A Systolic Accelerator for Power System Security Assessment

This paper presents an application of a neural-network algorithm, the Kohonen feature map, to security monitoring in power transmission systems, and its implementation on parallel hardware. The computational requirements of the application led to the development of an SIMD processor array dedicated to neural networks. The heart of this system is a systolic array of simple processing elements (PEs). A VLSI custom chip containing a sub-array of 2 2 PEs has been built and the full system may contain up to 40 40 PEs. The system is designed to sustain sufficient instruction and data flows to keep the utilization rate close to 100 %.