Search

In the context of energy systems, managing the complex interplay between diverse power sources and dynamic demands is crucial. With a focus on smart grid technology, continuously innovating artificial intelligence (AI) algorithms, such as deep learning, reinforcement learning, and large language model technologies, have been or have the potential to be leveraged to predict energy consumption patterns, enhance grid operation, and manage distributed energy resources efficiently. These capabilities are essential to meet the requirements of perception, cognition, decision-making, and deduction in energy systems. Nevertheless, there are some critical challenges in efficiency, interpretability, transferability, stability, economy, and robustness. To overcome these challenges, we propose critical potential directions in future research, including reasonable sample generation, training models with small datasets, enhancing transfer ability, combining with physics models, collective generative pre-trained transformer-agents, multiple foundation models, and improving system robustness, to make advancing AI technologies more suitable for practical engineering.

The state-of-the-art centralized machine learning applied to fault identification trains the collected data from edge devices on the cloud server due to the limitation of computing resources on edge. However, data leakage possibility increases considerably when sharing data with other devices on the cloud server, while training performance may degrade without data sharing. The study proposes a federated fault identification scheme, named DBFed-LSTM, by combining the distribution-balanced federated learning with the attention-based bidirectional long short-term memory, which can efficiently transfer training processes from the cloud server to edge devices. Under data privacy protections, local devices and the cloud server are specialized for storage and calculation as well as for updating the global model of learning vital time-frequency characteristics, respectively. Given that different device data for monitoring a small probability event are generally non-independent identically distributed (non-IID), a global-model pre-training method and improved focal loss are accordingly proposed. It is verified by the case study that the DBFed-LSTM can be effectively implemented to challenge centralized training with data sharing while preserving privacy and alleviating cloud server computation pressure even for non-IID data. Furthermore, it represents a much preferable performance and robust model to centralized training without data sharing.

This letter proposes two novel effective transfer learning (TL) methods for power system stability assessment (SA) under distinct scenarios: cross-fault, where different types of faults are considered, and cross-scale, which accounts for varying system knowledge levels. Addressing the challenges faced in scenarios with few limited labeled SA data, our proposed data-driven SA models aim to transfer to the different but related scenarios by leveraging numerous instances from fully knowledge database and few labeled instances from the limited knowledge database. Moreover, a significant feature of our approach is the incorporation of the Extreme Learning Machine, a rapid neural network-based learning algorithm. Preliminary testing showcases an improvement of more than 24% in SA accuracy, especially for large-scale cross-scale transfer, demonstrating the efficacy of our TL techniques while maintaining computational efficiency.

In this paper, a live layered video multicast framework based on unmanned aerial vehicle (UAV) relays is presented for on-orthogonal multiple access (NOMA) networks, aiming to maximize the aggregated video reception quality for cell-edge users. A visualizable graph model is constructed to characterize the coupling of the deployment of UAVs and their association with multicast groups with interference cancellation. Based on this graph model, the problem of maximizing multicast group video reception quality is modeled as a clique-based nonlinear integer programming and decoupled into UAV placement and UAV-group association subproblem and subchannel allocation subproblem. For mathematical traceability, the former is transformed into a unique maximum weight clique problem under the available UAV number constraint, solved by an improved branch-and-bound algorithm. For the latter, a heuristic resource-matching strategy is designed to obtain a near-optimal subchannel allocation with low computational complexity. Simulation results demonstrate that the proposed scheme outperforms mainstream benchmarks in terms of aggregate peak signal-to-noise ratio (PSNR), spectrum utilization, network coverage, and adaptability.

To address privacy concerns of state-of-the-art centralized machine learning in non-intrusive load monitoring (NILM) applications, the adoption of federated learning (FL) has emerged as a solution to transfer training processes from cloud servers to edge devices. Nevertheless, conventional FL encounters several challenges including architecture safety, incentive mechanism, computing cost, and personalization. To overcome these challenges, the study proposes a self-motivated decentralized FL scheme for NILM, named blockchain-based clustered FL, by combining blockchain mechanism with clustered FL, incentivizing suitable clients to participate in FL by offering rewards based on data size and model performance. Under NILM-related differential privacy protections, the Laplace noise is injected into the first layer of neural networks in the blockchain-based clustered FL, and a decay factor is employed to mitigate the adverse effects of excessive noise on performance. Lightweight training techniques such as data quantization and weight pruning are employed to reduce computational complexity. Furthermore, a clustering approach is utilized to create multiple global models, thereby enhancing the model personalization degree. It is verified by the case study that the blockchain-based clustered FL outperforms the conventional FL in both accuracy and operation risk, and offers much superior performance and a more robust model compared to local training.