Energy Informatics最新文献

When it comes to smart power grid (SPG) reliability and energy balancing, multi-objective energy optimization is a must. Uncertainty and several competing criteria on the demand and generation sides make multi-objective optimization difficult. Selecting a model capable of resolving scheduling issues related to loads and dispersed energy sources is, therefore, essential. This study details a concept for optimizing the SPG’s operating cost and pollutant emissions using renewable electricity. Renewable energy sources, such as solar photovoltaic and wind power, are inherently unpredictable and subject to change. Uncertainty around renewable energy is handled by the suggested approach via the use of a probability density function (PDF). In order to address a multi-objective optimization (MOCO) issue, the model that was built relies on a MOCO method. A benchmark model for energy management and control is used to verify the performance of the suggested model, which is a multi-objective deep reinforcement learning (DRL) method. According to the results, MOCO reduces operating costs by 15% and environmental emissions by 8%. The results show that compared to the comparison models, the proposed model achieves the aims better.

This paper presents a new and multidisciplinary systematic analysis of floating photovoltaic (FPV) systems that integrates recent advances in computational modelling and intelligent optimization to address persistent issues with performance, hydrodynamics, and adaptability. The review is organized according to five main goals: (i) to publish experimental and empirical results in FPV literature; (ii) to develop a unified computational approach that combines CFD and ML; (iii) to assess system improvements through multi-scale hydrodynamic modelling and AI-driven adjustments; (iv) to introduce the Bidirectional Conceptual Feedback Loop (BCFL) as a dynamic optimization model; and (v) to develop a scalable, climate-resilient FPV model for the global energy transition. Scopus, Web of Science, Google Scholar, ScienceDirect, SpringerLink, and Taylor & Francis were the sources of 404 research publications in all. 189 high-impact publications were found through a careful curation of online databases, with a focus on computational innovations, machine learning (ML)-based optimization, and hydrodynamic analysis. Following a strict inclusion and exclusion process and using Mendeley reference management software to remove duplicate records during the screening stage, authors evaluated a collection of high-impact literature, technology developments, and verified empirical data related to mooring systems, wave-wind interactions, structural stability, predictive analytics, and digital twin environments. According to the synthesis, real-time adaptation, predictive defect detection, and optimized energy yield are made possible by the clever fusion of CFD and ML, especially in dynamic aquatic environments. In order to meet the demands of both climate resilience and the scaling of renewable energy, FPV platforms must become cyber-physical, self-optimizing systems. This paper introduces a paradigm shift by using a methodical and theoretical approach to review and incorporate empirical research, advanced simulation, and AI-driven system intelligence. Future FPV development can be revolutionized by the proposed BCFL paradigm, which makes it easier to move from isolated innovation to integrative, flexible, and globally replicable FPV system design.

To address the “secondary dispatch” problem in alpine virtual power plants caused by uncertainties in decentralized resource allocation, we develop an algorithm for aggregating dispatch parameters of distributed resources to achieve real-time load-demand matching. Based on alpine power generation resources, we design a specialized virtual power plant structure and analyze its market trading applications. For the actual operation of the decentralized resources in the alpine virtual power plant, we determined the power provided by the alpine virtual power plant to the electric power system as well as the adjustable power capacity and other scheduling parameters, and then designed the dispatch model objective with the decentralized resource power and scheduling parameters based on the imitator dynamic algorithm. The model incorporates constraints based on these parameters to enable effective aggregation of adjustable power ranges for both individual resources and the entire virtual power plant, while ensuring compliance with all power constraints. This approach enhances scheduling flexibility and resolves the grid-side secondary dispatch issue. An improved ant colony algorithm based on continuous optimization was used to solve the aggregation parameters. Experimental results demonstrate superior solution performance, with the aggregated parameters increasing wind farm planned output by over 12 MW across different periods. This significantly boosts power delivery to the main grid, provides more stable supply, and improves virtual power plant revenue.

With the increasing share of distributed renewable energy sources the need arises to store excess energy and/or to shift demands to match the given supply. To coordinate multiple suppliers and demands in a local energy-system different control approaches can be used. This publication introduces a framework called MASSIVE that aims to coordinate multiple participants in a district energy-system. The energy-system is controlled in a distributed way by using a multiagent approach that is scheduled by a market-mechanism. This market-mechanism allows to coordinate many individual agents with only few restrictions by using pricing mechanisms. This offers an incentive for the agents to adapt their power consumption to best match the forecasted power supply. However, the agents are free to follow this incentive or ignore it depending on the value of the incentive. The individual agents are flexible in the internal approach to forecast power supply or demand, allowing easy development of agents using individual algorithms. The coordination takes place using a market-mechanism that is similar to the day-ahead market. It, however, is run multiple times a day to form a rolling horizon, making it less sensitive to forecasting errors. The market approach furthermore exhibits a nearly linear scalability with regard to the duration of the market clearing. On the used computer, the creation and solving of the linear optimization-problem is performed in less than one minute for approximately 1500 participating agents. Therefore, this approach is capable of real-time use and can be used in real-world applications.

The rapid growth of photovoltaic (PV) installation poses a major challenge for the energy transition in Germany. A key concern is that the increasing number of PV systems can create overloads in the low voltage grid, particularly in areas with high concentrations of installations. To better estimate the adoption of industry sized PV systems, a recommendation framework is introduced to assess the probability of adoption for specific companies. The presented framework utilizes openly available data and a hierarchical clustering approach to predict the likelihood of PV adoption for a company. Predicting PV adoption for companies allows identification of potential bottlenecks in the energy grid. As a recommendation system, it can be leveraged to promote PV systems more effectively, targeting areas with high adoption potential and optimizing grid infrastructure planning. In order to achieve that, openly available data sources have been acquired through web scraping. Company data then have been clustered using a hierarchical agglomerative approach. The recall value for the installation prediction showed an average performance of 0.75, which is found sufficient for an elaborated estimate of PV adoption.

This research focuses on investigating predictive analytics for renewable energy systems, specifically developing advanced forecasting models for solar photovoltaic (PV) power generation and non-dispatchable load consumption. To address the challenges associated with the intermittent and variable nature of solar energy, an innovative hybrid model is proposed. Specifically, this research integrates the K-nearest neighbor (KNN) classification method and genetic algorithm (GA) to optimize a backpropagation neural network (BPNN). This novel approach significantly enhances the precision of short-term solar photovoltaic power generation forecasting, enabling more accurate predictions of power output. This study proposed a prediction algorithm for non-dispatchable loads based on an online learning long short-term memory (LSTM) network. The algorithm determines whether to update parameters in the LSTM network through an online learning strategy by evaluating the root mean square error (RMSE) between prediction results and actual power consumption. The KNN-MBP algorithm reduces the RMSE by 50.36% compared to the MBP algorithm through weather classification. The KNN-GA-MBP algorithm demonstrates the best prediction performance among the three algorithms, with an RMSE of only 0.39 kW, this represents a 43.37% improvement in RMSE over the KNN-MBP algorithm and a 71.89% improvement over the MBP algorithm.

This study proposes a two-level multi-objective particle swarm optimization (MPSO) framework, enhanced by a novel mutation mechanism, to optimize energy management in stochastic dynamic distribution network reconfiguration (DDNR). The hierarchical model addresses real-time decision-making under uncertainty by minimizing power losses at Level 1 through optimal switching configurations, and simultaneously reducing operating costs and Energy Not Supplied (ENS) at Level 2 by leveraging distributed generation (DG) and electric vehicles (EV) with the Eliminating Zone method to manage uncertainties in demand and market prices. The three objectives—losses, costs, and ENS—are integrated into a non-dominated solution set to balance trade-offs. Simulation on a 95-node test network shows that the proposed MPSO outperforms conventional methods (PSO, SFLA, GWO), achieving a 25% reduction in static distribution network reconfiguration losses (from 540 kW to 449.51 kW), a 21% reduction in losses (from 39,695.45 kWh to 32,823.36 kWh), and a 35% decrease in ENS under dynamic reconfiguration. These quantitative results demonstrate the effectiveness of the proposed approach in enhancing energy efficiency, reducing costs, and improving reliability, supporting the development of sustainable and resilient smart grids.

With the widespread integration of renewable energy sources, power systems increasingly require enhanced flexibility and economic efficiency. To address the constraints imposed by high costs of conventional physical energy storage in virtual power plant planning, a bi-level expansion planning model incorporating virtual energy storage systems is proposed. Initially, a user behavior model for virtual energy storage is developed, where incentive and discount signal mechanisms are integrated to characterize charge-discharge response characteristics. Subsequently, a bi-level optimization model is established, wherein the upper level minimizes energy storage configuration costs through capacity allocation optimization, while the lower level maximizes operational revenue through energy storage scheduling strategy determination. To improve computational efficiency, a hybrid Grey Wolf Optimization algorithm is employed for model solution. The effectiveness of the proposed methodology is evaluated using an industrial park located in the southeast coastal region as a test case. Experimental results indicate that the virtual energy storage system achieved an equivalent storage capacity of 10.4 MWh, reducing total storage investment costs by 18.9% compared to physical-storage-only solutions. The proposed bi-level optimization model improves annual operational revenue by 97.9% and 55.9% compared to the baseline and single-level models, respectively. This approach effectively reduces energy storage investment costs while enhancing operational revenue of virtual power plants and system dispatch flexibility.

Positioning, coverage, and energy efficiency are essential for developing next-generation intelligent sensor networks. In wireless sensor networks (WSNs), the random deployment of sensor nodes (SNs) frequently results in suboptimal area coverage and excessive energy consumption, primarily due to overlapping sensing regions and redundant data transmissions. This research presents a Particle Swarm Optimization (PSO) algorithm to optimize the deployment of electronic information sensing nodes. The focus is on maximizing the monitored area while minimizing energy usage. A Scalable coverage-based particle swarm optimization (SCPSO) algorithm integrates a probabilistic coverage model based on Euclidean distance to detect coverage gaps and guide the optimal positioning of nodes, ensuring that each target within the region of interest is covered by at least one sensor. Data preprocessing, including Z-score normalization and Independent Component Analysis (ICA), ensures feature scaling and dimensionality reduction for improved model performance, enabling effective optimization. Experimental results under different key metrics included coverage rate (CR) for various numbers of nodes (0.9971) with 50 nodes, deployment (99.95%) with the best coverage, and computation time (0.008s), indicating significant performance improvements under optimized deployment configurations. These results highlight the effectiveness of swarm intelligence methods in enabling energy-efficient, performance-optimized deployment of electronic information sensing systems in intelligent WSNs.