Jiaxing Hengchuang Electric Design and Research Institute

Under the concept of intelligent power distribution network system, the operating status of transformers that undertake core functions in substations of power distribution network needs to be accurately judged. The use of voiceprint data, which usually includes the state characteristic information of the transformer, can help to achieve accurate judgment of the working status, fault identification and classification, and timely troubleshooting. Previous studies have shown that due to the distinct characteristics of voiceprint data when various faults occur in transformers, the accuracy of fault identification is high regardless of the method used. Therefore, seeking more efficient and accurate identification methods should have higher practical application value. In this paper, a relatively simple artificial neural network (ANN) structure, namely extreme learning machine (ELM) is employed for fault classification of transformers in substations of power distribution network. The object classified by the ELM is transformed from time-series voiceprint data to a frequency heatmap formed by superimposing several short time periods (5 seconds) cut by a running time (2 minutes) through spectral transformation, and different types of transformer faults are determined through color value identifying in each frequency component. The effectiveness of the proposed method in fault classification is demonstrated through numerical experiments based on transformer data from a substation in Jiaxing, China.

The conventional training of summarization models using maximum likelihood loss faces a challenge: the discrepancy between training and evaluation objectives. Training minimizes token-level loss, while evaluation focuses on sentence or document quality, leading to less effective results. To address this, we propose the Hybrid Loss Abstractive Summarization (HLAS) framework, which combines maximum likelihood loss with global sequence prediction metrics, aligning training with evaluation. Experiments on the BillSum and GovReport datasets show that HLAS significantly outperforms baseline models, achieving higher ROUGE scores and better summarization quality.

Abstract Al2O3 was incorporated into SiCf/SiC–B4C composites to improve their mechanical properties after oxidation. The results showed that Al2O3 reduced the weight gain of these composites, and the oxidized layer became smooth and dense. The fibers pull‐out lengths of the Al2O3‐modified composites were longer than that of virgin composites oxidized for the same amount of time. Compared with the virgin composites, the tensile strength retention rate of Al2O3‐modified composites was improved by 7.3% and 8.4% after oxidation times of 50 and 100 h, respectively. But Al2O3 modification led to 20.6% and 44.3% improvements in σonset (acoustic emission, AE onset stress) as well as 76.5% and 95.5% the improvements in σmin (first AE stress), respectively. Further investigation indicated that a higher proportion of high‐energy level events occurred in the Al2O3‐modified composites when the load just reached the damage thresholds stress compared with the virgin composites.

Electric vehicles have sprung up, and the number of small charging stations in cities is increasing. In order to achieve efficient operation and maintenance, economic management, reasonable location and capacity planning of charging stations, and reduce the impact on the power grid, it is necessary to carry out accurate ultra-short-term load forecasting. In view of the characteristics of the load data of small charging stations, such as time series, nonlinear, low power, and weak periodicity, this paper proposes an improved Northern Goshawk algorithm (INGO) to optimize support vector machine (SVM) for ultra-short-term load forecasting. Firstly, variational mode decomposition (VMD) is used to decompose the time series data of charging load into multiple modal components of different frequencies, so as to better extract the characteristics and structure of time series information. Then, combining the basic meteorological data and holiday data, the Northern Goshawk optimization (NGO) algorithm was improved by Tent mapping and Levy strategy, and the optimal SVM parameters of each component were found to obtain the optimal prediction model. Finally, the components predicted by each SVM were superimposed and combined to obtain the final load-predicted value. The reliability and effectiveness of the proposed method are verified and analyzed on the actual load data of a small charging station and compared with other forecasting models.

Porcelain insulators are widely used in high-voltage transmission lines. Over prolonged periods of operation, their insulation and mechanical performance inevitably deteriorate, which can lead to flashovers in the external insulation of transmission lines and even the explosion of porcelain insulators, causing power outages. In this study, a simulation was conducted using relevant finite element analysis software on a model of the 110 kV straight-line cathead tower suspension double-string insulators. The simulation considered the comprehensive effects of the tower structure, tower head shape, fittings, and conductors. Changes in the position and quantity of zero-value insulators were made to measure variations in the surrounding electric field strength and compared with the field strength distribution under normal operating conditions. The results indicate that the quantity and location of deteriorated insulators have a significant impact on the distribution of the surrounding electric field strength. When insulators near the high-voltage end reach zero value, there is a noticeable increase in the overall electric field strength of the insulator string. Deterior ation of insulators on the opposite side has limited impact on the field strength measured on this side. The research findings provide a theoretical basis for the diagnosis and detection of deteriorated insulators, offering valuable insights for the operation and optimization of insulators.

This paper proposes a delaminating and districting demand response load fair and economical distribution method. The main methods are as follows: firstly, according to the power supply area and voltage level where the load is located, the vertical correspondence between users, distribution transformers, lines and substations is sorted out, and user attribute files are established to form a delaminating and districting load resource pool; then, based on load resource pool and power grid regulation demand, a comprehensive objective function considering user credit rating and regulation cost is established, and constraints such as regulation demand, recovery time, duration and regulation range are introduced to form a demand response load distribution model. Finally, the mixed integer programming technology is used to solve the model, so as to obtain the load demand response scheme that meets the principle of fair and economic distribution. Example test results verify the correctness and effectiveness of this method.

In the existing power grid demand management, there is a lack of perfect demand side resource load response analysis and load schedulable capacity analysis. Aiming at the problem of load adjustable capacity in the power grid, an adjustable load adjustment capacity evaluation model is established to convert the load into load characteristic parameters, and the fuzzy c-means clustering algorithm based on peak density is used to process the load characteristic parameters to accurately identify the adjustable load; Aiming at several influencing factors of adjustable load, the adjustable capacity of adjustable load is explored by using multi-core function, and the capacity is evaluated by different indexes.

In this study, nitrifying bacteria were enriched in a membrane bioreactor (MBR, R1) and their bioaugmentation effectiveness was evaluated in another two MBRs (R2 and R3). Nitrifying activated sludge (NAS) with high nitrification activity of up to 3,000 mg-N/(L·d)-1 was successfully enriched in R1. The results showed that chemical oxygen demand concentration of 100-200 mg/L had no negative effect on NAS enrichment but reduced the ratio of bacterial nitrifiers. Moreover, the cell concentration of nitrifying bacteria in NAS, which was 3.1 × 1011 cells/L, was similar to that of the commercial bacterium agent. For the bioaugmentation test, the reactor inoculated with 14% NAS achieved a 23% higher NH4+-N removal efficiency than that of the uninoculated reactor. Along with the improvement of nitrification performance, the bacterial nitrifiers abundance and microbial richness remarkably increased after bioaugmentation. These results suggested that the MBR system could efficiently enrich nitrifying bacteria using organic carbon containing culture medium, and potentially act as a side-stream reactor to enhance the nitrification function of the wastewater treatment plant.