FMI Compliant Approach to Investigate the Impact of Communication to Islanded Microgrid Secondary Control
In multi-master islanded microgrids, the inverter controllers need to share the signals and to coordinate, in either centralized or distributed way, in order to operate properly and to assure a good functionality of the grid. The central controller i…
Authors: Tung Lam Nguyen, Quoc Tuan Tran, Raphael Caire
FMI Complian t Appro ach to Investigate th e Impact of Communicati on to Islan ded Microgrid Secondary Control Tung Lam NGUYE N 1,2 , Quoc-T uan TRAN 2* , Raphael C AIRE 1 , Yvon BES ANGER 1* , Tr an The HOANG 1 ,Van Hoa NGUYEN 1 1 University Grenob le Alpes, G2Elab, F-38 000 Grenoble, France CNRS, G2Elab , F-38000 Grenoble, France 2 CEA-INES, Le B ourget-du-l ac, France *Senior Me mber, IEEE tung-lam.ngu yen @g2elab.grenob le-inp.fr Abstract —In mu lti-master islanded m icrogrids, the inverter controllers need to share the signals and to coordinate, in either centralized or distributed way, in o rder to operate properly and to assure a g ood functionality of the grid. The c entral controller is used in centralized strategy. In distributed co ntrol, Multi-agent system (MAS) i s considered to be a suitable solution for coordination of such system. However the latency and disturbance of the netw ork may disturb the communication from central controller to local controllers or am ong agents or and negatively influence the grid operation. As a consequence, communication aspects need t o be properly a ddressed d uring the control design and assessment. In this paper, w e propose a holistic approach with co-simulation u sing Functional Mockup Interface (FMI) standard to validate the microgrid control system taking into account the communication networ k . A use- case of islanded microgrid frequency secondary control with MAS under consensus algorithm is implemented to demonstrate the impact of communication and to illustrate the proposed holistic approach. Index Terms —microgrid, h ierarchical control, func tion mockup interface, multi-agent syste m , communication netw ork, I. I NTRODUCTI ON Microgrids (MGs) are essenti al b locks to constitute a nd ar e considered to be one of the m ajor changes required in th e develop ment of t he power system. In ge neral, a MG consists of a clu ster of dis tributed generato rs (DG s), loa ds, ener gy storage systems and other equip ment, which can operate in islanded m ode or grid-connected, and can seamlessly tr ansfer between t hese two modes [1 ]. In t he isla nded mode, due to the lack o f t he b ulk grid as a refe rence, t he microgrid has to keep its stable oper ation by coord inating all the elements o f its own. The control st ructure of microgrid is typic ally hierarchica l and divided into th ree l evels including primary , secondary and tertiary . The primary control level only uses the local information to response quickly to the change of the system. On the cont rary , the sec ondary and tertiary control level requires the remote informati on. The inform ation is global or adjacent d epends on the microgrid cont rolled in ce ntra lized o r distribute d fashion res pectiv ely. A centralize d control microgrid requires a central contr oller to send the appr opriate signals to every local controller. Meanwhile, the microgri d with distributed control strategy only need s the commun ication between neighbor local controllers and the depen dent on th e central controlle r coul d be i gnored . The m ulti- agent sy stem method is widely used in the latter control strategy . I n either case, the power system control requires signal from o ther spots to operate correctly and any disturbances of comm unication netw ork (i.e. latency , packet loss) w ould negatively influ ence its functi onality and subsequently affect the performance and even stability of the system [2]. Therefore, communication aspects need to be tak en into conside ration in m icrogrid cont rol assessm ent[3]. Currently , the impacts of com munication on control in a microgrid are studi ed in several approaches . A typical approach is to use co-simulation method [4]. I n general, one simu lation e nvi ronment only supports only o ne specif ic domain (PowerFactory , Matlab/Sim ulink, Powerworld,… in power system do main or ns-2, ns-3, OMNET++,… in commun ication domain). Th erefore, th e comprehens ive investigati on in a hybrid s ystem needs the com bination of tw o or m ore simulato rs. However, the synchronizat ion o f the signals exchanged b etw een simulators a nd the appropriate applicati on programm ing interface (API) are still the challenges of the co-sim ulatio n m ethod. T he an other approach is t o de sign a sy stem with th e phy sical controllers and the r eal netw ork to transmit data [5]. The grid of the sy stem is run in real-tim e and is c ontrolle d by hardware cont rollers . By settin g up the real commu nication netw ork, the data is transfe rred in a natural w ay and the be havi or could be close to the system in reality. Howev er, due to the limitation of area o f testi ng laboratori es, the distances betw een the controllers could not be reflected exactly . In this paper , w e firstly propose a co-sim ulation based method t o validate the secondary control strateg ies in island microgrid taking into account the impact of the commu nication netw ork. I n this method, we use the standar d Function Mockup Interface ( FMI) to transfer the communication emulation in to the power system si mulation environment. Secondly , we apply the propos ed meth od to ch eck the perf ormance of a m ulti- master islanded microgrid in various sce nari os of the commun ication network. A multi-agent system platform is also developed in the dis tribut ed cont rol circum stance. The r est of the paper is org anized as follows. Section I I introduces t he pr imary and secondar y control in MG. T he introduction o f FMI a nd t he pro posed method is presented in Section III. In section I V, the expe rimental setup a nd res ults are sho wn to validate t he proposed m ethod. Sect ion VI concludes the pa per and outlines p ossible future d irections. II. S ECONDARY CONT ROL IN ISLANDED MI CROGRID The microgrid control with inverter- based DGs mainly refers to the inverter control due to the fact that it is commonly interface d with the prime energy of each DG via a power electroni c in verter . Therefore, the fre quency control in microgrid is regulated b y the coordin ation of inverte r controllers . The primary control, or local control, adjusts the frequency and amplitude of voltage reference provided to the inner control loop of voltag e source inv erter. Droop contr ol method is used t o control pow er sharing between DGs in MG without commun ication. The droop equ ation presenting the relations hip between fr equency and active pow er is: 0 0 ( ) P f f k P P = − − (1) where f 0 is rated frequency o f grid voltage and P 0 is the norm al value of real pow er. f is the actual m easured value of frequen cy when the DG is supplying real power of P . k P is the droop coefficient . DC AC Inner Control Loop Reference Voltage Esin(wt) P/Q Output Measurement ∆ fi m easur ement PI + - + + + + V f L f C f P Q 50 Secondary control Primary contro l δ f i δ V f i Figure 1. The inverte r controlle r The primary co ntrol maintai ns the voltage and frequenc y stability in MG. H owever, the frequency of microgrid after t he primary control deviates fr om the rated frequenc y. T his deviation will be eliminated in secondary control level. T he steady-state error is measured from the grid and compensated by a PI co ntroller. I n particular, t he secondar y control is computed as f P i K f K f δ = ∆ + ∆ ∫ (2) where K P a nd K i ar e t he control p arameter of PI controller, ∆ f is t he measured microgrid frequency deviation and δ f is t he secondar y control signal sent to primary co ntrol level and the n added to the co rrection given b y eac h loop controller in Equation 1. 0 0 ( ) P f f f k P P δ = − − + (3) The secondar y control level method could be divided into centralized and d istributed control. Centralized control schemes for po wer s ystems ar e common. I n the traditional centralized control algorith ms, a central controller is r equired to collect all i nformation fro m local controllers and p rocess a large amount of data, which suffers fro m computat ion stress and single -point-failure. In the case of the secondar y co ntrol, the comple ment frequency is sent to all primary controllers of DG s. Such an approa ch has the ad vantage of havin g a global overview o f t he whole system. Ho wever, such a system is not easily e xtendable, is far from being co mputationall y s calable and has the vulnera bility of a single point o f failure. Alternatively, the distributed control with a spar e communication net work do es not need a central controller and each u nit is controlled by its local and neighbor co ntrol system. T he d istinct f eature of t he distrib uted approach is that the infor mation involved in th e contro l algorithm is not glo bal, but ad jacent for any given unit. Also, the le ngth of the communication links i s ofte n shorter , which offers better and more reliable latency. Moreover, the r isk o f o verall syste m failure ca n be reduced, becaus e the system does not depe nd o n a sole central co ntroller. The algorithms of centralized and distributed secondar y control will b e disc ussed more detail i n the next sectio n wit h a specific test case o f a microgri d. Due to the necessar y of trans ferring da ta i n b oth case s o f secondar y control, which i s global in term of ce ntralized control or ad jacent in term of distributed contro l, the communication net work co uld affect t he c ontrol operatio n and maybe t he stability of the microgrid. T he communicati on network, i n rea lity, is i mperfection and constraint. Theref ore, the evaluation of a co ntrol syste m needs an e xtra considerati on of the operatio n of the transmission network. III. FMI COMPLIANT CO - SIMULATION A PPROACH TO INVESTI GATE THE IMPACT OF COMMUNI CATION SYSTEM A. Co-simulatio n of Power syste m and communication network While commu nication techn ology increases rapidly its penetrati on in the power system, there are a very limited number of methodologi es and tools that allows the operators to take into account both domain s in a holistic manner. In the domain o f smart g rid n ow adays , the co-simulati on a pproach is often use d t o c ouple a power sy stem si mulator and a commun ication simulato r. Co-simulation fram ew ork allows in general the joint and simultan eous sim ulation of m odels developed with d ifferent tools, in which the intermediat e results are exchang ed during sim ulation execution. T he works on co-sim ulation of hybrid system Po wer/Com can be found in [6][7]. Gen erally , w e acknowledge two structures of co- simu lation: • Ad-hoc co-sim ulation: coup ling directly one power system simulator and one communicati on network simu lator. • Co-sim ulation wit h the Master algorithm : a master algorithm (e.g. HLA [8]) o r a co-sim ulation framew ork (e.g. Mosaik [9], Ptolemy [10]) will orchestrate the process . T his m aster algorithm is responsible for sy nchronizing different tim elines of involved sim ulators and for directing the inf ormation exchange am ong sim ulator’s inputs/ou tputs . Co-sim ulation is still a diffic ult method for the electrica l engineerin g c ommu nity due to the necessity of synchronizin g both simu lation tools properly at runtime. While power system simu lation is no rmally continuous with the possi bility of eve nt detecti on a ssocia ted to value crossing a certain threshol d; commun ication network simulation is bas ed on disc rete events whose o ccurren ce usually stochastically d istri buted with respect to time. The simulato r provides an event sch eduler to record current sy stem time and process the events in an even t list. Moreover, the existing simulation tools offer limite d options of adequate Applicat ion Programm ing Interface (API) for externa l couplin g. B. Propo sition to integ rate communicatio n emulation in to the power system environ ment To investig ate the im pact of comm unication netw ork to power system , we propose in this paper a method using Function Mockup Interface (FMI) 1 standard wh ich allows interopera bility and reus ability of the models in co-simu lation framew orks. FMI is a standard des igned to p rovide a unifie d model execution interface for dy namic system m odels betw een modeling tools and simu lation tools . The idea is tha t tools generate an d exch ange models that a dhere to the F MI specificati on. Such models are called Functional Mock-up Units (FMUs). Since its release, FMI has received a signific ant attention from both tool vendors and users. According to the information on the official website, there a re curren tly ove r 101 tools that supp ort or plan to support FMI. There is a rea l and pressing need to be able t o ex port a nd im port d yn amic system models between existing tools , and also to be able to develo p custom sim ulation environm ents. Based on the FMI standard, w e build a communication FMU to simulate t he latency and to e mulate the according to packet loss of the considered comm unication. The latency c an be calculated based o n one-sided transmission or round trip time. D ue to the nat ure o f our syste ms o f i nterest, we co nsider only one-sided tra nsmission latenc y or the time for the si gnal from sender to reach the dest ination. T he model allo ws us to emulate the communication bet w een two points in a network. In our prob lem, latenc y is d efined as t he time interval betwe en the e mission of first bit fro m the sender an d the arri val of last bit of a sig nal in the destination. In order to represent the stochasticity, a white noise is added to the latency block formula. In re ality, the co mmunication is done via se veral intermediate servers with differ ent protocols and technologies in-between. The communicatio n FMU takes that infor mation on network topo logy into acco unt. A p rotocol lib rary is available to pr ovide the message transfer mod ule with information about t he emplo yed protocols. A ge neral co nfiguration o f a communication FM U ca n then be illustrated in Fi gure 2 . In this versio n, the tra nsmissio n speed a nd data r ate need to be manually d efined. I n futu re develop ment , t his infor mation co uld be auto matically acquir ed from the p rotocol lib rary. T he Co mmunication FM U is th en 1 https://www. fmi-standard. org/ integrated into the con sidered po w er syste m to evaluate the impact of co mmunication network to system perfor mance. In t he next sectio n, we ap ply the p roposed method to an islanded m icrogrid secondary c ontrol, u sing centralized and MAS appr oaches. Figure 2. General ele ments of a communicat ion FMU IV. I MPLEMENTA TION AND RESULTS A. Test-case de scription and set up In thi s p aper, w e apply th e propos ed h olistic approach t o a test-case of operati ng a m icro-gri d in island mode. The MG includes five distributed generators (D Gs) supply ing p ower for two loads. For the sake of simplicity , the primary sources of DGs are supposed to be an ideal DC source. Each DG is controlled by a local contr oller as Figur e 1. The controlle rs send ap propriate p uls es sign al to inv erters to keep th e frequency and voltage amplitude o f the grid in referen ce values. The DGs, which are controlled in the voltage/f requency mode (inverter is th e grid-for ming inverter), operate in parallel so coordinati on is r equired . The sy stem is st udied from the time a t 6 0 seconds wh en the Load 2 is tripped, i.e. the total load is diminished an d the control has the r esponsi bility to return th e frequen cy back t o norm al condit ion. We will valida te the control system of the descr ibe d microgrid taking into account the comm un ication network in both cases of the secondary control - centralized or d istri bute d. In the central ized approach as Figur e 3 . a , w e use a microgrid central controller (MGCC) to diffuse the value of com plement frequency to l ocal controll ers. Meanwhile, in the d istri buted approach as Figu re 3 . b , the Multi-ag ent sy stem (MAS ) with the consensus algorithm is used to exchange inform ation. We investigat e in fou r cases of the netw ork: - T he first ca se is the netw ork with ideal cond itions; the influence of comm unicati on netw ork could be ignor ed. - In the three remaining cases, the data transmi ssion is considered by adding the FMU co mm unication block. The network scenarios 1, 2 and 3 ar e sorted b y the decay level of quality , i.e. gr owin g latency (cf. Ta ble 1 ). The grid in this paper is si mulated in Matlab/Si mulink by using the Si mPowerSyste ms too lbox. In the libr ary o f Malab/Simulin k, there ar e b locks which ca n use to express the delay in transmission of da ta. Ho w ever, the variatio n of latency must be defined b y th e user. The ti me of delay, i n fact, is not always ea sy to estimate. It varies in a r ange i n a natural way dep ending on the pro perties of the network such as the transmitted distance, the pr otocol, the size of data, etc. The emulation o f the network, therefore , need a more p recise module. In our work, we developed the module in t he F MI standard and i ntegrate to t he c ontrol lo op of the model using the P ilot Support Package ( PSP). Instead of transferring directly, the data is sent a nd received through the FMU b lock. This could express the dela y i n the transmission bet ween two distinct po ints. The tra nsmission is therefore emulated i n the nearly natural w ay a nd could be applied to investigate an y systems which need the broad cast of data. 1) Centralized con trol Figur e 3 .a illustrates the islanded microgrid controlled in centralize d strateg y. T he MGCC is assumed to be l ocated at a point which has distances to the local co ntr ollers as Table 1 . The MGCC is in charge of the secondary co ntr ol. It has the commun ication links w ith all local cont rollers . T he measurem ent frequency at one point in the microgrid is sent to the MG CC. T he complement frequency is then calculated at the secon dary level and diffus ed to the primary control units which are put at th e loc al sides of the DGs. T ABLE 1. T HE DISTANCES OF COMMU NICATION FROM MGCC TO CONTROLLERS MGCC - Control ler 1 (c- 1) MGCC - Control ler 2 (c- 2) MGCC - Control ler 3 (c- 3) MGCC - Control ler 4 (c- 4) MGCC - Control ler 5 (c- 5) 4 km 6 km 8 km 2 km 5 km We inves tigate the system in four described cases where the distances remain the same as prescribed in Table 1. However, the other properties of the netw ork (i.e. Data rate, serializat ion speed and bandw idth) changed and caused th e latency in d ifferent r anges of values (Figure 4). We use the box-and-w hisker plot to illu strate the distributi on of the datasets of the delay tim e with 1000 random samples in each case. The frequency of the system is demonstrate d in Figure 5. In all cases, the control system brings the frequency bac k to the reference values after the var iation happens in the microgrid . This verified tha t the designed contr ol guarantee the sta bility of the operation of the grid with ideal commun ication and with considered comm unication n etw ork scenarios. 2) Distributed control We propose a s tructur e using a multi -agent sy stem that leverages the consensus algorith m in the c ontext of microgrid distribute d control. The multi-agent syst em la yer is added o n top of the control layer. The agents in the mu lti-agent system are put at lo cations of DG units a nd take the responsi bility of processing and exchanging information. The d iffere nce between tw o control strat egies is that th e distribute d contro l fashion requires the information from neighbor a ge nts. Therefore, the inter-u nit communication with the distances described in Table 2 coul d also af fect t o the cont rol sy stem. We also investiga ted the system in the four cases of commun ication netw ork. T he laten cy in inter-ag ent transmis sion in three cases of the co mm unication netw ork is illustrat ed in Fig ure 6. T ABLE 2. THE DISTANCES OF COMMUN ICATION BETWEEN AGENTS Age nt 1-A gent 4 (1-4, 4-1 ) Age nt 2-A gent 3 (2-3, 3-2 ) Age nt 3-A gent 4 (3-4, 4-3 ) Age nt 3-A gent 5 (3-5, 5-3 ) 4 km 6 km 8 km 2 km a) Ce ntr alize d co ntrol b) D istri bute d co ntrol Figure 3. The test-case co nfiguration Figure 4. The late ncy time when trans mitting data from t he MGCC to the controller s in the three test cases (a) (b) (c) (d) Figure 5. The fre quency whe n tripping load 2 a) without communi cation networ k, b) with latency test case 1, c) w ith latency test case 2, d) with la tency test case 3 Figure 6. The inter -agent latency time in the three test cases Agent layer takes the respo nsibility alike the microgrid central cont roller. The agent la yer sends sign als to control layer to bring the frequency back to the reference value. How ever, the frequen cy at DGs oscil lates in the transient period after a variation in the m icrogrid. Due to the grid opera tes in the multi-m aster control mode, the requirem ent is that the signals sent to a ll local controllers at the almost same time and those signals have the same value to ensure the property of the control opera tion. To reach these condit ions, w e applied the average consensus algorithm . The result is that the local controllers could get simultan eously the average o f all instantane ous fre quency dev iations at the out put of DGs. In the lite rature, the agent- base d distribut ed control in a microgrid is us ually im plemented by conne cting w ith a multi- agent platfo rm [11] such as JADE, ZEUS, aiom as, etc. The interface w hich connects the simulation o f the gri d system to the platform s could be a barrier in studying . In this paper, w e built the agent sys tem platform right in the simulation of the grid. This platform is created ba sed on the id ea of integ rati ng the communication FMU into the power simulation environm ent. Figure 7 presents the agent designed in Matlab/Sim ulink . More details of the applicati on of the consensus algorithm could be foun d in [12]. The m ain idea of the pr ogress in each agent is that it gets the instant measu rement fro m the s ystem and returns the mean value to the controller. The progress is calculated in an iterativ e w ay. Suppose that N k den otes th e s et of the agent has the comm unication with agent k , a denotes the matrix calculated followi ng the Me tro polis rule [13], j i f denotes the v alues of calculated frequency at it erator i of agent j . The pseudo code of the process insi de agent k is describe d as follow ing: 1) The iterato r i = 1 2) The agent receives the instantane ous value of the frequency at output of the corresp onding DG k i meas f f = 3) The a gent sends the signa ls including the current iterator and agent value t o the neighbors 4) The agent r eceives the sign als from the neig hbors 5) If the agent receives signal at iteration i from every single ag ent in its neigh bor set then i = i + 1 1 * * k k kk k kj j i i i j N f a f a f + ∈ = + ∑ Else m ove to Step 4 6) If i = n consensus then m ove to Ste p 6 Else m ove to Step 3 7) Send the agen t value to the corresp onding c ontrolle r Move to Ste p 1 Figure 7. An agent i n Matlab/Simulink The controller gets signals when the correspon ding agent reaches the consensus or average value at Ste p 7. This means that the frequen cy at th e input of the controlle r is updated a ft er n consensu s iterators. The d elay in transmi ssion, therefore, depen ds on not only the latency when sending a nd receiving data between agents but also th e proces sing tim e inside agents. The frequency of the system is demonstrated in Figure 8. In the test case w ithout network latency or with the high quality network (Network 1 and Network 2 ), after a sh ort variation due to the disturbanc e in the microgrid, the fre quency was restored c orrectl y. Nevertheless, in Network 3 sy stem, the frequency was no longer stable. In this case, the d urati on of transmittin g data is mu ch more than the other cases and therefore it increases significantly the time of co nsens us process in m ulti-agen t s ystem . As a consequence, the Proportional- Integ ral (PI) control in the secondary level w orks improperly issuing w rong or late decision and the frequen cy could not r eturn to th e static r ated value. Therefore the cont rol system needs to be adjusted to fulf ill the stabil ity requir ement of the sys tem. In particula r, w e tu ned parameters of the PI block by increasing the Integral factor K i in E quati on 2. The frequency of the modifie d system is shown in Figure 9. Although the transient process takes longer tim e, the fluctuati on of the frequency is eliminated in all cases. This demonstrates the significan t effect of commu nication ne twork on the perfo rmance of the control system in m icrogrids. The optimal setting of PI co ntr oller therefore should take into account the d elay from the comm unication and should be robust in case of l atency m odifications. (a) (b) (c) (d) Figure 8. The fre quency whe n tripping load 2 a) without communi cation networ k, b) with latency test case 1, c) w ith latency test case 2, d) with la tency test case 3 (a) (b) Figure 9. The fre quency of the modifi ed control sy stem when tripping load 2 a) with latency tes t case 2, b) w ith latency test case 3 V. CONCLU SION This paper provided a holist ic approach to validate the control system in microgrids with the consideration of the commun ication netw ork. The comm unication m odel is developed in the FMI stan dard an d int egrated into the simu lation o f the microgrid. T he two domains of pow er and commun ication could be studied simul taneously in the same simu lator platform . A test-case of microgrid with five inverter-based DG s experimen ted in b oth cont rol ap proaches: centra lization and distributi on. A multi-agent platform is also built in Matlab/Sim ulink to implement the consensus algorithm in t he distribute d strategy . T he results show that the operati on of a microgrid is influenced by the pe rforman ce of comm unication netw ork. The evalu ation o f the stabili ty of the control system hence nee ds the consi derati on of the data tr ansm ission. The meth od used in this paper takes the advantages of FMI in the in teroperabil ity and r eusabi lity of the model. It could be used to study in m any cases of the c ontrol system with various conditions of the communication netw ork. I n the future research, the designed FMU could be extend ed to take into account the impac t o f cyber- security and packet loss to the system or the ev aluation the s ensitivity of Power Hardware-in- the-loop t esting w ith comm unicati on delay . A CKNOWLEDGMENT This work is supported by the H2020 Erigrid project, Grant Agreem ent No. 654113, w ith partial financial su pport from Vietnamese g overnm ent. The pa rticipati on of G2Elab and CEA-INES is also partially supported by the Carn ot I nst itute “Energies du Futur” under the PPInter op II pr oject (w ww .energiesdufutur .eu). R EFERENCES [1] N. Hatziargyriou, Microgrids: Ar chitectures a nd Control . 201 4. [2] J. Baill ieul and P. J . Antsaklis, “Co ntrol an d communication challenges in networked re al-time sy stems,” Proc. IEEE , vol. 95, no. 1, pp. 9 –28, 2007. [3] V. H. Nguyen, Q. T . Tran, and Y. Be sanger, “SCADA as a service approach for interoper ability of micro-gr id platforms,” Sustain. Energy Grids Netw ., vol. 8, pp. 26–36, De c. 2016. [4] K. Mets, J. A. Ojea, and C. Develder, “Combining p ow er and communication n etwo rk simulati on for cost-effective smart grid analysis,” IEEE Comm un. Surv. Tutorials , vo l. 31, no. 3, pp. 1–25, 2014. [5] T. L. Nguye n, Q. Tr an, R. Caire , C. G avriluta, an d V. H. Nguyen, “Agent Based Distributed Control o f Islanded Microgrid – Real-Time Cyber-P hysical Implementation,” Proceeding of the IEEE PES International Confer ence ISGT Europe 20 17 , Torino, I talia, 2017.. [6] S. C. M uelle r et al , “Interfacing Power S yste m a nd I CT Simu lators: Challenge s, State-of-the-Art, and C ase Studies,” IE EE Trans. Smart Grid , vol. PP, no. 99, pp. 1–1, 2016. [7] V. H. Nguyen et al , "Using Power-Hardware- in-the-loop Experiments together with Co-sim ulation in a holistic approach for cyber-physical energy system validation", Proceeding of the IEEE PES International Conference ISGT E urope 2017 , Tor ino, Italia, 2017. [8] S. I . S. C . (SISC), “IEEE Standard for Modeling and Simulation High Level Architecture (H LA ) - Framewor k an d Rules,” IEEE Std. 1516- 2000 . p. i-22, 200 0. [9] J. Dede, K. Kuladinithi, A. F örster, O. Nannen, a nd S. Lehnhoff, “OMNeT ++ an d mosaik: Enabling Si mulation of Smart Gri d Communications,” CoRR , vol. abs/ 1509.0, pp. 1–4, 2015. [10] J. Eker, J. W. Janneck, E. A. L ee, J. L iu, X. Liu, J. L udvig, S. Neuendorffe r, S. Sachs, and Y. Xiong, “Taming heterogeneity - T he ptolemy approac h,” Proc. IEEE , vol . 91, no. 1, pp. 127–143, 2003. [11] S. D. J. McArthur e t al , “Multi-Agent Systems for Power Engineering Applications-Part I : C oncepts, A pproaches, a nd Te chnical Chall enges,” IEEE Trans. Power Syst. , vol. 22, no. 4, p p. 1743–1752, 2007. [12] R. Olfati-Saber, J. A. Fax, and R. M . Murray, “Consensus and cooperation in networked multi-agent systems,” Proc. IEEE , vol. 95, no. 1, pp. 215–233, 20 07. [13] A. H. Sayed, “Adaptive networks,” Proc. IEEE , v ol. 102, no. 4, pp. 460–497, 2014.
Original Paper
Loading high-quality paper...
Comments & Academic Discussion
Loading comments...
Leave a Comment