Search

The accurate physical parameters of distribution lines are important for power grid analysis, and are essential to the stable and economic operation of the distribution network. In view of the subjectivity and hysteresis of traditional distribution network topology parameter recognition methods, this paper proposes a Dynamic Bayesian Network (DBN)-based distribution line parameter estimation model. The influence of the external environment and distribution operating mode is considered in the model. According to the relations between the parameters and the influencing factors, a DBN model is established to simulate the dynamic characteristics of the line parameters. Then the BP algorithm is applied to infer distribution line parameters with the DBN model. Finally, a case study based on a 10kV single-connected distribution network is used to prove that the proposed model has high accuracy and robustness in parameter identification.

The topologies of distribution networks change frequently due to their operation mode reconfigurations and modification projects. The models of the modified networks are difficult to maintain. Thus, it is challenging to obtain accurate models of real-time distribution networks. In this paper, we propose a data-driven approach to identify the connectivity relationship between the distribution transformers by using the historical data collected from smart meters of the SCADA system. The correlation analysis and linear regression algorithms are implemented on historical measurements of each distribution transformers and generate the inputs of Multi-Layer Perceptron (MLP) classifier. Therefore, distribution transformers connectivity recognition is converted to a 0–1 classification problem. Historical topology of distribution network is employed for neural network training. The MLP classifier-based method is demonstrated through simulation on real-world distribution network topology to prove its validity.

The accurate topology of distribution network has a significant impact on power flow calculation, reliability evaluation and system operation. However, due to frequently reconfiguration and modification of switches, the topology is hard to maintain. In this paper, a novel two-stage method is proposed to identify the connection relationships between distribution transformers and feeders. In stage 1, a new correlation indicator is adopted to screen out the unsure distribution transformers which are doubted of being incorrectly connected. In stage 2, unsure distribution transformers are assigned to the most likely feeders, and then incorrectly connected distribution transformers are found out by clustering correction. The proposed two-stage method can help operators to correct the topology errors. Cases studies in actual distribution network verified the effectiveness of the proposed method and correlation indicator.

Vortex electromagnetic (EM) waves, which carry orbital angular momentum (OAM), offers a new degree of freedom for radar detection, although to date few researches on vortex EM waves scattering characteristics for typical targets have been reported. In this paper, we investigated the backscattering characteristics of vortex EM waves for targets, i.e., corner reflectors, and flat plate. Experiments are carried out in an anechoic chamber to measure the radar cross section (RCS) of the targets, and the results indicate that for corner reflectors the RCS fluctuations of vortex EM waves are similar to that of the plane EM wave, whereas for the flat plate, RCS varies with OAM mode. The significance of this work exists in its pioneering character, which paves the way for revealing EM scattering characteristics in OAM domain.

Voltage fluctuations in distribution networks are usually caused by the stochastic volatility of photovoltaic generation system and load. The traditional deterministic power flow calculation method is difficult to assess the voltage risk. Moreover, traditional serial computing technologies cannot provide rapid voltage risk assessment of large-scale distribution networks. To address these problems, the probabilistic characteristics of PV generation and load fluctuations are first analyzed. Then, the corresponding probabilistic model is established based on the Newton-Raphson algorithm. Finally, an indicator to evaluate voltage risk is proposed by using the Monte-Carlo simulation. In order to meet the computing speed requirement imposed by large-scale distribution networks, a parallel computing platform based on Apache Spark is designed, and a parallel implementation of the Monte-Carlo simulation on Spark Resilience Distributed Datasets (RDD) is established. Simulation results based on real-world data illustrate the effectiveness of the proposed voltage risk assessment for large-scale distribution power networks and its computational efficiency.

Vortex electromagnetic (EM) waves with orbital angular momentum (OAM) offer a new degree of freedom for radar detection. Although to date few researches on target scattering characteristics of vortex EM waves have been reported. In this paper, the backscattering characteristics of vortex EM waves for a perfectly conducting wedge is investigated experimentally using the radar cross section (RCS) calibration measurement method. The experimental results show that for vortex EM waves, due to the specific spatial distribution, the RCS fluctuation of vortex EM waves is different from that of plane EM waves, and it varies with the OAM mode of the incident wave. The RCS of OAM beams is higher than that of the plane wave in some incident direction. This phenomenon indicates that vortex EM waves are more effective than plane EM waves to detect the low RCS target in some direction. And vortex EM waves of different OAM modes may be helpful to detect the low RCS target by increasing the RCS of the target.

Considering the complexity of data analysis caused by increasing urban load data, an Apache Spark based urban load analysis and forecasting technology is proposed in this paper. Firstly, the large-scale data processing platform is designed. Secondly, statistics of urban load data are processed by transformation and action in the form of resilient distributed datasets (RDDs). Thirdly, load forecasting model based on Dynamic Bayesian Network (DBN) is built for short-term urban load forecasting. GraphX is used for graph-parallel computation. Finally, the result of load analysis is presented and forecasting results of DBN are compared with other models. The comparison outcome shows that the DBN model implemented in Spark has better accuracy in load forecasting.

A polarization and channel calibration technique used in dual polarization eight channel mono-pulse radar system is introduced in this paper. System error model is proposed. The method is based on equipments test and microwave chamber test without any calibrators. Cross-polarization error and amplitude-phase imbalances between different transmitting and receiving channels as well as different antenna sub-arrays can be extracted by using this method.

With the development of Internet, the amount of Information has been rapidly growing which is spread widely. In order to improve the value and accuracy of science information that is pushed in this paper, an intelligence dichotomous method for science information categorization to identify science information from massive Web news is presents. During the experiment, 85.3% recognition rate of the recognition non-tech news are realized and 82.9% accuracy rate, the results show that the method can effectively identify Web science information news and reduce the amount of independent news.

Repeated pattern is a common phenomenon in query result pages of deep web sites. The deep web back-end data can be accessed by mining repeated patterns. So far, most of the algorithms of discovering repeated pattern use traditional web information extraction methods. But the recall percentage and accuracy are not high. How to obtain the repeated pattern accurately and completely is still a difficulty. We propose a method based on the largest block strategy to discover such pattern. The core of the method is using the largest block strategy to discover the repeated pattern layer. We can quickly navigate to the region of the entity data, and then analyze the sub tree in this area, finally, get the simplified repeated pattern of the deep web site. According to the results of the experiment, this method can get the repeated pattern data more accurately and more completely than the traditional methods. It can also address the multi-pattern problem which has not been solved yet in other methods.

The flow shop scheduling problem is a branch of production scheduling, which is among the hardest combinatorial optimization problem, and represents an important area in production scheduling. In this paper, a new effective hybrid algorithm is proposed which is based on plant growth simulation algorithm for permutation flow shop scheduling with the criterion to minimize the maximum completion time (makespan). In the algorithm, a directed graph which is based on the characteristic of flow shop scheduling problem representation is developed, and an efficient exchangeable node set is proposed and incorporated into plant growth simulation algorithm. Simulation results based on well-known benchmarks and comparisons with standard genetic algorithm demonstrate the effectiveness of the proposed bionic algorithm.