A Four-Timescale Continuous-Time Model with Sliding-Window Calibration for State-of-Charge Prediction on Battery-Powered Devices

Predicting how long a battery-powered device will run before its state of charge (SOC) reaches a usable limit is a forward problem in continuous time: the quantity of interest is a first-passage time of a state driven by inputs whose timescales differ by about seven orders of magnitude, from processor load within seconds to capacity loss over the service life. A model resolved on one timescale misplaces the others. This paper develops a continuous-time model in which one state equation is closed by four terms, each owning a band: a discharge-power decomposition into baseline, user and environmental parts, an hourly integration whose power coefficients are recalibrated on a sliding window, an aging coefficient on the day-to-month band, and an Arrhenius–solid-electrolyte-interphase capacity trajectory beyond a month. Sixteen public datasets supply the coefficients and the usage statistics, and nine of them carry the parameters the model needs and serve as validation runs. Across the nine runs the mean absolute error is 6.0786 and the root mean square error is 7.0300 in SOC percentage points. The predicted mean falls below the observed mean in every run, and across the five tabulated runs the gap is 4.20 to 5.68 points, with that offset accounting for 97% to 99% of the mean absolute error in four of them, which places the dominant error in the calibration rather than in the structure. At full charge the predicted endurance runs from 3.0 h under extreme usage to 11.9 h in low-power mode, a factor of 4.0 that a user can move within by changing settings alone.

Authors

Institutions

Publication Details

Journal
Academic Journal of Applied Sciences
Published
2026-11-20
DOI
https://doi.org/10.54097/dwfm4n93
Primary Topic
Battery Modeling and Management
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Four-Timescale Continuous-Time Model with Sliding-Window Calibration for State-of-Charge Prediction on Battery-Powered Devices

Haiyang Yu
Academic Journal of Applied Sciences
Battery Modeling and Management
article

A Four-Timescale Continuous-Time Model with Sliding-Window Calibration for State-of-Charge Prediction on Battery-Powered Devices

Haiyang Yu
article en

Abstract

Predicting how long a battery-powered device will run before its state of charge (SOC) reaches a usable limit is a forward problem in continuous time: the quantity of interest is a first-passage time of a state driven by inputs whose timescales differ by about seven orders of magnitude, from processor load within seconds to capacity loss over the service life. A model resolved on one timescale misplaces the others. This paper develops a continuous-time model in which one state equation is closed by four terms, each owning a band: a discharge-power decomposition into baseline, user and environmental parts, an hourly integration whose power coefficients are recalibrated on a sliding window, an aging coefficient on the day-to-month band, and an Arrhenius–solid-electrolyte-interphase capacity trajectory beyond a month. Sixteen public datasets supply the coefficients and the usage statistics, and nine of them carry the parameters the model needs and serve as validation runs. Across the nine runs the mean absolute error is 6.0786 and the root mean square error is 7.0300 in SOC percentage points. The predicted mean falls below the observed mean in every run, and across the five tabulated runs the gap is 4.20 to 5.68 points, with that offset accounting for 97% to 99% of the mean absolute error in four of them, which places the dominant error in the calibration rather than in the structure. At full charge the predicted endurance runs from 3.0 h under extreme usage to 11.9 h in low-power mode, a factor of 4.0 that a user can move within by changing settings alone.

Academic Journal of Applied SciencesVol. 4(1)
Nanjing University of Science and Technology (CN)
Battery Modeling and Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
A Four-Timescale Continuous-Time Model with Sliding-Window Calibration for State-of-Charge Prediction on Battery-Powered Devices — Haiyang Yu · Academic Journal of Applied Sciences (2026) | TGRS Research Map | TGRS