In existing methods for processor energy estimation, the information on the internal state of processor architecture, i.e., (de)activation of specific modules such as pipeline stages and the cache, is available via a cycle-accurate simulator. Despite good accuracy, above methods are too slow due to the complexity of cycle-accurate simulation. This paper describes a new technique for estimating the energy consumption of processors by predicting the internal information of processor architecture through a cycle-approximate simulator. We have evaluated our approach in comparison with an existing energy estimation method using a cycle-accurate ARM v5 architecture simulator. Experimental results shows that the proposed method achieves 97% accuracy and 28 times speed-up on average.