Comparison of AC and DC Bus Interconnections of Energy Storage Systems in PV Power Plants with Oversized PV Generator
Markku Järvelä, Seppo Valkealahti
Tampere University, Electrical Engineering, Tampere, Finland
Abstract — Oversizing the photovoltaic (PV) generator improves the profitability of PV power plants, but a downside is energy losses due to power clipping. If an energy storage system (ESS) is needed, connecting it to the DC side enables the utilization of the clipped energy. We compared the AC and DC bus interconnections of ESS in a typical building block of megawatt- scale PV power plants. The DC bus interconnection increases the energy yield, but also the utilization of the ESS. The operation of the ESS is affected by the production profile of the PV generator, ESS energy capacity and DC-to-AC power ratio.
Index Terms — energy storage systems, optimal design, power plant design, PV generator.
I.INTRODUCTION
DC-to-AC power ratio is defined as the ratio of the nominal power of the photovoltaic (PV) generator to the nominal power of the inverter or transformer. Recently, oversizing PV generators has become economically rational. This is largely because PV panel prices are falling faster than other system components. Oversizing brings many economic and operational benefits. Most notably, the PV power plant will operate longer at its nominal power. This leads into higher energy yield with respect to the rated system power, more constant grid feeding power and better utilization of all AC side components, which further improves the overall yield-to-cost ratio. Currently, the levelized cost of electricity can be minimized by DC-to-AC power ratios of 1.3 – 1.6 [1]. One obvious downside of oversizing the PV generator is the clipping of power at peak hours of production. The higher the DC-to-AC power ratio, the larger is the lost energy yield.
Energy storage systems (ESSs) are needed if time shifting of production is required. Some grid codes set limits for the ramp- rates of grid-feeding power [2]. This is to mitigate the effects of fast PV power fluctuations to the grid caused by cloud shadows.
Geographic dispersion of PV panels smoothens the variation of power. However, the diameter of megawatt-scale PV generators is up to some hundreds of meters and the speed of cloud shadows can be tens of meters per second [3]. Therefore, even the largest PV generators can ramp-down power from the clear-sky nominal power to 20% in tens of seconds, which is considerably faster than the ramp-rates required in grid codes.
In practice, ESSs are needed to limit the ramp-rates in PV applications.
Electrochemical energy storages, such as lithium-ion batteries, are well suited for stationary applications. Lithium- ion batteries have fast response times, low self-discharge rates
and high efficiencies [4]. In ramp-rate control application, the required energy capacities are often rather low when compared to the power capacities. For example, when considering a 1 MW PV power plant, approximately 0.8 MW of power and 0.08 MWh of energy capacity is needed from the ESS to comply with the 10 %/min ramp-rate limit relative to the nominal power [5]. This translates to discharge C-rate of about 10. High charging and discharging currents reduce the lifetime of conventional lithium-ion batteries [6], and therefore energy capacity must be oversized to improve the lifetime of the battery. If the ESS is connected to the AC bus of the inverter or somewhere else in the PV power plant, the inverter acts as a bottleneck, and if the PV generator is oversized, part of the power must be curtailed. By interconnection to the DC bus, the excess energy capacity of the ESS can be utilized to pass on the curtailed energy, which will increase the total energy production of the PV power plant.
DC bus interconnection of the ESS requires some additional components. To prevent the back-feed of current in low irradiance conditions, it must be possible to disconnect the PV generator. In practice, this can be done by installing contactors or blocking diodes between the PV generator and the inverter.
The operating voltage range of the ESS DC/DC converter must be similar to the operating voltage range of the PV generator and the DC side of the inverter.
In this paper, we created a simulation model to analyze the operation of a PV power plant equipped with an ESS to limit the ramp-rates of the grid-feeding power. The novelty is the comparison between AC and DC bus interconnections of the ESS when the PV generator is oversized. We used real irradiance measurement data to determine the PV generator power and calculated the energy yields using combinations of different ESS energy capacities and DC-to-AC power ratios.
We also categorized the daily PV power production profiles to make results more applicable to different climatic conditions.
The DC bus interconnection of the ESS improves the energy yield, but also increases the utilization of the ESS thus reducing its lifetime. The operation of the ESS depends on the PV power production profile. Determining the optimal ESS energy capacity and DC-to-AC power ratio is straightforward for a clear-sky day. However, the ESSs with relatively small energy capacities can increase the energy yield more on a partially cloudy day than on a clear-sky day, making it challenging to the determine optimum DC-to-AC power ratio and ESS energy capacity.
II.METHODS
A. Data
This study is based on irradiance measurement data of the solar PV research power plant of the Tampere University located in Northern Europe [7]. Irradiance was measured with a photo-diode based pyranometer installed at a fixed angle of 45° facing nearly due south. The sampling frequency was 1 Hz.
Prevailing weather type in the area contains many cloudy days. Therefore, to make the simulation results universally more applicable, we identified clear-sky days to analyze their production profiles separately. Multiple methods have been proposed to detect the clear-sky periods [8], but in this paper we categorized the days by using the following simple method.
Because early morning and late evening hours are not that relevant for energy production, we analyzed six hours around the solar noon and the measured irradiance had to be 99% of the time over 90% of the expected clear-sky irradiance.
The data set consisted of 691 days measured between 1st of April and 30th of September from 2015 to 2018. A total of 47 days was categorized as clear-sky days.
B. Simulation and control models
The simulation models are presented in Fig. 1. The ESS is connected to either the DC or AC bus of the inverter. Other inverter and PV power plant topologies also exists, but these two topologies were chosen because they represent the case where there are no bottlenecks between the PV generator and the ESS, and the case where the inverter acts as a bottleneck so that the power of the PV generator is clipped during peak production hours.
Fig. 1. Single-line diagrams of the simulation models where the ESS is connected to the (a) DC and (b) AC bus of the inverter.
We used the irradiance data to calculate the output power of the PV generator by using the low-pass filter method [9]. The low-pass filter method considers the PV generator land area to calculate the effect of spatial smoothing to the irradiance. The effective irradiance Gs affecting the PV generator land area is
ܩ௦ሺݐሻ ൌ ܩሺݐሻ
߬ݏ ͳǡ ሺͳሻ
where G is the irradiance, and s is the Laplace transfer function.
߬ is the low-pass filter time constant, which is defined as
߬ ൌ ξܣ
ʹߨ ή ܽǡ ሺʹሻ
where A is the land-area of PV generator, and a is an experimental coefficient. The irradiance of 1000 W/m2 was used as a base value to calculate the PV generator power from the effective irradiance. The PV cell temperature was not considered when calculating the PV generator output power.
Simulation parameters are presented in Table I. The transformer power defines the grid connection power and it was chosen to match the power ratings of typical building blocks used in utility-scale PV power plants. The inverter power is the same as the transformer power. The PV generator power is varying so that the DC-to-AC power ratio varies between 1.0 and 2.5. The limit for positive and negative ramps is 10 %/min of the grid connection power. The minimum ESS energy capacity is the minimum level required to achieve the set ramp- rate limit and the maximum is heavily oversized to improve the lifetime of the ESS. To get technology independent results, losses are not implemented in the simulation model.
Ramp-rate control algorithms can be divided into three categories: moving average and exponential smoothing based methods, filter based methods, and ramp-rate based methods [10]. In this paper, a ramp-rate based control algorithm was used. Grid feeding power was limited to comply with the ramp- rate limit. The ramp-rate limit compliance was achieved by limiting the grid-feeding power to a level where the state of charge (SoC) of the ESS is enough to ramp down the power in case the PV generator is shut down in similar manner as presented in [11]. To maximize the energy yield, the ESS is discharged whenever possible. Because of this, the ESS is empty in the beginning and at the end of the day. Therefore, a comparison of the energy fed into the grid and the energy cycled in the ESS is possible between individual days.
The operation of the control algorithm on a partially cloudy day is exemplified in Fig. 2. The DC-to-AC power ratio is 1.5, the ESS is connected to the DC bus and its energy capacity is 0.15 MWh. If the SoC allows, the ESS is discharged (Pess) keeping the power fed to the grid (Pgrid) at rated power. When the SoC drops close to the allowed minimum value, the power fed to the grid varies according to the PV generator power (Ppvg) and the SoC of the ESS. In Fig. 2, it is worth noting how ESS smoothens the power fed to the grid keeping it at high values compared to the highly fluctuating PV generator power.
TABLE I SIMULATION PARAMETERS
Parameter Value
Inverter and transformer powers, Pinv, Pgrid 1.0 MW PV generator power, Ppvg 1.0 – 2.5 MW
Ramp-rate limit 0.1 MW/min
ESS energy capacity, Eess 0.2 – 3.0 MWh
Fig. 2. A 20-minute sample of the PV power plant’s operation on a partially cloudy day.
III.RESULTS
We analyzed the operation of the PV plant during 691 days from late spring to early autumn from the energy yield and the ESS utilization points of view. The daily variations in the production profiles were taken into consideration by analyzing clear-sky days separately and by finding the production profiles that caused the biggest differences in energy yield between the DC and AC bus interconnections of the ESS.
It should be noted that due to the fixed mounting angle of pyranometers, the peak value of the clear-sky day’s irradiance in June and July is on average 920 W/m2. In addition, the power of the PV generator was calculated only based on the land area of the PV generator and the measured irradiance so that the temperature of the PV panels was not considered. Therefore, when interpreting the results, the DC-to-AC power ratio should be scaled according to location specific temperature and irradiance conditions.
A. Energy yield
Differences in energy fed to the grid between the ESSs connected to the DC (EDC) and the AC (EAC) bus were analyzed by comparing the relative and absolute values. The relative difference was examined by the ratio between energy yields (EDC/EAC) and the absolute difference is the difference between energy yields (EDC-EAC).
The average daily energy yield differences between the DC and AC bus interconnections are presented in Fig. 3 as a function of DC-to-AC power ratio for different ESS energy capacities during the whole observation period. With 1.0 power ratio, the difference between DC and AC bus interconnections is practically nonexistent. The power ratio must be around 1.1 to have a noticeable difference in energy yield. This is because the clear-sky irradiance value peaks at 920 W/m2 in June and July, and therefore the tiny difference with 1.0 power ratio is mainly due to the over irradiance events caused by cloud enhancement phenomenon. When the power ratio is increased,
the energy yield difference increases almost linearly until the ESS energy capacity starts to limit further increase.
Fig. 3. Daily average energy yield difference between the DC and AC bus interconnections of the ESS as a function of DC-to-AC power ratio for different ESS energy capacities during the whole observation period.
The energy yield ratios between the DC and AC bus interconnections of the ESS are presented in Fig. 4 as a function of DC-to-AC power ratios for different ESS energy capacities.
The yield ratio forms an envelope curve for the maximum feasible energy capacity of the ESS with increasing DC-to-AC power ratio. As is also visible in Figs. 3 and 4, the relative amount of energy gained by connecting the ESS to the DC bus decreases with increasing size of the ESS at high DC-to-AC power rations, i.e., the profitability of the DC bus connected ESS decreases with increasing storage capacity.
Fig. 4. EDC/EAC yield ratios as a function of DC-to-AC power ratio for different ESS energy capacities during the whole observation period.
B. Production profile
The analysis of energy yields of the whole observation period gives a good general assessment of how the DC bus interconnection compares to the AC bus interconnection under Nordic conditions. However, the production profile of the PV generator greatly affects the operation of the PV power plant.
During clear-sky days, production profiles vary only slightly, which is mainly due to seasonal changes in sun path. During a partially cloudy day, the number of irradiance transitions can be several hundred [3], which will cause numerous charge-discharge cycles throughout the day.
The average daily energy yield differences between the AC and DC bus interconnections on clear-sky days of the observation period are presented in Fig. 5 for different ESS energy capacities as a function of DC-to-AC power ratio. The DC-to-AC power ratio must be at least 1.1 before there is any noticeable difference in energy yields. Further increase of the power ratio increases the yield difference almost linearly until the energy capacity of the ESS starts to limit the energy yield.
On clear-sky days, the maximum possible daily energy yield difference between the two topologies is close to the ESS energy capacity. The additional energy yield is also directly proportional to the energy capacity of the ESS at high DC-to- AC power ratios.
Fig. 5. Daily average energy yield difference between the DC and AC bus interconnections of the ESS as a function of DC-to-AC power ratio for different ESS energy capacities on clear-sky days of the observation period.
The average daily energy yield ratios between DC and AC bus interconnections on clear-sky days of the observation period are presented in Fig. 6. For clear-sky days, a distinct DC-to-AC power ratio can be detected, where the EDC/EAC yield ratio reaches its maximum. Therefore, for regions with mostly clear-sky days, it is straightforward to determine the optimum ESS energy capacity for each DC-to-AC power ratio of the PV plant to maximize the energy yield. Larger ESS energy capacity would not affect the energy yield, but it might improve the lifetime of the ESS.
Fig. 6. EDC/EAC yield ratios as a function of DC-to-AC power ratio for different ESS energy capacities during clear-sky days of the observation period.
On partially cloudy days, the PV generator output power can ramp up and down several times per day and the ESS is utilized throughout the day to compensate the power fluctuations. Fig. 7 presents the maximum daily energy yield difference between DC and AC bus interconnections for different ESS energy capacities as a function of DC-to-AC power ratio. With 1.0 DC-to-AC power ratio, DC and AC bus interconnections have a small, but noticeable difference in the energy yields. This is due to several over irradiance events on a single day caused by cloud enhancement phenomenon. The maximum energy yield difference follows quite closely the yield difference on clear- sky days (the envelop curve in Fig. 5) until the ESS energy capacity is reached. However, unlike on clear-sky days, further increase of the DC-to-AC power ratio still increases the yield difference. This is due to power fluctuations on partially cloudy days, which allows the ESS to discharge during the periods of low insolation. Therefore, the ESSs with relatively small energy
Fig. 7. Maximum daily energy yield difference between the DC and AC bus interconnections as a function of DC-to-AC power ratio for different ESS energy capacities during the whole observation period.
capacities can have several charge-discharge cycles per day and the yield increase can be substantially larger than on clear-sky days. The results presented in Figs. 3 to 7 demonstrate that the local daily PV power production profiles must be considered closely when sizing the ESS.
The operation of a 0.2 MWh ESS connected to the DC and AC bus of the inverter on a partially cloudy day is compared in Fig. 8. The DC-to-AC power ratio of the PV plant is 1.5. The DC bus interconnection of the ESS increased the daily energy yield by 1.07 MWh. The PV generator power fluctuations are distributed quite evenly throughout the peak production hours, and therefore there are several charge-discharge cycles. During this day, the ESS reached its maximum energy capacity once so that 0.19 MWh of energy had to be curtailed. The energy cycling in the ESS connected to the DC bus was 1.47 MWh and in the ESS connected to the AC bus 0.90 MWh.
Fig. 8. Operation of the ESS connected to the DC and AC bus of the inverter on a partially cloudy day on 17th of June 2015. The ESS energy capacity is 0.2 MWh and the DC-to-AC power ratio is 1.5.
Even though ESSs with relatively small energy capacities can increase the energy yield more on partially cloudy day, the total daily energy yield of such PV power plants cannot be larger than on clear-sky days. To maximize the total energy yield, the PV power plant should operate against its power limit and the ESS SoC should be full in the evening before the PV generator power drops below the inverter power. On partially cloudy days, this can be achieved if the PV generator power fluctuations are distributed evenly throughout the peak production hours. On clear-sky days, the PV power plant is obviously operating against its power limit, and therefore the total grid-feeding energy is at least the same as on partially cloudy days, while the share of the lost energy is different.
C. Energy storage system utilization and cycling
In ramp-rate control applications, the minimum energy capacity requirement can be derived from the worst fluctuation model and it varies as a function of ramp-rate limit [5]. When the ESS is connected to the AC bus of the inverter, it is only
utilized to compensate the fast PV power fluctuations. In this case, the ESS utilization depends on the production profile, ramp-rate limit, and the control algorithm.
Fig. 9 presents the average and maximum energy cycled daily through the ESS connected to the AC bus as a function of the DC-to-AC ratio during the whole observation period and the average energy cycled also during clear sky days. In addition, average and maximum values are shown for ramp-rate limits of 5 and 20%/min of the rated power during the whole observation period. The effect of different irradiance profiles is clearly visible. On clear-sky days, the utilization of ESS is mainly due to preparation for potential power fluctuations. Therefore, regardless of the power ratio, the average cycled energy on a clear-sky day is approximately the amount of energy needed to ramp-down the grid-feeding power in case the PV generator is suddenly disconnected. The average daily energy cycled through the ESS was approximately 0.3 MWh for the 10% per minute ramp rate limit and. This energy is approximately 4 to 5% of the daily average grid feeding energy. The maximum energy cycling for one day was 1.0 MWh for the 10%/min ramp-rate limit at power rations above 1.5. The smoother the ramp-rate limit requirement is, the more the ESS is utilized.
Fig. 9. Average and maximum energy cycled daily through the AC bus interconnected ESS as a function of the DC-to-AC ratio for different ramp rate limits during the whole observation period. The average daily energy cycled through the ESS on clear-sky days is also shown for the ramp-rate limit of 10%/min.
Fig. 10 compares the average energy cycled daily through the ESS connected to the DC and AC bus as a function of DC-to-AC power ratio for different ESS energy capacities. The energy capacity of the ESS connected to the AC bus does not affect the amount of cycled energy, which is almost constant as a function of the DC-to-AC power ratio. With 1.0 power ratio, the difference in cycled energy between the DC and AC bus interconnection is almost nonexistent, but with increasing power ratio, the DC bus interconnection increases the utilization of the ESS. The utilization of the DC bus interconnected ESS does not increase linearly with increasing
energy yield. That is, the cycled energy is less than the sum of cycled energy needed to compensate the power fluctuations of the PV generator (AC bus interconnection) and the increase of the total energy yield (AC vs. DC bus interconnection). This can be observed from Fig. 8, where the DC bus interconnection increased the energy yield 1.07 MWh, but the ESS utilization increased only 0.58 MWh. This is because rather than ramping the grid-feeding power up and down, the excess energy stored in the ESS can be used to keep the grid-feeding power constant.
Fig. 10. Average energy cycled daily through the ESS connected to the DC and AC bus of the inverter as a function of DC-to-AC power ratio for different ESS energy capacities during the whole observation period.
IV.CONCLUSION
We have compared the operation of AC and DC bus interconnections of the ESS in PV power plants with oversized PV generator from the energy yield and ESS utilization point of views. We have created a simulation model of a typical building block used in utility scale PV power plants where the ESS is primarily used to compensate for the PV generator power fluctuations when limiting the ramp rates in power fed to the grid. Actual irradiance measurement data was used to determinate the PV generator power. We varied the DC-to-AC power ratio and ESS energy capacity, and we studied the operation of the ESS with different production profiles.
If the PV generator is oversized, connecting the ESS to the DC bus will increase the total energy yield. Both the EDC/EAC
yield ratio and EDC-EAC yield difference depends on daily production profile type, DC-to-AC power ratio and ESS energy capacity. They both increase with increasing energy capacity of the ESS at high DC-to-AC power ratios. For clear-sky days, it is possible to determine an optimum ratio between the ESS energy capacity and PV generator power for each DC-to-AC power ratio. However, on partially cloudy days an ESS with a relatively small energy capacity increases the energy yield more than on clear-sky days. This is because on partially cloudy days, there can be several charge-discharge cycles. The DC bus
interconnection increases the utilization of the ESS, which will affect its lifetime. In addition, to prevent the back-feed of current to the PV generator, DC bus interconnection of the ESS requires some additional components to the DC side.
REFERENCES
[1] K. Zipp, “Why array oversizing makes financial sense,” Solar Power World, 2018.
[2] V. Gevorgian, S. Booth, V. Gevorgian, and S. Booth,
“Review of PREPA Technical Requirements for Interconnecting Wind and Solar Generation Review of PREPA Technical Requirements for Interconnecting Wind and Solar Generation,” no. November, 2013.
[3] K. Lappalainen and S. Valkealahti, “Analysis of shading periods caused by moving clouds,” Sol.
Energy, vol. 135, pp. 188–196, 2016.
[4] R. Amirante, E. Cassone, E. Distaso, and P.
Tamburrano, “Overview on recent developments in energy storage : Mechanical , electrochemical and hydrogen technologies,” Energy Convers. Manag., vol. 132, pp. 372–387, 2017.
[5] J. Marcos, O. Storkël, L. Marroyo, M. Garcia, and E.
Lorenzo, “Storage requirements for PV power ramp- rate control,” Sol. Energy, vol. 99, pp. 28–35, Jan.
2014.
[6] S. M. Rezvanizaniani, Z. Liu, Y. Chen, and J. Lee,
“Review and recent advances in battery health monitoring and prognostics technologies for electric vehicle ( EV ) safety and mobility,” J. Power Sources, vol. 256, pp. 110–124, 2014.
[7] D. Torres Lobera, A. Mäki, J. Huusari, K.
Lappalainen, T. Suntio, and S. Valkealahti,
“Operation of TUT solar pv power station research plant under partial shading caused by snow and buildings,” Int. J. Photoenergy, vol. 2013, 2013.
[8] M. J. Reno, I. Lillo-bravo, and M. A. Silva-p,
“Identifying periods of clear sky direct normal irradiance,” vol. 113, pp. 756–763, 2017.
[9] J. Marcos, L. Marroyo, E. Lorenzo, D. Alvira, and E.
Izco, “From irradiance to output power fluctuations:
the PV plant as a low pass filter,” Prog. Photovoltaics Res. Appl., vol. 19, no. 5, pp. 505–510, Aug. 2011.
[10] S. Sukumar, M. Marsadek, K. R. Agileswari, and H.
Mokhlis, “Ramp-rate control smoothing methods to control output power fluctuations from solar photovoltaic ( PV ) sources — A review,” J. Energy Storage, vol. 20, no. April, pp. 218–229, 2018.
[11] I. de la Parra, J. Marcos, M. García, and L. Marroyo,
“Control strategies to use the minimum energy storage requirement for PV power ramp-rate control,” Sol.
Energy, vol. 111, pp. 332–343, Jan. 2015.
