Search

This article proposes the federated-learning-based second-order recurrent neural network (SORNN–FL) for distributed frequency control of a multimicrogrid (MMG) system. The problem is conceptualized as proportional–integral derivative consensus control using the Lyapunov function, which depicts the behavior of recurrent neural network (RNN). The parameters of the controller are obtained by deriving the dynamics of energy function to update the state of neurons. During the learning process, FL of weights is adopted to improve the performance of the network by increasing the communication among the neurons. The effectiveness of the developed SORNN–FL controller is tested for the MMG system, and its performance is compared with those of the MMG system using SORNN, RNN, and the conventional method of tuning. The results obtained reveal that the proposed higher order control gives an improved performance during uncertainties and communication failure. Specifically, we use a differential privacy mechanism to secure the data from false data injection attacks by the invaders. The various extensive cases studied on the MMG system demonstrate that the proposed privacy-preserving-based SORNN–FL scheme is robust and efficient. Even if a communication failure and a malicious attack occur in the MMG system, our proposed control scheme can regulate the system frequency and maintain system stability. The real-time validation has been demonstrated at the 0.4-kV microgrid test-bed system in the laboratory.

The economic operation of lithium-ion battery energy storage in electricity markets requires optimally balancing the tradeoff between maximizing the revenue from energy arbitrage and minimizing the capacity loss due to usage. This optimal balance can be achieved by incorporating the stress due to the depth of discharge and battery temperatures in the optimal dispatch framework. However, the stress models are nonlinear and the quantification of partial charge–discharge cycles requires the rainflow cycle counting algorithm, which does not have an analytical form. Considering the challenges, a set of physics-inspired sufficient conditions are developed to handle the nonanalytical form of the rainflow algorithm and to consider cell-level temperatures. The proposed stress cognizant optimal battery dispatch (SC-OBD) framework is applied to a battery participating in both the day-ahead and real-time balancing market. A model predictive control-based framework is proposed to handle uncertain electricity prices in the real-time market and to guarantee the fulfilment of day-ahead market commitments. The numerical results indicate that the proposed SC-OBD can efficiently utilize the cooling to reduce degradation with/without modifying the market-benchmark dispatch.

A growing number of microgrid (MG) systems ipate in market trading as prosumers to improve their own operation profits, as well as the flexibility of power supply. To handle the MG dispatch problem with uncertainties in both load demands and electricity prices, this paper provides a data-driven robust optimization (RO) approach. Firstly, the correlations between the uncertainties of demands and prices are verified via the analysis of historical monitoring data from a real-world case. Based on the data-fitting for the rectangular and ellipsoidal intervals, a new data-driven demand-price uncertainty set with correlations is built to eliminate these unreasonable scenarios effectively. Secondly, a two-stage RO dispatch model is established considering multiple uncertainties of renewable-load power and prices in MGs. The bilinear objective term brings huge obstacles to the solution of the RO model, therefore a novel nested iterative algorithm is further exploited. The bilinear terms of power and price in the objective function are linearized by a binary expansion approach. Besides, to avoid the enumerations of numerous binary recourses in problem solving, the block coordinate descent (BCD) theory is adopted to co-optimize different types of variables, thereby enhancing the computational tractability. Numerical simulations indicate the validity and superiority of the developed RO model and the solution algorithm.

Partial discharge (PD) of overhead lines is an indication of imminent dielectric breakdown and a cause of insulation degradation. Efficient PD detection is the significant foundation of electrical system maintenance. This article proposes a transformer-based multilevel filtering (TBMF) framework for PD detection. It creates the multilevel filtering mechanism to be robust to large-scale industrial measurements contaminated with a variety of background noises and plenty of invalid information. The primary filtering innovatively creates the principle of possible PD measurements to replace feature extraction and reduce manual intervention. For the first time, multiple transformer-based algorithms are introduced to the PD detection field to process the possible PD measurements without relying on the sequence order. The secondary filtering then refines the segmentation-level results from the primary filtering and outputs the overall detection results. Multiple numerical algorithms, artificial intelligence models, and intelligent metaheuristic optimization have been adopted as methodologies of the secondary filtering. The TBMF framework is experimentally verified by extensive field trial data of medium-voltage overhead power lines. Its detection accuracy reaches 96.1$\%$, which outperforms other techniques in the literature. It provides an economic and complete PD detection solution to maintain the economical and safe operation of power systems.

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.

Predicting the degradation behaviors is challenging and essential for prognostics and health management for proton exchange membrane fuel cells (PEMFCs). However, existing methods based on data-driven or model-based methods can face the problem of significant performance inconsistencies in different prediction stages. We investigate the cause and attribute it to the ignorance of the voltage recovery phenomena of PEMFCs observed during the frequent start-stop processes during practical applications. A novel prognostic method is proposed to provide a more comprehensive analysis of PEMFC aging that integrates data-driven and model-based methods. Specifically, a physics-based aging model considering voltage recovery (PA-VR) is first reported as a model-based method to enhance the prediction effect at voltage mutation points. Then, the moving window method with iterative function is used to combine the data-driven method with the PA-VR model, which realizes the online update of model parameters. Finally, the weightings on individual approaches are dynamically determined at different stages throughout the PEMFC lifecycle. The proposed hybrid method achieves an effective improvement in prediction performance by combining the overall degradation trend predicted by the PA-VR model and the local dynamic characteristics predicted by the data-driven method.

Voltage stability assessment is essential for maintaining reliable power grid operations. Stability assessment approaches using deep learning address the shortfalls of the traditional time-domain simulation-based approaches caused by increased system complexity. However, deep learning models are shown to be vulnerable to adversarial examples in the field of computer vision. While this vulnerability has been noticed by the power grid cybersecurity research, the domain-specific analysis on the requirements imposed upon effective attack implementation is still lacking. Although these attack requirements are usually reasonable in computer vision tasks, they can be stringent in the context of power grids. In this paper, we conduct a systematic investigation on the attack requirements and credibility of six representative adversarial example attacks based on a voltage stability assessment application for the New England 10-machine 39-bus power system. We show that (1) compromising about half the transmission system buses’ voltage traces is a rule-of-thumb attack requirement; (2) the universal adversarial perturbations regardless of the original clean voltage trajectory possess the same credibility as the widely studied false data injection attacks on power grid state estimation, while the input-specific adversarial perturbations are less credible; (3) the prevailing strong adversarial training thwarts the universal perturbations but fails in defending certain input-specific perturbations. To advance defense to cope with both universal and input-specific adversarial examples, we propose a new approach that simultaneously estimates the predictive uncertainty of any given input of voltage trajectory and thwarts the attacks effectively.

This paper presents a blockchain-based decentralized frequency control of an islanded microgrid (MG) using a novel federated learning fractional order recurrent neural network (FL-FORNN). A self-adaptive proportional-integral-derivative (PID) controller is proposed using FL-FORNN to handle the uncertainties of power generation by the prosumers while participating in the trading of the islanded MG. In this work, the frequency oscillations during peer-to-peer (P2P) energy trading are regulated through the appropriate design of an adaptive controller and the implementation of a smart contract participation matrix (SCPM) for balancing the generation and consumer demand. A Proof-of-Authority (PoA) private blockchain-based SCPM has been developed to secure the contract among peers. The SCPM provides the power demand information from a consumer to a prosumer who participates in the P2P energy market. It also provides power reference to the prosumer distributed generation as a supplementary control, in addition to the secondary frequency control of the system. To study this, an islanded MG comprising four source nodes (prosumers) and three demand nodes (consumers) are tested with the implementation of the PoA-based blockchain framework. With its SCPM calculation running in the blockchain framework, the decentralized control of prosumers is implemented by interfacing the OPAL-RT with Raspberry Pi devices. The robustness of the proposed FL-FORNN controller has been verified with the FORNN tuned PID controller. Furthermore, a comparative analysis is made with the centralized scheme and traditional decentralized frequency control technique.

The distribution of measurement noise applied in practical power system state estimation (PSSE) can deviate from the assumed Gaussian model, and the performance of an estimator becomes bad if the Gaussian model is still used. This article proposes a new distributed robust PSSE method for multiarea power systems. The non-Gaussian model is utilized to fit the measurement noise distribution to reach high model accuracy. The proposed distributed method is derived based on the maximum correntropy criterion to further reduce the impact of non-Gaussian measurement noise and outliers. The influence function combined with the finite-time average consensus algorithm is used to implement the proposed distributed robust method in a fully distributed manner. Simulations conducted on the IEEE 30-bus, 118-bus and 300-bus systems and the Polish 2383-bus system demonstrate the robustness and effectiveness of the proposed distributed method. Each local area can get the system-wide robust state estimation solution by only using local information and small amounts of data from neighboring areas. Our proposed distributed robust method has at least 12% improvement in reducing the mean squared error under Gaussian-Uniform noise. It is verified the communication network applied for the proposed method is fairly flexible while the existing distributed approaches use a fixed communication network when the power systems are partitioned. The simulation results also verify the robustness of the proposed method to bad data and communication failure.