Transformer-based correction scheme for short-term bus load prediction in holidays

To tackle the problem of inaccurate short-term bus load prediction,especially during holidays,a Transformer-based scheme with tailored architectural enhancements is proposed.First,the input data are clustered to reduce complexity and capture inherent characteristics more effectively.Gated residual connections are then employed to selectively propagate salient features across layers,while an attention mechanism focuses on identifying prominent patterns in multivariate time-series data.Ultimately,a pre-trained structure is incorporated to reduce computational complexity.Experimental results based on extensive data show that the proposed scheme achieves improved prediction accuracy over comparative algorithms by at least 32.00%consistently across all buses evaluated,and the fitting effect of holiday load curves is outstanding.Meanwhile,the pre-trained structure drastically reduces the training time of the proposed algorithm by more than 65.75%.The proposed scheme can efficiently predict bus load results while enhancing robustness for holiday predictions,making it better adapted to real-world prediction scenarios.

Effect of Thermal Cycling under Load on Martensite Transformation and Two-way Shape Memory Effect in a TiNi Alloy

The effect of thermal cycling under loading on martensitic transformation and two-way shape memory effect was investigated for Ti-49.8 at, pet Ni alloy. It is shown that M(s), and M(f) temperature increase with increasing the number of cycles, while A(s) and A(f) temperature decrease during thermal cycling. The total strain at and permanent strain epsilon (p) increase with increasing applied stress and number of cycles. The two-way shape memory effect can be improved by proper thermal cycling training under loading, while excessively high applied stress results in the deterioration of TWSME. The reason for the changes in martensitic transformation characteristics and two-way shape memory effect during thermal cycling under loading is discussed based on the analysis of microstructure by TEM observations.

Phase Transformations in ZrO_2 after Shock Loading

The effects of shock loading on the morphology,grain growth during heating and phase transforma- tion of ZrO_2 have been investigated.It is shown that shock loading may be efficiently used to modify submicron ceramic powders with nanocrystalline structure.After shock loading,the critical diameter of ZrO_2 particles transformed from tetragonal to monoclinic decreased due to stored strain energy. Annealing of powders resulted in reversible transformation to the tetragonal without considerable grain growth up to 1200℃.

Transformation Matrix for Combined Loads Applied to Thin-Walled Structures

This paper transforms combined loads, applied at an arbitrary point of a thin-walled open section beam, to the shear centre of the cross-section of the beam. Therein, a generalized transformation matrix for loads with respect to the shear centre is derived, this accounting for the bimoments that develop due to the way the combined loads are applied. This and the authors’ earlier paper (World Journal of Mechanics 2021, 11, 205-236) provide a full solution to the theory of thin-walled, open-section structures bearing combined loading. The earlier work identified arbitrary loading with the section’s area properties that are necessary to axial and shear stress calculations within the structure’s thin walls. In the previous paper attention is paid to the relevant axes of loading and to the transformations of loading required between axes for stress calculations arising from tension/compression, bending, torsion and shear. The derivation of the general transformation matrix applies to all types of loadings including, axial tensile and compression forces, transverse shear, longitudinal bending. One application, representing all these load cases, is given of a simple channel cantilever with an eccentrically located end load.

Space Transformation-Based Interdependency Modelling for Probabilistic Load Flow Analysis of Power Systems

Dependence among random input variables affects importantly the results of probabilistic load flow(PLF),system economic operation,and system security.To solve this problem,the main objectiveness of the paper is to analyze the performance of several schemes for simulating correlated variables combined with the point estimate method(PEM).Unlike the existing works that considering one single scheme combined with Monte Carlo simulation(MCS) or PEM,by neglecting the correlation among random input variables,four schemes were presented for disposing the dependence of correlated random variables,including Nataf transformation /polynomial normal transformation(PINT) combined with orthogonal transformation(OT) / elementary transformation(ET).Combining with the 2m+1 approach of PEM,a space transformation-based formulation was proposed and adopted for solving the PLF.The proposed approach is applied in the modified IEEE 30-bus system while considering correlated wind generations and load demands.Numerical results show the effectiveness of the proposed approach compared with those obtained from the MCS.Results also show that the scheme of combining Nataf transformation and ET with PEM provides the best performance.

基于生成对抗Transformer的电力负荷数据异常检测

电力负荷异常数据将给电力系统规划、负荷预测以及用能分析等带来较大的负面影响,因此亟须对负荷数据异常进行检测与识别。首先,针对电力负荷数据异常分类、原因及其特征开展分析。其次,改进传统Transformer编码器结构,采用多头注意力层代替掩码多头注意力层,同时移除前馈网络,以提高模型对负荷时序序列的全局注意力。基于生成对抗网络(generative adversarial networks,GAN)生成器与判别器的博弈结构,提出一种改进的GAN-Transformer模型,以更好地捕捉趋势性特征并加速模型收敛。然后,引入多阶段映射与训练方法,综合焦点分数打分机制,通过分阶段负荷序列重构帮助模型更好地提取负荷数据异常特征。最后,算例分析结果表明,GAN-Transformer模型在负荷数据异常检测精确率、召回率、F_(1)值以及训练时间方面均具有更优的性能,验证了所提方法的有效性和优越性。文中研究工作为基于深度学习进一步实现电力负荷数据异常分类与数据修复提供了有益参考。

Rotor broken bar fault diagnosis for induction motors based on double PQ transformation

A new rotor broken bar fault diagnosis method for induction motors based on the double PQ transformation is pre-sented. By distinguishing the different patterns of the PQ components in the PQ plane,the rotor broken bar fault can be detected. The magnitude of power component directly resulted from rotor fault is used as the fault indicator and the distance between the point of no-load condition and the center of the ellipse as its normalization value. Based on these,the fault severity factor which is completely independent of the inertia and load level is defined. Moreover,a method to reliably discriminate between rotor faults and periodic load fluctuation is presented. Experimental results from a 4 kW induction motor demonstrated the validity of the proposed method.

Design of Non-contact On-load Automatic Regulating Voltage Transformer

At present, an automatic-mechanic contact tap-changer is widely used in power system, but it can not frequently operate. In addition, arc will occur when the switch changes. In order to solve these two problems, this paper presented an automatic on-load voltage-regulating distributing transformer which employed non-contact solid-state relay as tap-changer, and mainly introduced its structure, basic principal, design method of each key link and experimental results. Laboratory simulation experiments informed that the scheme was feasible. It was a smooth and effective experiment device, which was practical in application.

TRIP Steel Retained Austenite Transformation under Cyclic V-bending Deformation

To investigate the transformation behavior of TRIP steel retained austenite under cyclic load, cyclic V-bending deformation of low carbon Si-Mn TRIP600 was studied by experiment and finite element in this paper. The results showed that, under cyclic V-bending deformation, retained austenite in TRIP steel transformed into martensite gradually with the increasing of bending times, and for the symmetrical characteristic, upper surface and lower surface presented the same transformation tendency. From the first to the fourth V-bending deformatiort, retained austenite volume fraction decreased nearly linearly and then attained saturation step by step. Compressive stress state was helpful for martensite transformation than tension stress state with V-bending deformation, and strain magnitude was the determining factor for retaining anstenite martensitic transformation. With the increasing of bending times effective stress increased and the relationship between maximum effective stress and bending times was nearly linear. Effective stress and effective strain distribution were non-uniform, the maximum effective stress and effective strain were present in the center of the samples. The relationships between retained austenite and V-bending times, and retained austenite with effective strain were set up as Eqs.（1）-（5）. The relationship was typical quadric function, decreased linearly for the initial deformation and attained saturation finally.

Empirical Thermal Investigation of Oil-Immersed Distribution Transformer under Various Loading Conditions

The distribution transformer is the mainstay of the power system.Its internal temperature study is desirable for its safe operation in the power system.The purpose of the present study is to determine direct comprehensive thermal distribution in the distribution transformers for different loading conditions.To achieve this goal,the temperature distribution in the oil,core,and windings are studied at each loading.An experimental study is performed with a 10/0.38 kV,10 kVA oil–immersed transformer equipped with forty–two PT100 sensors(PTs)for temperature measurement installed inside during its manufacturing process.All possible locations for the hottest spot temperature(HST)are considered that made by finite element analysis(FEA)simulation and losses calculations.A resistive load is made to achieve 80%to 120%loading of the test transformer for this experiment.Working temperature is measured in each part of the transformer at all provided loading conditions.It is observed that temperature varies with loading throughout the transformer,and a detailed map of temperature is obtained in the whole test transformer.From these results,the HST stays in the critical section of the primary winding at all loading conditions.This work is helpful to understand the complete internal temperature layout and the location of the HST in distribution transformers.

Management of Charging Load of Electric Vehicles for Optimal Capacity Utilisation of Distribution Transformers

A de-centralised load management technique exploiting the flexibility in the charging of Electric Vehicles (EVs) is presented. Two charging regimes are assumed. The Controlled Charging Regime (CCR) between 16:30 hours and 06:00 hours of the next day and the Uncontrolled Charging Regime (UCR) between 06:00 hours and 16:30 hours of the same day. During the CCR, the charging of EVs is coordinated and controlled by means of a wireless two-way communication link between EV Smart Charge Controllers (EVSCCs) at EV owners’ premises and the EV Load Controller (EVLC) at the local LV distribution substation. The EVLC sorts the EVs batteries in ascending order of their states of charge (SoC) and sends command signals for charging to as many EVs as the transformer could allow at that interval based on the condition of the transformer as analysed by the Distribution Transformer Monitor (DTM). A real and typical urban LV area distribution network in Great Britain (GB) is used as the case study. The technique is applied on the LV area when its transformer is carrying the future load demand of the area on a typical winter weekday in the year 2050. To achieve the load management, load demand of the LV area network is decomposed into Non-EV load and EV load. The load on the transformer is managed by varying the EV load in an optimisation objective function which maximises the capacity utilisation of the transformer subject to operational constraints and non-disruption of daily trips of EV owners. Results show that with the proposed load management technique, LV distribution networks could accommodate high uptake of EVs without compromising the useful normal life expectancy of distribution transformers before the need for capacity reinforcement.

基于变量选择与Transformer模型的中长期电力负荷预测方法

准确且有效的负荷预测对于电力系统的实时运行和调度非常重要。提出了一种融合变量选择与稀疏Transformer模型的预测方法,将静态变量和时序变量作为输入,充分发挥静态变量在全局时间范围内的信息增强作用,基于门控机制设计变量分权组件,根据变量与预测结果的相关性,赋予变量不同的权重。设计了双层编码结构,进行时序特征提取,对注意力进行稀疏处理,通过多变量输入对未来时刻负荷进行预测。基于真实电力负荷数据的实验表明,本文模型能够提高中长期负荷预测精度和效率。

Unbalanced Load Simulation of Electronic Power Transformer in Distribution System

A control scheme of electronic power transformer (EPT) in a three-phase four-wire distribution system, which included an input section, an isolating section and an output section, was researched under unbalanced loads. The simple and appropriate control scheme was developed through analyzing the system requirements of the primary side and the load requirements of the secondary side. In the input section, a dual-loop control in synchronous rotating d-q coordinates was introduced, and in the output section, a dual-loop control based on instantaneous output voltage was used. Load characteristics of EPT were investigated by using Matlab/Simulink software. Simulation results showed that, with the proposed control scheme, the EPT has good performances and the sinusoidal input current and constant output voltage can be realized under both balanced and unbalanced loads.

基于多尺度时空图卷积网络与Transformer融合的多节点短期电力负荷预测方法

深度学习的发展为处理电力系统中海量的负荷数据提供了良好的基础。然而,现有的负荷预测方法大多采用历史负荷序列的时间相关性构建模型,没有同时考虑相邻节点之间存在的空间耦合特性和外部因素的影响。由于图卷积神经网络在挖掘电力系统拓扑结构中的空间特征上具有巨大潜力,因此,该文提出一种基于属性增强的多尺度时空图卷积神经网络与Transformer融合的电力系统多节点负荷预测方法。首先,将外部因素建模为动态属性和静态属性,设计属性增强单元对这些因素进行编码,并利用快速最大互信息系数量化各节点负荷的动态耦合信息。其次,采用多尺度时空图卷积网络挖掘节点间的短期时空特征,同时采用Transformer补充挖掘各节点负荷的长期时域特征。最后,使用门控融合层对两个模型进行融合。在纽约公开负荷数据集上的实验结果表明,所提方法能够充分挖掘多节点负荷数据中的时空耦合特性,具有更高的预测精度和稳定性。

A chaotic hierarchical encryption/watermark embedding scheme for multi-medical images based on row-column confusion and closed-loop bi-directional diffusion

Security during remote transmission has been an important concern for researchers in recent years.In this paper,a hierarchical encryption multi-image encryption scheme for people with different security levels is designed,and a multiimage encryption(MIE)algorithm with row and column confusion and closed-loop bi-directional diffusion is adopted in the paper.While ensuring secure communication of medical image information,people with different security levels have different levels of decryption keys,and differentiated visual effects can be obtained by using the strong sensitivity of chaotic keys.The highest security level can obtain decrypted images without watermarks,and at the same time,patient information and copyright attribution can be verified by obtaining watermark images.The experimental results show that the scheme is sufficiently secure as an MIE scheme with visualized differences and the encryption and decryption efficiency is significantly improved compared to other works.

Investigating Periodic Dependencies to Improve Short-Term Load Forecasting

With a further increase in energy flexibility for customers,short-term load forecasting is essential to provide benchmarks for economic dispatch and real-time alerts in power grids.The electrical load series exhibit periodic patterns and share high associations with metrological data.However,current studies have merely focused on point-wise models and failed to sufficiently investigate the periodic patterns of load series,which hinders the further improvement of short-term load forecasting accuracy.Therefore,this paper improved Autoformer to extract the periodic patterns of load series and learn a representative feature from deep decomposition and reconstruction.In addition,a novel multi-factor attention mechanism was proposed to handle multi-source metrological and numerical weather prediction data and thus correct the forecasted electrical load.The paper also compared the proposed model with various competitive models.As the experimental results reveal,the proposed model outperforms the benchmark models and maintains stability on various types of load consumers.

面向用电负荷分解的特征融合与Transformer模型

针对目前非侵入式负荷分解中存在的特征提取不充分、分解精度较低等问题,本文提出了一种基于特征融合与Transformer的负荷分解模型MulTrm。首先使用滑动窗口法对负荷数据进行处理,增加训练样本数量;接着采用多个不同尺寸的卷积块提取总负荷功率值的多尺度特征并进行融合,同时结合总负荷序列中的位置特征,以获取更加丰富的特征信息;然后通过Transformer中的多头自注意力机制扩大感受野,以更好地捕获用电负荷序列中蕴含的长距离依赖关系,从而提高模型的分解精度;最后通过反卷积层和全连接层将特征映射为电器负荷序列,实现负荷分解。通过在REDD数据集和UK-DALE数据集上进行实验,验证了MulTrm模型的有效性。

基于SF-Transformer的智能教育平台短期电力负荷预测研究

建设智能教育平台是推动教育智能化的一个重要过程,但智能教育平台依赖的人工智能模型在训练过程中会消耗大量电力,因此,开展短期电力负荷预测对建设智能教育平台具有重要意义.针对在考虑多个属性开展短期电力负荷预测时,由于部分属性与电力负荷数据的相关性不强并且Transformer无法捕捉电力负荷数据的时间相关性,而导致电力负荷预测不够准确的问题,基于SR(Székely and Rizzo)距离相关系数、融合时间定位编码和Transformer,提出了一种短期电力负荷预测模型SF-Transformer.SF-Transformer通过SR距离相关系数对影响电力负荷数据的属性进行筛选,选择与电力负荷数据之间SR距离相关系数较大的属性.SF-Transformer采用一种全局时间编码与局部位置编码相结合的融合时间定位编码,有助于模型全面获取电力负荷数据的时间定位信息.在数据集上开展了实验,实验结果表明SF-Transformer与其他模型相比,在两种时长上进行电力负荷预测具有更低的均方根误差和平均绝对误差.

考虑综合需求响应的Transformer-图神经网络综合能源系统多元负荷短期预测

为提高在需求响应情境下,综合能源系统的多元负荷短期预测精度,基于消费者心理学、响应不确定性原理、耦合响应原理,构建了考虑综合需求响应的Transformer-图神经网络(Trans-GNN)预测模型。通过响应不确定性随电价差产生的变化规律和消费者心理学原理,量化在不同概率条件下的电力需求响应结果。通过耦合响应原理,求解包含冷、热耦合响应的综合需求响应信号,最终利用注意力机制将综合需求响应信号引入Trans-GNN预测模型,提高网络模型在需求响应情境下的多元负荷预测能力。算例分析结果表明,该模型能有效地提高预测精度,为计及综合需求响应的多元负荷预测研究提供了一定的理论基础。

一种改进Transformer的电力负荷预测方法

负荷预测是电网系统中很多应用的关键部分,具有重要作用。然而,由于电网负荷的非线性、时变性和不确定性,使得准确预测负荷具有一定的挑战。充分挖掘负荷序列的潜在特征是提升预测准确率的关键。文中认为在特征提取时应该充分利用负荷序列的位置信息、趋势性、周期性和时间信息,同时还应构建更深层次的神经网络框架进行特征挖掘。因此,提出了基于特征嵌入和Transformer框架的负荷预测模型。该模型由特征嵌入层、Transformer层和预测层组成。在特征嵌入层,模型首先对历史负荷的位置信息、趋势性、周期性和时间信息进行特征嵌入,然后再与天气信息进行融合,得到特征向量。Transformer层则接受历史序列的特征向量并挖掘序列的非线性时序依赖关系。预测层通过全连接网络实现负荷预测。从实验结果来看,文中模型的预测性能优于对比模型,体现了该模型的可行性和有效性。