Distribution system state estimation (DSSE) is a valuable step for DSOs toward tackling the challenges of transitioning to a more sustainable energy system and the evolution and proliferation of electric cars and power electronic devices. However, on the LV level, implementation has only taken place in a few pilot projects. In this paper, an LV DSSE method is presented and implemented in four real Hungarian LV supply areas, according to well-defined scenarios. Pseudo-measurement datasets are generated from AACs and SLPs, which have been used in different combinations on networks built with different accuracies in terms of load placement. The paper focuses on the critical aspects of finding accurate and coherent information on network topology with automated management of information systems, real LV network implementation for power flow calculation and managing portions of the network characterized by uncertain or inconsistent line lengths. A refining algorithm is implemented for the integrated network information system (INIS) models. The published method estimates node voltages with a relative error of less than 1% when using AACs, and a meter-placement method to reduce the maximum value of relative errors in future scenarios is also presented. It is shown that the observation of node voltages can be improved with the usage of AACs and SLPs, and with optimal meter placement.

Effective infrastructure monitoring is a priority in all technical fields in this century. In high-voltage transmission networks, line inspection is one such task. Fault detection of insulators is crucial, and object detection algorithms can handle this problem. This work presents a comparison of You Only Look Once architectures. The different subtypes of the last three generations (v3, v4, and v5) are compared in terms of losses, precision, recall, and mean average precision on an open-source, augmented dataset of normal and defective insulators from the State Grid Corporation of China. The primary focus of this work is a comprehensive subtype analysis, providing a useful resource for academics and industry professionals involved in insulator detection and surveillance projects. This study aims to enhance the monitoring of insulator health and maintenance for industries relying on power grid stability. YOLOv5 subtypes are found to be the most suitable for this computer vision task, considering their mean average precision, which ranges between 98.1 and 99.0%, and a frame per second rate between 27.1 and 212.8, depending on the architecture size. While their predecessors are faster, they are less accurate. It is also discovered that, for all generations, normal-sized and large architectures generally demonstrate better accuracy. However, small architectures are noted for their significantly faster processing speeds.

Automated license plate recognition (LPR) systems have garnered substantial attention within the field of intelligent transportation systems, owing to their pivotal role in facilitating toll collection, enhancing traffic management, and ensuring operational efficiency. Despite recent breakthroughs in convolutional and recurrent neural network architectures, Polish LPR remains underexplored, with most existing approaches relying on conventional optical character recognition. This study proposes a hybrid convolutional neural network – recurrent neural network (CNN-RNN) model equipped with a Thin-Plate Spline (TPS) transformation module, a ResNet-based feature extractor, a bidirectional Long Short-Term Memory (LSTM) sequence model, and an attention-based decoder to address the unique challenges of Polish license plates. The model is trained on a high-difficulty dataset, comprising real-world images without explicit character-level bounding boxes. Empirical evaluations underscore the efficacy of the proposed system, with competitive accuracy and normalized edit distance scores achieved on Polish, Czech, Hungarian, and Slovak datasets. Additionally, transfer learning from closely related Central European plate formats to Polish data demonstrates marked improvements in convergence and overall performance. Further validation on a challenging video-based dataset reveals the robustness of the proposed approach, evidencing its potential applicability in real-world scenarios and highlighting majority voting as an effective strategy to enhance system reliability under variable conditions.

Soil fauna is crucial for carbon cycling, controlling organic matter decomposition and contributing to ecosystem services ^1–7 . Soil microarthropods are top-down regulators in the decomposer food web and serve as fundamental indicators of soil health. Yet, routine field-level monitoring remains restricted to resource-intensive research, leaving a critical gap in arable land management and policy. To address this, we developed Edapholog ^® extractor, a fully automated laboratory device that detects and identifies live soil microarthropods via real-time, AI-based image analysis, enabling taxonomic identification. Here, we demonstrate its accuracy across 319 arable fields spanning ten European countries. We show that computer vision can support ecological interpretation in arable systems and long-term studies on conservation tillage and cover cropping, offering a scalable tool for integrating soil biodiversity metrics into regenerative agriculture and carbon farming. We validated the system against classical taxonomy and found that across ∼35,000 microarthropod individuals, the device achieved an overall accuracy of 86%, sensitivity of 75%, and specificity of 99% compared to manual identifications. Community composition analyses revealed high similarities (83%), with minimal richness differences (7%) and low species replacement (13%) across countries, indicating that the AI does not introduce taxonomic bias. When applied in a long-term field experiment, the system detected significant taxon-specific responses to conservation tillage, with effect sizes ranging from 0.5 to 4. Total abundance, richness, and a soil biological health index were 39%, 47%, and 150% higher, respectively, under conservation tillage compared to conventional ploughing. These effects were statistically consistent between the automated and classical methods. However, while manual microscopy required several hours per sample, the AI-based system delivered immediate results without the need for taxonomic expertise. Edapholog ^® extractor offers exciting opportunities for rapid, scalable soil biodiversity monitoring for future sustainable land management.

Soil invertebrates play a key role in ecosystem functioning, yet their monitoring remains constrained by labour‐intensive extraction and taxonomic identification procedures. Consequently, large‐scale and high‐frequency assessments of soil biodiversity remain difficult despite increasing demand for standardized monitoring tools. Here, we present an AI‐driven soil extractor that combines classical dry extraction with real‐time computer vision and deep‐learning‐based classification of soil microarthropods. Organisms are detected and classified while moving through the device, allowing automated quantification of taxon densities while individuals remain alive and available for subsequent analyses. System performance was evaluated using a validation framework that separately assessed AI classification accuracy, the integrated detection and transport workflow, and extraction performance relative to conventional Berlese funnels. Automated image classification achieved an overall accuracy of 80.9% (95% CI: 79.9%–81.8%; Cohen's κ = 0.767) when compared with expert image identification. Automated detections and manual identification of the same extracts showed high agreement in community composition, with a median Bray–Curtis similarity of 0.80, indicating limited mechanical bias introduced by specimen transport. Abundance estimates obtained with the automated system closely matched those produced by conventional Berlese extraction, with only minor taxon‐specific differences. Controlled experiments further demonstrated high mechanical reliability, with detection success exceeding 97% and transport success exceeding 92%. The system provides a scalable approach for quantitative monitoring of soil microarthropod communities while preserving live specimens for downstream analyses. By integrating extraction, image acquisition, classification and specimen collection into a single workflow, the AI‐driven soil extractor offers a practical tool for high‐throughput soil mesofauna monitoring.