Abstract: With rapid development of ultra-high voltage (UHV) power grid in China, a single fault may lead to serious blackout accidents. However, existing fault alarm functions cannot meet power grid dispatching requirement. A fault analysis mechanism based on combination of alarm reliability and data availability is established to compress, fuse, filter and abstract alarm data from various data sources. The improved fault alarm analysis results are visualized and easily accessible by operators. Application results show that proposed smart power grid alarm techniques can significantly improve both alarm reporting rate and accuracy. It provides powerful support for power grid fault prediction and processing.