In view of the characteristics of smart grid

such as instantaneity

volatility

disorder

etc.

a real-time stream processing framework of big data in the smart grid is proposed. The framework can realize data collection

data buffer and flow calculation

so as to meet the fast processing need

such as condition monitoring anomaly detection and analysis of electric behaviors. In order to solve the problem of inconsistency in the speed of data acquisition and flow calculation

the data source change is monitored

while the data are collected in real-time and buffered by using the message subscription mode. A Storm-based sliding window processing approach to monitoring the data stream is proposed to process the condition monitoring data stream in batches within the specified time

while ensuring continuous data computation and detecting abnormality by threshold determination. Experimental results show that

under the condition of a definite cluster size

appropriately changing the number of working processes and the number of concurrent execution threads can increase the tuple throughput of sliding window and improve the real-time processing efficiency of condition monitoring anomaly detection. This work is supported by National Natural Science Foundation of China(No.61074078).