The main circuit faults of the four-quadrant rectifier and the corresponding influences were analyzed and summarized，focusing on the analysis of the voltage and current parameter equations in the pre-charging process of the intermediate DC link，and the estimation method of the converter input current and the intermediate DC link voltage was proposed. The estimated parameters were compared with the actual detection results，and the corresponding fault detection can be carried out by using the performance of different faults in the pre-charging process. Through simulation and experiment，different faults were simulated and analyzed，and the results show that the fault detection method proposed is effective and feasible.

Modular multi-level matrix converter（M3C）is a kind of new topology which can convert one specified electric frequency to another directly，it is applicable to high voltage and large capacity scene. In the fractional transmission system，the complete flow was introduced to achieve power frequency grid fault ride-through. The arm power state equation of M3C in double α-β frame was introduced when power frequency voltage and fractional frequency voltage is unsymmetrical. The control strategy of switching on braking resistor in the fractional frequency side based on the average voltage of all sub-module capacitor was introduced，in order to consume redundant power in the fractional frequency side. Positive and negative sequence reactive current based on the power frequency positive voltage dips and negative voltage component were output. In addition，when power frequency positive sequence voltage is too low，diagonal balancing control was converted to the fractional frequency side from the power frequency side，which can realize the balance of all arm sub-module capacitor voltage in the process of power frequency grid fault ride-though. At last，a 3.3 kV/3.3 kV/3 MV·A M3C model was built in Matlab/Simulink，the effectiveness of power frequency grid fault ride-through strategy was validated.

Compared with large photovoltaic systems with isolation transformers，distributed photovoltaic systems have small power and volume. In order to improve efficiency，non-insulated inverters are often used in design，but photovoltaic leakage current endangering system safety is also generated. Taking the single-phase full-bridge photovoltaic system as an example，the generation mechanism of photovoltaic leakage current was analyzed in detail through in-depth discussion of the problem of photovoltaic leakage current. Subsequently，the waveform characteristics of the photovoltaic leakage current entering the distribution network were analyzed，and a photovoltaic leakage current monitoring method based on Hausdorff distance was proposed，which simply and efficiently solved the problem of photovoltaic leakage current monitoring and protection. Finally，the feasibility of the proposed method was verified by simulation analysis.

In response to the problem of slow response speed of main steam pressure in the coordinated control system of thermal power units，the control effect of the coordinated control system was improved by blowing out some of the stored powder in the pulverizing system. Based on the traditional milling system model，considering the influence of primary air flow，a milling system model based on primary air flow was established，which has been verified to reflect the dynamic characteristics of primary air flow on the milling system. On the basis of the improved model of milling system，a primary air flow dynamic compensation system was designed，which combined with the original coordinated control system of the unit to form a coordinated control system based on the primary air flow dynamic compensation，which realized the utilization of powder storage inside the coal mill. The simulation results show that the designed system effectively utilizes the stored powder in the coal mill by changing the primary air flow rate. On the premise of ensuring rapid response to load changes，it has a good control effect on the main steam pressure，improves the response speed of the main steam pressure，and improves the control quality of the coordinated control system.

Wireless power transfer（WPT）load voltage，current and other state information need to be sent to the original side for closed-loop control and monitoring processing，the secondary side to the original side of the communication is very important. Firstly，the topology and working principle of capacitive modulated WPT system were introduced. With or without capacitance modulation，the system topology was changed，the original and secondary side topologies were series/series（S/S）structure when unmodulated，and the original and secondary side topologies were series/series-parallel（S/SP）structure during modulation. The switching of the modulation capacitor formed two states to transmit communication information，and the change of the state can be reflected in the primary current. Secondly，the amplitude-shift keying（ASK） demodulation circuit was used to process the primary current to restore the communication signal. In order to improve the demodulation accuracy and anti-interference ability，the ASK demodulation circuit was improved. Finally，the effectiveness of the scheme was verified by simulation and experiments，and the experimental results show that the system can accurately complete communication from the secondary side to the primary side at a power of 100 W and a communication frequency of 2~5 kHz.

To address the issue of over/under voltage in the distribution network due to the scale access of distributed generators and flexible loads under the "double carbon" target，an active distribution networks cooperative voltage control strategy based on the active support of electric vehicles（EVs）aggregation was proposed，and the regulation effect of flexible load aggregation was fully utilized to support the safe and efficient operation of distribution networks. First，an EVs aggregation prediction model based on the concept of "travel chain" was proposed to improve the accuracy of EVs aggregation prediction. Then，aiming at the voltage problem of multiple nodes，according to the voltage-power sensitivity，a coordinated voltage regulation strategy based on "EVs aggregation，photovoltaic multifunctional inverter and reactive power management equipment " was proposed to promote the effective regulation of node voltage，reduce the configuration cost of reactive power control equipment，and ensure the safe power supply of the system. Finally，through the comparison of different methods and multi-scenario tests，the validity of the proposed EVs aggregation prediction method and the coordinated voltage regulation strategy was verified.

Aiming at the inherent problems of large common-mode voltage and heavy computation in model predictive current control of the six-phase permanent magnet synchronous motor（PMSM），a model predictive current control method for an open-end winding six-phase PMSM was proposed. Firstly，the neutral point of the six-phase PMSM was opened and two six-phase voltage source inverters were connected to form an open-end winding structure fed by a dual inverter structure，in order to offset the common-mode voltage between inverters. Secondly，a wide range of voltage vectors were fed by a dual inverter，from which the vector with zero common-mode voltage property was selected. On this basis，the harmonic current in the x-y plane and the utilization rate of bus voltage were considered，and the vector set was reduced to a control set containing seven voltage vectors，which reduces the computational burden. Finally，compared with the existing model predictive control method of traditional six-phase PMSM，the effectiveness of the proposed method was proved.

Under the change of island operation conditions，the microgrid will cause that the digital twin（DT）model cannot accurately match the source domain model when migrating，which will lead to low migration efficiency and other problems.For this reason，an adaptive migration method of digital twins model under the scenario of micro grid island operation condition changes was proposed. By analyzing the operation characteristics of the microgrid under the island operation condition，the digital twin model was established，and the time-varying operation condition of the microgrid island was used to match the calculation source domain model. The adaptive migration of the digital twin model was realized by reducing the distribution difference of the source domain data under different island operation conditions of the microgrid. The experimental results show that some characteristic data such as current load and actual current have high migration effect and high migration efficiency，which verifies that the proposed migration method can adapt to different micro grid island operation conditions，and has good practicability.

Under the power grid dispatch mode of "source follows load"，different AC power supply modes are required in the market. To meet market demand，an uninterruptible power supply（UPS） system compatible with various AC power supply modes was designed. The modes of the system were selected by switching devices. According to the selected mode，different methods of driving control were carried out，and thus the compatibility of various AC power supply modes was realized. To improve the reliability of uninterruptible power supply，a fault-tolerant mechanism with compatible modes was also proposed. Finally，an experiment was conducted on a 40 kV·A/36 kW prototype. Experimental results demonstrate that the system can be compatible with various AC power supply modes，by which the correctness of the design was verified.

In order to improve the accuracy of intelligent diagnosis of reactor mechanical fault，according to the correlation characteristics between reactor vibration signal and mechanical state，a vibration diagnosis method of reactor mechanical fault based on stacked auto-encoder（SAE） was proposed. Firstly，the original vibration signal of reactor was decomposed by wavelet packet decomposition algorithm，and the time-frequency energy matrix of the signal was extracted. Then，the diagnosis model of reactor mechanical fault based on SAE was built，the deep feature mining of the time-frequency energy matrix was completed through unsupervised self-learning，and the identification of reactor mechanical fault was realized through supervised fine-tuning. Finally，vibration data of 10 kV oil immersed reactor under different mechanical states was used to train the fault identification model and optimize the super parameters. The numerical results show that the proposed method can identify reactor mechanical fault better than the traditional vibration signal identification method，and the accuracy can reach 98%.