Accurate topology identification and line parameter estimation are essential for monitoring and control of Low-Voltage Distribution Grids (LVDGs), yet practical deployments are severely challenged by low observability and dense measurement noise. In this paper, we propose a three-stage Prior Knowledge-Integrated Adaptive Search (PKIAS) framework for joint topology and parameter identification. The first stage formulates the objective that integrates power-flow physical constraints with available prior knowledge, reducing reliance on high-quality data while balancing data-model consistency and physical plausibility. The second stage introduces a Balanced Adaptive Search method to optimize the objective formulated in Stage one. By adaptively combining the Stochastic Fractal Search algorithm with chaos, refinement mechanism, and regression algorithm, this stage enhances both the reliability and robustness of the search process. With the best-found solution obtained in Stage two, Stage three employs the Complex Hierarchical Clustering model to recover the complete topology and branch parameters. The PKIAS framework thus addresses large-scale LVDGs under low observability and dense noise by leveraging prior knowledge for improved robustness. Tests on a practical urban LVDG demonstrate that the PKIAS method reliably reconstructs the true topology, attains high-fidelity parameter estimates with strong noise robustness, and provides high-granularity models of the LVDG.
The rapid proliferation of distributed energy resources (DERs) aggregated by virtual power plants (VPPs) elevates cybersecurity risk by expanding the attack surface. A coordinated cyberattack has the potential to disrupt distribution-system, bulk-system, and market operations. Existing cyberphysical defense strategies largely assume access to grid network models and operating conditions, which may not be available to third-party aggregators. This paper proposes a novel hybrid, perturbation-based cyber-physical defense for VPP-operated DERs that blends game-theoretic planning with model-free reinforcement learning (RL) under a Bayesian belief-update framework. Topology agnosticism is defined from the VPP information and implementation perspective: the VPP does not require access to the feeder topology, line parameters, switching configuration, or full system state. This work advances the state of the art through three contributions: (1) a multi-output Gaussian process with Hamiltonian Monte Carlo for discrete attacker-type inference, probability of detection, and confidence of detection for modeling the belief distribution; (2) moment-matched Gaussian propagation of posterior beliefs across episodes; and (3) an RL policy that adaptively integrates Stackelberg leader (VPP) actions with exploration. Furthermore, we establish equilibrium existence and derive an episode-level closed-loop utility bound connecting attacker-belief uncertainty with GT–RL action alignment. Evaluations on IEEE 13- and 123-Bus feeders with DER portfolios demonstrate effective detection under attacker-model uncertainty and evaluated feederre-configuration conditions, while requiring no explicit feeder-topology knowledge. Overall, the results indicate that combining bi-level game-theoretic planning, nonparametric Bayesian inference, and adaptive RL is a promising direction for VPP cybersecurity.
Identifying the power-grid components whose loss can cause the greatest disruption poses a significant challenge as simultaneous outages can interact in ways that node-by-node, topology-only analyses cannot represent. Nevertheless, vulnerability studies commonly rely on classical single-node centralities, all-pairs shortest paths, and purely structural graph models. We address these limitations within a unified centrality framework that combines electrical information, group-level evaluation, and operationally relevant source–target restrictions. Specifically, we compare classical and electrical centralities, extend reactance- and current-flow-based measures to groups of nodes, and formulate source–target variants that restrict betweenness computations to generator–load interactions. We evaluate the resulting methods on benchmark grids of varying sizes under generator and node attacks using complementary operational and structural impact measures. Group centralities turn out to be most beneficial when removals alter network topology, whereas source–target restrictions improve attack quality in most settings and substantially reduce the computational cost of group betweenness on larger grids. These results demonstrate that jointly accounting for electrical flow, collective outages, and generator–load roles improves the identification of critical component sets and makes advanced vulnerability analysis more practical on larger systems.
Time-synchronization (TS) attacks on Phasor Measurement Units (PMUs) can pose a serious risk to Wide-Area Monitoring, Protection, and Control (WAMPAC) applications, in particular Voltage Stability Monitoring (VSM). On this basis, this article develops optimization-driven TS attack models that target major VSM indicators, such as the Impedance Stability Index (ISI), while remaining undetected by conventional Largest Normalized Residual (LNR) Bad-Data Detection (BDD) schemes. Two attacker models are formulated. First, a Single-Level (SL) formulation computes coordinated multi-PMU timing offsets that maximally degrade the operator-perceived ISI under AC powerflow feasibility and LNR-BDD stealth constraints. Second, an operator-aware bi-level formulation anticipates practical voltage-control responses (reactive support and shunt actions) triggered by the biased ISI, producing a divergence between perceived and true stability margins and potentially inducing counterproductive corrective actions. To defend against such timing-induced deception, we propose a supervised, topology-aware Graph Neural Network (GNN) detector trained on physics-informed simulated snapshots, with training-only conditional Generative Adversarial Network (GAN) augmentation to mitigate class imbalance. The detector provides node-level attribution via feature saliency to support posthoc localization of compromised regions/PMUs. Simulations on the IEEE 30-bus and 118-bus systems show that the proposed attacks can fabricate a false voltage-stability crisis, while the topology-aware GNN detector outperforms system-level baselines and effectively flags stealthy attacks across diverse loading conditions and attacked-PMU sets.
Transitional weather causes abrupt changes in meteorological patterns, significantly reducing the accuracy of day-ahead photovoltaic (PV) power forecasting. To address this, we propose a multimodal Seq2Seq forecasting framework enhanced with implicit category projection. The framework features: (1) an implicit category projection embedding layer that classifies PV generation profiles via density-weighted clustering and a supervised pretraining category projection block, overcoming the assumption that input series and prediction horizon share the same category; (2) a time-aligned multimodal encoder that performs temporal projection of PV categories through bit-level feature embedding gate and integrates meteorological features via a time-step-wise feature fusion MLP, effectively handling heterogeneous PV type and weather sequence. Experiments on PVOD and DKASC datasets show the method achieves superior performance under transitional weather conditions.
