Integration of a mean-torque diesel engine model into a hardware-in-the-loop shipboard network simulation using lambda-tuning

Transcription

1 Integration of a mean-torque dieel engine model into a hardware-in-the-loop hipboard network imulation uing lambda-tuning A. J. Rocoe*, I. M. Elder*, J. E. Hill, G. M. Burt* *Intitute for Energy and Environment, Univerity of Strathclyde, Glagow, G XW Andrew.Rocoe@eee.trath.ac.uk Roll-Royce plc, Derby, UK Keyword: marine electrical ytem/technologie, power ytem control, power ytem tability, power ytem reliability, imulation. Abtract Thi paper decribe the creation of a hardware-in-the-loop (HIL) environment for ue in evaluating network architecture, control concept and equipment for ue within marine electrical ytem. The environment allow a caled hardware network to be connected to a imulation of a multi-megawatt marine dieel prime-mover, coupled via a ynchronou generator. Thi allow All-Electric marine cenario to be invetigated without large-cale hardware trial. The method of cloing the loop between imulation and hardware i decribed, with particular reference to the control of the laboratory ynchronou machine which repreent the imulated generator(). The fidelity of the HIL imulation i progreively improved in thi paper. Firtly a fater and more powerful field drive i implemented to improve voltage tracking. Secondly the phae tracking i improved by uing two neted PIDA (proportional integral derivative acceleration) controller for torque control, tuned uing lambda-tuning. The HIL environment i teted uing a cenario involving a large contant-power load tep. Thi both provide a very evere tet of the HIL environment, and alo reveal the potentially advere effect of contant-power load within marine power ytem. Introduction Recent year have een increaing interet in the naval and commercial hipping indutrie in concept uch a the All-Electric Ship, in which an electric propulion ytem i combined with the power network erving other electrical load. While thee technologie offer ignificant economic and environmental benefit over traditional architecture, hipbuilder, owner and regulatory authoritie mut be able to determine that the reulting hip will

2 meet the exiting and forthcoming tandard of reliability and afety, a well a delivering the promied benefit in maintainability and reduced fuel ue. One approach that ha been adopted in meeting thi challenge i the contruction of a full-cale hardware demontrator []. While thi approach ha obviou advantage, it alo ha everal drawback, including high cot, limited flexibility, and the inability to evaluate equipment at the deign tage. Alternatively, evaluation of deign through imulation can be adopted, which remove ome of the cot and inflexibility of an all-hardware approach, but require ufficiently accurate model of all component which are to be included. Where equipment or phenomena are poorly undertood or poorly documented, thi can be a ignificant barrier. A third approach i to adopt a reduced-cale Hardware-in-the-Loop (HIL) approach [2] [3] [4] [5] [6] in which phyical machine, drive, cable, etc. are ued to repreent certain part of the ytem, while other are repreented uing imulation model. Key advantage of thi approach, which i decribed in more detail in [7], include the ability to evaluate the actual performance of phyical drive, controller, etc., a well a the option to cale the output from large imulated device uch a prime mover in a way which would not be poible in hardware. Thi paper decribe the adaption and integration of a model of a large multi-megawatt dieel engine, a developed by the manufacturer, into a kilowatt-cale marine network hardware demontrator. To improve the fidelity of the HIL experiment, pre-exiting laboratory HIL capability decribed in [7] i augmented; firtly with more powerful field control, and econdly uing a dual neted control loop uing PIDA controller, which are tuned in turn uing the Iambda tuning approach. To demontrate the effectivene of the approach, meaurement of the HIL network behaviour are compared with equivalent imulation. The tet cenario ued i a large contant-power load tep. It i hown that good agreement in both electrical and mechanical behaviour can achieved in the HIL imulation. 2 Choice of tet cenario: contant power load tep For marine veel, propeller characteritic are defined from a erie of tank tet. Perhap the mot famou, and bet documented, of thee are the Wageningen (or Troot) B erie. Thee give the propeller torque a a function of the propeller peed and the axial peed of the propeller through the water [8]. For implified, quai-teady-tate,

3 calculation, the propeller characteritic (torque a a function of propeller peed) can be approximated a a quarelaw in the firt quadrant, i.e. the propeller puhing the veel through the water in the forward direction. At low veel peed, it i uual to control the propeller at a contant peed via a propulion motor peed-control loop. However, at higher veel peed, the repone of thi type of control cheme i unatifactory. Small change in veel peed, due to a dynamic change in veel reitance, lead to large change in the torque (and electrical power demand) demanded via the propulion motor. Thi in turn lead to change in electrical ytem frequency, and i particularly problematic at higher ea tate. The electrical generator are governed to maintain contant frequency, and therefore the engine fuel demand alo fluctuate. Overall, thi caue extra tre and wear on the prime mover. Therefore, a contant-power propulion mode i ued at higher veel peed. Change in the veel reitance are then manifet through change in the veel peed which are moothed through the veel inertia, which i uually large. In modern veel with an All-Electric deign, uch contant-power drive ytem are relatively eay to implement, uing power electronic. However, the impact that a tep change in load applied through a contant-power drive can have on a marine power ytem i large. A an illutration, Fig. and Fig. 2 how the reulting generator torque and frequency (peed) from four different load tep of 0.5pu (per-unit): Contant power of 0.5 pu Contant torque of 0.5 pu Torque proportional to electrical frequency Torque proportional to electrical frequency quared Clearly, the contant-power load tep caue the bigget perturbation to the power ytem. Thi i becaue the torque at the generator haft i inverely dependent upon engine peed, for a given power, and therefore rie a the frequency (engine peed) fall, exacerbating the frequency dip. Sudden introduction of load in contant torque mode, or where torque fall when peed (frequency) fall, provide much maller perturbation. In the following ection, the contant-power load tep i tudied, ince it repreent the mot intereting and challenging cenario, with ROCOF (rate of change of frequency) rate of up to 5 Hz/. In the cenario, a reitive 0.5pu (per-unit) load tep i applied to an unloaded engine/generator which i initially at it nominal peed (pu,

4 500rpm, giving 50Hz). The input to the engine model are the peed reference (et to an un-drooped value of pu for a 50Hz target), and the generator torque. The per-unit torque i calculated by the per-unit load power divided by the per-unit frequency. Two cae of thi cenario are preented and compared. In the firt cae ( Pure Simulation ), the engine model plu the electrical load are imulated. In thi cae, the electrical load i et to a perfect tep-function. In the econd cae, ( HIL ), the engine model i imulated, but the electrical load i applied in hardware uing a reitive loadbank, The loadbank i inenitive to frequency, and therefore, o long a the ytem voltage remain at pu, the reitive load behave a a contant-power load. In both cae, it i aumed that the generator maintain a contant pu voltage during the tep, due to effective AVR action. In reality, an actual AVR may truggle to maintain nominal voltage during the evere tet preented in thi tudy, in the ame way that the HIL hardware cannot reduce voltage excurion to exactly zero. However, the aumption of perfect regulation i ueful becaue it provide both the harhet tet of the dieel engine and the HIL hardware, and alo it allow imple frequency-inenitive reitive loadbank to be ued in hardware to provide the contant-power load tep. In the hardware (HIL) tet, the dieel engine model (which model a machine at the 4 MW cale) wa caled uch that it nominal pu power rating wa 37.4kW. Thu, the actual reitive load applied wa nominally a reitive 8.7kW (0.5 pu) at 400V (pu), 3-phae, 50Hz. The caling to the multi-mw model i achieved in thi cae by imply multiplying the meaured electrical power flow in hardware by the factor 4000/37.4, to derive the power (and hence torque via knowledge of the rotational peed) which would have been extracted from the full-cale dieel. 3 Architecture of imulation and HIL implementation 3. Simulation of dieel engine The mean-torque dieel engine model ued in thi paper i a ignificant piece of proprietary MATLAB Simulink code, upplied by a European manufacturer of marine reciprocating engine. The tructure of thi imulation model approximately follow that decribed in [9]. A impler model, developed pecifically for HIL application, i preented in [0]. The imulation model ued in thi paper include a governor controller, the coupling dynamic of the engine to the generator haft, and alo the inertia of, and torque applied to, the generator. The imulation model doe not include

5 magnetic phenomena or the electrical behaviour of the generator; the electrical behaviour i determined by the phyical ynchronou generator. The dieel model wa originally developed in the continuou mode within Simulink, and o thi ha been converted for ue with dicrete imulation (and HIL application) by replacing all filter, differential and integral with digital equivalent. For thi paper, a time tep of 2m i ued, ince it matche that ued in the HIL application. 3.2 Hardware in the loop implementation The dieel engine model can be placed within a HIL environment (Fig. 3). In thi cae, the aim i to control a real 80kVA ynchronou motor-generator o that it behave with the ame peed/torque and inertial repone a the model of the dieel engine and coupled generator. Thi allow an entire laboratory network of load, interconnector, breaker, and maller generator to be connected. The power hardware network can thu virtually driven by the model of the dieel/generator, and the model of the dieel/generator become loaded by the network. The cloure of thi feedback loop create a HIL environment. It hould be noted that while thi paper refer pecifically to cenario where the imulated part conit of a dieel reciprocating engine, the hardware-in-theloop deign, functionality and fidelity i applicable to many other type of imulated prime mover, machine or electrical network. When entire electrical network are required to be imulated, an appropriate electrical real-time imulator i required [7]. The 80kVA motor-generator i driven by a fat reponding DC motor coupled to a thyritor drive. The eential detail of thi implementation are decribed in [7]. However, in thi cae the imulation of the dieel engine can be executed on the ame computer a the HIL control (the imulator and controller in Fig. 3). Alo, the dieel engine model return a peed output a a repone to a torque input, and thi peed output mut be integrated to provide the phae of V * N. Thi integration include an arbitrary (contant) phae offet value which i eentially a free variable. For thee two reaon, o long a the ampled value of V N, I N are meaured carefully with matched antialiaing filter and made coherently (or proceed to be coherent a in [7]), then there i eentially zero loop delay. Thi ignificantly implifie the implementation compared to previou work in [7]. 4 Incremental improvement to HIL fidelity The 0.5pu contant-power load tep wa firt applied to the HIL environment uing identical field and torque control for the 80kVA generator a were originally ued in [7]. Specifically, the field control ued a rotary exciter

6 and the torque control ued a control algorithm hown in Fig. 2 of [7]. It wa hown in [7] that although the rotary exciter had ufficient power to maintain teady-tate operation, it wa lacking in ufficient overhead voltage capability to enable the toughet HIL cenario to be tracked accurately. Thi i borne out in the firt et of reult, hown in Fig. 4, and by the original line on Fig. to Fig. 3. In particular, the poor voltage tracking (Fig. ) lead to a drop in power flow relative to the pure imulation 4. Improving the accuracy of the voltage tracking at the hared node To improve the voltage tracking, the rotary exciter for the 80kVA generator wa removed from the HIL hardware, and replaced with a large olid-tate DC power upply with a high witching frequency and fat repone. Thi ha adequate DC current capability to maintain teady-tate terminal voltage at the machine rating. More ignificantly, it ha a much higher available DC voltage than the rotary exciter. Thi enable the field current to be increaed much more rapidly during load change, allowing terminal voltage to be regulated much better. The particular DC upply ued during thi tudy i only capable of generating poitive voltage. For the cenario preented, thi i acceptable. For cenario involving tracking of voltage dip, a bidirectional DC ource capable of negative voltage output and revere power flow would be highly deirable, to allow forcible collape of the field. The field control contain a PID (proportional-integral-derivative) controller, and the control gain can be increaed uing the DC upply, due to the fater repone of the field uing thi hardware. Uing the improved field control lead to the dahed line on Fig. to Fig. 3. Notably, the voltage tracking error i much reduced, and therefore the power flow in the HIL experiment i much cloer to the pure imulation cae, than with the original etup. 4.2 Improving the accuracy of the phae tracking at the hared node In [7], the torque control to the DC motor (driving the 80kVA generator) wa derived uing a phae-locking control hown in Fig. 2 of [7]. Thi conited of a PID controller to control frequency, uing an additional low-pa filter on the differential control (only) to limit the effect of meaurement noie [], augmented with a imple proportional-only controller to enable tracking of phae. In addition, the value of the control parameter were obtained by hand tuning in the laboratory. The value of thee original parameter are hown in Table. To improve the torque control, the lambda tuning approach ha been taken, which i a variant of IMC (internal model control) tuning [2]. To enable thi, the parameter for the generator were meaured through a erie of

7 tet, uch a pin-down tet (to meaure inertia and friction), and tep-change in command torque (to meaure drive repone). Knowledge of thee parameter, combined with knowledge of the meaurement algorithm, allow the ytem to be modelled to the required level of accuracy. The model tructure i hown in Fig. 5. The right-hand loop of Fig. 5 i a conventional loop for controlling frequency to a given target. The left-hand loop augment the control ytem with the unconventional (but required) control of generator phae to achieve a given target. In the model, thi input can be regarded a a zero input for OLTF (open-loop tranfer function) tability analyi, or can be ued to imulate the effect of meaurement noie at different frequencie in a CLTF (cloed-loop tranfer function) analyi. To implify the control ytem, the right-hand frequency control loop could be opened during phae control, leaving only the left-hand loop active. However, thi mean that both the integration (/) tage H() the generator inertia, and R() the frequency-to-phae tranformation, would be preent in the OLTF. Thi mean that the phae lag of the OLTF would be 80 even at DC, becoming even more lagged at higher frequencie due to the action of low-pa filter and meaurement time. Stabiliing uch a loop preent a ignificant problem requiring large amount of differential gain. Thu, the control i eaier to tabilie if both the control loop are cacaded, ince the effect of the inertial lag can be reduced by cloing the frequency-control loop. The round-trip (command to meaured value) cloed-loop repone for the ytem, uing the original hand-tuned parameter (Table ) are hown in Fig. 6. The phae loop repone repreent the entire control ytem of Fig. 5, wherea the frequency loop repone repreent the inner frequency-control loop. The phae loop ha a bandwidth of about 2 Hz. The OLTF (open loop tranfer function) i not hown graphically, but the forecat gain margin i 5.5dB and the phae margin i 42. To improve the repone, lambda-tuning wa ued [2]. Target were et for the deired cloed-loop command-to-meaured-value repone of the frequency control loop, and then the phae control loop. The target are defined uch that the repone hould ideally behave a firt-order low-pa filter repone to command ignal. Firtly, the CLTF for the frequency-control loop: C C F F D D H HL () H a F HL

8 where a i the target firt-order repone time. Secondly, the repone for the phae-control loop: b P R Q F HL H D C HL H D C C Q R Q F HL H D C HL H D C C Q F F P F F P ' ' (2) where b i the target firt-order repone time. Auming that () can be atified, then (2) can be implified to: b P R a C R a C P P (3) () and (3) can be olved, yielding: a f fa a f Hfd Lfd d f H d f L H L C F ) (4) and b p pb b p ap p a C P 0 2 (5) Equation (4) and (5) define the parameter for the two controller C F and C P. Notably thee controller are unconventional PIDA controller, combined with low-pa filter element. The acceleration term (in 2 ) dramatically aid the tability, ince they counter the phae lag. However, the rik in a practical ytem i that they (and indeed even the differential control) introduce large amount of noie due to the differentiation tage. It i only poible to ue thee term, even with reduced magnitude, due to the good noie reduction of the meaurement algorithm [3] Setting the actual control in practice involved the following tep, recogniing that Fig. 5 i only an etimation of the actual ytem, and that the imple target () and (2) do not fully define the repone required in all cenario: Chooing repone time a and b

9 evaluating (4), (5) and then examining OLTF (tability), CLTF (command-to-output) and repone (command-to-meaured-value) plot uing MATLAB Trial of the choen etting in the hardware implementation, making mall change to the etting of a and b Reduction in the actual proportion of acceleration ( 2 ) control ued, from thoe uggeted in (4) and (5), to limit the repone to meaurement noie, a hown in table. Increaing the amount of integral gain in C F to improve the initial ettling to a new frequency, a hown in table. Repetition of trial of tep 2-4, in a range of cenario, until the bet behaviour i achieved. The final et of control parameter are hown below in Table, together with the original hand-tuned value. The reulting bode plot of the OLTF and the command-to-output CLTF are hown in Fig. 7 and Fig. 8. Forecat gain margin i 9dB and phae margin i 32. Indeed, the hardware ytem i fat reponding and on the limit of tability with thi configuration. It i certainly found that if the Ka acceleration term (in 2 ) are et to zero, the hardware i untable. Converely, if the Ka term are increaed to their theoretical value, noie become intolerable at the torque command output due to the increaed CLTF repone at high frequencie. The reulting repone (command to meaured value) are hown in Fig. 9. The phae loop bandwidth i increaed from about 2 Hz to about 6Hz. Fig. 9 alo how the predicted deviation from the target lambda-tuned repone, caued mainly by the ¼ reduction in the Ka gain to limit the effect of noie. The 0x boot to the integral gain of C F caue almot no viible change to the repone hown on thi figure. Uing the improved control, the HIL environment i now able to track the pure imulation cae much more accurately. Fig. 0 how the frequency profile, with the HIL reult now tracking the imulation reult much more cloely than before (Fig. 0 to Fig. 3). The ue of the lambda-tuning approach and the redeign of the torque control allow the peak phae tracking error to be reduced from 30 to 5. Of note in the preented cenario i the large frequency dip. Thi i due to the behaviour of the turbo-charger which take time to pin up, epecially when engine peed drop in repone to the load tep. The turbo preure (in

10 gauge bar) can be extracted from the HIL model in real time, and i hown in Fig. 4. Thi how excellent agreement between the pure imulation and the HIL reult. 4.3 Summary of incremental improvement The incremental improvement in the fidelity of the hardware-in-the-loop environment, from the tarting point of the baic control ued in [7], are ummaried in Table 2. The incremental improvement are firt the more powerful and fater-acting field control (ection 4.), followed by the improved phae tracking uing lambda-tuning (ection 4.2). The improvement in fidelity are alo hown in Fig. 0 to Fig Concluion A neted pair of PIDA (proportional-integral-derivative-acceleration) controller have been tuned uing lambda-tuning. Thee can be ued to tightly match the phae of a large ynchronou machine to the phae of a imulated machine. Thi allow a imulation of a multi-megawatt dieel engine to be interfaced with phyical hardware to create a hardware in the loop (HIL) imulation environment, uitable for ue with marine power ytem cenario. The PIDA controller allow the phyical generator to track the phae of the imulated prime-mover to within 0-5 for cenario with rate of change of frequency up to 5 Hz/. The phae tracking i much better than thi figure for le dynamic cenario. The phae tracking accuracy cannot be ignificantly improved from the preent performance, due to the phyical limitation of the drive ytem and machine inertia. However, ome further marginal improvement might be made by reducing the meaurement algorithm time (Fig. 5). Care would be needed, however, ince noie output from thee algorithm will degrade the control ignal, which make ue of unconventional control term in 2 to increae the control gain without becoming untable. For ue within a HIL environment, it i alo important that the field control for the ynchronou generator be able to rapidly increae or decreae the field current, to track the imulation. Thi require a field drive with ufficiently large voltage drive capability. Improved performance wa hown in thi paper by uing a more capable field drive. A bidirectional field drive would alo allow cenario uch a harp voltage dip to be modelled. Finally, the cenario of a tep in contant-power demand i of particular interet in that the reulting torque can be larger than expected if frequency drop appreciably. Since contant-power electrical load (e.g. olid-tate motor drive) are becoming more prevalent, uch cenario are important in any All-Electric Ship deign.

11 Acknowledgement The author would like to acknowledge the funding and upport provided by Roll-Royce plc, which enabled thi work to be undertaken. Reference [] Butcher, M.S., Mattick, D., and Cheong, W.J.: 'Informing the AC veru DC debate - the Electric Ship Technology Demontrator'. All Electric Ship (AES) Veraille, France, 2005 [2] Armtrong, M., Atkinon, D.J., Jack, A.G., and Turner, S.: 'Power ytem emulation uing a real time, 45 kw, virtual power ytem'. European Conference on Power Electronic and Application, 2005 [3] Langton, J., Suryanarayanan, S., Steurer, M., Andru, M., Woodruff, S., and Ribeiro, P.F.: 'Experience with the imulation of a notional all-electric hip integrated power ytem on a large-cale high-peed electromagnetic tranient imulator'. Power Engineering Society General Meeting, 8-22 June 2006 [4] Liu, Y., Steurer, M., and Ribeiro, P.: 'A novel approach to power quality aement: Real time hardwarein-the-loop tet bed', IEEE Tranaction on Power Delivery, 2005, 20, (2), pp [5] Qian, L., Liu, L., and Carte, D.A.: 'A Reconfigurable and Flexible Experimental Footprint for Control Validation in Power Electronic and Power Sytem Reearch'. Power Electronic Specialit Conference (PESC) Orlando, USA, 2007 [6] Woodruff, S., Boenig, H., Bogdan, F., Fike, T., Peteren, L., Sloderbeck, M., Snitchler, G., and Steurer, M.: 'Teting a 5 MW high-temperature uperconducting propulion motor'. Electric Ship Technologie Sympoium, July 2005, pp [7] Rocoe, A.J., Mackay, A., Burt, G.M., and McDonald, J.R.: 'Architecture of a network-in-the-loop environment for characterizing AC power ytem behavior', IEEE Tranaction on Indutrial Electronic, 200, 57, (4), pp [8] Hill, J.E., Kinon, A.S., Eraut, N.D., French, C., and Tumilty, R.M.: 'Motor Load Train Calculation for a Land Baed Demontrator for Marine Electrical Network (DMEN)'. IEE International Conference on Power Electronic, Machine and Drive (PEMD), Edinburgh, UK, 2004 [9] Kao, M.H., and Mokwa, J.J.: 'Turbocharged Dieel-Engine Modeling for Nonlinear Engine Control and State Etimation', Journal of Dynamic Sytem Meaurement and Control-Tranaction of the Ame, 995, 7, (), pp [0] Cooper., A.R., Morrow, D.J., and Chamber, K.D.R.: 'A Turbocharged Dieel Generator Set Model'. Univeritie Power Engineering Conference (UPEC) Glagow, UK, 2009 [] Kritianon, B., and Lennarton, B.: 'Robut tuning of PI and PID controller - Uing derivative action depite enor noie', IEEE Control Sytem Magazine, 2006, 26, (), pp [2] Lennarton, B., and Kritianon, B.: 'Evaluation and tuning of robut PID controller', IET Control Theory and Application, 2009, 3, (3), pp [3] Rocoe, A.J., Burt, G.M., and McDonald, J.R.: 'Frequency and fundamental ignal meaurement algorithm for ditributed control and protection application', IET Generation, Tranmiion & Ditribution 2009, 3, (5), pp

12 Fig. Generator torque due to 0.5pu load tep in four different mode Fig. 2 Generator frequency due to 0.5pu load tep in four different mode Key Phyical 3-phae cable connection Control ignal/data Meaurement of voltage or current Simulation boundary Nomenclature I G Vector of 3-phae current from 80kVA generator I N Vector of 3-phae current flowing into hardware V N Vector of 3-phae voltage at hared node in hardware * V N Vector of 3-phae voltage at hared node in imulation Local cloed-loop control to force hared node voltage M ~ 80 kva motor-generator Fixed load I G V N Shared node I N Laboratory power hardware network Real-Time Simulator & Controller Dieel-generator model calculate generator peed (frequency), due to electrical power output (I N, V N) and governor action. Integrate frequency to a target phae, and thu create a three-phae voltage et V N* Control 80kVA genet to lock phae and magnitude of the actual V N triplet to the imulated V N* Fig. 3 Hardware-in-the-loop (HIL) environment for the dieel/generator model

13 Fig. 4 Generator frequency : Pure imulation and HIL uing original etup Zero, or yntheied noie Phae command P() Actual Phae ( ) F nom = 50 Hz + - Meaured Phae ( ) p Phae error ( ) Phae meaurement repone.5 cycle averaging p = 0.09 R() 360 F nom Q() Phae change per econd ( /) pu revolution error, time Fnom 360 Target pu frequency f F nom Q () C P () Meaured pu frequency F() Actual frequency Frequency meaurement repone 9 cycle averaging f = 0. 2H H() H =.20 L L() Fig. 5 Torque control ytem pu frequency error Friction factor L = C F () + - Feedforward throttle d D() + + Drive repone d = 0.03 pu torque command

14 Fig. 6 Command to meaured-value repone of the phae loop and frequency loop: hand-tuned parameter Decription Parameter Adjutment to (4) & (5) Final Value Target frequency CLTF repone time a 0.5 ec Target phae CLTF repone time b 0.02 ec Original etup (Hand-tuned) Frequency loop C F Ki Booted by 0x Kp Kd.352 Ka Reduced to ¼x Frequency loop C F low-pa filter cutoff fa f a 66.6 m 2.39 Hz Phae loop C P Ki m 50 Hz (on Kd control only) Kp x0.04 = 4.4 Kd Ka Reduced to ¼x Phae loop C P low-pa filter cutoff pb p b. m 4.3 Hz Table Control parameter for DC motor torque control

15 Fig. 7 OLTF and CLTF bode plot (Gain) for lambda-tuned torque control Fig. 8 OLTF and CLTF bode plot (Phae) for lambda-tuned torque control Fig. 9 Command to meaured-value repone of the phae loop and frequency loop: lambda-tuned parameter

16 Fig. 0 Generator frequency : Pure imulation and HIL uing improved field and torque control Fig. Voltage tracking for 0.5pu contant-power tep HIL experiment Fig. 2 Phae tracking for 0.5pu contant-power tep HIL experiment

17 Fig. 3 Power flow for 0.5pu contant-power tep HIL experiment Fig. 4 Turbo preure Decription Maximum voltage tracking error (pu) Maximum frequencytracking error (Hz) Maximum phae tracking error (degree) Maximum power flow tracking error (pu) Original etup a per [7] Improved voltage tracking (80kVA field control) Improved voltage tracking and DC motor control Hz 0.6 Hz 0.6 Hz Table 2 Incremental improvement in the fidelity of the 0.5pu contant-power tep HIL experiment