Time-series forecasting in the energy sector is a labor-intensive process requiring expertise in multiple areas, such as data preprocessing, feature engineering, and neural network optimization. Although large language models offer automation potential, existing large-language-model-based multi-agent systems lack specialized structures for forecasting workflows, balancing computational cost and tool capability remains challenging. To address these issues, we propose an Energy-domain End-to-end Automated Forecasting Multi-Agent System, an on-premises end-to-end forecasting framework built upon a novel Department-Collaborative Multi-agent System Structure, specifically fine-tuned for energy forecasting applications. Within the Energy-domain End-to-end Automated Forecasting Multi-Agent System dominated by Department-Collaborative Multi-agent System Structure, two core and innovative modules are introduced: (i) Structured Behavioral Knowledge Distillation, which enables small-parameter large language models to operate feature engineering tools, facilitating lightweight and efficient on-premises deployment; and (ii) the Alignment-to-Selection Framework, which automates neural network selection by integrating architectural knowledge with historical performance data. Extensive experimental results demonstrate that the novel Energy-domain End-to-end Automated Forecasting Multi-Agent System significantly reduces manual effort, while maintaining high forecasting accuracy across diverse datasets.

Distribution system state estimation (DSSE) plays a crucial role in the real-time monitoring, control, and operation of distribution networks. Besides intensive computational requirements, conventional DSSE methods need high-quality measurements to obtain accurate states, whereas missing values often occur due to sensor failures or communication delays. To address these challenging issues, a forecast-then-estimate framework of edge learning is proposed for DSSE, leveraging large language models (LLMs) to forecast missing measurements and provide pseudo-measurements. First, natural language-based prompts and measurement sequences are integrated by the proposed LLM to learn patterns from historical data and provide accurate forecasting results. Second, a convolutional layer-based neural network model is introduced to improve the robustness of state estimation under missing measurement. Third, to alleviate the overfitting of the deep-learning-based DSSE, it is reformulated as a multitask learning framework containing shared and task-specific layers. The uncertainty weighting algorithm is applied to find the optimal weights to balance different tasks. The numerical simulation on the Simbench case is used to demonstrate the effectiveness of the proposed forecast-then-estimate framework.

The federated learning (FL) technique can provide a promising solution for the timely training of a deep learning model with the critical requirement of privacy protection. However, the existing FL frameworks still confront challenging issues including heterogeneous data sources, edge device heterogeneity, sensitive information leakage, nonconvex loss, and communication resource constraints which place obstacles in terms of practicality. In this article, first, a federated multitask learning (FedMTL) approach is introduced to reformulate the FL model as a multiobjective optimization problem which results in federated multigradient descent algorithm (FedMGDA) with a better model personalization against data heterogeneity and Byzantine attack. Second, a new semi-asynchronous model aggregation method is developed to asynchronously aggregate small partial clients for compensating impacts of the straggler and staleness. Third, a distributed differential privacy technique is applied to enhance the privacy protection of asynchronous FedMGDA with the convergence guarantee where the convergence analysis of differentially private asynchronous federated multiple gradient descent algorithm (DP-AsynFedMGDA) is studied for both the convex and the nonconvex loss functions. Empirical examples and comparative studies are presented to illustrate the effectiveness of the proposed DP-AsynFedMGDA.

Many machine learning models are naturally multitask, which may involve regression and classification tasks, in which they can be trained by the multitask network to yield a more generalized model with the aid of correlated features. When these learning models are deployed on Internet-of-Things devices, the computation efficiency and the privacy of the data can pose a significant challenge to developing a federated learning (FL) algorithm for both higher learning performance and better privacy protection. In this article, a new FL framework is proposed for a class of multitask learning problems with hard parameter-sharing model through which the learning tasks are reformulated as a multiobjective optimization problem for better performance. Specifically, the stochastic multiple gradient descent approach and differential privacy are integrated into this FL algorithm for achieving a Pareto optimality that obtains a good tradeoff among different learning tasks while providing data protection. The outstanding performance of this algorithm is demonstrated by the empirical experiments on multiMINIST, the Chinese city parking dataset, and Cityscapes dataset.

Federated learning techniques provide a promising framework for collaboratively training a machine learning model without sharing users’ data, and delivering a security solution to guarantee privacy during the model training of IoT devices. Nonetheless, challenges posed by data heterogeneity and communication resource constraints make it difficult to develop an efficient federated learning algorithm in terms of the low order of convergence rate. It could significantly deteriorate the quality of service for critical machine learning tasks, e.g., facial recognition, which requires an edge-ready, low-power, low-latency training algorithm. To address these challenges, a communication-efficient federated learning approach is proposed in this paper where the momentum technique is leveraged to accelerate the convergence rate while largely reducing the communication requirements. First, a federated multi-task learning framework by which the learning tasks are reformulated by the multi-objective optimization problem is introduced to address the data heterogeneity. The multiple gradient descent algorithm is harnessed to find the common gradient descending direction for all participants so that the common features can be learned and no sacrifice on each clients’ performance. Second, to reduce communication costs, a local momentum technique with global information is developed to speed up the convergence rate, where the convergence analysis of the proposed method under non-convex case is studied. It is proved that the proposed local momentum can actually achieve the same acceleration as the global momentum, whereas it is more robust than algorithms that solely rely on the acceleration by the global momentum. Third, the generalization of the proposed acceleration approach is investigated which is demonstrated by the accelerated variation of FedAvg. Finally, the performance of the proposed method on the learning model accuracy, convergence rate, and robustness to data heterogeneity, is investigated by empirical experiments on four public datasets, while a real-world IoT platform is constructed to demonstrate the communication efficiency of the proposed method.

Real-time License Plate Recognition plays a significant role in traffic congestion control and road safety monitoring. Practically, a network camera may capture multiple license plates in one frame while the data collected by different network cameras cannot be shared due to privacy concern. In this paper, a federated learning framework is introduced to simultaneously detect multiple license plates over different network cameras through semantic communication. Specifically, to achieve a high efficiency of multiple license plates recognition in real time, the semantic segmentation model is applied to locally extract the important features of an image with multiple license plates. And then, an autoencoder is developed to carry out the semantic encoding which translates the meaningful information. Moreover, a multi-task learning approach for multiple license plates recognition is proposed through a multi-objective optimization technique which can train the license plate recognition model with stronger generalization. To improve the reliability, an asynchronous federated learning algorithm is also considered to ensure the training process can be tolerant to the transmission delay. Empirical experiments on the Chinese City Parking Dataset (CCPD) show that the proposed approach can effectively improve the recognition performance while providing robust service.

Fuel cell prognostic has been considered an efficient way for the health management of fuel cell system and contribute to its lifespan. Nevertheless, accurate fuel cell prognostic is a challenge task because the lifespan of fuel cell is influenced by the operating conditions and there could be reversible aging due to the electrochemical reaction nature of fuel cell. In addition, conventional data-driven prognostic method trains a prediction model for each working condition without leveraging the common aging features for various conditions and thus may yield a less generalized model. What's more, training a promising data-driven prognostic model require large amount of data, which is hard to be collected by one institution due to the high cost and long-term degradation nature of fuel cell, whereas collecting aging data from different entities face the challenges of privacy concern. To address the above-mentioned issue, this paper proposes a private reversible aging-aware method. To address the reversible aging issue, a hint function based method is proposed to add indication terms in the input feature to help the model better deal with the reversible aging. After that, a multitask learning approach is considered to enable the institution to learn shared representations of the aging data, where the multitask learning is reformulated by federated learning to protect the data privacy of each institution. Experiment on the PHM2014 data challenge prove the effectiveness and robustness of the proposed method.

With the recent progress of AI, large amount of data generated in distributed Internet of Things (IoT) devices can be used to build different kinds of models that are helpful to improve people’s daily life. For example, language models can improve the speech recognition performance. Federated learning enables the distributed clients to jointly learn a model with data preserve in local, which provide a promising solution to leverage the massive data. However, in federated learning, the model learned by the local devices need to be repeatedly transmit to the server, which poses communication overhead. To tackle the communication issue, this paper proposes a communication efficient federated learning framework that utilize local momentum term to accelerate the convergence speed. Convergence guarantee under non-convex case is provided. Experiment on EMNIST and CIFAR10 dataset demonstrate that proposed method can effectively increase the convergence speed.

As a renewable and efficient power source, proton exchange membrane fuel cells (PEMFCs) are receiving more and more attention from the world, but it still has shortcomings with poor durability and insufficient reliability, so the purpose of this paper is to effectively improve the reliability and durability of PEMFCs by a data-driven fault diagnosis method. In this method, principal component analysis (PCA) is first adopted to reduce the dimensionality of fault data. Then, a classification method named eXtreme Gradient Boosting (XGBoost) which based on boosting algorithm is used to classify these data. In the end, the experiment is proposed to verify the diagnostic performance of this method. The result shows that the method can effectively identify four healthy states of membrane drying failure, hydrogen leakage failure, normal state as well as unknown state, the diagnostic accuracy can reach 99.72%, and the diagnosis period is 0.17751 s, which is suitable for online implementation.

This paper presents an observer based power switch open-circuit fault diagnosis method for a two-phase interleaved boost converter (IBC). Motivated by the simplicity and robustness of immersion and invariant (I&I) theory, I&I observers are adopted for residual generation. The inductor current and output voltage used in controllers are selected for input voltage estimation, which avoids the use of extra sensors. Furtherly, to eliminate the unexpected effects, which are caused by parameter uncertainty, an adaptive threshold is designed. The simulation results obtained under different operation situations have demonstrated that the preselected threshold has a strong robustness against internal and external disturbances with no false alarms. What' more, the simulation results also show that the switch open-circuit diagnosis is equal to one switch period under various conditions. Thus, the effectiveness of the proposed diagnosis method is validated rigorously.

The power converters can play an important role in fuel cell systems, and robust control design of the converters can improve both the system efficiency and stability. In this paper, an advanced control strategy of a two-phase interleaved boost converter for fuel cell applications is proposed and implemented, which consists of the outer voltage regulation loop and the inner current tracking loop. The current inner loop is designed based on the Super-Twisting sliding mode control to track its reference current tightly. The active disturbance rejection control (ADRC) is applied to the voltage outer loop to regulate the output voltage and generate the reference current for the inner loop. The stability of the control loops is proved through the Lyapunov stability theory and Routh-Hurwitz criteria. Finally, the effectiveness of the proposed control strategy is demonstrated by numerical simulation results, which indicates that strong robustness can be achieved when compared with traditional dual-loop PI controller.