Sequence Training and Data Shuffling to Enhance the Accuracy of Recurrent Neural Network Based Battery Voltage Models
Bibliographic record
Abstract
Battery terminal voltage modelling is crucial for various applications, including electric vehicles, renewable energy systems, and portable electronics. Terminal voltage models are used to determine how a battery will respond under load and can be used to calculate run-time, power capability, and heat generation and as a component of state estimation approaches, such as for state of charge. Previous studies have shown better voltage modelling accuracy for long short-term memory (LSTM) recurrent neural networks than other traditional methods (e.g., equivalent circuit and electrochemical models). This study presents two new approaches – sequence training and data shuffling – to improve LSTM battery voltage models further, making them an even better candidate for the high-accuracy modelling of lithium-ion batteries. Because the LSTM memory captures information from past time steps, it must typically be trained using one series of continuous data. Instead, the proposed sequence training approach feeds a fixed window of prior data (e.g., 100 seconds) into the LSTM at each time step to initialize the memory states properly and then only uses the output at the current time step. With this method, the LSTM just requires the prior data window to be continuous, thereby allowing the handling of discontinuities. This also means that during the training process, the data can be shuffled randomly, enabling mini-batches to speed up the training significantly. When these approaches were applied, LSTM voltage estimation error was reduced by 22%, from 28.5 mV to 22.3 mV RMS error over four drive cycles and temperatures from -20 to 25°C.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".