Abstract:The fault identification of transformer based on the characteristics of dissolved gas contents in the oil is of great significance to its safe operation. The deep belief network (DBN) is selected as a fault identification model in this paper since it has unique for extracting features from sample data. Firstly, the original data set of dissolved gas in transformer oil is directly deployed as the inputs of the training model, and the DBN is processed through three intelligent search algorithms.Three important parameters of mid-batch processing, gradient descent learning rate, and number of neural units are intelligently optimized to solve the problem of a low fault recognition rate when input raw sample data are limited. It is shown that the performance of the proposed method is better than the particle swarm search (PSO) algorithm and genetic algorithm (GA) search optimization. The total recognition rate of CS-DBN is 4.2% higher than that of GA-DBN, 2.5% higher than PSO-DBN and 56.2% higher in evolution efficiency. It also has a good generalization performance.

Abstract:The extraction of abnormal heating regions in infrared images is the important prerequisite for the intelligence diagnosis of thermal state in the electrical equipment. For cable terminations, an automatic extraction method is proposed in this paper. Firstly, an adaptive wavelet threshold denoising method based on Maximum a Posteriori Estimation (MAP) is applied to remove the noise and improve the quality of infrared images. Then, the cable terminations in the images are identified and located by the deep learning network, and the interference information is eliminated. Finally, the Mean-Shift algorithm is employed to cluster the pixels of cable terminations. The abnormal heating regions are extracted on the basis of clustering results. It is shown that the proposed method is suitable for infrared images at different backgrounds and different shooting angles. After identifying and locating the cable terminations, the overheating regions can be extracted accurately. Comparing with some existing methods in efficiency and accuracy, the proposed method achieves a better performance.

Abstract:The PD pattern recognition is an important part of the insulation state evaluation of GIS. For the purpose of the accurate and efficient identification of PD signals, a new method of PD signal pattern recognition is proposed based on CNN optimized by the dual attention mechanism in this paper. Firstly, the GIS PD test platform is built, and four typical defects are set up in GIS chamber. The PD signals of different defects are collected by the UHF and ultrasonic detection respectively. Then, the data preprocessing is carried out based on the characteristics of the data obtained by methods mentioned above respectively. The feature space composed by the image features of UHF PD spectrum and the gram angle field density distribution of ultrasonic signal are constructed. Finally, the input image is extracted through the method of convolutional neural network optimized by double attention mechanism, and the results are predicted by a softmax classifier at the end of the network. It is shown that 97.57% recognition accuracy can be achieved by the fusion algorithm, which is higher than the recognition rate considering the single feature. The convolutional neural network optimized by the double attention mechanism is superior to the common algorithm in the aspects of the recognition rate, training speed and robustness.

Abstract:In order to improve the accuracy of load modeling to meet the requirements of power system simulation calculation accuracy, this paper proposes a load model parameter identification strategy based on the grey wolf optimization (GWO) algorithm from the perspective of overall measurement and identification method. The load model parameter identification strategy takes the substation bus voltage and voltage phase angle as the input when the power grid is disturbed, selects the classic load model of the induction motor parallel ZIP load. The strategy realizes the iterative optimization of the objective function through the gray wolf algorithm to obtain a set of optimal load model parameters, so that the model response can better fit the sample power curve. GWO algorithm has strong fast convergence ability and global search ability. Its application in load modeling parameter identification practice can effectively improve the identification accuracy. By establishing a power system simulation model in PSD-BPA software, two examples are simulated with the disturbance data at the substation bus as the input data of the load modeling. Simulation results show that GWO has obvious advantages in calculation accuracy and convergence speed compared with the commonly used particle swarm optimization algorithm.

Abstract:Most of the rated capacity design of transmission lines is based on conservative meteorological conditions such as high temperature and low wind speed, and the rated capacity is a fixed value. This conservative method often leads to insufficient utilization of line capacity. The method of dynamically adjusting the line capacity is to adaptively calculate the available capacity of the line by using the local meteorological parameters such as wind direction and wind speed and the grid load. In this paper, the local wind field is modeled based on hydrodynamics calculation, and the wind field modeling results are combined with the meteorological station data to estimate the available capacity of the line. The proposed method is tested on overhead lines in Shenyang suburb. The results show that the calculated available capacity is higher than the static rated capacity in 76% of the research time, and it is 22% higher than the static rated capacity, which verifies the effectiveness of the proposed method.

Abstract:In order to improve the accuracy of faulty feeder selection when the single-phase-grounding fault occurs in a resonant earthed system, the distribution features of phase-to-phase current in both faulty lines and non-faulty lines are analyzed in detail. It is found that in a small fault resistance rather than a high fault resistance, there is a significant difference between the variation of phase-to-phase current in faulty feeder and non-faulty feeder. If the compensation degree of arc-suppression coil is adjusted after the occurrence of the fault, the mutation of phase-to-phase transient current would be equal to the change of zero-sequence inductor current, which is more overt in the case of high impedance grounding faults. Therefore, a non-setting faulty line selection method is proposed, which integrates the mutation features of phase-to-phase current mutation in each line with compensation adjustment after the fault. The fault feature distance can also be defined by Euclidean distance so that the line with maximum value can be distinguished as the faulty line. Simulation results show that the proposed criteria requires no setting, and can fit in the scenarios with high transition resistance orsmall inception angle fault.

Abstract:In order to realize the adaptive optimization of the relay protection strategy of the distribution network, the protection is divided into the fault identification and exit decision-making process in terms of the fault handling process of the relay protection. On this basis, a two-level optimization method for relay protection strategy of the distribution network based on the fusion of operation mode and fault information is proposed. First of all, relying on the tracking of the running mode,the fixed values of the protection action in the fault identification process are set. Then, the fault information generated by the fault identification process is opened to the export decision-making process so as to realize the optimal setting of the export decision-making process. Finally, this paper focuses on the optimization method of the exit decision-making process integrating fault information. Through the fault topology, protection exit coordination and reclosing benefit analysis, the isolation action selection, exit fixed-value setting and reclosing strategy selection of protection exit decision-making for the distribution network relay protection in response to different fault scenarios are given to effectively improve the protection flexibility.

Abstract:The traditional fault location method based on transient value only utilizes local fault feature information and it has a poor adaptability to different fault conditions. Therefore, the time-frequency characteristics of transient voltage are analyzed at first. The similarity and difference of transient feature generated by different fault positions are considered and five fault features are involved in total. In time domain, the arrival time difference, amplitude, and polarity of each transient signal are extracted. In frequency domain, the energy ratio of low frequency to high frequency and the wave velocity of the main frequency components of each transient surge are selected. On this basis, a fault transient information fusion matrix is constructed according to the extracted fault features. Finally, the difference of different fault transient information matrix is analyzed quantitatively by utilizing a waveform correlation coefficient, and a novel single-ended fault location method based on fault transient information fusion is proposed and it can be adapted to different fault conditions since it has a simple principle and low requirement for sampling frequency. The simulation verifies reliability and practical value of this method.

Abstract:An integration framework of the distribution network cyber-physical system (CPS) is established in this paper. Firstly, the CPS model of the distribution substation is constructed according to the IEC 61850 standard and the interaction between the information layer and the physical layer is analyzed. Secondly, different types of physical layer fault propagation scenarios triggered by information elements are enumerated, classified and defined under the consideration of the abnormal operation conditions such as the delay, malfunction and refusal of cyber elements. By calculating the probability of different physical layer fault propagation scenarios during the fault clearing process, the coupling effect of distribution network information elements on the physical layer is quantitatively evaluated. A 10 kV distribution network feeder system in a certain area is taken as an example to carry out the calculation and analysis of the fault scenario. The failure of cyber elements expand the scope of physical layer failures, among which the communication link failure of the switch has the most significant influence.

Abstract:In order to improve the accuracy of power quality disturbance recognition and make up for the shortcomings of traditional single feature quantity pattern recognition methods that are easily disturbed and have low precision, a power quality disturbance pattern recognition method based on fuzzy cluster analysis is proposed. The method uses Hilbert-Huang transformation (HHT) to extract corresponding disturbance feature quantities from various types of power quality disturbance signals, and then performs fuzzy clustering analysis on the extracted feature quantities to accurately classify these power quality disturbance signals into photovoltaic disturbances and public grid disturbances one by one. At the same time, a power quality disturbance identification process based on fuzzy cluster analysis is established. Simulation results show that this method overcomes the limitations of the traditional single-feature pattern recognition method, optimizes the recognition effect of disturbance signals, improves the recognition efficiency, and has high recognition accuracy and strong anti-noise ability.

Abstract:Benefit from the high proportion of natural gas application and the carbon absorption capacity a of P2G (power to gas, P2G), the integrated energy system has become an emerging platform for wind accommodation and low-carbon emissions.This paper presents a low-carbon operation dispatch model for integrated electrical-gas energy systems considering wind power curtailment.The minimum operation cost of power unit, wind power curtailment and carbon emissions are chosen as the objective. Meanwhile,the concept of P2G start-stop of wind curtailment and carbon trading market is introduced to enable the system accommodate more wind power and reduce carbon emissions.This model is solved by the multi-objective particle CPLEX solver from the YALMIP toolbox and simulated under 3 different scenarios to verify that the proposed model has the ability to absorb the redundant wind power and significantly reduce carbon emissions. Finally, carbon emissions and economic benefits of P2G with/without wind curtailment start-stop device are compared. It is shown that system operator should set wind curtailment start-stop device if the carbon trading price is high so as to avoid carbon emission increasing from P2G during daytime when there is no wind curtailment.

Abstract:Demand response (DR) has played an important role in promoting the integration distribution generations (DG) into the park integrated energy system (PIES). However, the uncertainty of DGs becomes a huge challenge to the economic operation of PIES.In view of this, a PIES economic dispatch model is proposed on the basis of the information gap decision theory (IGDT).First of all, with the objective function of minimizing the operating cost of the PIES, an economic dispatch model of PIES with DR is constructed with the consideration of cooling/heating/electricity.On this basis, the IGDT is utilized to deal with the uncertainties of DR and DG. Since the traditional IGDT is only suitable for single-factor deviation coefficients, different weights are set to DR and DG to take all the uncertainties into account when analyzing the impact of economic operations of PIES.For different types of decision makers, a risk aversion robustness model and a risk seeker opportuneness model were established to meet the needs of different types of decision-making. Then, according to the GAMS software, the equivalent deviation coefficients are calculated for the uncertainties corresponding to the two different decision-making models, with the consideration of different operating cost targets. Finally, the validity of the proposed model is verified by a case study, which provides a quantitative decision-making basis for the planner to make PIES scheduling plans.

Abstract:Considering the economic and low-carbon performance of regional integrated energy systems, a method of low-carbon operation strategy of regional integrated energy systems based on the Q learning is proposed. Firstly, the basic operation model of such regional integrated energy system is constructed on the basis of the energy hub. Then, taking the minimum daily operating cost as the objective function, including the carbon dioxide treatment cost, a low-carbon economic operation strategy of regional integrated energy system is proposed. Then, the low-carbon economy operation strategy is modeled through the Markov decision problems, and the improved Q learning is utilized to solve thoseproblems. The simulation results verify the effectiveness of Q learning algorithm for solving operation strategies in the regional integrated energy system. It is shown that the proposed operation strategy can give full play to the multi-energy complementary advantage, and realize the economic and low-carbon objectives during operation of regional integrated energy system.

Abstract:In traditional power system state estimation (PS-SE), the iterative step size of the state correction equation is generally fixed. But this method often fails to converge effectively because of the low data quality and complex network conditions. For the purpose of solving this problem and improving the suitability of state estimation, the classical logic function is reconstructed to find the generating function which is naturally suitable for the high-quality numerical iteration of state estimation on the image. Then, this function is considered as the step size control factor, and the step size factor can be adjusted intelligently by controlling parameters. After that, the weight factor function introdueced to make the algorithm perform variable weight operation in the iterative process and the influence from the bad data can be reduced then. Compared with the analytical method in terms of an adjustable step size, this method has the characteristics of low coupling in model and strong portability. Consider an IEEE30 node system as example. It is found that the proposed algorithm is superior to the traditional fixed step size method in terms of numerical stability, computation efficiency, and estimated quality when the measurement has bad data and the power system is under quasi ill-conditioned and ill-conditioned.

Abstract:To meet the flight guarantees and the charging requirements of Civil Aviation Specific Electric Vehicle (CASEV), a charging control strategy is proposed in this paper. Firstly, the relation between the number of real-time flights and the number of CASEV, and the impacts of the number of real-time flights on the critical threshold of electric special vehicle charging power are analyzed. Then, a orderly mathematical model of CASEV charging strategy, which is subjected to the number of flights and charging capacity of the electric vehicle, is established which takes the minimum cost of charging as the objective function. At last, based on the actual operation data from an existing airport electric special vehicle, the charging behaviors of airport electric vehicles under the ordered charging and out-of-order charging strategies are simulated. It is shown that the proposed orderly charging strategy can decrease the impacts on the power supply grid from the charging for CASEV, and also decrease the cost of charging operations.

Abstract:In order to effectively utilize demand-side resources and renewable energy power generation, the virtual power plants can be defined as commercial virtual power plants and technical virtual power plants according to their different characteristics and then a bi-layer optimization model of virtual power plants combining the commercial layer and the technical layer is established on the basis. Among them, the upper-level commercial virtual power plant manages user loads with the goal of maximizing user-side benefits. Different types of controllable loads is optimized on the basis of the established time-of-use electricity price and the scheduling role of price-based and incentive-based demand response is comprehensively utilized. Meanwhile, the technology-based virtual power plant would manage the wind-photovoltaic-fuel-storage combined power system and the lower-level objective is minimizing the output cost of distributed power sources on the basis of satisfying the upper-level scheduling results. The proposed method takes into account the benefits for both demand side and power generation side simultaneously. Finally, the dispatching strategy of the virtual power plant is obtained by applying the CPLEX solver. The economic benefits of the virtual power plant in different scenarios are compared to verify the rationality of the established model.

Abstract:For active distribution networks (ADN) with a high proportion of photovoltaic-energy storage system (PV-ESS) units, a distributed coordinated control strategy is proposed to track the power reference value at the grid connection point (PCC) in this paper. The voltage violations caused by the mismatch between photovoltaic output curves and local load curves can be mitigated. Firstly, the strategy generates the reference value of the change in the PV-ESS units' load rates based on the power difference between the reference value and the actual value at the distribution network PCC. Then, a pining consensus algorithm is utlized to adjust the active/reactive load rates of ESS and inverter. A proportional based power distribution is achieved and the power at the PCC of ADN can track the reference value. In addition, an average consensus algorithm is introduced to calculate the average SOC of the system, and each ESS corrects the power reference according to its SOC to achieve the balanced control of SOC. Finally, a simulation ADN system containing PV-ESS units is built in Matlab/Simulink software to verify the effectiveness of the proposed strategy.

Abstract:The logistics and distribution problems of electric energy measuring instruments include demand prediction, distribution center location, customer distribution, etc. An optimization strategy for electric energy measuring instru-ment distribution network is proposed based on an improved immune genetic algorithm and Holt-Winters model in this paper. Firstly, the Holt-winters model is applied to predict the monthly demand in the next year. The golden section method can be utilized to search for the optimal smoothing coefficient to improve prediction accuracy. Then, the Gauss projection method is employed to transform the longitude and latitude of customers into Cartesian coordinates, and the objective function of minimum transportation costs is constructed. The antibody concentration of the immune genetic algorithm is improved based on the similarity and vector distance between populations. Finally, the power meter dis-tribution problem of metering center-distribution center-customer in a provincial power grid is included as an example. The improved immune genetic algorithm solves the objective function and selects the optimal distribution center loca-tion and customer distribution scheme. It is shown that the improved immune genetic algorithm has higher conver-gence efficiency and the ability to avoid local convergence. The proposed optimization strategy of the electric energy measuring instrument distribution network has a certain reference value to reduce the distribution costs.

Abstract:In order to evaluate the operation status of electric energy metering devices (EEMD) more comprehensively and accurately, a state evaluation method of the EEMD is proposed on the basis of Fuzzy Analytic Hierarchy Process. The data of each state variable of the EEMD can be obtained by the Internet of Things, and are further used to construct a fuzzy judgment matrix for solving the weight of each layer element. The final score of the target device is calculated by the weighted summation of the weight and the index scores of each layer, and then the interval of the score is used to determine whether each part of the EEMD is normally operating. Through qualitative and quantitative analysis, the proposed method can systematically and effectively reveal the relationship between the final evaluation target and each level. Finally, the effectiveness and accuracy of the method are verified through realistic cases.

Abstract:Energy storage systems have excellent power regulation and frequency control ability, so they play an important role in absorbing new energy. The AGC control strategy of the whole station and energy storage unit of Zhejiang power grid-side energy storage power stations is introduced. The AGC control strategy is optimized based on battery energy efficiency, and a load distribution strategy considering the battery energy consumption factor is proposed. At the AGC site of an electrochemical energy storage power station, the conventional equal proportion distribution strategy is used to test the AGC of the energy storage unit and the whole station, respectively. The test results show that the regulation rate, response time, and steady-state error of AGC control of the energy storage unit are significantly improved. Then, AGC tests are carried out by using equal proportion distribution, equal margin distribution, and optimal distribution strategies. The test results show that after using the load distribution strategy considering the battery energy consumption factor, the load distribution results of PCS in the energy storage unit are different, which is conducive to the rapid load response of the energy storage system.

Abstract:A transformer DGA fault diagnosis method is proposed based on the random forest feature optimization and multi-scale cooperative mutation particle swarm limit learning machine for the problems that different input characteristics effects the diagnosis results and the low accuracy of particle swarm algorithm optimization limit learning machine. Firstly, the candidate feature set is established on the basis of the DGA data in the fault sample. The random forest algorithm is utilized to calculate the feature importance scores and rank them in a descending order. The optimal input features are then selected by the sequence forward selection method. Next, aiming at the problem of difficult parameter selection of extreme learning machine, a multi-scale cooperative mutation particle swarm optimization algorithm is introduced for optimization. Finally, the method is compared for the diagnostic performance with the IEC three-ratio method and different combinations of extreme learning machines. An example shows that the proposed method has higher diagnostic accuracy.

Abstract:Under weak grid conditions, the system generates resonance due to the coupling between parallel inverters and between inverters and the grid. At the same time, the grid impedance in the actual system and the line impedance from the inverter to the grid connection point cause the resonant point to shift, aggravating the system resonance instability. Under the background, a multi-inverter parallel mathematical model considering grid impedance and line impedance is established on the basis of a single grid-connected inverter firstly. Then, the resonance formation mechanism of the parallel system under weak grid is discussed and the influence of grid and line impedance on system resonance is analyzed. Meanwhile, a resonance suppression method is proposed by combining the full feedforward of the incoming current and the virtual admittance in parallel at the PCC point. Finally, three parallel system simulation models of T-type three-level inverters based on LCL filters are built in Simulink. The simulation results show that this method can effectively suppress the inherent resonance of the LCL grid-connected inverter and the resonance caused by the weak grid. In addition, the method can also effectively improve the system stability and enhance the robustness of the multi-machine parallel system to changes in line impedance and grid impedance.

Abstract:In order to solve the on-line detection problem of insulator insulation resistance for UHVDC transmission lines, a new method of measuring insulation resistance of UHVDC line insulators by using live robots is proposed, and a low zero value insulator detection robot for transmission lines is developed. By using the method of equipotential operation, a bypass circuit for the tested insulator is built with the arms and body of the robot. The interference of leakage current to the tested circuit is eliminated and the measurement accuracy is improved by using this method. The measurement, storage and transmission of insulator resistance information are completed by using the mechanical structure design of the robot itself and the cooperation of hardware circuit. Meanwhile, the discrete fitting function is established to accurately calculate the data, thus the accuracy of the test results is ensured. Finally, a prototype is developed for comparative test. The test results show that the robot measurement method can accurately measure the insulators resistance in operation, and the error is controlled within 5%. Then, the operating condition of insulators can be judged, and the online warning of low-zero insulators can be realized.

Abstract:The magnetic resonant coupling based wireless power transfer utilizes the high-frequency alternating magnetic field coupling fundamentally. Once a metallic foreign body invades, the eddy current effect generated by the foreign body may bring considerable security risks to the system. At present, the detection methods of metallic foreign body in this wireless power transfer system are mainly based on power loss and thermal effect caused by foreign body. Its dominate disadvantages relay mainly on the high delay and complex parameter extraction. Under the background, a new detection system of metallic foreign body location based on three-stage coil is proposed in this paper. The first and second stage coils are composed of two parallel rectangular unit detection coils. The second stages coils are arranged perpendicular to each other to achieve the simple positioning of metallic foreign body in a quarter of the charging range. The third stage coil is composed of four isosceles triangular unit detection coils that just cover the transmitting coil. The locating detection of metallic foreign body within one eighth of the charging range is then realized. Finally, a magnetic resonant coupling based wireless power transfer system with an area of 340 mm×280 mm transmitting coil is applied for verification. It is shown that a square iron block with side length as small as 20 mm can be detected by the system when the threshold voltage is 1.4 V.

Abstract:In order to improve the level of electricity consumption, it is an inevitable trend to use artificial intelligence technology to provide active service to electricity customers. Under the background, an active customer service recommendation method is proposed based on LSTM-Attention fusion considering the lack of research on active customer service in the power industry. The proposed method can effectively solve the problems of gradient mass and gradient explosion in the service recommendation of a single deep learning model. Firstly, a model is established for extracting potential service demands of customers from electric power complaint work orders. Then, an active service recommendation method is obtained for electric power customers based on the LSTM-Attention fusion algorithm. Finally, an electric power customer complaint work order in one city is included to verify the algorithm and model. It is shown that this method is effective.