1 / 34

Graph Based Next Generation Energy Management System

Graph Based Next Generation Energy Management System. Renchang Dai GEIRI North America June 18, 2019. Next Generation EMS Roadmap. 05-The final goal is Robot EMS 02-The industry is at the stage to make EMS full functional 03-The critical path to meet the gap is faster than real-time EMS.

spencerm
Télécharger la présentation

Graph Based Next Generation Energy Management System

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Graph Based Next Generation • Energy Management System • Renchang Dai • GEIRI North America • June 18, 2019

  2. Next Generation EMS Roadmap • 05-The final goal is Robot EMS • 02-The industry is at the stage to make EMS full functional • 03-The critical path to meet the gap is faster than real-time EMS • Next Generation EMSGoals: • Provides real-time, proactive, intelligent, and predictable operation system in control center. • Employs graph database, parallel computation,natural language processing, deep learning,and situation awareness and autonomous dispatch to drive anlaytical EMS to intelligent/robot EMS.

  3. Faster-than-Real-Time EMS • Why We Need it Fast Change of Operation in Minutes Due to the Renewable Intermittence Power Electronics Dominated Large and Complex System Current EMS cycle delays responses to the cascding events driving the blackout Today's View > 20 seconds Needed View < 0.5 seconds Sequence of Cascading Events in the 2011 Southwest Blackout in the US

  4. Faster-than-Real-Time EMS • Why We Need it • Subseconds analytical processing time enabling the critical functions: • Security Constrained Automatic Generation Controls • On-line SPS Arming Decision • Multi-time Point Analysis Pathway to Speed Improvements in Analytical Decision Making • The analytical processing time needs to be reduced, from tens of seconds to subseconds, to move from monitoring and visualization to automatic controls. • The need for fast and predictive analytics is amplified by physical and cyber attack on critical infrastructure. • US DOE requires to develop State Estimation at 0.5 seconds speed for medium size system. DOE funded $220 Millions to Faster Real-time Analytical Tools.

  5. Faster-than-Real-Time EMS • What We Achieved Source: GEIRINA EMS Prototype Source: Commerical EMS

  6. Faster Than Real-time EMS

  7. RDB for Power System Modeling • Physical System • Nodes are connected by edges • Connectivity is naturally a graph • Relational Database • Use table structure • Not support unstructured data • Attribute relations modelled by separated tables • Use commonly shared key values to represent data relationships • Issues of Relational Database for Power System Modeling • Join intensive queries for the whole database invite large computation time • Maintain small portion of system requires multiple table update • Time consuming to support recursive queries and parallel queries

  8. RDB and GDB Comparison Modeling Different Data Structure Comparison Operation Efficiency Comparison • Power System, as a Graph Physically, is a Good Fit for Graph Database to Model • Power System Analysis Needs Graph Computing to Parallelize

  9. RDB based Power System Computing Core Calculation Output Traversal Relational Database Data Preprocess • Issues of Relational Database for Power System Computing • Need loop through branch table and bus table to create connectivity • Complicated to support linear equation parallel computing • Map solved variables to bus voltages and branch flows inviting time consuming output traversal

  10. RDB Time Decomposition • A Real Case of2650 bus system Data source: Beijing Kedong Electric Power Control System Co., LTD. • Observations: • PF Data Input Processing and Output Traversal cost 94.5% of the total time • SE Data Input Processing and Output Traversal cost 64.3% of the total time • Need a new platform to integrate data management, calculation, and visualization

  11. Graph Potentials - Parallel Computing - PF • Using graph database model to save the time of • Input Processing • Ouptut Traversal (s) • Further improve Core Calculation efficiency by • Graph based Nodal Parallel Computing • Parallel LU Solver Hierarchical Parallel Computing Time Decomposition (ms) • Observations: • Matrix formation and P/Q calculation cost 45% of the total time which can be nodal parallelized. • Matrix factorization and F/B substitution take 51% of the total time which can be hierarchical parallelized.

  12. Graph Potentials - Parallel Computing - SE • Using graph database model to save the time of • Input Processing • Ouptut Traversal • Further improve Core Calculation efficiency by • Graph based Nodal Parallel Computing • Parallel LU Solver Hierarchical Parallel Computing (s) Time Decomposition (ms) • Observations: • Gain matrix formulation and right-hand-side vector update take ~60% of core computation time which can be implemented by nodal parallel computing. • Gain-matrix factorization and forward/backward substitution cost ~20% of time which can be hierarchically parallelized.

  13. GDB for Power System Modeling • Relational Database • Use table structure • Attribute relations modelled by separated tables • Need to update multiple table to maintain system • Hard to support recursive queries and parallel queries Original Data Document • Physical System • Edges are connecting by nodes • Connectivity is naturally a graph • Graph Database • Use graph structure with edges and nodes • Store data by attributes of nodes and edges • Support parallel computing • Easy to maintain large system Relational Database: Data model is a collection of interlinked tables. Graph Database: Data model is a multi-relational graph.

  14. GDB based Power System Computing 4 1 3 2 5 V1(100,50,0,0) V4(0,0,40,5) V3(0,0,45,15) E34(0.01,0.03) E13(0.08,0.24) E23(0.06,0.25) E12(0.02,0.06) E24(0.06,0.18) E45(0.08,0.24) E25(0.04,0.12) V5(0,0,60,10) V2(150,80,20,10) • RDB Based Computing • GDB Based Computing • Advantages of Graph Computing for Power System • Integrate system modeling, core computing, and result visualization in graph database • Data input preprocessing and result output traversal are not need • Change calculation approach from serial computing to parallel queries • Implement a suite of computation queries as library for power system applications

  15. GDB based EMS Architecture 4 1 3 2 5 V1(100,50,0,0) V4(0,0,40,5) V3(0,0,45,15) E34(0.01,0.03) E13(0.08,0.24) E23(0.06,0.25) E12(0.02,0.06) E24(0.06,0.18) E45(0.08,0.24) E25(0.04,0.12) V5(0,0,60,10) V2(150,80,20,10)

  16. Graph Computing 3 7 1 3 7 1 6 8 6 8 4 10 4 10 9 2 9 2 5 5 Graph Nodal-Hierarchical Parallel Factorization Symbolic Factorization Numerical Factorization Forward/Backward Sub Ordering Graph Structuring Graph Computing Graph Transforming Nodal Parallel Hierarchical Parallel G(A) G+(A) T(A) Node Partition Graph Structuring: G(A)Degrees Calculating, Node Optimal Ordering, Elimination Tree building Graph Transforming: Factorization Graph Building, Graph Attributes Updating Graph Computing:Forward/Backward Substituting on Graph Perform Computing on Graph in Parallel other than on Matrix in Series

  17. Graph Computing: A Real Case • Relational Database • Preprocessing: Search bus tables and branch tables to find connectivities. • Output Traversal:Map solved variables to bus voltage and branch flow. • Graph Database • No preprocessing. Connectivity is predefined in graph database. • No output traversal. Solved bus voltage and branch flow are attributes of nodes and edges in graph database. Relational Database Graph Database and Computing 2650 Bus System Model in GDB

  18. GDB for Node-Breaker Model Node – Breaker Graph Model • Based on Base-value, Substation, Bus, AC line, Unit, Transformer, Load, Compensator, Converter, DC line, Island, Topo-node, Breaker and Disconnector are modeled by vertices • Common Information Model (CIM) Substation representation in one-line diagram and CIM/E Substation representation in one-line diagram and CIM/E Substation modeling in CIMGDB Substation modeling in CIMGDB Graph Topology Processing Bridge Node-Breaker to Bus-Branch Model

  19. Graph Based Faster-Than-Real-Time EMS Platform • Complete EMS applications at SCADA sampling rate (5 seconds) • SE/PF/CA including topology processing is faster than real time, completed within 5 seconds • Visualize applicatoin status, start time, and execution time • Voltage heat map and operational serverity index show overall operational risk • Substation diagram is automatically dynamically drawn and shows the detailed operations

  20. Graph Based Testing Results: Newton-Raphson PF (ms) Testing System 1: 2650 Bus System

  21. Graph Based Testing Results: Newton-Raphson PF Testing System 2: MP10790 System (ms)

  22. Graph Based Testing Results: Fast Decoupled PF (ms) Testing System 1: 2650 Bus System

  23. Graph Based Testing Results: Fast Decoupled PF Testing System 2: MP10790 System (ms)

  24. Substation One-Line Diagram AutomaticGeneration and Visualization-From Graph to One-Line Diagram

  25. Substation One-Line Diagram Automatic Generation CIM/E Model Graph Model One-line Diagram From Graph to One-line Diagram Steps • Bus-bar Configuration Pattern Recognition • Branch Recursive • Bus-bar and Branch Layout Optimization

  26. Bus-bar Configuration Pattern Recognition and Layout Pattern Recognition • Identify Voltage Islands • Count the Number of Bus-bars • Recognize Bus-bar Configuration Pattern Bus-bar Layout • Positioning by Transformer Number and Type • Sizing by Branches and Breakers • Spacing by Configuration Pattern

  27. Branch Layout Branch Layout • Recursive Solution • Depth First Search • Spacing by Numbers

  28. Substation Level Layout Optimization Problem • Minimize the Length of Connections • Minimize the Number of Intersections • Constraints on the Space of Inlets and Outlets

  29. Potential Applications Transient Stability Analysis Electromechanical Transient Stability Graph Computing Sequential Method + Electromagnetic Transient Stability

  30. Conclusions • EMS cycle shall be speeded up as power system is significantly evolving to be larger and more complex with more power electronics, higher uncertainty, and faster events. • Fast and predictive analytics are critical to respond the cascading events, avoid the severe blackouts, and enable the advanced system automatic control. • High performance computing is critical to meet the gap on the pathway to EMS Robot. • Graph database and graph parallel computing are promising to achieve Faster-than-Real-Time EMS. • One-line diagram automatic generation tool saves modeling time from months to hours.

  31. Thank You! Q&A

  32. Back Up

  33. Table 4-2 IEEE 118 bus system result comparison

  34. 国网1万节点系统测试结果(速度比较) MP10790PQ-分解法(毫秒) 四川2650PQ-分解法(毫秒)

More Related