Tackling Climate Change with Machine Learning

ACM Computing Surveys, Journal Year: 2022, Volume and Issue: 55(2), P. 1 - 96

Published: Feb. 7, 2022

Climate change is one of the greatest challenges facing humanity, and we, as machine learning (ML) experts, may wonder how we can help. Here describe ML be a powerful tool in reducing greenhouse gas emissions helping society adapt to changing climate. From smart grids disaster management, identify high impact problems where existing gaps filled by ML, collaboration with other fields. Our recommendations encompass exciting research questions well promising business opportunities. We call on community join global effort against climate change.

Language: Английский

---

Artificial intelligence and machine learning approaches to energy demand-side response: A systematic review

Renewable and Sustainable Energy Reviews, Journal Year: 2020, Volume and Issue: 130, P. 109899 - 109899

Published: June 10, 2020

Recent years have seen an increasing interest in Demand Response (DR) as a means to provide flexibility, and hence improve the reliability of energy systems cost-effective way. Yet, high complexity tasks associated with DR, combined their use large-scale data frequent need for near real-time decisions, that Artificial Intelligence (AI) Machine Learning (ML) — branch AI recently emerged key technologies enabling demand-side response. methods can be used tackle various challenges, ranging from selecting optimal set consumers respond, learning attributes preferences, dynamic pricing, scheduling control devices, how incentivise participants DR schemes reward them fair economically efficient This work provides overview utilised applications, based on systematic review over 160 papers, 40 companies commercial initiatives, 21 projects. The papers are classified regards both AI/ML algorithm(s) application area DR. Next, initiatives presented (including start-ups established companies) innovation projects, where been paper concludes discussion advantages potential limitations reviewed techniques different tasks, outlines directions future research this fast-growing area.

Language: Английский

---

Peer-to-peer energy systems for connected communities: A review of recent advances and emerging challenges

Applied Energy, Journal Year: 2020, Volume and Issue: 282, P. 116131 - 116131

Published: Nov. 10, 2020

Language: Английский

---

Reinforcement learning for building controls: The opportunities and challenges

Applied Energy, Journal Year: 2020, Volume and Issue: 269, P. 115036 - 115036

Published: May 12, 2020

Language: Английский

---

Data-driven probabilistic machine learning in sustainable smart energy/smart energy systems: Key developments, challenges, and future research opportunities in the context of smart grid paradigm

Renewable and Sustainable Energy Reviews, Journal Year: 2022, Volume and Issue: 160, P. 112128 - 112128

Published: March 5, 2022

Language: Английский

---

Deep Reinforcement Learning for Smart Home Energy Management

IEEE Internet of Things Journal, Journal Year: 2019, Volume and Issue: 7(4), P. 2751 - 2762

Published: Dec. 3, 2019

We investigate an energy cost minimization problem for a smart home in the absence of building thermal dynamics model with consideration comfortable temperature range. Due to existence uncertainty, parameter uncertainty (e.g., renewable generation output, nonshiftable power demand, outdoor temperature, and electricity price), temporally coupled operational constraints, it is very challenging design optimal management algorithm scheduling heating, ventilation, air conditioning systems storage home. To address challenge, we first formulate above as Markov decision process, then propose based on deep deterministic policy gradients. It worth mentioning that proposed does not require prior knowledge uncertain parameters model. The simulation results real-world traces demonstrate effectiveness robustness algorithm.

Language: Английский

---

Adaptive Power System Emergency Control Using Deep Reinforcement Learning

IEEE Transactions on Smart Grid, Journal Year: 2019, Volume and Issue: 11(2), P. 1171 - 1182

Published: Aug. 6, 2019

Power system emergency control is generally regarded as the last safety net for grid security and resiliency. Existing schemes are usually designed offline based on either conceived "worst" case scenario or a few typical operation scenarios. These facing significant adaptiveness robustness issues increasing uncertainties variations occur in modern electrical grids. To address these challenges, this paper developed novel adaptive using deep reinforcement learning (DRL) by leveraging high-dimensional feature extraction non-linear generalization capabilities of DRL complex power systems. Furthermore, an open-source platform named Reinforcement Learning Grid Control (RLGC) has been first time to assist development benchmarking algorithms control. Details DRL-based generator dynamic braking under-voltage load shedding presented. Robustness method different simulation scenarios, model parameter uncertainty noise observations investigated. Extensive studies performed both two-area, four-machine IEEE 39-bus have demonstrated excellent performance proposed schemes.

Language: Английский

---

A Multi-Agent Reinforcement Learning-Based Data-Driven Method for Home Energy Management

IEEE Transactions on Smart Grid, Journal Year: 2020, Volume and Issue: 11(4), P. 3201 - 3211

Published: Feb. 4, 2020

This paper proposes a novel framework for home energy management (HEM) based on reinforcement learning in achieving efficient home-based demand response (DR). The concerned hour-ahead consumption scheduling problem is duly formulated as finite Markov decision process (FMDP) with discrete time steps. To tackle this problem, data-driven method neural network (NN) and Q-learning algorithm developed, which achieves superior performance cost-effective schedules HEM system. Specifically, real data of electricity price solar photovoltaic (PV) generation are timely processed uncertainty prediction by extreme machine (ELM) the rolling windows. decisions household appliances electric vehicles (EVs) can be subsequently obtained through newly developed framework, objective dual, i.e., to minimize bill well DR induced dissatisfaction. Simulations performed residential house level multiple appliances, an EV several PV panels. test results demonstrate effectiveness proposed framework.

Language: Английский

---

Machine learning driven smart electric power systems: Current trends and new perspectives

Applied Energy, Journal Year: 2020, Volume and Issue: 272, P. 115237 - 115237

Published: June 2, 2020

Language: Английский

---

Reinforcement Learning and Its Applications in Modern Power and Energy Systems: A Review

Journal of Modern Power Systems and Clean Energy, Journal Year: 2020, Volume and Issue: 8(6), P. 1029 - 1042

Published: Jan. 1, 2020

With the growing integration of distributed energy resources (DERs), flexible loads, and other emerging technologies, there are increasing complexities uncertainties for modern power systems. This brings great challenges to operation control. Besides, with deployment advanced sensor smart meters, a large number data generated, which opportunities novel data-driven methods deal complicated control issues. Among them, reinforcement learning (RL) is one most widely promoted optimization problems. paper provides comprehensive literature review RL in terms basic ideas, various types algorithms, their applications The further works also discussed.

Language: Английский

---