Zhiyu Pan

Energy-domain tables often face significant challenges in semantic interoperability because semantically related concepts may be represented with inconsistent terminology, units, and structural formats across data sources. Large language models provide a promising solution due to their strong contextual reasoning ability, but their predictions can vary across models and simple aggregation strategies are often insufficient to ensure robust ontology-guided annotation. In this paper, we propose a multi-LLM framework for ontology-based semantic annotation of energy-domain tables. The framework combines ontology preprocessing, multi-model candidate generation, and a confidence-aware aggregation mechanism, termed Multi-LLM Soft Voting (MSV), to improve the reliability of semantic prediction. It supports both CTA-only setting and joint CTA+CPA setting, and further investigates how different ontology representations affect annotation quality and computational efficiency. We evaluate the proposed framework on real energy-domain tables using multiple commercial LLMs. The experimental results show that MSV consistently improves annotation performance over the baseline, while structured ontology input further reduces both computational time and LLM token cost.

We present our development of a co-simulation scenario ontology that provides a semantic framework for describing, sharing, comparing, and executing simulation scenarios, thereby enhancing interoperability and reproducibility.The ontology is created through a structured development process following the Linked Open Terms methodology.

Data anonymization is essential for data privacy, enabling organizations, especially in the energy sector with IoT data, to comply with regulations while using sensitive information. Despite many existing anonymization methods and tools, it lacks clarity on integrating these tools into a complete process and ensuring effective collaboration with data providers. To overcome these gaps, this paper proposes a collaborative data anonymization framework that efficiently chains open-source tools to streamline the process and improve data provider involvement. The findings indicates while these tools can detect sensitive information and generate anonymizeddata, there are still limitations in metadata detection.

This paper examines artificial intelligence (AI) bias in power systems applications through systematic analysis of three critical use cases: load forecasting, predictive maintenance, and ontology matching for system interoperability. While AI solutions show great potential for addressing complex power system challenges, they face adoption barriers due to biases that compromise fairness, reliability, and operational performance. Our investigation demonstrates how different bias types—including data representation, algorithmic, and sampling biases—manifest in power systems contexts, directly affecting grid efficiency, resource allocation, and socioeconomic equity across the electrical power and energy domain. For each use case, we provide quantitative evidence of bias impact and propose targeted mitigation strategies that emphasize data diversity, ensemble methods, explainable AI techniques, and fairness-aware algorithms. By establishing a comprehensive taxonomy of bias types relevant to power systems and developing practical mitigation frameworks, this work bridges the critical gap between abstract bias concepts and real-world power system applications. The resulting framework provides a structured approach for developing equitable, robust AI systems that align with power systems’ operational requirements while accelerating the responsible adoption of AI in safety-critical infrastructure.

Data integration in the energy sector, which refers to the process of combining and harmonizing data from multiple heterogeneous sources, is becoming increasingly difficult due to the growing volume of heterogeneous data. Schema matching plays a crucial role in this process by giving each representation a unique identity by matching raw energy data to a generic data model. This study uses an energy domain language model to automate schema matching, reducing manual effort in integrating heterogeneous data. We developed two energy domain language models, Energy BERT and Energy Sentence Bert, and trained them using an open-source scientific corpus. The comparison of the developed models with the baseline model using real-life energy domain data shows that Energy BERT and Energy Sentence Bert models significantly improve the accuracy of schema matching.

With the ongoing digital transformation and multi-domain interaction occurring in the buildings, a huge amount of heterogeneous data is generated and stored on a daily basis. To take advantage of the gathered data and help better decision makings, suitable methods are needed to meet the demand for building operations and reinvestment planning. Ontology, which provides not only the vocabulary of a certain domain but also the relationship between each other has been used in multiple engineering fields to manage heterogeneous data. A plethora of ontology development methodologies have been developed in the last decade, whereas those methods are still really time-consuming and in a low degree of automation. In this paper, we approach the problem by first presenting a semi-automatic ontology development framework that integrates existing automatic ontology tools and reuses existing ontology and data model. Based on this framework, we create a building energy management ontology and evaluate the data coverage of several real-life data sets.

Journal of Advanced Management Science (JOAMS)