As the scale of power systems continues to increase, measurement data of power system grows rapidly. However, the process of massive data collection, measurement, transmission, and storage in power systems may lose data, which seriously threatens the safety of the power grid. Aiming at the problem of missing data in power systems, this paper proposes a missing data recovery method based on long short-term memory (LSTM) networks. Firstly, since the LSTM network can extract the characteristics of the timing law of measurement data, a double-layer and full connection LSTM network architecture is presented, which uses known data to map missing data. Furthermore, in order to improve the recovery accuracy of missing data in different states, a state identification method based on random forest and a recovery strategy considering the location of the missing data are developed. Finally, case studies by using simulation data and measurement data verify the effectiveness and accuracy of the method proposed in this paper, and the proposed method can significantly improve the data quality of the power systems without system topology modeling.
The identification method based on data driving for identifying abnormal data in power grid has become the focus of research in the field of power grid security. Due to the small number of abnormal data in the actual statistical data of power generation, it is extremely difficult to identify abnormal data through data mining. This paper proposes an improved generative adversarial network and isolated forests based abnormal data identification method. Firstly, by using WGAN alternating training generator and discriminator, the distribution characteristics of power generation statistical data are learned and samples are generated, which will generate abnormal samples to enhance the original abnormal samples. According to the accuracy of outlier data identification, the expansion ratio of outlier samples is determined. Then the isolated forest algorithm is used to identify the abnormal data on the expanded balanced dataset. Finally, an evaluation index system consisting of accuracy, recall and precision is selected to evaluate and compare the identification effects of models before and after category equalization. The results show that the proposed method can effectively improve the classification preference of the identification model for most classes and improve the overall identification accuracy.
In view of the current need to further explore the feasibility of the application of reinforcement learning in demand-side energy management and user-side demand response, this paper proposes a demand-side energy management method for building clusters on the basis of reinforcement learning. Firstly, the building clusters are used as the terminal energy load carrier to construct the demand-side energy management framework of the building clusters. Secondly, according to the virtual energy storage characteristics of the intelligent buildings, a novel intelligent building thermal resistance-capacity (R-C) thermal balance model and user flexibility load model are constructed, and a demand-side energy management model based on reinforcement learning is constructed by combining Q-learning algorithm. Finally, the effectiveness and practicability of the proposed theoretical method are verified by comparing the results of demand-side energy management and the performance of the algorithm through actual simulation cases.
For countries with a relatively high proportion of coal-fired power units, the implementation of flexibility transformation is one of the key means to enhance the flexibility of the power system. However, the role and operation mode of the system have changed after the unit transformation, and the original modeling method and scheduling strategy are no longer applicable. This paper proposes an optimized dispatching method for the electro-thermal energy integrated system considering the flexibility and transformation of thermal power differentiation. Firstly, the related technologies and characteristics of the flexible transformation of the two types of thermal power units are compared and analyzed. The apex-convex combination method is used to uniformly model the multi-stage peak shaving of the pure condensing unit and the multi-mode operation process of the cogeneration unit. Secondly, with the minimum total operating cost of the system as the objective function, and the electricity and heat balance, unit operation mode, unit climbing and start-up and stop as constraints, a mixed integer programming model for optimal dispatching of electro-thermal energy integrated systems is established. Finally, an example is used to verify the validity and applicability of the model. The results show that the model proposed in this paper can greatly improve the wind power accommodation capacity of the system, reduce the operating cost of the system, and can effectively reduce the peak load of the power grid in the "Three North" area.
This paper proposes a method to calculate the maximum allowable access capacity of distributed photovoltaic (PV) generation in an active distribution network (ADN), in order to facilitate the usage of the distributed PV on the premise of system security and stability. The objective is to maximize the allowable access capacity of distributed PV. Several electrical constraints, including power flow equations, bus voltage limits, branch current limits, etc., are taken into account. Furthermore, active management (AM) and demand-side management (DSM) techniques including on-load voltage regulation, reactive power compensation, control of the battery energy storage system, network reconfiguration and load curtailment are applied to increase the access capacity under the worse uncertainty condition. A two-stage robust optimization model is established. The model can be decomposed into a master problem and a subproblem in the form of mixed integer second order cones, and solved by column-and-constraint generation algorithm. A case study is implemented on the modified IEEE 33-node system to verify the validation of the proposed model and algorithm. The maximum allowable access capacity of distributed PV can be obtained. The enhancement of the result is achieved through AM and DSM techniques.
The energy Internet is regarded as the development form of smart grid in the future by realizing horizontal multi-energy complementation and vertical high coordination of "source-grid-load-storage" to achieve efficient energy utilization. In order to evaluate the benefits of multi-energy storage configuration in regional energy Internet (REI) multi-energy collaborative optimization, the architecture of regional energy Internet is analyzed firstly, and then on the basis of the multi-energy storage model, multi-energy coupling device model and energy router (ER) model, a multi-energy coupling energy storage system model is constructed. Then, the benefit evaluation model of energy storage is proposed from three aspects of peak-valley difference benefit, environmental benefit and energy-loss reduction benefit, and the corresponding evaluation indices are given. Finally, a commercial REI with electricity, gas and heat load is taken as an example to calculate the energy storage benefit, and different capacity and different types of energy consumption are analyzed and the effectiveness of the proposed index is verified. The results provide guidance for the allocation of energy storage in REI and the establishment of relevant market mechanism.
Under the background of energy Internet, the power grid in China is witnessing an increasing penetration level of renewable energy power generation and the AC/DC ultra-high voltage (UHV) interconnection between regional power grids, which increases greatly the system reserve requirement for contingency and system power fluctuation. At the same time, reserve sources from demand side and energy storage will become important complement to conventional generators. The current management mode of ancillary service categorization and management based on "two rules" can hardly satisfy management requirements of future power grid reserve services. In this respect, a new reserve categorization system is proposed for the energy internet from the perspective of controlling process. The response time and duration requirements are specified. The new reserve categorization system is compatible with diversified reserve sources from generators, networks, loads, and energy storage. The types of reserve services they can provide are also discussed. The corresponding market-oriented reserve configuration suggestions and related problems to implement the new reserve system are put forward.
Commutation failure (CF) is one of the most common faults in traditional HVDC system. Effective prediction of CF is beneficial to the safety and stability of the power system. The physical-driven prediction method can effectively reflect the causal law but it is difficult to establish a precise model. Data-driven prediction method has the advantage of efficient training, but the prediction accuracy depends on a large number of high-quality training samples. Combining the advantage of physical-driven and data-driven methods, a CF prediction method is proposed. In the part of physical-driven, the inherent response of the power system is transformed from time-domain to frequency-domain to obtain the predicted commutation voltage. Then the predicted DC current can be obtained according to the superposition theorem. Finally, the predicted extinction angle can be calculated according to the commutation mechanism. In the part of data-driven, the amplitude and phase of each harmonic of the commutation voltage are taken as the input characteristics, and the extinction angle predicted by the physical-driven method can be modified. According to the results of the test system built in electromagnetic transient simulation software, the validation of the proposed method is verified.
High-proportion integrating of intermittent new energy has a great impact on the stability of distribution network. A control strategy based on double-layer control to improve the voltage stability of distribution network is proposed. The upper-level controls the voltage stabilization to determine the total active/reactive power of the energy storage power station group for each node to meet the voltage safety operation and the minimum voltage deviation. The lower-level control optimizes the distribution of upper-level operation results, and allocates the power of each energy storage power station considering the operation economy of power grid and the utilization rate of new energy. The control process is implemented in multiple time scales, and the operation results are uploaded to the energy storage detection and control module, which finally determines the working mode and power distribution of each energy storage unit according to the real-time status of each energy storage station. IEEE 33-bus system is used to verify the proposed control method. The result of energy storage power distribution is calculated by particle swarm optimization algorithm, which proves the rationality and effectiveness of implementing double-layer control in multiple time-scales.
Under the policy of "New Infrastructure", the development of photovoltaic (PV) and battery energy storage (BES) technologies supplies reliable support for large scale charging piles integration into grid. This paper proposes an optimal configuration method for BES and PV in charging station. Firstly, capacity degradation, lifetime loss, and dynamic efficiency are integrated to the discharging and charging characteristics of BES to establish a refined BES model. According to this model, a mixed integer non-linear programming (MINLP) model is developed to minimize the cost of charging station. By separating variables, the MINLP is transferred to a double-layer model. The optimal solution of external layer is determined through genetic algorithm (GA), and the inner layer optimization can be solved by commercial solvers. Lastly, the double-layer model is simulated in a case study, and the simulation results validate the feasibility of the proposed model and solution methods, which could be a reference for PV and BES configuration in electric vehicle charging station.
This paper presents a method to evaluate the ability of private electric vehicle (EV) to participate in power grid regulation based on EV trip chain. According to the characteristics of users’ trip chain, the trip chain of private EVs is simulated during working days and non-working days, and the spatial distribution variation characteristics of EVs in various functional areas of the city are obtained. Considering the key factors affecting users’ participation in power grid regulation, the capacity model of single private EV is established according to the SOC constraint and power constraint of EV, and then the regulation capacity model of cluster EVs in different urban functional areas is established. Finally through the simulation analysis, the regulation ability of private EVs during working days and non-working days is evaluated and compared. Furthermore, the regulatory capacity changes of private EVs under the willingness to participate in regulation are considered. The results show that the change of the regulatory capability of private EVs is related to the change of EV number in a functional area of a city, and the decrease of the proportion of users’ willingness to participate makes the overall regulatory capability decrease, but the change feature is still dominated by the change feature of EV spatial distribution.
The hybrid multi-terminal HVDC transmission system combines the advantages of conventional HVDC and flexible HVDC transmission technology, which can realize the functions of high-capacity power transmission, fault ride-through and suppression of commutation failure while reducing the economic cost. However, the hybrid multi-terminal system topology is relatively complex, and the fully controlled power electronic equipment has a low ability of enduring over current. Therefore, a kind of line protection strategy based on transient current correlation coefficient for hybrid multi-terminal HVDC transmission is proposed. Current in the three ports of T-type tandem busbar is measured to calculate correlation coefficient, which can accurately identify the fault range. The simulation results show that the proposed protection scheme does not require communication, and the protection can be triggered within 1 ms. Meanwhile, it can withstand the interference of 300 Ω transition resistance and 20 dB white noise. Therefore, the proposed method can better meet the protection requirements of the hybrid multi-terminal HVDC transmission system.
Due to the short electrical distance of the inverter stations in multi-infeed HVDC system, the interaction between DC and AC and between DC and DC are more complex. When the AC system is weak, insufficient reactive power support is provided during the fault recovery period, which is easy to cause subsequent commutation failure of multi-infeed HVDC system. Configuration of reactive power compensation devices, optimization of control parameters and other measures are conducive to improve the fault recovery characteristics of HVDC system. This paper proposes a parameter optimization strategy of VDCOL to suppress subsequent commutation failure of multi-infeed HVDC system. The reactive power output of STATCOM is used to measure the recovery degree of HVDC system. Combined with DC voltage as input signal, the input and output characteristics of VDCOL can be adaptively adjusted according to the fault development process, so as to reduce the probability of subsequent commutation failure. A dual-infeed HVDC system model with STATCOM is built in PSCAD/EMTDC, the simulation results verify the effectiveness of the proposed strategy.
In order to solve the problem of voltage imbalance in three-phase three-leg off-grid inverter, a fractional order proportional integral (PI) quasi proportional resonant voltage control strategy with virtual impedance is proposed. In the d-q rotating coordinates, the fractional order PI tracking control is designed for the positive sequence component of the fundamental wave, and the quasi-proportional resonant tracking control is designed for the negative sequence component of the fundamental wave. Because of the difference of bandwidth between fractional order PI and quasi proportional resonant controller, the compound voltage controller does not need the separation of positive and negative sequence of voltage, which avoids the complexity of program implementation and the problem of delay and precision degradation caused by separation. Considering the influence of unbalanced voltage drop on the line impedance on the common bus voltage, the negative-sequence virtual impedance is designed to reduce the total negative sequence impedance, thus realizing the three-phase voltage balance at the point of common coupling. Finally, the effectiveness of the proposed control strategy is verified by simulation and experiment.
Insulator pollution flashover is a main disaster of electrical power system. A large number of artificial pollution tests are investigated under different contamination levels (different soluble contaminants densities or dust densities). According to experiment data, seven acoustic signal characteristics are extracted and analyzed. According to the conclusion, the general regression neural network (GRNN) model of risk degree prediction is established, in which the seven acoustic signal characteristics are as the inputs with the risk degrees used as outputs. It is found that the prediction accuracy is affected by soluble contaminants density mostly. The results show that the greater the soluble contaminants density, the smaller the acoustic signal characteristics’randomness, and the better prediction accuracy can be obtained. The conclusion of this paper provides reference for acoustic monitoring of insulators in different regions with different pollution levels.
Author Login
Peer Review
