论文部分内容阅读
利用人工神经网络具有从大量不完整数据中逐步获取知识并进行复杂目标优化的能力,提出一种基于改进的BP神经网络预测功耗可管理器件空闲时间长度的自适应动态电源管理模型(AMBA).该模型根据空闲时间长度预测值选择最佳低功耗模式,并且通过预测性能评估结果动态调整神经网络的网络权值,从而能够自适应功耗可管理器件负载特性的变化.仿真结果表明了该模型的有效性,预测正确率可以达到80%左右,平均功耗比指数平均预测方法降低了30%,比Timeout策略降低了37%.
Based on the ability of artificial neural network to acquire knowledge gradually from a large amount of incomplete data and to optimize complex targets, an adaptive dynamic power management model (AMBA) based on improved BP neural network to predict the idle time of power manageable devices is proposed. The model chooses the best low-power mode according to the predicted value of idle time, and dynamically adjusts the network weights of the neural network by predicting the performance evaluation results, so as to be able to adaptively change the load characteristics of the power manageable device.The simulation results show that The validity of this model can be predicted to be about 80%. The average power consumption is 30% lower than the exponential average forecasting method and 37% lower than the Timeout strategy.