Sustainable Energy and Artificial Intelligence

Abstract Frequency stability is the fundamental problem of keeping the PS's frequency within a reasonable range. The PS's stability would be threatened by frequency variation brought on by load disruptions and uncertainty pertaining to PS characteristics. The PS's LFC system is crucial for controlling and maintaining frequency stability. This work uses the adaptive Fuzzy-FOPID (FFOPID) controller in the LFC structure to address uncertainties and disturbances while improving frequency stability in a two-area PS (TAPS). Additionally, RL has been used to enhance the adaptive FFOPID controller's performance. The adaptive FFOPID controller's performance against the uncertainties and disruptions of the TAPS has been enhanced by RL, which has also increased the control system's environmental adaptability. In a number of scenarios, the suggested control (FFOPID-RL) method has been compared to FFOPID, PDSMC, and SMC control methods. The findings indicate that the control method is better at lowering frequency deviations, settling times, and damping frequency oscillations associated with the TAPS

Abstract The deployment of AI-enabled battery digital twins for battery management systems (BMS) and DC fast-charging control is often approached as a purely technical task focused on improving prediction accuracy and charging speed. This study argues that such deployments are inherently socio-technical, shaped by boundary judgments about system purpose, beneficiaries, measures of success, decision authority, resources, expertise, and legitimacy. Using Critical Systems Heuristics (CSH), we conducted semi-structured interviews with 22 experts and stakeholders spanning BMS/battery engineering, fast-charging operations, system integration, and safety/regulatory perspectives. Interview transcripts were analyzed using template analysis aligned with the four CSH dimensions (motivation, control, expertise, and legitimacy) to contrast “what is” versus “what ought to be” boundary judgments. Results show that current practice is predominantly throughput-driven, with battery longevity, thermal safety, and grid impacts treated as secondary constraints; decision authority is fragmented across vendors, OEM-side BMS logic, operators, and utilities; and resource allocation underinvests in calibration, data governance, cybersecurity, and continuous validation. The boundary critique was translated into a structured set of CSH-informed design requirements, emphasizing health-aware charging control, explicit safety envelopes and fail-safe modes, calibrated sensing and data-quality gates, drift detection and continuous validation, auditable decision logging, and multi-stakeholder governance and representation mechanisms. The study contributes a boundary-aware pathway for designing and governing trustworthy digital-twin-enabled fast charging beyond algorithmic performance.

Abstract This paper presents a dual-mode artificial intelligence-based control approach for robust regulation of DC-DC two-switch forward converters under wide input voltage fluctuations and dynamic load conditions. Two neural network-driven strategies are explored: (1) an adaptive PI controller whose gains are optimized offline via a genetic algorithm and predicted using a feedforward neural network, and (2) an online neural network controller that directly computes the duty cycle in real time using online learning. Extensive simulation studies compare these neural approaches against classical PI and adaptive fuzzy controllers using various quantitative performance metrics, including steady-state error, overshoot, peak time, settling time, and integral error indices (IAE, ISE, ITAE). Simulation results under nominal conditions and multiple disturbance scenarios—including sudden load changes and input voltage drops—demonstrate that both AI-based approaches outperform conventional PI and fuzzy controllers. While the offline-trained adaptive PI controller excels in nominal conditions with minimal steady-state error and overshoot, the online neural network controller offers superior dynamic adaptability and robustness in real-time scenarios. Comparative performance metrics and radar-based multi-criteria evaluations confirm the suitability of these intelligent control strategies for real-time power electronics applications requiring high precision and resilience

Abstract The accelerating integration of artificial intelligence (AI) into sustainable energy policy presents both transformative opportunities and complex challenges for achieving equity and innovation in the global energy transition. While AI-driven frameworks promise enhanced efficiency, predictive analytics, and optimized resource allocation across renewable energy systems, prevailing policy approaches often overlook the underlying boundary judgments, stakeholder inclusivity, and socio-technical complexities inherent in these transitions. This study applies Critical Systems Heuristics (CSH) to systematically examine the assumptions, values, and power dynamics embedded within AI-assisted policy making for sustainable energy. Employing a qualitative research design that combines a systematic literature review with expert interviews from policy, technology, and community sectors, the analysis utilizes the twelve CSH boundary questions to contrast current “is” framings with normative “ought” expectations for just and innovative energy governance. Findings reveal that dominant AI-enabled policies frequently prioritize technical optimization and economic growth, with limited mechanisms for participatory decision-making, transparency, or recognition of marginalized communities. In contrast, stakeholders advocate for policy frameworks that explicitly address distributive and procedural justice, foster adaptive learning, ensure algorithmic transparency, and embed equity as a core design principle. The study proposes a CSH-informed conceptual framework to bridge these gaps highlighting pathways for inclusive stakeholder engagement, ethical AI deployment, and systemic innovation in sustainable energy governance. The results underscore the necessity of integrating critical systems thinking with digital innovation to advance socially legitimate and resilient energy transitions.

Abstract The pharmaceutical supply chain is a critical and energy-intensive network, particularly due to temperature-sensitive products requiring continuous refrigeration. Ensuring product quality and safety while minimizing energy consumption has become an essential challenge. Despite growing interest in digital technologies for pharmaceutical logistics, a structured synthesis of AI-driven energy optimization and IoT-enabled sustainability in cold chains remains lacking. This review systematically analyzes the application of the Internet of Things (IoT), Artificial Intelligence (AI), and complementary digital technologies in enhancing energy efficiency, operational performance, and sustainability of pharmaceutical cold chains. A total of 76 peer-reviewed studies—including empirical investigations, case studies, architecture proposals, optimization frameworks, and conceptual models—were examined to identify technological trends, methodological approaches, and practical outcomes. Findings highlight how IoT-enabled sensors, Radio Frequency Identification (RFID) systems, and cloud-based platforms facilitate real-time monitoring, intelligent inventory management, and predictive decision-making. Integration with AI supports demand forecasting, anomaly detection, and energy-aware cold-chain optimization, while blockchain ensures secure, traceable, and tamper-resistant recording of events. Results indicate that digital solutions can reduce energy consumption, minimize waste, improve regulatory compliance, and enhance operational resilience. Challenges such as cybersecurity, standardization gaps, high implementation costs, and organizational barriers remain critical considerations. By synthesizing technological innovations, operational strategies, and sustainability implications, this review provides a structured roadmap for AI- and IoT-enabled energy-efficient supply chain management, offering actionable guidance for researchers, policymakers, and industry practitioners seeking sustainable and smart pharmaceutical logistics solutions.

Abstract Facial expressions constitute one of the most important forms of non-verbal communication, enabling humans to convey emotions and establish effective interpersonal interactions. In recent years, facial emotion recognition (FER) has significantly benefited from deep learning techniques, particularly architectures capable of modeling both spatial and temporal information in video sequences. Motivated by the dynamic nature of facial expressions, this study proposes and evaluates two hybrid deep learning frameworks for video-based emotion recognition: GRBM-ConvLSTM2D and GRBM-3DCNN. In both configurations, a Gaussian Restricted Boltzmann Machine (GRBM) is employed as a preliminary feature extraction layer to enhance spatiotemporal representation learning. Implemented in Python, the proposed models are designed to classify seven emotional categories: surprise, happiness, anger, sadness, fear, disgust, and contempt. The experimental evaluation was conducted on the benchmark CK+ dataset. The results demonstrate that both proposed architectures achieve strong recognition performance, with the GRBM-ConvLSTM2D model attaining a validation accuracy of 91.16% and the GRBM-3DCNN model achieving 88.80%, confirming the effectiveness and robustness of the proposed hybrid frameworks for video-based facial emotion recognition.