles to investigate the current research on the application of ontologies in MAS for the energy domain, and the main discussion in the 24 articles can be divided into five categories: 1) definition of agent MAS, 2) MAS applied in energy domains 3) defined ontologies in the energy domain, 4) MAS Design and architectures, and 5) Ontology in the MAS development.
