Wind turbine dynamic shading: The effects on combined solar and wind farms

In the different energy scenarios, a large role is foreseen for deployment of large-scale solar and wind energy on land and water. Morris et al.'s prediction for 2050 is wind and solar contribution of about 10 000 EJ in the global electricity production of total 41 000 EJ, meaning roughly 24%.¹ The contribution in the Global Primary Energy supply is about half of that. DNV has a higher estimate with a share of 11% for wind and 17% for solar, in the world primary energy supply.² In their New Energy Outlook 2022, Bloomberg shows an almost double contribution with a share for wind of 36% and for solar of 29% in the electricity demand. The share in the global primary energy demand will be lower, but substantial.³ However, grid congestion problems drastically slow down the implementation in several countries. Rausch et al.⁴ and Schermeyer et al.⁵ explain that the decentralized generation of renewable energy is causing stress on the electricity grid in Germany as it is often generated far from the location where the energy is needed. This mismatch between generation and demand is also mentioned by Hitaj.⁶ A solution to the problem is by using different ways of congestion management as described by Yusoff et al.⁷ and Nandini et al.⁸ Increasing the grid capacity is another solution to solve congestion problems, but this will take many years due to various existing regulations. However, some short-term solutions are possible as follows:

1. Curtailment (limiting the electricity generation from solar and wind farms to stay below the rated grid capacity). This is explained in the work by Tanno in which a curtailment solution is described that ensures equality in the generation as well as maximization of the generation.⁹ A large study in the US looked at the electricity generation of ∼2100 utility-scale wind and solar plants including the curtailment in these parks for congestion management purposes.¹⁰ Also Schermeyer et al. describe a new approach for curtailment as a solution for congestion management.¹¹ 

2. Cable pooling (using a single connection cable for more than one power plant, e.g., a solar and wind farm). Mertens, e.g., explored the possibilities and effects of cable pooling when combing solar and wind farms.¹² It shows that the combinations offers a more reliable supply and better match with demand. Golroodbari et al.¹³ looked at the techno-financial aspects of cable pooling. They conclude that combining solar and wind on one cable can be beneficial, but that it strongly depends on the nominal power delivered to the grid and the costs of the PV system.

While the first solution often requires regulatory changes before it can be implemented, the second solution, cable pooling, can be used immediately. In this option, the grid connection is shared between a solar farm and a wind farm. Weschenfelder et al. performed a literature survey on the generation profile of solar and wind energy in time and came to the conclusion that solar and wind energy are to a large extent complementary.¹⁴ This is also observed by Vattenfall in one of their combined solar and wind farms (see Fig. 1). In that specific situation, only during 2% of the time do solar and wind farms operate simultaneously at more than 75% of their full capacity.

CSWFs also have other benefits, for various actors, which are as follows:

• Power consumer: lower costs for renewable energy generation, lower grid costs

• CSWF developer: no additional grid connection waiting time and costs, lower costs for land

• Grid operator: more constant grid load, less grid reinforcement needed.

For these reasons, grid operators promote cable pooling between solar and wind farms.

However, it imposes some disadvantages. The proximity of wind turbines does not only introduce the (slowly moving) “static” shade of the turbine tower but also dynamic shades on the PV system by the moving rotor blades. The frequency of these dynamic shades is in the range of 0.25–1.2 Hz as can be found in the literature by Kerner et al. (0.25–0.9 Hz),¹⁵ Cooper (0.65–0.8 Hz),¹⁶ and Arani et al. (0.25–1.2 Hz).¹⁷ It is known that a static or slowly moving shadow does affect the power output of a PV system, but less is known about the effect of the dynamic shade.

This is acknowledged by the paper of Zakki et al.¹⁸ in which they state that depending on the maximum power point tracking (MPPT) algorithm of the inverter (indirect or direct), the algorithm might be fast, but not able to extract the maximum amount of power. For this reason, they have developed a distributed algorithm that is a combination between a direct and indirect method and is able to track the MPP at 0.3 ms with an output efficiency of 99.96%. Such MPPT could help to maximize the output of CSWP. Also Veerasamy et al.¹⁹ developed a new tracking algorithm to increase tracking speed and efficiency.

There is also some work done to predict the output of the solar part in a CSWP. Agrawal et al. take only the static part of the wind turbine shadow into account.²⁰ A similar approach is used by Mamia et al. who use a static shadow based on the blade length.²¹ Shanghavi et al. calculate the shading over 1 m² in the park every second, assuming an elliptical shadow from the turbine blades.²² Unfortunately, the paper does not mention anything about the park layout and the inverter algorithm.

In the literature, many MPPT algorithms are described, all having advantages and disadvantages under various conditions. A comprehensive review is given by Bhukya²³ including detailed descriptions of the algorithms perturb and observe (P&O),²⁴ incremental conductance(INC),²⁵ hill climbing (HC),²⁶ current sweep (CS),²⁷ extremum seeking control (ESC),²⁸ ripple correlation control (RCC)²⁹ and sliding mode control (SMC) on parameter tuning, tracking accuracy, efficiency, and complexity. Hayder et al.³⁰ describe other algorithms like artificial neural network (ANN)³¹ controllers, which show fast response time under dynamic irradiance, and they report on a new combined algorithm that is a combination between a neural network and a P&O method. The disadvantage of the particle swarm optimization (PSO) algorithm, as studied by Shaikh,³² is that it might end up in a local optimum and fail to find the global optimum. For this reason, Deng et al. suggest a new approach with time-varying compression factor.³³ Deboucha et al., on the other hand, introduce a voltage track optimizer (VTO) MPPT algorithm.³⁴ Others like Rizzo et al. propose a two-step approach to deal with the rapidly varying conditions.³⁵ An interesting approach is shown by Kumar et al. who describe an MPPT algorithm for the combined solar and wind farm.³⁶ However, all these approaches are shown for a time step of more than several hundreds of milliseconds, whereas the changes in irradiance due to passing wind turbine blades is much faster. None of the cited studies compare the behavior of the actual MPPT algorithms with measurements in the field.

In earlier small scale experiments, it has been shown that these dynamic shades can cause fluctuations in power that are higher than the power loss caused by a “static” shade of the same size.¹⁹ In addition to those power losses, the quickly fluctuating currents and voltages also lead to dynamic loads on the power electronic equipment. This can lead to a shorter lifetime of power electronic equipment, thus causing accelerated system failures and higher maintenance costs.³⁷ 

This is the first time that (i) dynamic shading effects of CSWF are investigated in the field, and (ii) effects of dynamic shading are investigated experimentally at a millisecond scale. The outline of the paper is as follow: Sec. II describes the PV plant and the monitoring system. Section III describes the results of the measurements, and the discussion of the results and conclusions are given in Sec. IV.

The measurements were done from 14 August 2022 until 2 February 2023.

The measurements were done on a solar farm consisting of 115.232 solar panels (total 38 MW_DC) and 163 string inverters (total 30 MW_AC). Each string consists of 32 in series connected 60 full-cell solar panels of 325–330 Wp each. The solar panels are mounted in portrait orientation in two rows on top of each other. The inverters have nine maximum power point trackers on which 2–3 strings are connected in parallel per MPPT. The tip height of the wind turbine is 150 m. An overview of the solar farm is given in Fig. 2.

Data from inverters are not accurate and not logged fast enough to evaluate the influence of the shade of the wind turbines on the operation of the solar farm. Therefore, a monitoring system (Fig. 3) was designed and connected to an inverter on two parts of the solar farm, namely, on PV modules with shade and on PV modules without shade of the wind turbine, as shown in Fig. 2 by a red and green area, respectively.

The purpose of the monitoring system was twofold.

The first purpose was to determine the effect of the shade of the turbines on the DC and AC yield, including the deviation from the MPP at the DC input of the inverter. This is done by logging with DC power meters of the DC voltage, current, power, and energy of the top string and the parallel connection of the top and bottom string (Fig. 3) as connected to one of the inputs of the inverter. AC power meters log the three phase AC voltage, current, power, and energy and reference cells for the in-plane irradiance and temperature (Fig. 4). The data from the meters and sensors were logged simultaneously every 6 s. The power meters were supplied by Accuenergy, consisting of AcuDC243 DC power meters and AC Acuvim II 0.1% accurate AC power meters in combination with Eleq 4Q4809 class 0.5 current transformers and Eleq SVU105 class 0.2 voltage transformers.

The second purpose of the logging was to determine the reaction of the MPPT of the inverter, the voltage of the modules, and the transients at the inverter due to the fast moving shades of the wind turbine blades. This was done by logging the voltage and current of the top string and parallel connection of the top and bottom string at the DC input of the inverter, the voltage of the PV modules at the beginning, middle, and end of the top string and the irradiance near these PV modules (Fig. 4). The signals were sampled simultaneously every millisecond, using a parallel measurement system of National Instruments consisting of a NI-Crio-9031 controller and NI-9202 I/O units in combination with P41000 Knick high voltage galvanic isolated transducers.

The accuracy of the monitoring system was confirmed with a Yokogawa WT1800 power meter and Delta SM1500-CP-30 15 kW bi-directional DC power supplies as source and load. The response time of the reference cell was determined by lightning the reference cell with a power LED connected to a NF BP4610 programmable power supply and a Rohde & Scharwz RTO2014 oscilloscope.

A camera was mounted on the tower of the wind turbine to have a general view of the shade on the PV modules and the direct/indirect light conditions. The ambient temperature was measured with an ambient temperature sensor located near the PV modules. The temperature sensors in the reference cells were taken as approximation for the temperature of the PV modules, to avoid the attachment of temperature sensors to the back of the PV modules, which could possibly weaken the backsheet of the PV modules, resulting in long-term degradation. The approximation of the module temperature by the reference cell temperature was tested in our outdoor test setup in Petten by comparing the reference cell temperature with PV module backsheet temperatures as a function of the irradiance and wind speed. The temperature of the reference cell was 0–4 °C higher than the backsheet temperature of the PV modules for irradiance from 0 to 1000 W/m². This confirms the temperature difference of a cell and the backsheet of a specially prepared PV module.³⁸ Therefore, it was concluded that the reference cell temperature is a good approximation for the cell temperature of the PV module.

Figure 5 gives the location of the monitored strings and the reference cells of the system with shade of the wind turbine. The wind turbine is located in the South–West area of these panels, in front of the area that is seen in Fig. 5. The shade of the tower and wind turbine blades are clearly visible.

In Fig. 6, the daily electricity generation per installed kWp of the top string is given for the months 8 and 9 of 2022 for the inverters with and without shade of the wind turbine. The electricity generation is almost equal for days with mainly indirect light during the time that the wind turbine can give shade on the solar panels, like on the 17th, 19th, and 26th of August and 8th, 16th, 18th, 23th, 24th, and 26th of September. This shows that the normalized electricity generation of the inverter without shade can be used as a reference system for the inverter with shade. On sunny days, a considerable loss in energy is measured due to the shade of the wind turbine. The measurements of September 1 will be explained in more detail as an example of the effect of the shade on the performance and MPPT on a sunny day.

In Fig. 7, the timeseries of the irradiance of the reference cell in the middle of the string is given in time on 1 September. The drop in irradiance just before 1000 h CET is due to the shade of the tower of the wind mill. The changes in irradiance between 1300 and 1400 h CET are due to clouds, resulting in periods with low irradiance and high irradiance by cloud enhancement.

In Fig. 8, the voltage and current output of the top string of the unshaded reference system is given in time. The voltage and current are as expected from the measured irradiance profile without shade, having only a disturbance in current and voltage by the clouds between 1300 and 1400 h CET. The lower voltage around noon is mainly because the module temperature is about 33 °C higher than in the morning.

To determine whether the inverter is operated around the MPP, Fig. 9 shows the current as a function of the voltage. In this case, current and voltage were both temperature corrected to 25 °C, which allows the plot to be compared directly with the datasheet of the solar panel. The temperature coefficient of the Impp was assumed to be equal to the datasheet coefficient of the Isc (0.048/°C), and the temperature coefficient of the Vmpp of −0.42 °C/°C was calculated from the datasheet coefficient of the Pmpp (−0.37 °C) and the assumed coefficient of the Impp. The temperature corrected IV curve is comparable with the datasheet of the solar panel, including a slight decrease in voltage at higher current by the Ohmic losses in the solar panel. Therefore, it was concluded that the temperature correction is correct and the MPPT of the inverter for the reference system without shade operates correctly.

In Fig. 10, the voltage and current output of the system with shade is given in time between 0800 and 1100 h CET, the main period with shade from the wind turbine. The voltage varies between 500 and 1170 V. In Fig. 11, the current is given as a function of the voltage between 0800 and 1100 h CET to understand the possible deviation from MPP due to the shade.

Four areas with possible deviation from the MPP are distinguished in Figs. 10 and 11, which are as follows:

1. Increase of the voltage from Vmpp to 1150 V

2. Switch from small stepwise to high continuous voltage sweep MPPT

3. Maximum voltage sweep from 500 to 900 V

4. Increase of voltage from 800 to 970 V

The effect of the shade of the blades on the MPPT will be discussed in the next paragraph using the fast log data.

In Fig. 12, the position of the shade of the wind turbine blade on the monitored string is given at 0826 h CET by the dashed circle. It is clearly seen that the monitored string is only shaded by the wind turbine blades, which turns clockwise.

First, the middle reference cell is shaded, directly followed by the reference cell in the beginning of the string (Fig. 13). The shade of the blade does not reach the reference cell at the end of the string. The shade is due to the clear sky conditions, and the irradiance drops from 600 W/m² down to 100–200 W/m² every 2.5 s. The shade of the blade is almost horizontal over the string and takes only about 0.3 s to traverse the string, resulting in some part of the string being in shaded conditions 12% of the time. The period without shade is longer, namely, about 2.2 s, and the string is unshaded 88% of the time.

In Fig. 14, the power, current, and voltage is given at the input of the inverter.

The shade of the blades result in a drop in current and power every 2.5 s, slightly later than the dip in the measured irradiance since the strings are located just below the reference cells, and the blades turn clockwise (see Fig. 4). The inverter tries to keep the voltage constant during the period of shade, but there is a limited overshoot and undershoot of around 3 and −10 V, respectively.

Under steady irradiance conditions (not shown), the MPPT of the inverter has a very regular pattern, namely, a nominal voltage for 4 s, step-up of 9 V for 2 s, back to nominal voltage for 4 s, step-down of 9 V for 2 s, back to nominal voltage for 4 s, etc.

Under non-steady irradiance conditions due to the shade of the wind turbine blades, the step size is increased to 18 V, and the voltage is constant during only 1 s, as shown by the dashed line in Fig. 14. During the period of 5 s, the voltage goes three steps up and two steps down. The overall voltage increase is 18 V per 5 s. This cycle occurs ten times in a row, by which the voltage increases from 970 to 1150 V in 50 s and stays around this high level for about 20 min.

The voltage of the string is about 180 V higher than the Vmpp of the unshaded string. In shaded conditions, the Vmpp of a string can be lower than the sum of the Vmpp of the modules when modules are bypassed to maximize the power output of the string. However, if the voltage of the string is higher than the sum of the Vmpp of modules, the string will always be out of Pmpp as evident from the curve of the power as a function of the voltage of the string.

In fact, even in unshaded conditions that occur 88% of the time, the DC power at the input of the inverter can be reduced by up to 50% due to the excess voltage (Fig. 15). Also the power drop during the period of shade of the blades is less at Vmpp than at higher voltages. Therefore, the power of the string could have been doubled if the inverter had stayed at the Vmpp of 970 V compared with 1150 V.

The reason for excess voltage of the MPPT of the inverter is most probably due to the interference of the shade of the blades and the change in voltage of the MPPT: every third step-up in voltage of the MPPT occurs just before the disappearance of the shade of a wind turbine blade (Fig. 14). The MPPT is not able to function correctly since this occurs simultaneously every 5 s.

At around 0915 h CET, the MPPT algorithm changed from stepwise changes in voltage of 18 V to large continuous changes in voltage, which reached the largest voltage change around 0930 h CET (see Fig. 10).

In Fig. 5, the location of the shade at 0930 h CET is shown. The shade of the tower is in between the reference cells at the beginning and middle of the string. These shaded PV modules are bypassed, causing around 15%–20% of the string voltage to be lost. Therefore, the Vmpp of the string without shade of the blades is about 850 V.

The wind turbine blades shade the PV modules of the monitored string from 307 to 308 s, while from 308 to 309, the PV modules are not shaded by the blades (Fig. 17). When the shade starts at 307 s at the end of the string, the inverter reacts by lowering the voltage down to 500 V with a rate of 3000 V/s to keep the current around Impp for the highest power output for strings with shaded PV modules (Fig. 16). Since the minimum voltage of the inverter is limited to 500 V, the current drops to 10A, which is below the Impp. After 0.6 s, the number of shaded PV modules decreases, and the current starts to rise at a string voltage of 500 V. When the current becomes higher than Impp, the inverter starts to increase the voltage (A in Fig. 16) back to Vmpp. In order to keep the current around Impp, the voltage of the MPPT should increase fast to follow the increasing number of PV modules without shade. However, the rise of the MPPT in voltage is too slow, causing the current of the string to increase above Impp and become constant at around 15 A (B in Fig. 16).

The maximum current of around 15 A is below the maximum current per MPPT of the inverter (26A) and, therefore, the inverter is not limiting the current. The current is limited by the maximum current of the PV modules. The Isc of the PV modules at STC is 10.3A (see the inset of Fig. 9). The irradiance without shade is around 750 W/m² (Fig. 17). Therefore, the maximum current of the parallel string at given irradiance is around 2 × 10.3 × 750/1000 = 15.4 A. The modules are not shortened but operated at a voltage below Vmpp. From the inset of Fig. 9, it is clear that the current is close to Isc for a voltage up to around 85% of the Vmpp, causing the constant current (B in Fig. 16) in case the voltage of the MPPT is more than 15% lower than Vmpp.

The modules that are shaded by the wind turbine blades are not able to produce enough current and are therefore shortened by the bypass diodes in order to prevent direct damage of these PV modules (A in Fig. 18). However, also the unshaded PV module at the end of the string has shortened substrings (B in Fig. 18), probably because of the high current by which one or more cells in the string of the module is not able to produce the high current B of Fig. 16. This can occur if a PV module is operated close to the Isc, where the smallest differences in irradiance on a cell of the module, or differences between the cells, can result in loss of a cell so that a whole substring per bypass diode is lost.

During shaded operation of the PV system due to the wind turbine blades, IR measurements show that some cells in the PV modules have an increased temperature (Fig. 19). We speculate that the cells with elevated temperature are the ones that do not generate sufficient power under the high current conditions, resulting from the too low voltage of the inverter (B in Fig. 18).

The change in the MPPT algorithm from small stepwise changes in voltage to large continuous changes in voltage occurred during the period with longer shading durations (see Fig. 17 compared with Fig. 13). This change in the MPPT algorithm for the longer shade period from 307 to 308 s of Fig. 17 improves the power output but appears to be suboptimal.

First, the drop in voltage is limited to 500 V (as given in the datasheet) by the MPPT software and probably also the hardware of the inverter, which causes the current to decrease from Impp to around 10 A. At a slightly lower voltage, the current should have been around Impp to yield higher power output.

Second, the voltage ramp-up of the inverter is not fast enough, resulting in a current higher than Impp. The increase in the voltage lags behind the shade, and, as a result, parts of the unshaded PV modules were forced in bypass, resulting in hot cells and lower power output of the total string.

In Fig. 20, an improved voltage control is proposed. The proposal is based on the assumption that the string of PV modules will have the highest power output at the Impp of the unshaded PV modules. At this condition, the unshaded PV modules have the highest power output, while the shaded PV modules are bypassed. At a higher current, the unshaded PV modules will have a lower power output, and the shaded PV modules are also bypassed, resulting in a lower power output of the string compared with the current at Impp. The same holds for a lower current. Only at a very low current, the shaded PV modules are not bypassed anymore, but at this current, the power output of the total string is lower than at Impp by the lower power output of the unshaded PV modules.

The proposed voltage is not based on a simulation but on reasoning how to keep the current at Impp, considering the shade on the PV modules by the tower and turbine blades, the IV curve of the PV module, and the measured voltage and current of the string.

The current is kept close to Impp by having a voltage lower than 500 V under severe shade conditions and a faster ramp-up of the voltage in the period that the shade disappears. This will result in a higher power output and will also prevent the issue of hot unshaded cells since the current is kept close to Impp.

The voltage increases from 800 to 970 V in period 4 of Fig. 10. In this period, the shade of the tower moves away from the PV module at the end of the string, by which less PV modules are fully shaded, and the MPPT stepwise increases the voltage. This is a correct approach as only the shade of the blades give a short disturbance in the current.

In Fig. 6, the normalized DC yield in Month 8 and 9 is given of the top string of the monitored rows of the systems with and without shading by the wind turbine. The normalized daily DC yield loss of the shaded string in the given period is between 0% and 13%, depending on the weather conditions, wind direction, and wind speed. For months 8 and 9, the total DC yield loss of the monitored string is 6.0%. The AC yield loss for the monitored inverter is 5.7%, close to the yield loss of the monitored string.

In Fig. 21, the relative normalized DC yield of the top string is given for the whole measurement period from August 2022 until February 2023. The relative yield loss is lower in the autumn and winter months but still considerable. The effect on the total loss is small since the energy production in the autumn and winter months is low. For the total measurement period, the DC and AC yield loss are 5.3% and 5.8%, respectively.

The yield loss depends on many factors, like the series connection of the PV modules, orientation of the PV farm, distance and orientation to the wind turbine, dimensions of the wind turbine, type of inverter and settings, etc. The given loss is therefore only an anecdotal result for the monitored PV farm and location and can differ for other PV farms.

Measurements in a combined solar wind farm with a logging rate of 6 s (slow log) and milliseconds (fast log) were presented. It was shown that the shadow of the wind turbine results in an energy loss of about 6% for the given period, park configuration, PV modules, inverter type, and setting.

A detailed analysis of the measurements at a millisecond logging rate of the local irradiance and PV module voltage, voltage and current at the string level, and at the input of the inverter gave clear insight into the MPPT of the inverter under various conditions. The MPPT appeared to be quite close to the MPP by changing the step size in voltage and time on constant voltage, but also by changing from stepwise MPPT to continuous larger change in voltage under fast dynamic conditions by the shadow of the wind turbine blades.

Under certain conditions, the MPPT did not function correctly. In the case of stepwise MPPT, the voltage of the MPPT became higher than the Vmpp in case the MPPT sampling frequency interfered with the frequency of the disturbance by the shade of the blades, which resulted in substantial power losses of up to 50%. In the case of continuous voltage tracking, the minimum voltage of the input of the inverter was limiting the maximum power output. Also, the voltage ramp-up rate appeared to be too slow, and, as a result, power was lost, and unshaded PV modules were forced in reverse, possibly resulting in hot cells. Directions for an improved voltage tracking pattern were proposed to avoid hot cells and to have higher power output.

As indicated, the resulting energy loss mentioned is specific for the park under investigation and only for that part of the park affected by the shadow of the rotating blades of the wind turbine. In order to determine the overall loss of the dynamic shade of the wind turbines and to determine if it is significant in a combined solar and wind farm, a time-resolved shading model will be needed in combination with a yield model for the solar farm. The development of such a model is the next step in which the results of the current work will be used to validate the model. This model can help to design and test new MPPT algorithms as well as aid the design of combined solar and wind farms.

This work is a part of the SolarWind project (Reference No. TEUE119001) and was financed by the Dutch Ministry of Economic Affairs and Climate through RVO.

The authors have no conflicts to disclose.

Nico J. Dekker: Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Lenneke H. Slooff: Project administration (lead); Supervision (lead); Writing – original draft (equal); Writing – review & editing (equal). Mark Jansen: Investigation (supporting). Gertjan de Graaff: Investigation (equal). Jaco Hovius: Resources (equal). Rudi Jonkman: Funding acquisition (supporting). Jesper Zuurbier: Resources (equal). Jan Pronk: Resources (equal).

The data that support the findings of this study are available within the article and its supplementary material.