Status : Verified
| Personal Name | Amarga, Ydron Paul C. |
|---|---|
| Resource Title | Online state of charge estimation for wireless sensor network nodes |
| Date Issued | June 2021 |
| Abstract | Existing state of charge (SOC) estimation methods for wireless sensor network (WSN) nodes primarily use mathematical models with parameters that are not updated on the fly. This results in erroneous SOC values that degrade the performance of energy-aware optimizations. Furthermore, these SOC estimation methods are often calibrated on empirical datasets, which discourages rapid WSN deployment. As a solution, this thesis proposes "Collaborative Nodes and Model Parameter Recalibration" (CNMR) and its variants which exploit collaboration between different nodes of the WSN for battery profiling and model parameter identification. This approach has not been explored elsewhere, to the best of our knowledge. CNMR recalibrates both the Equivalent Circuit Model (ECM) and segmented Open Circuit Voltage (OCV) model without the need for extensive datasets. Moreover, modifications on CNMR are implemented to improve its performance in terms of SOC estimation accuracy and sensor network lifetime. Partial CNMR I recalibrates its segmented-OCV model but uses numerical algorithms for subspace state-space identification (N 4SID) and prediction error minimization (PEM) for its predetermined battery model parameters. Finally, Partial CNMR II recalibrates its battery model and uses piece-wise least squares regression for the predetermined segmented-OCV model. The methods are assessed through experimental setups with full-length and variable-length battery datasets based on the loading profiles of WSN applications. The evaluations and hypothesis tests through Bayes factors (BF) show that these methods outperform non-collaborative SOC estimation approaches in low-current applications with variable length battery datasets. |
| Degree Course | MS Electrical Engineering |
| Language | English |
| Keyword | Bayes factors; Kalman filter; least squares regression; state of charge; wireless sensor networks |
| Material Type | Thesis/Dissertation |
Preliminary Pages
3.89 Mb