Rough knowledge enhanced dueling deep Q-network for household integrated demand response optimization

Article Engineering, Electrical & Electronic

DRL-HEMS: Deep Reinforcement Learning Agent for Demand Response in Home Energy Management Systems Considering Customers and Operators Perspectives

Aya A. Amer et al.

Summary: This article proposes a data-driven multi-objective DRL-HEMS solution that optimizes household energy consumption, reduces electricity cost, and considers resident comfort and transformer loading condition.

IEEE TRANSACTIONS ON SMART GRID (2023)

Add to Collection

Article Energy & Fuels

Online transfer learning strategy for enhancing the scalability and deployment of deep reinforcement learning control in smart buildings

Davide Coraci et al.

Summary: In recent years, advanced control strategies based on Deep Reinforcement Learning (DRL) have been effective in optimizing the management of energy systems in buildings. However, the scalability and implementation of DRL controllers are limited by the time required for convergence. This paper proposes an Online Transfer Learning (OTL) strategy that transfers a DRL control policy from a source building to various target buildings in a simulation environment.

APPLIED ENERGY (2023)

Add to Collection

Review Computer Science, Artificial Intelligence

Reinforcement learning algorithms: A brief survey

Ashish Kumar Shakya et al.

Summary: Reinforcement Learning (RL) is a machine learning technique that learns sequential decision-making in complex problems through continuous interaction with a dynamic environment. Deep reinforcement learning has successfully solved previously considered impossible tasks, such as learning to play video games from pixel information. RL agents can surpass human performance without human intervention. This review provides an overview of RL principles, methods, and applications, serving as a reference for researchers entering the field. It covers model-free RL algorithms, deep RL algorithms using function approximation, and also touches on model-based and multi-agent RL approaches, as well as potential research directions.

EXPERT SYSTEMS WITH APPLICATIONS (2023)

Add to Collection

Article Engineering, Electrical & Electronic

Meta-Reinforcement Learning-Based Transferable Scheduling Strategy for Energy Management

Luolin Xiong et al.

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS (2023)

Add to Collection

Article Construction & Building Technology

Design of type-2 fuzzy logic controller in a smart home energy management system with a combination of renewable energy and an electric vehicle

Ali Beheshtikhoo et al.

Summary: This paper aims to design two type-2 fuzzy logic controllers in the demand-side energy management system to efficiently use renewable energy sources in a smart home. The proposed method results in a reduction in electricity costs by 71.5% and a decrease in the peak-to-average ratio by 64.6%.

JOURNAL OF BUILDING ENGINEERING (2023)

Add to Collection

Article Thermodynamics

Modeling and energy dynamic control for a ZEH via hybrid model-based deep reinforcement learning

Yanxue Li et al.

Summary: An efficient and flexible energy management strategy is crucial for energy conservation in the building sector. This study proposes a hybrid model-based reinforcement learning framework that uses short-term monitored data to optimize indoor thermal comfort and energy cost-saving performance. Simulation results demonstrate the efficiency and superiority of the proposed framework, with the D3QN agent achieving over 30% cost savings compared to measurement results.

ENERGY (2023)

Add to Collection

Article Thermodynamics

Optimization model for home energy management system of rural dwellings

Zhijia Huang et al.

Summary: This paper proposes a home energy management system method considering the working state of flexible load to address the issue of mismatch between home energy supply and demand. The research results demonstrate that the system with flexible load scheduling can reduce the electricity purchase cost and increase the utilization rate of renewable energy.

ENERGY (2023)

Add to Collection

Article Green & Sustainable Science & Technology

The usage of 10-fold cross-validation and grid search to enhance ML methods performance in solar farm power generation prediction

Seyed Matin Malakouti et al.

Summary: Accurate projections of future solar energy generation are crucial for improving the competitiveness of solar power plants and reducing dependence on fossil fuels. In this study, we analyzed databases from California spanning 2019 to 2021 and utilized 10-fold cross-validation and Grid search to enhance the performance of decision tree, light gradient boosting machine, and an extra tree in Solar Farm Power Generation Prediction.

CLEANER ENGINEERING AND TECHNOLOGY (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Supervised assisted deep reinforcement learning for emergency voltage control of power systems

Xiaoshuang Li et al.

Summary: This paper proposes a novel hybrid emergency voltage control method that combines expert experience and machine intelligence. The expert experience is extracted through a behavioral cloning model, and the deep reinforcement learning method is applied to discover and learn new knowledge autonomously. Experiments validate the effectiveness and applicability of the proposed method.

NEUROCOMPUTING (2022)

Add to Collection

Article Automation & Control Systems

Smart Train Operation Algorithms Based on Expert Knowledge and Reinforcement Learning

Kaichen Zhou et al.

Summary: This article proposes two smart train operation algorithms that integrate expert knowledge with reinforcement learning algorithms to achieve continuous action control and multi-objective optimization in subway systems, while ensuring comfort, punctuality, and safety.

IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS (2022)

Add to Collection

Review Construction & Building Technology

A taxonomy of machine learning applications for virtual power plants and home/building energy management systems

Seppo Sierla et al.

Summary: A virtual power plant is an information and communications technology system that enhances renewable power generation, aggregates Distributed Energy Resources (DERs), and monetizes them based on energy contracts or markets. It also includes functionalities such as load forecasting, market price prediction, and asset management related to the distributed energy resources. Machine learning is used to optimize, forecast, and classify these functionalities.

AUTOMATION IN CONSTRUCTION (2022)

Add to Collection

Article Construction & Building Technology

A novel forecasting based scheduling method for household energy management system based on deep reinforcement learning

Mifeng Ren et al.

Summary: The article proposes a new forecasting based optimization method for real-time scheduling, adopts a deep reinforcement learning approach for optimal HEMS dispatching, and introduces a new neural network for predicting outdoor temperature. Experimental results demonstrate that the method effectively reduces user costs while maintaining user satisfaction.

SUSTAINABLE CITIES AND SOCIETY (2022)

Add to Collection

Article Energy & Fuels

Deep reinforcement learning-based joint load scheduling for household multi-energy system

Liyuan Zhao et al.

Summary: This paper proposes a joint load scheduling strategy for household multi-energy system (HMES) aiming at minimizing residents' energy cost while maintaining thermal comfort. By formulating the HMES scheduling problem as a Markov decision process and proposing a deep reinforcement learning-based approach, it successfully tackles the uncertainties in energy price, photovoltaic generation, outdoor temperature, and residents' hot water demand.

APPLIED ENERGY (2022)

Add to Collection

Article Energy & Fuels

Predicting residential electricity consumption patterns based on smart meter and household data: A case study from the Republic of Ireland

Zhifeng Guo et al.

Summary: In this study, machine learning algorithms were used to investigate residential electricity consumption for households in the Republic of Ireland. The findings revealed that temperature, day of week, and month of year have a significant impact on consumption. The prevalence of different load profiles varied between weekdays and weekends as well as seasonally. Key factors associated with annual load profiles include household makeup, size, and occupation of the primary income earner. The study also discussed the policy and management implications of the findings, as well as proposed future research directions.

UTILITIES POLICY (2022)

Add to Collection

Article Construction & Building Technology

Community decision-makers' choice of multi-objective scheduling strategy for integrated energy considering multiple uncertainties and demand response

Gao Fangjie et al.

Summary: A multi-objective optimal scheduling model is proposed to address the impact of uncertainty on scheduling strategy in community integrated energy systems. The model incorporates a source-load uncertainty model and a comprehensive demand response model, and utilizes entropy weight method and Muirhead mean operator to determine the best strategy. The model has been validated using a multi-scene case and has shown effectiveness in reducing resident cost, carbon treatment amount, and improving energy supplier profit and renewable energy utilization rate.

SUSTAINABLE CITIES AND SOCIETY (2022)

Add to Collection

Article Construction & Building Technology

Optimal control method of HVAC based on multi-agent deep reinforcement learning

Qiming Fu et al.

Summary: This paper proposes a Multi-Agent deep reinforcement learning method for building Cooling Water System Control (MA-CWSC) to optimize the load distribution, cooling tower fan frequency, and cooling water pump frequency of different types of chillers. Experimental results show that the MA-CWSC method achieves significant energy saving performance and has a faster learning rate compared to other control methods.

ENERGY AND BUILDINGS (2022)

Add to Collection

Article Engineering, Electrical & Electronic

A multi-objective multi-agent deep reinforcement learning approach to residential appliance scheduling

Junlin Lu et al.

Summary: This research introduces a tunable deep reinforcement learning algorithm for demand-side management of household appliances. Compared to rule-based approaches, the tunable Deep Q-Network algorithm shows significant improvements in cost, peak power, and punctuality objectives, without the need for retraining, by generating scheduling policies for different preferences.

IET SMART GRID (2022)

Add to Collection

Article Construction & Building Technology

A Multi-objective Optimization Framework for Dynamic Planning of Energy Hub Considering Integrated Demand Response Program

Amir Ahmarinejad

Summary: This paper presents a multi-objective optimization framework for long-term planning of energy hub, considering equipment degradation and integrated demand response programs. Simulation results show that the integrated demand response program has reduced total costs, losses, and emissions, while also leading to an increase in total costs.

SUSTAINABLE CITIES AND SOCIETY (2021)

Add to Collection

Article Energy & Fuels

Robust Optimization of the Flexibility-constrained Energy Management Problem for a Smart Home with Rooftop Photovoltaic and an Energy Storage

S. Bahramara

Summary: The paper proposes a flexibility-constrained energy management model for smart homes equipped with rooftop PVs and electrical energy storage to limit the power trading ramp rate with the main grid, increase system flexibility, and model uncertain energy market prices through robust optimization. The results demonstrate the model's ability to minimize operating costs for smart homes within the constraint, while a sensitivity analysis is conducted to investigate the impact of flexibility constraints and robustness parameters on outcomes.

JOURNAL OF ENERGY STORAGE (2021)

Add to Collection

Article Energy & Fuels