Implementation of EN-Elman Backprop for Dynamic Power Management

Power utilization has become a major issue in portable designs, since its battery storage is less compared to its usage. One of the popular techniques to solve this problem is to use Dynamic Power Management (DPM) at the system level. Dynamic power management is a technique used to save power when the system is idle. Earlier it was assumed that the prediction can be done only in long range dependent systems which may be a random process or short range dependent. But a single user will not work similarly the next time, so a single assumption will not hold good. To overcome the above assumptions, we propose an Elman Model which uses Moving Average, Elman Back prop network and random walk model to predict the idle period. Here we use Artificial Neural Network (ANN) in which we train the neurons in a particular way the user desires, replacing neurons by time series we can calculate how much power is saved. Our enhanced version consequently reduces energy consumption, noise and cooling requirements. There by we achieve prolong battery life. By simulation we can show that this method achieves higher power saving compared to other methods.