Design and Evaluation of a New Converter Control ... - IEEE Xplore

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. 1

Design and Evaluation of a New Converter Control Strategy for Near Shore Tidal Turbines Mahda J. Jahromi, Student Member, IEEE, Ali I. Maswood and K.J. Tseng, Senior Members, IEEE

 Abstract— while tidal stream patterns of natural channels are essentially consistent and smooth, coastal tidal flows experience much more radical speed variations. This paper proposes a new converter control strategy for maximum power extraction of the fast changing currents that are typically seen in open waters and near the shore. The proposed method makes use of variable speed generation and has the ability to adapt to the nature of different tidal flows; thus enabling efficient exploitation of resources that require faster dynamic responses. The main motivation behind this study has been exploration of the possibility of combined off-shore wind-tidal co-generation. Aspects discussed in the paper include interdependencies between the injection of the fluctuating generated power, grid-tie connection, and maximum power capture with a view to addressing energy conversion and power quality requirements. Detailed insights into this novel power generation system are provided with a case study that involves a method for Maximum Power Point Tracking (MPPT) and power injection controls in a non-stiff network. Index Terms—Tidal Power, Variable Speed Generation, Maximum Power Point Tracking (MPPT), Voltage Regulation

M

I. INTRODUCTION

arine or oceanic energy refers to the energy carried by ocean tides, waves, salinity, and ocean temperature differences. Among the aforementioned types, this study focuses on the oceanic tidal stream power which began to draw attention in the mid-70s after the first oil crisis. The periodic rise and fall of the water along ocean coasts and their rhythmic daily patterns are familiar phenomena which are caused for the most part by the relative positions of the moon and Earth. Tidal currents occur in conjunction with the rise and fall of the tide. In other words the vertical motion of the tides along with the rotation of the earth causes the water to move horizontally, creating currents. Such recurrent cycles of current buildup along with the high density of water make tidal stream power a renewable resource with extraordinary energy density when compared to other renewables.

Consequently development of this emerging technology presents unprecedented opportunities to increase power extraction and efficiency that are yet to be explored. The lunar semidiurnal tide, with two high and two low waters each day is the principal world tide. The amount of displaced water in a lunar cycle each day is tremendous, however it happens very gradually over a course of several hours. Consequently in the open ocean where the water enjoys a vast basin the tidal currents caused by the rise and fall of water, have lower velocities. In addition to the lunar tides at these so called nearshore areas, stream patterns are also influenced by propagation and reflection of tidal waves. As the wave propagates into shallower water, its wave speed decreases and the energy contained between crests is compressed both into a smaller depth and a shorter wavelength; thus the tide height and the tidal stream strength must increase accordingly. Once reaching the shore the water returns (usually at an angle) to the sea causing reverse currents. This phenomenon can also be explained through the Bernoulli potential and kinetic energy conservation equations and the continuity of mass principle [1]. One must note that this water flooding and ebbing occurs more frequently than the lunar tides, imposing regular and rapid speed variations on near-shore and coastal currents. On the other hand in water straits, estuaries and natural channels where water streams are restricted in a much narrower space, the current speed variations are essentially shaped by the lunar tides with wave propagation having less effect. In these areas tidal streams are of much higher speed and rectilinear, meaning they flow back and forth in a somewhat straight line, changing speed and direction much more gradually than the coastal flows. Consequently most of the existing tidal turbines are designed and installed in channel type areas where a broad strait connects two areas of the sea.

Manuscript received on April 23, 2012 and was accepted for publication on November 19, 2012. Copyright (c) 2009 IEEE. Personal use of this material is permitted. However, permission to use this material for any other purposes must be obtained from the IEEE by sending a request to [email protected]. Mahda J. Jahromi did his Ph.D. at the school of EEE, Nanyang Technological University, Singapore. He is now with the E.E department of Endurance Wind Power, Vancouver BC, Canada. (email: [email protected]) Ali I. Maswood and K.J. Tseng are with the Department of Electrical and Electronics Engineering, Nanyang Technological University, Singapore.

Fig. 1: Different types of tidal stream speed profiles

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].


As illustrated in Fig. 1, while channel type tidal currents have very gradual and consistent speed variations that also happen to be very predictable [2], in spite of their similar overall trend, speed changes in costal currents are more spontaneous. Actual recorded current speed profiles illustrated in the following figures, confirm the aforementioned characteristics. Figure 2 depicts the current speed profile that was recorded at Martinez-Amorco station in November 2010. This station is located in the strait that connects the San Pablo bay in California to the bays of Suisun and Grizzly. As can be seen from the figure once the current builds up, the tidal stream pattern in that channel is quit consistent with minor speed variations. On the other hand figure 3 shows the speed profile of a harbor near the Ship Island in Mississippi; apparently although the general tidal flow pattern is still observable, the instantaneous speed variations have increased.

Fig. 2. A channel type current speed profile, recorded at Martinez-Amorco station, 1/11/2010 [3]

Fig. 3: A near-shore type current speed profile, recorded at Day Marker station, 7/11/2010 [3]

While speed variations intensify as the distance to the shore decreases, coastal flows on the whole are apparently weaker than the channel type streams. This fact raises a fundamental question of whether it’s feasible to exploit such tidal currents in the first place. The current trend in the tidal industry can also be used as a bear witness to support the argument, since nearly all tidal power companies find natural channels and water straits to be more attractive for installing their turbines. Be that as it may, the concept of offshore wind power and the advantages that co-generation brings with itself, substantiates the need for a closer study of near shore currents and ways of increasing the productivity of near-shore tidal turbines. In addition to that the same technology can also be applied to other devices that experience similar boisterous conditions. The physics of the existing Tidal In-Stream Energy Conversion (TISEC) systems is very similar, in principle, to the kinetic energy conversion systems in the wind industry; resulting in an inevitable, though superficial, resemblance of many of the TISEC devices to wind turbines. The stream power that a tidal turbine extracts is proportional to the density of seawater and cube of its speed:

½

.

1

is the cross-sectional area of the flow Here intercepted by the turbine; / is the stream velocity and is the water density in ( / ). Normally 934 and it’s this high water density (compared to that of the air) that causes the high power potential. However a turbine can only harness part of this power and so a performance is included to represent the percentage of the coefficient stream power that the turbine extracts. This factor is limited to 16⁄27 ≅ 59% by the well-known Betz law. For wind turbines, have typical values in the range of 25-30% while for tidal turbines, is estimated to be within 35-50% [4]. Based on (1), it can be inferred that the available power of both the channel type (slow dynamics) and near-shore type (fast dynamics) would present similar profiles to that of the speed. From an electrical point of view the variability presented by these marine resources demands special control strategies that could maintain the turbine Tip Speed Ratio (TSR) at its optimum under different flow velocities. A controllable TSR calls for variable speed generation in which the speed of the rotor is varied according to the speed changes of the flow. Maximum power capture through variable speed generation is a field that has gained attention in the wind power industry [5-8], however, MPPT in tidal in-stream systems and in particular for near-shore turbines remain a rather less documented field. Moreover, this relatively new field of study also requires addressing of other critical aspects such as maintaining the power quality and avoiding voltage fluctuations while injecting variable amounts of high active power into the network. Series/parallel reactive power compensators (e.g. DVR and STATCOM) or fault ride through capabilities have already been proposed for other renewables, particularly wind turbines [9-12]. However unlike wind power, repetitive nature of tidal currents provide opportunities for novel Point of Common Coupling (PCC) voltage regulation schemes through the grid-tie converter reactive power control, a feature that is also studied in this paper. This paper addresses the power extraction and power transfer problems of near-shore tidal turbines in an integral manner. Based on the characteristics of the resource, a maximum power extraction technique is proposed (referred to as Controlled K method) and compared to a conventional MPPT (referred to as the Constant K) method. A PCC voltage regulation scheme is also investigated to mitigate voltage fluctuations that result from interactions between the instantaneous power delivery, dc link, and power extracted from the turbine. A detailed case study and a design example are presented based on computer models of a tidal turbine that employs a Squirrel Cage Induction Generator (SCIG), and connects to the grid through a full-scale converter interface. As a result of the investigation, the advantages of the proposed Controlled K method become evident for fast changing flow patterns. This is a timely contribution to the emerging field of tidal power generation.



Fig. F 4: Schematic diagram of a grid connected tidal tuurbine and its full sscale converter.

TIDALL IN-STREAM ENERGY CONVEERSION The overall sy ystem under study is depicteed in Fig. 4. The T AC C/DC/AC system in one piecce realizes a fu ull scale converter unnit that conneccts the SCIG--turbine set to o a non-stiff grid g (shhown by its eq quivalent Thev venin model) th hrough submarrine cab ables and a step s up transfformer. Based d on the con ntrol obbjective the system is divideed into two sections; a variaable freequency rectifiier at the mach hine side that bu urdens the task k of MPPT and macchine flux reg gulation, and the t grid impo osed d AC voltages and a freequency invertter that regulates the DC and is referred to in the diagram ass the Hybrid Power P Port (HP PP). dulation is used d to Ass illustrated in the figure, spaace vector mod opperate the conv verters, resultin ng in an indep pendent controll of P and Q by the grid-tie g inverter and a decoup pled control of the CIG torque and d flux by the machine m side recctifier. SC Both the inveerter and rectifier units are Voltage Sourrced Coonverters (VSC C) that are conn nected back to o back through the DC C link. Altho ough the conv ventional 2-leevel VSC is the doominant buildin ng block of a wide w range of devices d in mediium annd high powerr applications, 3-level VSC Cs require low wer sw witch voltage rating r and ressult in lower THD levels. But B coontrolling multiilevel converteers is a bit moree complex as th hey neeed an additional voltage eq qualizing conttroller. There are diffferent types of multilevel VS SC systems eacch with their cons annd pros, such as a the Cascadeed H-Bridge (C CHB), the Fly ying Caapacitor (FC),, and the Neutral N Point Clamped (NP PC) coonverter. In th his study a 3-level (NPC C) converter was w preeferred since unlike u the CBH H it does not require r a separrate DC C source and d hence avoiids a bulky and nonstand dard traansformer and d is relatively y cheaper thaan the FC ty ype coonverters [13]. Figure 5 show ws the schemattic diagram of the 3-L Level NPC co onverter employed in the stud dy. The converter is operated und der PWM with w third harrmonics injecttion witching strateg gy that enables a better utiliization of the DC sw buus voltage wh hile preventing g over-modulaation; under this t schheme the bus voltage can bee reduced to 86% 8 of it nomiinal

value [[14, 15]. Thee switching llosses associaated to the converteer are modeledd as a currentt source ( ) attached to the DC side [16], w while the switcches are consiidered to be semi-ideeal with on--state resistannces and voltage drops, reflectinng their condduction lossess. Figure 6 depicts the capacitoor voltage equualizing system m that is designned to seize any parttial DC voltagee drift. Only onne voltage balaancer is used in the ovverall system aas voltage balaancing is perforrmed merely by the H HPP converterr. As the figuure illustrates, the DC bus capacitoors voltage diffference i.e. is proccessed by a Low Paass Filter (LP PF) before beiing used as thhe feedback signal. T The direction of the HPP power flow dettermines the sign of tthe controller ooutput. Thereaafter, any capaccitor voltage drift is compensated by means of changing the modulation C voltages beecome equal again. The index uuntil the DC controll er details aloong with othher employed values are presenteed in the Appenndix.

Fig. 5: Scchematic diagram of the 3-Level NP PC converter and itts switch model

Fiig. 6: Control blocck diagram of the DC bus voltage eqqualizer



I. DESIGNING A HYBRID POWER PORT The grid-tie VSC operating at line frequency along with its DC link capacitor is referred to as the Hybrid Power Port (HPP).The objective of the HPP controller is to regulate the active (P) and reactive (Q) power flow that is exchanged with the AC system. For simplicity, in this part the turbinegenerator set along with its AC/DC variable frequency converter is considered as a black box that exchanges a timewith the DC side of the system. A varying power capacitor link is used to maintain the DC voltage, implying that the DC bus voltage is not imposed and so needs to be regulated. In view of that the converter control system is designed to be capable of a stable bidirectional power exchange between the DC bus and the AC network. There are two approaches to controlling P and Q in a VSC system. One is called the voltage-mode control in which the active and reactive powers are controlled, respectively, by the phase angle and the amplitude of the converter AC side terminal voltage (with respect to the PCC bus voltage); and the other is referred to as the current-mode control in which the active and reactive powers are controlled by the phase angle and the amplitude of the converter line current (with respect to the PCC bus voltage). The voltage-mode control is simple and has a low number of control loops; however, its main shortcoming is that there is no control loop closed on the converter line current, and consequently, the converter is not protected against overcurrents. Therefore the converter current may undergo large excursions if the power commands are rapidly changed or faults take place in the AC system. Apart from the overcurrent issue, other advantages of the currentmode control include robustness against system parameter variations and higher control precision [17]. Due to the aforementioned advantages and in order to realize a decoupled active and reactive power control, a dq-frame based current mode control strategy is employed. Accordingly a suitable PLL was designed to continuously generate the grid voltage needed for dq components as well as the rotation angle abc/dq vector transformation. In order to synchronize the HPP with the grid, the PLL measuring inputs had to be obtained from the PCC; while the current measurements were taken from the LV side, where the HPP is located, Fig. 4. Consequently the phase shift introduced by the ⁄Δ transformer along with voltage and current scaling had to be taken into account when designing the PLL, adapting it to the vector group and turns ratio of the transformer. Ultimately the active and reactive powers delivered to the grid at the PCC can be formulated as: 3 2 3 2

2

where , represent the HV side (grid) dq-frame voltages (PLL outputs) and , represent the LV side (converter controlled) dq-frame currents. Since the grid voltage cannot be

is constant, however once the

controlled,

vectors so that PLL reaches its steady state, it adjusts , thus simplifying the equation set (2) to: 0 and 3 2 3 2

⇒

2 3 2 3

3

4

In other words a proper PLL operation, leads to manipulation of and by means of controlling the HPP converter and currents respectively. By rearranging (3) and introducing reference variables, the reference signals of the HPP, dq-frame current controller can be formulated as presented in (4). Consequently, as long as the current control system provides fast reference tracking, and can be independently controlled by their respective references. It must be noted that (PLL output) in steady state is a DC value. Thus provided that and are adjusted smoothly (within several cycles of the line and signals could also be considered as frequency), DC quantities. As a result, the current control system in dqframe deals with DC variables which can be addressed with a basic PI controller. Figures 7 and 8 illustrate , reference generating systems employed in the the HPP. The generated ( , signals, are then used alongside with the PLL outputted ( , values to form the ( , references that are required by the converter current controller system of Fig. 9. In other words, while the dq-frame current controllers form the inner loop of the HPP controller, the active and reactive power controllers (figures 7 and 8) form the outer loops, generating the active and reactive power references (according to eq. 4) by processing the DC and AC bus voltage errors. The control systems depicted in figures 7 and 8 are designed with respect to their overall closed loop dynamics. of Fig. 7 is a Therefore while the active power controller second order controller, designed for a stable bidirectional power flow within the HPP power rating, the reactive power controller on the other hand is a basic proportional integrator with a conservatively large phase margin and a relatively small bandwidth ensuring a decoupled and stable closed loop operation. The smaller bandwidth compared to that of the closed loop dq-axis current controller is to minimize their mutual impact. One must note that the reactive power controller in this case is designed with respect to the facts that there is no direct load on the PCC bus, and the impact of the reactive power on the PCC bus voltage is much greater than that of the active power. The inputs of these controllers are also assigned in accordance to the overall



syystem dynamicss. For examplee, considering the t power balance eqquation of the system shown n in (5) (wheere is the HPP H terrminal power and a is the e converter pow wer loss) the HPP H syystem can be seen s as a dynaamic model in which is the staate variable and a is the control inputt while aand are the distu urbances. 1 2 ⇒

5

2

Hence, instead d of

and

, their sq quared values, i.e.

III. VECTOR CO ONTROL OF A VARIABLE SPEE ED SCIG The ppower port of the previous ssection was strructured as a DC capaacitor link connnected to a griid imposed freqquency NPC converteer. In this secttion the same kind of NPC converter is controll ed to establishh a bidirectionnal power flow w between a variablee speed induction generator aand the DC linkk. Due to the relativelly large machine inductannce, no additional series inductannce is requireed to reduce the effect oof switching harmoniics. Thus as deepicted in Fig.. 4, the variablle frequency converteer is directly connected to the machine stator. The electricaal and mechannical equationss governing thhe dynamics of a sym mmetrical threee phase inductiion machine are:

and are employed. This makes the system lin near wiith respect to the t state variaable, leading to o a reduced orrder coontroller and a relatively r betteer dynamic perrformance. At the sam me time the instantaneous power generaated by the tiidal geenerator, i.e. ) is used d as a feed forward inputt to im mprove the sysstem responsess to any abrup pt changes in the inccoming powerr. This feature results in a much m steadier DC buus voltage und der rapid speeed/power variaations which are coommon in tidal wave [18], wind and near-sh hore turbines. The reactive power p controlller inputs (Fig g. 8) on the otther haand are selected d to be the diffference betweeen the rms vallues off the PCC and grid Thevenin voltages, i.e. . CC bus voltag ge regulation is performed with w Thherefore the PC resspect to , which can be b obtained fro om a dispatch hing centre. Similar to t the active power regulattor of Fig. 7, the i also limited within the HP PP correspond ding reaactive power is poower ratings. For controller details and parameter vallues reffer to the Appeendix.

Fig. 7. Con ntrol block diagram m of the active pow wer controller

6 1

1

7 3 2

.

1

1

∗

8

where tthe stator andd rotor leakaage factors and are defined in (8) with and repressenting the stattor and rotor inductannces, and thhe magnetizing inductance. In a squirrel cage orr a wound rottor asynchronoous machine w whose rotor terminalls are short ccircuited, 0, and is usually not measuraable. However by introducinng a fictitious sspace phasor current defined as ≜ 1 , it can b e shown that tthe machine ellectric torque ssimplifies to the folloowing equationn [16]. 3 2 1

Fig. 8. Conttrol block diagram of the reactive power controller

9

Whatt (9) imposes iss that if whhich is basicallly equivalent to the machine maagnetizing currrent and thuus flux, is machine torque becomes a maintainned at a constaant value, the m linear ffunction of the q-axis component off the stator current . Therefoore in this stuudy is reguulated at the machinee nominal maggnetizing curreent value correesponding to its ratedd voltage and frrequency [19], i.e.: F Fig. 9: Block diagrram of the current mode controller in n dq frame. and d t inductance vaalue seen between the reepresent the networrk frequency and total HPP and the PC CC, respectively.

2 32 1

10



It can also be sho own that:

11 1

whhere: 1

12 1 ,

osing (11) into its real and imagin nary By decompo coomponents the equation set (13) ( is obtaineed that forms the baasis of SCIG rotor-flux obsserver [20]. Fiig. 10 shows the reaalization of (13) in a roto or-flux observeer model thatt is em mployed in th his study. Th his configurattion is generaally refferred to as thee current modell flux observerr.

Fig. 11: B Block diagram of the SCIG Flux annd Torque controlleers in dq frame

13 1

For the most part, regulation n of is only y exercised eitther duuring the system m startup proccess, or when the machine iss at staandstill during which 0. 0 This way, flux f establishm ment is ensured beforee any torque demand. d The outputs of the flux f hen used as feeedback signals for obbserver block (Fig. 10) are th thee flux and torq que controllerss as well as ab bc/dq conversiion. Ass can be seen n in Fig. 10, is fed to the t observer as a a sepparate input which w implies th he need for an n encoder. Fig. 11 deepicts the blocck diagram for fo the complete SCIG vecctor coontrol system, in the rotor field f coordinaates. It should be nooted that while the d componeent of the stato or current contrrols thee machine flu ux, the q comp ponent (which h is a function n of booth ) controls thee machine elecctrical torque. For and paarameter valuess and controllerr details refer to t the Appendiix. d in figures 4 and a 11, the con ntrol inputs to the As illustrated vaariable frequen ncy rectifier co ontrol system are the mach hine eleectrical torquee command ( ) as well w as the flux f refference . It is assum med that once the t flux reference is set to the maachine nominaal value (10) it remains fairly coonstant resultin ng in chang ging essentially y with . At the sam me time Fiig. 11 show ws that oncce the flux is esttablished, and a consequently are mainly m controllled byy . Thus, at various stages can bee manipulated d to reggulate the macchine speed, reesulting in an indirect i controll of thee power exchaange between the mechanicaal turbine and the eleectrical machin ne.

Fig. 10: Scheematic diagram of the rotor-flux observer for SCIG

Fig . 12: Performancee coefficient versuus TSR, characterisstic curve

III. MA AXIMUM POWER R EXTRACTION N Therre are variouus methods available forr achieving maximuum power exttraction in the wind and ssolar power industriees. They incluude methods such as the hhill-climbing search ccontrol, power signal feedbacck or the use off fuzzy logic controll ers [21-23]. H However so far few articles (if any) in the T methods beiing proposed oor examined literaturre show MPPT for mariine in-stream aand particularlyy near-shore tidal turbines. In this ppaper, an effecctive and simpple solution is proposed in which a conventionaal MPPT routtine is enhancced with an acceleraation boosting technique designed to addreess the rapid speed vvariations of thhe coastal andd near-shore cuurrents. The proposeed MPPT methhod will be tested against the original MPPT rroutine whichh is consideredd as a benchm mark in the analysiss. It’s worth m mentioning thhat the accelerration boost conceptt can also bbe implemennted using otther MPPT techniquues; a topic thaat is consideredd as part of thee team future studies. As m mentioned befoore, the portioon of the availlable stream power thhat a tidal turbbine can extract is given by (11), where is a highhly nonlinear function of the turbine Tip Speed Ratio andd blade pitch anngle ). Fig. 12 shows the , versus characteeristic curve off the employedd turbine for vaarious values 0 of . It can be seen thhat 0; as increeases also increasees until it reacches a peak vaalue at 5( 0), and thereafter, anyy further increaase in resultss in the drop of vaalue. Noting thhat the peak vaalue of , i.e.. is largest w when the pitchh angle 0; if is increassed, peak value drrops and so ddoes the turbinne power . U Usually is



regulated to keep the turbine working within its rated limits, however for simplicity, throughout this study the turbine is assumed to operate at a fixed pitch angle, i.e. 0. To keep the turbine spinning at must , the generator speed change at different water speeds , so that the tip speed ratio . Conventionally, this i.e. . ⁄ , is maintained at can be assured if the machine electric torque is commanded to change proportionally to the square of its rotor speed [24]. .

0.5

_

curves are assumed to Moreover, although the turbine be static, in practice formation of fouling over the blades and other surfaces results in a gradual change of the turbine characteristics. Even though application of anti-fouling paint or transient high energy voltage/acoustic pulses has been proven to be effective, there still is no long-term solution to this problem. This is why ease of installation/disengagement is an issue with high priority in tidal turbine structural designing. Consequently allowing for an adjustable K also results in an improved system adaptability and robustness to the changes in turbine characteristics.

14

Equation set (14) explains the conventional MPPT method in which represents the machine nominal frequency, is the machine nominal power, represents the mechanical torque exerted on the machine shaft, N is the gearbox ratio and is the turbine radius. The principal behind this is that if , then . Consequently, / is positive and so the turbine-generator set accelerates, and increases. Similarly, if , then , / is negative, and so decreases. Therefore, the optimum operating point is quite stable and a steady state can be reached where / 0, , and . Although this method is simple and robust, the system needs adequate time to reach and settle at the optimum speed. Thus the traditional MPPT technique presents limited dynamic response making it ineffective when subjected to rapid speed variations. To overcome this problem, the machine should accelerate/decelerate faster. It can be seen ⁄ is a from (14) that the machine acceleration function of J (the moment of inertia) and . Taking into account the much larger mechanical time constants compared to the electrical ones, can be varied while remains essentially unchanged. Thus, by commanding to change to a smaller value than what the conventional method proposes (i.e. ) a greater acceleration can be created. On the contrary, if the machine is required to decelerate, can be commanded to change to a value that is greater than . . This way, the gap between and is widened at the instants of abrupt speed variations, thus leading to a higher machine acceleration/deceleration. It must be noticed that this change only takes place during speed variations and at all other times the torque coefficient returns . Fig. 13 depicts the proposed control system. This to value under controller is designed to operate at the normal conditions while adjusting the K value when the TSR deviates from its optimal value. As is seen from the figure, while a saturation block limits K to negative values (keeps the machine in a generator mode), the deadzone function eliminates any small fluctuations in K once a steady state is reached.

Fig. 13: Proposed K controller for a better MPPT under rapid speed variations

IV. CASE STUDIES In this section the TISEC system shown in Fig. 4 will be examined with the proposed and conventional MPPT strategies under both slow and rapid speed variations that in turn resemble natural channel and coastal type of flow patterns. For this study a standard 1.68MW, wound-rotor, asynchronous machine whose rotor terminals are short circuited is considered to be connected to a tidal turbine through a gearbox with a gear ration of 10. Since tidal flow speeds are usually less than 5 ⁄ , a directly coupled induction generator would need to operate at very low rpms undermining its efficiency. Hence a small gear box with a constant gear ratio is employed to increase the generator rotation speed. The power exchange between the tidal turbine and the AC grid is controlled through the grid-tie converter (HPP) which is linked to the on-shore step-up transformer by means of a 480V sub-sea transmission cable. Although the choice of a LV link seems counter-productive with respect to copper losses, the choices for voltage levels are limited to the converter capabilities as well as local legislations as utilization of high voltage undersea systems especially in areas with possibility of public access is prohibited. Be that as it may, should a higher voltage submarine line be used, a medium voltage DC line would be more beneficial than a HVAC, since inverter output voltage levels are limited and employing an underwater transformer is not very appealing. The submarine DC transmission system is essentially a good topic for further studies in multi-machine systems since in addition to loss reduction it also allows for networking of various tidal generators that are spinning at different speeds. The system in Fig. 4 starts by activation of the HPP converter, which takes place prior to that of the machine side. This allows for the initial capacitor voltage buildup and machine energization through intake of utility power. To make better use of the simulation run time, the DC link capacitors are assumed to have an initial voltage close to their nominal value. It’s worth mentioning that the nominal DC link capacitor voltage is determined based on several factors such as the AC side peak voltage, the expected power levels and the



am mount of reversse voltage thatt switch modulles can withstaand. Coonsidering thee plots of figu ures 14 and 15 the follow wing seqquence of ev vents take place during thee system starttup. Att 0.1 , gaating signals of the HPP P converter are unnblocked and all a its correspo onding controlllers are activatted. Thhis causes the controller c to co ommand a neg gative , (i.ee. to im mport active po ower from thee AC system to the DC siide) chharging the cap pacitors. Conseequently, by 0.15 , the DC vooltage is regulated at 2500V V, (Fig. 22). The T machine-sside coonverter on thee other hand is activated at a 0.2 . Once activated the flux f compensaator steps up , which h is b , regulating to its no ominal value. For rappidly tracked by prootection purpo oses the machin ne stator dq cu urrents are limiited to 1200A. Ass can be seeen from Fig. 14, while 12 200 , risees towards its reference valu ue at a rather low l paace due to the large rotor tim me constant (ssee also Fig. 10). 1 Onnce reaches the 141A (eq. 10) set point, the flux f coontroller reducees from the maximum m value of 120 00A to about the sam me steady state value v of 141 . At this point the b is com mplete. The required r mach hine maachine flux buildup ennergization po ower is obtain ned from thee grid which is refflected by the negative tran nsferred powerr plot of Fig. F 155. Until this stage, since th he machine waas stationary, the eleectrical and mechanical m torq ques as well as were zeero. Chhanges in thesee variables take place after 0.5 , once the tiddal flow begin ns causing a grradual move of o the drive traain. Thhis can be seen n as the slowly increasing in Fig. 15. Fig. F 166 shows the cu urrent speed prrofile that is used in this pap per. Thhe speed profille is designed to resemble a so called “seemidiuurnal” tidal prrofile, having two high and two low tidess in onne cycle (a lun nar day). Mostt areas of the world except the W Western coasts of o America, So outh East Asiaa and Middle East E exxperience simillar kind of tidaal flows. Furth hermore compaared to an actual tidall flow, here thee corresponding g time periods are ware fidelity limitations. Baased scaaled down to meet the softw onn the speed vaariation two sccenarios are considered. In the firrst scenario thee speed profilee of Fig. 17 is assumed to occcur in 1/1000 of a lu unar day which h is approximaately 100 secon nds. Hoowever in thee second scenaario the same speed profilee is coonsidered to taake place 10 times faster, (i.e. in only 10 secconds). In oth her words, the first case-stud dy is designed d to miimic the flow pattern p of a naatural channel where changess in speed happen relatively r slow wly while the second scenaario s currents, where the speeed variations are appplies to near shore muuch more raapid. This way w an impaartial and beetter coomparison can be made bettween the perfformances as the exxerted flows on nly differ in theeir occurring tiime spans. It must m bee noted that since the torquee demand takees place after the maachine energiization the system behav vior during the iniitialization perriod in both scenarios is basiically the samee as whhat was explain ned previously y. Caase 1: Slow Ch hanging Speedss (Channel Typ pe Currents) Fig. 17 presents the extractted . as well as the griid transferred powers under each of the two MP PPT strrategies for the channel typee tidal currentt. The extractaable

For the sake of clarity a power iis plotted as a reference. F moving average filterr is used to sm mooth the annd graphs. As cann be seen froom the plots the performannces of the PPT schemes are almost Controllled and Connstant K, MP identicaal, with small superiority off the Controlleed K routine during tthe startup andd saddle regionns where the sspeed is low and the flow is changing direction vvery fast. This can be seen more eevidently in tthe plots of Fig. 18. It’s worth mentionning that sincce actual speeed changes oof a natural channel take place at a much slowerr pace than thiss simulation, the perfformance of thee conventionall Constant K sccheme could practicaally improve evven further. Thhus with regardds to currents with sllow speed vaariations the performance of the two methodss can be considdered more or lless the same.

Fig. 14: Machine maggnetization processses during the starttup period

Fig. 15: P Plots of , SCIG tthree phase currents and Geneerator Speed durring machine startuup period

F Fig. 16: Tidal flow w speed profile useed in the two case studies

Case 2: Rapid Changiing Speeds (Neear Shore Currrents) Figurre 19 shows thhe extracted powers under thee Controlled



annd Constant K, MPPT strategies s plottted against the exxtractable power when the speed changess occur 10 tim mes fasster. As can bee seen from Fig. 20, even th hough there arre 4 driifts in the streeam pattern (Fiig. 16), the Co onstant K scheeme miisses most of th he power in the first two driffts; the Controllled K method on thee other hand is capable of cap pturing part of the poower available in the second drift, displayiing the advanttage off this scheme during startup ps. However compared to the preevious example this scenariio depicts a much m weaker and a sloower startup which w is mainly y due to the high h inertia of the driive train preveenting the rotorr speed to incrrease promptly y; in othher words the system simply y does not havee adequate timee to caatch-up with the changing currrent. Apart fro om the startup for dK thee rest of the siimulation the superiority of the Controlled schheme is quit noticeable, paarticularly from m the plotss of Figg. 20.

allow thhe machine entter a motoring state for a shoort period of time whhenever subjectted to sudden speed increasees. This way, the systtem takes som me of the poweer stored in thee DC link to catch-upp with the fasst changing tiides, resulting in a better maintainned tip speed ratio and pow wer capturing pperformance. Howeveer in such a casse any gain in ppower capturinng efficiency needs too be weightedd against the inncrease of the system cost and com mplexity, parrticularly since rapid channge of high electricaal currents is reequired.

Fig. 17: Extracteed and Transfeerred Powers , under both MPPT T schemes plotted ag gainst the Total Extractable Power, channel type curren nts

Fig. 19: E Extracted Powers under both M MPPT schemes plotted against the Total Extraactable Power, coastal type currents

nce Coefficient plots for Consta ant and Controlled d Fiig. 18: Performan MP PPT schemes underr channel type currrents

Although the proposed system in this casse performs mu uch beetter than the co onventional MPPT M method, as expected wh hen coompared to the previous sccenario where the same speed proofile was hap ppening more gradually, th he overall pow wer cap apturing efficiency is not goo od, indicating that t the portion n of thee available po ower that can be captured in n coastal areass is muuch lower than n that of the straaits and water channels. m thaat in this stu udy was not It’s worth mentioning alllowed to take on o positive vallues, thus prev venting the systtem froom entering a motoring m mode. So an altern native approach h to adddress the low power capturiing efficiency here could bee to

Anothher noticeablee point in Figg. 20 is the shift in the curvees while rremains the sam me which is m mainly due to the rellatively largee turbine-gennerator mechaanical time constantts. Subsequenttly the delay in the machinee responding time whhich was also ppresent in the previous scennario is more visible iin this graph ddue to the smaaller simulationn period and thus thee higher resoluttion of the plott.

Fig. 20: Performance Coeefficient plotts for Constant andd Controlled K MPPT schhemes under coasttal type currents

A. Com mparisons

Fig. 21: P Power capturing effficiency differencces under the two M MPPT schemes



The surface area a under the , and power plotss of figgures 17 and 19 represent the correspon nding extractab ble, cap aptured and tran nsferred energy y sums respecttively. Due to the shhifts in the captured powers in i both scenariios, energy ratther u to calcullate the efficiiencies. Thus the thaan power is used ovverall power caapturing efficiiency under eaach of the MP PPT strrategies is calcu ulated as: %

whhere:

3 2

.

100 15 1

. 1 2

evaluateed under the Controlled and Constantt K MPPT methodss for coastaal and channel type speeed frofiles respectivvely. Figures 22-25 show w how the H HPP system regulatees the AC and DC vooltages underr diffeerent circumsstances.

16 1

using the trapezo oidal numerical integration method. m Since is s caalculated with respect to the PCC bus volttage and LV side he calculated efficiencies take cuurrent measureements (16), th intto account th he mechanical losses as weell as the lossses peertaining to thee converters an nd the submarrine line. The bar graaph of Fig. 21 1 summarizes the overall effficiencies for the tw wo MPPT schem mes under each h scenario. The efficiency gaains in the figure can c be interprreted as the ability a to han ndle n a sudden chan nge geenerator accelerration and deceeleration when off speed occurss. The results show that un nder rapid speed d outperforms the chhanges the proposed Controllled K method coonventional Co onstant K routiine. However when the systtem opperates under gradual speeed variationss, both methods peerform more or less similarly. Therefore, the Controlled d K strrategy not onlly is applicab ble to water channels c but also a suuitable for bidirrectional tidal generators thaat are intended for cooastal and neear-shore tidaal farms exp periencing more m spontaneous speeed changes. d Power Port (HPP) (H B. Performance of the Hybrid The need for the reactive po ower control iss more noticeaable nstable networrks. The stabillity of a netw work in electrically un caan be charactterized either by its equiivalent Theveenin mpedance or itss short circuit level. l The high her the impedance im or alternatively the lower thee short circuitt level, the more m v variability. The netw work vuulnerable is thee network to voltage em mployed in this study was an 11 KV netw work with a sh hort cirrcuit level of ju ust 6MW. Acco ording to [25] the standard sh hort cirrcuit level off an 11KV network shou uld be 200M MW, inddicating that th he network em mployed here was w much weaaker thaan the averagee network. Thiis small short circuit level was w inttentionally asssumed to illustrate the effeectiveness of the prooposed HPP reactive r power handling sy ystem even un nder opperation in excceptionally weak networks. To overcome the lim mitations of th he conventional DC power ports which are meerely designed d to handle actiive power, the HPP described d in thiis study was designed to exchange acttive and reacttive poowers (like a STATCOM), S so s that the PC CC bus voltagee is maaintained withiin acceptable limits while vaariable amountss of active power iss transferred to t the grid. Subsequently S the peerformance of the HPP systtem presented in section IIII is

Fig. 22. D DC bus voltage proofile under the Controlled and Consstant K schemes

Fig. 23: Performaance of the DC voltage equalizing syystem

Fiig. 24: %Δ (withh respect to ) plotted against and Constant K, M MPPT scheme, chaannel type currentts

.

Figg. 25: %Δ (withh respect to ) plotted against and Controlled K, MPPT scheme, N Near-shore currentts

.



Fig. 22 shows the DC bus voltages plotted for the case one and two scenarios. In both cases the HPP intaks active power from the grid to increase the DC bus volatge from an initial value of 2300V to the nominal 2500V setpoint. Due to its smaller simulation time span the V buildup at startup is more clearly seen on the left hand side plot. After this initialization period both plots show that the nominal DC voltage is well maintained at different stages of the system operation irrespective of the status of the power exchange. Fig. 23 on the other hand shows the performance of the DC voltage equalizing system introduced in section II, which is also part of the HPP control system. The two plots in this ) figure show the DC capacitor voltage differences (i.e. in each of the two scenarios. It is observed that in both cases the voltage equalizing system renders the capacitor voltage balancing effectively, as the DC voltage differences are relatively insignificant. Plots of figures 24 and 25 on the other hand show how the HPP system, handels the AC voltages. Both figures show the PCC bus injected active (Ps) and reactive (Qs) powers under, near unity power factor (PF=0.98) and Voltage Regulated (V.R) conditions. The pertaining voltage differences, i.e. is also depicted under each of the two Δ MPPT schemes. As can be seen from the results, while the near unity power factor scheme results in voltage swells that vary by the amount of extracted/trasferred power, the proposed HPP reactive power control has resulted in an average voltage deviation of less than 1%. Hence, variable amounts of active power can be transferred to the local network without causing voltage distortion. In other words assigning a small portion of the grid tie converter power capacity to reactive power control has resulted in an effective PCC bus voltage regulation without the need for any auxilary STATCOM or any similar equipments. V. CONCLUSIONS Due to the variable nature of tidal streams, extraction and transfer of tidal power brings about many challenges such as the need for efficient power extraction and PCC voltage regulation strategies. In this paper a design solution was proposed and evaluated for coastal and channel type tidal streams that differ mostly by the rate of speed variation. Several simulation examples as well as comparison studies were included to illustrate the effectiveness of the proposed system under different scenarios. It was found that by widening the gap between mechanical and electrical torques at instances of rapid speed change the machine rotor can be made to accelerate/decelerate faster, thus tracking the changing tides more closely. The results of the study show that although the proposed MPPT system performs more or less similarly to a conventional scheme when employed for a channel type speed profile, the power extraction efficiencies are well different when it comes to coastal type currents that change speed and direction much more rapidly. In this case, although the proposed MPPT system requires an extra flow

meter for monitoring the speed of the incoming streams, the practical difficulties associated with employing such a device are offset by the increase in power capture. At the same time it was shown that, assigning a small portion of the grid-tie converter power capacity to reactive power adjustment, counteracts the effect of active power injection on the PCC bus voltage. Therefore variable amounts of active power can be transferred to the local network while maintaining the PCC bus voltage within the specified limits. The shortcomings of the system proposed in this study include the higher electric current and torque variations that are the result of faster generator acceleration/decelerations as well as the dependency of the proposed MPPT system on generator and flow metering devices which can reduce the system reliability and increase the maintenances efforts. Addressing of these issues is part of the team objectives in future studies. APPENDIX Grid Voltage 11 KV Frequency 60 Hz SCL 6MVA , ⁄

Line Parameters L 50 μH R 2 mΩ ω 377 rad/s

8

Transformer Converters Yg/Δ1 Type: 3 Phase, NPC, 3 Level Sn 2.5MVA , f 60 Hz r 1 mΩ LV 480 Ph Ph rms Forward Threshold Voltage Vd 1 V HV 11KV Ph Ph rms Carrier Frequency 1680 Hz R R 0.0325 pu Modulation type: PWM 3rd Harmonic Injection L L 0.818 pu C 9625 μF , V _ 2500 V Rm Lm 500 pu Machine Turbine & Gear Box Voltage 2.3 KV , Power 2250 hp Radius 5m C 0.5 at β 0 f 60 Hz 29 mΩ , R 22 mΩ R λ 5 L 35.2 mH , L 34.6 mH , L 35.2 mH τ 0.5 s 1.213 s , τ 1.6 s σ 0.03384 τ Gear Ratio N 10 J 63.87 Kg. m , No of Poles 4 96% HPP Controllers (GIF-VSC) s

H

1870

H

8000 , s

20 2075s 1 . 0005 0.2 s

K s

K s G

s

1

3.26 s

36 s 568520 s 166s 6890 1505s 568520 s 962s 232325 s s

s

Machine side Controllers (VF-VSC) H G s PI



186

s

I

0.625 s s

I

s

800 s

500 s

DC Voltage Equalizer LPF

13.6 s 333 s

s s

3ω 3ω

377

⁄

6.5

10

A lunar day is slightly longer than a solar day and lasts 24 hours and 50 minutes, i.e. 89400 seconds

ACKNOWLEDGMENT The author would like to thank Prof. Martin Ordonez for his valuable comments and insights.



REFER RENCES [1]

[2] [3]

[4] [5]

[6]

[7]

[8]

[9]

[100]

[111]

[122]

[133]

[144]

[155] [166] [177]

[188]

G. H. B. Polagye, "M Methodology fo or Estimating Tidal T P by Tidal T Current Energy Resourcees and Power Production m Energy Conversion (T TISEC) Devicces," In-Stream EPRISepttember 29, 2006 2006. M. J. Jah hromi, A. I. Masswood, and K. J. Tseng, "Long Term T Prediction n of Tidal Currrents," Systems Journal, IEEE, vol. PP, pp. 1--1, 2010. CO-OPS. Real-Time Actiive Current Stattion Data - Sta ation Availaable: Selection.. http://tideesandcurrents.noaaa.gov/cdata/Station nList?type=Cur rent+Dataa&filter=active L. Myerrs and A. S. Bahaj, "Power output performaance characteriistics of a horizzontal axis mariine current turbiine," Renewablle Energy, vol. 31,, pp. 197-208, Feb b 2006. S. Moriimoto, T. Nak kamura, and Y. Y Takeda, "Po ower maximizaation control of o variable-speed d wind generaation system using permanen nt magnet syncchronous generattor," 0, pp. 11-19, Jan n 30 Electricall Engineering in Japan, vol. 150 2005. h to C. M. Hong and W. M. Lin, "Inteelligent approach m power point tracking control strategy for maximum variable-sspeed wind turrbine generation n system," Eneergy, vol. 35, pp. 2440-2447, Jun n 2010. nuolo, G. Petronee, S. V. Araujo, C. Cecati, E. FriisF G. Spagn Madsen, E. Gubia, D. Hissel, H M. Jasin nski, W. Knapp, M. Liserre, P. Rodriguez, R. Teodorescu, and P. Zachaarias, ble Energy Operation and Conv version Schemes: A "Renewab Summary y of Discussions During the Sem minar on Renew wable Energy Systems," Indusstrial Electronicss Magazine, IE EEE, vol. 4, pp. 38-51, 2010. G B. Singh,, S. S. Murthy y, and N. Kish hore, P. K. Goel, "Isolated Wind-Hydro Hybrid System Using Cage C y Storage," Ind dustrial Electron nics, Generatorrs and Battery IEEE Tra ansactions on, vol. 58, pp. 1141-1153 3, 2011. H. Chong, A. Q. Huang, M. E. Baran, S. Bhattacharya, W. n, A. L. Johnson n, and A. A. Edris, Litzenberrger, L. Anderson "STATCO OM Impact Stud dy on the Integration of a Large Wind Faarm into a Weeak Loop Poweer System," Eneergy Conversio on, IEEE Transsactions on, voll. 23, pp. 226-233, 2008. S. W. Mohod M and M. V. Aware, "A STATCOM-Con ntrol Scheme for Grid Conneccted Wind Energy y System for Po ower S Journal, IEEE, vol. 4, pp. Quality Improvement," Systems 346-352, 2010. S. Zhang g, K. J. Tseng, and S. S. Choii, "Statistical volltage quality assessment meth hod for grids with wind po ower n," Renewable Power P Generation, IET, vol. 4, pp. generation 43-54, 20 010. M. Alberrdi, M. Amundaarain, A. J. Garrrido, I. Garrido, and F. J. Maseda, "F Fault-Ride-Through h Capability of ng-Water-Column--Based Waave-Power-Generaation Oscillatin Plants Equipped With Doubly D Fed Indu uction Generator and Airflow Control," Industtrial Electronics, IEEE Transacttions on, vol. 58, pp. 1501-1517, 2011. J. Rodrig guez, S. Bernet, B. Wu, J. O. Pontt, P and S. Ko ouro, "Multilev vel voltage-so ource-converter topologies for industrial medium-voltag ge drives," Ieeee Transactions on Industriall Electronics, vol. 54, pp. 2930-2945 5, Dec 2007. J. A. Ho ouldsworth and D. D A. Grant, "Th he Use of Harm monic Distortion n to Increase thee Output Voltagee of a 3-Phase Pwm P Inverter," Ieee Transactiions on Industryy Applications, vol. 224-1228, 1984. 20, pp. 12 B. M. Wilamowski W and J. D. Irwin, Po ower electronics and motor driives, 2nd ed. Boca Raton, FL: CRC Press, P 2011. A. Yazd dani and R. Iraavani, Voltage-sourced converterss in power systems s : mod deling, control, and applicatiions. Hoboken,, N.J.: IEEE Press//John Wiley, 2010 0. M. P. Kazmierkowski K and L. Malesan ni, "Current con ntrol techniquees for three-phase voltage-sourcee PWM converrters: A survey y," Ieee Transacctions on Industrrial Electronics, vol. 45, pp. 69 91-703, Oct 1998. Z. Zhou,, W. Knapp, J. MacEnri, H. C. Sorensen, E. Friis F Madsen, I. Masters, and a P. Igic, "Permanent maagnet

[19] [20] [21]

[22]

[23]

[24]

[25]

generator contr trol and electriccal system connfiguration for Wave Dragon MW wave energy take-off system," in Industrial Electtronics, 2008. IS ISIE 2008. IEEE E International Symposium on, 22008, pp. 1580-1585. W. Leonhard, C Control of electrrical drives, 3rd ed. ed. Berlin ; London: Springger, 2001. W. Leonhard, Control of electtrical drives, 3rdd ed. Berlin ; New York: Sprinnger, 2001. Q. C. Wang and L. C. Chaang, "An intelliggent maximum power extractioon algorithm forr inverter-based variable speed wind turbine systems," Ieeee Transactions on Power Electronics, vol.. 19, pp. 1242-12449, Sep 2004. K. Ohyama, S. Arinaga, and T. Shigemizuu, "Sensorless vector controllled converter for variable speed wind generation systtem using an innduction generator," Electrical Engineering in JJapan, vol. 159, ppp. 62-75, Jun 20077. M. G. Simoes, B. K. Bose, annd R. J. Spiegell, "Design and performance eevaluation of a fuzzy-logic-baased variablespeed wind generation systtem," Ieee Traansactions on Industry Applicaations, vol. 33, pp. 956-965, Jul-Augg 1997. R. Datta and V. T. Rangaanathan, "Variablle-speed wind power generati on using doublyy fed wound rrotor induction machine - A comparison withh alternative scchemes," IEEE Transactions onn Energy Conveersion, vol. 17, pp. 414-421, Sep 2002. Society., IEEE M. H. J. Bolleen, IEEE Industrry Applications S Power Electronnics Society., aand IEEE Poweer Engineering Society., Undeerstanding powerr quality probleems : voltage sags and interrup uptions. New Yorkk: IEEE Press, 1999.

BIOGRAPHIEES Mah hda J. Jahromi ((S’08) was born inn Shiraz – Iran. He received his B.Enng. degree from Y Yazd University and his M.Eng (Firstt Class) from Shiiraz University, D. at the electrical and electronics Irann. He did his Ph.D engiineering departm ment of Nanyang Technological Uniiversity, Singaporee. Since July 2012 he’s been with the electrical engineering team of Enndurance Wind wer, a wind turbinee manufacturer in V Vancouver BC, Pow Cannada.

Ali I. Maswood (S’885–M’88–SM’96) obtained his B M. Eng. Degrees with first class from Moscow & M Pow wer Engineering Institute, and Ph.D. from Conncordia Universitty, Montreal, Caanada. Having taugght in Canada forr a number of yeears, he joined Nannyang Technologiccal University, Sinngapore in 1991 wheere currently he iss an Associate Proofessor. He is a seniior member of IEE EE, actively involvved in the local IAS S/PELS chapter annd in the steeringg committee of the IEEE Power E Electronics & ddrives (PEDS) con ference. ng-Jet Tseng (S’85–M’88–SM’98)) was born in Kin Singgapore. He receivved his B.Eng. (F First Class) and M.E Eng. Degrees froom the National University of Singgapore, and his Ph.D. degree froom Cambridge Uniiversity, U.K. He iis currently the heaad of the power eng ineering divisionn at Nanyang Technological Uniiversity, Singaporee. He has held a nuumber of major app ointments in proffessional societiess, including the Chaair of IEEE Singaapore Section in 2005 and was awaarded the IEEE Thhird Millennium M Medal.
