برنامه های شبیه سازی شبکه برق توسعه یافته با استفاده از چارچوب رایانش عملکرد بالای GridPACKTM Power grid simulation applications developed using the GridPACKTM high performance computing framework

۱٫ Introduction Despite the complexity of the modern power grid, commercial software for simulating its behaviors is mainly designed for sequential computation on single-processor workstations. High performance computing (HPC), which is capable of tackling much larger problems and solving them in much shorter time, has made few inroads into power grid simulation for real-world applications. A main reason that HPC is not more widespread in power grid simulation is that the complexity and cost of developing parallel algorithms and implementing them in parallel software are high. The need for high performance computing has been recognized for decades, and many researchers have devoted efforts to this topic. These prior efforts examined the HPC topic from individual application perspectives without a systematic view on how applications can benefit from a common set of software modules. For example, references [1–۵] discuss parallel implementation of a state estimation application; [6–۹] introduce massive static contingency analysis that utilizes a dynamic load-balancing scheme, and [10–۱۴] discuss multiple implementations of dynamic simulation. As a result, each of these efforts required significant parallel programming skills as well as power engineering expertise, and the resulted outcome is a one-off implementation. The barrier to adopting these HPC applications is high. We recognize that developing HPC-based power grid applications typically requires strong parallel programming skills that are outside the focus of most power grid modelers and power grid modelers are more interested in developing their modeling capabilities than improving performance using complex programming techniques. It is therefore highly desirable to provide a common set of reusable parallel software modules on a unified framework supporting multiple applications and thus to lower the barrier to the adoption of HPC for power grid simulation. Power system applications typically contain many common components such as the formulation of an admittance matrix (i.e. Y-bus matrix) [15], or contain component simulations that couple with each other. In addition, some common functions such as solving the power flow equations or integrating differential-algebraic equations for dynamic simulation, are used across multiple applications. For example, power flow solution is an important part of contingency analysis, dynamic simulation, and other applications. Reproducing these functions in different applications leads to duplication of efforts and reduces programmers’ productivity. Efforts are underway to promote software reuse and simplify the process of developing parallel software codes for grid applications. The GridPACK, as a result of such efforts, provides open-source software for use by the power grid community to accelerate the adoption of high performance computing. The goal of GridPACK is to simplify the task of developing parallel software for power grid applications that can run on high performance architectures. GridPACK accomplishes this goal by encapsulating many function modules of power grid computation in high level abstractions that hide from users the details of data communication and distribution, which are necessary for parallel software implementation. Application developers can focus on the physics and mathematics of framing and modeling their problems, without being overwhelmed by the bookkeeping required to distribute data over multiple processors and to access it in a distributed context. These abstractions also make it simpler for users to access high performance computing libraries, as well as promoting software reuse and reducing the costs of software development and maintenance. This paper will describe the development and performance of several power system applications to illustrate the functionality and benefits of GridPACK. The rest of this paper is organized as follows: Section 2 provides a brief description of GridPACK. Section 3 discusses the example implementation of several applications using GridPACK. Section 4 concludes the paper with a discussion of future work.