Search

At present, several provinces' power supply companies have established harmonic monitoring platform in China, by real-time monitoring the voltage and current of power grid, then upload these data to the main station to analyze and settle. The platform can collect about 1 TB(1TB=1000GB) data every day[12], these data contain all the indexes of power quality, but we can only use this platform to do some basal data analysis work, but not processing analysis with multi-indexes, so that it does not show strong supporting effort to the power grid operation. With the development of energy internet and the third industry, more and more unstructured data are impacting the power grid operation, how to make better use of big data technology to deal with the data from monitoring platform has became the focus of the power grid company. In this paper, some introduces of contents and characteristics of harmonic real-time monitoring platform and big data technology has been made, by comparing the way of using and settling data of platform to the big data technology, meanwhile by combining with the characteristics of power quality data, we analyze the combination of the two prospects. At last, the development direction of the advanced application for the harmonic line monitoring system is discussed in this paper.

This paper improves Itai and Rodeh's algorithm for finding a shortest cycle in an undirected unweighted graph. Given an undirected unweighted graph G with n vertices,the algorithm can determine a cycle whose length is at most k+1 where k is the length of the shortest cycle in G. The results of the experiment show that the improved algorithm runs faster than the original one when the input graph meets some special features.

Given an undirected graph with vertex and edge weights and a subset of vertices called terminals, the bounded size graph clustering (BSGC) problem is to partition the vertices into clusters of size at most a given budget such that each cluster contains at most one terminal and the total size of the clusters is minimized, where the size of a cluster is defined as the total vertex weight in the subset plus the total edge weight at the boundary of the cluster. For BSGC on trees, we present a pseudo-polynomial time exact algorithm, which can be modified via rounding to yield a (1+\epsilon)-approximation in polynomial time violating the given budget by a 1+\epsilon factor.

Given a graph G=(V,E) with real positive edge weights, where each edge (u, v) is labeled either + or – depending on whether u and v have been deemed to be similar or dissimilar, the problem of correlation clustering with l clusters is to partition the vertices of G into at most l clusters to minimize the total weight of + edges between clusters and – edges within clusters. This problem for general graphs is APX-complete. A fixed-parameter tractable (FPT) algorithm is presented, where the parameter k is defined as the maximum, over all blocks B of the graph, of the number of vertices that must be removed from B to obtain a forest.

The k-means clustering is one of the most popular schemes for discovering clusters in data. Its aim is to minimize the mean squared distance from each data to its nearest center. A lot of variants of Lloyd's heuristic have been studied. Unfortunately, many research results haven't given any approximate ratio. In this paper, an algorithm is presented which can obtain the optimal clustering with the ration of at most 2. The main idea of this algorithm is that k points are selected by means of a very simple, randomized seeding technique and then the local search is implemented to improve the accuracy. Some examples are selected to verify our algorithm and got better results both the speed and accuracy than the former methods. The main algorithmic contribution is that the input points are used as the candidate sets to obtain the optimal clustering with a constant ratio by means of local search technique and one method of selecting initial points.

A new kind of neural networks on solving shortest paths problem is presented, and the stability of the neural networks is proved. For a directed or undirected graph, the proposed neural network is always convergent to its single equilibrium, so that the shortest paths between every pair of nodes of the oriented graph are produced. This paper makes a new study on solving the optimization but not NP-hard problems by neural networks.

This paper gives an improved sufficient condition of ensuring the stability of delayed cellular neural networks (DCNN) by means of the analysis of a correspondent Lyapunov function. This result improved the limit given by Civalleri and gives an optional upper limit of the stability of DCNN.

It is difficult to reduce an NMR image's ringing artifacts and obtain better images. This paper uses the method of dyadic wavelets and self-adaptive threshold to process the images that have the ringing artifacts. The test results prove that the method is effective and it can process the NMR image's ringing artifacts. Finally the paper discusses using the fuzzy method.

In this paper the character of the associative memory is discussed emphatically from the viewpoint of the storage capacity and the associative ability of the network. The associative memory function can all be realized by utilizing the feedback network on competition. The competing classifying (CC) algorithm is presented here. Compared with any other associative memories that already exist, the neural network determined by the algorithm has very good performances such as high error-correcting capability, large storage capacity, no "rejected-point" and no spurious attractive center. The obtained network exceeds the storage capacity of HAM and BP networks.

The character of the associative memory is discussed emphatically from the viewpoint of the storage capacity and the associative ability of the network. The associative memory function can all be realized by utilizing the feedback network or the forward-feed network. The new algorithm, based on adaptive competing-classifying (ACC) algorithm, is presented here. Compared with other associative memories, the neural network determined by the algorithm has high error-correcting capability, large storage capacity, no "rejected-point" and no spurious attractive center. ACC algorithm can decrease the competition-number more than CC algorithm. The obtained network has the very strong ability of correcting errors and exceeds the storage capacity of HAM and BP networks.