High-precision short-term photovoltaic (PV) power prediction has become a critical technology in ensuring grid accommodation capacity, optimizing dispatching decisions, and enhancing plant economic benefits. This paper proposes a long short-term memory (LSTM)-based short-term PV power prediction method with the genetic algorithm (GA)-optimized adaptive fusion of space-based cloud imagery and ground-based meteorological data. The effective integration of satellite cloud imagery is conducted in the PV power prediction system, and the proposed method addresses the issues of low accuracy, poor robustness, and inadequate adaptation to complex weather associated with using a single type of meteorological data for PV power prediction. The multi-source heterogeneous data are preprocessed through outlier detection and missing value imputation. Spearman correlation analysis is employed to identify meteorological attributes highly correlated with PV power output. A dedicated dataset compatible with LSTM algorithm-based prediction models is constructed. An LSTM prediction model with a GA algorithm-based adaptive multi-source heterogeneous data fusion method is proposed, and the ability to construct a precise short-term PV power prediction model is demonstrated. Experimental results demonstrate that the proposed method outperforms single-source LSTM, single-source CNN-LSTM, and dual-source CNN-Transformer models in prediction accuracy, achieving an RMSE of 0.807 kWh and an MAPE of 6.74% on a critical test day. The proposed method enables real-time precision forecasting for grid dispatch centers and lightweight edge deployment at PV plants, enhancing renewable energy integration while effectively mitigating grid instability from power fluctuations.

High penetration of distributed photovoltaics (PVs) increases overvoltage risk in distribution networks under strong solar generation and uncertain load demand. Existing voltage control methods can mitigate overvoltage, but their online use may be limited by repeated optimization, iterative coordination, or offline training requirements. This paper presents a risk-priority-driven online voltage violation control method for PV clusters. Source-load scenarios are mapped to nodal voltage risks through voltage sensitivity analysis, where violation probability and conditional violation severity are jointly considered. The nodal risks are aggregated into cluster-level priorities, and the inverter commands are dynamically determined according to residual overvoltage and sensitivity-normalized reactive power allocation. Active power curtailment is configured only as a fallback under residual high-risk conditions. In the IEEE 33-bus case study, the proposed method reduces the maximum voltage violation probability by 97.0% and the system voltage risk by 99.91% compared with the no-control case. It keeps the maximum 95th-percentile voltage at 1.0483 p.u., below the 1.05 p.u. limit, while using 39.3% less reactive power absorption than uniform PV control and avoiding active power curtailment.

Limited by the computational performance limits of the CPU(Central Processing Unit), the traditional Spark architecture struggles to achieve high throughput and low latency under the dual pressure of a large data scale and real-time requirements in centralized control systems. This work uses a publicly available CNC(Computer Numerical Control) milling dataset as a functional validation proxy for time-series data processing, then extends validation to a large-scale synthetic power transmission grid dataset. Furthermore, Spark-GPU(Graphics Processing Unit) collaboration suffers from load balancing failure due to heterogeneous resource scheduling and communication overhead, thus failing to unleash its performance potential. This paper proposes a Spark-GPU fusion acceleration technology path. The path consists of three key components: first, it integrates the RAPIDS accelerator; second, it designs a GPU-aware partitioning and task co-scheduling strategy; and third, it optimizes the zero-copy data path. Together, these components realize an integrated collaboration of heterogeneous resources. Validation on real-world datasets yields the following results. In real-time aggregation scenarios, the proposed solution improves throughput by a factor of 3.7 over the pure CPU baseline and reduces end-to-end latency by 62%. Compared with the basic GPU solution, GPU utilization rises from 51.7% to 72.3%, representing a relative improvement of 39.8%. Furthermore, the solution meets industrial-grade high availability requirements. This research significantly improves the processing throughput and reduces end-to-end latency in typical centralized control scenarios, thus providing a feasible technical route for demanding concurrent centralized control scenarios such as electric power industry manufacturing with high real-time demands.

Air conditioning loads play a critical role in maintaining the supply-demand balance of building microgrids (BMGs), yet their distributed nature and volatile response may undermine secure and stable operation. This paper proposes a day-ahead and real-time aggregated control strategy for BMG air conditioning loads with user privacy protection. First, an approximate aggregation model is developed based on building heat transfer characteristics, and the aggregated response potential is evaluated by jointly considering user comfort and willingness. Second, without sharing fine-grained user information, a Building Microgrid Operator (BMO)-Load Aggregator (LA) day-ahead distributed-scheduling model is formulated and solved using the alternating direction method of multipliers (ADMM). Finally, to address load fluctuations caused by heterogeneous initial indoor temperature distributions, a real-time control strategy based on State-Queueing (SQ) temperature-state pre-transfer is proposed. Case studies show that, compared with the baseline scheme, the proposed method reduces the system operating cost from CNY 50,694.58 to CNY 47,131.64, a 7% decrease, and decreases load shedding from 1466.35 kWh to 257.31 kWh, an 82% decrease. Meanwhile, the real-time control effectively suppresses power fluctuations in the early control stage, thereby improving both economic performance and response smoothness.

To address the widening peak-to-valley difference and imbalanced charging station utilization caused by disorderly charging of large-scale electric vehicles, this paper proposes an energy-optimized charging station allocation method based on the Ant Colony Optimization (ACO) algorithm. First, a multi-objective optimization model incorporating SOC, power constraints, and time-of-use electricity pricing is constructed in the Python environment. This model aims to minimize daily operational costs of charging stations, user waiting time penalties, and the mean square deviation of grid load. Second, an adaptive pheromone decay factor and a heuristic dynamic adjustment strategy are introduced into the classical ant colony algorithm. This enhances global search capabilities and convergence speed, effectively achieving spatiotemporal balance of charging loads and economical system operation.

A vulnerability surface modeling method based on dual intensity metrics is proposed to assess the impact of typhoons and heavy rainfall disasters on the distribution pole-tower conductor system. A three-dimensional finite-element model is developed for a typical “three-pole four-conductor” distribution line, considering the uncertainties in both load-side and structural-side parameters. A spatially coherent turbulent wind field is generated using the Davenport spectrum and harmonic superposition method, while an equivalent rain load is derived based on raindrop spectrum integration. Nonlinear dynamic time-history analysis is then conducted under multiple combinations of basic wind speeds and rainfall intensities, extracting engineering demand parameters such as conductor axial tension and pole-base bending moments. Based on probabilistic demand analysis, the relationship between engineering demand parameters and dual intensity measures is regressed in the logarithmic domain to construct bivariate fragility surfaces for both the conductors and the poles. Critical failure curves are obtained by intersecting the fragility surfaces with the 10% exceedance probability level, enabling rapid classification of structural risk under the joint effects of wind and rain. The results show that the regression model provides a high fit, effectively revealing that wind speed is the dominant control factor, while rainfall intensity serves as a secondary amplifying factor. The resulting critical failure curves can be directly used as operation and maintenance warning thresholds and can be coupled with observed and forecast meteorological data for time-varying risk assessment. These findings provide methodological support and engineering guidance for risk assessment, operation and maintenance decision-making, and resilience enhancement of distribution networks under multi-hazard coupling.

With the large-scale integration of distributed photovoltaic generation (PV) into modern distribution networks, the inherent stochasticity and volatility of renewable energy outputs have imposed non-negligible impacts on the secure and economic operation of power systems. Conventional probabilistic power flow (PPF) methods are limited in accurately modeling source–load uncertainty and, more importantly, in capturing complex nonlinear and time-varying dependence among multiple renewable energy sources. To address these issues, this paper proposes a novel PPF calculation framework based on advanced source-load modeling and time-varying D-vine Copula. Firstly, an enhanced finite mixture Beta model and a Gaussian cluster mixture model are developed to characterize the uncertainty of PV output and load demand, respectively. Secondly, a time-varying D-vine Copula model based on the generalized autoregressive score framework is constructed. And a two-stage regularized profile likelihood estimation method is proposed to estimate correlation parameters, capturing the dynamic nonlinear dependence among multiple PV generators. Finally, the de-randomized Sobol sequence-based Quasi-Monte Carlo method is adopted to perform stochastic power flow calculation. Simulation results on a real-world 129-bus distribution system in East China verify the accuracy and effectiveness of the proposed method.

As the participation of photovoltaic-storage systems (PVSS) in the energy and frequency regulation ancillary service markets continues to increase, the market risks caused by photovoltaic output uncertainty will directly affect photovoltaic integration efficiency and the provision of system flexibility, thereby having a significant impact on the sustainable development of power systems. Therefore, studying the risk decision-making of PVSS in the energy and frequency regulation markets is of great importance for supporting the sustainable development of power systems. First, to address the issue where the existing studies regard PVSS as a price taker and fail to reflect the impact of bids on clearing prices and awarded quantities, this paper constructs a market bidding framework in which PVSS acts as a price-maker. Second, in response to the revenue volatility and tail risk caused by PV uncertainty, and the fact that existing CVaR-based bidding studies focus mainly on a single energy market, this paper introduces CVaR into the price-maker (Stackelberg) bidding framework and constructs a two-stage bi-level risk decision model for PVSS. Finally, using the Karush-Kuhn-Tucker (KKT) conditions and the strong duality theorem, the bi-level nonlinear optimization model is transformed into a solvable single-level mixed-integer linear programming (MILP) problem. A simulation study based on data from a PV-storage power generation system in Northwestern China shows that compared to PV systems participating only in the energy market and PVSS participating only in the energy market, PVSS participation in both the energy and frequency regulation joint markets results in an expected net revenue increase of approximately 45.9% and 26.3%, respectively. When the risk aversion coefficient, beta, increases from 0 to 20, the expected net revenue decreases slightly by about 0.4%, while CVaR increases by about 3.4%, effectively measuring the revenue at different risk levels.

In the absence of auxiliary influencing factors, net load point forecasting suffers from insufficient accuracy and difficulty in quantifying uncertainty. To address these issues, this paper proposes a probabilistic net load forecasting method based on an improved crested porcupine optimizer (CPO)-tuned temporal convolutional network and bidirectional long short-term memory (TCN-BiLSTM) model, combined with residual sparse variational Gaussian process (SVGP) modeling. First, the CPO is improved by incorporating a hybrid chaotic initialization and a hybrid Lévy-normal time-varying step size. Meanwhile, a point forecasting model combining TCN and BiLSTM is constructed. Then the improved CPO is employed to optimize the parameters of this forecasting model, and short-term net load point forecasts are obtained. Finally, SVGP regression is applied to the prediction residuals, yielding the predictive mean and variance of the residuals. The probabilistic net load forecasts are generated by combining these estimates with the point forecasts. Experimental analysis on real-world data from a region in East China demonstrates that the proposed method effectively improves net load forecasting accuracy and reliably quantifies the uncertainty of net load in power systems. It thus provides robust technical support for power system operation and scheduling, and holds clear engineering significance.

This study proposes a multi-source heterogeneous data fusion method based on joint Kalman filtering to solve the problems of large data fusion errors and low efficiency in traditional distribution network data processing, thereby improving the accuracy and fusion effect of energy data in power regulation systems. Firstly, by constructing a parameter estimation optimization model and utilizing improved chaos algorithm and Markov Monte Carlo algorithm to fill the missing data in the power dispatch system, the accuracy of parameter processing has been improved. Subsequently, a comparative analysis was conducted on three fusion algorithms: joint Kalman filter, time-domain Gaussian process model, and neural network, and fusion tests were conducted on key parameters. The results indicate that the joint Kalman filter performs the best in terms of fusion error, especially in key indicators such as root mean square error. Specifically, in the process of power data fusion, the joint Kalman filter algorithm has almost no fluctuation in weight during iteration, and with a relative error of 2

With the increasing frequency of extreme weather events, distribution networks are facing compounded impacts from typhoon-rainstorm coupled disasters. This paper develops an integrated disaster model that combines wind field simulation with inundation distribution characteristics to quantify failure probabilities of critical nodes and lines. A Monte Carlo approach is employed to simulate the fault evolution and recovery process under typhoon-rainstorm coupling. Based on a multi-dimensional evaluation framework—including load loss rate, critical load outage duration, islanding continuity, and resilience index - the resilience performance of distribution networks is systematically assessed under both single and coupled hazard scenarios. Simulation results indicate that while a single rainstorm has a relatively limited impact, its combination with a typhoon significantly amplifies system losses, particularly in critical node areas. Proper deployment of distributed generation and energy storage can effectively reduce outage duration and enhance overall resilience. This study provides a practical modelling approach and empirical insights for resilience assessment of distribution networks in multi-hazard contexts.

With the rapid increase in electric vehicle (EV) ownership, charging loads have an increasingly significant impact on the operation of urban distribution networks. Due to the high randomness, volatility, and heterogeneity of EV charging behavior, traditional load modeling methods struggle to fully capture its characteristics. To address this issue, this paper proposes an EV charging load behavior analysis method based on the integration of ISODATA clustering and feature selection. First, to overcome the limitation of requiring a predefined number of clusters in traditional K-means, an adaptive Iterative Self-Organizing Data Analysis Technique (ISODATA) algorithm is introduced to automatically identify typical charging behavior patterns. Then, a multi-dimensional feature indicator system is constructed, and the Spearman correlation coefficient method is employed to select the optimal subset of features. Finally, quantitative user profiling is performed based on the selected features, and visualized using radar charts. Case studies demonstrate that the proposed method effectively improves clustering accuracy, feature representativeness, and profile interpretability, providing a valuable reference for EV load modeling, classification forecasting, and differentiated control strategies.

Accurate probabilistic power flow calculation is the basis for safe operation and optimal dispatch of high-renewable power systems, but the contradiction between computational accuracy and efficiency has long restricted its practical application. To address this issue, a probabilistic power flow method based on truncated R-vine copula and de-randomized Halton sequence quasi-Monte Carlo (QMC) is proposed. The truncated R-vine is used to construct the joint distribution of uncertain variables, which retains key dependencies while significantly reducing modeling complexity. The de-randomized Halton sequence with low-discrepancy property is adopted for efficient sampling, achieving far higher accuracy than random Monte Carlo simulation with fewer samples. Experimental results on the IEEE 39-bus test system show that the proposed method outperforms traditional R-vine, C-vine and D-vine based QMC methods, achieving a good balance between calculation accuracy and efficiency.

With the increasing penetration of renewable energy into integrated energy systems (IES), the intensifying source-load mismatch has led to real-time power imbalance in off-grid IES clusters, decrease in system energy utilization efficiency and power supply reliability, thereby affecting the safe and stable operation of IES clusters. To address this gap, by fully utilizing the information transmission and energy complementarity between IES, a two-stage distributed collaborative optimization strategy for IES clusters is proposed in this paper. In the first stage, the load capacity of IES is evaluated based on the load margin, and a power allocation model based on a consensus algorithm is established. The real-time power of each controllable unit of IES is obtained through iterative processes, the reasonable output and minimum adjustment cost of multiple IES devices can be achieved, and the total power command of the electric-hydrogen hybrid energy storage system (HESS) is transmitted to the second stage. In the second stage, a HESS energy management strategy is developed considering the operational characteristics of alkaline electrolyzers. Furthermore, dynamic adjustment rule for charging and discharging power weighting factors is designed based on a logistic function, and a multi-mode real-time power allocation strategy for HESS is proposed, adjusting the output of two types of energy storage in real time. Finally, numerical simulation verified the feasibility and superiority of the proposed power allocation strategy of IES clusters.

This paper addresses the current issue of charging load forecasting that does not consider the bidirectional coupling of electric vehicle (EV) and charging pile load power, and proposes a bidirectional coupling prediction algorithm for vehicle-grid load power based on knowledge graph and BiLSTM. Firstly, in response to the vast amount of data from EVs and charging piles, as well as the difficulty in integrating data between EVs, power grids, and road networks, this paper constructs a vehicle-roadnetwork coupled three-domain knowledge graph fusion architecture. Secondly, this paper proposes a bidirectional coupling prediction algorithm for vehicle-grid load power based on knowledge graph and BiLSTM to enhance the accuracy of charging pile load forecasting. Finally, this paper conducts a case study analysis for the proposed model, verifying through a comparison with single load forecasting that the model proposed in this paper can quickly achieve accurate prediction of charging load power, thereby further improving the new energy consumption rate.

While the grid-forming inverter can stabilize the weak power system, it still suffers from oscillation problems when connected to a stiff grid. An impedance reshaping based on current controller reshaping is proposed to enhance stability in this paper. Finally, experimental results based on the hardware-in-loop platform are provided to verify the effectiveness of the theoretical analysis and the proposed method for impedance reshaping.

The safe and stable operation of power system is significantly challenged by extreme weather. To effectively reduce the damage caused by extreme weather conditions to the power system, it is essential to identify the weak links in the power system under extreme weather conditions. A vulnerability assessment method is proposed in this paper, which provides a quantifiable decision-making tool for improving the resilience of power system under extreme weather conditions. First, a typical storm scenario is modeled. We introduce a dual-parameter Weibull distribution to describe the variation of line failure probability. Moreover, a piecewise failure probability calculation method is employed to quantify the dynamic impacts of storm conditions on the power system. Then, we use the direct current optimal power flow model (DC-OPF) to calculate the load shedding and power flow changes. Next, to evaluate the operating status of power system, a comprehensive evaluation index system that considers both safety and optimization is proposed. Finally, the non-sequential Monte Carlo method is applied to generate evaluation data, and the analytic hierarchy process (AHP) is utilized to calculate the comprehensive evaluation index to identify the weak links on the system. Simulation results based on the IEEE 39-bus system show the effectiveness of the proposed method.

To mitigate the impact of photovoltaic (PV) power generation uncertainty on power systems and accurately depict the PV output range, this paper proposes a quantile regression probabilistic prediction model (TCN-QRBiLSTM) integrating a Temporal Convolutional Network (TCN) and Bidirectional Long Short-Term Memory (BiLSTM). First, the historical dataset is divided into three weather scenarios (sunny, cloudy, and rainy) to generate training and test samples under the same weather conditions. Second, a TCN is used to extract local temporal features, and BiLSTM captures the bidirectional temporal dependencies between power and meteorological data. To address the non-differentiable issue of traditional interval prediction quantile loss functions, the Huber norm is introduced as an approximate replacement for the original loss function by constructing a differentiable improved Quantile Regression (QR) model to generate confidence intervals. Finally, Kernel Density Estimation (KDE) is integrated to output probability density prediction results. Taking a distributed PV power station in East China as the research object, using data from July to September 2022 (15 min resolution, 4128 samples), comparative verification with TCN-QRLSTM and QRBiLSTM models shows that under a 90% confidence level, the Prediction Interval Coverage Probability (PICP) of the proposed model under sunny/cloudy/rainy weather reaches 0.9901, 0.9553, 0.9674, respectively, which is 0.56–3.85% higher than that of comparative models; the Percentage Interval Normalized Average Width (PINAW) is 0.1432, 0.1364, 0.1246, respectively, which is 1.35–6.49% lower than that of comparative models; the comprehensive interval evaluation index (I) is the smallest; and the Bayesian Information Criterion (BIC) is the lowest under all three weather conditions. The results demonstrate that the model can effectively quantify and mitigate PV power generation uncertainty, verifying its reliability and superiority in short-term PV power probabilistic prediction, and it has practical significance for ensuring the safe and economical operation of power grids with high PV penetration.

To address the real-time monitoring demand of Centralized Control Stations (CCS) for smart grid key equipment and solve the bottlenecks of traditional ultrasonic detection in GIS basin-type insulator applications, this paper proposes a physics-guided industrial B-ultrasound diagnosis framework. Drawing on the “Study on Ultrasonic Propagation Characteristics in Basin-type Insulators” report, an annular phased array system is designed to eliminate mechanical rotation, with the optimal detection frequency determined via Lamb wave attenuation simulation. Integrating Full-Matrix Capture (FMC)-Total Focusing Method (TFM) imaging, waveform accumulation analysis, and a lightweight AI model (MobileNetV2 + YOLOv5n), the framework achieves millimeter-level defect positioning (±O.5 mm) and 97.2% defect recognition rate for cracks/bubbles. Detection results are uploaded to CCS via MQTT protocol, meeting CCS's real-time and data integration requirements.

In order to ensure safe and stable operations of microgrid in a complex working environment, running modes of inverters need to be switched according to actual conditions between PQ control and virtual synchronous generator (VSG) control. In the switching process of traditional inverters, there were sudden changes of control instructions, as a consequence, heavy current pulse and voltage distortions occur, which endangers the stability and safety of microgrid power supply. This paper presents a smooth switching control strategy of $\text{P Q}$ control and VSG control based on power compensation. By using the power compensation, the control instruction after switching is consistent with the control instruction before switching, and the smooth mode switching between PQ and VSG control is realized.