Dc power distributions have been the trend to replace conventional ac distributions in data centers for their reduced conversion steps and higher efficiency, and 48 V VR has been proposed in recent years to reduce the large bus-bar I2R loss. However, conventional 12 V VR is still widely used for its mature design, and when replacing the 12 V VR with the 48 V VR, the converter before the VR usually needs to be redesigned for different output voltage, which results in a high cost during the update iteration. In this paper, a high efficiency high power density series resonant converter (SRC) dc transformer (DCX) with modular structure and different output voltage adaption is proposed. By using low voltage devices with low NFoM (NFoM=Rdson·Cotr), the conduction loss of the MOSFETs can be reduced. A chain series-parallel structure is proposed for current and voltage balancing and both 48 V and 12 V output voltage can be achieved with the same components. With the same key parameters, 1 MHz380V/48V and 380 V/12 V SRC DCX prototypes are built by only changing the connection structures. Peak efficiency over 98.0% is achieved and the power density attains more than 800W/in3under a full load of 1800 W.

The intermediate dc bus voltage in modern data center backend power supply is evolving from conventional 12V to 48 V. It still requires the voltage regulator modules(VRM) to feed the terminal loads such as memory and computing units operating with very high current (> 100 A/module) and very low logic voltage (0.8 V-1.8 V). This makes it challenging to optimize the design of load-side VRMs with quadrupled input voltage. This paper comprehensively reviews the state-of-the-art 48 V VRMs and categorizes them according to passive component utilization. The first category is inductive solution which is further divided into coupled-inductor-based converters and transformer-based converters. The second category named capacitive solution is fur-ther divided into resonant switched-capacitor-based converters(Resonant SCC) and hybrid switched-capacitor-based converters(Hybrid SCC). Typical topologies are discussed， analyzed and summarized to perform a comprehensive performance compari-son， such that the characteristics of different VRMs can be man-ifested. Some design considerations are also given to facilitate the design of the practical prototypes. Moreover， opportunities and challenges in the future data center power system are presented to provide technical insights.

Lightning is one of the main causes of voltage sags, and it is of great significance for subsequent analysis and governance to accurately evaluate the severity of the sag events caused by lightning. There are many uncertain factors between lightning fault events and voltage sag events. To evaluate the severity of voltage sag events caused by lightning, a data-driven self-learning evaluation method for voltage sag severity is proposed. According to a large number of online monitoring data of lightning positioning system and power quality monitoring system, the association rule mining algorithm based on incremental learning is used. And the rules are kept updating through the accumulation of historical data, which may give it the abilities of self-learning. The empirical analysis is carried out based on the monitoring data of a regional power grid. The results show that the method in this paper can accurately mine more valuable rules in reality and solve the problem of low efficiency of mining algorithm when the database changes dynamically.

The switched-capacitor multilevel inverters(SC-MLIs) are the popular type of multilevel inverter. This kind of inverter topology uses the on-off states of switches to control the charging and discharging of capacitors to achieve multilevel out-put. Most SCMLIs make use of an H-bridge structure to change the polarity of the output voltage， which cause the switches to withstand the peak of the output voltage. The H-bridge is replaced by two half-bridges on both sides of the proposed inverters， and the maximum voltage stress (MVS) on switches in half bridge is kept within 2Vdc， as well as in the extended structure. Therefore， the voltage stress of the switches is greatly reduced. In addition， the topology has a modular structure， which makes the expansion and modulation of the topology simple， while achieving a higher voltage gain. Moreover， with the growth of output levels， the MVS of the switches in the topology remains unchanged， which has good practical application scenarios. In this study， the correctness and feasibility of the topology have been verified by experiments.

This paper analyzes and develops a multi-variable& multi-constraint design optimization approach with the goal of minimizing power losses in a 2 MHz LLC resonant converter for next-generation data center applications. For a thorough co-design of a multi-MHz resonant converter, intricately curated performance constraints and associated design-based trade-offs are presented. In addition, accurate characterization, and parametric minimization of the AC resistance by optimal selection of transformer winding configuration, while achieving a controllable leakage flux for the resonant inductor integration into the high frequency planar transformer (HFPT) thereby reducing the effective winding losses by 6%. An all-GaN based 700 W, high power density(6.2W/cm3) experimental proof-of-concept was built for a conversion from a variable input bus voltage (380-420 V) to 12 V stiff output at a resonant frequency of 2 MHz. The results portrayed a steady state peak efficiency of 95.65%, with an improvement of 2.2% over the state-of-the-art (SOA) operable at MHz frequency.

Multi-resonant switched capacitor converter can make efficient use of active and passive components， and has two characteristics of high efficiency and highly power density. Therefore， we propose a 9：1 cascaded multi-resonant switched capacitor converter and further explore ways to improve the performance of the converter in this paper. On the one hand， by analyzing the coupling relationship between the first and second circuit topology， we propose a method to reduce the intermediate decoupling capacitance. On the other hand， by adjusting the dead time of the control signal， the zero-voltage switch (ZVS) of most switching devices is realized， and the efficiency of the converter is improved. Therefore， a 48 V-5 V resonant converter prototype with rated power of 120W，power density of 330 W/in3，peak efficiency of 98.1% and maximum output current of 23.7 A is designed in this paper. From 20% to full load， the efficiency is al-ways maintained at more than 92% (including driving loss)， and most of the loss is reflected in the conduction path， reflecting great optimal space and application potential.

When the switching frequency is above 100 kHz， the winding loss of the round wire magnetic component will increase sharply due to skin effect and proximity effect， which greatly limits its wide application. On the contrary， the planar magnetic component can not only work at such high frequency， but also suitable for high power density applications for its low profile. However， in order to avoid core saturation and change the inductance， the air gap is usually required in the design of the planar magnetic component， which will produce fringing effect and further increase the winding loss. Based on Roshen's assump-tions， this paper analyzes the planar magnetic component with center air gap. Firstly， the Schwartz-Christoffel mapping is used to establish the mapping relationship between the air gap fringing region and the strip region. Then the magnetic field intensity in complex form is obtained by the complex potential function and the one in vector form is given， which is represented by the strip region coordinate. Finally， the magnetic field intensity verification and the thermal verification are carried out. The method pro-posed in this paper provides effective guidance for the design of planar magnetic components operating at high frequency.

Cascaded Isolated DC-DC Converters (IDCs) is a popular topology for battery energy storage system in data center application with the advantage of galvanic isolation， higher effi-ciency and independent control capability. However， the absence of normal operation under unbalance conditions among battery modules is a big issue. To this end， considering the voltage-power relationship and limited voltage gain of IDC， a distributed multi-mode control is proposed in this paper. Each IDC can be switched flexibly between multiple operation modes to extend the range of system optimization operation. Individual power regulation and autonomous distribution of output voltage are achieved without real-time communication under the proposed control strategy. Meanwhile， boundary operation conditions are analyzed in detail. Furthermore， a parameter design procedure for cascaded IDCs based battery energy storage system is also proposed considering the upper/lower gain limits of IDCs and three kinds of boundary operation conditions. In this way， modular design is achieved and the operation efficiency and robustness are improved. A simula-tion system and a down-scaled experimental system is constructed to verify the effectiveness of the proposed control strategy.

A new isolated bridgeless PFC converter based on active-clamped SEPIC is proposed. Compared to conventional active-clamped SEPIC PFC converter， the proposed converter reduces conduction loss by removing two diodes in the line cur-rent path， while preserving the advantages of the active-clamped SEPIC such as zero-voltage switching (ZVS) and active voltage clamp for all the high frequency switches， soft turn-off of the out-put diodes and well damped control-to-input-current response. In addition， the peak voltage stress on the output diode is inherently clamped by the center-tapped transformer configuration. Exper-imental results from a 3.3 kW prototype verified the operation of the converter and demonstrated peak efficiency of 97.26%with 140 kHz switching frequency. The power density of the sin-gle-phase PFC module is 101 W/in'。

As the number of internet data centers (IDC) increases， the energy consumption of modern data centers is growing rapidly， which takes an increasing part of the global electricity consumption. In implementation， to cover the electricity consumption and reduce the electricity bill， data centers are often deployed with renewable energy， energy storage system (ESS) and conventional generators， which has brought the characteristics of both flexible electricity load and micro grid to the data centers. Therefore， the data center micro grids can adjust its operations according to the real time electricity price and cut down the operational cost. However， the rising of electricity will force the data center to start up the conventional generators， which leads to the incensement of operational cost and carbon tax. In view of this， we propose a coordinated operation scheduling method of data centers and electricity retailers based on cooperative game theory. In this method， a cooperative game union comprised of the data center and the electricity retailer is established to minimize the total cost. And the extra income of the alliance is distributed using the Shapley value method. A simulation is launched on Gurobi software， and the results have illustrated that the proposed method can reduce the data center operational cost by 5%.