In batch processes, online steady state identification (SSID) is important for ensuring the quality consistence of final products. This paperpresents a robust method for batch process SSID by the use of ensemble empirical mode decomposition (EEMD) and statistical test. First, EEMD and moving-window technique areadoptedto decompose batch process signalsinto multiple intrinsic mode function (IMF) components in real time. Then, by computingthe instantaneous frequencies of each IMFthroughthe generalized zero-crossing (GZC) method, the IMFs are divided into three levels corresponding to high-frequency noise, intra-batch variation, and inter-batch trend, respectively. By utilizing the inter-batch trend instead of the original signal in SSID, the identification results are robustto measurement noise and process disturbance. Injection molding, a typical batch process, is usedto demonstratethe proposed method.