False data injection (FDI) attacks can mislead the system operator to conduct incorrect dispatch decisions, causing cyber-induced physical line overloads. However, traditional false data is constructed either further from normal data and easy to detect, or not effective to overload multiple lines. To improve both attack stealth and overload impacts, this paper proposes a bilevel cyber-induced overloads (CIO) mechanism that can cause a predefined number of multi-line overloads considering post-attack economic dispatch, where the injected false data is minimised to improve the attack stealth. Within this mechanism, a detailed CIO attack model is formulated that explicitly incorporates the post-attack economic dispatch, enabling it to design more practical attack strategies. One advanced feature of this CIO attack model is the optimal selection of overloaded lines for the bilevel optimization of cyberattack resources and post-attack impacts in terms of line overloads, operation costs, and load loss. To solve the proposed model, it is converted into a single-level nonlinear model by a strong duality theory, and then CIO attacks are discretised to convert this model into a mixed-integer linear programming (MILP) problem. Case studies conducted on an IEEE 14-bus power system and an industrial 126-bus equivalent power system validate the superiority and effectiveness of our proposed approach.

Data-driven online dynamic security assessment offers system operators a computationally efficient approach for monitoring system dynamics. However, the challenges of processing multisource spatial–temporal data from different measurement systems remain unsolved, thus resulting in potentially biased results. In addition, most existing data-driven dynamic security assessment methods that focus on state estimation/prediction overlook the fault location identification, which is important to real-time decision-making. To address the above limitations, an advanced online dynamic security assessment, which learns system dynamics and fault characteristics from multisource spatial–temporal data, is developed. Considering the challenge posed by different sampling rates and sensor numbers, global and local spatial–temporal data from various measurement systems are modeled as graphs with different numbers of nodes and edges. Then, two different sets of graph neural networks are customized to learn global and local spatial–temporal features, respectively. With the learned multisource spatial–temporal features, a Fourier neural operator-based dynamics trajectory predictor and a multilayer perceptron-based fault location identifier are developed for the advanced online dynamic security assessment. Case studies on the IEEE 39 bus system and the IEEE 118 bus system validate the effectiveness and efficiency of the developed online dynamic security assessment.

Dynamic distribution network reconfiguration (DDNR) is a widely used technique for the secure and economic operation of power distribution networks (PDNs), especially in the presence of high-penetration renewable energy sources (RESs). DDNR is realized by controlling the on/off status of remotely controlled switches (RCSs) equipped at power lines in PDNs to optimize power flows. Thanks to the enhanced data availability of PDNs, data-driven solutions to DDNR, such as deep reinforcement learning (DRL), have gained growing attention recently. However, DDNR solves a sequence of combinatorial problems featuring a vast and sparse action space incurred by a so-called "radiality constraint," which is highly challenging for DRLs to handle. Existing DRL methods are either unscalable to large-scale problems or potentially restrict optimality. Hence, we propose a sequential masking strategy to decompose its complex action space into a sequence of maskable sub-action spaces. A gated recurrent unit (GRU)-based agent and an adapted soft actor critic (SAC) algorithm are designed accordingly, producing a data-efficient, safety-guaranteed, and scalable DRL solution to the DDNR problem. Comprehensive comparisons with existing data-driven methods and model-based benchmarks are conducted via various case studies, demonstrating the advantages of the proposed method in both algorithmic performance and scalability.

Two main tasks of online dynamic security assessment (DSA) are real-time state monitoring and postfault transient trajectory prediction. For the first task, the model-based methods are too computationally intensive for real-time state estimation/prediction. As for the second task, the existing deep learning-based methods always require massive training data and fail to provide accurate long-term prediction given the limited fault-on data. To solve the abovementioned problems, a novel online DSA based on physics-guided deep learning models is developed. First, a deep Koopman operator (DKO) is developed to model the system dynamics. Thanks to the linear prediction capability of the DKO, it can be used not only to predict future states from previous states but also to develop a deep Koopman Kalman filter to estimate states from measurements. Then, a physics-informed neural network (PINN) is developed. By integrating the swing equation into the model training, the trained PINN can predict the postfault transient trajectory directly from the fault-on data without waiting for the postfault initial state trajectory or external data. Comparison experiments are conducted on the IEEE 39 bus system and the IEEE 118 bus system to verify the effectiveness and efficiency of the developed approach.

Aeroengine blisks operate in a harsh working environment and are prone to low cycle fatigue (LCF) failure. The probabilistic LCF life prediction considering multiple uncertainties needs to be performed for reliability assessment. To consider the combined effects of heterogeneous uncertainties, this article employs a unified reliability assessment method by processing the uncertainties simultaneously. To overcome the extremely time-consuming limitation of probabilistic finite-element model simulation, this article develops an ensemble generalized constraint neural network (EGCNN)-based unified reliability assessment method. The developed EGCNN surrogate model can conduct efficient, accurate, interpretable, and robust reliability assessments with nonlinear fitting capability, knowledge interpretability, and premature avoidance ability. The developed EGCNN-based unified reliability assessment method can also be applied to other assets and failure mechanisms, providing a new reliability-based design optimization tool.

Situation awareness (SA) has been recognized as a critical guarantee for the stable and secure operation of electric power systems, especially under complex uncertainties after renewable energy integration. In this article, an artificial-intelligence-powered solution is presented to reach a full realization of SA covering perception, comprehension, and prediction, the last of which is more advanced but challenging and hence has not been discussed in any literature before. A novel SA model is proposed by aggregating two powerful deep learning structures: convolutional neural network (CNN) and long short-term memory (LSTM) recurrent neural network. The proposed CNN-LSTM model has superiority to achieve collaborative data mining on spatiotemporal measurement data, i.e., to learn both spatial and temporal features simultaneously from phasor measurement units data. Two functional branches are designed within the SA model: a contingency locator to detect the exact fault location at present and a stability predictor to predict stability status of the system in the future. Test results have shown high performance (accuracy) of the model even on a low level of data adequacy. The proposed SA model can promisingly facilitate very fast postfault actions by the system operators to prevent the power system from any unstable operational status.

This article proposes an optimal coordinated scheduling of electric vehicles (EVs) for a virtual power plant (VPP) considering communication reliability. Recent advancements on wireless technologies offer flexible communication solutions with wide coverage and low-cost deployment for smart grid. Nevertheless, the imperfect communication may deteriorate the monitoring and controlling performance of distributed energy resources. An interactive approach is presented for combined optimization of dynamic spectrum allocation and EV scheduling in the VPP to coordinate charging/discharging strategies of massive and dispersed EVs. In the proposed approach, a dynamic partitioning model of the multi-user multi-channel cognitive radio is used to cope with the vehicle-to-grid (V2G) communication issue due to variable EV parking behaviors, and a two-stage V2G dispatch scheme is proposed for the wind-solar-EV VPP to maximize its overall daily profit. Furthermore, the effects of packet loss probability on the VPP scheduling performance and battery degradation cost are thoroughly analyzed and investigated. Comparative studies have been implemented to demonstrate the superior performance of the proposed methodology under various imperfect communication conditions.

Current maintenance mode for high-speed railway (HSR) power equipment is so outdated that can hardly adapt to the high-standard modern HSR. Therefore, a new possibility is proposed in this article to update the obsoleting maintenance mode of the HSR power equipment by adopting both predictive maintenance and proactive maintenance. With the combination of data-driven (predictive) and model-based (proactive) approaches, two principal constituents-the sample generator and the maintenance predictor-are designed. The maintenance predictor which is powered by the long short-term memory recurrent neural network is developed to realize the goal of predictive maintenance. The sample generator which is formulated by the physical degradation and failure model of HSR power equipment is proposed toward the goal of proactive maintenance. Test results on a gas-insulated switchgear have shown the powerful collaboration between the generator and the predictor, to not only accurately predict future maintenance timing of the switchgear based on historical sample data, but also enrich the data supply proactively to deal with potential data deficiency problems.

The probabilistic design of complex structure usually involves the features of numerous components, multiple disciplines, nonlinearity, and transients and, thus, requires lots of simulations as well. To enhance the modeling efficiency and simulation performance for the dynamic probabilistic analysis of the multicomponent structure, we propose an improved decomposed-coordinated Kriging modeling strategy (IDCKMS), by integrating decomposed-coordinated (DC) strategy, extremum response surface method (ERSM), genetic algorithm (GA), and Kriging surrogate model. The GA is used to resolve the maximum-likelihood equation and achieve the optimal values of the Kriging hyperparameter θ. The ERSM is utilized to resolve the response process of outputs in surrogate modeling by extracting the extremum values. The DC strategy is used to coordinate the output responses of analytical objectives. The probabilistic analysis of an aeroengine high-pressure turbine blisk with blade and disk is conducted to validate the effectiveness and feasibility of this developed method, by considering the fluid- thermal-structural interaction. In respect of this investigation, we see that the reliability of turbine blisk is 0.9976 as the allowable value of radial deformation is 2.319 × 10-3m. In terms of the sensitivity analysis, the highest impact on turbine blisk radial deformation is of gas temperature, followed by angular speed, inlet velocity, material density, outlet pressure, and inlet pressure. By the comparison of methods, including the DC surrogate modeling method (DCSMM) with quadratic polynomial, the DCSMM with Kriging, and the direct simulation with finite-element model, from the model-fitting features and simulation performance perspectives, we discover that the developed IDCKMS is superior to the other three methods in the precision and efficiency of modeling and simulation. The efforts of this article provide a highly efficient and highly accurate technique for the dynamic probabilistic analysis of complex structure and enrich reliability theory.

With continually increasing complexities of power systems, transient stability analysis as an important task for system security operation becomes very time consuming and thus an efficient analysis tool is urgently needed. In this paper, a spatially paralleled hybrid approach combining the high-order Taylor series algorithm and the block bordered diagonal form (BBDF) was proposed to improve the computational efficiency of power system transient stability analysis. The proposed approach only exchanged high-order derivatives of generator voltages and currents between partitioned power network subsystems and the coordinated-bus cluster based on BBDF, and enhanced the triangular factorization recursive utilization rate of admittance matrix using the Taylor series algorithm with a large integration step. Finally, the proposed spatially paralleled hybrid approach was tested in the New England 39-bus, IEEE 145-bus and an expanded 580-bus systems, and simulation results have validated the proposed approach is very effective and computationally efficient for power system transient stability analysis.