Objectives: This study aims to improve the accuracy and consistency of Named Entity Recognition (NER) systems by addressing the limitations of transformer-based models that rely on independent token-level predictions. It focuses on exploring how structured prediction techniques can enhance sequence labeling performance. Methods: A comparative evaluation is carried out using three hybrid architectures: BERT combined with CRF, BiLSTM combined with CRF, and BERT integrated with a Transformer layer. All models are tested within a consistent experimental setup using the CoNLL-2003 dataset. The methodology emphasizes combining contextual representations with sequence-level optimization through Conditional Random Fields. Findings: The results indicate that the BERT + CRF model delivers the highest performance, achieving an F1-score of 0.94. In comparison, BERT + Transformer attains 0.81, while BiLSTM + CRF reaches 0.72. These outcomes demonstrate that incorporating CRF significantly improves label sequence coherence and overall extraction effectiveness. Novelty: The work presents a unified and systematic comparison of hybrid NER models, highlighting the importance of structured decoding in modern architectures. It offers valuable insights into balancing model performance and computational efficiency, supporting the development of reliable and scalable NER solutions for practical applications. The other NER models often fail to classify multi-token organization entities and entities with ambiguous contextual meaning correctly. The BERT + CRF model got better boundary detection by structured sequence optimization.
Accurate wind resource assessment is a fundamental requirement for wind farm planning, turbine selection, and reliable estimation of long-term energy production. Mesoscale wind atlas frameworks, particularly the New European Wind Atlas (NEWA) and the Global Wind Atlas (GWA), are increasingly used to derive site-specific wind climatology and support preliminary feasibility studies. However, their predictive performance is strongly influenced by terrain complexity, surface roughness, and atmospheric flow conditions. In this study, a systematic and comparative assessment of wind resource characterization and annual energy yield estimation is conducted using NEWA and GWA across four representative environments: forested terrain, mountainous terrain, offshore marine conditions, and normal inland terrain. Height-dependent wind statistics were extracted from both atlases and processed using Rayleigh and Weibull parameterization, Hellmann power-law extrapolation, and probabilistic turbine power curve integration to estimate hub-height wind conditions and annual energy yields. The resulting predictions were validated against reference datasets from the Bavarian Energy Atlas and official offshore production data, and relative errors were quantified to evaluate model reliability. In parallel, a bibliometric analysis of peer-reviewed literature published between 2002 and 2026 was performed to examine research trends, thematic evolution, and methodological developments related to wind atlas applications and wind energy yield assessment, providing a broader scientific context for the study. The results indicate that both atlases perform well under homogeneous inland and offshore conditions, while prediction errors increase in forested and mountainous regions due to unresolved microscale flow effects and enhanced surface roughness. The findings highlight the terrain-dependent strengths and limitations of NEWA and GWA and emphasize the need for hybrid correction approaches and site-specific calibration to improve prediction accuracy. Overall, the study provides a validated methodological framework for the application of wind atlas data in technical and pre-investment wind energy assessments.
This paper proposes an advanced neural network-based Modelling and control approach for an 8 MW grid-connected photovoltaic (PV) system, leveraging real-world irradiance and temperature data to enhance power extraction efficiency. Traditional maximum power point tracking (MPPT) methods, while effective, exhibit limitations in dynamic response and accuracy under rapidly varying environmental conditions. To address this, a The LSTM-based NN (LMST-NN) is trained on historical solar irradiance and temperature datasets to predict the optimal operating points of the PV array, replacing conventional MPPT algorithms. The system architecture comprises a PV array, a DC-DC boost converter (elevating voltage to 11 kV), and a three-phase inverter synchronized with the utility grid using a phase-locked loop (PLL). The neural network’s predictions dynamically adjust the boost converter’s duty cycle, ensuring maximum power transfer under fluctuating conditions. Simulation in MATLAB/Simulink validates the system’s performance, demonstrating superior prediction accuracy compared to traditional techniques. Additionally, the integration of an LCL filter maintains total harmonic distortion (THD) below 1%, complying with IEEE 519 standards. Key metrics, including grid synchronization stability (frequency deviation < 0.05 Hz), converter efficiency (97.5%), and transient response, are rigorously evaluated. The results highlight the neural network’s robustness in optimizing power generation while ensuring seamless grid integration, offering a promising alternative to conventional MPPT for large-scale PV systems.
Driver fatigue while travelling for large distances and at night is the most common cause of accidents because of the driver's loss of attention or slow reaction. In order to solve this problem, it will be necessary to create a system that would be able to check the state of the driver's health and react accordingly. In this paper, an algorithm that would be capable of detecting the level of sleep of the driver and make lane switching has been proposed. A computer camera would constantly be monitoring the driver's eyesight. Due to this technology, the driver's eyes would be under control, and his/her fatigued condition can be detected in order to warn the driver about sound and lights. In case the driver ignores the warning, this technology would be activated by switching to the lane where the process of control is performed by the Arduino Uno microcontroller.
In today’s time, rivers & lakes are becoming more polluted. Day by day it increasing due to plastic waste, floating materials, improper disposal of waste & soluble impurities also. This problem becomes more serious during big spiritual events like Mahakumbh or other Utsav’s, where so many peoples are gathered & huge amount waste is generated in huge amount. To solve this problem, we have designed an automatic solar power surveillance boat with waste segregator which works on Sun’s energy i.e., solar energy (sunlight). The boat has a solar panel, so overall system works on it. It also collects floating waste from the rivers or any other water surface & store it in a container which is mounted on the boat. In Addition to this, the main Feature of this project is waste segregation, where waste is Seperateded on the basic of Biodegradable & non- Biodegradable Materials. Our project mainly addresses the improper disposal of waste, illegal dumping, reduce the human involvement & improve durability. It Merges, conveyor Mechanism, Segregation unit, Bluetooth Control & Microcontroller – Based Automation for Efficient Waste Management. The Boat is Controlled using Arduino .it can work in both automatic & remote mode. the main aim of our project is to clean & Sort Water bodies & promote renewable energy & facilitate renewable purchase obligations.
