Addressing the gap where existing microgrid research prioritizes economic performance over demand-side carbon emissions， this paper develops a multi-agent collaborative optimization model for park-level microgrids incorporating electricity-carbon coupling mechanisms. Building upon conventional electricity-use optimization， a carbon-responsibility optimization strategy based on carbon emission factors is introduced. This approach enables the flexible transfer and rational allocation of carbon responsibility to drive low-carbon microgrid operation. Furthermore， an equivalent energy storage （EES） model for electric vehicle （EV） clusters is established to leverage demand-side flexibility， thereby reducing the microgrid’s reliance on dedicated energy storage systems. The collaborative optimization is formulated using potential game theory. Case studies demonstrate that the proposed strategy effectively enhances both the economic viability and carbon performance of each agent. Additionally， the EV-based EES significantly promotes renewable energy accommodation and optimizes storage capacity configuration， providing a feasible methodology and decision support for the low-carbon operation of park-level microgrids.
The frequent occurrence of extreme icing disasters poses a severe threat to the secure operation of power grids， presenting significant challenges for the deployment of emergency resources. To accurately forecast emergency resource demands during icing disasters， this paper proposes an emergency resource demand forecasting method for power grid under icing disasters based on WGAN-Stacking （Wasserstein generative adversarial networks and Stacking ensemble learning）. First， multi-source meteorological， geographical， and grid data are integrated. A WGAN is utilized to generate fault samples， and the quality of these generated samples is evaluated using the Fréchet Inception Distance （FID） and Kernel Inception Distance （KID） to alleviate data imbalance issues. Second， a Stacking ensemble learning framework is constructed， employing random forest， gradient boosting decision trees， and support vector machines as base learners， with a feed-forward neural network acting as the meta-learner； this framework leverages the complementary strengths of multiple models to enhance robustness and accuracy of fault forecasting. Based on the forecasted fault locations and historical patterns of emergency repair unit allocation， region-specific and category-specific forecasting of emergency resource demands—specifically for repair personnel and vehicles—is achieved. Finally， a case study is conducted using icing disaster data from Zhejiang Province in 2024. The results demonstrate that the proposed method effectively forecasts emergency resource demands under icing disaster conditions， providing practical guidance for emergency resource allocation.
Existing transient stability assessment （TSA） methods based on artificial neural networks commonly employ features from only a few time instants as inputs， making them unable to capture the temporal characteristics of fault evolution and thereby limiting the assessment accuracy. To address this issue， a TSA method based on a temporal convolutional network with an attention mechanism （TCN-AM） and multi-stage features is proposed. Information from multiple time instants before， during， and after faults is collected to construct multi-stage feature sequences that characterize the dynamic evolution of power systems. The causal dilated convolution operation of TCN is employed to effectively extract the temporal dependencies embedded in the feature sequences. The severely disturbed generators are identified， while irrelevant generator features are eliminated for feature dimensionality reduction. Minority class samples are supplemented to improve sample balance， and an improved focal loss function is adopted to enhance the model’s attention to unstable samples. Furthermore， a multi-head self-attention mechanism is introduced to adaptively focus on and enhance high-value critical features. The proposed method is validated on the IEEE 39-bus system integrated with wind power generation. The results demonstrate that the proposed model achieves high TSA accuracy， fast training speed， and good interpretability.
To elucidate the mechanisms by which different impulse current waveforms impact the degradation rate of electrical performance of zinc oxide （ZnO） varistor valves， and to explore a unified characterization method for the degradation states along with its engineering applicability， impulse tests using 4/10 μs and 8/20 μs current waveforms were conducted on D85 varistor valves under identical energy injection level. The results indicate that the degradation patterns of the macroscopic electrical performance and microscopic grain boundary parameters of the varistor valves are consistent under both impulse waveforms， with the DC reference voltage variation rate （ΔU1mA） being the most sensitive to the degree of degradation. An analysis of the causes behind the variance in degradation rates reveals that the residual voltage and the energy injection rate likely influence the reduction speed of the microscopic grain boundary barrier height and depletion layer width， thereby inducing differences in the macroscopic electrical performance degradation rates. Furthermore， the correspondence between ΔU1mA and the grain boundary barrier height remains consistent across both impulse waveforms， demonstrating that ΔU1mA can serve as an equivalent evaluation basis for the impulse withstand capability of ZnO varistor valves. In the future， the impulse degradation state of varistor valves could be characterized and forecasted online by integrating the offline ΔU1mA measurement with online monitoring indicators， such as impulse count， waveform characteristics， and power-frequency leakage current.
To achieve low-carbon and economically efficient operation of integrated energy systems while enhancing supply-demand flexibility， a coordinated optimal dispatch strategy integrating electric vehicles （EVs） and energy supply resources is proposed. An EV participation willingness model for vehicle-to-grid （V2G） services is first established， based on which an orderly charging and discharging strategy is developed according to state-of-charge （SOC） intervals. A flexible thermoelectric output model is constructed by incorporating a Kalina cycle and a reversible solid oxide cell （rSOC）. Subsequently， a low-carbon economic dispatch model for the integrated energy system is formulated with the objective of reducing total operating cost， while accounting for EV charging and discharging， supply-side flexibility， and tiered carbon trading. Simulation results demonstrate that， compared with the conventional fixed heat-to-power ratio operating mode， the proposed strategy effectively reduces both carbon emissions and operating costs while enhancing flexibility on both the supply and demand sides. The strategy also provides a foundation for the future integration and coordinated utilization of diverse flexibility resources.
