http://hdl.handle.net/10453/148699 2026-05-17T06:37:43Z http://hdl.handle.net/10453/194984 Title: Pathways to environmental sustainability through energy efficiency: A strategic next energy vision for sustainable development by 2050 Authors: Mujeeb, A; Oladigbolu, J; Bakare, MS; Ibrahim, AA Abstract: As the global push for carbon neutrality accelerates, energy efficiency has become essential for sustainable development, especially for nations like Nigeria that face rising energy demands and significant environmental challenges. This study explores how integrating energy efficiency with carbon neutrality can support Nigeria s strategic energy goals while offering global lessons for other countries facing similar challenges, focusing on key sectors, including industry, transport, and power generation. The study systematically examines the impacts of renewable energy (RE) technologies, like solar, wind, and hydropower alongside policy reforms, technological innovations, and demand-side management strategies to advance energy efficiency in Nigeria. Key findings include the identification of strategic policy frameworks, technological solutions, and the transformative role of green hydrogen in decarbonizing hard-to-electrify sectors. The study also emphasizes the importance of international climate finance, decentralized RE systems like solar mini-grids for improving energy access, and economic opportunities for job creation in the RE sector. Furthermore, it highlights the need for behavioral changes, community engagement, and consistent policy implementation to address infrastructure gaps and drive energy efficiency goals. The novelty of this research lies in its scenario-based analysis of Nigeria s low-carbon transition, detailing both the opportunities and challenges, such as policy inconsistencies, infrastructure deficits, and financial constraints. The findings stress the importance of international collaboration, technological advancements, and targeted investments to overcome these challenges. By offering actionable insights and strategic recommendations, this study provides a roadmap for policymakers, industry stakeholders, and researchers to drive Nigeria towards a sustainable, carbon-neutral future by 2050. 2025-12-01T00:00:00Z http://hdl.handle.net/10453/194961 Title: Enhancing community battery operation via deep learning-based behind-the-meter PV identification Authors: Li, X; Wang, Y; Gao, Y; Alhelou, HH Abstract: The increasing integration of photovoltaic (PV) systems into residential households presents significant challenges for efficient energy management. Our study aims to address these challenges by leveraging deep learning-based behind-the-meter (BIM) PV identification and optimizing community battery operations for enhanced energy management. To achieve this, we first propose a deep learning model (i.e., RCA-BERT) to identify the BTM PV and investigate the load demand patterns through non-intrusive load monitoring (NILM). Specifically, the model is capable of disaggregating meter data to obtain PV generation, relevant PV capacities, and appliance-level load. Afterwards, based on the reliable NILM results, we employ community battery optimization model to reduce expenditures for both consumers and Distribution System Operators (DSOs). The objective of the optimization model is to minimize the electricity cost for a community equipped with PV and a community battery system. To validate our approach, real-world datasets were employed in experimental studies. The results demonstrate that our proposed deep learning model can accurately estimate BTM PV generations across different PV system capacities, achieving an Eacc range of 96.7 to 97.5 . Furthermore, comparative studies were conducted in three distinct scenarios (lowest, median, highest temperature across weekly time horizon within one year) to evaluate the performance of the community battery optimization model. These experiments reveal that our optimized model effectively reduces electricity purchase costs, with savings ranging from 26.7 to 33.9 , depending on the temperature conditions considered. 2025-12-01T00:00:00Z http://hdl.handle.net/10453/194938 Title: Hybrid DPFM and SPWM Control Method for DCM Dual-Leg Integrated Boost Inverter Authors: Feng, S; Lu, DDC; Siwakoti, YP; Hassan, W; Alam, MM; Hu, J Abstract: This paper presents a grid-connected control strategy for a dual-leg integrated-boost (DLIB) inverter that interfaces two renewable sources to the AC grid. A hybrid scheme is adopted: sinusoidal PWM (SPWM) on the H-bridge provides precise grid-current regulation and DC-link control, while dualpulse frequency modulation (DPFM) on the two boost legs enables frequency-controlled MPPT and enlarges the practical discontinuous-conduction-mode (DCM) region. The method achieves independent per-leg power regulation with limited interaction with the grid-side control loops. A small-signal statespace model is developed and used to design a dual-loop SPWM controller consisting of an outer DC-link voltage PI regulator and an inner grid-current PR regulator. DPFM on each boost leg uses a perturb-and-observe (P&O) MPPT, followed by a PI compensator that schedules the switching-frequency setpoint for each leg. PLECS simulations verify stable grid synchronization, accurate current tracking and DC-link regulation, and effective MPPT-based power sharing between the two sources. 2025-01-01T00:00:00Z http://hdl.handle.net/10453/194919 Title: pFedRAG: A Personalized Federated Retrieval-Augmented Generation System with Depth-Adaptive Tiered Embedding Tuning Authors: He, H; Yuan, X; Wu, K; Liu, RP; Ni, W Abstract: Large Language Models (LLMs) can undergo hallucinations in specialized domains, and standard Retrieval-Augmented Generation (RAG) often falters due to general-purpose embeddings ill-suited for domain-specific terminology. Though domain-specific fine-tuning enhances retrieval, centralizing data introduces privacy risks. The use of federated learning (FL) can alleviate this to some extent, but faces challenges of data heterogeneity, poor personalization, and expensive training data generation. We propose pFedRAG, a novel Personalized Federated RAG framework, which enables efficient collaborative fine-tuning of embedding models to address these challenges. The key contribution is a new Depth-Adaptive Tiered Embedding (DATE) architecture, which comprises a Global Shared Layer, combined using FL to capture common knowledge, and a Personalized Layer with adjustable depth tailored for local data and training results of each client. The depth is locally controlled based on crafted metrics and scoring criteria. Also, pFedRAG incorporates a fully client-side pipeline leveraging local small LLMs and vector database filtering to construct high-quality query-document pairs. Experiments on diverse medical non-IID document datasets demonstrate that pFedRAG significantly reduces communication costs, handles data heterogeneity, and improves retrieval performance. Human evaluations confirm the enhanced response quality of pFedRAG. 2025-01-01T00:00:00Z