To address the limitations of insufficient accuracy in traditional methods and deployment challenges of large-scale deep learning models in transformer bolt loosening diagnosis, this paper proposes a visual vibration diagnostic scheme based on an innovative lightweight MS-ST-LiteFormer network. High-speed video sequences are converted into vibration signals using the Steerable Complex Pyramid (SCP) phase-difference accumulation method, followed by Singular Value Decomposition (SVD) noise reduction to construct a multi-dimensional feature set. The MS-ST-LiteFormer achieves time-frequency feature complementarity via dual-modal input fusion, captures subtle loosening features using multi-scale spatio-temporal attention, and reduces deployment costs with lightweight strategies (depthwise separable convolutions, channel pruning). Experimental validation on oil-immersed transformers demonstrates that MS-ST-LiteFormer significantly outperforms traditional machine learning and classic deep learning models. Leveraging non-contact detection and lightweight characteristics, this scheme adapts to in-situ edge deployment, providing core technical support for intelligent transformer bolt loosening diagnosis.

Integrated Energy Systems (IES) have emerged as a key approach to enhancing energy efficiency, promoting renewable energy utilization, and strengthening multi-energy complementarity. To improve the operational stability and flexibility capacity of integrated energy systems, a two-step robust optimization framework considering energy storage planning, the demand response of the upper-level power grid, and internal flexible load resources is proposed. In the first step, a mixed-integer linear programming (MILP) model is developed to determine the optimal capacity and power ratings of electrical and thermal energy storage devices. This planning stage aims to enhance the system's flexibility capacity by smoothing load fluctuations and improving its capability to respond to grid dispatch signals. In the second step, based on the planned storage configuration, a robust scheduling model is formulated to optimize system operation under uncertainties in photovoltaic output and multi-energy loads. Both grid-level demand response participation and internal flexible load regulation within the IES are explicitly considered. Correlated uncertainty sets are introduced to capture the physical coupling between renewable generation and energy demands, and the resulting min–max problem is reformulated into a tractable MILP using duality theory. A case study of Anhui Province's integrated energy system demonstrates the effectiveness of coordinating energy storage planning with robust dispatch in enhancing the economic efficiency, reliability, and demand response performance of integrated energy systems.

Loosening of fastening bolts in power transformers can induce severe faults such as structural resonance and winding deformation, posing a serious threat to the safe and stable operation of the power grid. To address the limitations of conventional contact-based vibration detection methods—namely complex installation procedures, potential safety risks, and insufficient representation of global modal characteristics—this paper proposes a non-contact vibration detection method for transformer bolts based on amplitude-enhanced video phase motion estimation. The proposed method first performs multiscale and multidirectional time-frequency decomposition of the video sequence using a two-dimensional Gabor filter bank. Subpixel-level motion fields are then estimated based on the phase constancy principle. To suppress noise, local amplitude responses are introduced as confidence weights, enabling the reconstruction of time-domain displacement signals. Finally, vibration characteristics are extracted through frequency-domain analysis to identify the bolt condition. To validate the effectiveness of the proposed approach, an oil-immersed transformer experimental platform was established. Vibration video data of bolts in both tightened and loosened states were acquired under no-load and load operating conditions. Feature parameters, including the mean value, peak-to-peak value, fundamental frequency ratio, and the ratio of odd harmonics at 50 Hz, were extracted and analyzed. Experimental results demonstrate that, in the loosened state, the absolute mean value increases, the peak-to-peak value rises significantly, the fundamental frequency ratio decreases, and the proportion of odd harmonics at 50 Hz increases; moreover, these feature differences are more pronounced under load conditions. Across all four operating conditions, the fundamental frequency remains stable at approximately 100 Hz, confirming the reliability of the experimental data. The proposed method is easy to implement and exhibits strong robustness to interference, providing efficient technical support for predictive maintenance of transformer fastening bolts.

Microgrids have broad prospects under the ‘double carbon’ goal. Traditional centralized optimization has the problem of information islands, and multi-agent system (MAS) has become a new solution for distributed coordination due to its autonomy and parallelism. This paper proposes a novel multi-agent reinforcement learning (MARL) framework for optimal scheduling in electrical-thermal microgrids. By integrating the Group Relative Policy Optimization (GRPO) algorithm, we enable distributed coordination among microgrid operators, electrical/thermal load agents, and flexible resources (EV / AC aggregators), achieving cost minimization and renewable energy utilization maximization without centralized control.

Due to the large-scale integration of distributed photovoltaic (PV) systems, the operational volatility of distribution networks has increased sharply, resulting in changes to the power flow direction. This has led to issues such as voltage limit violations, reverse power flow, and challenges in PV accommodation. This project focuses on the regulation potential of flexible distribution networks, represented by intelligent soft switches and mobile energy storage systems, to investigate dispatch optimization methods for flexible distribution networks based on energy storage technologies. First, various devices within flexible distribution networks are incorporated into mobile storage units, leading to the development of an intelligent soft switch system with mobile storage capabilities. Second, by comprehensively considering constraints such as power flow limits, operational constraints of flexible devices, and operational constraints of energy storage units, a dispatch optimization method for distribution networks with mobile energy storage is proposed, aiming to minimize system operating costs. Finally, based on the IEEE 33-bus system, case studies are conducted to validate the effectiveness of the proposed methods and demonstrate their capability to enhance the economic operation of distribution systems under conditions of high PV penetration.

In flexible distribution systems, the strong uncertainty of generation and load demand poses challenges for energy interaction and resource coordination. However, existing energy interaction strategies generally focus only on economic benefits, neglecting safety performance, and are insufficient to ensure the reliable operation of the system. To address these issues, this paper proposes an energy interaction strategy for multi-prosumer flexible distribution systems, considering the economic benefits of all parties and the voltage safety of the system. First, a multi-agent energy interaction framework based on the Stackelberg game is established, and a bi-level optimization model for the distribution network operator and prosumers is constructed. Second, the paper innovatively introduces soft open point-based power flow control technology into the energy trading market. Then, the KKT conditions, dual theory, linearization, and relaxation techniques are applied to transform the original bi-level game problem into a single-level mixed-integer second-order cone programming problem, improving computational efficiency. Finally, the improved IEEE 33-bus distribution system is simulated and compared with two other scenarios. The results show that the proposed strategy can significantly improve the economic and safety performance of the energy interaction system, optimize the power flow distribution, and effectively enhance power quality. The approach offers a promising solution to the growing challenges of managing distributed energy resources in the context of flexible and reliable grid operation.

As distributed photovoltaics rapidly increase in distribution networks, their dispersed nature often leads to voltage violations, undermining power supply stability and quality. Traditional transformers have poor voltage regulation; on - load tap - changers lack smooth adjustment and reliability, struggling in smart grids, while costly power electronic transformers hinder large - scale use. This study thus develops a 10 kV electromagnetic - electronic hybrid distribution transformer prototype: its high - voltage side connects to 10 kV grids, the low voltage side has three single - phase outputs with 200 V and 40 V windings. A bidirectional totem-pole power factor correction(PFC) rectifier cuts losses, and a full - bridge inverter with an inductor capacitor(LC) filter outputs high - quality sine waves, ensuring stable phase and voltage when in series with 40Valternating current(AC).

The traveling wavefront calibration is an important part of fault location for transmission lines, which determines the accuracy of the fault location. In this article, we propose a machine vision-based wavefront calibration method to improve the fault location accuracy. Firstly, the cumulative sum of fault current signal is obtained to detect the starting point, which helps extract the image of the fault current. Then, the Radon transform is applied to the obtained image, and each wavefront corresponds to a most sensitive angle. Finally, according to the angles and projection lines, the arriving time of each wavefront can be calculated. By this means, the fault current data is transformed into image processing, and more accuracy arriving times of the wave fronts can be obtained.

Installed capacity of converter integrated distributed energy resources (CIDER), such as wind power and photovoltaic (PV), has increased rapidly, leading to the increasingly serious problem of harmonics in the power system. To effectively monitor the integration of existing distributed power sources into the distribution network and quantify their impact, it is necessary to trace the sources of distributed power with random injections and quantify the influence of each distributed power source at the point of common coupling (PCC). This paper would decompose the power injection data from CIDER into fundamental and non-fundamental apparent powers, and the latter part will be used as quantification indicator for the impact of CIDER. A dynamic awareness and power tracing method would used to quantify the impact of each distributed power source at the PCC through multiple CIDER measurements, achieving precise perception of CIDER impacts.

Fault localization technology based on traveling wave theory has been widely applied in transmission lines. When applied to distribution network sites, due to their complex grid structures, relatively short line lengths, severe load fluctuations, and high line noise, corresponding improvement work needs to be carried out further. To achieve a better performance of traveling wave technology applying in power distribution network, this paper proposes a pulse characteristics extraction method based on the three-dimensional phase space trajectory analysis. Firstly, the eigenvalues analysis and KPCA method are used to compresses the traveling wave data and identifies suspected pulse sequences. Then, through the nonlinear time series analysis, the traveling wave sequences are transformed into three-dimensional phase space trajectories. Subsequently, this study also proposes the characteristic quantities that represent phase space trajectories. Finally, the effectiveness of the proposed method is validated through experimental data, whose characteristics are obtained by proposed method and used to fault type identification successfully.

With the development of the power system, the requirements for the performance of distribution transformers are increasing day by day. Traditional transformers have problems such as poor voltage regulation performance and easy voltage overrun, which seriously affect the stability and quality of power supply, which brings challenges to the safe and reliable operation of the power grid. This paper proposes an electromagnetic-electronic hybrid distribution transformer, which combines a traditional transformer and a partial power electronic converter to solve the problems of poor voltage regulation performance and voltage overrun of the traditional transformer. Meanwhile, this paper introduces its system theory, including the design method of multi-winding electromagnetic transformers and the topology and control structure of power electronic converters. This paper also constructs an AC-DC-AC converter voltage regulation system based on fully controlled power semiconductor devices, which regulates the voltage by PWM pulse width modulation of the rectifier-inverter circuit. Simulation results confirm the feasibility of the designed electromagnetic-electronic hybrid distribution transformer.

In the transactional processes within a multi-building microgrid system, it is imperative to safeguard stakeholders’ interests and ensure stable, economically efficient operation. Therefore, this paper proposes an integrated interactive control of distribution systems with multi-building microgrids based on game theory. Initially, an interactive framework encompassing superior power grids, distribution network operators, and multi-building microgrids is proposed. This framework establishes a master–slave game model between distribution network operators and multi-building microgrids. Additionally, it introduces the concept of Soft Open Points between building microgrids to enhance system operational safety while also reducing economic costs for distribution network operators. Subsequently, to maintain solution accuracy and optimality, it employs linearization and cone relaxation methods to transform the original model into a mixed-integer second-order cone programming model. Furthermore, it enhances an iterative search method and the price mechanism based on supply and demand ratios. The revised price mechanism serves to boost the participation of building microgrid users in energy regulation, while the iterative search method effectively resolves the equilibrium point of interest among all game participants. Finally, a simulation analysis is conducted on an IEEE-33 test system, and the effectiveness of the proposed strategy is verified by comparing three schemes.

Power equipment such as compressors and generators are not only complex in electrical structure, but also precise in mechanical design. During their operation, it is easy to produce mechanical vibration. Therefore, it is very important to carry out vibration detection on the power equipment, which helps to prevent equipment failure during operation and ensure the safe and stable operation of the power system. Based on this, this paper proposes a power equipment vibration detection method based on acoustic image and visible light image fusion. In the data preprocessing stage, shallow data fusion technology is used. This fusion technology integrates data from different sensors or data sources at a lower level. Then, the information collected by the visible light image feature extraction module and the sound information acquisition module is fused at the feature level. Finally, the vibration state of the power equipment is detected by the Transformer-based conditional detection head. The test results show that the method provided in this paper effectively improves the ability to detect the vibration state of power equipment, and can effectively meet the needs of equipment vibration state detection in power system safety construction.

Microphone arrays have the potential to detect partial discharge phenomena in electric equipment non-contact and flexibly. However, the existing methods do not fully consider the data characteristics of microphone arrays, and there is a paucity of research on the identification of local discharge types. to address the varying acoustic characteristics of different fault types, the microphone array data is initially processed to obtain the respective Mel spectrograms. Subsequently, the features are extracted using a deep convolutional neural network (DCNN), and the local discharge types are classified. Based on this approach, a local discharge type recognition model for acoustic array data is proposed, which is founded on the extraction of Mel spectrogram features.

In order to better meet the electricity demand of residents and reasonably arrange dispatching management, this paper proposes a hybrid neural network model based on the mechanism of CNN-GRU for low voltage and overload trans-former area prediction. Firstly, the original data is preprocessed, including missing value completion and data normalization, the model is established and learning parameters are set; Then, the pro-posed method was verified by a single transformer district in a certain area, and compared with a single neural network, the robustness greatly improved; Finally, the data of all transformer districts in three cities in North China, Central China and South China in 2022 were used for comparison, with an average accuracy of 84.4%, providing help for the safe and stable operation of power systems and references for transformer district prediction research.

The demand for energy storage equipment capable of achieving low energy loss at extremely high temperatures is growing in fields such as electric vehicles and aerospace.Therefore,developing high-temperature resistant and high-performance dielectric energy storage materials is urgently needed.Polyetherimide is the top choice for high-temperature-resistant energy storage thin films due to its high breakdown strength and superior thermal stability.Incorporating nanoparticles into polyetherimide is an effective way to further reduce high-temperature loss.As one of the main forms of energy loss,researching the internal mechanism of conductance loss is valuable.PEI/SiO2 nanocomposite dielectrics with varying doping contents are prepared to clarify the effect of nanoparticle doping on dielectric conductivity.The variable temperature conductivity test results reveal that doped nanoparticles can effectively reduce the conductivity.The space charge limited current model and jump conductance model are used to analyze the variable temperature conductance characteristics of polymer nanocomposite dielectrics.It is evident that the change in doping content does not affect the local state distribution shape in the polymer,indicating that the conductance characteristics of the interface region and substrate in the nanocomposite dielectrics remain consistent.The analysis focuses on the temperature dependence of polymer nanocomposite dielectric conductivity.It has been observed that the temperature dependence adheres to the Arrhenius equation and Meyer-Neldel compensation,suggesting that the traps in nanocomposite dielectrics function in the same way as the polymer matrix.The incorporation of nanoparticles increases the energy level of the local state through a reduction in the compliance of the molecular chain in the interface region,which is depicted in the schematic of the folded-chain fringed micellar grain model.This decreases carrier mobility,manifested as a reduction in dielectric conductivity and energy loss.

This study proposes a method for fault section identification in distribution networks based on variational mode decomposition (VMD) and chaotic oscillators to address the difficulty of effectively identifying electricalcharacteristic quantities during single-phase ground faults. First, the zero-sequence current signals of each line are processed through the VMD algorithm, yielding a series of intrinsic mode functions. Second, an excellent denoising and smoothing model is established to balance the contradictory characteristics of waveform similarity and smoothness. Optimal intrinsic mode functions are chosen to realize the smooth processing of zero-sequence current signals. Finally, the smoothened zero-sequence current signals are inputted into a chaotic oscillator system to obtain phase portraits for each line. Furthermore, the chaotic state of the phase portraits is determined by calculating the Euclidean distance of image texture parameters, enabling fault section identification. Simulations demonstrate that the proposed method is unaffected by factors such as transient resistance, fault closing angle, and noise interference, and it has a high accuracy rate for fault section identification.

With the advancement of urbanization. There is an increasing trend of construction machinery around the power channel, and it is necessary to drive relevant machinery out of the vicinity of the transmission channel to maintain safe and stable operation of the transmission channel. In response to the above requirements, this article proposes a construction machinery detection method based on the fusion of acoustic images and visible light images. This method first uses cross-correlation algorithm to construct acoustic images, and then fuses visible light and acoustic image features at the feature level; Finally, the fused features are input into the DETR head based on the separation of content and spatial information to achieve adaptive detection of construction machinery. The experimental results show that the proposed method effectively improves the detection accuracy of hidden targets, with an AP50 of 77.6%, which can meet the needs of mechanical detection in transmission channel construction.

The flow of contaminated transformer oil has a significant influence on the formation of bridges composed of impurity particles and in turn affects its breakdown progress, but the relevant phenomenon may be not well interpreted by the classical suspended solid particle mechanism. In this article, an oil circulation device was built, and the development of cellulosic bridges in slightly nonuniform electric field at different flow velocities was recorded by a high-speed camera. The results show that, when dc voltage was applied, the time required to form cellulosic bridges across the oil gap (named complete bridge) first decreased and then increased with the rise of flow velocity. It indicates that the low-speed oil flow could accelerate the formation of complete bridge in the dc field, while the high-speed oil flow delays or prevents it. When ac voltage was applied, the length of the longest partial bridge (it did not bridge two electrodes) and the number of cellulose particles accumulated in the gap both increased first and then decreased with the rise of flow velocity. Finally, the results were analyzed according to the force conditions of particles. This article contributes to understand the effect of oil flow on bridging characteristics of cellulose particles and lays the foundation for the subsequent studies of breakdown mechanism about flowing contaminated liquid dielectrics.

In this article, we propose a P2P based distributed dual loop energy management method with voltage regulation capability. In the inner loop process, various prosumers engage in P2P multilateral negotiation and iteration through energy sharing cooperation without considering the physical constraints of the distribution network to achieve the optimal energy trading. Then, the distribution network operator (DNO) calculates the node voltage of the distribution network in the outer ring, updates the price of energy interaction between prosumers and the superior power grid, and guides over/under voltage producers to make decision changes to maintain the voltage security of the distribution network. In addition, we use the alternating direction method of multipliers (ADMM) to update the electricity consumption decisions of consumers. Finally, the effectiveness of the proposed method was verified through IEEE 33 bus simulation.