Showing papers presented at "International Symposium on Industrial Electronics in 2019"

Challenges of developing a digital twin model of renewable energy generators

Abstract: Although the recently wide propagated trend mainly known as digital twin is an essential step for optimal design and reliable functioning of large generators in the future, but no serious strategy and comprehensive study have been yet proposed. There is not even a standard description of this concept. This paper discusses the necessity and challenges of digital twin models of large renewable energy generators and introduces a comprehensive modeling strategy for developing a multi-domain live simulation platform of large generators for wind and hydro power plants.

Model Predictive Frequency Control of Low Inertia Microgrids

Abstract: In isolated power systems with low rotational inertia, fast-frequency control strategies are required to maintain frequency stability. Furthermore, with limited resources in such isolated systems, the deployed control strategies have to provide the flexibility to handle operational constraints so the controller is optimal from a technical as well as an economical point-of-view. In this paper, a model predictive control (MPC) approach is proposed to maintain the frequency stability of these low inertia power systems, such as microgrids. Given a predictive model of the system, MPC computes control actions by recursively solving a finite-horizon, online optimization problem that satisfies peak power output and ramp-rate constraints. MATLAB/Simulink based simulations show the effectiveness of the controller to reduce frequency deviations and the rate-of-change-of-frequency (ROCOF) of the system. By proper selection of controller parameters, desired performance can be achieved while respecting the physical constraints on inverter peak power and/or ramp-rates.

Empirical Evaluation of Exponentially Weighted Moving Averages for Simple Linear Thermal Modeling of Permanent Magnet Synchronous Machines

Abstract: Permanent magnet synchronous machines (PMSMs) are a popular choice in many traction drive applications due to their high energy and power density and moderate assembly costs. However, electric motor thermal robustness in general is harmed by the lack of accurate temperature monitoring capabilities such that safe operation is ensured through oversized materials at the cost of its effective utilization. Classic thermal modeling is conducted through lumped-parameter thermal networks (LPTNs), which help to estimate internal component temperatures rather precisely but also require expertise in choosing model parameters and lack physical interpretability as soon as their degrees of freedom are curtailed in order to meet the real-time requirement. In this work, it is shown that, as an alternative to LPTNs, linear regression achieves similar predictive performance with low computational complexity as long as input representations are preprocessed with exponentially weighted moving averages. Thus, domain knowledge becomes neglectable, and estimation performance depends entirely on collected data and considered input representations. Furthermore, dependence on data quantity and data diversification is examined in order to assess the minimal mandatory amount of test bench measurements.

LSTM-based Short-term Load Forecasting for Building Electricity Consumption

Abstract: The unprecedented level of flexibility in energy management is required to ensure the balance of real-time energy production and consumption. Accurate short-term load forecasting (STLF) is vital for making the intelligent operation scheme. However, conventional forecasting techniques may not meet the increasingly demanding precision in load forecasting. This paper presents a novel energy load forecasting methodology based on Recurrent Neural Network (RNN), specifically Long Short-term Memory (LSTM) algorithms. The proposed LSTM-based model was trained and tested on a benchmark dataset which contained electricity consumption data for different kinds of buildings in America with onehour resolution. The comparative models including multi-layer perceptron neural network (MLP), random forest (RF), and kernelized support vector machine (SVM) was also tested on the same dataset. The week-ahead forecasting results have shown that the proposed LSTM-based model outdoes the three comparative models in nine of twelve months.

A Deep Reinforcement Learning Method for Mobile Robot Collision Avoidance based on Double DQN

Abstract: We propose a deep reinforcement learning method based on Double Q-learning Network(DDQN) to enable mobile robots to learn collision avoidance and navigation capabilities autonomously. Information such as target position, obstacle size and position is taken as input, and the direction of movement of the robot is taken as an output. Traditional mobile robots usually requires real-time accurate and fast Simultaneous Localization And Mapping(SLAM) technology for global navigation. We aim at the scenario that after an initial globally feasible path is established, the path could be split into finite segments of sub-goals, and the proposed method focuses on using deep reinforcement learning to control the robots reaching the subgoals in sequence. Experiments show that the proposed method can navigate the mobile robots to desired target position without colliding with any obstacle and other moving robots, and the method is successfully implied on a physical robot platform. In addition, the method is a non-global path planning method, which greatly reduces the computational cost.

A Review of Artificial Intelligence Based Intrusion Detection for Software-Defined Wireless Sensor Networks

Abstract: Wireless communications and Wireless Sensor Networks (WSNs) are intensively used in manufacturing industries, in medical devices, for the determination of the position and for the guidance of military drones and bombs. Given the scope of utilization of WSNs, the security of wireless communications is a very critical problem that must be tackled accordingly. A Software-Defined Wireless Sensor Network (SDWSN) is realized by infusing a Software Defined Network (SDN) model in a WSN. In this paper, the cryptography schemes as well as the security threats related to SDWSNs are identified and the Artificial Intelligence (AI) techniques used to detect intrusions in SDWSNs are presented. It is shown that a two-level security model combining cryptography schemes and AI techniques can be used to fight malicious attacks against SDWSNs.

Nine-Level Packed U-Cell (PUC9) Inverter Topology with Single-DC-Source and Effective Voltage Balancing of Auxiliary Capacitors

Abstract: The 9-Level Packed U-cell (PUC9) inverter topology is presented in this paper. It uses eight switches, one DC voltage source and two auxiliary capacitors to generate a 9-level voltage at the output. Obtaining two small size of flying capacitors through proper voltage balancing technique integrated into the modulation unit and without using the complicated external control system that may increase its reliability is the innovation of this article. This topology offers a low-cost inverter compared to the other similar 9-level inverters due to the reduced number of devices and simple voltage controller. PUC9 topology is chosen by trade off between redundancy of states and voltage levels at the output of inverter by selection of suitable proportion of voltage at capacitors than each other and DC source. The stand-alone operation of the PUC9 inverter with implemented voltage balancing technique is investigated through simulation analysis in Matlab-Simulink and results are discussed in details.

Maximizing Vanadium Redox Flow Battery Efficiency: Strategies of Flow Rate Control

Abstract: Vanadium redox flow batteries (VRFBs) are one of the most promising technologies for large-scale energy storage due to their flexible energy and power capacity configurations. The energy losses evaluation assumes a very important rule on the VRFB characterization in order increase the efficiency of the battery. Very few papers describe the relations between hydraulic, electrical and chemical contributions to the system energy losses, especially in a large size VRFB system. In the first part a fluid dynamics characterization of a 9kW / 27 kWh VRFB test facility has been conducted. In particular, we will consider the internal resistance as the sum of an ohmic and a transport resistance. Secondly, an overall loss assessment based on both numerical and experimental results has been carried out. Finally, some improvements in the battery management strategy and in stack engineering are proposed, that results from this work and can help the future designer to develop more efficient VRFB stack with a compact design.

Current Based Model Predictive Control for DC Capacitor Optimization in Grid-Connected and Stand-Alone Nine-Level Packed U-Cell Inverter

Abstract: This paper investigates the application of Model Predictive Control (MPC) for controlling the load current and balance capacitors voltages of a nine-level Packed U-Cell inverter in both grid-connected and stand-alone modes of operations. PUC inverter has the advantage of minimum manufacturing cost compared to other multilevel topologies, but it needs bulky dc capacitors. Therefore, this paper proposes the MPC as a proper solution to optimize the capacitors size of nine-level PUC inverter, which makes it more interesting for various industrial applications particularly for single-phase grid-connected devices. For this purpose, structure of nine-level PUC is firstly investigated in terms of switching states, advantages and disadvantages. Afterwards, MPC control method has been developed and designed based on the mathematical equations of nine-level PUC topology. The proposed control loop is thereafter simulated using Matlab/Simulink software and the obtained results prove the desired load current reference tracking and capacitors voltages balancing with optimized capacitance; while minimum Total Harmonic Distortion (THD) and low capacitors voltages ripple have been observed.

Semantic Navigation Mapping from Aerial Multispectral Imagery

Abstract: The emergence of Unmanned Aerial Vehicles (UAV) in the Precision Agriculture (PA) domain allowed decision support systems to have access to aerial images of the terrain surface. By exploiting multispectral aerial imagery, crop health analysis and terrain classification and mapping is possible. Therefore, this work proposes an open-source ROS-based (Robot Operating System) framework, capable of handling multispectral imagery and exploit it for terrain classification, building semantic maps structured by layers of vegetation, water, soil and rocks. The obtained experimental results were validated in the scope of several research projects funded by the Portuguese Rural Development Plan PDR2020, with success rates between 70% and 90%.

A Decentralized Adaptive Control Method for Frequency Regulation and Power Sharing in Autonomous Microgrids

Abstract: In this paper, a novel decentralized control structure is proposed to compensate voltage and frequency deviations of an ac microgrid (MG) with higher bandwidth compared to the conventional control structure with no need for a communication network. This approach is realized by firstly employing finite control set model predictive control of voltage source converter at the primary control level. Then, an adaptive droop control is presented to keep the voltage and frequency of the MG stable in steady state and serve as a secondary level of hierarchical control. Therefore, the MG voltage and frequency are restored to the nominal value with a decentralized communication-free control structure. Simulation results verify the accurate frequency and voltage restoration as well as fast power-sharing during the transient and steady-state performance with no need for communication infrastructure.

Applications of artifical intelligence in power electronics

Abstract: In this paper, various areas of artificial intelligence as well as their applications on power electronics, transmission lines and micro-grids have been introduced. In the first part of this paper, a brief history of artificial intelligence is given. Furthermore, the role of artificial intelligence in different industries is explained. Finally, the most popular algorithms of artificial intelligence are introduced and compared. In the second part of this paper, the role of power electronics is explained. Moreover, the ever-increasing complexity of control, fault detection and diagnosis for power electronic devices are introduced. Finally, it is emphasized that how artificial intelligence can be widely used to overcome the complex problems linked with power electronics and their applications in various areas especially the renewable energies.

Deep Learning for Ground Reaction Force Data Analysis: Application to Wide-Area Floor Sensing

Abstract: Deep learning methods are proposed to process and fuse raw spatiotemporal ground reaction forces (GRF) to accurately categorize gait pattern. These methods are based on convolutional neural network and long short-term memory networks architectures to learn spatiotemporal features, automatically end-to-end from raw GRF sensor signals. In a case study on Parkinson's disease (PD) data, spatiotemporal signals of gait for PD patient and healthy subjects are processed and classified, resulting an effective gait pattern classification with a precision performance of 96%. Deep learning considerably achieved better classification results, compared to the shallow learning methods with the handcrafted features. This implies that for the purpose of automatic decision-making, it is beneficial to utilize deep learning methods to analyse GRF. This insight is portable across a range of industrial tasks that involve complex spatiotemporal GRF signals classification. The proposed models are computationally efficient and able to achieve high classification precision from a large set of GRF signals.

Intelligent fault diagnosis of wind turbine gearbox based on Long short-term memory networks

Abstract: Gearbox with complex structure is one of the most fragile components of wind turbines. Fault diagnosis of gearbox is crucial to reduce unexpected downtime and economic losses. This paper proposes an intelligent fault diagnosis method based on the Long Short-term Memory (LSTM) networks. Firstly, the multi- accelerometers vibration signals are divided into data segments. Then the common time domain features are extracted from these data segments. After that, these features are fed into the LSTM networks for fault pattern classification. The proposed method has no requirement for well-selected features, and also classifies the fault type accurately. The performance of the proposed method is validated by the multi- accelerometers vibration signals from wind turbine driven test rig. Through comparing with support vector machine (SVM) method, the superiority of the proposed method is verified. Moreover, the impact of different data segments on classification results is analyzed in this paper.

A Comprehensive Review of Thermal Design and Analysis of Traction Motors

Abstract: Traction motors used in electric vehicles have some definite operational requirements including high torque at low speed and constant high power at high speeds. These high torque and power requirements contribute directly to higher heat generation in the motors. Thus, an effective thermal design is needed during motor design for improved thermal performance. This review paper mainly focuses on different types of effective methods of thermal design analyzing the thermal performance of traction motors. In broad view, this review paper presents a brief discussion on comparison of different methods of thermal design and analysis for different types of traction motors. The final goal of the comparison is to show how improved thermal design can be achieved through optimum electromagnetic and structural design characteristics for different traction motor geometries.

Virtual Reality Based Maintenance Training Effectiveness Measures – a Novel Approach for Rail Industry

Abstract: This paper investigates engineering education and technical training optimization techniques based on augmented and virtual reality (AVR) technologies. Different learning methods are discussed in detail, with respect to identified enhanced learning outcomes such as higher engagement level, knowledge retention, and applied practical competency. Integration of AVR technology has a proven potential to elevate engagement and memory retention capacity according to the distinguished literature discussed in this paper. The primary objective of this research is establishing an evaluation framework that effectively measure the learning experience. Two experimental groups of novice trainee students were exposed to a traditional, class room based training program and virtual reality (VR) based training program. Both programs deliver a maintenance training activity of a chosen mechanism (sub-system), which will be used as preliminary results. An evaluation procedure was developed to assess each trainee to gather important data such as time consumption, accuracy, aptitude, performance level and intuition. A mathematical model was formulated to assess and index the individuals from the collected data, to establish their individual cognitive and psychomotor skill index (CPSI). This numerical rating allows assessment of workers over various tasks, and it can be used as an indicator of the competencies of trainees. Through integrated VR training, individuals were shown to have higher engagement during the training process, as their understanding of the scenario improved. This allowed them to work faster, more efficiently and more accurately with a shorter overall training time.

A Stackelberg Game Modelling Approach for Aggregator Pricing and Electric Vehicle Charging

Abstract: Electric vehicle (EV) has been growing rapidly around the world in the recent years. With the development of EVs, the aggregator will play an important role of intermediary agent between a wholesale energy market and end EV owners. This paper proposes a Stackelberg game model based on the grid electricity price and the aggregator pricing mechanism, in which the upper level maximizes the aggregator’ s profit, while the lower level minimizes the charging cost of each EV according to the prices. An optimization method is developed to calculate equilibrium of the game model by quadratic programming and iteration algorithm. Case studies with the proposed game model demonstrate the impacts of aggregator’s pricing mechanism.

Photovoltaic System Based On The PUC9 Inverter

Abstract: A new competitive photovoltaic system based on the PUC9 converter is presented in this paper. It uses a reduced count of semiconductors and passive components. These advantages are performed using a control technique which allows the balancing of capacitors voltages in open loop operation. The proposed system allows a power conversion with high energetic efficiency which is due to the reduced harmonics impact of the PUC9 inverter. The Perturb & Observe algorithm is performed to extract the maximum power of the PV panel. The whole photovoltaic system is based on the association of this MPPT technique and the proposed balancing technique. Consequently, the load voltage and current are nearly sinusoidal. Therefore, there is no need for additional investment in any filters.

A review of sliding mode observers based on equivalent circuit model for battery SoC estimation

Abstract: Battery technology is a major technical bottleneck with electric vehicles (EVs). It is necessary to perform state of charge (SoC) estimation in order to ensure battery safe usage and reduce its average lifecycle cost. Sliding mode observer (SMO) has been used widely in battery SoC estimation owing to its simplicity and robustness to both parameter variations and external disturbances. The SMO uses a switching function of the model error as feedback to drive estimated states to a hypersurface where there is no difference between measured and estimated output exactly. In this paper, three kinds of SMOs based on equivalent circuit model (ECM) for SoC estimation in the existing literatures are reviewed. Their difference in the structures and principles are discussed in the hope of providing some inspirations to the design of efficient SMO based SoC estimation methods.

Active Fault-Clearing on Long-Distance Overhead Lines using a Hybrid Modular Multilevel Converter

Abstract: Hybrid modular multilevel converters (MMCs) that consist of a combination of half-bridge submodules (HBSMs) and full-bridge submodules (FBSMs) can block dc fault currents. Because the excessive energy stored in the dc lines must be absorbed by the capacitors of FBSMs, the FBSM overvoltage may exceed the acceptable level in the case of long-distance lines when the converter blocking approach is used. An active clearing approach is studied to clear fault current by adjusting the dc-link voltage using the negative voltage states of FBSMs. Thereby, the excess energy stored in the lines can be transmitted into the ac grid instead of being absorbed by FBSM capacitors because the hybrid MMC continuously operates during the fault-clearing process. A model for estimating the fault-clearing time of the active clearing process is proposed. Considering the expected fault-clearing time, the required number of FBSMs in the active clearing approach is analyzed. The analytical analysis method and the proposed fault-clearing strategy are verified by the simulation results.

Reactive Power Support in Medium Voltage Networks by Coordinated Control of Distributed Generators in Dispatchable Low-Voltage Microgrid

Abstract: This paper proposes a strategy to coordinate distributed generators in low-voltage micro-grids to support its medium-voltage point-of-common coupling (PCC) with reactive power. The well-known voltage–reactive power function (volt–var) is applied directly at the PCC by a master controller. This approach enables this system to contribute with reactive power support to the medium voltage network as conventional capacitor banks that are switched on and off depending on the load/voltage profile. The power-based control strategy is used to coordinate the distributed generators in the microgrid in order to regulate the active/reactive power injection and ancillary services, which allows the active and reactive power flow at the PCC to be controlled according to the need of the distribution system operator. Simulation results based on a typical urban three-phase low-voltage network show that this method enables the microgrid to operate as a single-controllable entity, regulating the voltage and increasing the capacity of the medium voltage network to transfer active power.

Using blockchain to enhance the security of fog-assisted crowdsensing systems

Abstract: Recent years has witnessed a boom in fog-assisted crowdsensing, which exploits powerful sensing capabilities of various mobile devices or vehicles distributed in large-scale areas to efficiently gather information and make better decisions. However, the fog-assisted crowdsensing system is totally open, which provides the opportunity for malicious individuals or organizations to launch different attacks. In order to cope with the security threats from participants, a blockchain-based crowdsensing framework is proposed, which helps check the authentication of submitted sensor data and resists record tempering. Moreover, a bitcoin-based reward delivery scheme is designed to prevent requesters from denying payments. The sensing capability differences between users are considered in our design. Through security analysis and simulations evaluation, the performance superiority of the proposed framework and reward delivery scheme is demonstrated, in terms of malicious behaviour detection, user utility and sensor data quality.

A Modified IRPT Control Method for a Shunt Active Power Filter for Unbalanced Conditions

Abstract: The presence of power electronics devices on the power grid can cause power quality problems such as harmonic distortion and power system unbalance. Shunt active power filters can be employed to eliminate these power quality problems. Some control systems for the filters do not operate effectively under unbalanced power system conditions. Traditional methods of extracting the signals utilized in instantaneous reactive power theory (IRPT) have weak performance under unbalanced conditions. Therefore, a new extraction method is introduced with improved performance under unbalanced conditions. An improved control method, based on the proposed extraction method, is also presented. Simulation and experimental results validate the improved performance of this control method.

Analysis of SDN-Based Security Challenges and Solution Approaches for SDWSN Usage

Abstract: In recent years, wireless sensor networks (WSNs) and software defined networks (SDN) have witnessed overwhelming research interests in both industry and the academia. Most of these researches have been devoted to address several security challenges in the traditional WSN to mitigate network-based threats and attacks. SDN emanated as one such solutions which has been adopted in WSN to address its inherent network inflexibility, a paradigm called Software-Defined Wireless Sensor Networks (SDWSN). Albeit, SDWSN emerged as a consolidated solution, there is no guarantee that it is totally immune to current dynamic and complex security threats and vulnerabilities that grow explosively on a daily basis. Therefore, this paper aim to explore and analyze some of the security challenges faced by SDWSN. A comprehensive analysis is conducted using the content analysis approach on existing SDN researches in order to identify the existing security techniques, their impacts, and the drawbacks of each approach. The implication is on directing the research interest through having insights into current security challenges.

Virtual Synchronous Generator with Simplified Single-Axis Damper Winding

Abstract: Thanks to the improvements in power electronics, a higher integration of renewable energy sources into the electric grid using power converters is possible. Virtual Synchronous Generators (VSGs) seem to be a valid control solution for grid-tied inverters, guaranteeing not only virtual inertia features, but also several grid services, such as reactive grid support during faults and current harmonics compensation. This paper deals with the damping of the mechanical part of VSG models using only a damper winding in the q-axis. A comparison is made between two of the most widespread damping methods presented in literature, and the one proposed by the authors, that shows better features. An analytical model for the q-axis damper is presented, and an experimental comparison and validation are carried out on a 15kVA grid-connected inverter.

Architecture for Safe Human-Robot Collaboration: Multi-Modal Communication in Virtual Reality for Efficient Task Execution

Abstract: Task-sharing and Human-Robot Collaboration has gained increased attention with the widespread commissioning and usage of collaborative robots. However, recent studies show that the fenceless collaborative robots are not as harmless as they look like. In order to study Human-Robot Interaction scenarios, in a safe manner, we propose to execute the scenario in Virtual Reality simulation and afterwards implement it in real robotic applications (supervised from VR). In addition, this simulated world allows ad-hoc modifications and easy prototyping of different multi-modal communication forms. In this paper we propose an architecture for safe human-robot collaboration and describe a use-case: task of nut screwing, which is executed by the human and the robot together. The nut is hold by the human and the screw is screwed into the nut by the robot (as this part is the repetitive part of the task). The task can be executed in the VR simulation with different input and feedback channels (multi-modal) in order to identify the most efficient communication way between the human and the robot. The different input and output channels are presented in detail.

A Brief Review on State-of-the-art Grid-connected Inverters for Photovoltaic Applications

Abstract: The most important issue in Photovoltaic (PV) systems is increasing of the total system efficiency. Since solar cells have a limited and low efficiency (about 20~30%), the most effective method to increase the total system efficiency is to enhance the efficiency of interface inverters. Hence, finding a proper inverter for these systems among various kinds of existed inverters is of utmost importance among PV systems designers. Many review papers have been published and tried to gather different kinds of grid-connected inverters. But, in this article a brief review of state-of-the-art, high-efficient, and grid-connected inverters have been gathered, discussed, and compared to each other. However, former reviews just focused on typical structures which are completely out-of-date. All of the structures reviewed in this article, are trying to increase efficiency by using different techniques and structures.

Effective informational entropy reduction in multi-robot systems based on real-time TVS

Abstract: Paper covers the questions of informational entropy reduction and effectiveness improvement of multi-robot systems (MRS) in densely cluttered terrain. The presented approach uses the author’s real-time laser technical vision systems (TVS) as the main tool for environmental sensing. The article proposes the MRS collaboration approach during the indoor environment digitalization while sharing information about detected obstacles.

A Memory-Optimized and Energy-Efficient CNN Acceleration Architecture Based on FPGA

Abstract: The development of Convolutional Neural Network (CNN) contributes to breakthroughs made in the field of artificial intelligence. Compared with traditional algorithms, CNN has merits in speed and accuracy concerning detection, identification and classification. GPU is of great popularity for implementing CNN on account of its computational capacity. However, its high power consumption limits the application in the embedded field. Recently, researchers accelerate CNN utilizing Field Programmable Gate Arrays (FPGA) which is demonstrated more energy-efficient than GPU, and is suitable for the applications of embedded systems. Although FPGA has the superiority in low power consumption, powerful parallel computing and high flexibility, bandwidth and memory accessing become the bottleneck of CNN accelerator design. In this paper, a novel memory-optimized and energy-efficient CNN accelerating architecture is proposed. The paper analyzes the on-chip memory and off-chip memory resources of FPGA, and proposes a memory optimization solution using specially mixed operation of FIFO and ping-pong. To ensure accuracy, a folat-16 CNN model is used to test the framework, and evaluated on Xilinx ZCU102 platform which has both Arm-Core and FPGA on one chip. After testing the VGG-16 Net and a FCN Net with 500MB weights, the architecture is 10 times faster than CPU, and has better energy-efficiency than GPU does.

Image-Based Visual Servoing of a Quadrotor with Improved Visibility Using Model Predictive Control

Abstract: In this paper, an observer-based model predictive control scheme is proposed for the image-based visual servoing (IBVS) of a quadrotor. The control objective is to regulate the relative pose of the quadrotor to a planar ground target using a minimal sensor set, i.e., a monocular camera and an inertia measurement unit (IMU). Image features are deliberately selected to decouple the roll/pitch motion from the image kinematics, such that the overall system dynamics possesses a cascaded structure. This structural property enables the simplification of the controller design by adopting a dual-loop control structure. During visual servoing, the target object could potentially leave the field of view (FOV) of the camera, which causes failure of control input generation. To improve the visibility, a model predictive control (MPC) controller is developed to explicitly bound the roll and pitch angles to alleviate the feature loss due to large rotation. In addition, a high-gain observer is designed to facilitate the estimation of the linear velocity, which is unavailable due to the absence of the global positioning system (GPS). In the end, simulation results are presented to verify the effectiveness of the proposed control scheme.