Risk-aware hybrid reinforcement learning and robust optimization for reliable PV-battery scheduling under uncertainty

Abstract The increasing penetration of photovoltaic (PV) generation and battery energy storage systems (BESSs) has intensified the need for intelligent scheduling strategies that can simultaneously reduce operating cost, preserve battery health, and ensure operational reliability under significant uncertainty. In practical grid-connected PV–battery systems, the stochastic nature of solar power, load demand, and electricity prices often leads to constraint violations and high tail-risk costs when conventional optimization or purely data-driven methods are used. To address this challenge, this paper proposes a risk-sensitive hybrid framework that integrates Soft Actor-Critic (SAC), Conditional Value-at-Risk (CVaR), conformal uncertainty estimation (CUE), and robust optimization (RO) in a unified energy management architecture. The SAC agent learns a stochastic dispatch policy, while CVaR explicitly penalizes unfavorable cost outcomes and CUE provides distribution-free uncertainty bounds for uncertain inputs. These bounds are then embedded into a robust optimization layer to guarantee feasibility under adverse operating conditions. In addition, an adaptive fusion mechanism combines the outputs of the learning-based and robust controllers to balance economic performance and reliability. Simulation results for a typical residential PV–battery system demonstrate that the proposed method achieves the best trade-off among cost, risk, and constraint satisfaction. Compared with a rule-based baseline, the full model reduces the SOC violation rate from 4.8% to below 1%, while also lowering the total operating cost and CVaR. Moreover, the proposed framework outperforms the RO-only variant by achieving a lower average cost without sacrificing robustness, and it consistently provides the highest reliability across independent runs. These results confirm that the proposed hybrid approach offers a practical and scalable solution for secure and economical PV–battery scheduling under uncertainty.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-15
DOI
https://doi.org/10.1038/s41598-026-70541-2
Primary Topic
Smart Grid Energy Management
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Risk-aware hybrid reinforcement learning and robust optimization for reliable PV-battery scheduling under uncertainty

Mehrdad Tarafdar Hagh, S.H. Hosseini, Bareq Kawwam Ahmed
5 links
Scientific Reports
Smart Grid Energy Management
article

Risk-aware hybrid reinforcement learning and robust optimization for reliable PV-battery scheduling under uncertainty

Mehrdad Tarafdar Hagh, S.H. Hosseini, Bareq Kawwam Ahmed
article en

Abstract

Abstract The increasing penetration of photovoltaic (PV) generation and battery energy storage systems (BESSs) has intensified the need for intelligent scheduling strategies that can simultaneously reduce operating cost, preserve battery health, and ensure operational reliability under significant uncertainty. In practical grid-connected PV–battery systems, the stochastic nature of solar power, load demand, and electricity prices often leads to constraint violations and high tail-risk costs when conventional optimization or purely data-driven methods are used. To address this challenge, this paper proposes a risk-sensitive hybrid framework that integrates Soft Actor-Critic (SAC), Conditional Value-at-Risk (CVaR), conformal uncertainty estimation (CUE), and robust optimization (RO) in a unified energy management architecture. The SAC agent learns a stochastic dispatch policy, while CVaR explicitly penalizes unfavorable cost outcomes and CUE provides distribution-free uncertainty bounds for uncertain inputs. These bounds are then embedded into a robust optimization layer to guarantee feasibility under adverse operating conditions. In addition, an adaptive fusion mechanism combines the outputs of the learning-based and robust controllers to balance economic performance and reliability. Simulation results for a typical residential PV–battery system demonstrate that the proposed method achieves the best trade-off among cost, risk, and constraint satisfaction. Compared with a rule-based baseline, the full model reduces the SOC violation rate from 4.8% to below 1%, while also lowering the total operating cost and CVaR. Moreover, the proposed framework outperforms the RO-only variant by achieving a lower average cost without sacrificing robustness, and it consistently provides the highest reliability across independent runs. These results confirm that the proposed hybrid approach offers a practical and scalable solution for secure and economical PV–battery scheduling under uncertainty.

Scientific Reports
University of Tabriz (IR)
University of Tabriz
Affordable and clean energy, Industry, innovation and infrastructure, Climate action
Smart Grid Energy 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