The multi-chamber arc extinguishing gap has attracted attention in lightning protection of distribution networks. During its operation, the arc sprays out and exchanges heat with air convection to form a gas jet, which forms a shear layer with the surrounding fluid velocity difference, which may evolve into a vortex ring structure, promoting gas cooling and insulation recovery after the arc. The formation of vortex rings is influenced by the chamber structure, and optimizing the structure can increase the probability of vortex ring formation and enhance the arc extinguishing effect. Therefore, the dissipation process of post-arc gas is observed by building a schlieren observation platform. The effects of chambers with different structure parameters on the formation of post-arc gas vortex ring and gas density recovery are investigated, and the effect of vortex ring on promoting post-arc gas cooling is analyzed. Based on the schlieren observation results, the formation process of post-arc gas vortex ring is simulated and analyzed by finite element simulation analysis software. It is proved that the reasonable setting of chamber outlet diameter can increase the probability to form post-arc gas vortex ring, and the vortex ring entrains ambient cold air into the hot kernel of the post-arc gas to promote the cooling and insulation strength recovery of the post-arc gas.

Transformer overload is easy to form bubbles in the oil-paper insulation system and lead to a failure. However, there is only a transformer hot-spot temperature limit of 140℃, which fails to fully consider the effect of the actual operating state on the initial temperature of bubble formation. It is difficult to fully utilize the maximum efficiency of the equipment. To solve this problem, the domestic and foreign tests of the initial temperature of bubble formation of the oil-paper insulation are synthesized and a calculation method for the initial warning temperature of transformer bubble formation applicable to actual working conditions is proposed. According to the operating parameters such as service life of the transformer, moisture content in the oil and the altitude, the initial temperature of bubble formation of the inner oil-paper insulation system of the transformer is estimated. Then, 90% of the initial temperature of bubble formation is determined to be early warning temperature to characterize the maximum load capacity that the transformer can bear. Based on Bayesian network, a short-term load prediction model is established. Then, combined with the calculation method of winding hot-spot temperature, the short-term prediction of transformer hot spot temperature is realized. A transformer short-term overload early warning method is proposed， which is confirmed to consider fully the actual operating condition of the transformers, issue warning in advance for future overload situations, guide relevant operation and maintenance departments to carry out transformer overload shutdown and load transfer work, and reduce the occurrence of transformer shutdown and insulation failure.

Aiming at the difficulty in selecting transformer fault features and the low accuracy of diagnosis model, a hybrid fault feature selection method is proposed, and the deep hybrid kernel limit learning machine （DHKELM） is optimized by the improved northern Goshawk optimization algorithm （INGO） to realize transformer fault diagnosis. Firstly, a 24-dimensional transformer fault feature set is constructed based on correlation ratio method. From the perspective of linear correlation and nonlinear correlation, Pearson correlation coefficient and mutual information method are used to filter out the features with low correlation. Secondly, Logistic chaotic mapping, stochastic reverse learning and adaptive t-distribution mutation are introduced to improve NGO algorithm, so as to improve its optimization performance. Then, INGO algorithm is used to filter the retained features for the second time to obtain the optimal input features. Finally, the automatic encoder of the extreme learning machine is introduced into the hybrid core extreme learning machine, and the DHKELM diagnosis model is established. The initial parameters of the DHKELM model are optimized by INGO, and the INGO-DHKELM transformer fault diagnosis model is completed. Experiments show that compared with the conventional feature selection method, the input features selected by the hybrid fault feature selection method can effectively improve the diagnosis accuracy. Compared with other optimized diagnosis models, INGO-DHKELM has higher accuracy and better stability.

For the operation optimization of hybrid multi-type energy storage, all equipment is usually treated as an equivalent whole, without considering factors such as differences in the operating characteristics of each energy storage system, This approach ultimately results in low equipment utilization and poor economic operation capability. Firstly, an optimization model is established with the objectives of maximizing renewable energy consumption and minimizing grid exchange capacity. Secondly, considering the dynamic characteristics of charging and discharging, operating cost features, and operating states of charge （SOC） settings of multi-type energy storage systems, a two-layer decoupled operational optimization scheme is proposed to adapt to the coordinated operation of multi-type energy storage in distributed energy systems （DESs）, The scheme also incorporates SOC consistency constraints of energy-based energy storage system. Finally, based on data from a distributed renewable energy system industrial demonstration park under construction, the proposed scheme is validated and analyzed through numerical examples. The results show that this scheme can ensure the synchronization of charging and discharging of multi-type energy storage systems within a DES, while also enabling clustering of energy storage systems with the same operating characteristics. This method effectively reduces computational complexity, enables the coordination and optimization of multi-type energy storage, and thereby promotes renewable energy consumption in distributed energy systems.

As a high proportion of renewable energy is connected, the reliable and stable operation of distribution network is affected by multiple uncertain factors. Therefore, an energy storage siting and sizing optimization method is proposed that considers the reliability and vulnerability of distribution network. Firstly, taking into account the uncertainty of wind and photovoltaic power and load, the improved K-means clustering algorithm and synchronous backtracking elimination algorithm are used to obtain the uncertain operation scene set. Then, the idle capacity and power of energy storage are optimized to participate in the energy market and frequency modulation auxiliary service market, and the operation plan of energy storage system is revised. Secondly, an energy storage multi-objective programming model is established, which takes the annual comprehensive cost of distribution network, the expected value of the system power shortage and the comprehensive vulnerability index of distribution network as the optimization objectives. At the same time, considering the availability of energy storage system on the basis of load transfer, the reliability evaluation method of distribution network based on optimization model is used to calculate the reliability index. Finally, the multi-objective differential evolution algorithm based on Tent chaotic mapping and adaptive mutation strategy is used to solve the model. The advantages of the proposed model and algorithm are verified by setting different schemes through the numerical simulation analysis of IEEE 33-node distribution system.

In order to fully excavate the active support capability of distributed photovoltaic （PV） and energy storage system （ESS） and improve the economic benefits and power supply quality of the distribution network, a coordinated optimal strategy for day-ahead and intra-day operation of distribution network considering the active support distributed PV and ESS is proposed. Considering the dispatch cost risk caused by PV and load prediction error, the conditional value-at-risk （CVaR） method is used to quantify the cost risk. Day-ahead optimization model takes the minima of operation cost and cost risk of distribution network as the optimization objective, and formulates day-ahead optimization scheme to provide reference for intra-day optimization. The intra-day optimization model dispatches the active and reactive power of PV and ESS. The minimum weighted voltage deviation, the line loss and the active power deviation are taken as the optimization objective to improve the power supply quality of the distribution network to ensure the economy of intra-day operation as well as improve the quality of power supply in the distribution network. The proposed model is transformed into a two-order cone programming model which is easy to solve, and the proposed strategy is verified by an IEEE 33 node example. Results show that the proposed coordinated optimal strategy fully exploits the active support capability of PV and ESS, effectively reduces the operating cost of the distribution network, and has an obvious easing effect on the voltage violation problem.

The accurate forecast of photovoltaic power is of great significance for the security, stability and economic operation of the power grid. Therefore, a day-ahead photovoltaic power forecasting method is proposed. The method of wavelet transform （WT） is used to decompose numerical weather prediction （NWP） data and photovoltaic power data into frequency data with time information, eliminating the influence of randomness and volatility in data information on forecasting accuracy. Convolutional neural network （CNN） model is used to deeply excavate the seasonal characteristics and spatial correlation characteristics of input data, and bi-directional long-short term memory （BiLSTM） model is used to obtain the temporal correlation of input data series. A day-ahead photovoltaic power forecasting model based on WT-CNN-BiLSTM is constructed. Taking a certain photovoltaic power station as the calculation object, the forecasting results of WT-CNN-BiLSTM model, CNN-BiLSTM model, LSTM model, GRU model and PSO-BP model are compared and analyzed under different seasons and climatic conditions. The calculation results show that the forecasting accuracy of WT-CNN-BiLSTM model is higher than that of other models.

Aiming at the problem that unsteady harmonic parameters are difficult to be accurately detected, this paper improves variational mode decomposition by means of the energy contribution rate of modal components, extractes modal components by adaptive decomposition of harmonic signals, and constructes a new K-M mutual convolution window using Kaiser window and maximum side-lobe decay window. The correction formula of bispectral line unsteady harmonic parameters based on K-M mutual convolution window is derived, and the detection method of unsteady harmonic parameters based on adaptive variational modal decomposition （VMD） and a new K-M mutual convolution window is proposed, and the experimental platform of unsteady harmonic parameters analysis based on digital signal processer （DSP）is developed. Simulation analysis and actual measurement results show that the proposed method can effectively and accurately detect the unsteady harmonic parameters under the interference of noise and fundamental frequency fluctuation. Compared with the traditional harmonic analysis method, the proposed method is suitable for unsteady harmonic analysis and has high precision of harmonic detection.

The mixed integer linear programming （MILP） model for large-scale security constrained unit commitment （SCUC） problem is difficult to solve because of its high-dimensional and non-convex characteristics, especially when considering post-contingency security constraints, the scale of model increases rapidly, and MILP algorithm often encounters the bottleneck of decreasing convergence gap. In order to meet the requirements in spot market clearing process, a fast solving method based on warm-start is proposed to improve the solve speed of SCUC. To reduce the scale of model and speed up the convergence process, the method starts with a feasible solution, integer variables are fixed based on locational marginal price and unit benefit analysis, also the non-binding security constraints are reduced. Simulation results show that the proposed method can greatly reduce the scale of SCUC model. Especially for the large-scale SCUC problem with post-contingency security constraints, the method can effectively overcome the bottleneck of decreasing convergence gap and significantly improve the solution efficiency.

Demand side response can bring considerable benefits to the power grid, and research on economic dispatch optimization methods for power systems that takes into account demand response has become a topic of concern for domestic and foreign scholars. A differential evolution algorithm based on a stochastic black hole model, called the differential evolution hole algorithm （S-DEH）, has been proposed to solve the economic dispatching problem of power systems considering demand-side response. This algorithm bases on the original differential evolution algorithm combined with the stochastic black hole model, which is highly sensitive to initial sequence, stable and reliable, and requires little computation. It can select the best control parameters for different problems, reduce artificial operation errors and the time required to adjust the optimization algorithm parameters, and has high solving efficiency. At the same time, the calculation model adopted in this paper is to solve the social welfare maximization problem（SWMP） of the power system in a single period considering the demand-side response to verify the performance of the algorithm. By solving the SWMP scheduling problem in a multi-period, the algorithm can still maintain efficient performance when dealing with complex optimization problems of high dimensions.

Due to the influence of the resistances and inductances of the tripping coil and relay, as well as the grounding capacitance of DC system, the tripping circuit of the substation circuit breaker forms an RLC second-order circuit after a single point of grounding occurs. The oscillation of the grounding current at one-point may cause the coil or relay to trip incorrectly and cause the circuit breaker to trip. The composition of the tripping circuit and the risk of one-point grounding are analyzed, the influences of the resistance, capacitance and inductance of the tripping circuit on the grounding current within the conventional value range are also analyzed, and the value ranges of the resistance, capacitance and inductance of the tripping circuit are proposed under the condition of ensuring that one-point grounding does not misoperate through methods such as particle swarm optimization and multi-dimensional variable analysis. The necessity of further standardizing the resistance and inductance values of relays, optimizing the operating current of relays, and controlling the equivalent grounding capacitance of DC systems is demonstrated, which provides the guidance and reference for the selection of substation equipment components and the normalized control of DC system grounding capacitance.

Identifying electricity theft user with correlation between electricity usage of user and line loss of associated feeder could facilitate electricity theft detection with low false positive rate. However, most existing approaches have stringent requirement on stability of time series of load data, which hinder engineering application of these approaches. A high-loss line theft user identification method based on segmented dynamic time bending distance is proposed. Firstly, the heuristic segmentation algorithm is used to transform the data of each user′s power consumption sequence and line loss power consumption sequence to achieve feature extraction and data reduction. Secondly, dynamic time bending distance is employed to find out the user′s power consumption most similar to the line loss power consumption pattern, and the linkage between them is analyzed. Finally, the corresponding user corresponding to the power consumption in the most similar pattern of line loss and the same fluctuation direction is designated as the suspected user of electricity theft. Based on the actual data of high-loss lines, the simulation results show that the proposed method has better accuracy and lower false positive rate than the comparison method.

During the audible noise test of UHV HVDC transmission lines, the sudden interference of the external environment will make the experimental data doped with more invalid data, which seriously affects the subsequent data analysis. In this paper, a method based on attention mechanism (AM) and long short-term memory network-light gradient boosting machine (LSTM-LightGBM) is proposed to clean the invalid data of the transmission lines with audible noise. Firstly, feature extraction is carried out based on LSTM neural network, aiming at the characteristics of nonlinear and high-dimensional temporal redundancy of audible noise data. At the same time, the feature dimension attention mechanism is introduced, and the weights are allocated adaptively to describe the expressive ability of key feature information. Then, LightGBM is used to classify the extracted features and detect invalid data. Then, the measured audible noise data of an UHV HVDC transmission line is analyzed experimentally. The results show that the detection accuracy rate of this method is 95.55%, the recall rate is 97.73%, and the score of F1 is 0.9663, which are superior to the comparison experimental model. Finally, the invalid data is deleted and filled with the mean interpolation method. After the invalid data is cleaned, the 50% value and 95% value of the data remain basically unchanged. Only the maximum value and 5% value of the invalid data are reduced. This method has certain reference significance for improving the reliability of audible noise data of transmission lines.

Compared with traditional measurement methods such as current and voltage, the image of the partial arc on ice surface captured by schlieren technology is more intuitive. Its image processing and refractive index field reconstruction are the basis for quantitative analysis of physical characteristics such as geometric size, density and particle distribution of discharge. Based on the optical deflection characteristics of the discharge channel, the principle of measuring the discharge channel by schlieren imaging technology is introduced. A set of image processing and refractive index field numerical reconstruction method for the discharge channel on the icing surface is proposed. The accuracy of the method is verified by numerical experiments. The results show that the designed schlieren system has a spatial resolution of 10 ~ 100 μm / pixel and a temporal resolution 1 ~ 38.45 μs. The proposed refractive index field reconstruction method has high accuracy and can effectively obtain the refractive index field in each discharge stage of ice partial arc formation.

In order to solve the problem of direct survey of cable well caused by water, a direct survey technique for the internal state of cable wells using sonar imaging based on the remotely operated vehicle（ROV） platforms is proposed. The terminal sliding mode control is used as the closed loop control of ROV. By using the particle swarm optimization algorithm to optimize the parameters and eliminate the influence of the disturbance observer on the uncertainty, the accurate control of the ROV 's walking position in the cable well is achieved, and the sonar imaging technology is used to realize the direct observation of the cable laying state in the cable well. The test results show that the improved sliding mode control model can realize the accurate control of the ROV's walking position, with the control accuracy of 0.1 m. The sonar imaging sensor can realize the measurement of the cable distribution status and cable size in the cable well, with the measurement accuracy better than 5%. The proposed sonar survey technology for cable wells based on ROVs can directly survey the internal state of cable wells, greatly reducing the survey work of operators and having high practical application value in production.

The electrical characteristics of doubly fed induction generators （DFIGs） are easily affected by changes in power grid voltage throughout the entire process of low voltage ride through （LVRT）. However, existing researches mostly focuse on the electrical characteristics of DFIGs during faults, without considering the differences in electrical characteristics of DFIGs in different scenarios after fault clearance. There is also limited research on the power characteristics of DFIGs throughout the entire process of LVRT, and the analysis of electrical characteristics throughout the entire process of LVRT is not yet comprehensive. For this purpose, the operation status of DFIG in the entire LVRT process is studied. Before the fault is cleared, the entire process of LVRT is divided into two stages: fault occurrence and reactive power priority control startup. After the fault is cleared, it is analyzed in two situations: secondary use of the pry bar and active power ramp recovery. On this basis, the expressions for the stator side output current and power throughout the entire process of LVRT are derived, and a DFIG simulation model is established in PSCAD. The time-domain simulation results closely match the analytical expression waveform, demonstrating the effectiveness of the proposed method.

To address the poor output issues of voltage quality and dynamic performance of inverters under nonlinear loads, a multi- voltage harmonics (MVH) control strategy is proposed based on traditional dual-loop control. Firstly, the mechanism of output voltage distortion in a three-phase four-wire inverter with dual-loop control under nonlinear loads is revealed, and the distribution characteristics of the output voltage harmonics in the inverter are analyzed. Secondly, a MVH control method is proposed for each phase inverter, in which each low order voltage harmonic is independently controlled in the corresponding synchronous coordinate system. Based on the characteristics of different voltage harmonics, a method for constructing the 3rd harmonic dq component of voltage based on the decoupled dual synchronous coordinate system and the 5th and 7th harmonic dq components of voltage based on the decoupled multi synchronous coordinate system are proposed to eliminate the AC component of each dq axis signal and improve the performance of voltage harmonic control. Then, a unified mathematical model of each voltage harmonic control loop is established in the static coordinate system, and the design method of the MVH controller parameters is analyzed. The output impedance characteristics and robustness of the proposed inverter control system are discussed. Finally, a 200 kVA T-type three-level inverter platform is built to verify the correctness and effectiveness of the proposed MVH control strategy by experimental results.