2020, 44(24):1-10. DOI: 10.7500/AEPS20200720009
Abstract:The new generation of artificial intelligence technology and its application represented by deep learning and reinforcement learning are the research hotspots in the field of power systems. Artificial intelligence technology has the advantages of independence of physical mechanism, high calculation speed and high discrimination efficiency. However, the inherent disadvantages of artificial intelligence, such as poor interpretability and weak stability, restrict its application in some scenarios of power systems. In this paper, the application of new generation artificial intelligence technology in power system load and renewable energy forecasting, fault diagnosis, on-line stability assessment, frequency and voltage optimal control and power grid operation mode formulation are summarized and analyzed. This paper summarizes the existing research problems and points out that the application of artificial intelligence technology should be problem-oriented, scenario-based and application-targeted. Finally, the future application of artificial intelligence technology in dispatching and operation of power system is prospected.
2020, 44(24):11-18. DOI: 10.7500/AEPS20201016002
Abstract:From August 14 to August 15, 2020, a rotating power outage occurred in California, USA. From August 17 to August 18 and from September 5 to September 6 of the same year, the California power grid entered a state of emergency, affecting the normal electricity usage of at least 810 000 residents. Accident analysis report of the California Independent System Operator (CAISO)pointed out that the accident was caused by a series of factors. Based on the public reports of the CAISO, this paper sorts out and summarizes the development process of the accident, and analyzes the causes. Finally, combined with the actual situation of the power grid in China, suggestions on prevention of extreme weather, focusing on normalizing demand response, making steady progress toward the carbon neutral goal in 2060, and improving cross-regional power dispatching capability are put forward.
2020, 44(24):19-27. DOI: 10.7500/AEPS20191230009
Abstract:There still exist problems in the current risk analysis on cyber security such as the inaccuracy of fundamental probability and the lack of severity models. First, based on the framework of the power monitoring system in the substation, this paper proposes a “perimeter-device protection” model to describe the cyber threat of intrusion and improve the preliminary invasion probability model with the machine learning results based on historical data. Cyber-net as a specific state transition diagram is achieved based on the network security topology of general stochastic Petri net (GSPN) and power monitoring system. Then, a comprehensive risk estimation model for cyber security is proposed based on system operation results and the Markov chain steady-state probability to quantify the correlation between information and physical security of the power system. The accuracy of the risk model is improved by transient fault calculation. Finally, the feasibility and effectiveness of the model are verified by simulation and sensitivity analysis, and the feasible defense strategies and potential application scenarios are summarized.
2020, 44(24):28-35. DOI: 10.7500/AEPS20200505018
Abstract:In smart grid, accurate data acquisition is the basis of the whole system security and economic operation. With the deepening integration of cyber-physical systems, the requirements of various big data applications and real-time control tasks for the acquisition of high-frequency data are keeping rising. However, increasing the data sampling frequency will inevitably cause higher data communication and storage burden to the system. In this paper, a deep learning based super-resolution perception technology is proposed to recover accurate high-frequency data from low-frequency sampled sensor data. Specifically, a deep end-to-end super-resolution perception method based on the gated recurrent unit (GRU) network is proposed, which includes feature extraction, relationship inference and information reconstruction. In the feature extraction part, an one-dimensional convolution network is used to extract the features of low-frequency data. In the relation inference part, the GRU network is used to learn the obtained features to infer the internal relationship between the low-frequency data and the high-frequency data. In the information reconstruction part, the inferential information is reconstructed with the full connection layer to obtain the corresponding high-frequency data. The proposed method is used for super-resolution perception of the power data of residential and industrial users and the voltage data of transmission lines, and the recovered high-frequency data is used for load state identification. The example results show that the proposed method can accurately and effectively recover the lost information of low-frequency data and improve the accuracy of practical applications such as load identification.
2020, 44(24):36-43. DOI: 10.7500/AEPS20200324007
Abstract:Aiming at the complexity of on-line computation of transient stability preventive control in power systems, this paper proposes a transient stability preventive control method based on the generative adversarial network. By modeling the transient stability preventive control as a sample space mapping problem, a data-driven method is adopted to establish mapping by a training generative model from the transient instability operation space to the transient stable operation space. The model improves the transient stability margin of the power system by adjusting the active power output of the generators to meet the checking requirements of the transient stability for operation points in the power grid. Compared with the traditional optimization modeling methods, the proposed method solves the control strategy through feedforward inference of the neural network without iteration, which greatly improves the solving efficiency. The test results based on the New England 39-bus system verify the feasibility and effectiveness of the proposed method.
2020, 44(24):44-52. DOI: 10.7500/AEPS20200521003
Abstract:There exist misdetection (misclassification of unstable samples into stable samples) and false alarm (misclassification of stable samples into unstable samples) by the transient stability prediction method based on artificial intelligence, which is a major obstacle to practical engineering applications. In response to this deficiency, this paper proposes a two-stage power system transient stability prediction method based on the convolutional neural network (CNN) considering the cost of misdetection and false alarm. At the first stage, the corresponding sliding time window input features are trained to obtain different layers of integrated CNN models, and the credibility index for each output layer is established. Then, the credibility threshold optimization selection problem is transformed into a multi-objective optimization problem. This stage could minimize or even eliminate the misdetection and output credible samples with high credibility as soon as possible. At the second stage, an emergency control start-up strategy based on multi-criteria fusion for the credible unstable samples predicted at the hierarchical prediction stage is proposed to reduce the actual loss caused by false alarm. The analysis of a simulation system shows that the proposed method can minimize or even eliminate the misdetection at the minimum cost, and promote the probability of practical engineering application of transient stability prediction results based on artificial intelligence method.
2020, 44(24):53-59. DOI: 10.7500/AEPS20200313007
Abstract:In view of the problem that the frequency regulation reserve capacity of wind power is idle for a long time and does not fully serve the operation of power grid, the in-depth study on the optimal configuration method for the frequency regulation reserve capacity of large-scale wind power has important impact on the power grid frequency regulation ability, the economy of wind power generation and the peak regulation of power grid. Based on this, the optimal configuration strategy for the reserve capacity of large-scale wind power participating in the primary frequency regulation is studied. According to the daily load forecast curve, by setting larger frequency regulation base point power in the peak period, smaller base point power in the valley period, and base point power in other periods in proportion to the peak load, the dynamic reserve capacity of wind power frequency regulation can be configured in each period of one day. According to the determined dynamic reserve capacity in each period, the reference speed and reference pitch angle are solved based on the primary frequency regulation strategy of speed control or pitch control, so that the wind turbine unit can dynamically adjust the reserve capacity of primary frequency regulation at any wind speed (greater or lower than the rated wind speed). The results show that the proposed strategy can not only meet the demand of grid frequency regulation, but also effectively promote the effect of peak shaving and valley filling and improve the economy of wind power generation in the scenario of large-scale wind power integration.
2020, 44(24):60-67. DOI: 10.7500/AEPS20200414001
Abstract:Offshore micro integrated energy systems are the foundation of offshore oil and gas engineering. In order to reasonably evaluate operation risks and ensure the safe development of marine resources, this paper proposes a risk assessment scheme for offshore micro integrated energy system based on the analysis of matter-energy flow. By adopting a two-state Markov model for the key equipment to describe its on-off status, risk impact factors are introduced. Then a risk output model is built, and the matter-energy conversion relationship of the output equipment under risk is analyzed. A risk transfer model based on the generalized matter-energy flow model is proposed to describe the dynamic behavior of risk in the system from the perspective of matter-energy flow. In addition, according to the system structure and operation characteristics, a risk index for micro integrated energy system and its evaluation procedure are designed. Taking an offshore oil and gas platform in the Bohai Sea of China as an example, simulation experiments are performed to verify the effectiveness and rationality of the proposed model and method.
2020, 44(24):68-76. DOI: 10.7500/AEPS20200420003
Abstract:In an attempt to solve the joint clearing problem of integrated energy system and external grids in the environment of spot markets and distribution side deregulation, a two-stage clearing method including day-ahead and real-time stages is proposed considering the mutual balance of interests between multiple entities. In the day-ahead clearing stage, multiple rounds of iterative bidding is implemented considering the interest-driven of all entities, and incorporating factors into the day-ahead bidding process, including mutual insurance, reserve and peak shifting. In the real-time clearing stage, a triple energy supply system of producers, superior power grid and spinning reserve providers is formed by coordinating internal and external reserve, energy supply refinement and interest distribution. Moreover, the clearing decision models for producers, operators, and load aggregators are built. Finally, the effects of the demand response, mutual insurance contract and spinning reserve market on peak shifting and internal and external reserve are analyzed by the case simulation, which shows that the method can realize the supply and demand balance and interest balance among the entities.
2020, 44(24):77-88. DOI: 10.7500/AEPS20200601001
Abstract:In order to give full play to the advantages of decentralized dispatching of multiple microgrids (MMGs) trading, protect the privacy of each sub-microgrid, and perform efficient and fast calculations, this paper proposes a multi-time scale trading mechanism of the grid-connected MMGs system and an optimization algorithm of trading strategy based on deep learning. Firstly, an internal electricity pricing model of MMGs is established, which can dynamically adjust the internal trading price according to the changes of supply and demand situations in MMGs, and make the internal trade among the MMGs more economical than the direct trade between the microgrid and the distribution network, thus encourages each sub-microgrid to participate in internal trade. Secondly, a day-ahead and intra-day trading mechanism of “quoted volume without quotation” is established. In the day-ahead trading, the trading plan and trading price are formed through iterations of the trading power and the internal trading prices, and the day-ahead power trade is cleared; in the intra-day trading, each sub-microgrid only declares the trading quantity of the unbalanced power once, and it is cleared directly after the declaration. In addition, based on the deviation of the expected and actual power exchange between the MMGs system and the distribution network, a compensation scheme is proposed to reduce the influence of the power fluctuations on the operation of distribution network. Then, based on the generated intra-day trading sample data, the deep neural network algorithm is introduced to train and learn the trading strategies of each sub-microgrid, so that the sub-microgrids can quickly and accurately obtain its own optimal power trading plan during the intra-day trading stage. Finally, an example is given to verify the effectiveness of the proposed model and algorithm.
2020, 44(24):89-95. DOI: 10.7500/AEPS20200513004
Abstract:Aiming at the power fluctuation of multi-photovoltaic (multi-PV) DC distribution network systems in different application scenarios, this paper proposes a distributed multi-PV coordinated control strategy based on the discrete consensus algorithm. First, the information of the power deviation and the operation modes can be exchanged between neighbor PV controllers. The average value of power deviation and operation modes can be obtained through the discrete consensus algorithm. Then, the operation mode instructions and power reference instructions can be updated and the power deviation is balanced by utilizing power complementary between PVs. PVs can operate in the maximum power tracking mode or constant power mode adaptively. In such a way, power balance will be guaranteed even if the solar irradiance and rated capacity of PVs are uneven. Finally, the simulation results verify the effectiveness of the proposed control strategy.
2020, 44(24):96-104. DOI: 10.7500/AEPS20200602002
Abstract:Aiming at the modular multilevel converter based multi-terminal flexible direct current (MMC-MTDC) transmission system, this paper proposes a scheme of topology selection and parameter configuration considering the multi-dimension characteristics of application scenarios. An economic optimization model for the whole life cycle of the MMC-MTDC transmission system considering topological differences is established with the scenario characteristics of three dimensions, i.e., temporal characteristic, spatial distribution and resource capacity. The modified second-order cone programming (SOCP) relaxation method is proposed to transform the nonconvex secondary optimization into SOCP. The economically optimal topologies in various application scenarios are obtained, and the configuration scheme of parameters such as the output of relevant equipment and line capacity is formulated. Based on the engineering parameters of Zhangbei flexible DC grid and Zhoushan MTDC transmission in China, this paper designs the four-terminal and five-terminal MMC-MTDC transmission systems in two typical application scenarios with long and short transmission distances. The economic analysis with three topologies is carried out from three aspects, i.e., total economic benefit, the consumption rate of renewable energy, and network loss in the whole life cycle. The effectiveness of the proposed topology selection method is validated.
2020, 44(24):105-110. DOI: 10.7500/AEPS20200423004
Abstract:The rapid detection and removal of faults are the key to improving the operation reliability of DC microgrids. The current differential protection can remove the fault quickly and selectively. But it is greatly affected by the short-circuit impedance, and it may refuse to operate in the cases of the high impedance short circuit. For ring DC microgrids, a differential protection based on change rates of bus power is proposed. The power change rates of both sides of the bus are taken to formulate the differential value. The differential value of the change rates of bus power is bigger than the operation setting in the case of internal fault, and the protection will operate and remove the fault line. The differential value of the power change rates is proportional to the square of short-circuit current and short-circuit impedance. Compared with protection schemes based on current, it has faster fault identification speed and higher protection sensitivity. The simulation results show that the proposed protection scheme has better operation speed and sensitivity, and improves the reliability and stability of the ring DC microgrid.
2020, 44(24):111-118. DOI: 10.7500/AEPS20191206002
Abstract:Large-scale multi-infeed direct current (MIDC) systems lead to the relatively weak voltage support capability in receiving end AC systems, resulting in severe voltage stability issues. Although, it is critical to enhance the strength of the AC system, it may exacerbate the short-circuit current level. The structure of the receiving-end AC system affects both system strength and short-circuit current level. Since optimizing the grid structure of the MIDC system can deal with the contradiction between system strength and short-circuit current level, this paper proposes a grid structure optimization method based on the sensitivity analysis of generalized short-circuit ratio (gSCR) and the branch addition method. Firstly, the relationship between the gSCR and the short-circuit current level is explored. Secondly, the impact of AC system structure adjustment on the two indicators is studied, and the comprehensive benefits of structure adjustment after coordination of the two indicators are analyzed. Finally, the receiving-end DC terminals are classified, and corresponding optimization methods for receiving AC system structures are given. Simulation results verify the validity and effectiveness of the proposed method.
2020, 44(24):119-125. DOI: 10.7500/AEPS20191228001
Abstract:The spot network has the characteristic of high power supply reliability. It is conductive to the consumption of distributed energy resource (DER). However, the existing structure and its methods for control and protection limit its further development. This paper proposes the structure of the AC/DC hybrid spot network based on the combination of the advantage of the multi-power supply of the spot network and the idea of the AC/DC hybrid. It gives the network operation mode in typical working scenarios. For the dynamic balance of energy in different operation modes, a unified control strategy based on overall power shortage is proposed. It can realize the unified control of the converter in different operation modes and their switching process. Finally, simulation analysis verifies the effectiveness of the proposed structure and unified control strategy.
2020, 44(24):126-133. DOI: 10.7500/AEPS20191214004
Abstract:When multiple energy storage units operate in parallel, the traditional load distribution control technology, such as the current-sharing method, is difficult to fully utilize the residual capacity of each energy storage unit. Therefore, how to realize the load distribution among energy storage units according to the difference of residual capacity has become one of the research hotspots of energy storage device control technology. In this paper, the load coefficient of the energy storage unit is defined by using its output current and residual capacity. By combining the load coefficient with the load distribution control circuit which has the structure of double signal buses, a load distribution control strategy based on residual capacity matching is designed to improve the utilization rate of the residual capacity of the energy storage device. In order to enhance the operation flexibility of energy storage units, a flexible thermal switching control method for parallel operation of energy storage units is formed based on the proposed load distribution control strategy. Simulation and experimental results verify the feasibility and effectiveness of the proposed control strategy and method.
2020, 44(24):134-144. DOI: 10.7500/AEPS20200501002
Abstract:The virtual synchronous generator (VSG) control can simulate the operation characteristics of the synchronous generator (SG), which helps the inverter power supply overcome the shortcomings of low inertia and low damping. However, the VSG control may cause the inverter power supply to participate in the low-frequency oscillation between machines or regions of a large power grid. Aiming at this problem, this paper develops a small-signal model of the VSG-SG interconnected system for the study of low-frequency oscillation characteristics of the power grid. Based on the model, the difference and connection between the VSG and SG influencing the low-frequency oscillation modes of the system are studied to reveal the mechanism of VSG affecting the low-frequency oscillation of power grid. The influence of various control parts in VSG on the low-frequency oscillation mode is also analyzed, which indicates the low-frequency oscillation risks of different types of VSGs. Finally, experiments on the RT-LAB platform are conducted to verify the results of the small-signal model analysis.
2020, 44(24):145-150. DOI: 10.7500/AEPS20200108008
Abstract:In recent years, the second harmonic content of the transformer current is lower than the protection setting value during the no-load transformer energization test, which leads to the problem of mal-operation of differential protection. Aiming at the problem, two straight lines with different slopes are used to fit the magnetization characteristic curves of transformers. Then, the expressions for the second harmonic content of magnetizing inrush currents with different no-load voltages are derived. Based on the finite element analysis software, this paper develops and verifies the simulation models of a 14 kVA single-phase transformer and a 325 MVA three-phase transformer. The validated model is simulated under the condition of no-load transformer energization. The study shows that: considering the non-linear characteristics of the beginning of the core magnetization curve, the magnetizing inrush current contains a small amount of the second harmonic when the no-load voltage is less than the critical voltage; the second harmonic content of the magnetizing inrush current reaches its peak when the no-load voltage is equal to the critical voltage; with the increasing of no-load voltage level, the second harmonic content of magnetizing inrush current shows an increasing trend before decreasing.
2020, 44(24):151-160. DOI: 10.7500/AEPS20200406001
Abstract:Modular multilevel converter (MMC) can realize DC fault current clearing by improving the sub-module topology, but most sub-modules do not have the capacitor voltage self-balancing capability. Based on the full-bridge sub-module, a novel sub-module named oblique-connection full-bridge sub-module (OCFBSM) with fault current self-clearing capability and module capacitor voltage self-balancing capability is derived. The OCFBSM is composed of two full-bridge sub-modules by shifting combination. Under normal working conditions, it can operate in bypass, series and parallel states according to the connection relationship of two capacitors, which leads to the voltage balance of module capacitors without additional voltage balancing control. In the case of DC short-circuit fault, the OCFBSM can automatically clear the DC fault current by reversely putting two capacitors into the fault circuit. The effectiveness of the OCFBSM in DC fault current clearing and voltage self-balancing is verified by simulation results based on MATLAB/Simulink. Moreover, capacitor voltages are balanced after the fault blocking of sub-modules, which is conducive to the restart of MMC.
2020, 44(24):161-168. DOI: 10.7500/AEPS20191228004
Abstract:The development of power systems has put forward new requirements for the use of power system knowledge. In order to achieve automatic extraction of unstructured text knowledge of power grid dispatch and control, the deep learning model of attention based bidirectional long- and short-term memory networks and conditional random fields is proposed in this paper. The deep learning model extracts grid operation rules and grid accident processing procedures from text data such as dispatching protocols. The experimental results show that the accuracy, recall rate, and F1 score of the proposed model on the corpus are 91.00%, 89.98%, and 90.49%, respectively, and the results are slightly better than the other three models. F1 evaluation is performed on the training set and test set, respectively, and the difference in recognition accuracy is small. This shows that no over-fitting occurs during model learning, and the proposed deep learning model has good generalization ability.
2020, 44(24):169-177. DOI: 10.7500/AEPS20191118001
Abstract:Existing hydropower stations mostly adopt the constant-speed generation systems, which will not only cause a decrease in periodic hydro-energy capture efficiency but also deteriorate the unit operation environment. The variable-speed operation (VSO) of the hydraulic turbine is the key to overcome the above-mentioned drawbacks. This paper reviews the application and development of the variable-speed constant-frequency (VSCF) technology in the hydropower generation field. Firstly, the common types of hydropower stations and the principles of constant-speed generation systems are described, and the advantages and disadvantages of VSO systems are induced. Then, the main realization methods and operating principles of the VSCF technology are introduced; the characteristics and differences of each VSO system are compared; and the applicability and prospect of each system are analyzed. Finally, the emphases and difficulties of VSCF hydropower generation technologies are analyzed, and the future development trends are prospected.