Demand Side Management and Frequency Support within Microgrid

contenu protégé  Document accessible sous conditions - vous devez vous connecter ou vous enregistrer pour accéder à ou acquérir ce document.
Prix : 10,00 € TVA 20,0% comprise (8,33 € hors TVA) - Accès libre pour les ayants-droit


Demand Side Management and Frequency Support within Microgrid


302.26 Ko


Creative Commons Aucune (Tous droits réservés)
<resource  xmlns:xsi=""
        <identifier identifierType="DOI">10.23723/1301:2017-3/19478</identifier><creators><creator><creatorName>Catalin Gavriluta</creatorName></creator><creator><creatorName>Thomas Morris</creatorName></creator><creator><creatorName>Da Wang</creatorName></creator><creator><creatorName>Gaspard Lebel</creatorName></creator><creator><creatorName>Cedric Boudinet</creatorName></creator><creator><creatorName>Vincent Debusschere</creatorName></creator><creator><creatorName>Raphael Caire</creatorName></creator></creators><titles>
            <title>Demand Side Management and Frequency Support within Microgrid</title></titles>
        <resourceType resourceTypeGeneral="Text">Text</resourceType><dates>
	    <date dateType="Created">Tue 18 Jul 2017</date>
	    <date dateType="Updated">Sat 22 Jul 2017</date>
            <date dateType="Submitted">Wed 19 Sep 2018</date>
	    <alternateIdentifier alternateIdentifierType="bitstream">979740fadf8bc68b7d064b73949231f6016628c8</alternateIdentifier>
            <description descriptionType="Abstract"></description>

REE N°3/2017 Z 105 LES MICROGRIDS (PARTIE 2) DOSSIER 2 Demand Side Management and Frequency Support within Microgrid A test bench to validate ICT architecture and to dump oscillations Thomas Morris, Da Wang, Gaspard Lebel, Catalin Gavriluta, Cedric Boudinet, Vincent Debusschere and Raphael Caire Researchers at Univ. Grenoble Alpes, CNRS, Grenoble INP, G2Elab, F-38000 Grenoble, France Climate change mitigation policies in microgrids commonly rely on the replacement of fossil fuel generation assets by Renewable Energy Resources (RES-E) and the introduction of storage units. Most RES-E are inter- mittent e.g. photovoltaics and wind power. These new gene- rating units induce challenges both in terms of stability and operational planning. As such, the research and industrial communities have made solutions to these issues a primary focus of inquiry since the mid 2000s. These rely in particu- lar on the deployment of information and communication technology (ICT) in power system assets, from the microgrid operation centre to the distributed energy resource (DER) units. Among the range of possibilities of microgrids, this article aims at developing a solution to handle the issue of frequency stability of microgrids. The concept developed is based on the measurement in real time of the system frequency on-site. The stability is complex to guarantee as the inertia of such systems is weak and the associated cost of reserves is high. A new concept has been elaborated to ensure an optimal coordination of electric loads switching on and off, based on frequency deviation. Whilst this tech- nique is favourable from a financial point of view, it raises several challenges in terms of stability, as observed in real time simulations. These simulations demonstrated the dyna- mics of microgrids in which a complete ICT infrastructure has been added for coordination purposes. These simula- tions were conducted at the Grenoble Electrical Engineering Laboratory. ABSTRACT Dans le domaine des microgrids, les politiques entreprises pour lutter contre le changement climatique et les problèmes d’acheminement de ressources fossiles reposent communément sur le remplacement des moyens de production de type fioul par de nouveaux moyens de type renouvelables ainsi que l’ajout de dispo- sitifs de stockage. Les énergies renouvelables sont le plus souvent intermittentes (énergies éolienne et photovoltaïque principalement) et posent des problèmes d’incertitude et de planification opérationnelle. Les mondes de la recherche et de l’industrie se sont struc- turés depuis le milieu des années 2000 afin d’apporter des réponses aux problèmes rencontrés et d’offrir aux pays en voie de développement électrique des solutions fiables et accessibles. Cette réponse passe notamment par le dé- ploiement de technologies de l’information et de la com- munication (TIC) dans les réseaux électriques, des centres de contrôle jusqu’au sein même des moyens de production et de consommation distribués. Parmi le champ des pos- sibles du microgrid, cet article traite d’une proposition de participation des charges locales à la stabilité en fréquence des réseaux afin de mettre en place un réglage primaire de fréquence distribué sur l’ensemble des composants actifs du système pour réduire au maximum le coût des réserves. En effet, la stabilité est alors complexe à assu- rer du fait de la faible inertie des microgrids et les coûts élevés des moyens de fourniture de réserve primaire (par les groupes électrogènes classiques). Il a ainsi été élaboré un concept de coordination de charges électriques déles- tables distribuées, qui se déconnectent et se reconnectent de manière autonome sur le réseau au gré des variations de fréquence mesurées sur site. Ces modulations de puis- sance répondent à un schéma préétabli qui dépend de la consommation électrique effective de chacune des charges. Si cette solution semble séduisante du fait du coût réduit des moyens de réglage, elle n’en reste pas moins complexe du point de vue de la stabilité comme le montrent les simu- lations temps réelles réalisées au sein du Laboratoire de génie électrique de Grenoble. RÉSUMÉ 106 Z REE N°3/2017 LES MICROGRIDS (PARTIE 2)DOSSIER 2 Introduction With the necessity of reducing carbon dioxide emissions to avoid catastrophic climate change, and the eventual exhaus- tion of fossil fuel reserves, renewable energy penetration levels continue to grow in power networks across the world. Two of the most dominant renewable energy technologies (RET), wind and solar, pose challenges to the grid, owing to their variability and uncertainty. One such problem is the threat these intermit- tent sources pose to frequency stability. Frequency deviations and fluctuations have the potential to lower supply quality, damage network devices and in extreme cases, lead to power system collapse. Conventionally, various frequency regulation measures, generation dis- patch and under frequency load shed- ding (UFLS) are employed to rectify power imbalance and maintain sys- tem frequency in an acceptable range. Under increasing penetration levels of RET, traditional frequency controls lose their effectiveness. Additional reserves are required to counter the power imba- lance caused by RET intermittence. In addition, the traditional UFLS sheds a predetermined amount of load when the frequency falls below a certain thres- hold value, by disconnecting primary substations’ feeders. The consequence of this is that RETs connected to these feeders are also cut off. Therefore, control effects of load shedding are jeo- pardized or even become negative if the feeders output the power. This could happen in a subsystem with high pene- tration of renewable energy. Clearly new, intelligent and efficient solutions for frequency security are required. One such promising solution to the aforementioned challenges of frequency stability in a high RET penetration grid is that of demand regulation (DR). Tradi- tionally, frequency deviation and power balancing have been addressed on the supply side, by the dispatch of a pool of flexible generators. Many loads in the network are capable of time shifting, tem- porary interruption, or modulation. DR utilizes advanced communication, mea- surement and control devices in distribu- tion networks to modulate pools of these flexible loads to rectify power imbalance, complementing traditional frequency regulations. In such DR programs, consu- mers can adjust their consumption beha- viours in a future time horizon according to time varying prices provided by Dis- tribution System Operators (DSOs), with the DSOs utilizing contracted flexibilities of consumption to cope with power fluc- tuations and deviations in normal, critical and emergent operation states, to main- tain the frequency profile in admissible ranges [1]. In response to the question of DR, G2Elab has developed a novel distri- buted and coordinated balancing and control scheme for the local energy mar- ket in the framework of the European DREAM project [2], known as Primary Support. Flexibilities of generation or consumption from numerous customers are collected, aggregated and transfer- red to local DSO agents. Once a signi- ficant disturbance causes the frequency to deviate outside predetermined thres- holds, DSOs will activate these flexibi- lities. With high IT penetration, these reserves can be activated automati- cally, triggered by frequency events as a frequency controlled reserve (FCR). Primary Support has been imple- mented and tested at G2Elab. This paper introduces its ICT structure and the results of simulations conducted using the hardware-in-the-loop (HIL) test platform OPAL-RT. DR Scheme Description Developed at the G2ELab at Gre- noble INP, Primary Support is a coordina- tion platform for the power modulation of a pool of flexible distributed energy resources (DER) for frequency support [2]. Primary Support takes operation and business data, and calculates two parameters for each DER in its pool of flexible DERs, namely: Figure 1: Stepwise DER shedding. REE N°3/2017 Z 107 Demand Side Management and Frequency Support within Microgrid to which the DER modulates i.e. if the local frequency falls below the thres- hold value (for a load, the opposite direction for a generator), the load will modulate; lows the determination of the amount of modulation necessary. These thresholds are unique to each load, and thus invoke shedding in a stepwise, sequential order, as can be seen in figure 1. The order of shedding is merit based and determined on a range of criteria such as cost and flexibility. Primary support is targeted at the FCR and UFLS range of frequency sup- port and supports domestic, tertiary and some small industrial loads. The current system consists of four layers in a hierarchical system. These control layers include: 1. A pool of flexible DERs; 2. DER controller (slave device) e.g. exter- nal microcontroller or inherent switch or modulation device in a smart load or generator; 3. Master controller; - port platform. In the current configuration, the mas- ter unit reads the system information from the slave units describing each DER. The master unit sends the infor- mation to Primary Support, which then calculates the frequency thresholds and them back to the master unit. The master unit distributes the thresholds and gradients as well as current grid frequency to the respective slave units, which modulate the DERs accordingly. Figure 2 highlights this structure as it was set up in the laboratory, with Rasp- berry Pis (RPi) employed as the master and slave units. Power System Modelling In order to validate the feasibility and effectiveness of the Primary Support control system in responding to tran- sients and stabilising the frequency of the grid, extensive testing was required. The testing of the system for UFLS in a real grid is impractical owing to the dif- ficulty and cost of having a power sys- tem with real life components set-up necessitating simulation. The power electronics found in a grid sophisticated enough to employ DR introduces high- frequency PWM signals and switching devices, which are both time consuming and unsuitable for studying long pheno- mena. Real time simulators together with Hardware in the Loop (HIL) solve this problem and were thus employed in this study. The modelling of the power sys- tem to be interfaced with the hardware employing Primary Support was of vital Figure 2: Laboratory testing structure and hierarchy. 108 Z REE N°3/2017 LES MICROGRIDS (PARTIE 2)DOSSIER 2 importance to the relevance and accu- racy of test results. In the case of tes- ting the system for its capacity in a UFLS scheme, load-frequency dynamics were required to be modelled accurately, lea- ding to the employment of a system frequency response (SFR) model. The SFR model developed by Aik and outlined in [4] was used as the buil- ding blocks of the real-time simulation constructed in Matlab Simulink. The 2nd order model is displayed in figure 3, where 6Pd represents the imbalance (in the imbalanced block), H the system inertia, and D the system load damping. The dominant generation is assumed to be reheat-type steam boiler plant, with its associated dynamic capability of immediate total spinning reserve capa- city dispatch. Simulation Microgrid It was desirable to test the sys- tem in a configuration based on a real world grid. The grid simulated was the Kuching Island field test, which was ori- ginally represented by [4] with a third order representation and further deve- loped by [2] at the G2ELab. The Primary Support system was tested within this microgrid environment. Kuching Island is an almost insulated power system, with generation domi- nated by coal-fired steam and diesel plants. Power import through its inter- connector is maximised in the model. The model displayed in figure 3 repre- sents a third order SFR model. It was composed of the four following blocks: The system base was equal to 282 MW, and the generator ramp limited to 8%. Test Bench Choice The test bench used for testing the system was the Opal-RT OP5600 HIL system, in tandem with Matlab Simu- link and a range of Raspberry Pi micro- computers. The aforementioned power system model of the Kuching micro- grid was developed with Matlab Simu- link on a standard PC. It was run as a real-time simulation on the OP5600 providing as well the interface with the controllers. The Raspberry Pi microcom- puters were employed as both the mas- ter and the slave devices (figure 2). The slave devices held state information on the loads they were simulating being connected to, whilst the master unit communicated with the Primary Sup- port coordination platform and changed the slaves accordingly. When the state of a slave device’s load was modulated, it would communicate this with Opal- RT which updated its load information accordingly. All communication and data structures were IEC61850 compliant. In addition to a number of simulated loads in Opal-RT, some real life loads including lights and a fan were connected to the test bed through power amplifiers. This allowed the testing of the system with real life delays. For the testing, the base power pro- duction was reduced, such as would be seen in the case of a generator fai- lure or disconnection, thereby indu- cing a power imbalance. The level of base load generation was modulated up and down to observe the frequency Figure 3: Simulink model of the simulation from [4]. REE N°3/2017 Z 109 Demand Side Management and Frequency Support within Microgrid response and the Primary Support response. Results In both simulations 1 and 2, displayed in figures 4 and 5 respectively, the inter- connector cable between mainland China and Kuching Island is cut at t = 6 resul- ting in a 0.25 p.u. reduction in base load power production. In both simulations, the drop in frequency can be observed, falling to 49.4 Hz in both cases. After this event, the Kuching Island is disconnected and thus runs as a microgrid system. In the first case (figure 4), the virtual loads, indicated in green in the lower picture, are quickly shed, in excess of 0.65 p.u.. This results in a frequency up- swing at around t = 9, which triggers a large reconnection event, bringing the active load to 0.75 p.u.. The frequency upswing is quickly attenuated, remaining below 50 Hz. The frequency is close to the threshold at this point and wavers as the real power fluctuates. A minor load is triggered for reconnection at t = 54 from the slight frequency fluctuations, trigge- ring a frequency drop. The frequency crosses the threshold at t = 55 and the load is once more disconnected. The frequency starts to rise once more, trig- gering a small disconnection event. This trend of fluctuating frequency triggering events, continues until the end of the simulation. Figure 4: Results presenting oscillations. Figure 5: Modified set-up. 110 Z REE N°3/2017 LES MICROGRIDS (PARTIE 2)DOSSIER 2 In simulation 5, the load shedding regime was altered to introduce stag- gered reconnection of loads, with kp = 1.1 x 10-4 . It can be seen that after the frequency drop at t = 5, the frequency rapidly climbs. Rather than causing a large reconnection event, just 0.1 p.u. of load is introduced in the first second. This allows the frequency to continue rising. After another second, t = 11, another load block is reconnected. This causes the frequency to drop, dipping slightly below 50 Hz at t = 17, and then rising slightly to stabilise below 50 Hz, and well within a reasonable operation range. Conclusion Frequency oscillations, caused by dynamic responses of UFLS after a frequency contingency, are common challenges for active demand response. Indeed, in simulation 1, a sawtooth trend in the frequency response was observed, which could be of concern for the use of Primary Support in grids connection of flexible DERs could be of annoyance to users and damaging to equipment. This sawtooth trend was caused by the frequency overshoot of the system, which triggered events which would not have occurred in a system of higher inertia. This highlights the risk of low inertial grids to such phenomenon. The staggered reconnection of loads during a frequency upswing, as trialled in simulation 2, was success- ful in eradicating the connection oscil- lations in the system for the shedding regime. It is noted that further simula- tions, with a grid of lower inertia and a greater disturbance for example, might still display oscillations despite stag- gered reconnection. It is also possible that staggered reconnection be insuf- ficient to stabilise the frequency drop in a transient of great enough magni- LES AUTEURS Thomas Morris a réalisé son stage de master sur la thématique des gestion- naires de microgrids et des structures de communication compatibles 61850 validées sur de la simulation temps réel. Il est maintenant consultant et chercheur en Australie. Da Wang a été chercheur confirmé et Work Package Leader du projet européen Dream au G2Elab après une thèse à l’université de Liège sur la stabilité des réseaux électriques. Il est maintenant chercheur à l’université de Delft. Gaspard Lebel a été chercheur junior sur des aspects de profitabilité de la notion de centrale virtuelle et a rédigé une thèse principalement sur l’apport des res- sources locales pour la stabilité des réseaux électriques. Il est maintenant mana- ger dans le privé. Catalin Gavriluta est chercheur confirmé au G2Elab sur des aspects concernant l'apport de la communication sans fils pour les réseaux électriques intelligents et associé au projet ANR Sogreen après une thèse sur les réseau MTDC basse tension en Espagne. Cedric Boudinet est ingénieur permanent sur des aspects logiciel embarqué pour la simulation temps réel des réseaux électriques intelligents après six années de conseil en développement logiciel. Vincent Debuschère est enseignant à l’Ecole nationale supérieure de l’énergie, de l’eau et de l’environnement et chercheur au Laboratoire de génie électrique de Grenoble sur les microgrids, les charges non conventionnelles et la gestion de l’intermittence dans les réseaux de distribution. Raphael Caire est enseignant à l’Ecole nationale supérieure de l’énergie, de l’eau et de l’environnement et chercheur au Laboratoire de génie électrique de Gre- noble sur les réseaux électriques intelligents, tant sur les aspects automatisation que planification des réseaux actifs et flexibles et sécurisation des infrastructures vitales. REE N°3/2017 Z 111 Demand Side Management and Frequency Support within Microgrid tude. Future work will involve the deter- mination of the rate and magnitude of reconnection staggering, potentially correlated to inertia levels and distur- bance magnitude. The results of the system testing pro- ved the effectiveness of the Primary Support system for FCR. The sawtooth - non was observed in the low inertial grid. Future work will involve the design of a control system to limit frequency overshoot and connection oscillations. Another option could be the expan- sion of the system to include generator invocation for under frequency events, which would allow the system to control microgrids. References [1] Le Ren. C. C, Luu Ngoc. A, Ta Wei. L and Wei Jen. L. Incorporating Demand Response With Spinning Reserve to Realize an adaptive Frequency Restoration Plan for sys- tem contingencies. IEEE transactions on Smart Grid, Vol 3(3), September 2012, pp.1145-1153. [2] G. Lebel, “Distributed Energy Re- sources coordination toward the supply of ancillary services in real- time”, Thèse Univ. Grenoble Alps - Grenoble INP soutenue le 26 avril 2016. [3] G. Lebel, R. Caire, N. Hadjsaid, « Procédé de stabilisation d’un ré- seau électrique par délestage de charges », brevet européen étendu EP 3084910 A1. [4] Aik DL. A general-order system frequency response model incor- porating load shedding: analytic modeling and applications. IEEE Tran- sactions on Power Systems. 2006 May; 21(2): 709-17.