Zhipeng Yuan

Accurate forecasting of building cooling loads in building energy management systems is vital for effective building energy management. In recent years, various Transformer-based deep neural network architecture have been introduced for load forecasting. However, existing approaches often fail to adequately account for the effects of passenger-flow variations on cooling demand, and struggle to jointly capture short-term fluctuations and long-term temporal regularities within a unified modeling framework, which ultimately limits forecasting accuracy. To address this challenge, this study investigates the impacts of meteorological conditions and airport passenger flow on the cooling load of airport terminals and a novel dual-branch hybrid approach based on Dual-Branch Time Series Forecasting (DBTSF) is proposed for airport terminal cooling load forecasting (CLF). Inspired by the iTransformer architecture and the Temporal Convolutional Network model, the proposed model synergistically integrates Temporal Convolutional Network (TCN) with an Improved iTransformer , namely Rckanformer, to capture both local temporal patterns and long-range dependencies. For the Rckanformer, we propose a convolutional autocorrelation attention mechanism to replace the traditional self-attention mechanism, strengthening the modeling of periodicity and long-term correlations in cooling load series. Then, a Kolmogorov–Arnold Networks (KAN) -based feed-forward component is introduced to replace the standard multilayer perceptron (MLP), enhancing the representation of complex nonlinear relationships in high-dimensional inputs. Furthermore, the Synchrosqueezed Wavelet Transform (SWT) is used to denoise the raw data. To validate the performance of the proposed DBTSF, experiments are conducted on a real-world dataset collected from Chongqing Jiangbei International Airport to forecast the cooling load for the next 24 hours. The results demonstrate that the proposed method consistently outperforms the eight baseline models. Compared with the best-performing baseline Autoformer, the MAE, RMSE, and MAPE are reduced by 9.33%, 7.40%, and 17.30%, respectively, while R^2 is improved by 0.0180. The generalization of the model is further verified on public data sets. The code is available at the GitHub repository: https://github.com/ABlin1/DBTSF

Aphids are among the most destructive pests that threaten global crop yields, harming crops through feeding and virus transmission. Accurate detection of aphids in fields is a crucial step in implementing sustainable agricultural pest management. However, the tiny size of aphids and the complex image background present significant challenges for accurate identification and classification for in-field detection. In response to the challenges, this study proposes a lightweight real-time object detection model, Aphid-YOLO (A-YOLO), for in-field aphid identification and counting. Specifically, a tiny path aggregation network with C2f-CG modules is proposed to enhance the detection ability of tiny objects while maintaining a low computational cost through efficiently fusing multilayer features. For model training, a normalized Wasserstein distance loss function is adopted to address the optimization challenges caused by the tiny size of aphids. In addition, an optimized data augmentation method, Mosaic9, is introduced to enrich training samples and positive supervised signals for addressing the classification challenge of tiny aphids. To validate the effectiveness of A-YOLO, this study conducts comprehensive experiments on an aphid detection dataset with images collected by hand-held devices from a complex field environment. Experimental results demonstrate that A-YOLO achieves outstanding detection efficiency, with an mAP@0.5 of 83.4%, an mAP@0.5:0.95 of 33.7%, an inference speed of 72 FPS, and a model size of 30.6 MB. Compared to the YOLOv8m model employing traditional Mosaic data augmentation, the proposed method improves mAP@0.5 by 5.8%, mAP@0.5:0.95 by 2.7%, increases inference speed by 5 FPS, and reduces model size by 38.4% .

Pests significantly threaten global agricultural production, which causes severe yield losses through feeding and virus transmission. To mitigate yield losses caused by pests, timely and precise pest management practices are critical. Although previous efforts have advanced automated solutions for real-time environmental monitoring in agriculture, implementing precise pest management decision-making and suggestion generation remains a challenge due to complex reasoning processes in practice. In response, an enhanced pest management system, PEZEGO, is proposed to provide precise management suggestions through multimodal environmental data, a fine-tuned open vocabulary detector (OVD), and large language models (LLMs). Specifically, a mobile application and low-cost Internet of Things (IoT) devices are developed to capture images and environmental information. A hybrid convolutional low-rank adaptation method (HCLoRA) is proposed to fine-tune pretrained OVDs, enabling zero-shot pest detection for converting images to pest species and quantity information. In addition, a structured data-based retrieval augmented generation (SRAG) workflow for LLMs is proposed to provide precise pest management suggestions through automatically extracted agriculture management knowledge and Chain-of-Thought. The effectiveness of PEZEGO is validated in a case study of pest management in the U.K., including pest detection in field scenarios and management suggestion generation. Compared to advanced model fine-tuning methods, HCLoRA for YOLOWorld achieves the highest detection performance with $0.1759~AP^{h}$ on pest detection. Additionally, the proposed SRAG workflow demonstrates the ability to support pest management with a 68.7% average F1 score for knowledge extraction and 77.33% accuracy for suggestion generation. Eventually, a mobile application demonstrates the practical effectiveness of the proposed system.

To enhance the robustness and adaptability of autonomous formation flight, this study develops a fast fixed-time distributed neural formation control-based disturbance observer for multiple quadrotor unmanned aerial vehicles with nonlinear dynamics, input saturation, and unknown disturbances. First, a fast fixed-time stable system is presented to enhance convergence speed. Building on this system, a distributed state observer is proposed to smoothly estimate the state of the leading quadrotor. Second, an adaptive fixed-time disturbance observer is proposed, which resolves chattering issues, enables rapid disturbance estimation, and updates the observer gain. To address the problem of input saturation, a modified anti-saturation method is developed via a Gaussian function and an auxiliary compensation system, achieving high saturation-approximation accuracy and mitigating the impact of approximation error. In addition, a low-computation adaptive neural law is proposed to handle nonlinear dynamics. This mechanism features fewer learning parameters that are independent of the network weight dimensions, thereby reducing computational demands. Theoretical analysis shows that the control errors converge to near the origin within a fixed time. Finally, comparative simulations are conducted to validate the effectiveness and superiority of the presented control strategy

Effective pest management is complex due to the need for accurate, context-specific decisions. Recent advancements in large language models (LLMs) open new possibilities for addressing these challenges by providing sophisticated, adaptive knowledge acquisition and reasoning. However, existing LLM-based pest management approaches often rely on a single-agent paradigm, which can limit their capacity to incorporate diverse external information, engage in systematic validation, and address complex, threshold-driven decisions. To overcome these limitations, we introduce PestMA, an LLM-based multi-agent system (MAS) designed to generate reliable and evidence-based pest management advice. Building on an editorial paradigm, PestMA features three specialized agents, an Editor for synthesizing pest management recommendations, a Retriever for gathering relevant external data, and a Validator for ensuring correctness. Evaluations on real-world pest scenarios demonstrate that PestMA achieves an initial accuracy of 86.8% for pest management decisions, which increases to 92.6% after validation. These results underscore the value of collaborative agent-based workflows in refining and validating decisions, highlighting the potential of LLM-based multi-agent systems to automate and enhance pest management processes.

Wheat yellow rust, caused by the fungus Puccinia striiformis, is a critical disease affecting wheat crops across Britain, leading to significant yield losses and economic consequences. Given the rapid environmental changes and the evolving virulence of pathogens, there is a growing need for innovative approaches to predict and manage such diseases over the long term. This study explores the feasibility of using deep learning models to predict outbreaks of wheat yellow rust in British fields, offering a proactive approach to disease management. We construct a yellow rust dataset with historial weather information and disease indicator acrossing multiple regions in England. We employ two poweful deep learning models, including fully connected neural networks and long short-term memory to develop predictive models capable of recognizing patterns and predicting future disease outbreaks.The models are trained and validated in a randomly sliced datasets. The performance of these models with different predictive time steps are evaluated based on their accuracy, precision, recall, and F1-score. Preliminary results indicate that deep learning models can effectively capture the complex interactions between multiple factors influencing disease dynamics, demonstrating a promising capacity to forecast wheat yellow rust with considerable accuracy. Specifically, the fully-connected neural network achieved 83.65% accuracy in a disease prediction task with 6 month predictive time step setup. These findings highlight the potential of deep learning to transform disease management strategies, enabling earlier and more precise interventions. Our study provides a methodological framework for employing deep learning in agricultural settings but also opens avenues for future research to enhance the robustness and applicability of predictive models in combating crop diseases globally.