Robustness of predictive energy harvesting systems: Analysis and adaptive prediction scaling

Abstract

Internet of Things (IoT) systems can rely on energy harvesting to extend battery lifetimes or even render batteries obsolete. Such systems employ an energy scheduler to optimise their behaviour and thus performance by adapting the system's operation. Predictive models of harvesting sources, which are inherently non‐deterministic and consequently challenging to predict, are often necessary for the scheduler to optimise performance. Because the inaccurate predictions are utilised by the scheduler, the predictive model's accuracy inevitably impacts the scheduler and system performance. This fact has largely been overlooked in the vast amount of available results on energy schedulers and predictors for harvesting‐based systems. The authors systematically describe the effect prediction errors have on the scheduler and thus system performance by defining a novel robustness metric. To alleviate the severe impact prediction errors can have on the system performance, the authors propose an adaptive prediction scaling method that learns from the local environment and system behaviour. The authors demonstrate the concept of robustness with datasets from both outdoor and indoor scenarios. In addition, the authors highlight the improvement and overhead of the proposed adaptive prediction scaling method for both scenarios. It improves a non‐robust system's performance by up to 13.8 times in a real‐world setting.

Optimal Neural Network for Predicting Solar Energy in Sensor Units Based on a Cascaded Input/Structure Direct Optimization

“…Thus, more execution time and more computational efforts would be required. As a solution, ANNs were applied offline with those sensors (i.e., implemented by a device with large computational efforts like a PC followed by extracting the resulted weights and biases Journal of Sensors to be inserted in the microcontroller as fixed numbers for a specific location) [33][34][35].…”

Predictive Energy-Aware Adaptive Sampling with Deep Reinforcement Learning

“…Although energy harvesting wireless sensor nodes should manage energy consumption well to prevent service disruption, finding an optimal energy management strategy at design time is difficult because the amount of energy available for harvesting changes over time [3]. For example, in the case of solar energy, the solar irradiance may change dramatically depending on the weather [4].…”