You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301352728

Assessing the influence of the temporal resolution of electrical load and PV


generation profiles on self-consumption and sizing of PV-battery systems

Article  in  Applied Energy · July 2016


DOI: 10.1016/j.apenergy.2016.04.050

CITATIONS READS

143 367

4 authors, including:

Tobias Beck Hendrik Kondziella


Bosch University of Leipzig
2 PUBLICATIONS   207 CITATIONS    19 PUBLICATIONS   624 CITATIONS   

SEE PROFILE SEE PROFILE

Thomas Bruckner
University of Leipzig
171 PUBLICATIONS   2,637 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IRPact: Integrated Resource Planning and interACTion View project

PVact - PhotoVoltaic diffusion and interACTion View project

All content following this page was uploaded by Tobias Beck on 05 September 2017.

The user has requested enhancement of the downloaded file.


Assessing the influence of the temporal resolution of electrical load and PV generation
profiles on self-consumption and sizing of PV-battery systems

T. Becka,∗, H. Kondziellab , G. Huarda , T. Brucknerb,c


a Corporate Sector Research and Advance Engineering, Robert Bosch GmbH, Robert-Bosch-Campus 1, 71272 Renningen, Germany
b Fraunhofer Center for International Management and Knowledge Economy, Sustainability Management and Infrastructure Economics,
Neumarkt 9-19, 04109 Leipzig, Germany
c Institute for Infrastructure and Resources Management, University of Leipzig, Grimmaische Straße 12, D-04109 Leipzig, Germany

Abstract
The interest in self-consumption of electricity generated by rooftop photovoltaic systems has grown in recent years,
fueled by decreasing levelized costs of electricity and feed-in tariffs as well as increasing end customer electricity prices in
the residential sector. This also fostered research on grid-connected PV-battery storage systems, which are a promising
technology to increase self-consumption. In this paper a mixed-integer linear optimization model of a PV-battery system
that minimizes the total discounted operating and investment costs is developed. The model is employed to study the
effect of the temporal resolution of electrical load and PV generation profiles on the rate of self-consumption and the
optimal sizing of PV and PV-battery systems. In contrast to previous studies high resolution (10 s) measured input
data for both PV generation and electrical load profiles is used for the analysis. The data was obtained by smart meter
measurements in 25 different households in Germany. It is shown that the temporal resolution of load profiles is more
critical for the accuracy of the determination of self-consumption rates than the resolution of the PV generation. For
PV-systems without additional storage accurate results can be obtained by using 15 min solar irradiation data. The
required accuracy for the electrical load profiles depends strongly on the load profile characteristics. While good results
can be obtained with 60 s for all electrical load profiles, 15 min data can still be sufficient for load profiles that do not
exhibit most of their electricity consumption at power levels above 2 kW. For PV-battery systems the influence of the
temporal resolution on the rate of self-consumption becomes less distinct. Depending on the load profile, temporal
resolutions between 5 min and 60 min yield good results. For optimal sizing of the PV power and the storage capacity a
resolution of 60 min is found to be sufficient. For the sizing of the battery inverter power of the storage system, a finer
temporal resolution of at least 300 s is necessary.
Keywords: photovoltaics (PV), PV-battery system, home battery storage, electrical household load profile, optimal
storage sizing

1. Introduction end customer electricity prices especially in the residential


sector on the other hand [11].
1.1. Motivation In this paper a mixed integer linear optimization model
In recent years the interest in self-consumption of elec- that determines the optimum system configuration (grid
tricity generated by rooftop photovoltaic (PV) systems has only, PV only, PV-battery) and the optimal sizing of its
grown due to its economical attractiveness. This also lead components under the given economic assumptions is de-
to an increased research interest in grid-connected battery veloped. The influence of the temporal resolution of elec-
storage systems [1–3], while in the past such systems were trical load and PV generation profiles on both the esti-
mainly applied in off-grid applications [4, 5]. In Germany mation of self-consumption rates and the optimal sizing of
this was pushed by an update of the renewable-energy-act PV and PV-battery is investigated with an emphasis on
(EEG) [6] that incentivized self-consumption between 2009 the load profile characteristics.
and 2012, rendering it economically more attractive than General relationships between the required temporal
grid feed-in [1, 7, 8]. Beyond the introduction of dedicated resolution of the input data and the accuracy of the final
incentives, the business case of self-consumption benefits results could help to increase the quality of future system
from decreasing levelized costs of electricity (LCOE) and designs while decreasing the required computational effort.
feed-in tariffs (FIT) on the one hand [9, 10], and increasing Furthermore the results are believed to provide assistance
in the reduction of uncertainties in the economical and
technical evaluation of self-consumption driven PV and
∗ Corresponding author, Tel.: +49 711 811-27701
PV-battery systems.
Email address: Tobias.Beck3@de.bosch.com (T. Beck)

Preprint submitted to Elsevier January 30, 2017


1.2. Literature review to detailed technical analysis of the power electronics and
In this section a brief overview of the scientific liter- battery behavior, which require dynamic models, simpli-
ature on PV-battery systems for residential applications fied energy flow models are applicable for operation of the
is presented, as it is the most common and promising ap- overall system [29]. Approaches comprise linear and mixed
proach to increase self-consumption of locally generated integer-linear programming [30, 31], dynamic programing
electricity [12]. In contrast to battery storage, the poten- [32], model predictive control [18] as well as neural net-
tial of demand-side management to increase self-consumption works [33]. The importance of forecast based operating
is regarded as rather low [13]. In addition to technical re- strategies is pointed out by Weniger et al. [34]. Masa-Bote
search areas like power electronics and battery aging, a lot et al. [35] proved that advanced forecasting methods com-
of research was done on the value of PV-battery storage bined with local energy storage can lead to a significant
systems for the customer as well as for the electricity grid. improvement in the quality of the forecast of grid feed-in
Common topologies for PV-battery systems include DC- by PV systems.
and AC-coupled systems [14]. Besides the DC/AC or the Besides the technological challenges described in the
AC/DC inverter (depending on the topology), the main previous paragraphs, the influence of self-consumption on
components of the storage system are the DC/DC battery the economics of PV and PV-battery systems have a pro-
converter and the battery. Detailed simulations for the nounced position in the scientific literature. This is es-
power electronics of a high efficiency DC/DC converter pecially important for the implementation of novel tech-
were carried out by Bragard et al. [15]. Nge et al. [16] nological concepts and modeling approaches in residential
proposed a simplified approach to optimize the control energy systems based on renewable energies. Early model
strategy with regard to the inverter efficiency. Detailed based approaches include estimation of self-consumption
simulations on the aging behavior of lithium ion batter- and self-sufficiency rates as well as sizing [1–3]. The im-
ies have been carried out [17], as well as approaches to portance of realistic load profiles is pointed out by Linssen
incorporate battery aging into the control and operating et al. [36], who also recommend further research on the
strategies [18–20]. impact of high temporal resolution load and generation
In addition to the economical benefit by increasing self- profiles. Systematic analysis of self-consumption rates for
consumption, the main driver for decentralized electricity different locations and load profiles in Germany were car-
storage applications is the possibility of reducing the peak ried out by Tjaden et al. [37]. With respect to optimal siz-
loading of the electricity grids. However it can be shown ing different approaches can be categorized. Besides exem-
that storage systems that only maximize self-consumption plary calculations, sensitivity analysis [1, 38, 39] and rather
don’t constitute a benefit for the electricity grids [21, 22]. heuristic methods [3, 40], some publications also employ a
Therefore different operating modes and strategies have linear optimization approach [30, 41]. This methodology
been proposed to overcome that problem and simulations is also frequently used for sizing of microgrids [42, 43] and
show that a grid-friendly operation of PV-battery sys- is particularly useful in finding the optimal system com-
tems can yield a significant benefit for the electrical grids position. Optimum community storage configurations are
[21, 23, 24]. The provision of ancillary services by PV- investigated in [44]. All mentioned approaches focus on
battery systems can also be a valuable contribution to the the optimal sizing of a PV-battery system, instead of find-
integration of large amounts of PV into the electricity grids ing the optimal system configuration (grid only, PV only,
[25]. PV-battery) and the optimal sizing of the system compo-
PV-battery systems have been also subject to research nents.
with respect to political boundary conditions and with re-
spect to their environmental impact. The economic effects 1.3. Temporal resolution of load and PV profiles
of different electricity tariffs are investigated by Parra and All scientific works mentioned so far have to rely on
Patel [26] and Jargstorf et al. [27], which highlights the im- load profiles of the residential electricity consumption and
portance of political and regulatory boundary conditions PV generation profiles as input data. Typically it is very
for the economics of self-consumption. McKenna et al. [28] hard to obtain measured load profiles, especially in tempo-
point out the negative environmental impact of lead-acid ral resolutions in the order of magnitude of seconds. The
batteries (due to production and use over lifetime) as well same holds true for PV profiles as irradiation data for arbi-
as their poor economic efficiency. trary locations can typically be obtained with a temporal
Optimized operating strategies have been a widely re- resolution of 15 min at the most. It is therefore crucial to
searched topic on PV-battery systems. The papers em- gain detailed knowledge on the influence of the temporal
ploy different modeling approaches and often apply fore- resolution of the input data on the results of simulations
cast based strategies. Besides contributions to grid topics and optimization works.
(see previous paragraph), the focus lies on the optimization Several papers deal with the influence of the tempo-
of the rate of self-consumption of storage systems. In the ral resolution on PV self-consumption. Wille-Haussmann
future this might be encouraged due to variable electricity et al. [45] analyzed the error on self-consumption rates
prices, variable feed-in tariffs or feed-in limits. In contrast for PV systems. The temporal resolution of the PV pro-
files was varied between 10 s and 15 min and resulted in
2
an overestimation of the rate of self-consumption of up to the temporal resolution on the optimal system configura-
2 %. However the influence of the temporal resolution of tion and sizing of PV and PV-battery systems is studied
the load profiles was not studied and synthetic instead of by means of a mixed integer linear programming (MILP)
measured load profiles were used. optimization model. The main research questions can be
Cao and Sirén [46] proposed a number of indices to summarized as follows:
express the matching capabilities between on-site energy
generation and demand and carried out a detailed case- • What is the influence of the temporal resolution of
study. Synthetic electrical load profiles and real irradia- the electrical load and PV generation profiles on the
tion data in 1 min temporal resolution was used, however rate of self-consumption of PV and PV-battery sys-
only single days and not a whole year could be analyzed. tems?
The authors conclude that with a coarser temporal resolu- • How does the temporal resolution of the electrical
tion the self-consumption is overestimated. However they load and PV generation profile influence the optimal
don’t recommend a preferred resolution for future works, sizing of PV and PV-battery systems?
as the results vary significantly between different simula-
tions. This could be a result of the fact that only single • How do the characteristics of the electrical load pro-
days were analyzed. Extreme load or weather conditions files affect the errors due to the changing temporal
are more likely to be overestimated by such an analysis. resolution?
Additionally the impact of the PV and load profiles is not
The remainder of this paper is divided into four sec-
discussed separately, which does not allow one to attribute
tions. Section 2 provides an overview of the optimization
the results to the type of load profile. Adding a battery to
model and the electrical household load and PV genera-
the system resulted in a remarkable decrease in the influ-
tion profiles used for the analysis. Section 3 presents the
ence of the temporal resolution.
results on the influence of the temporal resolution on the
Wyrsch, N. et al. [47] analyze the impact of both the
rate of self-consumption for PV and PV-battery systems.
load profile and the PV generation profile separately for
Subsequently, the effect of the temporal resolution on the
PV and PV-battery systems. High resolution load data
optimal investment decision in PV and PV-battery sys-
(6 − 12 s) and PV data (60 s) is used, however only one
tems is investigated in section 4. Section 5 completes the
electrical load profile was used, which limits generalization
paper with concluding remarks and recommendations for
of the results. The authors recommend a resolution of at
appropriate temporal resolutions.
least 30 s for the load profile and 10 min for the PV profile
for PV systems which both yield errors of less than 3 %.
For the PV-battery system they found the influence of the 2. Input data and optimization model
temporal resolution to be almost negligible.
In contrast to the three previous publications, Braun This section provides an overview of the electrical load
et al. [48] come to the conclusion that the rate of self- profiles and the PV generation profiles used as input data
consumption is underestimated by an increase of the simu- for this study. Subsequently the optimization model devel-
lation time step. No detailed description of the input data oped and employed for this study is presented, followed by
is provided, as the study didn’t focus on the influence of a brief description of the general economic assumptions.
the temporal resolution, but rather on the applicability of
the employed simulation model. Therefore the reasons for 2.1. Input data
the discrepancy to the remaining publications cannot be The input data for this study was obtained by smart
ascertained. meter measurements in 25 different households (subsequently
denoted by HH) across Germany during the years of 2010-
1.4. Goal and structure of the paper 2013. Electrical load profiles are available for all house-
In this paper the influence of the temporal resolution holds with a temporal resolution of the measurements of
of electrical load as well as PV generation profiles on the approximately 2s. Generation profiles for PV are also
rate of self-consumed electricity is investigated for PV as available for some households. For better data handling
well as PV-battery systems. This allows us to draw con- the measured data was rescaled to an equally spaced tem-
clusions about the role of simultaneousness of PV gener- poral resolution of 10s.
ation and electrical load as well as about the role of the
flexibility added through a battery storage system. In the 2.1.1. Electrical load profiles
previous section it was shown that the scientific literature Seven load profiles covering a broad range of energy
on this topic exhibits some limitations and inconsistencies consumptions (3053 − 6230 kWh/a) and residents (2 − 5)
[45–48]. In contrast to these previous publications high were chosen for this study (see table 1). Only profiles
resolution (10 s) measured input data is used for both the with less than 5 % missing data during one year were used
electrical load and the PV generation. To account for the and missing data and measurement errors were corrected
impact of different user types, measured data for seven manually based on previous and subsequent days in the
different households is used. Additionally, the impact of data sets. HH17 is equipped with a direct-electrical hot
3
water heater, which yields a much higher peak power of 2.2.1. Target function
up to 20 kW compared to all other load profiles. This is The target of the optimization is the minimization of
mentioned here in advance with regard to results presented total discounted electricity costs. The formulation of the
in sections 3 and 4. total annual operation costs is given in equation (1).

C total = C grid + C op,fix − Rev (1)


Table 1: Overview of electrical load profiles of seven exemplary
households (obtained by smart meter measurements; original It consists of the costs for electricity from the grid
temporal resolution: 2 s, rescaled to equally spaced 10 s) C grid , fixed costs for operation and maintenance C op,fix
Identifier
Annual electricity
Residents
and the revenues Rev due to a feed-in tariff. All com-
consumption [kWh/a] ponents of the target function are written in upper case
HH1 3053 4 letters. The detailed formulations of the components of
HH4 3702 4 the target function can be obtained from equations (2),
HH5 3454 2 (3) and (4).
HH6 4317 4 X grid2load
HH8 4740 5 C grid = Pt · cgrid (2)
HH13 5818 2 t
HH17 6230 4
C op,fix = cop,fix,pv + cop,fix,bat (3)
X pv2grid
Rev = Pt · cfit,pv (4)
2.1.2. PV generation profile t
Experimental data was obtained by one of the house- The variables Ptgrid2load and Ptpv2grid denote the power
holds of the smart meter measurement campaign. For fur- that is consumed from the grid or fed into the grid respec-
ther processing the profile was normalized by the rated tively. All cost terms refer to averaged annualized costs.
peak power (5.2 kWp ) which subsequently allows rescaling Terms written in lower case letters (cgrid , cfit,pv , cop,fix,pv ,
to the desired rated peak power. The annual energy yield cop,fix,bat ) denote fixed and variable costs that relate to a
of the non-normalized data is 5250 kWh. The advantage specific technology. The annualization is carried out ac-
of the experimental data is the high temporal resolution, cording to the following two equations by multiplying the
which makes it suitable for the analysis of the influence of total discounted costs c0 with the annuity factor af . The
the temporal resolution carried out in this paper. For siz- definition of the annuity factor refers to the annuity in
ing applications, however it is often not possible to obtain advance, the discount rate is denoted by i.
such detailed data for arbitrary locations, as it is typically
available in a temporal resolution of at most 15 min. t=T
X ct

c0 = (5)
(1 + i)t
2.2. Model description t=0

The study is carried out by applying a mixed integer (1 + i)


T −1
·i
linear programming model written in GAMS (General Al- af = T
(6)
gebraic Modeling System) with CPLEX used as the solver (1 + i) − 1
for the resulting optimization problem. The model can be The value of self-consumed electricity is implicitly ac-
run in two different modes, operation and investment & counted for by the reduction of electricity purchases from
operation. The optimization of the annual operation ap- the grid. As long as the electricity price exceeds the feed-in
plies a purely linear approach, whereas the optimization tariff, self-consumption of generated electricity is econom-
of the investment decision requires binary variables to ac- ically more attractive than feeding it to the grid. Thus
count for the fixed investment costs. the optimization model maximizes self-consumption and
Profiles for the electrical load and the PV generation only feeds electricity to the grid if it can’t be consumed lo-
have to be supplied as input data. The optimization model cally. This can happen if the generation exceeds the load
then calculates for each time step the optimal power flows for a system without battery or the battery is already fully
between the PV system, the battery, the electrical grid and charged.
the electrical load of the household. All power flows are The target function for investment mode additionally
defined as positive variables. In investment & operation incorporates variable and fixed investment costs (C inv,var ,
mode, the model also decides on the system configuration C inv,fix ). Subject to the investment decision are the rated
(grid only, PV, PV-battery) and the optimal sizing of the peak power of the PV system P rat,pv , the rated storage
main components (PV power, battery capacity, battery capacity of the battery Caprat,bat and the rated power of
inverter power). the battery inverter P rat,bat . The updated target function
for investment mode can then be formulated as follows:

C total = C grid + C op,fix + C inv,var + C inv,fix − Rev (7)


4
The variable and fixed investment costs (C inv,var , C inv,fix )
Ptpv2grid
are calculated according to equations (8) and (9) respec- Grid
tively. The variable investment costs are obtained by mul-
tiplying the investment variables (P rat,pv , Caprat,bat , P rat,bat )
Ptgrid2load
with their corresponding specific costs, where cinv,var,pv de-
note the costs per PV power, cinv,var,cap,bat the costs per
rated battery capacity and cinv,var,p,bat the costs per rated Ptbat,out · η bat,out
Ptpv2load Balance of
PV Ptbat,in /η bat,in Battery
power of the battery. power flows

C inv,var = P rat,pv · cinv,var,pv + P rat,bat · cinv,var,p,bat Dt


rat,bat inv,var,cap,bat
(8)
+ Cap ·c
End-Use
C inv,fix = Inv pv · C inv,fix,pv + Inv bat · C inv,fix,bat (9)
The binary variable indicating the investment decision Figure 1: Visualization of possible power flows between PV
in an PV system Inv pv or a battery system Inv bat are electricity generation, battery storage, electricity grid and end-
subject to equations (10) and (11). use (electrical household load), according to the load balance
equation (12)
P rat,pv − P max,pv · Inv pv ≤ 0 (10)

Caprat,bat − Capmax,bat · Inv bat ≤ 0 (11)


2.2.3. PV system
If an investment decision in a PV or battery system is The output of the PV generator is the AC power pro-
made, the corresponding binary variable Inv pv or Inv bat duced by the PV system. Therefore it consists of the PV
is set to 1 and the fixed investment costs are accounted modules and the PV inverter. As described before, the
for in equation (9). If no investment is carried out, the PV generation is not calculated within the model but the
model sets the binary variables to 0 and the fixed costs profile is provided exogenously. Therefore the profiles can
are not accounted for. These formulations also ensure that be taken from experimental measurements or from a sim-
the maximum power P max,pv for the PV system and the ulation model, which makes the model very well suited for
maximum capacity Capmax,bat for the battery cannot be the analysis carried out in this paper. The PV generation
exceeded. The installed PV power is limited by the avail- Ptpv is calculated by multiplying the normalized PV gen-
able roof area, whereas the battery capacity is typically eration profile profilepv rat,pv
t with the rated peak power P
only limited by the available volume of the system. according to equation (13).
2.2.2. Load balance equation Ptpv ≤ P rat,pv · profilepv
t (13)
The main constraint of the optimization model is the
load balance equation. It ensures that the electricity de- As the power from the PV generator can be consumed
mand Dt of the household is satisfied at all times, accord- locally (directly or via the battery) or fed into the grid,
ing to equation (12). the different power flows have to satisfy the following con-
straint as well.
Dt = Ptpv2load + Ptgrid2load
(12) Ptpv = Ptpv2grid + Ptpv2load (14)
P bat,in
− tbat,in + Ptbat,out · η bat,out
η 2.2.4. Storage system
The electricity demand of the household can be satis- The battery storage system is modeled according to the
fied by PV (Ptpv2load ), by the battery (Ptbat,out ) or by the following set of equations. The storage constraint enforces
the conservation of stored electricity between time steps,
grid (Ptgrid2load ). Excess PV power can be stored in the
where the state of charge of the battery (SOC) is defined
battery (Ptbat,in ) or fed into the grid (Ptpv2grid ). The struc-
as the ratio between the stored electrical energy and the
ture of the load balance constraint implies an AC coupled
rated capacity of the battery.
PV-battery system. A visualization of the different power
flows is provided by figure 1. SOCt+1 = SOCt · η bat
The load balance equation describes the power flows   (15)
between the different system components. Additionally + Ptbat,in · ∆t − Ptbat,out · ∆t /Caprat,bat
there are specific constraints for the PV generator and the
battery storage, which are described in the following two The power flows in and out of the battery are denoted
sections. by Ptbat,in and Ptbat,out and are defined as positive vari-
ables. Charge and discharge efficiencies are accounted for
5
in equation (12) and also include the efficiency of the bat- 3. Influence of the temporal resolution of load and
tery inverter. Additionally a certain percentage of the PV profiles on the level of self-consumption
stored electricity is lost due to the self-discharge of the
battery. The self-discharge rate is modeled via η bat . The In this section the optimization model is run in oper-
self-discharge of lithium ion batteries is neglected in this ation mode only to investigate the influence of the tem-
analysis as it is very low (0.1 − 0.3 %/day [49]). poral resolution of the input load and PV profiles on the
To keep the model linear, charging and discharging of self-consumption for a PV and a PV-battery system. As
the battery within one time step is allowed. However ac- long as the electricity price is higher than the feed-in tar-
cording to constraint (16) the sum of the charge and dis- iff, the resulting operating strategy is to maximize self-
charge power must not exceed the rated power of the bat- consumption.
tery inverter within one time step (see also [50] for a similar A PV system with a rated power of 5 kW is chosen
formulation). as standard for this analysis. The standard PV-battery
system has a nominal capacity of 3 kWh and a battery in-
Ptbat,in + Ptbat,out ≤ P rat,bat (16) verter power of 1.5 kW. Additionally, PV powers of 2.5 kW
and 10 kW and battery inverter powers of 3 kW and 5 kW
Under normal operating conditions the model has no are employed for parameter variations. An overview of the
incentive for simultaneous charging and discharging and systems is presented in table 3.
therefore constraint (16) will limit either Ptbat,in or Ptbat,out
to the maximum rated power of the battery inverter P rat,bat .
Table 3: Overview on analyzed systems (PV, PV-battery)
Equation (17) initializes the storage content at the first
time step of the optimization period to the minimum state System P rat,pv [kW] Caprat,bat [kWh] P rat,bat [kW]
of charge in percent of the rated capacity. PV 2.5, 5, 10 n.a. n.a.
PV-Bat 5 3 1.5, 3, 5
SOCt=1 = SOC min · Caprat,bat (17)

Minimum and maximum storage level of the battery Before the presentation of the results, the rate of self-
are enforced by constraints (18) and (19). consumption and its relative error are introduced. The
rate of self-consumption is defined as the electricity gen-
SOCt ≥ SOC min · Caprat,bat (18) erated by PV that is consumed locally by the household
E pv,consumed divided by the overall PV generation E pv,generated :
SOCt ≤ SOC max · Caprat,bat (19)
E pv,consumed
For the battery system the SOC of the battery the Self-consumption (SC) = (20)
E pv,generated
lower limit of the usable capacity is SOC min = 10% and
the upper limit is SOC max = 90%. Charging η bat,in and The relative error of the rate of self-consumption with re-
discharging efficiency η bat,out are assumed to be 95% for spect to the finest or baseline temporal resolution is defined
this study. as follows:
|SCT − SCTbaseline |
2.3. General economic parameters Relative error SC = (21)
SCTbaseline
As the model is driven by the minimization of the total
discounted costs, certain assumptions regarding the eco- For subsequent analyses it is important to understand
nomic parameters have to be made. An overview of the the difference between the optimization time step t and
general economic assumptions is presented in table 2. the temporal resolution T of the input profiles. The opti-
mization time step is the discretization of the optimization
model. The temporal resolution of the input profiles de-
Table 2: General economic assumptions
scribes the quality of the input data. For example if the
Parameter Value Unit temporal resolution of the load profile is T = 300 s and
Interest rate 5 %/a the optimization time step ∆t = 10 s, the value of the load
Inflation 2 %/a profile doesn’t change for 30 time steps.
Electricity price 0.34 e2020 /kWh
Price increase 1.4 %/a 3.1. PV system
Feed-in tariff 0.05 e2020 To study the influence of the temporal resolution of
the PV and the electrical load profile separately, two sets
The assumptions made in table 2 apply to the model of calculations are performed. The optimization model is
run both in operation and operation & investment mode. always run with an optimization time step of ∆t = 10 s.
The detailed assumptions regarding the investment costs For the first analysis the temporal resolution of the load
of the system components are not in the scope of the sub- profile is fixed to T = 10 s and the temporal resolution
sequent analysis and are therefore not presented here. of the PV profile is varied between 10 s − 3600 s. For the
6
Relative error self-consumption [%]
40 3
HH1
Self-consumption [%]

35 HH4
2 HH5
30 HH6
HH8
25 HH13
1
HH17
20

15 0
10 30 60 300 900 3600 10 30 60 300 900 3600
T [s] T [s]
(a) Influence of temporal resolution of the PV generation on the (b) Relative error of the rate of self-consumption for the variation
rate of self-consumption of temporal resolution of the PV generation

Figure 2: Variation of the temporal resolution of the PV generation for a PV system (P rat,pv = 5 kW; temporal resolution of load
profiles fixed to T = 10 s , optimization time step ∆t = 10 s)

45 Relative error self-consumption [%] 20


HH1
40
Self-consumption [%]

15 HH4
35 HH5
HH6
30 10 HH8
25 HH13
5 HH17
20
15 0
10 30 60 300 900 3600 10 30 60 300 900 3600
T [s] T [s]
(a) Influence of the temporal resolution of the electrical load on (b) Relative error of the rate of self-consumption for the variation
the rate of self-consumption of temporal resolution of the electrical load

Figure 3: Variation of the temporal resolution of the load profile for a PV system (P rat,pv = 5 kW; temporal resolution of PV
generation fixed to T = 10 s , optimization time step ∆t = 10 s)

second analysis the PV profile is fixed and the load profile the influence of the temporal resolution of the electrical
is varied. load profile starts to be immediately visible at T = 30 s.
The influence of the temporal resolution of the PV pro- For coarser resolutions than T = 60 s the overestimation
file on the rate of self-consumption is shown in figure 2a. of self-consumption steadily increases up to an overestima-
The rate of self-consumption obviously depends on the tion of as much as seven percentage points for load profile
load profile, with its different characteristics (e.g. annual HH17. For the remaining load profiles the error amounts
energy consumption, user behavior). Up to T = 60 s there up to three percentage points at a temporal resolution of
is no influence of the temporal resolution observable. For T = 3600 s.
larger resolutions the self-consumption is slightly overesti- Once again the relative error of the rate of self-consumption
mated, however even for T = 3600 s the error is below one provides a clearer picture of the accuracy of the simula-
percentage point for all load profiles. tions. Up to a resolution of T = 60 s the relative error
A good measure of the accuracy of the results is the rel- remains smaller than 2 % for all load profiles. For reso-
ative error of the rate of self-consumption, which is plotted lutions up to T = 900 s it remains smaller than 6 % for
in figure 2b. By definition the relative error for T = 10 s all load profiles except HH17. For the coarsest resolution
is zero. For T = 300 s the relative error is still less than of T = 3600 s the error amounts to as much as 20 % for
1 % and even for the very coarse resolution of T = 3600 s HH17 and up to 11 % for the remaining households. A
it is below 3 % for all load profiles. distinct influence of the different households on the error
The influence of the temporal resolution of the electri- can be observed. It has to be kept in mind that the profile
cal load profile on the rate of self-consumption is shown in HH17 is somewhat different to the remaining load profiles
figure 3a. In contrast to the variation of the PV resolution, because of the fact that hot water generation is carried out

7
Relative error self-consumption [%]
25
60 HH1,PV=5kW
Self-consumption [%]

20 HH1,P=2.5kW
HH1,P=10kW
40 15 HH6,P=5kW
HH6,P=2.5kW
10 HH6,P=10kW
20 HH17,P=5kW
5
HH17,P=2.5kW
HH17,P=10kW
0 0
10 30 60 300 900 3600 10 30 60 300 900 3600
T [s] T [s]
(a) Influence of the temporal resolution of the electrical load on (b) Relative error of the rate of self-consumption for the varia-
the rate of self-consumption for different PV powers P rat,pv = tion of temporal resolution of the electrical load for different PV
2.5, 5, 10 kW powers P rat,pv = 2.5, 5, 10 kW

Figure 4: Variation of the temporal resolution of the load profile for selected households (HH1, HH6, HH17) for different PV
powers (P rat,pv = 2.5, 5, 10 kW; temporal resolution of PV generation fixed to T = 10 s , optimization time step ∆t = 10 s)

by an electrical hot water boiler, resulting in peak powers households as for the variation of the PV power. An in-
up to 20 kW in the original temporal resolution (T = 10 s). crease of the inverter power from 1.5 kW to 5 kW only has
In figure 4 the influence of different levels of PV gener- a significant influence on the level of self-consumption for
ation is shown for selected households. Figure 4a clearly HH17 (see figure 6a) and yields a decrease of the corre-
shows that the level of self-consumption depends on the sponding relative error in self-consumption from more than
rated power of the PV system. The relative error in the 8 % to 5 % at 3600 s in figure 6b.
self-consumption decreases with increasing PV power for
all load profiles (see figure 4b). Only a small influence can 3.3. Discussion of the results
be observed for HH1, while a larger influence is observed In this section the results of the previous two sections
for HH17 for 10 kW and especially for HH6 where the rel- are discussed briefly. This is carried out with special re-
ative error drops from nearly 15 % to 7.5 % at 3600 s. gard to the implications for the optimization of the invest-
ment decision which is subject to section 4.
3.2. PV-battery system
In this section a PV-battery system is analyzed. Fig-
3.3.1. Load profile characteristics
ure 5 shows the results of the simultaneous variation of the
To gain a deeper understanding of the results, the load
temporal resolution of both the PV and the load profile.
duration curve and the power vs. energy curve for the
This is justified by the results obtained in the previous
seven household load profiles are presented in figure 7.
section, where the errors could be separately attributed to
The load profiles are shown in the original resolution of
the load and the generation profiles.
T = 10 s (solid lines) and in the coarsest resolution of
In contrast to the corresponding analysis for the PV
T = 3600 s (dashed lines) to illustrate the deformation of
system, the self-consumption hardly changes up to an tem-
the curves due to the time averaging effect. According to
poral resolution of T = 900 s for all households except
the load duration curves from figure 7a, the profiles can
HH17 (see figure 5a). For household HH17 the influence
be categorized in two groups. While a clear deformation
of the temporal resolution is much more distinct, which is
of the load duration curve is visible for HH6, HH13 and
in line with the results for the PV system without storage.
HH17, the influence is less pronounced for HH1, HH4, HH5
For the coarsest temporal resolution the self-consumption
and HH8. The same effect can be seen in figure 7b, where
is overestimated by four percentage points for HH17 and
the remaining annual energy above a certain power level
up to one percentage point for the remaining households.
is presented. It can also be observed that for HH6, HH13
If one excludes profile HH17, the relative error of the
a lot of the annual energy consumption is above the power
self-consumption stays below 1 % for all load profiles for
level of 1 kW and for HH17 even above 2 kW.
temporal resolutions up to 900 s. For the coarsest reso-
lution included in this analysis (T = 3600 s), the relative
3.3.2. PV system
error only slightly exceeds 2 % for profiles HH6 and HH13.
Once again HH17 takes an exceptional position, with a It can be observed that the rate of self-consumption is
relative error around 5 % for T = 900 s and more than 8 % less sensitive on the temporal resolution of the PV profile
for T = 3600 s. than of the temporal resolution of the load profile. This
The influence of a variation of the battery inverter can be explained with the larger and more frequent fluctu-
power is presented in figure 6 for the same selection of ations in the electrical load profiles. As already described

8
Relative error self-consumption [%]
55 10
HH1
Self-consumption [%]

50 8 HH4
HH5
45 6 HH6
HH8
40 4 HH13
HH17
35 2

30 0
10 30 60 300 900 3600 10 30 60 300 900 3600
T [s] T [s]
(a) Influence of the temporal resolution of PV generation and (b) Relative error of the self-consumption of the temporal reso-
electrical load profiles on self-consumption lution of PV generation and electrical load profiles

Figure 5: Simultaneous variation of the temporal resolution of the PV and electrical load profile for a PV-battery system (P rat,pv =
5 kW, Caprat,bat = 3 kWh, P rat,bat = 1.5 kW; temporal resolution equals optimization time step T = ∆t )

55 Relative error self-consumption [%] 10


HH1,P=1.5kW
Self-consumption [%]

50 8 HH1,P=3kW
HH1,P=5kW
45 6 HH6,P=1.5kW
HH6,P=3kW
40 4 HH6,P=5kW
HH17,P=1.5kW
35 2
HH17,P=3kW
HH17,P=5kW
30 0
10 30 60 300 900 3600 10 30 60 300 900 3600
T [s] T [s]
(a) Influence of the temporal resolution of PV generation and (b) Relative error of the self-consumption of the temporal reso-
electrical load profiles on self-consumption for different battery lution of PV generation and electrical load profiles for different
inverter powers P rat,bat = 1.5, 3, 5 kW battery inverter powers P rat,bat = 1.5, 3, 5 kW

Figure 6: Simultaneous variation of the temporal resolution of the PV and electrical load profile for a PV-battery system for
different battery inverter powers P rat,bat = 1.5, 3, 5 kW and selected households (HH1, HH6, HH17; Caprat,bat = 3 kWh; temporal
resolution equals optimization time step T = ∆t )

in section 2.1.2, high resolution irradiation data is typi- ergy curves differ a lot. In HH8 most energy consumption
cally not available for arbitrary locations. Common data takes place below 1 kW, while in HH6 a large portion of
sources [51] provide 15 min data, which seems to be suffi- the consumption takes place above 1 kW. Among all load
cient for the analysis of PV systems due to the very small profiles this leads to the second largest errors for HH6 and
error (around 1 %) implied by a temporal resolution of to the second smallest errors for HH8. In general it can be
T = 900 s of the PV profile (see figure 2) . stated that the more energy is consumed above approx-
Figures 3 and 4 indicate that the error in self-consumption imately 1 kW, the higher are the corresponding relative
does not only depend on the temporal resolution but also errors.
on the characteristics of the electrical load profiles. Es-
pecially HH17 shows much higher errors than the remain- 3.3.3. PV-battery system
ing households. The reason for the different behavior of Compared to the corresponding analysis for a PV sys-
this household can be obtained from the inverted power tem (see figures 2 - 4) the influence of the temporal reso-
to energy curve illustrated in figure 7b. The deformation lution of the input profiles is clearly less pronounced for a
of the curve is clearly visible for power levels above ap- PV-battery system (see figure 5). This can be attributed
proximately 1 kW. Additionally more than 3000 kWh of to the additional flexibility, which is introduced by means
the annual energy consumption take place at power lev- of the battery storage system. When adding a storage sys-
els above 2 kW. Similar observations can be made for HH6 tem the simultaneousness of electrical load and PV gener-
and HH8. Both households show a similar annual energy ation becomes less important, as electricity can be shifted
consumption, however their load duration and power to en- between time steps.
9
4 6000
HH1,T=10s

Remaining Energy [kWh]


3 HH4,T=10s
HH5,T=10s
Power [kW]

4000
HH6,T=10s
2 HH8,T=10s
HH13,T=10s 2000
1 HH17,T=10s
HH1,T=3600s
0 0
0 2000 4000 6000 8000 0 1 2 3 4 5 6
Time [h] Power [kW]
(b) Inverted power to energy curve (annual energy above a certain
(a) Load duration curve (duration of power levels during one year)
power level)

Figure 7: Detailed information about the electrical household load profiles for minimum and maximum temporal resolution
considered in this study (solid lines: T = 10 s, dashed lines: T = 3600 s)

For HH17 the battery does not reduce the effect as load duration and power to energy curves are stronger af-
strongly as for the remaining profiles. This results from the fected by time averaging in power ranges relevant to self-
fact that the battery inverter power is only 1.5 kW, which consumption.
is relatively low with regard to the peak loads of HH17.
Larger battery inverter powers (P rat,bat = 3 and 5 kW) 3.4. Comparison of the results to the literature
reduce the relative error in self-consumption, however due A detailed overview of the results from previous publi-
to the characteristics of HH17 (see figure 7) there still re- cations on the influence of the temporal resolution of load
main differences to the other households. Theoretically a and PV profiles was carried out in section 1.3. Braun et al.
battery with an even higher battery inverter power and [48] concluded, that an increase in the optimization time
larger battery capacities could reduce the error even fur- step leads to an underestimation of the self-consumption.
ther (see [46] for an analysis carried out with no power This is in contradiction to the remaining publications [45–
limitation on the battery inverter). However, in real ap- 47] and the results obtained so far in this paper and is
plications battery inverters typically exhibit powers in the therefore not discussed further. Possible reasons for this
range applied here. The results for optimally sized PV- discrepancy were already discussed in detail in the litera-
battery systems obtained in the next section, also lead to ture review in section 1.3.
rather small battery inverter sizes (see also section 4.2). Wyrsch, N. et al. [47] recommend a temporal resolution
of 10 min for the PV profile and 30 s for the electrical load
3.3.4. Recommendations for the temporal resolution for profile for an error below 3 % for the PV-systems and sim-
modeling purposes ilar errors of 2 % are reported in Wille-Haussmann et al.
The results discussed so far give confidence in using [45] for PV profiles at a temporal resolution of 15 min. This
15 min irradiation data, which can be typically obtained is in good agreement to the results obtained so far. The
for arbitrary locations. For most load profiles a temporal differences in the results for PV-battery systems from [47]
resolution of 15 min is sufficient to achieve acceptable rela- can be explained by the missing limitation of the battery
tive errors of less than 5 %. However, for load profiles that charging/discharging power as well as the fact that only
exhibit large amounts of the annual energy consumption one electric load profile was considered in their study.
at power levels above 2 kW, such as HH17, a temporal res- The most detailed investigation is carried out by Cao
olution of 60 s should not be exceeded for the load profiles and Sirén [46]. For single days they observe that the er-
in order to obtain reliable self-consumption rates through rors in self-consumption can vary extremely for different
simulations. For PV-battery systems the influence of the electrical load profiles and generation powers. Therefore
temporal resolution becomes almost negligible for all elec- they do not give a general recommendation on the required
trical load profiles except HH17. A temporal resolution of temporal resolution but provide general conclusions de-
300 s yields adequate results for all load profiles. However, pending on the characteristics of the generation and the
depending on the load profile characteristics, resolutions demand profiles instead. This is very helpful to identify
of up to 60 min can be still sufficient. cases where sufficient care has to be taken in order to avoid
For both PV and PV-Battery systems, the temporal large errors due to the chosen temporal resolution. The er-
resolution should be treated carefully for households, that rors observed in this study are much smaller, however as
show power-energy curves that exhibit a lot of the annual annual simulations are carried out, the results are not di-
energy consumption at power levels above 2 kW, as their rectly comparable. It is very likely that days with extreme

10
errors will be leveled out over the course of an entire year. 8
The load duration curves of the profiles used in this study
are smoother compared to the ones used in [46], where HH1
6 HH4
larger errors often occurred in regions of the electrical load

Prat,pv [kW]
profiles with long sharp spikes that lead to steep regions in HH5
4 HH6
the duration curve. The errors for annual simulations can
HH8
therefore be expected to be smaller for the smoother du- HH13
ration curves employed in this paper. The reason for the 2
HH17
differences in the load duration curves can originate from
the differences between daily and annual profiles as well as 0
from the differences between measured and synthetically 10 30 60 300 900 3600
T [s]
generated profiles.
For the PV-battery system larger errors are observed in Figure 8: Influence of simultaneous variation of the temporal
this paper compared to [46]. This can be attributed to the resolution of the PV and electrical load profile on optimal in-
stalled PV peak power for a PV system (temporal resolution
limitation of the battery converter power applied in this
equals optimization time step T = ∆t )
study. In figure 6 it is clearly shown that larger battery
converter powers lead to smaller errors in the rate of self-
consumption. For real applications and optimal sized PV-
battery systems, battery inverter powers are rather low 300 s the relative error stays well below 5 % for all house-
compared to the peak powers in the load profiles (see also holds except HH17. The largest relative error observed for
section 4.2), which motivates the limitation of the battery 3600 s amounts to 9 % for HH17, which corresponds to an
converter power in this study. absolute error of 500 W.

4. Influence of the temporal resolution of load and Table 4: Relative and absolute errors for optimal PV power
PV profiles on the optimal sizing of PV and P rat,pv for selected temporal resolutions (T = 300 s, 3600 s);
PV-battery systems the absolute error for the optimal PV power is defined as:
P rat,pv = |PTrat,pv − PTrat,pv rat,pv
=10 |, the relative error as: |PT −
rat,pv rat,pv
For the subsequent analysis the model is run in invest- PT =10 |/PT =10
ment & operation mode for economic parameters expected relative error [%] absolute error [W]
for the year 2020. To reduce calculation time, the smallest Identifier 300 s 3600 s 300 s 3600 s
time step for the optimization of the PV-battery system is
30 s, which is justified by the results obtained in section 3. HH1 0.3 5.9 5.9 107.9
Additionally, the system configuration for the PV-battery HH4 3.6 7.5 75.3 154.9
HH5 0.6 3.0 16.8 90.8
system is predefined by fixing the corresponding binary
HH6 1.5 2.5 54.1 88.5
variables to decrease the required computation time.
HH8 0.3 0.5 8.5 17.0
HH13 0.4 0.8 15.7 33.0
4.1. PV system HH17 5.8 9.0 319.6 499.6
In this scenario the model optimizes the investment de-
cision in a PV system subject to economic parameters for
the year 2020. The model can choose between obtaining 4.2. PV-battery system
all electricity from the grid or to invest in a PV system
In the following investigation the scope of the analysis
and determine its optimal size.
is expanded to the investment decision in a storage system.
The results for the optimization of the PV system are
In addition to the size of the PV installation, the model
shown in figure 8. For all households a PV system is
optimizes the size of the battery storage and the power of
installed and the optimal size strongly depends on the
the battery inverter.
load profile. Up to T = 300 s the optimal system size
The results for the PV power and the size of the bat-
is practically independent of the temporal resolution. For
tery are presented in figure 9. The trends of figure 9a
T = 900 s and T = 3600 s minor changes in the optimal
correspond very well to figure 8 and are described in the
PV power are visible, but don’t follow a clear trend. For
previous section. With the exception of HH17 no distinct
HH17 an increase in the optimal PV power is clearly vis-
influence of the temporal resolution on the optimal sys-
ible. For HH1 and HH5 the optimal system size slightly
tem size can be determined. A similar conclusion can be
increases, whereas it slightly decreases or stays constant
drawn from figure 9b. For temporal resolutions up to 900 s
for the remaining households.
almost no influence on the optimal system size is visible.
Table 4 gives an overview of the relative and absolute
For the coarsest temporal resolution T = 3600 s the opti-
errors of the optimal PV power for two selected temporal
mal storage size is slightly underestimated for all house-
resolutions, 300 s and 3600 s. For a temporal resolution of
holds. The relative and absolute errors for the PV power
11
10 8 3
HH1

Caprat,bat [kWh]
8 HH4

Prat,bat [kW]
Prat,pv [kW]

6 2 HH5
6 HH6
4 1 HH8
4 HH13
HH17
2 2 0
30 60 300 900 3600 30 60 300 900 3600 30 60 300 900 3600
T [s] T [s] T [s]
(a) Influence of temporal resolu- (b) Influence of temporal resolu- (c) Influence of the variation of the temporal resolution on opti-
tion on optimal rated PV peak tion on optimal storage size mization of battery inverter power
power

Figure 9: Influence of simultaneous variation of the temporal resolution of the PV and electrical load profile on optimization of
PV power and storage size for a PV-battery system (temporal resolution equals optimization time step T = ∆t )

and the battery capacity are given in table 5 for 300 s and previously discussed, the storage system adds flexibility to
3600 s. Compared to the results for a PV system without the entire system, which reduces the importance of simul-
storage (see section 4.1), the relative and absolute errors taneousness between the PV generation and the electrical
are even smaller for the PV-battery system. The relative load profile. The total amount of electricity that can be
errors of the battery capacity only exceed 1 % for HH17 self-consumed is limited by the daily PV electricity gen-
for a temporal resolution of 300 s. For 3600 s the largest eration and the electricity consumption, not so much by
relative errors observed are around 7 % for HH 6 and HH8. their simultaneousness. This also explains the major im-
The absolute errors of the optimal PV power is in the order pact that the temporal resolution has on the sizing of the
of magnitude of today’s typical PV module sizes (200 W- battery inverter power of the storage system. The addi-
300 W) for the coarsest temporal resolution of 3600 s. For tional flexibility the storage system provides can only be
the battery capacity a similar picture can be drawn with exploited if the battery inverter doesn’t become a bottle-
small absolute errors in the range of 100 Wh to 300 Wh neck. With a coarser time resolution the need for sufficient
(see table 5). inverter power for charging and discharging of the battery
For all load profiles the optimal power of the battery is therefore underestimated by the optimization model.
inverter is underestimated with increasing temporal res- The temporal resolution only has a small influence on
olution. This can be attributed to the fact that a lower the optimal rated PV power and battery capacity; how-
temporal resolution reduces the probability of occurrence ever the optimal power of the battery inverter is heavily
of peaks in the load profile. influenced by the temporal resolution. For sizing of the
The relative errors for a temporal resolution of 300 s are PV power and the storage capacity a temporal resolution
between 5 % and 15 % and increase up to 50 % for HH17 at of the input profiles of 15 min is recommended. For the
a resolution of 3600 s (see table 5). At 3600 s the optimal sizing of the battery inverter power, which is heavily influ-
inverter power is underestimated by 1363 W for HH17 and enced by the load profile, a finer temporal resolution of at
200 W to 500 W for the remaining households. least 300 s is recommended.

4.3. Discussion of the results


5. Conclusions
The influence of the temporal resolution on the optimal
sizing of the PV system as described in figure 8 proved to In this paper the effect of the temporal resolution of
be almost negligible for all analyzed temporal resolutions. electrical load and PV generation profiles on the estima-
This is surprising against the background of the influence tion of rates of self-consumption and optimal sizing of PV
of the temporal resolution on the self-consumption (see and PV-battery systems was studied. In contrast to pre-
figure 5) but can be explained by the smaller influence of vious publications [45–48] a strong emphasize was put on
the temporal resolution of the PV profiles compared to the measured input data for both PV generation and electri-
electrical load profiles on the overall accuracy. cal load profiles. For the analysis a mixed integer linear
The rather low influence of the temporal resolution on optimization model of a PV-battery system was developed
the optimal size of the installed PV power and the storage and employed. The model can be run in operation and in-
size is in line with the results of section 3. The influence on vestment & operation mode and minimizes the total dis-
the optimal sizing of the PV power is in fact even smaller counted costs to satisfy a household’s demand for electric-
for the PV-battery system compared to the PV system. As ity. The optimization of the operating strategy is based

12
Table 5: Relative and absolute errors for optimal PV power P rat,pv , battery capacity Caprat,bat and battery inverter power
P rat,bat for selected temporal resolutions (T = 300 s, 3600 s); the absolute error for the optimal PV power is defined as: P rat,pv =
|PTrat,pv − PTrat,pv rat,pv
=30 |, the relative error as: |PT − PTrat,pv rat,pv
=30 |/PT =30 (absolute and relative errors of Cap
rat,bat
and P rat,bat defined
correspondingly)
PV power P rat,pv Battery capacity Caprat,bat Battery inverter power P rat,bat
relative error [%] absolute error [W] relative error [%] absolute error [Wh] relative error [%] absolute error [W]
Identifier
300 s 3600 s 300 s 3600 s 300 s 3600 s 300 s 3600 s 300 s 3600 s 300 s 3600 s
HH1 0.5 1.0 15.6 29.5 0.5 1.0 16.7 31.2 5.3 21.3 42.6 171.6
HH4 0.2 0.6 6.9 20.4 0.4 2.6 14.3 102.5 9.6 33.4 104.3 361.9
HH5 0.2 2.1 5.9 76.8 0.5 5.3 13.1 150.1 8.3 45.0 95.7 522.0
HH6 0.1 1.9 5.9 84.3 0.6 7.5 21.4 290.7 9.0 36.8 128.2 521.6
HH8 1.2 0.5 45.1 18.5 4.9 7.0 125.7 179.4 12.2 35.4 102.1 294.9
HH13 0.3 4.1 14.8 217.0 0.4 2.1 20.1 115.0 8.6 31.9 145.7 542.4
HH17 2.3 3.2 162.5 224.0 1.7 2.9 88.7 147.4 16.3 53.1 419.9 1363.4

on a simplified linear programming approach, whereas the ing decisions mainly depending on electrical energy (PV
optimization of the investment decision follows the mixed power, storage capacity) rather than the characteristics of
integer programing paradigm. the load profile, a temporal resolution of the input pro-
The influence of the temporal resolution of the PV and files of 60 min is sufficient for most load profiles. For the
load profiles on the rate of self-consumption is studied in sizing of components that are strongly influenced by the
detail. The aim of this analysis is to gain more insight load profile characteristics, such as the power of the stor-
into the simultaneousness between the PV generation and age system, a finer temporal resolution of at least 300 s is
the electrical load profile. For PV systems without storage recommended.
the results indicate that the temporal resolution of the load The results obtained in this study can provide assis-
profiles has a much larger impact than the PV profiles. For tance in the reduction of uncertainties in the evaluation of
future analysis this gives confidence to use 15 min irradia- the economical feasibility of self-consumption driven PV
tion data, which can be typically obtained for arbitrary lo- and PV-battery systems and therefore reduce the associ-
cations. In order to obtain reliable self-consumption rates ated investment risks. Thanks to the detailed discussion
through simulations, the required temporal resolution of of the load profile characteristics, the results can be trans-
the electrical load profiles strongly depends on the char- fered to arbitrary electrical load profiles and decrease the
acteristics of the electrical load profile. While a temporal calculation time needed for simulation and optimization
resolution of 60 s yields good results for all load profiles calculations. Furthermore, the developed methodology for
in this study, 15 min data is still sufficient for load profiles optimal configuration and sizing can provide valuable in-
that do not exhibit most of their electricity consumption at formations for future system designs.
power levels above 2 kW. For PV-battery systems the influ-
ence of the temporal resolution becomes almost negligible
Acknowledgements
for most electrical load profiles, which can be attributed
to the additional flexibility introduced by the storage sys- This work was financially supported by the German
tem. For electrical load profiles with frequent peak loads Federal Ministry for Economic Affairs and Energy in the
and a large amount of energy consumption at power lev- framework of the publicly funded PV-HOST project (fund-
els above 2 kW (see HH17), the rate of self-consumption ing number 0325477A). The authors take full and sole re-
is sensitive to the temporal resolution for the PV-battery sponsibility for the content of this paper.
system as well. This effect is especially visible at low bat-
tery converter powers. A temporal resolution of 300 s leads
to errors of less than 5 % for all load profiles employed in References
this study, while temporal resolutions up to 60 min can [1] M. Braun, K. Büdenbender, D. Magnor, A. Jossen, Photo-
be still acceptable depending on the characteristics of the voltaic self-consumption in germany using lithium-ion storage
load profile. to increase self-consumed photovoltaic energy, in: Proceedings
The second part of the paper investigates the impli- of the 24th European Photovoltaic Solar Energy Conference,
Hamburg, 2009.
cations of the previous results on the configuration and [2] G. Mulder, E. Peeters, Storage development for households with
sizing of both PV and PV-battery systems. In contrast solar panels, in: Proceedings of the 20th International Confer-
to the previous results, the temporal resolution only has ence and Exhibition on Electricity Distribution, Prague, 2009.
[3] G. Mulder, F. D. Ridder, D. Six, Electricity storage for grid-
limited influence on the optimal rated PV power and bat-
connected household dwellings with PV panels, Solar Energy
tery capacity for all employed load profiles. However, the 84 (7) (2010) 1284–1293, doi:10.1016/j.solener.2010.04.005.
optimal power of the battery inverter is heavily influenced [4] D. U. Sauer, Optimierung des Einsatzes von Blei-Säure-
by the temporal resolution. It is concluded that for siz- Akkumulatoren in Photovoltaik-Hybrid-Systemen unter

13
spezieller Berücksichtigung der Batteriealterung, Ph.D. thesis, [23] K. Büdenbender, M. Braun, T. Stetz, P. Strauss, Multifunc-
Universität Ulm, 2003. tional PV systems offering additional functionalities and im-
[5] A. Jossen, J. Garche, D. U. Sauer, Operation conditions of bat- proving grid integration, Int. Journal of Distributed Energy Re-
teries in PV applications, Solar Energy 76 (6) (2004) 759–769, sources 7 (2011) 109?128.
doi:10.1016/j.solener.2003.12.013. [24] J. Moshövel, K.-P. Kairies, D. Magnor, M. Leuthold, M. Bost,
[6] Gesetz zur Neuregelung des Rechts der Erneuerbaren Energien S. Gährs, E. Szczechowicz, M. Cramer, D. U. Sauer, Analysis of
im Strombereich und zur Änderung damit zusammenhängender the maximal possible grid relief from PV-peak-power impacts by
Vorschriften, Bundesgesetzblatt Jahrgang 2008, 2008. using storage systems for increased self-consumption, Applied
[7] K. Büdenbender, H. Barth, M. Braun, P. Strauß, Eigenver- Energy 137 (2015) 567–575, doi:10.1016/j.apenergy.2014.07.021.
brauch: Wirtschaftlichkeit und Technik, neue Einspeiseregeln, [25] J. von Appen, M. Braun, T. Kneiske, A. Schmiegel, Ein-
in: Tagungsband 26. Symposium Photovoltaische Solarenergie, fluss von PV-Speichersystemen auf das Niederspannungsnetz,
Bad Staffelstein, 2011. in: Tagungsband 28. Symposium Photovoltaische Solarenergie,
[8] V. Quaschning, Einsatzmöglichkeiten und Potenziale der Pho- Bad Staffelstein, 2013.
tovoltaik ohne erhöhte EEG-Vergütung, in: Tagungsband [26] D. Parra, M. K. Patel, Effect of tariffs on the performance
27. Symposium Photovoltaische Solarenergie, Bad Staffelstein, and economic benefits of PV-coupled battery systems, Applied
2012. Energy 164 (2016) 175 – 187, doi:http://dx.doi.org/10.1016/j.
[9] C. Kost, J. Mayer, J. Thomsen, N. Hartmann, Stromgeste- apenergy.2015.11.037.
hungskosten Erneuerbare Energien, Studie des Fraunhofer- [27] J. Jargstorf, C. D. Jonghe, R. Belmans, Assessing the reflec-
Institut für Solare Energiesysteme ISE, Fraunhofer-Institut für tivity of residential grid tariffs for a user reaction through pho-
Solare Energiesysteme ISE, Freiburg, 2013. tovoltaics and battery storage, Sustainable Energy, Grids and
[10] C. Breyer, A. Gerlach, Global overview on grid-parity event Networks 1 (2015) 85 – 98, doi:http://dx.doi.org/10.1016/j.
dynamics, in: Proceedings of the 25th European Photovoltaic segan.2015.01.003.
Solar Energy Conference and Exhibition, Valencia, 2010. [28] E. McKenna, M. McManus, S. Cooper, M. Thomson, Eco-
[11] EUROSTAT, Electricity prices by type of user nrg_pc_204, nomic and environmental impact of lead-acid batteries in grid-
URL http://ec.europa.eu/eurostat/tgm/table.do?tab= connected domestic PV systems, Applied Energy 104 (2013) 239
table&init=1&language=en&pcode=ten00117&plugin=1, 2016. – 249, doi:http://dx.doi.org/10.1016/j.apenergy.2012.11.016.
[12] R. Luthander, J. Widén, D. Nilsson, J. Palm, Photovoltaic self- [29] D. Magnor, N. Soltau, M. Bragard, A.-U. Schmiegel, R. W.
consumption in buildings: A review, Applied Energy 142 (2015) De Doncker, D. U. Sauer, Analysis of the model dynamics
80–94, doi:10.1016/j.apenergy.2014.12.028. for the battery and battery converter in a grid connected
[13] J. Widén, Improved photovoltaic self-consumption with appli- 5 kW photovoltaic system, in: Proceedings of the 25th Eu-
ance scheduling in 200 single-family buildings, Applied Energy ropean Photovoltaic Solar Energy Conference and Exhibi-
126 (2014) 199–212, doi:10.1016/j.apenergy.2014.04.008. tion, Valencia, ISBN 3-936338-26-4, 3821 – 3827, doi:10.4229/
[14] F. Kever, Unterschiedliche Schaltungen, unterschiedliche Ef- 25thEUPVSEC2010-4CO.1.4, 2010.
fizienzen, pv magazine (2) (2013) 109–112. [30] T. H. Pham, F. Wurtz, S. Bacha, Optimal operation of a
[15] M. Bragard, N. Soltau, R. W. De Doncker, A. Schmiegel, Design PV based multi-source system and energy management for
and implementation of a 5 kW photovoltaic system with li-ion household application, IEEE, ISBN 978-1-4244-3506-7, 1–5, doi:
battery and additional DC-DC converter, in: Energy Conver- 10.1109/ICIT.2009.4939701, 2009.
sion Congress and Exposition (ECCE), 2010 IEEE, Atlanta, [31] C. Clastres, T. Ha Pham, F. Wurtz, S. Bacha, Optimal house-
ISBN 978-1-4244-5286-6, 2944–2949, doi:10.1109/ECCE.2010. hold energy management and participation in ancillary services
5618220, 2010. with PV production, Cahier de Recherche No. 10. v2, Grenoble,
[16] C. L. Nge, O.-M. Midtgard, L. Norum, Power management of 2009.
grid-connected photovoltaic inverter with storage battery, in: [32] Y. Riffonneau, S. Bacha, F. Barruel, S. Ploix, Optimal power
Proceedings of the IEEE PowerTech, Trondheim, doi:10.1109/ flow management for grid connected PV systems with batteries,
PTC.2011.6019415, 2011. IEEE Transactions on Sustainable Energy 2 (3) (2011) 309–320,
[17] D. Magnor, J. B. Gerschler, M. Ecker, P. Merck, D. U. Sauer, doi:10.1109/TSTE.2011.2114901.
Concept of a battery aging model for lithium-ion batteries con- [33] E. Matallanas, M. Castillo-Cagigal, A. Gutiérrez,
sidering the lifetime dependency on the operation strategy, in: F. Monasterio-Huelin, E. Caamaño-Martín, D. Masa,
Proceedings of the 24th European Photovoltaic Solar Energy J. Jiménez-Leube, Neural network controller for Active
Conference and Exhibition Conference, Hamburg, 2009. Demand-Side Management with PV energy in the resi-
[18] M. Koller, T. Borsche, A. Ulbig, G. Andersson, Defining a dential sector, Applied Energy 91 (1) (2012) 90–97, doi:
degradation cost function for optimal control of a battery en- 10.1016/j.apenergy.2011.09.004.
ergy storage system, in: Proceedings of the IEEE PowerTech [34] J. Weniger, J. Bergner, T. Tjaden, V. Quaschning, Bedeutung
Conference, Grenoble, 2013. von prognosebasierten Betriebsstrategien für die Netzintegra-
[19] S. X. Chen, H. B. Gooi, Scheduling of energy storage in a grid- tion von PV-Speichersystemen, in: Tagungsband 29. Sympo-
connected PV/battery system via SIMPLORER, IEEE, ISBN sium Photovoltaische Solarenergie, Kloster Banz, Bad Staffel-
978-1-4244-4546-2, doi:10.1109/TENCON.2009.5396150, 2009. stein, 2014.
[20] J. Li, M. A. Danzer, Optimal charge control strategies for sta- [35] D. Masa-Bote, M. Castillo-Cagigal, E. Matallanas, E. Caamano-
tionary photovoltaic battery systems, Journal of Power Sources Martín, A. Gutierrez, F. Monasterio-Huelín, J. Jiménez-Leube,
258 (2014) 365–373, doi:10.1016/j.jpowsour.2014.02.066. Improving photovoltaics grid integration through short time
[21] K. Büdenbender, M. Braun, A. Schmiegel, D. Magnor, J. C. forecasting and self-consumption, Applied Energy 125 (2014)
Marcel, Improving PV integration into the distribution grid. 103 – 113, doi:http://dx.doi.org/10.1016/j.apenergy.2014.03.
Contribution of multifunctional PV-battery systems to sta- 045.
bilised system operation, in: Proceedings of the 25th European [36] J. Linssen, P. Stenzel, J. Fleer, Techno-economic analysis of
Photovoltaic Solar Energy Conference and Exhibition, Valencia, photovoltaic battery systems and the influence of different con-
2010. sumer load profiles, Applied Energy (2015) –doi:http://dx.doi.
[22] ISE, Speicherstudie 2013 - Kurzgutachten zur Abschätzung org/10.1016/j.apenergy.2015.11.088.
und Einordnung energiewirtschaftlicher, ökonomischer und an- [37] T. Tjaden, J. Weniger, J. Bergner, F. Schnorr, V. Quaschn-
derer Effekte bei Förderung von objektgebunden elektrochemis- ing, Einfluss des Standorts und des Nutzerverhaltens auf die en-
chen Speichern, Studie im Auftrag des Bundesverband So- ergetische Bewertung von PV-Speichersystemen, in: Tagungs-
larwirtschaft (BSW-Solar), Fraunhofer-Institut für Solare En- band 29. Symposium Photovoltaische Solarenergie, Kloster
ergiesysteme ISE, 2013. Banz, Bad Staffelstein, 2014.

14
[38] G. Mulder, D. Six, B. Claessens, T. Broes, N. Omar, J. V.
Mierlo, The dimensioning of PV-battery systems depending on
the incentive and selling price conditions, Applied Energy 111
(2013) 1126–1135, doi:10.1016/j.apenergy.2013.03.059.
[39] J. Weniger, T. Tjaden, V. Quaschning, Sizing of residential PV
battery systems, Energy Procedia 46 (2014) 78–87, doi:10.1016/
j.egypro.2014.01.160.
[40] J. Kathan, M. Stifter, Increasing BIPV self-consumption
through electrical storage - feasible demand-coverage and di-
mensioning of the storage system, in: Proceedings of the 5th In-
ternational Renewable Energy Storage Conference, Berlin, 2010.
[41] B. Claessens, K. De Brabandere, G. Mulder, F. De Ridder,
Break-even analysis of PV-battery self-consumption systems, in:
Proceedings of the 5th International Renewable Energy Storage
Conference, Berlin, 2010.
[42] C. Marnay, G. Venkataramanan, M. Stadler, A. Siddiqui,
R. Firestone, B. Chandran, Optimal technology selection and
operation of commercial-building microgrids, IEEE Transac-
tions on Power Systems 23 (3) (2008) 975–982, doi:10.1109/
TPWRS.2008.922654.
[43] C. Marnay, M. Stadler, A. Siddiqui, N. DeForest, J. Donadee,
P. Bhattacharya, J. Lai, Applications of optimal building energy
system selection and operation, Proceedings of the Institution
of Mechanical Engineers, Part A: Journal of Power and Energy
227 (1) (2013) 82–93, doi:10.1177/0957650912468408.
[44] D. Parra, M. Gillott, S. A. Norman, G. S. Walker, Optimum
community energy storage system for PV energy time-shift, Ap-
plied Energy 137 (2015) 576–587, doi:10.1016/j.apenergy.2014.
08.060.
[45] B. Wille-Haussmann, J. Link, A. Sabo, C. Wittwer, Photo-
voltaik Eigenstromnutzung - Fluktuation von Strahlung und
Last, in: Tagungsband 27. Symposium Photovoltaische Solaren-
ergie, Kloster Banz, Bad Staffelstein, 606–611, 2012.
[46] S. Cao, K. Sirén, Impact of simulation time-resolution on the
matching of PV production and household electric demand, Ap-
plied Energy 128 (2014) 192–208, doi:10.1016/j.apenergy.2014.
04.075.
[47] Wyrsch, N., Riesen, Y., Ballif, C., Effect of the fluctuations
of PV production and electricity demand on the PV electricity
self-consumption, in: Proceedings of the 28th European Pho-
tovoltaic Solar Energy Conference and Exhibtion, Amsterdam,
doi:10.4229/28thEUPVSEC2013-5BV.7.77, 2013.
[48] M. Braun, K. Büdenbender, M. Landau, D. U. Sauer, D. Mag-
nor, A. U. Schmiegel, Charakterisierung von netzgekoppelten
PV-Batterie-Systemen - Verfahren zur vereinfachten Bestim-
mung der Performance, in: Tagungsband 25. Symposium Pho-
tovoltaische Solarenergie, Kloster Banz, Bad Staffelstein, 2010.
[49] J. Leadbetter, L. G. Swan, Selection of battery technology to
support grid-integrated renewable electricity, Journal of Power
Sources 216 (2012) 376–386, doi:10.1016/j.jpowsour.2012.05.
081.
[50] K.-H. Ahlert, Economics of Distributed Storage Systems - An
economic analysis of arbitrage-maximizing storage systems at
the end consumer level, Ph.D. thesis, Karlsruhe Institute of
Technology (KIT), 2010.
[51] EnergyPlus, EnergyPlus Weather Data, URL http://apps1.
eere.energy.gov/buildings/energyplus/weatherdata_about.
cfm, 2015.

15

View publication stats

You might also like