arXiv subjects

Renchang Dai

Publications and source records attributed to Renchang Dai.

22 records · Page 2Linked to original sources

arXiv2018

Exploration of Bi-Level PageRank Algorithm for Power Flow Analysis Using Graph Database

Compared with traditional relational database, graph database, GDB, is a natural expression of most real-world systems. Each node in the GDB is not only a storage unit, but also a logic operation unit to implement local computation in parallel. This paper firstly explores the feasibility of power system modeling using GDB. Then a brief introduction of the PageRank algorithm and the feasibility analysis of its application in GDB are presented. Then the proposed GDB based bilevel PageRank algorithm is developed from PageRank algorithm and Gauss Seidel methodology realize high performance parallel computation. MP 10790 case, and its extensions, MP 107900 and MP 1079000, are tested to verify the proposed method and investigate its parallelism in GDB. Besides, a provincial system, FJ case which include 1425 buses and 1922 branches, is also included in the case study to further prove the proposed algorithm effectiveness in real world.

cs.DC↗

arXiv2018

Graph based Platform for Electricity Market Study, Education and Training

With the further development of deregulated electricity market in many other countries around the world, a lot of challenges have been identified for market data management, network topology processing and fast market-clearance mechanism design. In this paper, a graph computing framework based on TigerGraph database is proposed to solve a security constrained unit commitment (SCUC) and security constrained economic dispatch (SCED) problem, with parallelized graph power flow (PGPF) and innovative LU decomposition techniques, for electricity market-clearance. It also provides a comprehensive visualization platform to demonstrate the market clearing results vividly, such as locational marginal price (LMP), and is able to be utilized for electricity market operators' education and training purpose.

eess.SP↗

arXiv2018

CIM/E Oriented Graph Database Model Architecture and Parallel Network Topology Processing

CIM/E is an easy and efficient electric power model exchange standard between different Energy Management System vendors. With the rapid growth of data size and system complexity, the traditional relational database is not the best option to store and process the data. In contrast, the graph database and graph computation show their potential advantages to handle the power system data and perform real-time data analytics and computation. The graph concept fits power grid data naturally because of the fundamental structure similarity. Vertex and edge in the graph database can act as both a parallel storage unit and a computation unit. In this paper, the CIM/E data is modeled into the graph database. Based on this model, the parallel network topology processing algorithm is established and conducted by applying graph computation. The modeling and parallel network topology processing have been demonstrated in the modified IEEE test cases and practical Sichuan power network. The processing efficiency is greatly improved using the proposed method.

cs.DC↗

arXiv2018

Exploration of Graph Computing in Power System State Estimation

With the increased complexity of power systems due to the integration of smart grid technologies and renewable energy resources, more frequent changes have been introduced to system status, and the traditional serial mode of state estimation algorithm cannot well meet the restrict time-constrained requirement for the future dynamic power grid, even with advanced computer hardware. To guarantee the grid reliability and minimize the impacts caused by system status fluctuations, a fast, even SCADA-rate, state estimator is urgently needed. In this paper, a graph based power system modeling is firstly explored and a graph computing based state estimation is proposed to speed up its performance. The power system is represented by a graph, which is a collection of vertices and edges, and the measurements are attributes of vertices and edges. Each vertex can independently implement local computation, like formulations of the node-based H matrix, gain matrix and righthand-side (RHS) vector, only with the information on its connected edges and neighboring vertices. Then, by taking advantages of graph database, these node-based data are conveniently collected and stored in the compressed sparse row (CSR) format avoiding the complexity and heaviness introduced by the sparse matrices. With communications and synchronization, centralized computation of solving the weighted least square (WLS) state estimation is completed with hierarchical parallel computing. The proposed strategy is implemented on a graph database platform. The testing results of IEEE 14-bus, IEEE 118-bus systems and a provincial system in China verify the accuracy and high-performance of the proposed methodology.

eess.SY↗