HEИндивидуална стипендия2023–2026

SmartHEM · Integrating reinforcement learning and predictive control for smart home energy management

„Хоризонт Европа“ — Действия „Мария Склодовска-Кюри“

Период: 2023-10-10 → 2026-01-09
Финансиране от ЕС: 222 728 €
Участници: 1
Схема: HORIZON-TMA-MSCA-PF-EF

Линиите свързват координатора с партньорите.

Накратко на български

Интелигентното управление на домашната енергия изследва оптимизирането на батериите за електромобили и бързото разряждане на стари литиево-йонни клетки за рециклиране. Това помага за намаляване на вредните емисии и разхода на енергия в сградите.

Този кратък обзор е генериран от изкуствен интелект

Кратко обяснение, генерирано от езиков модел по текста на CORDIS. Оригиналът е по-долу.

Резултати накратко

Integrating reinforcement learning and predictive control for smart home energy management

The world is currently witnessing a global urgency to revolutionise the energy sector and make it completely renewable. Buildings are responsible for approximately 40% of global energy consumption and 33% of greenhouse gas (GHG) emissions. To address this issue, improving building energy efficiency plays a key role in achieving the ambitious goal of carbon neutrality outlined in the European Green Deal. Specifically, smart buildings interacting with renewable energy sources, such as solar panels, power grids, and electric vehicles (EVs), are a prevailing concept to reduce both emissions and energy consumption. However, managing these systems is complicated due to the variable nature of renewable sources and stochastic user behaviours. Consequently, this project aims to address these challenges by developing advanced control and learning frameworks for battery systems and home energy management. The project provides in-depth solutions for three critical aspects of the energy ecosystem: the efficient recycling of retired batteries, the fast charging of operational EV batteries, and the intelligent management of home energy. First, regarding the end-of-life management of batteries, the project focuses on the pre-treatment phase of battery recycling. A robust model predictive control (MPC) framework has been developed for the fast discharging of retired lithium-ion batteries. The objective is to minimise the discharging time to improve recycling efficiency while strictly satisfying safety constraints, such as temperature limits, despite the uncertainties inherent in retired cells. Second, to facilitate the adoption of EVs, the project addresses the conflict between charging speed and battery longevity. A reinforcement learning (RL) based strategy has been designed for the fast charging of batteries. This approach considers the battery state of health (SoH) and the risk of lithium plating. By optimising the charging current in a health-aware manner, the algorithm aims to achieve safe and fast charging across the entire battery lifespan. Third, at the residential level, the project develops a smart home energy management system (HEMS) using deep reinforcement learning. This system manages the power flow between rooftop photovoltaic panels, stationary home batteries, and EVs under uncertain electricity prices and user driving schedules. The developed algorithm aims to minimise the total household electricity cost and the battery degradation cost simultaneously, while strictly maintaining occupant thermal comfort to ensure an economic and sustainable operation of the home energy network.

Текст от CORDIS, на английски · Данни: CORDIS, © Европейски съюз

Цел на проекта

To achieve carbon neutrality and reduce the dependence on fossil energy, solar panels, heat pumps and electric vehicles (EVs) are becoming commonplace in European homes. The abundant solar power, geothermal energy, as well as environmentally friendly electric vehicles, reduce the usage of fossil-based energy, while posing challenges to the home energy management system at the same time due to the intermittent feature of renewable energy and potential battery degradation from EVs. In this proposal, we propose a smart home management system that optimises the entire home energy cost while considering a set of constraints related to battery safety and reliability, driver/household demand and actuator. As an MSCA-PF fellow, Dr. Meng Yuan will receive crucial training at the Chalmers University of Technology and engage in battery power control and home energy optimisation works which span the areas of control theory and electrical engineering. Interaction between different areas will spur novel methodologies and fruitful outcomes including 1) an adaptive and robust power controller for EV battery systems, 2) a learning-based energy management algorithm that minimises the total energy cost while satisfying the household power supply, and 3) a safe learning-based control framework for home energy system management. The foreseeable results of the project will include a health-aware optimal control for vehicle battery systems that enables vehicle-to-home technology and prolongs the battery's lifetime; a new interdisciplinary energy management system combining machine learning and control algorithms to revolutionise energy sectors.

Оригинален текст от CORDIS (на английски).

Участници

CHALMERS TEKNISKA HOGSKOLA AB · GoteborgКоординаторШвеция