Energies

Research

22 pages, 5273 KiB

Open AccessArticle

Application of Machine Learning to Predict the Performance of an EMIPG Reactor Using Data from Numerical Simulations

by Owen Sedej, Eric Mbonimpa, Trevor Sleight and Jeremy Slagley

Energies 2022, 15(7), 2559; https://doi.org/10.3390/en15072559 - 31 Mar 2022

Cited by 1 | Viewed by 2173

Microwave-driven plasma gasification technology has the potential to produce clean energy from municipal and industrial solid wastes. It can generate temperatures above 2000 K (as high as 30,000 K) in a reactor, leading to complete combustion and reduction of toxic byproducts. Characterizing complex [...] Read more.

Microwave-driven plasma gasification technology has the potential to produce clean energy from municipal and industrial solid wastes. It can generate temperatures above 2000 K (as high as 30,000 K) in a reactor, leading to complete combustion and reduction of toxic byproducts. Characterizing complex processes inside such a system is however challenging. In previous studies, simulations using computational fluid dynamics (CFD) produced reproducible results, but the simulations are tedious and involve assumptions. In this study, we propose machine-learning models that can be used in tandem with CFD, to accelerate high-fidelity fluid simulation, improve turbulence modeling, and enhance reduced-order models. A two-dimensional microwave-driven plasma gasification reactor was developed in ANSYS (Ansys, Canonsburg, PA, USA) Fluent (a CFD tool), to create 644 (geometry and temperature) datasets for training six machine-learning (ML) models. When fed with just geometry datasets, these ML models were able to predict the proportion of the reactor area with temperature above 2000 K. This temperature level is considered a benchmark to prevent formation of undesirable byproducts. The ML model that achieved highest prediction accuracy was the feed forward neural network; the mean absolute error was 0.011. This novel machine-learning model can enable future optimization of experimental microwave plasma gasification systems for application in waste-to-energy. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

33 pages, 16154 KiB

Open AccessArticle

Implementation of ANN-Based Embedded Hybrid Power Filter Using HIL-Topology with Real-Time Data Visualization through Node-RED

by Raffay Rizwan, Jehangir Arshad, Ahmad Almogren, Mujtaba Hussain Jaffery, Adnan Yousaf, Ayesha Khan, Ateeq Ur Rehman and Muhammad Shafiq

Energies 2021, 14(21), 7127; https://doi.org/10.3390/en14217127 - 01 Nov 2021

Cited by 10 | Viewed by 2477

Abstract

Electrical power consumption and distribution and ensuring its quality are important for industries as the power sector mandates a clean and green process with the least possible carbon footprint and to avoid damage of expensive electrical components. The harmonics elimination has emerged as [...] Read more.

Electrical power consumption and distribution and ensuring its quality are important for industries as the power sector mandates a clean and green process with the least possible carbon footprint and to avoid damage of expensive electrical components. The harmonics elimination has emerged as a topic of prime importance for researchers and industry to realize the maintenance of power quality in the light of the 7th Sustainable Development Goals (SDGs). This paper implements a Hybrid Shunt Active Harmonic Power Filter (HSAHPF) to reduce harmonic pollution. An ANN-based control algorithm has been used to implement Hardware in the Loop (HIL) configuration, and the network is trained on the model of pq0 theory. The HIL configuration is applied to integrate a physical processor with the designed filter. In this configuration, an external microprocessor (Raspberry PI 3B+) has been employed as a primary data server for the ANN-based algorithm to provide reference current signals for HSAHPF. The ANN model uses backpropagation and gradient descent to predict output based on seven received inputs, i.e., 3-phase source voltages, 3-phase applied load currents, and the compensated voltage across the DC-link capacitors of the designed filter. Moreover, a real-time data visualization has been provided through an Application Programming Interface (API) of a JAVA script called Node-RED. The Node-RED also performs data transmission between SIMULINK and external processors through serial socket TCP/IP data communication for real-time data transceiving. Furthermore, we have demonstrated a real-time Supervisory Control and Data Acquisition (SCADA) system for testing HSAHPF using the topology based on HIL topology that enables the control algorithms to run on an embedded microprocessor for a physical system. The presented results validate the proposed design of the filter and the implementation of real-time system visualization. The statistical values show a significant decrease in Total Harmonic Distortion (THD) from 35.76% to 3.75%. These values perfectly lie within the set range of IEEE standard with improved stability time while bearing the computational overheads of the microprocessor. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

22 pages, 550 KiB

Open AccessArticle

Deep Reinforcement Learning for Hybrid Energy Storage Systems: Balancing Lead and Hydrogen Storage

by Louis Desportes, Inbar Fijalkow and Pierre Andry

Energies 2021, 14(15), 4706; https://doi.org/10.3390/en14154706 - 03 Aug 2021

Cited by 11 | Viewed by 2369

Abstract

We address the control of a hybrid energy storage system composed of a lead battery and hydrogen storage. Powered by photovoltaic panels, it feeds a partially islanded building. We aim to minimize building carbon emissions over a long-term period while ensuring that 35% [...] Read more.

We address the control of a hybrid energy storage system composed of a lead battery and hydrogen storage. Powered by photovoltaic panels, it feeds a partially islanded building. We aim to minimize building carbon emissions over a long-term period while ensuring that 35% of the building consumption is powered using energy produced on site. To achieve this long-term goal, we propose to learn a control policy as a function of the building and of the storage state using a Deep Reinforcement Learning approach. We reformulate the problem to reduce the action space dimension to one. This highly improves the proposed approach performance. Given the reformulation, we propose a new algorithm, DDPG

_{α_{rep}}

, using a Deep Deterministic Policy Gradient (DDPG) to learn the policy. Once learned, the storage control is performed using this policy. Simulations show that the higher the hydrogen storage efficiency, the more effective the learning. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

28 pages, 39943 KiB

Open AccessArticle

Spatially-Explicit Prediction of Capacity Density Advances Geographic Characterization of Wind Power Technical Potential

by Dylan Harrison-Atlas, Galen Maclaurin and Eric Lantz

Energies 2021, 14(12), 3609; https://doi.org/10.3390/en14123609 - 17 Jun 2021

Cited by 4 | Viewed by 2450

Abstract

Mounting interest in ambitious clean energy goals is exposing critical gaps in our understanding of onshore wind power potential. Conventional approaches to evaluating wind power technical potential at the national scale rely on coarse geographic representations of land area requirements for wind power. [...] Read more.

Mounting interest in ambitious clean energy goals is exposing critical gaps in our understanding of onshore wind power potential. Conventional approaches to evaluating wind power technical potential at the national scale rely on coarse geographic representations of land area requirements for wind power. These methods overlook sizable spatial variation in real-world capacity densities (i.e., nameplate power capacity per unit area) and assume that potential installation densities are uniform across space. Here, we propose a data-driven approach to overcome persistent challenges in characterizing localized deployment potentials over broad extents. We use machine learning to develop predictive relationships between observed capacity densities and geospatial variables. The model is validated against a comprehensive data set of United States (U.S.) wind facilities and subjected to interrogation techniques to reveal that key explanatory features behind geographic variation of capacity density are related to wind resource as well as urban accessibility and forest cover. We demonstrate application of the model by producing a high-resolution (2 km × 2 km) national map of capacity density for use in technical potential assessments for the United States. Our findings illustrate that this methodology offers meaningful improvements in the characterization of spatial aspects of technical potential, which are increasingly critical to draw reliable and actionable planning and research insights from renewable energy scenarios. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Graphical abstract

21 pages, 3354 KiB

Open AccessArticle

Flexible Transmission Network Expansion Planning Based on DQN Algorithm

by Yuhong Wang, Lei Chen, Hong Zhou, Xu Zhou, Zongsheng Zheng, Qi Zeng, Li Jiang and Liang Lu

Energies 2021, 14(7), 1944; https://doi.org/10.3390/en14071944 - 01 Apr 2021

Cited by 10 | Viewed by 2211

Abstract

Compared with static transmission network expansion planning (TNEP), multi-stage TNEP is more in line with the actual situation, but the modeling is also more complicated. This paper proposes a new multi-stage TNEP method based on the deep Q-network (DQN) algorithm, which can [...] Read more.

Compared with static transmission network expansion planning (TNEP), multi-stage TNEP is more in line with the actual situation, but the modeling is also more complicated. This paper proposes a new multi-stage TNEP method based on the deep Q-network (DQN) algorithm, which can solve the multi-stage TNEP problem based on a static TNEP model. The main purpose of this research is to provide grid planners with a simple and effective multi-stage TNEP method, which is able to flexibly adjust the network expansion scheme without replanning. The proposed method takes into account the construction sequence of lines in the planning and completes the adaptive planning of lines by utilizing the interactive learning characteristics of the DQN algorithm. In order to speed up the learning efficiency of the algorithm and enable the agent to have a better judgment on the reward of the line-building action, the prioritized experience replay (PER) strategy is added to the DQN algorithm. In addition, the economy, reliability, and flexibility of the expansion scheme are considered in order to evaluate the scheme more comprehensively. The fault severity of equipment is considered on the basis of the Monte Carlo method to obtain a more comprehensive system state simulation. Finally, extensive studies are conducted with IEEE 24-bus reliability test system, and the computational results demonstrate the effectiveness and adaptability of the proposed flexible TNEP method. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

21 pages, 5401 KiB

Open AccessArticle

Radio Frequency Based Wireless Charging for Unsupervised Clustered WSN: System Implementation and Experimental Evaluation

by Ala’ Khalifeh, Mai Saadeh, Khalid A. Darabkh and Prabagarane Nagaradjane

Energies 2021, 14(7), 1829; https://doi.org/10.3390/en14071829 - 25 Mar 2021

Cited by 8 | Viewed by 2356

Abstract

Wireless Charging (WC) is a promising technology that has recently attracted the research community and several companies. WC has a myriad of advantages and diverse applications especially in the emerging Internet of Things (IoT) and Wireless Sensor Networks (WSNs), where energy harvesting and [...] Read more.

Wireless Charging (WC) is a promising technology that has recently attracted the research community and several companies. WC has a myriad of advantages and diverse applications especially in the emerging Internet of Things (IoT) and Wireless Sensor Networks (WSNs), where energy harvesting and conservation are very crucial to prolonging network lifetime. Several companies have launched WC products and solutions and made them available to the end-users. This paper provides experimental and practical insights about this technology utilizing off-the-shelf (commercially available) products provided by Powercast Inc.; a pioneering company that has made their wireless charging kits and solutions available to the research and academic communities. In addition, a theoretical study of this technology is presented, where a close match between the theoretical and practical results is demonstrated. This will in turn assist the learners and technology adopters to better understand the technology and adopt it in various application scenarios. Furthermore, the paper presents the potential of using WC in unsupervised clustered WSN, where the Cluster Head (CH) node is proposed to be a mobile Unmanned Ground Vehicle (UGV) equipped with a wireless charging station. The UGV position is chosen to be in the centroid of the cluster in order to ensure that wireless charging takes place in the context of the cluster nodes efficiently. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Graphical abstract

23 pages, 5778 KiB

Open AccessArticle

Multivariable Unconstrained Pattern Search Method for Optimizing Digital PID Controllers Applied to Isolated Forward Converter

by Ghulam Abbas, Muhammad Usman Asad, Jason Gu, Salem Alelyani, Valentina E. Balas, Mohammad Rashid Hussain, Umar Farooq, Ahmed Bilal Awan, Ali Raza and Chunqi Chang

Energies 2021, 14(1), 77; https://doi.org/10.3390/en14010077 - 25 Dec 2020

Cited by 11 | Viewed by 2425

Abstract

Most of the traditional PID tuning methods are heuristic in nature. The heuristic approach-based tuned PID controllers show only nominal performance. In addition, in the case of a digital redesign approach, mapping of the heuristically-designed continuous-time PID controllers into discrete-time PID controllers and [...] Read more.

Most of the traditional PID tuning methods are heuristic in nature. The heuristic approach-based tuned PID controllers show only nominal performance. In addition, in the case of a digital redesign approach, mapping of the heuristically-designed continuous-time PID controllers into discrete-time PID controllers and in case of the direct digital design approach, mapping of the continuous-time plant (forward converter) into the discrete-time plant, results in frequency distortion (or warping). Besides this, nonlinear elements such as ADC and DAC, and delay in the digital control loop deteriorate the control performance. There is a need to tune conventionally-designed digital controllers to enhance performance. This paper proposes optimized discrete-time PID controllers for a forward DC–DC converter operating in continuous conduction mode (CCM). The considered conventional digital PID controllers designed on the basis of the digital redesign and direct digital approaches are tuned by one of the multivariable unconstrained pattern search methods named Hooke–Jeeves (H–J) search method to ensure excellent output voltage regulation performance against the changes in input voltage and load current. Numerical results show that the H–J-based optimized PID compensated forward converter system shows tremendous improvement in performance compared to its unoptimized counterpart and simulated annealing (SA)-based compensated system, thus justifying the applicability of the H–J method for enhancing the performance. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

18 pages, 5202 KiB

Open AccessArticle

Adaptive Takagi–Sugeno Fuzzy Model Predictive Control for Permanent Magnet Synchronous Generator-Based Hydrokinetic Turbine Systems

by Yu-Chen Lin, Valentina Emilia Balas, Ji-Fan Yang and Yu-Heng Chang

Energies 2020, 13(20), 5296; https://doi.org/10.3390/en13205296 - 12 Oct 2020

Cited by 6 | Viewed by 2161

Abstract

This paper presents a sensorless model predictive torque control strategy based on an adaptive Takagi–Sugeno (T–S) fuzzy model for the design of a six–phase permanent magnet synchronous generator (PMSG)–based hydrokinetic turbine systems (PMSG-HTs), which not only provides clean electric energy and stable energy-conversion [...] Read more.

This paper presents a sensorless model predictive torque control strategy based on an adaptive Takagi–Sugeno (T–S) fuzzy model for the design of a six–phase permanent magnet synchronous generator (PMSG)–based hydrokinetic turbine systems (PMSG-HTs), which not only provides clean electric energy and stable energy-conversion efficiency, but also improves the reliability and robustness of the electricity supply. An adaptive T–S fuzzy model is first formed to characterize the nonlinear system of the PMSG before a model predictive torque controller based on the T–S fuzzy model for the PMSG system is employed to indirectly control the stator current and the stator flux magnitude, which improves the performance in terms of anti–disturbance, and achieves maximum hydropower tracking. Finally, we consider two types of tidal current, namely the mixed semidiurnal tidal current and the northwest European shelf tidal current. The simulation results demonstrate that the proposed control strategy can significantly improve the voltage–support capacity, while ensuring the stable operation of the PMSG in hydrokinetic turbine systems, especially under uneven tidal current speed conditions. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Graphical abstract

15 pages, 5009 KiB

Open AccessArticle

Improving Load Forecasting of Electric Vehicle Charging Stations Through Missing Data Imputation

by Byungsung Lee, Haesung Lee and Hyun Ahn

Energies 2020, 13(18), 4893; https://doi.org/10.3390/en13184893 - 18 Sep 2020

Cited by 11 | Viewed by 2720

Abstract

As the penetration of electric vehicles (EVs) accelerates according to eco-friendly policies, the impact of electric vehicle charging demand on a power distribution network is becoming significant for reliable power system operation. In this regard, accurate power demand or load forecasting is of [...] Read more.

As the penetration of electric vehicles (EVs) accelerates according to eco-friendly policies, the impact of electric vehicle charging demand on a power distribution network is becoming significant for reliable power system operation. In this regard, accurate power demand or load forecasting is of great help not only for unit commitment problem considering demand response but also for long-term power system operation and planning. In this paper, we present a forecasting model of EV charging station load based on long short-term memory (LSTM). Besides, to improve the forecasting accuracy, we devise an imputation method for handling missing values in EV charging data. For the verification of the forecasting model and our imputation approach, performance comparison with several imputation techniques is conducted. The experimental results show that our imputation approach achieves significant improvements in forecasting accuracy on data with a high missing rate. In particular, compared to a strategy without applying imputation, the proposed imputation method results in reduced forecasting errors of up to 9.8%. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

12 pages, 2995 KiB

Open AccessArticle

Forecasting the Energy Consumption of an Actual Air Handling Unit and Absorption Chiller Using ANN Models

by Jee-Heon Kim, Nam-Chul Seong and Wonchang Choi

Energies 2020, 13(17), 4361; https://doi.org/10.3390/en13174361 - 24 Aug 2020

Cited by 16 | Viewed by 2929

Abstract

Air conditioning in buildings accounts for 60% of the total energy consumption. Therefore, accurate predictions of energy consumption are needed to properly manage the energy consumption of buildings. For this purpose, many studies have been conducted recently on the prediction of energy consumption [...] Read more.

Air conditioning in buildings accounts for 60% of the total energy consumption. Therefore, accurate predictions of energy consumption are needed to properly manage the energy consumption of buildings. For this purpose, many studies have been conducted recently on the prediction of energy consumption of buildings using machine learning techniques. The energy consumption of the air handling unit (AHU) and absorption chiller in an actual building’s air conditioning system is predicted in this paper using prediction models that are based on artificial neural networks (ANNs), which simply and accurately allow us to forecast energy consumption with limited variables. Using these ANN models, the energy usage of the AHU and chiller could be predicted by collecting a month’s worth of driving data during the summer cooling period. After the forecast models had been verified, the AHU prediction model showed performance in the ranges of 13.27% to 15.25% and 19.42% to 19.53% for the training period and testing period, respectively, and the mean bias error (MBE) ranges were 4.03% to 4.97% and 3.48% to 4.39% for the training period and testing period, respectively. The chiller prediction model satisfied the energy consumption forecast performance criteria presented by American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) guideline 14 (the measurement of energy and demand savings), with a performance of 24.64~25.58% and 7.12~29.39% in the training period and testing period, respectively, and MBE ranges of 2.59~3.40% and 1.35~2.87% in the training period and testing period, respectively. When the training period and testing period were combined for the AHU data, the actual energy usage forecast showed a lower error rate range of 0.22% to 1.11% for the training period and 0.17% to 2.44% for the testing period. For the chiller data, the error rate range was 0.22% to 2.12% for the entire training period, but was somewhat higher at 11.67% to 15.18% for the testing period. The study found that, even if the performance criteria were met, high accuracy results were not obtained, which was due to the poor data set quality. Although the forecast model based on artificial neural network can achieve relatively high-accuracy results with sufficient amounts of data, it is believed that this will require a thorough verification of the data used, as well as improvements in the predictive model to avoid overfitting and underfitting, to achieve such good results. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Graphical abstract

14 pages, 1788 KiB

Open AccessArticle

Data Augmentation for Electricity Theft Detection Using Conditional Variational Auto-Encoder

by Xuejiao Gong, Bo Tang, Ruijin Zhu, Wenlong Liao and Like Song

Energies 2020, 13(17), 4291; https://doi.org/10.3390/en13174291 - 19 Aug 2020

Cited by 35 | Viewed by 3110

Abstract

Due to the strong concealment of electricity theft and the limitation of inspection resources, the number of power theft samples mastered by the power department is insufficient, which limits the accuracy of power theft detection. Therefore, a data augmentation method for electricity theft [...] Read more.

Due to the strong concealment of electricity theft and the limitation of inspection resources, the number of power theft samples mastered by the power department is insufficient, which limits the accuracy of power theft detection. Therefore, a data augmentation method for electricity theft detection based on the conditional variational auto-encoder (CVAE) is proposed. Firstly, the stealing power curves are mapped into low dimensional latent variables by using the encoder composed of convolutional layers, and the new stealing power curves are reconstructed by the decoder composed of deconvolutional layers. Then, five typical attack models are proposed, and the convolutional neural network is constructed as a classifier according to the data characteristics of stealing power curves. Finally, the effectiveness and adaptability of the proposed method is verified by a smart meters’ data set from London. The simulation results show that the CVAE can take into account the shapes and distribution characteristics of samples at the same time, and the generated stealing power curves have the best effect on the performance improvement of the classifier than the traditional augmentation methods such as the random oversampling method, synthetic minority over-sampling technique, and conditional generative adversarial network. Moreover, it is suitable for different classifiers. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

15 pages, 2202 KiB

Open AccessArticle

EGCIR: Energy-Aware Graph Clustering and Intelligent Routing Using Supervised System in Wireless Sensor Networks

by Tanzila Saba, Khalid Haseeb, Ikram Ud Din, Ahmad Almogren, Ayman Altameem and Suliman Mohamed Fati

Energies 2020, 13(16), 4072; https://doi.org/10.3390/en13164072 - 06 Aug 2020

Cited by 19 | Viewed by 2437

Abstract

In recent times, the field of wireless sensor networks (WSNs) has attained a growing popularity in observing the environment due to its dynamic factors. Sensor data are gathered and forwarded to the base station (BS) through a wireless transmission medium. The data from [...] Read more.

In recent times, the field of wireless sensor networks (WSNs) has attained a growing popularity in observing the environment due to its dynamic factors. Sensor data are gathered and forwarded to the base station (BS) through a wireless transmission medium. The data from the BS is further distributed to end-users using the Internet for their post analysis and operations. However, all sensors except the BS have limited constraints in terms of memory, energy and computational resources that degrade the network performance concerning the network lifetime and trustworthy routing. Therefore, improving energy efficiency with reliable and secure transmissions is a valuable debate among researchers for critical applications based on low-powered sensor nodes. In addition, security plays a significant cause to achieve responsible communications among sensors due to their unfixed and variable infrastructures. Keeping in view the above-mentioned issues, this paper presents an energy-aware graph clustering and intelligent routing (EGCIR) using a supervised system for WSNs to balance the energy consumption and load distribution. Moreover, a secure and efficient key distribution in a hierarchy-based mechanism is adopted by the proposed solution to improve the network efficacy in terms of routes and links integrity. The experimental results demonstrated that the EGCIR protocol enhances the network throughput by an average of 14%, packet drop ratio by an average of 50%, energy consumption by an average of 13%, data latency by an average of 30.2% and data breaches by an average of 37.5% than other state-of-the-art protocols. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

26 pages, 1868 KiB

Open AccessArticle

Improving Thermochemical Energy Storage Dynamics Forecast with Physics-Inspired Neural Network Architecture

by Timothy Praditia, Thilo Walser, Sergey Oladyshkin and Wolfgang Nowak

Energies 2020, 13(15), 3873; https://doi.org/10.3390/en13153873 - 29 Jul 2020

Cited by 6 | Viewed by 3195

Abstract

Thermochemical Energy Storage (TCES), specifically the calcium oxide (CaO)/calcium hydroxide (Ca(OH)₂) system is a promising energy storage technology with relatively high energy density and low cost. However, the existing models available to predict the system’s internal states are computationally expensive. An [...] Read more.

Thermochemical Energy Storage (TCES), specifically the calcium oxide (CaO)/calcium hydroxide (Ca(OH)₂) system is a promising energy storage technology with relatively high energy density and low cost. However, the existing models available to predict the system’s internal states are computationally expensive. An accurate and real-time capable model is therefore still required to improve its operational control. In this work, we implement a Physics-Informed Neural Network (PINN) to predict the dynamics of the TCES internal state. Our proposed framework addresses three physical aspects to build the PINN: (1) we choose a Nonlinear Autoregressive Network with Exogeneous Inputs (NARX) with deeper recurrence to address the nonlinear latency; (2) we train the network in closed-loop to capture the long-term dynamics; and (3) we incorporate physical regularisation during its training, calculated based on discretized mole and energy balance equations. To train the network, we perform numerical simulations on an ensemble of system parameters to obtain synthetic data. Even though the suggested approach provides results with the error of

3.96

×

10^{- 4}

which is in the same range as the result without physical regularisation, it is superior compared to conventional Artificial Neural Network (ANN) strategies because it ensures physical plausibility of the predictions, even in a highly dynamic and nonlinear problem. Consequently, the suggested PINN can be further developed for more complicated analysis of the TCES system. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Graphical abstract

21 pages, 3394 KiB

Open AccessArticle

Localized Convolutional Neural Networks for Geospatial Wind Forecasting

by Arnas Uselis, Mantas Lukoševičius and Lukas Stasytis

Energies 2020, 13(13), 3440; https://doi.org/10.3390/en13133440 - 03 Jul 2020

Cited by 12 | Viewed by 3322

Abstract

Convolutional Neural Networks (CNN) possess many positive qualities when it comes to spatial raster data. Translation invariance enables CNNs to detect features regardless of their position in the scene. However, in some domains, like geospatial, not all locations are exactly equal. In this [...] Read more.

Convolutional Neural Networks (CNN) possess many positive qualities when it comes to spatial raster data. Translation invariance enables CNNs to detect features regardless of their position in the scene. However, in some domains, like geospatial, not all locations are exactly equal. In this work, we propose localized convolutional neural networks that enable convolutional architectures to learn local features in addition to the global ones. We investigate their instantiations in the form of learnable inputs, local weights, and a more general form. They can be added to any convolutional layers, easily end-to-end trained, introduce minimal additional complexity, and let CNNs retain most of their benefits to the extent that they are needed. In this work we address spatio-temporal prediction: test the effectiveness of our methods on a synthetic benchmark dataset and tackle three real-world wind prediction datasets. For one of them, we propose a method to spatially order the unordered data. We compare the recent state-of-the-art spatio-temporal prediction models on the same data. Models that use convolutional layers can be and are extended with our localizations. In all these cases our extensions improve the results, and thus often the state-of-the-art. We share all the code at a public repository. Full article

(This article belongs to the Special Issue Machine Learning and Deep Learning for Energy Systems)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Machine Learning and Deep Learning for Energy Systems

Share This Special Issue

Special Issue Editor

Special Issue Information

Keywords

Related Special Issue

Published Papers (14 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI