Data-driven Energy Management Research Articles

A data-driven solution for intelligent power allocation of connected hybrid electric vehicles inspired by offline deep reinforcement learning in V2X scenario

The proper power allocation between multiple energy sources is crucial for hybrid electric vehicles to guarantee energy economy. As a data-driven technique, offline deep reinforcement learning (DRL) solely exploits existing data to train energy management strategy (EMS), which becomes a promising solution for intelligent power allocation. However, current offline DRL-based strategies put high demands on the quality of datasets, and it is difficult to obtain numerous high-quality samples in practice. Thus, a bootstrapping error accumulation reduction (BEAR)-based strategy is proposed to enhance the energy-saving performance with different kinds of datasets. After that, based on the advanced V2X technology, a data-driven energy management updating framework is proposed to improve both fuel economy and adaptability of EMS via multi-updating. Specifically, the framework deploys multiple V2X-based buses to collect real-time information, and updates the strategy periodically making full use of offline data. The results show that the proposed BEAR-based EMS performs better than state-of-the-art offline EMSs in terms of fuel economy, especially realizing an improvement of 2.25% when training with mixed datasets. It is also validated that the offline EMS with the updating mechanism can reduce energy costs step by step under two different kinds of initial datasets.

Synergy in Action: Integrating Environmental Monitoring, Energy Efficiency, and IoT for Safer Shared Buildings

Shared public buildings have become centers of innovation, integrating advanced technologies to meet evolving societal needs. With a heightened emphasis on occupants’ health and well-being, these buildings serve as hubs for technological convergence, facilitating seamless connectivity and intelligent data analysis and management. Within this context, environmental monitoring emerges as a foundational element, pivotal to all aspects of building management. This article provides findings from the nationally funded RE-START project, which focuses on shared public buildings, with special regard to educational and medical facilities. The project explores enhanced indoor air quality monitoring, focusing on CO2 concentration that is directly correlated with occupancy, as a fundamental element for developing health and safety protocols, energy efficiency strategies, the integration of smart building technologies, and data-driven energy management. The intersection of environmental monitoring, energy efficiency, security, and IoT technologies in in indoor spaces is relevant. The outcomes of the study reveal the delicate nature of all the involved components, which need to be carefully developed in an integrated manner.

Battery-aware federated learning: Challenges and solutions

Smartphone battery life is a pivotal factor in consumers' purchasing decisions. Recent years have witnessed a surge in studies focusing on smartphone energy management, with data-driven energy management systems offering solutions to prolong battery life. Federated Learning (FL) emerges as a promising distributed learning algorithm, enabling wireless devices to upload locally trained models, fostering collaborative learning without exposing sensitive data. This paper explores the FL process, particularly the Federated Averaging (FedAvg) approach, which excels in scenarios with homogeneous data. In the era of burgeoning data generation, traditional cloud computing systems face limitations, driving the adoption of Edge Computing (EC), which processes data closer to its source, enhancing response times. To make FL efficient for e-commerce, resource constraints must be addressed. This involves techniques like local updates and model compression, which reduce communication overhead. However, FL brings challenges related to data distribution heterogeneity and privacy concerns. Solutions like differential privacy, encryption, and access control are discussed. In conclusion, this paper presents an overview of smartphone battery life, data-driven energy management, and the potential of FL, emphasizing its relevance in the age of EC. By addressing resource limitations and privacy issues, FL holds promise for efficient data processing.

DATA-DRIVEN ENERGY MANAGEMENT: REVIEW OF PRACTICES IN CANADA, USA, AND AFRICA

This research explores data-driven energy management practices in Canada, the USA, and Africa, offering a comparative analysis of successes, challenges, and future implications. Integrating smart grid technologies and supportive regulatory frameworks in North America has driven efficiency gains and sustainability. Conversely, Africa faces unique challenges, including infrastructural limitations, yet showcases localized successes in enhancing energy access. Lessons learned emphasize stakeholder collaboration and adaptable regulatory frameworks, providing valuable insights for global energy strategies. The study identifies gaps in technological infrastructure and recommends collaborative, context-specific solutions. As the global community moves towards sustainable energy futures, these findings contribute to a nuanced understanding, guiding policymakers, industry stakeholders, and researchers in shaping resilient and inclusive energy systems.
 Keywords: Data-Driven Energy Management, Comparative Analysis, Sustainability, Global Energy Landscape.

Data-Driven Energy Management of an Electric Vehicle Charging Station Using Deep Reinforcement Learning

Data-Driven Energy Management of an Electric Vehicle Charging Station Using Deep Reinforcement Learning

A data-driven energy management strategy for plug-in hybrid electric buses considering vehicle mass uncertainty

Conventional energy management strategies (EMSs) for hybrid electric vehicles are devised assuming the vehicle mass remains constant under dynamic driving conditions. However, the EMSs cannot adapt to different load conditions due to the dynamic change of vehicle mass. Investigating the characteristics of vehicle mass change plays significant roles in energy optimization, thereby improving overall efficiency and mobility. To this end, we propose an adaptive EMS for a plug-in hybrid electric bus (PHEB) based on artificial neutral network-Pontryagin minimum principle (ANN-PMP) by considering mass distribution characteristics. Firstly, the skew-normal distribution characteristics of PHEB mass are analyzed, and the distribution characteristic spectrum of vehicle mass is obtained based on the Monte Carlo method. Secondly, the influence of mass uncertainty on the PMP is analyzed, and then the ANN-PMP is devised by updating a dynamic co-state with ANN. Finally, an enhanced ANN-PMP (so-called ANN-PMP-CS) is proposed by combining the ANN-PMP strategy obtained by mass distribution training and CS (charge sustaining strategy) to include the no-load or full-load cases. The simulation results demonstrate that the ANN-PMP can adapt to mass change while ensuring that the final state-of-charge (SOC) convergence to the target value. We also observe that the fuel economy of ANN-PMP-CS is similar to that of the widely-used dynamic programming (DP) strategy. Compared with the charge depletion-charge sustaining (CD-CS) strategy, the fuel economy can be improved by about 46.93 % on average.

Framework for Project Sustainability for Power Installations Using Business Intelligence Approach: A Systematic Literature Review

The growing concerns about environmental sustainability and energy conservation have led to increased interest in optimizing electricity consumption and billing processes in various projects. This research article presents a comprehensive study on the development and application of Business Intelligence (BI) frameworks for enhancing project sustainability through data-driven energy management. Through the integration of BI tools and techniques, this research investigates the analysis of electricity consumption patterns, billing accuracy, and cost-effectiveness in diverse project contexts. The article emphasizes the significance of data preprocessing, statistical analysis, and predictive modelling in uncovering valuable insights to support informed decision-making. Additionally, the review examines the concept of project sustaibility, emphasizing its significance in achieving desired outcomes, meeting stakeholder expectations, and ensuring the project’s viability in an ever-changing environment. Traditional project management approaches often fail to adequately address sustainability concerns, leading to project failures or limited long-term impact. Hence, the review highlights the growing importance of leveraging BI-driven frameworks to enhance project sustainability in various sectors, including in Information and Communication Technology (ICT) domain. The Systematic literature review (SLR) method was used involving the scooping of 230 articles from over 8 global academic databases. With the use of exclusion criteria, only 61 articles were used in the study. The analysis of the articles shows that 57% were journal articles, 39% were conference proceedings, 2% were thesis/dissertations and 2% were generic. Within the scope of this literature review, key terms and keywords were identified to provide insights into the development of a novel BI-driven framework for project sustainability. Consequently, future research directions are identified to further explore the integration of renewable energy sources, AI and machine learning applications, and behaviour-based energy management strategies within BI frameworks for sustainable project outcomes. This review lays the foundation for future research endeavours in developing innovative BI-driven frameworks that foster sustainable practices and contribute to a greener and more resilient future across diverse industries and projects.

Data-driven energy management system for flexible operation of hydrogen/ammonia-based energy hub: A deep reinforcement learning approach

In the context of carbon neutrality, multi-energy systems are being designed to enhance the integration of renewable energy, and the deployment of large-scale energy storage solutions is vital to ensure a flexible and cost-effective system operation. Hydrogen and ammonia are two promising energy carriers that can serve as storage media, feedstock, and fuels. This study designed an energy hub that integrates multiple energy resources and energy storage methods. Two pathways, power-to-gas-to-power (P2X2P) and biomass-to-gas-to-power (B2X2P), which use hydrogen and ammonia as energy carriers, are proposed for the energy hub. Considering the flexibility and profitability of the energy hub, scheduling of the energy hub was controlled using model-based mixed-integer linear programming and model-free deep reinforcement learning methods. A modified double-deep Q-network framework was trained on a monthly dataset and executed on a yearly dataset. The results indicate that the B2X2P pathway outperforms the P2X2P pathway in terms of profitability, while the P2X2P pathway has greater operational flexibility. Because the optimal biomass-to-ammonia-to-power has a 550.76 MUSD net present value and a five-year discounted payback period, ammonia is best suited for mass production and storage. Moreover, the modified double deep Q network framework demonstrates superior generality for solving similar optimal scheduling problems.

A novel data-driven energy management strategy for fuel cell hybrid electric bus based on improved twin delayed deep deterministic policy gradient algorithm

With the prosperity of artificial intelligence in recent years, energy management strategies (EMSs) based on deep reinforcement learning (DRL) have become the mainstream approach to ensure efficient energy distribution of hybrid electric vehicles (HEVs). Moreover, in terms of developing sustainable urban public transportation systems, zero-emission fuel cell hybrid electric buses (FCHEBs) are more promising than hybrid electric buses (HEBs). Given that, this article proposes a novel data-driven EMS based on DRL to enhance the training efficiency of the proposed EMS while improving the fuel economy of an FCHEB. In this article, to optimize the power allocation of the FCHEB, an improved twin delayed deep deterministic policy gradient (TD3) algorithm combined with prioritized experience replay is innovatively formulated, and then a promising EMS based on it is proposed. Furthermore, to enhance the adaptability and improve the training efficiency of the proposed EMS, a stochastic training environment is established with massive real-world velocity data, and a pre-training method using the global optimal experience is designed. Simulation results show that the proposed EMS improves fuel economy by 5.87% compared with the EMS based on original TD3, reaching 97.15% of the global optimal dynamic programming method.

A Data-Driven Energy Management Strategy Based on Deep Reinforcement Learning for Microgrid Systems

A Data-Driven Energy Management Strategy Based on Deep Reinforcement Learning for Microgrid Systems

A data-driven energy management method for parallel PHEVs based on action dependent heuristic dynamic programming (ADHDP) model

Due to the strong self-learning ability and adaptability of neural network, adaptive dynamic programming (ADP) method is regarded as an effective method to improve the vehicle economy in the case of uncertain driving conditions. With the look-ahead information, a data-driven energy management method is proposed based on action dependent heuristic dynamic programming (ADHDP) algorithm. Firstly, based on statistical analysis of DP behavior, a reference SOC trajectory is generated to limit the electricity consumption, which is adjusted with the dynamically updated driving information. Then, a data-driven energy management control method is developed for HEV architectures, including information acquisition module, shift scheduling module, energy distribution module. The gear shift command is designed by enumeration and the power distribution is performed by multiple ADHDP models, the core of which is to determine the utility function, action network (ANN), critic network (CNN) and training process. To ensure practical driveability, the restrictions on gear shifting and engine starting-stopping are taken as additional conditions in ADHDP optimizing process. Finally, two case studies under different driving scenarios are given. Simulation results demonstrate that the proposed method gains a good performance in both optimality approximation and adaptivity to uncertain driving conditions.

Data-driven energy management in residential areas leveraging demand response

A distributed data-driven coordinated design is proposed to achieve efficient energy management of a residential grid, where controllable distributed resources such as electric vehicles (EVs) and thermostatically controlled loads (TCLs) are adjusted by balancing the end-use electricity cost, charging preference, and thermal comfort. The motivation for the control pattern is to minimize the total system cost by directly utilizing the measured input–output data instead of intractable model identification and state estimation. Firstly, we formulate a data based optimization problem with persistently exciting data sets and show the equivalence with the model-based problem. To protect the privacy of each consumer, we design a distributed pattern by the gradient of the augmented Lagrangian such that TCLs and EVs implement demand response individually. Moreover, the proposed algorithm is enhanced by a receding control scheme to tackle the uncertainties in the predictions. Standard test systems are used to illustrate the proposed design and demonstrate its effectiveness and benefits in the residential community.

Battery health-aware and naturalistic data-driven energy management for hybrid electric bus based on TD3 deep reinforcement learning algorithm

Energy management is critical to reduce energy consumption and extend the service life of hybrid power systems. This article proposes an energy management strategy based on deep reinforcement learning with awareness of battery health for an urban power-split hybrid electric bus. In this article, a specific driving cycle of the test bus route is constructed through a naturalistic data-driven method to evaluate the practical operating costs of the hybrid electric bus accurately. Furthermore, an energy management strategy based on twin delayed deep deterministic policy gradient algorithm considering battery health is innovatively designed to minimize the total operating cost with a tradeoff between fuel consumption and battery degradation. Finally, the superiority of the proposed strategy over other state-of-the-art deep reinforcement learning-based strategies including deep deterministic policy gradient and double deep Q-learning is validated. Simulation results show that the constructed driving cycle can effectively reflect the real traffic conditions of the test bus route, and the proposed strategy can reduce the total operating cost while extending the battery life efficiently. This article makes contribution to the reliable evaluation of the practical operating costs and the extension of the battery life for urban hybrid electric buses through deep reinforcement learning methods.

Novel data-driven energy management of a hybrid photovoltaic-reverse osmosis desalination system using deep reinforcement learning

This paper proposes a novel deep reinforcement learning-accelerated energy management system for a hybrid grid-connected photovoltaic-reverse osmosis-pressure retarded osmosis desalination plant. The energy management problem is formulated as a partially observable Markov decision process by using historical photovoltaic (PV) power data in order to cope with uncertainties related to the generation of solar power and provide more information regarding the true state of the system. The soft actor-critic (SAC) algorithm is employed as the core of the energy management system to maximize water production rate and contaminant removal efficiency while minimizing the supplied power from the external grid. We introduce 1-dimensional convolutional neural networks (1-D CNNs) to the actor, critic, and value function networks of the SAC algorithm to address the partial observability dilemma involved in PV-powered energy systems, extract essential features from the PV power time series, and achieve immensely improved performance ultimately. Furthermore, it is assumed that the proposed CNN-SAC algorithm does not have access to the current output power data of the PV system. The development of more practical energy management systems necessitates this assumption, and we demonstrate that the proposed method is capable of forecasting the current PV power data. The superiority of the CNN-SAC model is verified by comparing its learning performance and simulation results with those of four state-of-the-art deep reinforcement learning algorithms: Deep deterministic policy gradient (DDPG), proximal policy optimization (PPO), twin delayed DDPG (TD3), and vanilla SAC. The results show that the CNN-SAC model outperforms the benchmark methods in terms of effective solar energy exploitation and power scheduling, manifesting the necessity of exploiting historical PV power data and 1-D CNNs. Moreover, the CNN-SAC algorithm is benchmarked against a powerful energy management system we developed in our previous investigation by studying three scenarios, and it is demonstrated that considerable improvement in energy efficiency can be obtained without using any solar power generation forecasting algorithm. By conducting ablation studies, the critical contribution of the introduced 1-D CNN is demonstrated, and we highlight the significance of providing historical PV power data for substantial performance enhancement. The average and standard deviation of evaluation scores obtained during the last stages of training reveal that the 1-D CNN significantly improves the final performance and stability of the SAC algorithm. These results demonstrate that the modifications we detail in our investigation render deep reinforcement learning algorithms extremely powerful for the energy management of PV-powered microgrids, including PV-driven reverse osmosis desalination plants.

Data-driven energy management of isolated power systems under rapidly varying operating conditions

We propose an energy management algorithm for isolated industrial power systems that integrate uncertain renewable generation and energy storage. The proposed strategy is designed to ensure sustainable and cost-effective operations by managing the energy flows in the grid, and is structured so to cope with: (1) high levels of renewable power penetration, and (2) load profiles characterized by non-smooth patterns and irregular events (i.e., events such as those occurring from connections/disconnections of large scale equipment, or from large wind speed ramps). The proposed algorithm leverages a stochastic economic model predictive control (MPC) scheme capable of dealing simultaneously with the dispatch and scheduling of the local generation units. More precisely, the scheme embeds a mixed-integer linear programming (MILP) optimal control policy formulation together with a stochastic programming approach. Moreover, the optimization problem accounts for multiple techno-economical objectives, such as minimization of operational costs, battery degradation, and non-utilized energy. We test the algorithm on a case study of an isolated offshore Oil & Gas platform producing energy onsite with conventional gas turbines and a local wind farm, while integrating a battery energy storage system. The results show that the proposed approach can issue ensemble predictions that successfully capture the potential irregular variations just by using recent past information of the associated random variable, even when no particular sudden events are anticipated in the near-future (i.e., step changes/trend reversals). In this way, the approach provides useful future information for the optimal management of the grid. This effect is numerically quantified via simulations that compare the performance of the proposed stochastic optimization approach against its deterministic MPC version in several realistic operating conditions. The empirical results suggest that the stochastic version leads to better scheduling of the conventional generators, with up to 12.86% reductions of the operating cost, 2.56% reduction in fuel consumption and emissions, and 35.29% reduction in status transitions (on/off) of the gas turbines, while keeping dumped energy and battery degradation as low as possible.

A data-driven energy management strategy based on performance prediction for cascade refrigeration systems

The cascade refrigeration systems become the preferred choice in the field of cold storage due to its excellent performance under the condition of low evaporation temperature. Hence, a set of accurate performance prediction and energy management framework is essential for energy conservation and emission reduction. In this paper, a library of refrigeration components is established, based on which, two methods taking the data-driven model and the knowledge-driven model into account are presented to obtain optimal intermediate pressure in the cascade-condenser from the perspective of power and COP prediction respectively. Then, the effectiveness of the proposed optimal energy management strategy is demonstrated taking an actual CO2/NH3 cascade refrigeration system as a case study. Ten parameters are used to train GA-LSSVM model and 13014 on-site testing data points are randomly divided into the training set and the testing set. Two empirical formulas are used to calculate the isentropic efficiency of the NH3 and the CO2 compressors. A thermodynamic model is used as the knowledge-driven model to connect the calculation of the high-temperature stage with that of the low-temperature stage. The results show that these two methods are accurate enough for power and COP prediction. Based on the proposed methods, a set of data (1440 data points) from a typical workday is used to show the improvement by applying the optimal energy management strategy and the results show that the power consumption can be reduced by 10.52%.

Applying machine learning techniques to implement the technical requirements of energy management systems in accordance with ISO 50001:2018, an industrial case study

ABSTRACT This article developed an innovative data-driven framework utilizing machine learning models to implement effective energy management systems in the energy-intensive industries to comply with ISO 50001:2018. By implementing this three-level framework, the following standard’s requirements are fulfilled: (a) indicating the variables that significantly affect energy performance, (b) energy performance indicator and energy baseline development, (c) analysis of the current energy performance, (d) energy conservation opportunity quantification, (e) determining the energy target, and (f) defining the energy action plan. The proposed data-driven energy management approach was implemented in an industrial-scale ethane decarbonization unit in South Pars Gas Complex in Iran. The dataset consists of 52,560 records of unit states sampled throughout 2020 as a baseline period. A high-accuracy model with 99% R-squared is developed by applying the Gradient Boosting Machines algorithm to predict the unit’s energy consumption as an energy baseline. As a critical finding, the offline analysis reveals a great potential to save energy by 10.31% or 63,119 metric tons of yearly steam consumption decreasing and a reduction of 7,027 tons of carbon dioxide emissions per year. By adjusting the unit working modes to the best practices, as an action plan, the intended outcome would be achievable.

Data-driven energy management and velocity prediction for four-wheel-independent-driving electric vehicles

This paper proposes an online energy management and optimization method for four-wheel-independent-driving electric vehicles via stochastic model predictive control (SMPC), where velocity is predicted as the foundation to ensure feasibility and efficiency. By utilizing operating data of real-world electric vehicles from a big data platform, a data-driven Markov chain method is adopted to achieve vehicle velocity prediction in an accurate and reliable way. On top of the proposed method, real-time updates of the sample space and online substitution of the velocity-acceleration (V-A) state space can be realized, which mitigates problems of prediction interruption resulting from deficiency of sample state. Simulation results based on a constructed Hardware-in-Loop system indicate effectiveness of velocity prediction with root-mean-square error under 1.3 km/h. In the perspective of the energy conservation, the SMPC method can decrease energy consumption by 7.92% compared with traditional Rule-based methods, which is close to the optimization result of a conventional dynamic programming method. Further simulation and test results demonstrate that the proposed data-driven method is capable of realizing online accurate velocity prediction and energy management for real-world vehicles. • An energy management optimization method for four-wheel-independent-driving vehicles. • A data-driven Markov chain with is adopted for vehicle velocity prediction. • The stochastic model prediction control method is employed for torque allocation. • The proposed energy optimization method is verified through the Hardware-in-Loop test.

建物のデータ駆動型エネルギーマネジメントにおけるパターン分析による不具合検知に関する研究(第1報)照明消費エネルギーに対する不具合検知の検討

建物のデータ駆動型エネルギーマネジメントにおけるパターン分析による不具合検知に関する研究(第1報)照明消費エネルギーに対する不具合検知の検討

Data-Driven Energy Management System With Gaussian Process Forecasting and MPC for Interconnected Microgrids

Interest in predicting and optimising microgrid operation with a high proportion of variable renewable energy generation is growing. In this paper, we study and experimentally analyse the performance of a Gaussian-process regression forecasting and model predictive control algorithm in the context of interconnected microgrids. The scheme, which operated at six hours time horizon, achieved superior results with only a small deviation from the optimal operation calculated offline assuming perfect foresight. We also demonstrate that whilst a longer horizon provides a better solution in terms of lower cost of electricity, the battery cycling rate is also higher. Finally, we demonstrate improvements in renewable and load forecasts by sharing information between the microgrids.