Deep Learning-Assisted Energy Prediction Modeling for Energy Harvesting in Wireless Cognitive Radio Devices

Abstract

Cognitive radio is a technology that allows Secondary Users (SUs) to access vacant spectrum areas allocated to Primary Users (PUs) by dynamically adjusting their settings. However, the spectrum detection subsystem of SUs consumes battery power that could be used for transmission. This work aims to address the energy availability issue for cognitive radio devices by two methods: energy harvesting from the ambient environment and deep learning prediction of future energy levels. We compare three deep learning models: Long-Short Term Memory (LSTM), Convolutional Neural Network (CNN), and Convolutional Long-Short Term Memory (ConvLSTM) with three classic machine learning models: Artificial Neural Networks (ANN), Support Vector Regressor (SVR), and Extreme Gradient Boost (XGBoost). The results show that deep learning models outperform machine learning models across all datasets, with ConvLSTM being the best model with a Normalized Root Mean Squared Error (nRMSE) of 0.0632 and Mean Absolute Error (MAE) of 1.479, which are 8.80% and 9.04% better than the best machine learning model, ANN, with nRMSE of 0.0693 and MAE of 1.626.