Electric Power Data Retrieval Method Research

In the power system, with the digital and informational development of electric power industry, grid data sources diversified and complicated, A variety of distributed data sources are distributed in power generation, transmission, distribution, electricity and scheduling, marketing and other aspects, such as intelligent electric meter, sensor, information system which produce large amounts of data, prompting power data present the geometric progression growth. These data are one of the supporting elements to build a stable, reliable and efficient intelligent power grid. It not only reflects the characteristics of the internal laws of the industry, guide the power production and business management, but also reflects the development of economic and social conditions, is an important resource for the future development of electricity. In this paper, we mainly research and explore how to use the data analysis technology to quickly retrieve valuable information in the massive power data.First, the background and significance of the research on the power big data retrieval are briefly introduced. Then research the current status of research on the power of big data retrieval in domestic and foreign. At last induces and summarizes the power system data sources, data characteristics and key technologies.Secondly, in view of the characteristics of large quantity and high dimension of power grid data. Considering the efficiency of multi-dimension query in data retrieval is not high, and the retrieval result can not be multi-dimensional and integral matching. We put forward a data retrieval method for power grid section based on manifold ranking, which describe the grid section data as the corresponding points in a multidimensional vector space, and the retrieval results are obtained by the integral approximate manifold structure of the data. To make it have a high correlation with the source query to avoid the problem of high dimensional data query processing.Finally, at index building, taking the power data types into account, include both character attributes and numeric attributes, put forward a double layer hybrid index structure, which is suitable for electric power data. According to different types of data to establish different indexes in order to complete the demand for power data retrieval. In order to suit the needs of large data processing, achieve the parallel index based on the proposed index, and apply it to the distributed environment, greatly reducing the time of the query, speed up the query speed at the same time.