Electrically assisted manufacturing （EAM）is a promising and rapidly developing metal processing method.The power supply is a key sub-system for EAM，which needs to be designed properly.the model-based design of a low-voltage high-current pulse power supply used for EAM was proposed based on converter-level electro-thermal modeling.The thermal stress of key components was obtained by converter-level finite element simulations.A simplified thermal modeling method was proposed to reduce the computation burden of the finite element modeling（FEM） simulation to obtain the dynamic thermal profile under pulse current operation.The impact of the duration of the current pulse on the maximum temperature and temperature variations of MOSFETs was investigated based on the thermal model.A case study of a 10 V/500 A pulse power supply was presented to demonstrate the theoretical analyses and verification. The outcomes contribute to the design optimization and virtual prototyping of pulse power supplies for EAM applications.

A bidirectional isolated resonant converter with variable mode and wide voltage range was proposed. The proposed converter has the advantages of BOOST circuit and flyback current feed topology. By reusing the pre-storage and flyback energy feedback of the resonant inductor，it can improve the gain from the battery side to grid side and the power density of the converter，while also reducing the current circulation. The converter adopts fixed frequency pulse width modulation and has multiple working modes，which is suitable for wide range output. the working principle of the converter was first introduced and the output gain when the energy flows in different directions was analyzed. Finally，on the premise of meeting the charge and discharge of the battery in the home distributed energy storage system，the devices and control parameters were designed to verify the correctness of the theoretical analysis.

Aiming at the problem of large output voltage fluctuation and long recovery time of three-phase pulse-width modulation （PWM）rectifiers when the load changes，an improved single-neuron gradient learning control strategy was proposed. Due to the poor adaptability of the traditional PI controller parameters when the load changes，a single neuron PI control was adopted in the voltage outer loop，and the gradient descent method was used to adjust the weight parameters online. In order to avoid falling into a local optimal solution during the solution process，a stochastic gradient descent algorithm with restart function （SGDR）was used，and cosine annealing was used to change the learning rate of the weights to improve the convergence performance of the algorithm. Through Matlab and hardware-in-the-loop simulation experiments，the dynamic response performance of the voltage outer loop of the three-phase PWM rectifier under different control algorithms was compared and analyzed. The results show that the three-phase PWM rectifier controlled by the improved single neuron PI algorithm has smaller voltage fluctuation，faster dynamic response and more stable operating state when the load changes.

In order to analyze the power quality problem of actual power network under the influence of uncertain interference factors，a power quality detection and recognition method combining empirical wavelet transform（EWT）and improved S-transform was proposed. On the one hand，the frequency，amplitude and time parameters of the AM-FM component were accurately extracted by using the EWT joint normalization direct orthogonal（NDQ）algorithm and singular value decomposition（SVD）algorithm. On the other hand，considering the instantaneous amplitude fluctuation of the EWT algorithm in the high noise environment，the improved S-transform was introduced to extract the time-frequency information of power quality disturbances under the high noise interference. Finally，based on the disturbance feature vectors extracted by EWT and improved S transform，the power quality disturbance recognition classifier optimized by the support vector machine（SVM）based on improved particle swarm optimization（IPSO）algorithm was used to accurately identify the disturbance types. Simulation and experiments show that the average recognition accuracy of the proposed method is 93.23% in the case of composite disturbance recognition and classification，and it can accurately identify four kinds of measured disturbance signals.

In the grid load control and planning of distribution network，in order to optimize the grid structure of distribution network，such as line loss，power supply quality and load，a grid load control and planning technology based on differential evolution algorithm was proposed. For all kinds of power generation devices in the distribution grid，the power grid planning model was established，the initial population was set and constructed，and the appropriate differential initial vector was selected for mutation processing. On the premise of forming a new individual under the difference increment，in order to improve the diversity of the population of future generations，the differential individual was cross trained by using cross factors to form a new distribution individual. Finally，the greedy selection model was used to evaluate and retain the offspring population，control the active power of controllable load and uncontrollable load，and complete the grid planning. The experimental results show that the load control and planning of distribution grid using differential evolution algorithm are better，the robust performance is superior，and the iteration speed is faster.

AC-DC hybrid power grid can balance the power flow during the operation of the power system in a large range，which is conducive to improving the access capacity and access range of large-scale access of new energy to the power grid，which is an important trend in the development of modern power grid. In order to analyze the structural vulnerability of AC-DC hybrid system and avoid the occurrence of power grid outage，a rank-sum ratio （RSR）method was proposed to analyze the structural vulnerability of power grid. Firstly，the vulnerability index set was established based on the structural characteristics of the complex networks. Secondly，the RSR method combined with the subjective and objective evaluation method was used to obtain the comprehensive weight value of node vulnerability. Finally，to verify the validity of the proposed method，AC-DC mixed with EPRI-36 node system node vulnerability analysis based on an example，the results show that the method is feasible.

Reasonable planning of the active distribution network is an important part to improve the wind energy accommodation capability，however，the overuse of the wind power output and load timing characteristics increase the difficulty of model solving and have adverse effects on the optimal results. The Latin hypercube sampling （LHS）combined with the K-means clustering was employed to reduce the number of samples，thus a typical wind power and load multi-scenario model with higher calculation efficiency can be obtained. Considering the interests of wind power operators and the State Grid Corporation，a bi-level planning model of active distribution network considering the wind power timing characteristics was established. The upper level determines the wind power planning scheme with the goal of maximizing benefit of wind power operators，and the lower level optimizes the system operation state with the minimum loss of distribution network. The effectiveness verification of the planning mode was conducted based on the IEEE 33-bus distribution system. The results show that the loss cost of the distribution system is 260 400￥after planning based on the GA-PSO joint optimization algorithm，which is 5.03% and 0.77% lower than that of single GA algorithm and PSO algorithm respectively，and the scenario cost is reduced by 40 000￥compared to that of the results calculated by GA algorithm and PSO algorithm. Therefore，the validity of the planning model proposed was verified.

In view of the fact that the original high-dimensional nonlinear power flow model cannot be applied to the linear planning of distribution network，and the existing linear power flow model has the problem of weak universality，a calculation method of distribution network linear power flow was proposed considering the static characteristics of load voltage and PV nodes. Based on the power flow equation in polar coordinates，the proposed method decoupled the voltage amplitude and phase angle of the power flow equation using the characteristics of the distribution network. According to the control characteristics of PV nodes，a linear power flow calculation model with PV nodes was derived. The proposed model not only considered the static characteristics of PV nodes and load voltages，but also considered the adaptability to overload and weak loop networks. It could solve the voltage distribution of distribution networks without iteration. The simulation results show that the proposed method has high accuracy and versatility，and can be used for rapid analysis of distribution networks.

Under the background of limited resources and time，equipment maintenance is considered as an effective way to maintain the stable operation of the system. In order to allocate maintenance resources efficiently，the most critical equipment in the system should be identified first namely those that will cause significant consequences if they fail. Firstly，a new multicriteria decision-making（MCDM） scheme was proposed to identify the critical lines in the distribution network. Different from the previous analytic based hierarchy process，the best-worst method （BWM） was adopted to obtain the weight of the system reliability index according to the knowledge and judgment of experts. In addition，the fuzzy theory was introduced into the traditional best-worst method to overcome the general uncertainty in expert judgment and decision making. Finally，the technique for order preference by similarity to an ideal solution technology was used to prioritize the maintenance of IEEE14 distribution network lines. The proposed method can determine the priority of system maintenance more quickly and accurately.

With the accelerated construction of new power system，the harmonic characteristics of power grid are becoming more and more complex. It is of great significance to study the effective statistical management of harmonic data for evaluating the power quality of power grid. A statistical method of harmonic evaluation index based on maximum entropy principle was proposed. By recording and saving the average value and center distance of harmonic data，the maximum entropy principle was used to fit the probability distribution of harmonics，so as to save and identify the harmonic characteristics and facilitate data storage. Two harmonic evaluation criteria of harmonic 95% probability value and 99% probability value were obtained by using the fitted probability distribution，which ensures the accuracy and consistency of the index. Finally，the effectiveness of the proposed method was verified by the example analysis of harmonic measured data.

The state of the DC system of substation is directly related to the normal operation of the substation. A new method for ground fault detection in substation DC systems，which is a combination of double-tree complex wavelet transform and singular value decomposition，was proposed to achieve fast and accurate location of ground faults occurring in substation DC systems. Firstly，the method constructed a Hankel matrix to decompose the branch current signal through a dual-tree complex wavelet transform（DT-CWT）. Secondly，the Hankel matrix was decomposed by the singular value decomposition（SVD）method with the aim of obtaining a series of singular eigenvalues. Thirdly，the singular value difference spectrum was constructed using the adjacent singular value differences，and the number of singular values was retained by the maximum peak of the singular value difference spectrum. Finally，the low-frequency signal was reconstructed by the retained singular values. The analysis results of the algorithm show that the method can accurately extract the low-frequency AC signal from the branch current signal and achieve the accurate location of the DC system ground fault in the substation，which can largely reduce the influence of the ground capacitance on the detection accuracy.

The power transmission and transformation system is an important link of power transmission. In order to reduce the cost of operation and maintenance and improve the recognition effect，an application method of automatic identification and control of transmission and transformation vulnerable lines was proposed based on artificial intelligence （AI）technology. Using the greedy decision tree algorithm （ID3）of AI to measure the amount of information between different operation characteristics of the line，it got the line characteristics，introduced mutual information feature selection （MIFS）to optimize the decision tree，balanced the redundancy between the line operation characteristics by using the adjustment coefficient and penalty term，and introduced the weighted degree，measured the node weight and focused on the total active power of all connecting lines. Based on the comprehensive analysis of the permittivity and weighted degree，the comprehensive automatic identification index of vulnerable lines was obtained，the identification model was established，and finally the power flow transmission and load change of transmission and transformation lines were changed to achieve control. The experiment shows that the proposed method can accurately identify the location of vulnerable lines，which is effective and feasible，and its control application also effectively reduces the operation and maintenance costs.

In order to improve the debugging and fault monitoring and analysis capabilities of inverter products，the design scheme of network version process data acquisition and analysis software was proposed.The composition and implementation of the software system were discussed，and the technical means such as multi-thread，virtual memory，drawing class library and high-speed fiber synchronization were proposed to solve the technical problems of system operation. The field application of this software system shows that the system runs stably，has the characteristics of low cost，multi-channel transmission，low sampling period，high data accuracy，rich graphics functions，and strong scalability.