A novel control technique for on-load tap changer to enlarge the reactive power capability of wind power plants

This paper suggests a new control algorithm for On-Load Tap Changer (OLTC) of wind farm main transformer that is responsible for voltage regulation of collector network. The conventional control scheme adjusts the voltage of LV side of transformer without considering the wind turbine voltages. The voltage of end-string wind turbines may reach the upper or lower limits of Over or Under-Excited operation regions. Voltages close to upper or lower limits bounds the reactive power capability (generation or consumption) of wind farm due to dependency of wind turbine reactive power capability to voltage. Operation of wind turbine at extreme voltages may stress its power electronic devices. Furthermore, minor disturbances at either grid or power plant may trigger the protection system and interrupt generation due to vicinity of operation point to protection thresholds. The suggested technique integrates the OLTC control into the power plant controller and considers both LV side voltage of transformer and wind turbine voltages. By monitoring the wind turbine voltages, the voltage setpoint of OLTC is adjusted to provide voltages closer to middle of range for wind turbines that prevents the wind turbines from reaching the voltage limits. Dynamical simulations conducted in DigSILENT PowerFactory software for both Under-Excited and Over-Excited operation regions demonstrate desired performance for the suggested technique.

synchronous generators.On the other hand, the current technology of renewable energy sources do not possess the same characteristic and dynamic performance that is reflected by synchronous generators.Although, many of synchronous generator capabilities are emulated by renewable energy sources via implementing proper control loops, dealing with grid connection issues is still an open research area that needs more investigation and mature technical solutions for all types of converter-based renewable energy source, for example, solar, wind, storage etc. [4,5].
Weak grid is a result of aforementioned transition, since the installation location for synchronous generators differ from renewable energy sources.It means that the location of renewable energy source has to be moved to distant places far from the transmission system and/or load centers, where extracting the maximum energy from renewable energy sources is possible and more cost-efficient.As a result, the need for long transmission lines, causes appearance of high impedance between the grid and renewable energy sources, which is called weak grid [6][7][8].
The technology of wind turbines are gradually become matured in emulating the characteristics of synchronous generators.This performance improvement makes wind turbines a proper replacement for synchronous generators in response to fast-growing global warming concerns [1,4,6].Although, the technology performance of power electronic converters is promising at converter and wind turbine level, there are still some unaddressed shortcomings at plant level, that need more research investigation [7,9].Enhancement of wind turbine performance at plant level that is called grid code compliance, requires revision of technical performance of all wind farm elements, that is, power electronic converters, power plant controller, On-Load Tap Changer (OLTC), compensation devices etc. Dynamic characteristics of wind farm equipment, their distributed (local) control, the coordination between the local controllers via hierarchical (centralized power plant controller) control, and the optimized single line diagram of wind farm are among the key factors that determine the overall performance and robustness of wind farm connected to a time-varying grid connection.Among all of above key factors that can affect the performance of wind farm against various grid conditions, the OLTC is the main focus of this paper.The main objective of this study is to achieve a reliable and stable connection of wind farm to the grid that can utilize the maximum active power and reactive capability of wind farm with minimum sizing of compensation equipment.To this aim, integration of OLTC and power plant controller (as master) is the best solution, but since the power plant controller models are commercial and are not publicly available, the suggested technique is implemented as an external module besides of power plant controller.
There is another big difference between a synchronous generator and a wind farm of the same rating.For synchronous generators, the entire power is injected at a single point with a unique voltage, while wind farms consist of numerous wind turbines that are located relatively far from each other and have to inject their power at their own connection point at different voltages.It means that the power production is distributed and power injection into collection network is performed at numerous connection points with cable section impedance between them.Flowing current through the cable impedance changes the voltage of connection point of wind turbines.This voltage difference is worse for wind turbines that are connected at the end of string.Such wind turbines experience maximum voltage deviation from LV side of main transformer that is regulated around 1 pu by OLTC [10].Depending on the operation region (Over-Excited or Under-Excited) and active and reactive power set-points, the wind turbine voltage may reach the upper or lower limits and trigger the protection system of either collection network or wind turbines.
The OLTC of main transformer is designed to regulate the voltage of Point of Common Coupling (PCC) at 1 pu.Depending on the operation region of wind turbines (Capacitive or inductive), the wind turbine voltages lay down in only half of the full voltage range, that is, 0.9 pu to 1 pu or 1 pu to 1.1 pu.
In fact, improper control strategy of OLTC limits the wind turbine voltages to half of their permissible range.Therefore the voltage of end-string wind turbines may reach the limits that can cause saturation of reactive power for either generation or consumption.This phenomena becomes worse at full active power operation point.This deficiency narrows the reactive power capability of wind farm, reduces their grid connection flexibility and robustness and limits the target market for wind turbine manufacturer and wind farm developer.As a resolution for such issue, extra compensation devices are required for compliance of grid code requirement, which increases the CAPEX and OPEX of wind farm.
The OLTC generally measures and regulates the voltage at LV side of transformer by changing the tap position at HV side.In case of having multiple parallel transformers in service, since they measure the same LV side voltage, simultaneous tap changing action may happen by different parallel transformers.In case of having independent and autonomous operation of OLTCs, there is the risk of simultaneous tap change by numerous parallel transformers.Under such a circumstances, an over voltage condition might be converted to under voltage issue and vice versa due to over reaction of all parallel transformer OLTCs.Therefore, coordination of multiple OLTCs is essential to prevent undesired switching actions and their corresponding oscillations in the voltage profile [11].
Lack of such coordination may cause non identical tap position of transformers, hence different internal impedance and therefore unbalanced load sharing between parallel transformers that can reduce the life-time of transformers [12].The conventional and built-in OLTC control does not have any particular control policy for group operation.The integration control of OLTC in the power plant controller that is suggested in this paper can be a smart solution for all above mentioned issues.The total tap position changes can be equally distributed as much as possible between distinct transformers in service.Meanwhile, sufficient time delay can be considered after each individual tap change to prevent simultaneous and multiple tap changes until the voltage reaches the steady state, ready for next tap change action, if is needed.
International Standard (60214-2-2019) is recently released covering the requirements, references, definitions and relevant guides on selection of OLTC design in accordance with IEC 60214-1 or IEEE Std C57.131 for use in conjunction with the tapped winding of transformers or reactors [10].Reference [13] investigates the performance of three techniques: Reactive Power Control, OLTC and both as stand-alone (without coordination) to control the voltage of lower distribution network.The paper finds the latter more efficient especially for high penetration of renewable energy source.The drawback of this method is, first, lack of control coordination and second, only focusing on variation of active power and overlooking the reactive power capability of entire PV farm that is essential for grid code requirement compliance.
Reference [12] presents a comprehensive literature review on reactive power support devices such as OLTC transformers, FACTS, synchronous condensers that can improve the static and dynamic reactive power capability of renewable power plants.Combination of OLTCs with other reactive power support equipment is recommended to achieve efficient voltage regulation.A similar voltage regulation method is proposed in [14] via combination of OLTC and reactive power capability of wind turbines.The voltage control of collection network is performed using state estimation for OLTC.The voltage control at remote buses is done using their locally connected wind turbines, which can prevent over/under voltage at remote buses.For such a method, the hardware and software required for implementation of state estimation and the location of installation is in question.A modified particle swarm optimization approach is suggested in [15] for voltage management in distribution networks using a combination of STATCOM, OLTC and reactive capability of PV inverters.The proposed technique is validated against voltage imbalance and constraints, varying load and PV generation.Although, the performance of such a method might be promising, but from CAPEX and OPEX aspect of view, it might not be financially efficient and affordable.
A coordinating control strategy is suggested in [16] for OLTC in PV distribution system using daily profile of feeder voltages and tap position to control the PCC voltage.Dependency of the suggested scheme on daily profiles of tap position and feeder voltages is a challenge for achieving a reliable, robust and general control scheme.The paper [17] tries to map the feeder voltage and topology data to OLTC tap ratio changes to minimize the voltage deviation across the feeders using Markov Decision Process (MDP) and Reinforcement Learning (RL) algorithm.Another algorithm is suggested to estimate the voltage magnitude under different tap settings using linearized power flow model, which allows the RL algorithm to explore the state and action spaces offline without interrupting the system operation.This technique suffers from uncertainty in estimation accuracy of an offline mapping lookup table that may not be an efficient countermeasure for online control of nonlinear dynamics.
Unintended interactions between OLTC and reactive power controller at distributions systems is investigated in [11] for parallel operation of OLTC transformers in PV farms with different reactive power control strategies.The paper highlights reduction of life time of OLTC due to increment in the number of OLTC switching actions [12].The life time reduction is the consequence of lack of control coordination, which in our paper is suggested as an advantage of OLTC integration into power plant controller.Reference [11] clearly emphasizes on the contribution of current paper that addresses coordination of OLTC with entire power plant controller to achieve an efficient control strategy and higher OLTC life time as a positive side effect.
The methodology proposed in [18] determines the tap position to maximize either generation or consumption of reactive power considering uncertain levels of active power and voltage at point of common coupling.A mixed-integer nonlinear programming model is used for offline optimization of reactive power capability.The offline optimization is the main drawback of this method.Both the grid characteristics (impedance, inertia, short circuit power etc.) and power plant topology (out of service transformer and wind turbines) undergo time-varying changes.Such an offline method may not be the best solution for all possible scenarios of operating conditions.
The plant main transformers are manufactured with a standard and fixed number of tap positions, for example, ± 11 or 17.For most of wind farms, a part of tap position range may remain unused for ever.The main contribution of this paper is to utilize the unused tap position range of main transformer that can reduce the voltage deviation of wind turbines from 1 pu.Wind turbine voltages closer to nominal, provide wider reactive power capability and hence more flexibility in the grid code compliance at point of connection without need for installation of extra compensation equipment.On the other hand, the OLTC tap position is changed mechanically and all mechanical devices are designed for a particular number of operation actions as life-time.Always using the middle range tap positions may reduce the life time of corresponding contacts, while some of the lower and higher tap positions may never used.However, according to the probability theory, distribution of total number of tap changes between all possible positions may yield longer life time of OLTC mechanical parts.
The rest of this paper is organized as follows: The structure of a typical offshore wind farm is presented in Section 2. The reactive power capability and its grid code requirement are introduced in Section 3. The basics and performance of Conventional OLTC are described in Section 4. The proposed adaptive OLTC is discussed in detail and mathematically formulated in Section 5.The simulation results of both methods are compared in Section 6.Finally, the paper is concluded in Section 7.

WIND FARM CASE STUDY
Figure 1 represents the single line diagram of a typical offshore wind farm.The wind park consists of 4 strings with 5 wind turbines in each string.Submarine cables connect the strings to the PCC bus.The distance between the wind turbines relatively determines the length of cable sections, which is generally a trade-off between the wake effect and infrastructure costs.The cable length of 5-10 times of wind turbine rotor diameter is recommended as an optimal estimation.As can be seen in the Figure 1, all strings are connected to the PCC bus, where is the LV side of power plant main transformer.The PCC voltage is regulated by OLTC of main transformer, which is common for all strings.Therefore, the suggested technique is applied to the entire power plant.
where L 2 and R d stand for length of submarine cable and wind turbine Rotor diameter, respectively.The cross section of cables are determined using the built-in Cable Sizing tool available in DigSILENT, PowerFactory software.Cable Sizing tool utilizes the load flow and short circuit calculations, cable loading and voltage drop across the feeders for determining optimized

REACTIVE POWER CAPABILITY
Since the voltage dynamics are relatively local comparing to the frequency dynamics, the transmission system operator generally require contribution of grid connected generators into voltage regulation at their point of connection.It means that power plants should provide flexible and wide range of reactive power capability from generation to consumption to support grid voltage regulation.

Reactive power capability of wind turbine
Table 2 and Figure 2 represent a typical reactive power capability of wind turbine for both capacitive and inductive operation regions in terms of terminal voltage and active power in pu: Over-excited (capacitive): Production of reactive power, Q > 0. Under-excited (inductive): Absorption of reactive power, Q < 0.
There are some practical constraints associated with thermal limit of power electronic devices that are often overlooked in the modeling of wind turbines.The reactive power capability of wind turbines depend on both voltage and active power of operation point.It means that when the wind turbine reaches either the voltage or active power limits, the reactive power production capacity is depleted.This production saturation is worse at corner points, where the active power reaches 1 pu or the voltage reaches 0.90 or 1.10 pu as can be seen in Figure 2. Therefore, to utilize the maximum reactive power capability of wind turbine and hence the entire wind farm, the terminal voltage of wind turbines should be kept close to their nominal value of 1 pu as much as possible.Due to the existing cables in the collection network of wind farm and hence the impedance between wind turbines, provision of 1 pu voltage for all wind turbines at their terminals may not be possible in practice, especially in case of having long radial strings.
Different wind turbine manufactures produce various wind turbine types (Type 3: Doubly-Fed Induction Generator (DFIG) and Type 4: Permanent Magnet Synchronous Generator (PMSG).These types have their own variants considering power rating (Non/optimized active power), nominal frequency, and connection voltage that offer various reactive power capabilities.In some brands, the reactive power is a constant and predefined lookup table, but for some others a built-in code determines the reactive power limit online depending on the operation condition of wind turbine.Based on practical experience of working with different brands, the worst case scenario is considered for this study, in which the modeled wind turbine reaches its reactive power limits sooner than similar wind turbines from other brands, but good engineering practices in the plant level may resolve this issue (shortage in wind turbine reactive capability).By implementing the suggested technique, some of non grid code compliant wind turbines that are originally designed for connection to the strong grids (cheaper ones with limited reactive power capability) might be able to be grid code compliant.

Reactive power requirement at point of interconnection (POI)
Grid Code Compliance study investigates the performance of power plant under static (steady state), dynamic (transient) and faulty operation conditions while it is connected to the grid.Comprehensive tests including various conditions of wind power plant and grid are conducted via professional power system simulation platform to identify potentially grid connection issues.Reactive power capability is among the various tests that should be conducted to identify the level of grid support by power plant via provision of reactive power at POI. Assessment of reactive power capability for static (steady state) analysis include two main tests of PQ and UQ charts.
The PQ chart is a graphical representation of the reactive power capability of the power plant over the full range of active power dispatch and a specific voltage level at POI. Figure 3a indicates the grid code requirement for PQ chart of the power plant for different countries.The power plant is grid code compliant in a particular country if the reactive power capability of power plant fully covers the region surrounded by closed boundary (areas) of corresponding country (preferred with enough security margin).In other words, if the PQ capability of the power plant at POI is wider than the PQ chart requirement, the power plant meets the grid code requirement for POI voltage of 1 pu.
The UQ chart is another graphical representation of the reactive power capability of power plant over a specific range of POI voltage and a constant value of active power dispatch (Normally full power).Figure 3b indicates the grid code requirement for UQ capability chart of wind power plant.For UQ capability, the grid code requirement is similar to PQ chart, that is, the wind power plant must supply reactive power outside (wider) As can be seen in Table 2 and Figure 2, the reactive power capability of wind turbines directly depend on their active power and terminal voltage.The same applies to the reactive power that is exchanged between the grid and power plant at POI.The PCC voltage reference is determined power plant controller, which is calculated based on the wind turbine voltages.However, the PCC bus is located between the grid and wind turbines.According to the basics of electrical engineering, the source with less impedance, has more impact on the PCC voltage.Under strong grid condition, the PCC bus is electrically closer to the grid and therefore the grid voltage has higher impact on the PCC voltage, unless the weak grid condition applies, which is vice versa.However, the grid voltage is applied to the HV side of power plant main transformer and the OLTC task is to moderate the wind turbine voltage against the grid extreme voltages.Therefore the voltage at POI is dependent on the grid voltage and wind turbine active and reactive power.
One solution for enlarging the reactive power capability at POI is to improve the design of power electronic converter or considering some spinning reserve in the converter or selection of larger converter, which may not be financially affordable and is out of scope of this paper.Another solution is improving the voltage regulation at wind turbine terminals through power plant controller, OLTC or both via integration of OLTC controller into power plant controller or their coordination.In this paper, integration of OLTC controller in the power plant controller is of interest.

ON-LOAD TAP CHANGER (OLTC)
Power transformers equipped with OLTC are among the main components of industrial applications in power system for nearly 90 years.OLTCs are the indispensable part of regulating power transformers that are used for both generation and distribution system of electrical networks.The OLTC performs the phase shifting and/or voltage regulation at LV side of transformer inside a desired boundary by changing the tap position (transformer turn ratio) of primary winding under load condition without power interruption [12].The tap changing is generally performed at high voltage winding of transformer due to the following reasons: • The currents at high voltage winding is lower, the tap changer contacts can be smaller for current switching action.• The high voltage winding is wound outside the LV winding, access to the tapping connections is much easier.
In distribution systems, the OLTC is implemented as a stand-alone controller on the transformer, since OLTC is the last voltage control scheme before delivery of power to the customers.Therefore, for distribution system, upgrading the conventional OLTC is the only implementation option.Despite of the distribution system, for generation application such as wind farms, the power plant controller exists after the OLTC toward the low voltage side responsible for controlling of entire wind farm from MV to LV network.For such an application, integration and/or coordination of OLTC with power plant controller may provide more efficient control strategy.

Conventional OLTC (COLTC)
The main transformer at the offshore substation is responsible for regulation voltage at LV side of transformer, that is, 33 kV PCC bus as indicated in Figure 1.The control objective requires that the voltage at LV side should follow the voltage reference v re f .The control error tolerance is determined by voltage upper v max and lower v min limits.It means that if the voltage settles down between the boundary limits for a time period beyond the tap time constant T c , the tap position should be remained unchanged.Otherwise, for voltages above/below the upper/lower limit (v max /v min ) for a time period longer than the tap time constant T c , the tap position should be decreased/increased one step: where V min , V re f and V max are typically constant values for conventional OLTC controller as follows: The drawback of conventional OLTC controller is that the tuning set-points are constant and are determined regardless of operating region of wind farm, that is, Over-Excited (Capacitive) or Under-Excited (Inductive).Figure 4 compares the performance of Conventional and Adaptive OLTC control techniques for both capacitive and inductive operation regions.The vertical and horizontal axes show the voltage and reactive power of wind turbines in pu, respectively.The same reactive power set-point is applied to all wind turbines by power plant controller, but due to existence of cable impedance between wind turbines, different voltages appear at wind turbine terminals.For Conventional OLTC (Figure 4a), the PCC voltage is adjusted on 1 pu , while depending on the operation region, the voltage of end-string wind turbine may violate either the upper or lower limit.For those power plant controller brands, in which the wind turbine voltages are not involved in the control strategy, either the Low Voltage Ride Through (LVRT) or High Voltage Ride Through (HVRT) protection scheme might be triggered leading to trip of some or all wind turbines.Under weak grid condition the situation might be worse as more wind turbines may cross the voltage limits and can be considered as worst case scenario [6,7].
For those power plant controller brands, in which the measured voltage of some wind turbines (the first and the last in the string) or all wind turbines are taken into account in the control strategy, the reactive power set-point is reduced in such a way to keep the voltage of end-string wind turbine below/above the upper/lower limit.Although, for these smarter power plant controllers, trigger of LVRT or HVRT is avoided, reduction of reactive power set-point causes limited reactive power capability and hence less flexibility and robustness of wind farm for grid connection, especially under weak grid condition [6,7].A narrowed reactive power capability at plant level while full capacity is available at wind turbine level, may not be financially acceptable, as it may limit the market biding opportunities for the wind farm developer.This phenomena clearly reveals that a wind farm consisting of well designed wind turbine may not reflect desired performance at plant level.Therefore, coordination of all distinct and independent controllers are required to achieve an optimized, flexible and robust performance at plant level against recent grid connection issues.

PROPOSED ADAPTIVE OLTC (AOLTC)
Different solutions are suggested to the wind farm owner for meeting the grid code requirement.After investigation of wind turbine and power plant controller technical specifications, tuning the control parameters and protection settings are considered as the first and cheapest solution.If aforementioned firmware tuning fails in fixing the issue, designing and installation of supplementary compensation equipment is initiated, which may not be a cost-effective solution.The suggested technique is an alternative countermeasure that can fix the issue partially or completely, depending on the severity of condition.It means that if it does not fix the issue completely, it can at least reduce the sizing of required compensation equipment, which makes the solution efficient and financially affordable.
Figure 4b illustrates the performance of proposed Adaptive OLTC (AOLTC) that adjusts the voltage of LV side (PCC) at a dynamic value rather than a constant value as it is in the conventional OLTC (COLTC) method.In the new control scheme, the voltage reference is determined online based on the operating point of wind farm.For the Capacitive operation region, the wind turbines produce reactive power and as a result, the wind turbine terminal voltages are raised.Therefore, in order to prevent reaching the voltage of end string wind turbine to the upper limit, the voltage reference (v re f ) for PCC should be decreased by OLTC.The opposite trend applies for the Inductive region.
The main objective of control effort is to settle down the wind turbine voltages at 1 pu.Since there is a voltage difference between distinct wind turbines, the average voltage of all wind turbines (v av (t )) is feedback to the control scheme: where v i (t ) is the voltage of wind turbine i and n is the number of wind turbines.For those power plant controllers that have only access to two wind turbine voltages, that is, the closest and farthest wind turbines to the PCC, the average of these two voltages are considered.The average voltage deviation (Δv av (t )) of wind turbines from 1 pu can be calculated as below: The voltage reference should bring back the wind turbine voltages toward nominal voltage and distribute them around 1 pu as much as possible: In order to avoid trigger of voltage protection system at PCC due to voltage violation from limits, the deviation of voltage reference (v * re f (t )) from nominal voltage is limited to ±5% (±0.05 pu).
According to Equation ( 7) and as can be seen in Figure 4b, the proposed technique yields lower and higher v re f (t ) for Capacitive and Inductive operation regions, respectively.The v min (t ) and v max (t ) also need to be updated accordingly as below: Comparison of Figure 4a (Conventional OLTC) and Figure 4b (Adaptive OLTC) shows that the Adaptive approach shifts the wind turbine voltages toward the middle of permissible range for both operation region.Providing terminal voltages closer to 1 pu for more wind turbines causes generation or consumption of more reactive power by wind turbines, and hence yields wider reactive power capability.The voltage deviation at PCC (Δv(t )) from voltage reference (v re f (t )) can be calculated as follows: A dead-band area around the voltage reference (v re f (t )) is defined to prevent the tap form undesired switching, if the voltage settles down inside the upper (v max (t )) and lower (v min (t )) limits: where v + db (t ) and v − db (t ) are defined as below: To detect the proper direction of tap changing by means of increasing or decreasing, the sign(t ) variable is defined: Since the OLTC is operated by means of a drive mechanism, the differential equations are required to describe the dynamical variation of OLTC tap position using a state variable x tap(t ) : where T c is the time constant of OLTC tap changing.The total operation time period of an OLTC varies between 3 and 10 seconds, depending on the respective design and brand.The life time of mechanical parts of OLTC is around 1.2 million switching operations under load conditions with no maintenance required before 300,000 switching operations.The digitized tap position can be calculated from the integer part of its state variable x tap (t ), that is, [x tap (t )]: where ⌊x tap (t )⌋ and ⌈x tap (t )⌉ denote the floor and ceiling of x tap (t ), respectively.Both LV voltage measurement and tap changer actuator that are essential for implementation of suggested technique are available for conventional OLTC, therefore there is no need for extra equipment.For implementation, the control loop should be closed through power plant controller rather than local OLTC.To this aim, the control task needs to be added to the power plant controller.Merging OLTC task into power plant controller slightly increases its calculation burden, since the OLTC dynamics are relatively slow due to the existing time delay (in the range of seconds) between successive tap change actions.
Including the tap position and measured LV voltage signals into SCADA communication system might be the only practical implementation issue for suggested adaptive OLTC.Since the SCADA system is already available between the point of connection and power plant controller, this issue can be limited to inclusion of two more signals into the data exchange protocol.Based on SCADA field experience, such a flexibility to consider extra signals into SCADA communication system is possible, which makes the suggested technique more realistic.

SIMULATION RESULTS
The performance of Adaptive On-Load Tap Changer controller (AOLTC), is evaluated using worst case scenarios of operating points.The results are compared with Conventional On-Load Tap Changer controller (COLTC) in terms of steady state value of main transformer tap position, voltage and reactive power at wind turbine and PCC buses.In each scenario, the operating point of wind turbines are forced to move to the corner points indicated in Figure 2. The corner points are the worst case conditions, since the wind turbines are forced to operate at maximum active and reactive power close to the voltage limits for both Over-Excited and Under-Excited regions.

Scenario 1: Capacitive operating region
Figure 5 compares the performance of COLTC and AOLTC methods located at the first and second rows, respectively.The columns from left to right compare voltage, reactive power and tap position for aforementioned methods.In this scenario (S1), the wind farm starts from steady state of full active power production and no exchange of reactive power with the grid at POI bus, that is, (P,Q)= (1,0).Then the reactive power setpoint is changed to maximum (from Table 2 and Figure 2) at second 10.
Comparison of Figures 5a and 5d shows that the POI voltage profile is almost the same for both methods, but the main difference is in the PCC and wind turbine voltages.The PCC voltage is controlled around 1.00 and 0.95 pu by COLTC and AOLTC methods, respectively.For COLTC method (Figure 5a), higher PCC voltage causes some of wind turbine voltages reach the upper limit of 1.1 pu, while the wind turbine voltages are settled down in the range of 1.04 to 1.06 pu by AOLTC method (Figure 5d), far from the the upper limit.According to Table 2 and Figure 2, reaching the voltage of wind turbines 4 and 5 to the voltage limit, saturates their reactive power production as can be seen in Figure 5b, while none of wind turbines reach the voltage limit for AOLTC method (Figure 5e) and as a result, all wind turbines can reach their maximum reactive power production (Figure 5e).Figures 5c and 5f represent the tap position for COLTC and AOLTC methods, respectively.AOLTC method settles down the tap position at 11, while COLTC method yields position of 9.According to the basics of OLTC operation, higher tap position causes lower voltage at LV side of main transformer.
According to the almost all grid codes, the Transmission System Operators (TSO) require power plants to do not exchange reactive power with the grid, unless it is agreed.The studied scenarios consider the worst case condition in which the power plant is acquired to inject/absorb extreme reactive power.Such circumstances, that is, extreme reactive power and hence low voltage condition at PCC rarely occur.The aim is to investigate the power plant performance under such severe condition, identify the system vulnerability and take the best control strategy (best operation point/region) to prevent trigger of protection system.Table 3 compares the results of scenario 1 for COLTC and AOLTC methods.The AOLTC method causes higher reactive power injection to the grid (2% more) due to controlling the wind turbine voltages (maximum voltage < 1.063 pu) closer to the nominal value comparing to the COLTC method (maximum voltage =1.10 pu).Besides of wider reactive power capability, AOLTC keeps the wind turbine voltages far from the limits, which increases the reliability of power plant operation.In COLTC method, in which the wind turbine voltages are close to higher limit (1.10 pu), minor disturbances or faults in either grid or power plant may trigger the protection system and hence interruption of generation.
The suggested AOLTC technique provides more reactive power and this reactive power goes through the collection network cables to the grid.Therefore the losses is slightly increased.Table 3 shows 0.5% increase in active power losses in the cables, which is still quite low inside the permissible range of 3% recommended by Electrical Balance of Plant (EBoP).

Scenario 2: Inductive operating region
This scenario (S2) compares the performance of COLTC and AOLTC methods at worst-case operation point of Inductive region as indicated in Figure 6.The wind farm starts from the same steady state operation point as scenario 1, that is, full active power production and no exchange of reactive power with the grid at POI bus, that is, (P,Q)= (1,0).Then the reactive power set-point is changed to minimum that means maximum absorption of reactive power (from Table 2 and Figure 2) at second 10.In this scenario, the final steady state values are the same for both methods, since the wind turbine voltages are uniformly distributed around the nominal voltage.For this scenario, the performance difference is in the steady state period prior to the change of operation point to the corner point (from 0 to 10 sec).This scenario shows that the AOLTC method not only possesses the advantages of COLTC, but also provides much better performance for all operation conditions.Comparison of results for 0-10 sec, shows that the AOLTC approach provides much better voltage profile of wind turbines closer to the nominal voltage (1.02-1.04pu) comparing to COLTC method (1.05-1.08 pu).

CONCLUSION
In this paper a new control solution is suggested for OLTC controller of main transformer of wind farms by its integration into the power plant controller that is responsible for monitoring and control of entire wind farm.Considering the wind turbine voltages in the suggested technique yields wider reactive power capability of wind farm, and hence more robustness and flexibility for connection under various grid conditions.Moreover, the suggested adaptive OLTC method provide more reliable operation point by shifting the operation voltage of wind turbines toward nominal rating, far from the voltage limits that can trigger the protection system of either wind turbine or wind farm collector network under minor disturbances at either grid or power plant.This technique utilizes the full capacity of OLTC to provide a wider range of voltage for wind turbines under a united control scheme.Comparing the performance of suggested approach with Conventional OLTC via simulations conducted in PowerFactory demonstrates the efficiency of proposed technique.

FIGURE 1
FIGURE 1 Single-line diagram of offshore wind farm

FIGURE 3
FIGURE 3 Reactive power requirement at POI.(a) PQ chart, (b) UQ chart

FIGURE 4
FIGURE 4 Comparison of conventional and adaptive OLTC

TABLE 1
Type and cross-section of submarine cables

TABLE 3
Scenario1: Comparison of COLTC and AOLTC methods