Short-term scheduling of islanded microgrids with high renewable penetration must reconcile uncertainty, operating cost and post-contingency dynamic feasibility. This paper presents a sequential optimisation-simulation framework that integrates field-calibrated dynamic-security assessment with multistage stochastic dispatch. MVMO first identifies the PowerFactory dynamic parameters from synchrophasor records; correlated multivariate chronicles then represent demand and renewable uncertainty; and steady-state and RMS simulations determine the event-conditioned capability of the network and the grid-forming BESS. A security compiler converts the signed BESS trajectories into directional power and stored-energy margins, while a native Object Pascal extension of the SimSEE BESS actor enforces those margins before each stage cost is evaluated. Consequently, the optimiser searches within a security-qualified feasible set instead of producing an economic schedule that is screened afterwards. An ex-post reliability-worth layer values the selected coverage percentile without assigning an artificial security penalty to the dispatch objective. The Baltra-Santa Cruz study comprises 65,000 audited state-event cases and a separate 30,000-case fault campaign. The calibrated frequency and voltage RMSEs are 53.5 mHz and p.u. At , the upward/downward reserve requirements are 3.66/4.07 MW in Baltra and 1.83/2.03 MW in Santa Cruz. Across 100 aligned chronicles, the unconstrained and schedules for the same assets incur mean thermal-generation costs of $3568 per day and $4777 per day, respectively, while the latter retains an 82.7% renewable share. On the common RMS holdout, the proportion of dynamically insecure states decreases from 42% for the unconstrained schedule and 9% for a static-SCED reserve to 0% for the RMS-informed schedule.
Electric vehicles can provide valuable grid services through bidirectional charging, but coordinating these vehicle-to-everything (V2X) services across multiple buildings remains challenging. This paper develops a hierarchical cognitive digital twin framework that enables intelligent coordination of V2X services across networked building microgrids. The framework uses local digital twins at each building connected to a central coordinator that manages peer-to-peer (P2P) energy trading between buildings. A multi-agent reinforcement learning approach is developed that accounts for battery degradation to protect long-term battery health, with theoretical guarantees ensuring reliable convergence. Validated on IEEE 33-bus test systems with 3-20 buildings using real-world load and EV datasets, the framework achieves 47.7% reduction in daily operating costs compared to a no-V2X baseline, with 8.6%-16.4% improvement over state-of-the-art methods (GNN-RL, MAPPO, QMIX), with consistent performance confirmed across 10 independent trials (). Key benefits include automated adaptation to different electricity tariffs, resilient operation during grid outages and efficient scaling to larger building clusters.
The quality of measurement data in distribution networks is decisive for state estimation accuracy. To address quality issues such as anomalies, missing data, and noise in measurement data, this study proposes a method for bad data identification and reconstruction in distribution networks based on physical information and temporal attention networks. This method constructs a collaborative framework that integrates physical verification and data verification: first, physical verification based on power flow equation residuals is performed to identify obvious anomalies that violate physical laws; second, a long short-term memory network enhanced by a physically informed attention mechanism is employed for data verification to capture potential anomalies in time-series data. Identified anomalous data are reconstructed using model-predicted values, and a secondary physical verification step along with a closed-loop feedback mechanism is introduced to ensure that the reconstructed data satisfy both physical consistency and temporal rationality. To systematically verify the effectiveness of the proposed method, the extended Kalman filter algorithm is applied for unified state estimation using the reconstructed data. Case study analysis demonstrates that the proposed method can significantly enhance the accuracy, robustness, and ability to suppress abnormal data in state estimation.
Virtual synchronous generators (VSGs) can provide flexible inertia and damping support, but these two parameters are strongly coupled. Inappropriate parameter tuning may deteriorate dynamic performance and even destabilise the grid-connected system. Moreover, the short-circuit ratio (SCR), which characterises grid strength, further constrains the admissible ranges of VSG inertia and damping. To support adaptive VSG control and optimised dispatch, this paper establishes a three-dimensional adjustable operating domain for VSG parameters in the inertia-damping-SCR space. The proposed domain integrates the constraints associated with the inertia time constant, steady-state performance, power-control performance, large disturbances and small disturbances. Parameter constraints derived from a higher-order small-signal model are further compared with those obtained from the commonly used second-order VSG model, thereby clarifying their respective accuracy and generality. PSCAD/EMTDC simulation results show that the proposed operating domain identifies parameter combinations associated with acceptable and unacceptable dynamic performance and provides practical boundaries for VSG parameter tuning.
Declining inertia exposes power grids to high-impact, low-probability (HILP) composite contingencies, which can induce severe generation deficits or localised power surpluses during fragmentation. However, generation-first grid codes inadvertently strand the fast frequency response (FFR) potential of storage-first residential behind-the-meter (BTM) assets. To unlock these dormant reserves, this paper proposes the Bidirectional Adaptive Frequency Support (BAFS) proof-of-concept, which delivers symmetric primary frequency regulation by abstracting charging as 'dynamic absorption' and load offsetting as 'virtual injection', proving export-restricted assets can effectively bypass zero-export constraints. Validated via IEEE 39-bus simulations under a severe composite contingency (islanding followed by generator trip), a 20% BAFS penetration with 100-250 ms hardware latencies arrested rapid, zero-crossing frequency transients. BAFS bounded the over-frequency zenith to 1.00841 p.u. and elevated the under-frequency nadir by 171.9 mHz (to 0.99386 p.u.), safely expanding dynamic security margins. Modulating 115 MW required only 0.919 MWh, imposing a negligible state-of-charge burden. Furthermore, hybrid simulations confirm conceptual compatibility and synergistic benefits with the centralised virtual power plant (VPPs) approach. While VPPs require seconds to re-establish topological situational awareness, BAFS autonomously executes the zero-trust primary shock arrest before relinquishing control for secondary restoration, confirming BAFS as a practical, export-agnostic resilience solution.
