Enhancing Industrial Decarbonization Through Standardized Data Interoperability and PPO-Based DRL Multi-Energy Scheduling

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Demonstrates how standardized data schemas enable DRL-based multi-energy optimization with robust empirical gains in cost and carbon.

🏢実務担当者:Provides a concrete data pipeline and scheduling model to reduce energy costs and product carbon footprint in battery plants.

🏛政策担当者:Highlights data interoperability standards as a policy lever for industrial decarbonization and emerging battery regulation compliance.

Reducing the carbon footprint of lithium-ion battery manufacturing is a critical priority for the electric-vehicle transition, yet practical industrial energy optimization is often hindered by data silos and proprietary formats that limit interoperability and the deployment of learning-based scheduling and optimization. This paper presents a standardized data interoperability and PPO-based DRL multi-energy scheduling framework for battery manufacturing, built on the ANSI/ASHRAE 201 schema. The framework treats electricity and heat as a coupled system and employs deep reinforcement learning (DRL) with Proximal Policy Optimization (PPO) to jointly optimize energy cost and grid carbon intensity. To enable secure data openness, a semantic pseudonymization layer is introduced to protect sensitive asset identifiers while preserving statistical and physical characteristics required for high-fidelity learning. The framework is evaluated using real-world industrial time-series datasets (14 days) scaled to a 10MW-class production line while preserving temporal and statistical properties. The proposed data pipeline achieves an 80.6% reduction in Time-to-Model (TTM) and the PPO-based DRL scheduler reduces annual energy cost by 16.1%. The electrode pilot operation yields an observed localized carbon reduction of 0.71% under the evaluated operating conditions. Separately, based on a scenario-based Gigafactory-scale projection (30 GWh-class) that scales the validated scheduling logic under stated assumptions (including formation-stage energy recovery and HVAC/utility optimization potentials), the integrated carbon mitigation potential is projected to reach 9.80%, corresponding to 14,176.3 tCO2per year. In addition, the Product Carbon Footprint (PCF) improves by 763.2 g CO2/ton, supporting compliance with emerging regulations such as the EU Battery Passport. These results highlight standardized data openness as a practical foundation for scalable, carbon-aware energy orchestration in energy-intensive manufacturing.

• openalex https://doi.org/10.1109/access.2026.3660727first seen 2026-08-02 17:21:33