State-Adaptive and Tail-Risk-Aware Rolling Control of Wind–Solar–VRFB Systems with Experimentally Calibrated Power Envelopes
- 1School of Electrical Engineering, Guangzhou City University of Technology, Guangzhou 510800, China
Abstract
State-adaptive risk-aware rolling model predictive control (SA-RR-MPC) links experimentally calibrated battery action limits to grid-side tracking and ramp-risk costs in wind–solar–vanadium redox flow battery (VRFB) regulation. Protocol-matched PNNL cycling observations supply power envelopes conditioned on a continuous charge reference and measured capacity, while Mendeley pressure–flow measurements supply the normalized pump-loss shape. Pump-corrected net injection connects admitted stack actions to scenario conditional value at risk (CVaR). A two-pass linear program selects dispatch requests, and current-state projection provides minimum-correction execution with an explicit bound on the resulting ramp-risk perturbation. Whole-cycle calibration checks give 89.95% Q90 coverage, and the local power labels change by at most 0.205% under the common charge-reference comparison. A causal 184-day NSRDB/WIND Toolkit replay evaluates a nominal 20 MW/80 MWh VRFB under common empirical-envelope and energy protection for all storage methods. SA-RR-MPC reduces realized ramp-exceedance CVaR90 by 40.94% and request-to-execution correction energy by 65.64% relative to Fixed-MPC, with a 0.54% increase in tracking RMSE. Relative to a matched risk-aware controller differing only in predictive power bounds, correction energy falls by 58.48%. Two nonoverlapping 28-day weather windows and three paired scenario seeds retain 27.06–50.01% CVaR90 reductions relative to Fixed-MPC. Paired module tests identify complementary roles in request adaptation, net-power prediction, and tail-risk valuation. The 95th-percentile serial software-loop time is 18.54 ms over 624 cycles after 48 warm-up steps.
Related
- Finite element analysis-based flatness control and bending force optimization strategy for high rolling force conditions
- Adaptive Voiding Enables Interfacial Morphological Control at Low Stack Pressure in Solid-State Batteries
- Multi-sensor IoT data and AI-driven adaptive control model for library indoor environments
- Digital-Twin Attention-Augmented Adaptive Model Predictive Control for Degradation-Aware Residential Battery Energy Management
- An adaptive energy management and control framework for long-endurance UAV missions in dynamic environments: A reproducible simulation study
BibTeX
@article{cen2026stateadaptive,
title = {{State-Adaptive and Tail-Risk-Aware Rolling Control of Wind–Solar–VRFB Systems with Experimentally Calibrated Power Envelopes}},
author = {Yuanfa Cen and Ming Gao and Baichi Ou and Rushi Zhao and Xiaoqing Li},
journal = {Batteries},
year = {2026},
month = nov,
doi = {10.3390/batteries12110448},
url = {https://www.mdpi.com/2313-0105/12/11/448},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - State-Adaptive and Tail-Risk-Aware Rolling Control of Wind–Solar–VRFB Systems with Experimentally Calibrated Power Envelopes
AU - Yuanfa Cen
AU - Ming Gao
AU - Baichi Ou
AU - Rushi Zhao
AU - Xiaoqing Li
PY - 2026
DA - 2026/11/01
JO - Batteries
DO - 10.3390/batteries12110448
UR - https://www.mdpi.com/2313-0105/12/11/448
AB - State-adaptive risk-aware rolling model predictive control (SA-RR-MPC) links experimentally calibrated battery action limits to grid-side tracking and ramp-risk costs in wind–solar–vanadium redox flow battery (VRFB) regulation. Protocol-matched PNNL cycling observations supply power envelopes conditioned on a continuous charge reference and measured capacity, while Mendeley pressure–flow measurements supply the normalized pump-loss shape. Pump-corrected net injection connects admitted stack actions to scenario conditional value at risk (CVaR). A two-pass linear program selects dispatch requests, and current-state projection provides minimum-correction execution with an explicit bound on the resulting ramp-risk perturbation. Whole-cycle calibration checks give 89.95% Q90 coverage, and the local power labels change by at most 0.205% under the common charge-reference comparison. A causal 184-day NSRDB/WIND Toolkit replay evaluates a nominal 20 MW/80 MWh VRFB under common empirical-envelope and energy protection for all storage methods. SA-RR-MPC reduces realized ramp-exceedance CVaR90 by 40.94% and request-to-execution correction energy by 65.64% relative to Fixed-MPC, with a 0.54% increase in tracking RMSE. Relative to a matched risk-aware controller differing only in predictive power bounds, correction energy falls by 58.48%. Two nonoverlapping 28-day weather windows and three paired scenario seeds retain 27.06–50.01% CVaR90 reductions relative to Fixed-MPC. Paired module tests identify complementary roles in request adaptation, net-power prediction, and tail-risk valuation. The 95th-percentile serial software-loop time is 18.54 ms over 624 cycles after 48 warm-up steps.
SN - 2313-0105
ER -