针对城市电网中负荷波动大以及中长期尖峰负荷预测难度高的问题，本文提出一种融合蚁群优化算法与双向长短期记忆神经网络（ACO-BiLSTM）的中长期电网负荷预测模型。该方法通过引入蚁群算法对BiLSTM模型的关键参数进行动态优化，弥补了传统BiLSTM在处理时间序列数据时易陷入局部最优的问题，从而提升了模型在预测精度与稳定性方面的表现。模型能够自适应地调整网络结构参数，寻找更加优质的解空间，在提高模型学习效率的同时，增强对尖峰负荷变化趋势的响应能力。通过对A地区实际电网负荷数据的预测测试结果表明，所提出方法在预测准确率和响应速度上均优于传统BiLSTM模型，其中R²值达到0.99，平均绝对误差（MAE）较原模型下降了7.8%。此外，文中还系统分析了蚁群优化策略对不同时间尺度尖峰负荷预测性能的影响，进一步验证了其在电力负荷时序建模中的有效性。

The medium-voltage AC (MVAC)/MVDC hybrid distribution network attracts many interests for its advantages of high reliability, high power quality, and high compatibility of distributed generations and DC loads integration. Power-electronic-based converter stations are the key enablers to interconnect AC and DC distribution grids. At present, the industry has developed several standard bus configurations of substations that vary in complexity, cost, and reliability. However, the frameworks of the converter stations including not only bus configurations but also converter configurations, are also closely related to the initial cost and service reliability. Thus, referring to bus configurations of traditional AC substations, an overview of MVAC/MVDC converter stations’ frameworks and their corresponding features is presented in this paper. Finally, the application scenarios of each framework are analyzed.

With the development of the subway network, the research on traction network faults is becoming increasingly important. To ensure the reliability of the subway network, better study the grounding point faults, obtain the subway short-circuit current close to the real one, and study the action characteristics of the high-voltage DC circuit breaker used in the subway under the fault current, a DC large current generating device is proposed, which can input the amplitude, frequency, initial phase angle, etc. of the waveform to generate the corresponding waveform. Considering the ripple effect generated by high-frequency switching, a virtual impedance control technology was designed with the inductance of the filter increased, in order to reduce the ripple size and ensure the lightweight design of the equipment. The experimental results demonstrate that this scheme effectively suppresses the ripple size and aligns with expectations.

本文介绍了新建立的兆安级稠密等离子体焦点装置（Dense Plasma Focus, DPF）研制及实验进展。该装置具有MJ级储能的脉冲功率源，通过数十根低电感电缆汇流形成兆安培大电流传输至放电室负载。设计研制了具有同轴结构的Mather型DPF放电室，在百Pa量级低气压条件下初始击穿形成等离子体，并经过轴向加速、径向压缩、内爆箍缩等阶段形成稠密等离子体焦点。通过搭建可见光诊断、探针测量、电参数测试等多参数诊断方法，观察获得了兆安级稠密等离子体运动过程的参数特征和物理特性。该装置可实现兆安培电流加载条件下的稠密等离子体焦点产生，并辐射出高强度的软X射线或中子。同时，该装置还实现了紧凑小型化和可移动，具有良好的可靠性和稳定性。

LC filters (LCFs) play a crucial role in DC-DC converters, serving to filter out high-frequency switching noise and reduce input/output ripples. However, existing LCF solutions in converters rely heavily on a limited number of traditional configurations, with little exploration of new and potentially valuable configurations. To address this limitation, this paper proposes a systematic method to integrate DC-DC converters and second-order LC filters (SO-LCFs). The method begins with analyzing the behavior of inductors and capacitors in SO-LCFs to establish conditions under which an LC pair can function effectively as an SO-LCF. Then, with this condition, a three-step method is proposed to determine how to strategically incorporate an LC pair into a DC-DC converter to achieve effective current filtering without changing the converter's key performance, like voltage gain. The proposed method is successfully applied to Buck, Boost, and Buck-Boost DC-DC converters, resulting in 5, 5, and 6 distinct SO-LCF solutions, respectively. These solutions include not only existing configurations but also uncover several new ones, demonstrating that the proposed systematic method is more comprehensive and offers the potential for more valuable filtering solutions in practical applications.

分散供电的灵活性与浸没式液冷的高效散热，为超算中心提供高可靠、低能耗的供电解决方案。

在全球气候变化挑战下，中国在能源领域积极行动，从减排实践、理念政策、气候研究等多维度应对。能源领域实现绿色崛起，供应链优势凸显；国家战略融合减碳与适应，推进多领域行动并参与国际合作。

新能源场站并网点谐波电流包含场站发射和背景侵入两种分量，背景谐波侵入也会在并网点产生数值较高的谐波电流。目前谐波测试手段只能测试谐波电流数值，无法判定其方向。不考虑背景谐波侵入因素，仅通过谐波电流数值大小与限值进行比较来判定是否合格，可能会得出场站谐波电流超标的错误结论，进而影响场站并网。本文结合新能源场站系统结构和等值电路，研究了其阻抗特性和谐波潮流分布，并提出了新能源场站谐波电流实测超标时谐波溯源方法。当谐波超标由场站发射引起时，应采取治理措施保证本站谐波合格；如通过分析确认是背景引起时，应给出本站谐波合格的测试结论，保证新能源场站早日并网发电，并积极开展谐波溯源，按照“谁污染、谁治理”的原则督促干扰源用户对谐波进行就地治理。

本文研究针对并网式单相光伏逆变系统展开综合技术路径分析，采用双级非隔离式拓扑架构进行设计。其中前级选用Boost升压变换拓扑，后级逆变为全桥逆变回路。在最大功率跟踪环节，通过电导增量算法实现光伏阵列工作点的理论验证；并网控制单元使用了双闭环控制策略，并使用具有高临界击穿电场、低导通电阻和开关损耗的SiC MOSFET器件，该技术能有效提升电网同步稳定性，保障不同工况下的电能质量输出。最后通过对仿真波形的分析，确定该技术的有效性。

碳中和下氢能发展，可再生能源电解水制氢为主，IGBT制氢电源成主流，未来朝高效、离网/直流组网演进。