Search for: All records

This article proposes a matrix auto-transformer switched-capacitor dc–dc converter to achieve a high voltage conversion ratio, high efficiency, and high power density for 48-V data-center applications. On the high-voltage side, the proposed converter can fully leverage the benefits of high-performance low voltage stress devices similar to the multilevel modular switched-capacitor converter. Compared with the traditional isolated LLC converter with a matrix transformer, the proposed solution utilized a matrix autotransformer concept with merged primary and secondary side windings, thus leading to reduced transformer winding loss. The resonant inductor could be integrated into the transformer similar to the LLC converter. Because of the matrix autotransformer design, it can achieve a current doubler rectifier on the low voltage side. For less than 8-V low output voltage application, the current doubler rectifier design can fully utilize the best figure-of-merit 25-V device, which is more efficient than the full-bridge rectifier solution using two 25-V devices during the operation. All the devices can achieve zero voltage switching or zero current switching and can be naturally clamped without additional clamping circuits. A 500-W 48-V to 6-V dc–dc converter hardware prototype has been developed with optimized device selection and integrated matrix autotransformer design. Both simulation and experiment results have been provided to validate the features and benefits of the proposed converter. The maximum efficiency of the proposed converter can reach 98.33%. 

Switched Tank Converter(STC) is one kind of Resonant Switch Capacitor(ReSC) that can be considered as a good candidate for data center application with high power efficiency and high power density. On the other hand, LLC converter can also realize very good performance for low voltage application. Although STC can realize relatively higher efficiency than LLC converter in the light load since the core loss is saved, LLC can keep relatively higher efficiency in the heavy load than STC does since the conduction loss of LLC is smaller. The main reason is because transformer’s winding resistance is smaller than semiconductor devices’ resistance, and this is very important for high current application.In order to utilize the benefits of the STC and the LLC converter together, this paper proposes a family of the novel Switch Capacitor based Integrated Matrix Autotransformer LLC Converters (SCIMAC). The proposed converters share the same high voltage side circuit of the STC with low voltage stress devices. Different from the traditional LLC converter with an isolated transformer, the proposed SCIMAC utilizes one autotransformer with only the secondary side windings similar to LLC's secondary side. There are several advantages that can be realized of the SCIMAC: 1). Low figure of merit (FOM) devices can be adopted to realize higher efficiency due to the low voltage stress of the SCIMAC. 2). Higher power efficiency can be realized when compared with STC converter in heavy load because the resistance of the autotransformer’s windings is lower than semiconductor devices’ resistance. 3). The primary side winding loss of the transformer is saved to further increase the efficiency. 4) ZVS turning on can be realized by the magnetizing current of the core. 

Cascaded Connected Microinverter (CCM) system takes the advantage of adapting low voltage stress submodules to build the high voltage output, which makes it easier and safer to achieve for many applications. The distribution of active and reactive power in the CCM system has always been interdependent, resulting in additional communication components in different submodules. Such communication components within different submodules can be avoided by droop control. Droop control is widely adopted in parallel inverter system, and it originates from the synchronous generator that active power is controlled by adjusting synchronous generator's frequency and reactive power is controlled by adjusting its output voltage. However, the traditional droop control is not suitable for the cascaded microinverter inverter system. Therefore, it's necessary to modify the droop control to make it suitable for Cascaded Microinverter system, and therefore a control method called inverse droop control is adopted for cascaded inverter system under island mode. However, it requires a large feeder inductor when it's grid connected since every submodule works as voltage source inverter. In this paper, a duality control method that feedbacks each submodule's active power and reactive power to adjust its inductor current amplitude and frequency respectively is proposed. Compared with traditional cascaded inverter system that's controlled by inverse droop control method, the big line frequency feeder inductor is saved. 

Recent advances and achievements of artificial intelligence (AI) as well as deep and graph learning models have established their usefulness in biomedical applications, especially in drug–drug interactions (DDIs). DDIs refer to a change in the effect of one drug to the presence of another drug in the human body, which plays an essential role in drug discovery and clinical research. DDIs prediction through traditional clinical trials and experiments is an expensive and time-consuming process. To correctly apply the advanced AI and deep learning, the developer and user meet various challenges such as the availability and encoding of data resources, and the design of computational methods. This review summarizes chemical structure based, network based, natural language processing based and hybrid methods, providing an updated and accessible guide to the broad researchers and development community with different domain knowledge. We introduce widely used molecular representation and describe the theoretical frameworks of graph neural network models for representing molecular structures. We present the advantages and disadvantages of deep and graph learning methods by performing comparative experiments. We discuss the potential technical challenges and highlight future directions of deep and graph learning models for accelerating DDIs prediction.