In this paper, we develop a simple algorithm to determine the required number of generating units of wind-turbine generator and photovoltaic array, and the associated storage capacity for stand-alone hybrid microgrid. The algorithm is based on the observation that the state of charge of battery should be periodically invariant. The optimal sizing of hybrid microgrid is given in the sense that the life cycle cost of system is minimized while the given load power demand can be satisfied without load rejection. We also report a case study to show the efficacy of the developed algorithm.

For the islanded dc microgrids, the photovoltaic (PV) generators and storages are the main components to provide power and maintain the system stability. To avoid the dc bus voltage rise, the overcharge and overdischarge of storages, the coordination between generations and storages is critically important. In this paper, a decentralized control method is proposed to realize the generation-storage coordination based on the dc bus voltage signaling. First, a novel state-of-charge (SOC) based droop control strategy is designed for storages to provide dc bus voltage support and balance their SOCs, which can prevent the overuse of a certain storage. Second, an adaptive power control strategy is applied to PV generators, which can seamlessly regulate the output power of PV generators off maximum power point tracking mode according to the dc bus voltage. Combining these two strategies through the dc bus voltage signaling, the generation-storage coordination can be realized. Furthermore, the stability of the proposed control method is also analyzed. All the conclusions are verified by the real-time hardware-in-loop (HIL) tests.

To alleviate the impact of output uncertainty of renewable energy sources (RESs) in unbalanced active distribution networks (ADNs), this paper proposed a two-time scale robust optimization method for the multi-terminal soft open point (SOP). In the long-term scale, the operation points of the SOP were optimized by semidefinite programming (SDP) model to minimize system loss and mitigate voltage unbalance. An improved iterative cutting plane (ICP) method was proposed to strengthen the exactness of the SDP rank-one relaxation. In the short-term scale, $V^2-P$ and $V^2-Q$ droop control mode was implemented in the SOP to better respond to the output uncertainty of RESs. Then a slope robust optimization model was established based on long-term operation points and short-term forecast data to improve system robust security. To coordinate these two time scales, a co-optimization re-dispatch strategy was proposed in the short-term model so that the robust safety margin could be further expanded in extreme conditions. Case studies show that the proposed method can fully utilize the flexible power flow regulation ability of the multi-terminal SOP to reduce system loss, mitigate voltage unbalance and guarantee system robust security in the presence of RES output uncertainty.

The scale-up of AI models for analyzing large-scale power systems necessitates a thorough understanding of their scaling properties. Existing studies on these properties provide only partial insights, showing predictable decreases in loss function with increased model scales; yet no scaling law for power system AI models has been established, resulting in unpredictable performance. This letter introduces and explores the concept of “emergent abilities” in graph neural networks (GNN) used for analyzing large-scale power systems–a phenomenon where model performance improves dramatically once its scale exceeds a threshold. We further introduce an empirical power-law formula to quantify the relationship between this threshold and the power system size. Our theory precisely predicts the threshold for the emergence of these abilities in large-scale power systems, including both a synthetic 10,000-bus and a real-world 19,402-bus system.

In this paper, a novel stability analysis method is proposed to obtain the more accurate and complete results for the LCL-filtered single-stage PV generators. First, the complete model of the PV generators is established including the perturbation and observation (P&O) based power control and phase-locked loop (PLL), which are rarely considered by the existing stability analysis methods. Second, considering the nonlinear discontinuous link, the describing function (DF) method is adopted to analyze the whole system stability based on the established complete model. In this way, the accuracy and completeness of the stability analysis are enhanced. Furthermore, for the critically stable state, both the oscillation amplitude and frequency can be calculated accurately. At the same time, different influence factors are analyzed quantitatively including the operation points, grid strength and the interaction of multiple control loops. It is first revealed that the grid inductor has different influence on the system high-frequency and low-frequency stability. Based on these conclusions, some measures are put forward to enhance the stable margin of the PV generators and to suppress the low-frequency power oscillation. Finally, all the theoretical analyses are verified by the real-time hardware-in-loop (HIL) tests.

When renewable energy penetration increases and the power network becomes more complex, vulnerability and security concerns typically also become more prevalent. A proper evaluation method focused on risk prediction and failure identification is required in modern power systems. In our work, a novel evaluation framework with high efficiency and accuracy is proposed, where the influence of over/under-voltage on renewable energy generators is considered during the cascading failure risk evaluation process. A graph convolutional network and a long short-term memory network are applied to describe the complex network topological structure and extract the system's electrical features, respectively. The neural network model trained with offline simulation data can reduce the numerical effort of evaluating cascading failure risk in real-time. Finally, the proposed method is verified based on the benchmark of both an IEEE 39-bus test system and a real-world French power system. The results show that the proposed cascade neural network model can achieve better evaluation results with shorter training and evaluation times.