Validating grid‐forming capabilities of hybrid power park technologies in future OFTO networks
2020; Institution of Engineering and Technology; Volume: 14; Issue: 19 Linguagem: Inglês
10.1049/iet-rpg.2020.0732
ISSN1752-1424
AutoresAdam Dyśko, Richard Ierna, Mengran Yu, Agustí Egea‐Àlvarez, Andreas Avras, Li Can, Mark Horley, Campbell Booth, Helge Urdal,
Tópico(s)Islanding Detection in Power Systems
ResumoIET Renewable Power GenerationVolume 14, Issue 19 p. 3927-3935 Wind and Solar Workshop 2019Free Access Validating grid-forming capabilities of hybrid power park technologies in future OFTO networks Adam Dyśko, Corresponding Author Adam Dyśko a.dysko@strath.ac.uk orcid.org/0000-0002-3658-7566 EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this authorRichard Ierna, Richard Ierna Rivermere Systems Ltd, Chesnut Field, CV21 2PD Rugby, UKSearch for more papers by this authorMengran Yu, Mengran Yu National Grid ESO, National Grid House, Warwick Technology Park, Warwick, UKSearch for more papers by this authorAgustí Egea-Àlvarez, Agustí Egea-Àlvarez EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this authorAndreas Avras, Andreas Avras University of Strathclyde Power Networks Demonstration Centre, 62 Napier Rd, G68 0EF Cumbernauld, UKSearch for more papers by this authorCan Li, Can Li National Grid ESO, National Grid House, Warwick Technology Park, Warwick, UKSearch for more papers by this authorMark Horley, Mark Horley National Grid ESO, National Grid House, Warwick Technology Park, Warwick, UKSearch for more papers by this authorCampbell Booth, Campbell Booth EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this authorHelge Urdal, Helge Urdal EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this author Adam Dyśko, Corresponding Author Adam Dyśko a.dysko@strath.ac.uk orcid.org/0000-0002-3658-7566 EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this authorRichard Ierna, Richard Ierna Rivermere Systems Ltd, Chesnut Field, CV21 2PD Rugby, UKSearch for more papers by this authorMengran Yu, Mengran Yu National Grid ESO, National Grid House, Warwick Technology Park, Warwick, UKSearch for more papers by this authorAgustí Egea-Àlvarez, Agustí Egea-Àlvarez EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this authorAndreas Avras, Andreas Avras University of Strathclyde Power Networks Demonstration Centre, 62 Napier Rd, G68 0EF Cumbernauld, UKSearch for more papers by this authorCan Li, Can Li National Grid ESO, National Grid House, Warwick Technology Park, Warwick, UKSearch for more papers by this authorMark Horley, Mark Horley National Grid ESO, National Grid House, Warwick Technology Park, Warwick, UKSearch for more papers by this authorCampbell Booth, Campbell Booth EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this authorHelge Urdal, Helge Urdal EEE Department, University of Strathclyde, 204 George St, G1 1XW Glasgow, UKSearch for more papers by this author First published: 16 February 2021 https://doi.org/10.1049/iet-rpg.2020.0732Citations: 1AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract In recent years, there has been considerable interest in convertor-based generating solutions which to a greater or lesser extent mimic the behaviour of synchronous machines, thus overcoming many of the disadvantages of the existing technologies which are potentially destabilising at high penetration. Such solutions are frequently referred to as grid-forming convertors (GFCs). This study focuses on the application of GFC technologies in offshore windfarms, where installation, maintenance and/or modification of any offshore equipment is very expensive and carries greater commercial risks, requiring extensive testing and confidence building prior to deployment in real applications. This is time consuming and particularly significant for GB and where there are large quantities of offshore generation. Onshore solutions to stability are therefore desirable for off-shore transmission owners (OFTOs), especially, if they could be applied by retrofitting to existing conventional converter plant. Consequently, this study proposes and investigates the performance of hybrid solutions for offshore networks where the conventional STATCOM onshore unit is replaced by alternative options such as synchronous compensator and virtual synchronous machine converter of similar (or appropriate) rating with the aim of achieving grid-forming capability. A laboratory-scale implementation of the proposed control algorithm is also presented with selected validation test results. 1 Introduction This paper reports on the outcomes of the National Grid's 'Hybrid Grid-Forming Convertor' NIA (Network Innovation Allowance) project, which utilised a virtual synchronous machine (VSM) algorithm, also referred to as grid-forming converter (GFC) in this paper. The intention of the work was to improve the understanding of the implications of GFC proposals addressed through GC0100 Option 1 [1] and subsequently the VSM Expert Group [2]. The main objectives of the projects can be summarised as follows: (i) To design and test a VSM algorithm in line with general GFC/VSM principals such as GC0100 option 1 [1]; (ii) To establish which plant control principals, parameters and tests are particularly relevant to grid stability; (iii) To understand how grid-forming performance affects one of the possible convertor designs and strategies which might mitigate any negative effects; (iv) To establish whether it is possible to provide grid-forming performance from hybrid solutions (e.g. STATCOMs) where not all of the converters are grid forming. Table 1 shows a matrix of future anticipated GB transmission system, convertor growth inhibiters in the columns and the potential counter measures in the rows. The cells which intersect the columns and rows show which counter measures are capable of resolving the various inhibiters. It can be seen from Table 1 that only three counter measures are believed to be holistic, potentially solving all or most of the anticipated inhibiters, either on their own or in combination. This does not mean that the other counter measures investigated are not useful but would need to be combined with other solutions which uniquely solve other areas. Table 1. Future system inhibiters and counter measures Solution Estimated cost RoCoF Sync torque/power (voltage stability/Ref.) Prevent voltage collapse Prevent sub-Sync. Osc./SG compatible Hi Freq. stability RMS modelling Fault level Post fault over volts Harmonic and imbalance System level maturity Key No: - Doesn't Resolve Issue P - Potential I - Improves yes - Resolves Issue constrain asynchronous generation high I yes yes yes yes yes yes yes yes proven these technologies are or have the potential to be Grid Forming / Option 1 synchronous compensation or more Sync. Gen. at lower load high I yes yes yes yes yes yes yes yes proven VSM medium yes yes yes yes yes yes yes yes P modelled VSM0H low no yes yes no P P P yes P modelled has the potential to contribute but relies on the above Solutions synthetic inertia medium yes no no P no no no no no modelled other projects low yes P yes no no no P P no theoretical Preliminary results of the various contributions of this project (and a second associated NIA project) have been presented earlier in the form of five conference papers presented at the Wind Integration Workshop in October 2019 in Dublin [3-7]. Fig. 1 shows the overall block diagram of the controllers implemented within these two GFC focused projects. The implementation of the controllers and associated hardware differ slightly as each project focused on different aspects of the design but both are similar implementations and are discussed in the relevant papers as indicated in Fig. 1. Although not imperative, it is suggested that the readers familiarise themselves with the papers [3-7] to gain the fuller appreciation of the context and the technical details of this work. Fig. 1Open in figure viewerPowerPoint Simplified block diagram of potential VSM implementation From Fig. 1 we can see the converter design largely consists of six major blocks: Dispatcher and governor. VSM (inertia emulation and damping, dynamic braking, voltage control and power limiter). Impedance reducer. Vector current limiter. Harmonic and imbalance management. Convertor output stage and power electronics. In paper [5], the VSM model in combination with various wind farm models and network elements (transformers, lines, STATCOMs etc.) were utilised to build a realistic off-shore transmission owner (OFTO) network model, and test the proposed hybrid solutions against standalone GFC and synchronous machine (SM). The key value of this paper consists in the demonstration that the modified inverter control strategy can enhance the capabilities of the typical offshore windfarm installation by facilitating certain degree of grid-forming capability which can be achieved with existing STATCOM hardware with an addition of a battery unit. This solution is deemed both feasible as well as economical, as it can be applied either in new or existing offshore windfarms. The main benefit is that it addresses to some extent the issue of gradually falling system inertia. The paper provides initial yet realistic evaluation of this capability through detailed dynamic simulation and hardware prototype testing. The paper is organised as follows. A summary of the modelling and the findings of the original paper [5] is provided in Sections 2 and 3. This is followed by the findings of the laboratory testing of the real-time GFC prototype implemented on the Triphase convertors and Real-Time Digital Simulator (RTDS) hardware platform at the Power Network Demonstration Centre (PNDC) (Section 4). The key observations and conclusions are drawn in Section 5. 2 Building a testable OFTO and windfarm model 2.1 Motivation In recent years, there has been considerable interest in GFC internationally. In Europe, the ENTSO-E working group recently published ENTSO-E TG HP Report [8], and NG ESO developed the GC0100 option 1 proposals [1]. Additionally, a number of researchers and manufacturers have proposed a variety of solutions (e.g. see [9]). Some of these solutions have achieved a maturity level that has seen them move from the laboratory to field trials (e.g. see [10]). Whilst there has been considerable progress in recent years, uncertainties still remain with regard to specific convertor functionality requirements and testing, as well as manufacturer readiness to offer such solutions for offshore windfarms, as the financial risks are substantial. Considerably greater testing and confidence is therefore required. In latest NG proposals, GC0100 option 1 has been replaced with a more open and flexible arrangement allowing manufacturers and developers to offer differing GFC features and levels of services and NG to choose which it wishes to procure. Whilst it is hoped this will deliver quicker and cheaper GFC solutions to market, for the purpose of this project, GC0100 option 1 and its subsequent revisions, provide a better definition of a required objective. In addition, GC0100 was the dominant proposal at the time these projects were initiated, hence the frequent references to it. Offshore windfarms with an AC connection back to the mainland typically contain convertor equipment in the turbines located offshore and normally a STATCOM located in the onshore substation (in a minority of cases an SVC might be used instead). The STATCOM provides the voltage control at the point of connection (POC), i.e. connection to the mainland Grid. This work considers whether it is possible to leave the offshore equipment and converter control unchanged and provide the GFC capability for the offshore windfarm, just using the onshore convertor, i.e. near the POC. The effect of replacing the STATCOM with a synchronous compensator of similar rating to the GFC (used to replace the STACOM) has also been considered. In this paper, we refer to the mixed convertor solution as a hybrid grid-forming convertors (HGFC), as the equipment offshore is not grid forming but the plant onshore is. Offshore installations are of particular importance to GB where currently ∼8 GW proportion of its WTG population is located offshore and this figure is set to increase with most wind developments in England and Wales now occurring offshore. In addition to reducing the time to market by reducing the risk and testing required, such a solution has further potential benefit, being cheaper to install and maintain and yet still further benefits as it is retrofiTable and could potentially be used with a variety of technologies such as doubly-fed induction generator (DFIG's). 2.2 Typical topology of an offshore windfarm Fig. 2 shows the typical topology of an AC-connected offshore windfarm. Fig. 2Open in figure viewerPowerPoint Typical topology of an OFTO's network and associated offshore wind farm The main components of the example offshore windfarm presented here are: Three winding transformer 400/132/13 kV (211 and 120 MVA tertiary). STACOM and reactors (4 × 15 MVAr) and capacitors (3 × 15 MVAr) for voltage support at the POC. Compensation reactor (60 MVAr) for the cable. Harmonic filter (20 MVAr). Onshore (40 km) and offshore (50 km) 132 kV cables. Offshore compensation (if fitted – not used here). Two-winding transformer 132/33 kV (211 MVA). LV offshore collector grid (not modelled). Wind turbine generators (WTG) including convertors and transformers (modelled as a four-tap 211 MVA 33 kV to 690 V transformer and PowerFactory Static Generator rated at 210 MVA). For the purpose of modelling, the offshore WTGs are often aggregated into one single device, and this approach is also adopted here. Before considering whether a HGFC solution is workable, it was necessary to build a suitable offshore wind farm model and benchmark it by performing dynamic studies, subjecting it to a variety of scenarios including: Type A faults (140 ms 3 ph) and Type B faults (500 ms, and long voltage dips) Voltage steps, 1, 2 and 5% Frequency ramps 0.5 and 1 Hz/s Vector shifts ( 4.5 °, 9 ° and 18 °) Various other tests including, but not limited to, frequency sweeps, frequency perturbation, power limiter, islanding, and different combinations of equipment. Windfarm models were taken from the WECC and IEC types 3 and 4 standard models available in PowerFactory. These were then adapted to include the components for an OFTO network. The components general topology and parameter values for the power system components were taken from averaging typical values available from public sources such as the ETYS data [11]. This included the initial values of the STATCOM rating capacitors and reactors and so on. For the STATCOM dynamic model, a voltage droop controller with PI stabilising and PowerFactory Statgen power convertor was used. This was configured to provide voltage control at the POC, as required by the GB Grid Code. In contrast, the WEC and IEC turbine and power park controllers were setup to operate in constant PF/MVAr mode delivering approximately −20 MVAr's into the LV side of the 132/33 kV SGT, partially to offset the reactive power produced by the cable (the 132 kV winding typically absorbs 43 MVAr from the cable). The voltage at the 132 and 33 kV busbars was controlled through transformer tapping of the LV/MV side of the associated transformers. 3 Wind farm simulation and testing 3.1 Load flow tests To ensure OFTO network and windfarm had adequate tapping range on all transformers and sufficient reactive reserves, 16 combinations of active and reactive power, POC voltage and fault level were studied. This was done to ensure there were reserves both to maintain control and deliver the required reactive response for the entire operating range and for all operating conditions. The 16 conditions were derived by creating all possible combinations of the following: Max and Min (400 kV ±5%) volts at the POC between the OFTO and on-shore transmission system. Max reactive power import and export at the POC (0.95 lead and 0.95 lag). Max and Min fault level of ∼4500 and 400 MVA, respectively (this is controlled by series reactors placed between the POC and controlled infinite bus). Max and Min active power 200 MW (max), and 100 MW (min at 0.95 lead) and 40 MW (min at 0.95 lag) – from CC.6.3.2 in the GB Grid Code [12]. 3.2 Vector relationship between voltages Fig. 3 shows the relationship between the converter voltage (E), voltage at the POC (Vpoc) and the impedance between them, which is typically dominated by the filter and transformer reactance, denoted here as Xf and Xt (all quantities are in pu). Fig. 3Open in figure viewerPowerPoint Vector relationship between voltages If we assume that the resistance and other impedances are not significant and can be ignored, the vectors E and Vpoc form a triangle where the third voltage is the voltage across the impedance Xf + Xt and the angle between them is the operating angle of the converter (where Xf is the convertors internal filter and Xt is the transformer reactance). In the case of the algorithm developed here, E is the PWM voltage at the transistors as indicated in [13]. This vector diagram is superimposed onto the operating chart and we can see that the length of the vector Vpoc is roughly 1/(Xf + Xt) if we ignore the other impedance effects. Although not drawn to scale (Vpoc is normally longer) we see that changes in power are dominated by changes in delta and changes in reactive power by E and Vpoc. Converters typically have a lower coupling impedance to the network (Xf + Xt) than SM's and delta is therefore smaller, from the diagram we can see that this has two significant effects for GFC's. First, the GFC's are potentially considerably more responsive to vector shifts than SM's. Note however, SM's do have damper windings which provide some additional contribution to system events and there is no equivalent contribution in the algorithms presented in these papers. Second, a GFC or SM will lose synchronism when delta reaches 90° (the UEL for the SM) but in the case of the convertor this is outside the MVA limit operating circle provided the impedance is <1 pu. Whilst GFC's close to a fault or loss of power infeed and subsequent vector shift, may perform less favourably to SM's (i.e. if the current or active power limit is activated), those at intermediate distance where the overall impedance is lower could be more responsive potentially providing increased support. 3.3 System studies Fig. 4 shows the model used to study the OFTO network and associated offshore wind farm which consists of an 'almost' infinite bus bar controlled by a test converter whose output voltage and frequency can be modified to perform a variety of tests. Fig. 4Open in figure viewerPowerPoint Infinite bus model with all OFTO/WTG models tested Attached to this are a variety of models including a GFC and SM connected to the bus via a 12% impedance transformer for bench-marking performance and seven OFTO networks configured as WECC 3 and 4 and IEC 3 and 4 all with STATCOM's, WECC 4 with GFC (this is the HGFC solution), WECC 4 with SC, WECC 4 with 50% SC and 50% STATCOM. In all cases, the dynamics voltage support elements (i.e. the STATCOM, GFC or SC) of OFTO network were sized to 67 MVA with 3 × 15 MVA of capacitors and 4 × 15 MVA of reactors (in the case of the 50:50 STATCOM/SC system both were sized at 33.5 MVA). Whilst in practice it is accepted different proportions might be used for economic reasons or to avoid operational limits, they have been scaled to 67 MVA here to allow comparison of performance against rating and in the case of GFC, to allow 33% headroom at 0.95 power factor. From the studies performed it became apparent that whilst all the studies were useful in demonstrating different performance characteristics and in many cases compliance with existing grid codes, two or three studies/tests, in particular provide indication of grid-forming capability, namely: Vector shift Frequency ramp Frequency perturbation In the case of the vector shift study, the angle change is applied to the bus bar (although on site for compliance testing it might equally be applied to the GFC). From the graph in Fig. 5 we can see that the types 3 and 4 WECC and IEC wind farms show typical 'grid following' behaviour and provide no significant power injection (the four blue and green flat lines in the left hand graph) to resist the vector shift but the GFC (black left graph), HGFC (black right graph), SC (blue 100%, cyan 50% right graph) and SG (red left and right graph) provide varying degrees of response. Fig. 5Open in figure viewerPowerPoint Responses of differing control solutions to a 9 ° vector shift The quantity of response is proportional to the increase in power for the applied angle change (which was the same for each generator). The level of response to the change in vector is largely dictated by the connecting impedance between the GFC/SC voltage source (VS) and the POC voltage. The difference in frequency of the power swing, between the SC and the directly connected VSM and SG is due to the inertia which is set to 1.8 s in the SC and 6.25 s for the VSM and SG. If the SC inertia is increased to 6.25 s frequency of oscillation aligns with the VSM and SG. The inertia of the VSM in the HGFC is set considerably higher and is of the order of 10 s with damping parameter also altered although the algorithm is the same. Consequently, defining the response in terms of the inertia is not as straightforward as defining the overall response in terms of power produced for a given rate of change of frequency (RoCoF). The frequency ramp studies show how much equivalent inertial response is provided by each configuration of wind farm and OFTO. It is necessary to take some care when interpreting this result as the response curve shape is affected by contributions from the inertia, damping and droop governor, if active. Again, as can be seen in Fig. 6 there is no significant response from the standard WECC and IEC models (colours and graph format are as Fig. 5). Fig. 6Open in figure viewerPowerPoint Responses of differing control solutions to a frequency ramp The frequency perturbation test is detailed in [14] and not displayed here but we make readers aware of it because it is particularly useful for determining phase shifts and bandwidths of the various control system elements, e.g. where the governor response ends and the inertial response starts. It is particularly interesting to note that in both the vector shift and frequency ramp the HGFC solution outperforms the SC. Furthermore, fitting a solution where 50% of the rating is provided by a SC and the other by traditional STATCOM worsens the response to vector shifts although it may be more beneficial for traditional problems such as voltage support. 3.4 Critical impedances It is clear from the vector diagram, displayed in the previous section (Fig. 3), the impedances between the major voltage and power sources dominate the response to the vector shift studies. The lower the impedance the greater the response. The simplified diagram in Fig. 7 shows the equivalent circuit diagram with most significant impedances between the key components responsible for a HGFC. Fig. 7Open in figure viewerPowerPoint Simplified impedances for an offshore wind farm The performance level is determined by the amplitude of the response and whilst the HGFC model initially used, did not perform quite as well as reference SM or GFC, changing its filter or connecting transformer impedance, either in practice or artificially (by modifying the control system/software) improves performance. The following paragraphs discuss the effect of reducing the physical impedance but the paper [6] describes an algorithm which was applied to an RMS model and has the same effect as reducing the impedance. After publication of [5] which included the results shown in Fig. 6, it was noted that the VSM convertor's response to a frequency ramp was significantly higher than equivalent SM. Subsequent further investigation has revealed marked differences in the performance of the differing stabilising algorithms discussed in [4, 13] which provide damping to the 1 / 2 H s term. Whilst these algorithms can be tuned to give a similar response levels to a vector shift, their response to a frequency ramp varies significantly. The algorithm used in this paper is used and described in [13]. Unlike other algorithms it uses the rate of change of the internal frequency state variable rather than the actual measured system frequency. In [13], it was indicated that this could be a preferential because it is not susceptible to noise in the measured frequency signal. However, here we see a disadvantage as this stabilising signal requires the internal frequency to change before damping occurs and this delay results in greater difference in response, whereas other algorithms respond more promptly to external frequency changes. The response in Fig. 6 has demonstrated the importance of extensive testing with a wide variety of scenarios and the need for further work to optimise this particular aspect of the design, the stabilising algorithm, for all scenarios. The graph in Fig. 8 shows the HGFC response to the same test but here it is assumed that a converter solution is provided which utilises only 1% impedance in its output filter and the tertiary winding of the transformer is uprated from 120 to 150 MVA which effectively reduces its impedance by the same proportion. This design would rely heavily on the three-winding transformer to provide additional decoupling from the network and the current limiter and convertor protection would need careful consideration with such a low filter impedance. However, practicalities aside it can be seen, the response is improved outperforming the SG for 4.5 ° vector shift. Fig. 8Open in figure viewerPowerPoint Vector shift response with filter impedance reduced from 10 to 1% Likewise, because the HGFC inertia is typically programmed in software and the SC inertia, unless a flywheel is fitted, has a lower value (H = 1.8 s was used) than a synchronous generator, HGFC outperforms the SC in terms of inertia. It is evident in the results both in terms of the magnitude of power injected for the vector shift and RoCoF study results the HGFC performance is better. The 50:50 SC/STATCOM solution, as might be expected, provides less performance but it is not proportional as the tertiary winding impedance is still set to 120 MVA and not reduced. In addition to the potential performance benefits of HGFC, if fitted with batteries for energy storage, such systems can store or provide energy when not being used for grid-forming control. 3.5 Equivalent synchronous compensator performance The parameters used for the synchronous compensators were the same as the base case synchronous generator with the inertia and MVA being the only exceptions. The inertia and MVA of the synchronous compensator were set to 6.25 s and 211 MVA to match the synchronous generator and the system was retested. There was a considerable reduction in the frequency of the power oscillation and significant increase in the power produced for vector shifts. However, even with 211MVA rating for the synchronous compensators, the power swing for a 4.5° angle change was only about 50% of the HGFC or SM responses. The difference can be explained by increasing the rating of the three-winding transformer or reducing its impedance which significantly improves performance. Both increasing rating or reducing the impedance effectively reduces the overall impedance. The 211 MVA synchronous compensator is still outperformed by the HGFC for a 9° change, even though the HGFC limits its output power. To achieve a significant performance improvement, both synchronous compensator and transformer impedance have to be reduced. 4 Hardware implementation and laboratory testing The final stage of the GFC project undertaken by the University of Strathclyde involved power hardware-in-the loop (PHIL) implementation of the selected models and scenarios. To this end, an experimental testbed at the PNDC was setup. The main elements/tasks of this development included: A real-time RTDS model of a DFIG-based off-shore wind farm in the RSCAD environment, with a real-time controllable VS to represent 13 kV STATCOM bus allowing the connection of real VSM convertor. This was achieved using Triphase power converter platform configured as a controllable VS. A second Triphase convertor was programmed with VSM with Zero Inertia (VSM0H) and a conventional direct–quadrature current injection (DQCI) control logic to represent the STATCOM. The VSM0H control was extended to include an inertial element, to achieve a grid-forming behaviour, turning it into a VSM convertor. Each element was first tested on its own, then the two Triphase convertors were connected to the RTDS windfarm model and the complete system was tested with a specific subset of test scenarios. High-level arrangement, utilised in these tests, is presented in Fig. 9. More detailed description of the testbed is included in Sectio
Referência(s)