Emission-aware operation of residential PV-battery systems: the role of information structure
Abstract
Residential batteries can reduce the carbon emissions of households with photovoltaic (PV) generation, but how much depends on how they are operated under uncertain load, PV generation, and grid carbon intensity. We cast the operation of a residential PV-battery system as a sequential decision problem and compare, under a common and explicit evaluation protocol, a family of linear policies that differ in the information available at decision time, in how they anticipate the future, and in their adaptivity: a rule-based self-consumption policy, an open-loop stochastic schedule, affine feedback policies, receding-horizon policies with deterministic, scenario-based and two-stage lookahead models, and perfect-information benchmarks. Scenarios are generated by a joint principal component model of the three uncertain quantities, which supports conditional sampling during the day. Experiments on Italian data at 15-minute resolution comprise single-day evaluations on 10000 test scenarios in three seasons and a closed-loop simulation over more than eight months of real data. When the state of charge is treated consistently across policies, the self-consumption rule is within 3.2% of the perfect-foresight emissions in closed loop, whereas open-loop schedules and receding-horizon policies that decide before measuring the current period are about 20--35% above it. With real-time measurements, scenario-based receding-horizon control consistently outperforms its deterministic counterpart and, for households that rely more on the grid, reduces emissions by about 5% with respect to the rule. The benefit of emission-aware optimization is therefore determined mainly by the information structure of the controller and by the size of the system relative to the demand.
Related
- Stay, Reposition, or Recharge: Battery-Aware Operation of Robotaxi Fleets
- SN-ASMO: Satellite-Navigation Array Spatial-Manifold Precise Observation Theory A Mathematical Foundation for Observation Formation, Unified U(1) Geometry, Intrinsic Information, and Preservation of the RTK Integer Structure
- DOMIC: Provably Calibrated Detection of Dependence-Structure Change Points via Density-Operator Mutual Information
- Structure-Aware Online Optimization of the Age of Information with Unknown System Statistics
- The Sombor deletion correction term: extremal structure, exact coincidence classification, and the information content of edge decks, with applications to chemistry
BibTeX
@misc{sacilotto2026emissionaware,
title = {{Emission-aware operation of residential PV-battery systems: the role of information structure}},
author = {Anna Sacilotto and Ambrogio Maria Bernardelli and Stefano Gualandi},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2606.26648},
eprint = {2606.26648},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.26648},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Emission-aware operation of residential PV-battery systems: the role of information structure
AU - Anna Sacilotto
AU - Ambrogio Maria Bernardelli
AU - Stefano Gualandi
PY - 2026
DA - 2026/10/24
PB - arXiv
DO - 10.48550/arxiv.2606.26648
UR - https://arxiv.org/abs/2606.26648
AB - Residential batteries can reduce the carbon emissions of households with photovoltaic (PV) generation, but how much depends on how they are operated under uncertain load, PV generation, and grid carbon intensity. We cast the operation of a residential PV-battery system as a sequential decision problem and compare, under a common and explicit evaluation protocol, a family of linear policies that differ in the information available at decision time, in how they anticipate the future, and in their adaptivity: a rule-based self-consumption policy, an open-loop stochastic schedule, affine feedback policies, receding-horizon policies with deterministic, scenario-based and two-stage lookahead models, and perfect-information benchmarks. Scenarios are generated by a joint principal component model of the three uncertain quantities, which supports conditional sampling during the day. Experiments on Italian data at 15-minute resolution comprise single-day evaluations on 10000 test scenarios in three seasons and a closed-loop simulation over more than eight months of real data. When the state of charge is treated consistently across policies, the self-consumption rule is within 3.2% of the perfect-foresight emissions in closed loop, whereas open-loop schedules and receding-horizon policies that decide before measuring the current period are about 20--35% above it. With real-time measurements, scenario-based receding-horizon control consistently outperforms its deterministic counterpart and, for households that rely more on the grid, reduces emissions by about 5% with respect to the rule. The benefit of emission-aware optimization is therefore determined mainly by the information structure of the controller and by the size of the system relative to the demand.
ER -