Search

The transient power grid stability is greatly affected by the unpredictability of inverter-based resources of today's interconnected power grids. This article introduces an efficient transient stability status prediction method based on deep temporal convolutional networks (TCNs). A grey wolf optimizer (GWO) is utilized to fine-tune the TCN hyperparameters to improve the proposed model's accuracy. The proposed model provides critical information on the transient grid status in the early stages of fault occurrence, which may lead to taking the proper action. The proposed TCN-GWO uses both synchronously sampled values and synthetic values from various bus systems. In a postfault scenario, a copula of processing blocks is implemented to ensure the reliability of the proposed method where high-importance features are incorporated into the TCN-GWO model. The proposed algorithm unlocks scalability and system adaptability to operational variability by adopting numeric imputation and missing-data-tolerant techniques. The proposed algorithm is evaluated on the 68-bus system and the Northeastern United States 25k-bus synthetic test system with credible contingencies using the PowerWorld simulator. The obtained results prove the enhanced performance of the proposed technique over competitive state-of-the-art transient stability assessment methods under various contingencies with an overall accuracy of 99% within 0.64 s after the fault clearance.

Transient stability assessment (TSA) is critical to the reliable operation of a power system against severe fault conditions. In practice, TSA based on deep learning is preferable for its high accuracy but often overlooks challenges in maintaining data privacy while coping with network topology changes. This article proposes an innovative focal loss-based multihead attention convolutional network (FLACON) for accurate post-disturbance TSA under both symmetrical and asymmetrical smart grid faults. The proposed approach effectively incorporates cross-domain deep federated transfer learning (FTL) to leverage local operating data for TSA in a decentralized fashion. By introducing convolutional layers alongside multi-head attention mechanisms, the FLACON framework significantly improves learning efficiency across geographically distributed datasets. To address the challenge of class imbalance, the model integrates a balance factor-enhanced focal loss function. The FTL architecture enables decentralized model training across various clients, thus preserving data privacy and reducing the burden of communication overhead. To avoid the constant adjustment of hyperparameters, the FLACON employs an inductive transfer learning approach for hyperparameter tuning of the pre-trained model, markedly decreasing training time. Extensive experiments on datasets from the IEEE 39-bus system and the IEEE 68-bus system demonstrate FLACON's exceptional accuracy of 98.98% compared to some competitive alternatives.

Lithium-ion battery prognostics and health management (BPHM) systems are vital to the longevity, economy, and environmental friendliness of electric vehicles and energy storage systems. Recent advancements in deep learning (DL) techniques have shown promising results in addressing the challenges faced by the battery research and innovation community. This review article analyzes the mainstream developments in BPHM using DL techniques. The fundamental concepts of BPHM are discussed, followed by a detailed examination of the emerging DL techniques. A case study using a data-driven DLinear model for state of health estimation is introduced, achieving accurate forecasts with minimal data and high computational efficiency. Finally, the potential future pathways for research and development in BPHM are explored. This review offers a holistic understanding of emerging DL techniques in BPHM and provides valuable insights and guidance for future research endeavors.

The widespread penetration of inverter-based resources has profoundly impacted the electrical stability of power systems (PSs). Deepening grid integration of photovoltaic and wind systems is introducing unforeseen uncertainties for the electricity sector. As a cutting-edge machine learning technology, deep reinforcement learning (DRL) breakthroughs have been in the spotlight over the last few years with potential contributions to PS stability (PSS). The ubiquitous DRL architecture, by learning from the dynamism inherent in PSs, produces near-optimal actions for PSS. This article provides a rigorous review of the latest research efforts focused on DRL to derive PSS policies while accounting for the unique properties of power grids. Furthermore, this paper highlights the theoretical advantages and the key tradeoffs of the emerging DRL techniques as powerful tools for optimal power flow. For all methods outlined, a discussion on their bottlenecks, research challenges, and potential opportunities in large-scale PSS is also presented. This review aims to support research in this area of DRL algorithms to embrace PSS against unseen faults and different PS topologies.

For the utility to plan the resources accurately and balance the electricity supply and demand, accurate and timely forecasting is required. The proliferation of smart meters in the grids has resulted in an explosion of energy datasets. Processing such data is challenging and usually takes a longer time than the requirement of a short-term load forecast. The paper addresses this concern by utilizing parallel computing capabilities to minimize the execution time while maintaining highly accurate load forecasting models. In this paper, a thousand smart meter energy datasets are analyzed to perform day ahead, hourly short-term load forecast (STLF). The paper utilizes multi-processing to enhance the overall execution time of the forecasting models by submitting simultaneous jobs to all the processors available. The paper demonstrates the efficacy of the proposed approach through the choice of machine learning (ML) models, execution time, and scalability. The proposed approach is validated on real energy consumption data collected at distribution transformers' level in Spanish Electrical Grid. Decision trees have outperformed the other models accomplishing a tradeoff between model accuracy and execution time. The methodology takes only 4 minutes to train 1,000 transformers for an hourly day-ahead forecast of (~24 million records) utilizing 32 processors.

Forecasting of energy consumption in Smart Buildings (SB) and using the extracted information to plan and operate power generation are crucial elements of the Smart Grid (SG) energy management. Prediction of electrical loads and scheduling the generation resources to match the demand enable the utility to mitigate the energy generation cost. Different methodologies have been employed to predict energy consumption at different levels of distribution and transmission systems. In this paper, a novel hybrid deep learning model is proposed to predict energy consumption in smart buildings. The proposed framework consists of two stages, namely, data cleaning, and model building. The data cleaning phase applies pre-processing techniques to the raw data and adds additional features of lag values. In the model-building phase, the hybrid model is trained on the processed data. The hybrid deep learning (DL) model is based on the stacking of fully connected layers, and unidirectional Long Short Term Memory (LSTMs) on bi-directional LSTMs. The proposed model is designed to capture the temporal dependencies of energy consumption on dependent features and to be effective in terms of computational complexity, training time, and forecasting accuracy. The proposed model is evaluated on two benchmark energy consumption datasets yielding superior performance in terms of accuracy when compared with widely used hybrid models such as Convolutional (Conv) Neural Network-LSTM, ConvLSTM, LSTM encoder-decoder model, stacking models, etc. A mean absolute percentage error (MAPE) of 2.00% for case study 1 and a MAPE of 3.71% for case study 2 is obtained for the proposed forecasting DL model in comparison with LSTM-based models that yielded 7.80% MAPE and 5.099% MAPE for two datasets respectively. The proposed model has also been applied for multi-step week-ahead daily forecasting with an improvement of 8.368% and 20.99% in MAPE against the LSTM-based model for the utilized energy consumption datasets respectively.

Different aggregation levels of the electric grid’s big data can be helpful to develop highly accurate deep learning models for Short-term Load Forecasting (STLF) in electrical networks. Whilst different models are proposed for STLF, they are based on small historical datasets and are not scalable to process large amounts of big data as energy consumption data grow exponentially in large electric distribution networks. This paper proposes a novel hybrid clustering-based deep learning approach for STLF at the distribution transformers’ level with enhanced scalability. It investigates the gain in training time and the performance in terms of accuracy when clustering-based deep learning modeling is employed for STLF. A k-Medoid based algorithm is employed for clustering whereas the forecasting models are generated for different clusters of load profiles. The clustering of the distribution transformers is based on the similarity in energy consumption profile. This approach reduces the training time since it minimizes the number of models required for many distribution transformers. The developed deep neural network consists of six layers and employs Adam optimization using the TensorFlow framework. The STLF is a day-ahead hourly horizon forecasting. The accuracy of the proposed modeling is tested on a 1,000-transformer substation subset of the Spanish distribution electrical network data containing more than 24 million load records. The results reveal that the proposed model has superior performance when compared to the state-of-the-art STLF methodologies. The proposed approach delivers an improvement of around 44% in training time while maintaining accuracy using single-core processing as compared to non-clustering models.

Machine learning algorithms have been intensively applied to perform load forecasting to obtain better accuracies as compared to traditional statistical methods. However, with the huge increase in data size, sophisticated models have to be created which require big data platforms. Optimal and effective use of the available computational resources can be attained by maximizing the effective utilization of the cluster nodes. Parallel computing is demanded to allow for optimal resource utilization in dealing with smart grid big data. In this paper, a master-slave parallel computing paradigm is utilized and experimented with for load forecasting in a multi-AMI environment. The paper proposes a concurrent job scheduling algorithm in a multi-energy data source environment using Apache Spark. An efficient resource utilization strategy is proposed for submitting multiple Spark jobs to reduce job completion time. The optimal value of clustering is used in this paper to cluster the data into groups to be able to reduce the computational time additionally. Multiple tree-based machine learning algorithms are tested with parallel computation to evaluate the performance with tunable parameters on a real-world dataset. One thousand distribution transformers’ real data from Spain for three years are used to demonstrate the performance of the proposed methodology with a trade-off between accuracy and processing time.

Smart grids have been gradually replacing the traditional power grids since the last decade. Such transformation is linked to adding a large number of smart meters and other sources of information extraction units. This provides various opportunities associated with the collected big data. Hence, the triumph of the smart grid energy paradigm depends on the factor of big data analytics. This includes the effective acquisition, transmission, processing, visualization, interpretation, and utilization of big data. The paper provides deep insights into various big data technologies and discusses big data analytics in the context of the smart grid. The paper also presents the challenges and opportunities brought by the advent of machine learning and big data from smart grids.

Smart grids are re-engineering the electricity transmission and distribution system throughout the world. It is an amalgam of increased digital information with the electrical power grids. Managing the data generated from the grid efficiently is the key to successful knowledge extraction from the smart grid big data. Most of the scientific advancements are becoming data-driven and becoming an interesting area of research for data scientists. It is challenging the world computationally enough to develop new storage methods and data processing technologies. Managing big data involves data cleaning, integration of varied data sources, and decision-making applications. This paper focuses on the study of big data management and proposes a management process to help manage the data in the grid. Data management tools and techniques have been leveraged in understanding the sources and data types in the grid. The paper emphasizes the limitations of the existing solutions inclined towards applications of the smart grid big data.

Abstract This paper addresses the problem of DOA estimation using uniform linear array (ULA) antenna configurations. We propose a new low-cost method of multiple DOA estimation from very short data snapshots. The new estimator is based on the annihilating filter (AF) technique. It is non-data-aided (NDA) and does not impinge therefore on the whole throughput of the system. The noise components are assumed temporally and spatially white across the receiving antenna elements. The transmitted signals are also temporally and spatially white across the transmitting sources. The new method is compared in performance to the Cramér-Rao lower bound (CRLB), the root-MUSIC algorithm, the deterministic maximum likelihood estimator and another Bayesian method developed precisely for the single snapshot case. Simulations show that the new estimator performs well over a wide SNR range. Prominently, the main advantage of the new AF-based method is that it succeeds in accurately estimating the DOAs from short data snapshots and even from a single snapshot outperforming by far the state-of-the-art techniques both in DOA estimation accuracy and computational cost.