Energy Speaker Series - Module 4: Power System Optimization - Virtual Power Plants and Microgrids - MATLAB & Simulink
Video Player is loading.
Current Time 0:00
Duration 33:00
Loaded: 2.05%
Stream Type LIVE
Remaining Time 33:00
 
1x
  • Chapters
  • descriptions off, selected
  • captions off, selected
      Video length is 33:00

      Energy Speaker Series - Module 4: Power System Optimization - Virtual Power Plants and Microgrids

      Seungyup Baek, VGen
      Mahdieh Sadabadi, University of Sheffield

      Session 4.1: Development of Virtual Power Plant Management System using Machine Learning Technology – Seungyup Baek, VGen

      Recently, the interest in renewable energy is increased due to fossil fuel exhaustion and environment pollution. Also, South Korea Government took initiative in the investment on renewable energy by its Korea Renewable Energy 3020 Plan, setting a goal to produce 20% of its energy from renewable sources by 2030. When the plan that will dramatically grows the share of uncontrollable renewable energy resources in the power generation comes true, it will weaken security and controllability of power grid and could cause serious challenges for power grid operators. So VPP can be a solution that can solve that problem by integrating intermittent distributed energy resources (DER) through information and communication technologies to deliver single generation profile. There are two core technologies for VPP system, that we focus on: renewable energy forecasting technology and schedule optimization technology of electricity sales. In this presentation, works done by VGen Co., Ltd. on the development of VPP system using the various state-of-the-art technologies such as machine learning technologies, optimization technology and application compiler provided by MATLAB platform is introduced.

      Session 4.2: Distributed Control Approaches in DC Microgrids – Mahdieh Sadabadi, University of Sheffield

      Direct Current (DC) Microgrids are an appealing solution for increasing penetration of renewable energy sources with DC output-type, e.g. photovoltaics and fuel cells, as well as energy storage systems. Although DC microgrids hold significant promise, their widespread use has revealed major challenges from a control point of view, mainly associated with poor scalability and uncertainties in the energy supply and demands. Recent control developments in DC microgrid control systems are based on distributed control schemes, driven by advantages of the distributed control strategies compared to centralized control in terms of improved scalability, reliability, resiliency to a single point of failure, and reduced cost. The main objective of this talk is to develop a novel resilient distributed control approach with limited communication for achieving proportional load sharing and guaranteeing the rigorous robust stability of the microgrids with uncertain supply and demand. The proposed distributed controllers are designed in MATLAB and applied to a case study of a DC microgrid consisting of several distributed generation units in MATLAB/Simscape Electrical environment. 

      About the Presenters

      Seungyup Baek, CEO of VGen, South Korea

      Seungyup Baek received the B.S. and the M.S. degrees in metallurgical engineering from Seoul National University, Seoul, Republic of Korea, in 1993 and 1995, respectively, and Pd.D. degree in industrial engineering form Penn State University, PA, USA, in 2007. He is currently a CEO of VGen that speicializes in VPP business. He has also experience in photovoltaic industry at STX Solar, where he leaded R&D department to develop next generation solar cell and module technologies. Currently, he has the responsibility of developing VPP system using artificial technology and optimization technology. His current research interests are in the field of renewable energy, solar, VPP, artificial technology, deep learning, optimization algorithm.

      Dr. Mahdieh Sadabadi, Assistant Professor in the Department of Automatic Control and Systems Engineering, University of Sheffield, United Kingdom

      Dr. Mahdieh Sadabadi is currently an Assistant Professor in the Department of Automatic Control and Systems Engineering, University of Sheffield, United Kingdom. Prior to that, she was a Research Associate at the Department of Engineering, the University of Cambridge and affiliated with Trinity College in Cambridge. She was a Postdoctoral Fellow in the Division of Automatic Control at the Department of Electrical Engineering, Linkoping University in Sweden. She received her Ph.D. in Control Systems from Automatic Control Laboratory, Swiss Federal Institute of Technology in Lausanne (EPFL), Switzerland in February 2016. 

      Recorded: 3 Dec 2020

      Hi. I am Seungyup Baek from VGen in Korea. VGen, Asia, Korean small energy venture training future Korea first, top VPP provider.

      At first, thank you for having me in this MATLAB Energy Month and Speaker Series. Today, I will talk about my previous work using MATLAB to develop virtual power plant system based on machine learning.

      Contents are as follows. I will start to introduce my organization and VPP visions. And I will describe the whole story of project for VPP and will wrap up my talk with conclusion.

      First is introduction to organization and business. Speaker, me, is Seungyup Baek. My academic background is engineering, and I hold a degree in industrial engineering at Penn State University, USA.

      I have over 25 years experience in R&D at various fields, from automotive to photovoltaic. I started my own business five years ago. And VGen is my second venture company. My main interests are from AI, machine learning, simulation to photovoltaic and VPP.

      Vision of VGen is providing the society pursues clean energy with the VPP solution, contribute reducing greenhouse gases and new energy industrial development. I hope VGen to be Korea top first VPP platform provider and top three aggregator service provider. Now VGen is joined by 15 people.

      Second chapter is project overview. Nowadays there have been tremendous changes in global power industry. Those changes could be said to start from the changing power generation structure by introduction of many small renewable energy resources, such as wind, solar, biogas, et cetera.

      There are also changes in Korean power sector too. Korean government released Renewable Energy 3020 draft, which is aiming to produce 20% or less power from renewable sources by 2030. But it is sure that renewable energy causes instability of power grid.

      The renewable energy has many good things but it is essentially uncontrollable, which means that renewable energy has variation in power generation. When renewable energy might be stored over 10% of national power capacity, this uncontrollable characteristic of renewable energy can cause new problem of instability of power grid.

      One of the new businesses in Korea power industry is Demand Response market. Demand Response is defined as that allows end user to reduce their electricity usage during periods of higher power prices. In exchange, end users are compensated for decreasing their electricity use when requested by grid operator.

      There are two types. Peak load shift and demand reduction deal will reduce the need for new power plant construction, and finally reduce greenhouse gases.

      The objective of electricity grid operation is to make a balance between consumption and production at all times. When grid operation is at normal steady state, demand will be met with equivalent supply. By grid operator, a Korea grid operator is KPX.

      At some time, demand occurs from customers such as hotel, building, factory. The grid operator order the generator to supply the amount of electricity demand. The grid is maintained in normal steady state without any risk of blackout.

      But there are any risk to being in that state at any time? Today, when the demand fluctuated out of forecasting for generators operate [INAUDIBLE] to reduce the difference between demand and supply.

      In addition, if there is more change in demand, as I explained previous page, demand will react to recover the imbalance. Future as renewable energy increase, supply becomes more unstable. Then VPP will to react to reduce the difference between demand and supply.

      This page outlines the VPP concept. Conventional power industry is shown. When controllable production can match and make a balance to uncontrollable consumption to maintain stability of power grid.

      But by the introduction of renewable energy to production sector, a parallel production becomes uncontrollable, then grid can be always on the brink of breakoff. Then VPP concept has been proposed to handle both uncontrollable production and a parallel uncontrollable consumption.

      So there, VPP are going to get plenty of small uncontrollable DERs in order to operate as one big power plant. Key technologies of VPP is monitoring machine learning and optimization.

      There are two types of VPP. One is supply-based VPP, and the other is demand-based VPP. Both VPP needs testing and optimization techniques.

      Third chapter is project goals and challenges. Here, VPP concept is explained in more detail. VPP is developed by integrating renewable energy and digitization based on ICT technology to make new energy platform.

      For VPP, many small DERs should be aggregated to make them resources. VPP solution will process the forecasting of power generation and price of electricity. Then optimize the sales plan of electricity, and if there's charge and discharges gathered, to make total profit plus.

      Here, I show configuration of our VPP. VPP has five components; VPP engineering, VPP server, data collector, and training forecasting system. Last, optimization and trading system. Data collector connect to weather stations, solar system, and ESS to get the data.

      MATLAB is used to develop two main models, forecasting and optimization. The main challenge to VPP development are as follows. For training forecasting systems, there are three challenges.

      One is automation of data preprocess to training of AI-based casting model. Three, development of standalone application forecasting system. Four, optimization trading system. There are two challenges. One is optimization of power trading strategy. Two, development of standalone application for optimization system.

      First chapter is the how did we get there and leverage MathWorks? I designed the structural training forecasting system as shown. This system has three main functions; data preprocessed, hybrid artificial intelligence, model management.

      All the main component is generation big data. Main challenge for training forecasting system was served by MATLAB toolbar. Datastore is used for automation of big data process.

      Second, Machine Learning Toolbox and Deep Learning Toolbox are used for training AI-based forecasting model. Third, Application Compiler is used for development of standalone application forecasting system.

      Here, I show the structural optimization and trader system. This system has three main functions; data supply, mathematical optimization techniques, operation strategic management. Other main component is database connector.

      Main challenges for optimization and trader system was served by MATLAB Toolbox. Database Toolbox is used for connection to database. Second, Optimization Toolbox is used for power trading strategy optimization.

      Third, Application Compiler is used for development of standalone application for optimization system.

      Fifth chapter is achievements and outlook. We succeeded VPP version one development and validation of VPP at customer side. We could shorten development time for big system like VPP using MATLAB. In addition, we could make bases for commercialization to next second staging.

      Sixth chapter is further details on solutions adopted. Here I show the commercialization version of VPP, which is restored to our customer. With this tool, our customer can attain new business of small DERs and as it is changing in 2021 in Korea.

      I will wrap up my talk with conclusion as follows. A, we could develop VPP's two main functions; training and forecasting. Second is optimization in trader. B, we could shorten the project development time and reduce the need of human resources.

      We developed GUI of VPP using Appdesigner. The last one, we compiled and developed a standalone application using MATLAB Compiler. Thank you for hearing my presentation about virtual power plant using MATLAB.

      Good morning, everybody. Today, I'm going to talk about distributed control systems in converter-interfaced direct current microgrids.

      A evolution has been happening in electrical power systems, especially in the supply side of the energy systems by integrating renewable energy sources into the grid. The main reasons for this revolution are environmental and economic concerns, as well as the sensitivity of power grids to natural disasters.

      The three-hour blackout in England and Wales on August 2019 left more than a million people without power and caused transportation chaos and major disruption to other critical infrastructure.

      Microgrids are emerging technologies that can mitigate the consequence of electrical power outages and sustain the resilience of power grids especially during the emergencies. Moreover, renewable-based microgrids are widely recognized to be essential to reducing black carbon and CO2 emissions.

      Microgrids can be connected to the main grid or be disconnected from the grid once the blackout happens. Due to the reliability and security of microgrids and their independence from the main grids, they are being used in many critical loads, such as university, hospitals, military bases, data centers, and airports.

      The London City Airport is deploying microgrids in order to have a secure and sustainable electricity. Microgrids are highly available power supplies during natural disaster. An example is the Sendai Microgrid which supplied power during the earthquake in Japan in 2011.

      The main focus of today's talk will be on direct current microgrid and their control systems. We are going to talk about stability issues, control objectives, as well as cyber security for DC microgrids. We will use MATLAB for design validation and hardware in the loop verification of microgrid control systems.

      Direct current or DC microgrids are small scale electrical network composed of distributed generation units, DG units, with DC output, such as photovoltaics and fuses, DC loads and energy storage systems.

      Compared to AC microgrids, DC microgrids offer higher efficiency due to less conversion losses from sources to load. They have simpler control systems as there is no need to control frequency or reactive power as these two are one of the main challenges in the control of AC microgrids.

      Moreover, DC microgrids have wide application and they provide natural interface with DC output, renewable energy sources, DC loads, as well as energy storage systems.

      Stability is a key challenge in DC microgrids. Renewable energy sources are connected to the microgrids via DC to DC converters. They usually have fast dynamics and low inertia.

      The inertia of a DC microgrid is far less than the inertia of the conventional AC power grids with synchronous emissions. As a result, in the case there is a disturbance or a perturbation in DC microgrids. The sensitivity of the system to these disturbances increases, as a result, the stability is affected.

      Therefore, we need to develop efficient control systems for DC microgrid in order to guarantee the stability of the system and improve their robustness.

      Another key challenge in DC microgrid is related to plug-and-play operation of the distributed generation units. The term plug-and-play refers to the possibility of connecting or disconnecting of DC units with minimal efforts or human intervention.

      The main idea is to design the local controllers of the DC units in a plug-and-play or a scalable fashion such that the design relies on only information from each DG while the stability and desired performance of the overall DC microgrids.

      As a result, in the case of plug-in or plug-out of a unit, we don't need to retune all of the controllers, and the stability is preserved.

      In order to guarantee stable and efficient operation of DC microgrid, effective control methods should be developed. In general, the control of a DC microgrid is based on a hierarchical control system consisting of two main levels; higher level and lower level control.

      Higher level control is responsible for economical operation and power management system in DC microgrids. Higher level control provides the supervisory over lower level control. Lower level control is responsible for voltage regulation and load shedding in DC microgrid, and it has higher bandwidth than higher level control. In some literature, lower level control consists of secondary and primary control.

      The main control objectives in DC microgrids are stability. As we know, stability is the most important criterion in every physical systems, including microgrids. The second objective is voltage regulation. We would like to regulate the average voltage across the microgrid at specific setpoint provided by higher level control.

      Moreover, the total demand of electricity, the total load demand, must be proportionally shared amongst distributed generation units. In order to achieve these objectives, the controller structure in DC microgrids could be either centralized or distributed.

      In a centralized control as you can see in this figure, there is a central control unit. And the central control unit receives information from the local controllers of DG unit. The central control unit processes the information and sends back the commands to these local controllers.

      Central control systems require high bandwidth communication, and they have a single point of failure. Furthermore, they are not scalable. In order to improve the robustness and scalability of control systems for DC microgrids, we usually prefer distributed control systems than centralized control.

      In the distributed control system as you can see here, there is no central control unit. And the local controllers of DC units communicate amongst each other through the communication links according to the structure of distributed control.

      Consensus-based algorithms are common control approaches in DC microgrids. These approaches are mainly based on graph or network control theory. In these approaches, the communication and information exchange between the local controllers of the DG unit is based on specific metrics named Laplacian metrics.

      In the consensus-based approach, each local controller receives information from its own DG. The information is voltage at the point of interconnection and the current of the DC to DC converter. And as I said before, the information exchange between different local controllers is according to the Laplacian metrics.

      This consensus-based controller processes the information and sends back this signal, which is the control input. And it's the duty cycle of the DC to DC converters to the convertors of each DG unit.

      Microgrids are cyber-physical systems as they include two main layers, physical layer and cyber layer. The physical layer includes a large number of distributed generation units connected via distribution lines. And the cyber layer is related to distributed control algorithms and communication.

      There is sparse communication in the control layer. Moreover, there are communication links between control and physical layer.

      Much like any other cyber-physical systems, microgrids are prone to cyber-physical attacks. The attackers aim to destabilize DC microgrid and disrupt the normal operation of microgrids by injecting false data to actuators, sensors, or transmitted data. The concepts of cyber security and cyber attack in microgrid or in general in electrical power grids are very important.

      A real world example of the most recent cyber attack on power system is the attack on the Ukraine power grid in December 2015, which was known to be the world's first power outage caused by attackers.

      Therefore, resilience and the reliability of microgrids to cyber attack and cyber-physical attacks are essential.

      The term resilience is defined as the ability to prepare for and adapt to changing conditions and withstand and recover rapidly from disruption. This disruption could be any event, accident, or attacks in DC microgrids.

      In general, we can say that the probability of consequences in microgrids is a function of microgrid vulnerability and the appearance of attacks in microgrids. In order to decrease this probability, we need to enhance the resilience of microgrids to potential cyber attacks.

      Therefore, we need to develop innovative monitoring and control systems in DC microgrid such that the microgrid becomes resilient and robust to disturbances, uncertainties, and potential cyber attack.

      In general, there are two main approaches for this purpose. The first approach is based on attack detection-based algorithms. The idea is to use some approaches, data driven-based or model-based approaches to detect attack and mitigate the adverse effects of the attack.

      The second category is based on design of distributed control systems, which have a main property, and this property is about resilience to cyber attacks.

      In the second part of this talk, we will focus on the use of MATLAB for design and verification of distributed control systems in DC microgrids. In this slide, you can see a very simple model of two distributed generation units connecting via a power line in Simulink environment. I mean, in Simscape electrical environment.

      It's a very simple model of a DG unit. The idea is to show you how we can easily use Simulink to model a DG unit or a DC microgrid in Simscape electrical environment.

      As you can see here, the DG unit includes a DC to DC converter, as I model it here, and it's a buck converter. We have the model of the shunt capacitor. And here we model the load as a resistive load.

      So in the case we have more advanced loads such as constant power load, we can easily use MATLAB to model such loads. Different DG units are connected via distribution lines modeled as a resistive inductive network.

      And as I said before, the duty of the distributed control algorithm is to develop this control input, which is the duty cycle of the DC to DC converters. We use Hardware-in-the-Loop simulator to have real-time simulation or have experimental verification of microgrids.

      The real-time simulator such as OPAL-RT or Typhoon are fully integrated with MATLAB and Simulink. They enable our developed Simulink model to interact with the realistic model of DC microgrids in real-time. I mean, it could be a model of several DC to DC converter connecting to each other.

      Here in this slide is a case study of four DG units connecting in parallel. The main objective is to equally share the total load demand amongst the DG units, and regulate the voltage of the microgrid at 48 volts. Here, the microgrid is just a low voltage microgrid.

      As you can see here, the current is equally shared, and the voltage is regulated at the specific voltage with zero offset. Then at this specific time , there is a step low change. And as you can see, after this no change the current is still equally shared and the voltage is regulated at 48 volts with zero offset.

      The second case study which is just based on MATLAB simulations highlights the importance of development of resilient distributed control approach in DC microgrids. Here at this time, the current is equally shared among all of the distributed generation unit, and the voltage is regulated at 48 volts.

      Then we assume that at equal to 1, there is an attack on communication links and on the actuator of one of the converters. And the type of the attack is false data injection. As you can see, after happening, this attack, the current is no longer shared among the distributed generation units in an equality.

      And the voltage is not regulated at 48 volts, and there is statistic error. And this statistic error is about 5 volts in this case study.

      Then in the next case study, we repeated the same case study but we use a resilient distributed control approach. And we want to show you the importance of this resilient property in distributed control algorithms.

      Here, you can see after attack happens, the quantities are still equally regulated among four DG units. And the voltage is regulated at 48 volt. Here, the transient behavior is not obvious in this figure. You can see the transient behavior of the current here after the attack happens.

      It was the duty of the distributed controller to mitigate the effect of the attacks. And there is small transient, but after the transient the current is equally shared. In summary, distributed control systems offer a promising solution for the control of DC microgrid as they offer several advantages in terms of improved scalability, reliability, flexibility, and efficiency.

      However, distributed control methods are prone to cyberattack. And therefore, the resilience of DC microgrid to this cyberattack is essential for us, especially if we know that DC microgrids are used in many critical applications where high quality power and secure power are essential.

      Therefore, we need to develop some attack detection or develop some resilient distributed control approaches to enhance the resilience and reliability of DC microgrids to potential cyberattacks. Thank you for your attention.

      View more related videos