Detecting inter-turn short circuits in ultra-high voltage shunt reactors is challenging due to weak fault signatures. This paper establishes an accurate three-segment fault model and derives, for the first time, explicit analytical expressions for the short-circuit phase equivalent resistance $(\boldsymbol{R}_{\mathbf{eq}})$ as a function of both short-circuit turn ratio $(\boldsymbol{\alpha})$ and transition resistance $(\boldsymbol{R}_{\mathbf{k}})$. In particular, this model reveals specific $\boldsymbol{R}_{\mathbf{eq}}$ variation laws: monotonic decrease with $\boldsymbol{\alpha}$ for metal faults and non-monotonic behavior for transition resistance faults. Leveraging these insights, an innovative identification method is proposed. Its core innovation is a theory-driven, flexible threshold $(\boldsymbol{R}_{\mathbf{set}})$ calculation based solely on the derived $\boldsymbol{R}_{\mathbf{eq}}=\boldsymbol{f}(\boldsymbol{\alpha},\boldsymbol{R}_{\mathbf{k}})$ function and user-defined pa-rameters (minimum detectable $\boldsymbol{\alpha}_{\mathbf{min}}$ and maximum tolerable $\boldsymbol{R}_{\mathbf{k}\_\mathbf{max}}$), thereby reducing reliance on empirical values. Simulation results validate the proposed model and method, demonstrating superior sensitivity and ac-curacy, especially for small-turn inter-turn short circuits with transition resistance, compared to conventional ze-ro-sequence methods. This provides a robust theoretical foundation for practical fault detection.
With the increasing penetration of renewable energy, ensuring reliability and security of power supply has become a significant challenge. In this paper, a novel photovoltaic (PV) intraday power supply guarantee capability forecasting method is proposed. Different from the conventional PV power forecasting methods, it can provide diverse forecasting information, including guarantee power, low output power period, and power supply guarantee probability. PV guarantee power which represents a conservative prediction of PV power is forecasted based on convolutional neural network-bidirectional gated recurrent unit (CNN-BiGRU) model with a compound loss function. PV low output power period represents the period when the power gap between the theoretical maximum PV power and the actual PV power is higher than a specific threshold. It is forecasted using hybrid gradient boosting decision tree and logistic regression (HGBDTLR) model based on an improved spatiotemporal feature encoding method. The power supply guarantee probability which represents the risk of PV power shortage is obtained using quantile regression model based on gradient boosted regression tree (GBRT) considering different PV power supply demands. Furthermore, new indexes, including guarantee rate, guarantee energy ratio, success index, forecasting gain index, and probability forecasting accuracy are proposed to evaluate the forecasting performance. The effectiveness of the proposed method is verified based on actual operation data in a province in Northwest China.
Non-intrusive load monitoring (NILM) has gained widespread attention for improving residential energy efficiency by analyzing appliance-level energy consumption. Although machine learning-based NILM methods have demonstrated excellent performance, their effectiveness heavily relies on the availability of sufficient training data. Federated learning (FL) has emerged as a promising approach for collaboratively training NILM models by aggregating distributed knowledge across clients, effectively harnessing decentralized data reserves. However, due to significant data distribution discrepancies among clients, the global model aggregated through FL often deviates from the client-specific optimum. To address this challenge, this paper proposes a directed knowledge transfer-based personalized model learning method. In this method, clients acquire eligible models through peer-to-peer communication and perform cross-architecture knowledge transfer via knowledge distillation. Furthermore, a data-driven model trust evaluation mechanism is designed to pre-screen candidate models and guide directed knowledge transfer, thereby reducing communication overhead and improving transfer efficiency. Additionally, consistency learning is introduced to mitigate potential overfitting during personalized model training. Extensive experiments on three public datasets, PLAID, WHITED and HOUIDI, demonstrate that the proposed method achieves superior training efficiency and performance compared to existing methods.
With the continuous expansion of wind power integration, dual-sequence synchronization stability has become a critical concern for wind farms under asymmetrical grid faults. In particular, in multi-parallel systems, stability analysis is further complicated by various coupling effects. To address this challenge, a dual-sequence nonlinear dynamic model is developed to identify and characterize three types of coupling interactions. By incorporating the equal area criterion, the impact of coupling effects on dual-sequence synchronization stability under different scenarios is analyzed. On this basis, a method for calculating the coupling critical point is developed under accordance with grid code requirements, which can be utilized to determine appropriate current injection levels for wind farms. Finally, simulation results are presented to validate the effectiveness of the theoretical analysis and the proposed calculation method.
The reliability of All-DC wind farms is of great significance for enhancing the consumption capacity of renewable energy and promoting the construction and development of new power systems. This paper conducts a detailed analysis of the topological structure of All-DC wind farms. Based on the sequential Monte Carlo method, a reliability assessment model is constructed. From dimensions such as space and objects, reliability assessment indicators of hierarchical classes, object classes, time limit classes, and degree classes are defined, and a multi-level and multi-link reliability index system is established. At the same time, the cloud droplet influence value integrated cloud model method is introduced. By using the normal cloud generator, the randomness of index data are effectively processed, and a comprehensive assessment of the multi-dimensional reliability of All-DC wind farms is achieved. Finally, a 100 MW wind farm in Northwest China is taken as an example for simulation verification. The results fully demonstrate the effectiveness and superiority of the proposed method, when compared to conventional approaches. The proposed method can provide key technical support for the planning, operation, and maintenance of All-DC wind farms and the integration of new energy into the grid.
