Loss Model for Improved Efficiency
Characterization of DC Coupled PV-Battery System
Converters
Jolien Despeghel, Jeroen Tant and Johan Driesen
Department of Electrical Engineering (ESAT-ELECTA)
KU Leuven - EnergyVille
Leuven, Belgium
[email protected] Abstract—The use of battery energy storage systems (BESS) The measurement procedure to determine the efficiency of
is emerging in residential settings, complementary to the existing PV inverters is well established, by means of the European
presence of photovoltaic (PV) systems. BESS can play a vital role standard EN 50530 [7]. As a result, the information available
in distributed energy generation. The latest efficiency guidelines
for PV-battery systems, the Effzienzleitfaden presented by the on datasheets has evolved from a single (maximum) efficiency
German BVES only characterize the efficiency for singular power in the beginning to a more detailed efficiency curve. In Europe,
flows, which falls short in correctly representing the operational the European efficiency, ηEU , was introduced in 1991 [8]. Now
reality of the combinatory power flows present in a DC coupled with the emergence of PV-battery systems, there is a need to
PV-battery system. A loss model is proposed. The experimental define similar standards to evaluate their performance in terms
case study of a commercially available DC coupled PV-battery
converter system allows to perform a least-squares fitting using of efficiency.
experimental measurements. The proposed loss model accurately The challenge in developing a standard for the efficiency of
characterizes the efficiency of the converter both in modes of a PV-battery system, originates from the fact that the efficiency
singular and combinatory power flows. Comparison between of the system is dependent on the load. One system can be
the proposed model and a quadratic loss model based on the installed in different households, each with their own individ-
datasheet has shown that the data available on a datasheet is
insufficient to accurately model combinatory power flow losses. ual consumption profile. The different operational reality will
Index Terms—energy efficiency, loss measurement, energy have an impact on the overall system efficiency.
storage, converters The process of standardizing the efficiency of PV-battery
systems is still ongoing, with different concepts or approaches
I. I NTRODUCTION being proposed. In [9], an efficiency for storage systems is
presented similar to the European efficiency for PV systems.
Climate change is one of the biggest challenges faced by hu-
Aside from the power, the operational mode of the system is
manity to date. After the Paris Agreement, commitments have
taken into account to determine a weighted efficiency. Instead
been made, specified in nationally determined contributions
of the efficiency, a new performance indicator can also be
(NDCs), to limit global temperature rise to well below 2 °C
derived as in [10]. This indicator, the system performance
[1]. This calls for a decrease in CO2 emissions. One approach
index (SPI) is based on cost savings of a real versus an ideal
to achieve this goal is the energy transition. In Europe, part of
lossless system. Of course, to calculate cost savings, there is a
the implementation is found in the Revised Renewable Energy
need to determine the converter losses of a PV-battery system.
Directive, setting a target of at least 32% of renewable energy
This led to the creation of the German “Effizienzleitfaden für
by 2030 [2].
PV-Speichersysteme” in 2017, recently updated to version 2.0
Momentum is slowly building, and in combination with
[11] and made available in English as the “Efficiency guideline
falling costs of PV systems [3], [4], distributed generation
for PV storage systems”.
is on the rise. In Germany, residential consumers are
The challenge is more prominent, in DC coupled systems. A
incentivized to maximize self-consumption, which leads to
DC coupled PV-battery system (Fig. 1a) consists of two dc-dc
the use of a storage system [5]. At the same time, concerns
converters connecting both the PV panels and the battery of the
are voiced regarding the capacity of the distribution grid, for
system to a common DC bus, followed by a dc-ac converter or
which batteries could offer relief [6]. This tendency to install
inverter. Often it is not possible to measure the power flows
PV-battery systems gives way to new system configurations
at the DC bus point as the different converter components
and respective converters.
of the device are enclosed. The problem arises because of
the combined or combinatory power flows present in a DC
Interne Fondsen KU Leuven / Internal Funds KU Leuven coupled system, whereas in an AC coupled system (Fig. 1b),
978-1-7281-4878-6/19/$31.00 ©2019 IEEE 4740
Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2024 at 17:01:39 UTC from IEEE Xplore. Restrictions apply.
coupled. The efficiency is determined for different system
modes: charge (PV2BAT, AC2BAT), discharge (BAT2AC) and
direct consumption or grid injection (PV2AC). Specifically for
DC coupled battery systems, these singular power flows are
insufficient to determine the losses and respective efficiency,
as combinatory power flows occur quite frequently. Therefore,
in this paper a loss model is developed to characterize the
efficiency of the combinatory power flows.
The use of a loss model is not uncommon to characterize a
converter. In [12], a mathematical model for an inverter was
derived, where three terms of losses were identified: a constant
self-consumption loss, a loss linearly proportional to the power
and a loss proportional to the square of the power. This model,
where the loss PLOSS is a function of the input power PIN
is defined as
PLOSS = a + bPIN + cPIN 2 . (1)
(a) DC coupled system (b) AC coupled system
A similar quadratic mathematical model is discussed in [13],
though there the loss is a function of the output power.
Figure 1: Overview of PV-battery system topologies, including Starting from inverter loss models, where the loss is solely
singular power flows dependent on the input/output power, models have been de-
veloped which also take into account the influence of the dc
input voltage on the losses [14], [15]. For the purpose of this
only singular power flows occur. In an AC coupled system,
paper, the input voltages are assumed constant and the voltage
both the PV panels and battery are separately connected to
dependency is not taken into account.
the grid by means of an inverter, only encountering singular
power flows, thus the efficiency of the individual converters B. Mathematical models
can be determined, as is the case for PV systems.
This paper begins by developing a loss model to characterize The possible power flows in a DC coupled PV-battery
the converter used in a DC coupled PV-battery system. As system can be categorized as singular or combinatory power
an approximation of the proposed loss model, two additional flows. PV2AC, PV2BAT and BAT2AC are singular power
models are established based on the singular power flow data flows (see Fig. 1a), whereas PVB2G and PV2BG are combina-
available in a datasheet. It goes on to present an experimental tory power flows. In the case of PVB2G and PV2BG, PV2AC
case study in which measurements are performed to capture is respectively combined with BAT2AC and PV2BAT.
the power flows in various operating areas, both of singular and The loss model (1) is implemented as a building block in the
combinatory nature. The parameters of the proposed quadratic model of the DC coupled PV-battery system, which consists of
loss model are then determined by means of a least-squares three converters. The aim is to create a model which models
fitting. Finally, the proposed loss model is compared to the the losses as a function of the PV power PP V and the battery
two additional models in terms of their ability to accurately power PBAT , represented by
quantify the losses specifically for combinatory power flows.
PLOSS = f (PP V , PBAT ). (2)
II. DC COUPLED PV- BATTERY CONVERTER SYSTEM
MODEL The battery power PBAT indicates charging when PBAT >
0 and discharging when PBAT < 0.
A. Background
Equation (3) defines the proposed model, where charge
In the efficiency guideline [11], three singular power flows power PC and discharge power PD are assigned separate pa-
(PV2AC, PV2BAT and BAT2AC) are determined to charac- rameters, due to the asymmetric charge and discharge behavior
terize the efficiency. A singular power flow is defined as a of the battery. The constant parameter am will be different
power flow with a single source, where the power originates depending on the mode the system is in.
and a single sink, where the power is received. For PV2AC
and PV2BAT, this entails that the power originating from the PLOSS = am + b1 PP V + b2 PC + b3 PD +
PV panels respectively flows to the AC grid or charges the
c1 PP V 2 + c2 PC 2 + c3 PD 2 + d(PP V − PBAT )2 (3)
battery. In the case of BAT2AC, the battery is discharged and
power flows to the AC grid. To the author’s knowledge the where
guideline has not yet been adopted as a standard. (
The guideline distinguishes between three different types of PC , in modes PV2BAT and PV2BG
PBAT = (4)
system toplogies: AC coupled, DC coupled and PV generator PD , in modes BAT2AC and PVB2G
4741
Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2024 at 17:01:39 UTC from IEEE Xplore. Restrictions apply.
The set of parameters am through d can be fitted using
experimental measurement data.
In case the data only contains measurements of singular
power flows, the parameter d cannot be determined. The
least-squares fitting will result in an underdetermined system,
because for PV2AC and BAT2AC, either PBAT = 0 or
PP V = 0 and for PV2BAT, the term d(PP V − PBAT )2 will
be approximately zero.
As a datasheet of a DC coupled PV-battery system only
contains singular power flow data, additional models need to
be established. These models will approximate the proposed
loss model. Based on the datasheet there are two possible
options. The first, is a superposition of the singular quadratic
loss models. The second, is the fitting of a new loss model
based on the dataset of the singular power flows.
The superposition model combines the loss model of
PV2AC and PV2BAT or BAT2AC to achieve a loss model
for PV2BG and PVB2G. An example is given for PVB2G,
PLOSS,P V B2G,super =
Figure 2: Overview of the measurement setup and measured
aP V 2AC + bP V 2AC PP V + cP V 2AC PP V 2 + quantities for the case study of a DC coupled PV-battery
aBAT 2AC + bBAT 2AC PD + cBAT 2AC PD 2 , (5) system
with (a, b, c)P V 2AC and (a, b, c)BAT 2AC the parameters of the
quadratic loss model as defined in (1). tracker (MPPT) of the converter can find the maximum power
The fitting of the loss model based on the singular power point (MPP). Throughout the measurements, the MPP voltage
flows of the datasheet, or fitted model, results in the parameters VM P P was kept constant at 450 V.
of The aim of this setup is twofold. Firstly, the system is
PLOSS = am + b1 PP V + b2 PC + b3 PD + measured following the efficiency guideline [11], this allows to
calculate the losses and then determine the efficiency curve for
c1 PP V 2 + c2 PC 2 + c3 PD 2 . (6) each of the singular power flows. The experimental conditions
III. E XPERIMENTAL CASE STUDY are specified in the efficiency guideline, where the state of
charge is dependent on the system mode and the temperature
For the case study, a commercially available DC coupled
is kept within the prescribed range of 25 °C ± 5 °C. Secondly,
PV-battery system was investigated. To simulate the converter
the system is measured while varying the input powers in steps
system in normal operating conditions (as in Fig. 3), the PV
of 10%. This results in the fitting of the proposed loss model,
generation and AC load are emulated. The measurement setup
including the combinatory power flows.
thus consisted of the converter system (including the Li-ion
BESS), DC sources to emulate the I-V curve of the PV panels A. Efficiency measurements of singular power flows
and an AC load. The measurements were performed with
The efficiency of singular power flows is defined based on
the Yokogawa WT1806E power analyzer. An overview of the
measurements according to the efficiency guideline [11]. The
setup is presented in Figure 2.
losses are calculated, resulting in the efficiency
The specifications of the DC coupled PV-battery system are
listed in Table I. Based on the nominal operating conditions of PIN − PLOSS
η= . (7)
the system, the valid operating area can be defined (see Fig. PIN
3). For each singular power flow an efficiency curve is determined
Table I: Specifications of the case study system by means of a least-squares fitting of the losses as in (1).
The measurement procedure to define the efficiency of
Power [W] the singular power flows PV2AC, BAT2AC and PV2BAT is
PP V,nom 5200
PAC,nom 3680 based on the efficiency guideline. Input and output powers are
PBAT,nom 1500 measured for an input power of subsequently 100% (nominal),
75%, 50%, 30%, 25%, 20%, 10% and 5%. The nominal input
For the emulation of the PV generation two Delta Elec- power is dependent on the mode the system is in. For PV2BAT
tronika DC sources of type SM660-AR-11 (SM3300 series) and BAT2AC, the nominal input power will be limited to the
are used, which includes an interface to allow PV simulation. nominal battery power, while for PV2AC this is limited to the
Given the emulated I-V curve, the maximum powerpoint nominal AC power (see Table I).
4742
Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2024 at 17:01:39 UTC from IEEE Xplore. Restrictions apply.
(a) Dataset for least-squares fitting of singular power flows (b) Dataset for least-squares fitting of combinatory power flows
Figure 3: Overview of the experimental case study
Table II: Efficiency measurements of singular power flows
PV2AC, BAT2AC and PV2BAT
PP V /PP V 2AC,nom (–)
PV2AC
0.05 0.10 0.20 0.25 0.30 0.50 0.75 1.00
Efficiency η (%)
83.7 91.6 94.8 95.4 95.7 96.3 96.4 96.3
PBAT /PBAT 2AC,nom (–)
BAT2AC
0.05 0.10 0.20 0.25 0.30 0.50 0.75 1.00
Efficiency η (%)
78.3 86.9 91.8 92.9 93.9 94.9 95.3 95.5
PP V /PP V 2BAT,nom (–)
PV2BAT
0.05 0.10 0.20 0.25 0.30 0.50 0.75 1.00
Efficiency η (%)
64.6 80.4 89.4 90.8 92.4 94.3 95.4 95.9
Figure 4: Model-based efficiency curve and measured efficien-
The measured operating points for the three singular power cies of singular power flows PV2AC, PV2BAT and BAT2AC
flows are displayed in Fig. 3a. The resulting efficiency of
these points is presented in Table II. As Fig. 4 shows, the
efficiency curve based on the quadratic loss model is capable Table III: Overview of quadratic loss model errors for losses
of accurately representing the measured efficiency. and efficiency
Error Ploss η
B. Fitting of the proposed loss model Maximum (abs.) 15.57 W 1.94%
A measurement procedure is required to collect the nec- Minimum (abs.) 0.031 W 2.5 × 10−5 %
Average (abs.) ± 0.83 W ± 0.072%
essary data to fit the quadratic loss model, as proposed in Maximum (rel.) 18.26% 2.28%
(3). The measurement intervals as presented in the efficiency Average (rel.) ± 1.26% ± 0.078%
guideline [11] are rather sparse. Because of the combinatory
nature of the power flows, the complete operating area is
measured in a more detailed manner.
For this purpose, the PV power varies from 100% (or in Fig. 6. To verify whether or not the model is a good
nominal) power to 0% and the battery varies from nominal fit, the modelling errors are calculated. The set of measured
discharge to charge power, both in intervals of 10%. The datapoints is compared to the resulting model values. Table
measurement data are shown in Fig. 3b. III presents the maximum, minimum and average absolute
The least-squares fit of the quadratic loss model together error, as well as the maximum and average relative error for
with the measured losses of the combinatory power flows both the losses and the efficiency. With respect to the average
PVB2G and PV2BG are presented in Fig. 5. The correspond- measurement uncertainty of 7.75 W and 0.38% the proposed
ing model and measurements for the efficiency are presented model is capable of accurately modeling the converter system.
4743
Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2024 at 17:01:39 UTC from IEEE Xplore. Restrictions apply.
loss error, the absolute efficiency error is calculated. To
assess the weight of the error the measurement uncertainty
is included.
The mode PVB2G is displayed in Fig. 7. The superposition
model (red) clearly overestimates the losses. This is in part due
to the double counting of constant losses when superposing
separate loss models. The proposed model (green) is able to
achieve a tight fit with regards to the absolute losses, as well
as the absolute efficiency, and stays well within measurement
uncertainty (red band). While the fitted model (blue) comes
close, it is not capable of capturing the combinatory power
flows as accurately as the proposed model. A similar conclu-
sion can be drawn for the mode PV2BG (see Fig. 8). However,
for PVB2G, an irregularity occurs in the otherwise minimal
loss error. This loss error, which exceeds the measurement
uncertainty, is due to a change in the internal DC bus voltage,
indicating a significant influence of the voltage on converter
Figure 5: Verification of the modeled losses (mesh) with losses. Quantitatively, the maximum loss error relative to the
respect to the measured losses (dots) of the combinatory power maximum measured loss equals 22.67%, 26.51% and 6.68%
flows PVB2G and PV2BG for the superposition, fitted, and proposed model, respectively.
The reason why the proposed model is better able to
represent combinatory losses than the fitted model, is largely
due to the design of the proposed model. When comparing
(3) and (6), the proposed model includes an additional term,
d(PP V − PBAT )2 , to properly characterize the combinatory
power flows. The losses are still sufficiently different and as
such the datasheet lacks sufficient data to model the operation
of a DC coupled PV-battery converter system.
V. C ONCLUSION
A loss model has been proposed to estimate the losses of
combinatory power flows. The mathematical loss model was
based on the loss model of an inverter, which only encounters
singular power flows. This simple model acted as a building
block for the proposed loss model. Additionally, a quadratic
term of the difference of the model input powers was included.
After creating two models, one for singular power flows and
the proposed loss model for combinatory power flows, both
Figure 6: Verification of the modeled effciency (mesh) with
models were compared. The comparison was based on an
respect to the measured efficiency (dots) of the combinatory
independent dataset, where two singular power flow models
power flows PVB2G and PV2BG
were compared to the proposed model. The superposition
model, where singular power flow loss models are superposed
and the fitted model, where a loss model is fitted on a dataset
IV. C OMPARISON OF LOSS MODELS FOR COMBINATORY
solely consisting of singular power flows.
POWER FLOWS
The proposed loss model is capable of accurately modeling
Now that the proposed loss model has been fitted to better the losses throughout the operating area of a DC coupled PV-
estimate the combinatory power flow losses, a comparison battery converter system. In comparison to the approximating
is made between the loss model and the two models based models, the proposed model achieves a better estimation of the
on the singular power flows found in the datasheet. For the losses than the fitted model, with an error up to 23% versus
comparison an independent dataset is used. The different 7%. The superposition model is disregarded for future use, due
models for loss estimation are referred to as proposed model, to its global overestimation of the losses. To conclude, the data
superposition model and fitted model. provided in the datasheet is found to be insufficient, leading to
Both combinatory modes PVB2G and PV2BG are exam- a mismatch when modeling combinatory power flow losses.
ined using the three models. First the losses are calculated
according to the models and compared to the measured losses, R EFERENCES
resulting in the absolute loss error. In addition to the absolute [1] UNFCCC, “Paris agreement,” 2015.
4744
Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2024 at 17:01:39 UTC from IEEE Xplore. Restrictions apply.
Figure 7: Comparison of the proposed (green), superposition Figure 8: Comparison of the proposed (green), superposition
(red) and fitted (blue) model with the measured losses for (red) and fitted (blue) model with the measured losses for
PVB2G - including measurement uncertainty (red band) PV2BG - including measurement uncertainty (red band)
[2] E. Parliament and E. Council, “Directive (eu) 2018/2001 of the european inverters used in low voltage grid-connected PV systems: Inverters below
parliament and of the council of 11 december 2018 on the promotion 10 kW,” Renewable and Sustainable Energy Reviews, vol. 13, pp. 1541–
of the use of energy from renewable sources (text with eea relevance.),” 1550, aug 2009.
2018. [9] A. Piepenbrink, “Eu efficiency for home storage systems-a new and
[3] J. Hoppmann, J. Volland, T. S. Schmidt, and V. H. Hoffmann, “The simple procedure,” in 10th International Renewable Energy Storage
economic viability of battery storage for residential solar photovoltaic Conference (IRES 2016), 2016.
systems–a review and a simulation model,” Renewable and Sustainable [10] J. Weniger, T. Tjaden, J. Bergner, and V. Quaschning, “Emerging per-
Energy Reviews, vol. 39, pp. 1101–1118, 2014. formance issues of photovoltaic battery systems,” in Proceedings of the
[4] J. Weniger, J. Bergner, T. Tjaden, and V. Quaschning, “Economics 32nd European Photovoltaic Solar Energy Conference and Exhibition,
of residential pv battery systems in the self-consumption age,” in Munich, Germany, pp. 20–24, 2016.
29th European Photovoltaic Solar Energy Conference and Exhibition, [11] BVES - Bundesverband Energiespeicher e.V. and BSW - Bundesverband
pp. 2936–2938, 2014. Solarwirtschaft e.V.r, “Effizienzleitfaden für PV-Speichersysteme,” tech.
[5] M. Braun, K. Büdenbender, D. Magnor, and A. Jossen, “Photovoltaic rep., BVES and BSW Solar, April 2019.
self-consumption in germany: using lithium-ion storage to increase self- [12] K. Peippo and P. D. Lund, “Optimal sizing of solar array and inverter in
consumed photovoltaic energy,” in 24th European Photovoltaic Solar grid-connected photovoltaic systems,” Solar Energy Materials and Solar
Energy Conference (PVSEC), Hamburg, Germany, 2009. Cells, vol. 32, pp. 95–114, 1994.
[6] J. Moshövel, K.-P. Kairies, D. Magnor, M. Leuthold, M. Bost, S. Gährs, [13] A. El -Aal, J. Schmid, J. Bard, and P. Caselitz, “Modeling and optimizing
E. Szczechowicz, M. Cramer, and D. U. Sauer, “Analysis of the maximal the size of the power conditioning unit for photovoltaic systems,” Jour-
possible grid relief from pv-peak-power impacts by using storage sys- nal of Solar Energy Engineering-Transactions of the ASME, vol. 128,
tems for increased self-consumption,” Applied Energy, vol. 137, pp. 567– no. 1, pp. 40–44, 2006.
575, 2015. [14] F. P. Baumgartner, “Euro realo inverter efficiency: DC-voltage depen-
[7] R. Bründlinger, N. Henze, H. Häberlin, B. Burger, A. Bergmann, and dency,” in 20th European Photovoltaic Solar Energy Conference, 2005.
F. Baumgartner, “pren 50530–the new european standard for perfor- [15] A. Driesse, P. Jain, and S. Harrison, “Beyond the curves: Modeling the
mance characterisation of pv inverters,” in 24th European Photovoltaic electrical efficiency of photovoltaic inverters,” in 33rd IEEE Photovolatic
Solar Energy Conference, Hamburg, Germany, 2009. Specialists Conference, IEEE, 2008.
[8] V. Salas and E. Olı́as, “Overview of the state of technique for PV
4745
Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2024 at 17:01:39 UTC from IEEE Xplore. Restrictions apply.
