In the context of a high proportion of renewable energy integration, active splitting section search—one of the “three defense lines” of a power system—is crucial for the security, stability, and long-term sustainability of islanded grids. Addressing the random fluctuations of high-penetration wind power and the weakened voltage support capability caused by multi-infeed HVDC, this paper proposes a slow-coherency-based active splitting section optimization model that explicitly accounts for wind power uncertainty and multi-infeed DC stability constraints. First, a GMM-K-means method is applied to historical wind data to model, sample, and cluster scenarios, efficiently generating and reducing a representative set of typical wind outputs; this accurately captures wind uncertainty while lowering computational burden. Subsequently, an improved particle swarm optimizer enhanced by genetic operators is used to optimize a multi-dimensional coherency fitness function that incorporates a refined equivalent power index, frequency constraints, and connectivity requirements. Simulations on a modified New England 39-bus system demonstrate that the proposed model markedly enlarges the post-split voltage stability margin and effectively reduces power-flow shocks and power imbalance compared with existing methods. This research contributes to enhancing the sustainability and operational resilience of power systems under energy transition.

With increasing renewable and power-electronic penetration, emergency control must convert early post-fault measurements into feasible actions within a short latency budget. This paper proposes a response-driven online framework that uses the simplified branch transient transmission capacity (sBTTC) as a physically interpretable interface between causal stability forecasting, action-response prediction, and constrained control. A preceding masked Informer forecasts the no-control all-branch sBTTC trajectories from the first 1.00 s of measured response; a ConvLSTM then predicts generator-tripping recovery and the recovery associated with four load-shedding levels. These predictions are embedded in a weighted mixed-integer piecewise-linear model with stability-recovery, action-bound, and power-balance constraints. On the studied 100-bus renewable-rich AC/DC system, ConvLSTM achieved an RMSE of $7.193\times {10}^{-3}$ and an MAE of $5.674\times {10}^{-3}$ with 69.16 ms inference time. Its inference was 66.42% faster than Informer, while its RMSE was only 2.06% higher; relative to conventional LSTM, its RMSE and MAE were reduced by 28.21% and 21.53%, respectively. Across 62 grouped out-of-sample disturbances, the validation results give a 96.77% control success rate. In the representative disturbance, 900 MW of generation tripping and 740 MW of load shedding restored the nonlinear terminal sBTTC to 0.998, and the command was issued 1.36 s after fault inception. The framework provides an auditable forecast–response–decision chain; its additive approximation is restricted to the validated action range and uses an empirical 0.05 sBTTC recovery margin selected to exceed the observed 95th-percentile absolute error; this margin is not interpreted as a worst-case error bound.

This paper presents the PIML-agent, a physics-informed multi-agent strategy for transient voltage support in ultra-high-voltage direct-current (UHVDC) receiving-end systems. A stress-risk signal derived from a reduced-order receiving-end model is fused with local, neighborhood, and receiving-area voltage channels in a distributed control law whose coordination gains are learned by derivative-free policy search over randomized physics-model rollouts. On the IEEE 39-bus benchmark, with all controllers sharing one receiving-end model and identical actuator limits, the PIML-agent reduces the receiving-area voltage deviation integral by 87.67–97.48% relative to no control and 76.95–94.84% relative to local rule-based support across six fault, load, delayed-support, corridor, and UHVDC-recovery scenarios, and its six-scenario mean by 60.05% relative to a cooperative multi-agent controller. It shortens the average sub-0.90 pu violation duration from 3.62 s (local rule) to 0.56 s and restores the voltage above 0.95 pu within 0.80–2.14 s after fault clearing. It outperforms every baseline in all 100 unseen randomized disturbances, and gains transferred without retraining to the 140-bus NPCC system achieve a 94.80% reduction relative to no control. The advantage persists under gain perturbations, communication delay and loss, agent outage, measurement noise, stronger disturbances, reduced capacity, and model mismatch.

The increasing integration of inverter-based distributed energy resources (DERs) significantly reduces power system inertia, posing critical challenges to transient stability. Traditional fault ride-through strategies, relying on passive and localized rules, often fail to provide effective coordinated support in low-inertia grids. To address these limitations, this paper proposes a Transient Dynamic Features Adaptation Distributed Cooperative Control (TDA-DCC) framework. This approach integrates a dynamic context-aware policy network based on multi-head attention mechanisms to extract temporal features from local observations, allowing agents to anticipate transient dynamics rather than merely reacting to instantaneous states. A multi-agent deep deterministic policy gradient algorithm is employed to optimize a global multi-dimensional objective function encompassing frequency, voltage, and rotor angle stability. Furthermore, to ensure engineering reliability, a hybrid execution architecture is introduced, which embeds a deterministic safety monitor to switch between the intelligent policy and a robust backup controller during extreme anomalies. Case studies on a modified IEEE 39-bus system demonstrate that the proposed method significantly enhances transient stability margins and robustness against sensor failures compared to conventional baselines.