Electric Power Systems Research 189 (2020) 106763

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Mitigating the impact of distributed PV in a low-voltage grid using T

electricity tariffs
⁎
Jordan Holweger , Lionel Bloch, Christophe Ballif, Nicolas Wyrsch
Photovoltaics and thin film electronics laboratory (PV-LAB), École Polytechnique Fédérale de Lausanne (EPFL), Institute of Microengineering (IMT), Neuchâtel,
Switzerland

A R T I C LE I N FO A B S T R A C T

Keywords: A high share of distributed photovoltaic (PV) generation in low-voltage networks may lead to over-voltage, and
Grid stability line/transformer overloading. To mitigate these issues, we investigate how advanced electricity tariffs could
Mixed-integer linear programming ensure safe grid operation while enabling building owners to recover their investment in a PV and storage
Electricity tariff design system. We show that dynamic volumetric electricity prices trigger economic opportunities for large investments
Cost optimization
in PV and battery capacity but lead to more stress on the grid while capacity and block rate tariffs mitigate over-
Photovoltaic
voltage and decrease line loading issues. However, block rate tariffs significantly decrease the optimal PV in-
stallation size.

1. Introduction uncoordinated operation for years to come. Conversely, Hidalgo-Ro-

driguez et al. [3] studied two micro-grid coordinated control strategies
Significant investments in photovoltaic (PV) systems are expected in and one uncoordinated control strategy to achieve network balancing.
the coming years due to increasing economic interest and to their Further in this direction, Haschemipour et al. [4] proposed a control
contribution to clean energy generation. Reaching the full PV potential strategy for PV and battery systems to achieve voltage regulation, using
in urban environments could lead to a system in which a significant a multi-objective optimization. In this work, however, the grid ex-
fraction of the energy should be curtailed in order to cope with the change cost, i.e. the profitability for the building owner, is not con-
network operating constraints. Centralized control of PV curtailment by sidered as an objective function. The opposite approach of Sani Hassan
the distribution system operator (DSO) is a promising solution but [5] is to optimize the sizing and operation of an energy system on ty-
could raise concerns about intrusiveness and discourage (potential) pical days and to study the impact on the voltage profile under various
prosumers. An alternative approach is to propose novel electricity pri- technological scenarios. The latter apply only to a single multi-family
cing mechanisms that help to mitigate the impact of distributed storage building while the load of the other bus remains identical. The work of
and PV systems on the grid while allowing building owners to make Wang et al. [6] proposes to integrate the revenue from frequency
profitable investments. The purpose of this work is to evaluate which control ancillary services and reliability services into the objective
advanced electricity tariff creates the best trade-off between these two function. The model quantifies the value of such services.
objectives. The previously mentioned literature does not consider the use of
The impact of the high penetration of distributed renewable energy advanced tariff structures nor their impact on the system configura-
sources was investigated in [1] where it is shown how the network tions. Schreiber et al. [7] proposed electricity tariffs and evaluate their
topology and the way distributed PV is geographically located along the impact on the grid requirements for two different buildings under two
feeder have a strong impact on the voltage profile. Considering the technological scenarios. Deetjen et al. [8] introduced an integrated
deployment of net-zero-energy buildings, Arboleya et al. [2] studied the convex formulation for equipment sizing and optimal control for a
impact of various PV penetration levels on the voltage profile of a low- central utility plant in order to better integrate the surrounding rooftop
voltage grid. The authors of this study assumed that each building PV. The authors considered time-of-use electricity pricing to be the best
minimizes its energy bill, i.e. no coordination between building is solution for providing flexibility services, absorbing the excess PV
present, and argued that the network will have to face such generation. More tariff scenarios are considered in the study of Ren [9],

⁎
Corresponding author.
E-mail addresses: jordan.holweger@epfl.ch (J. Holweger), lionel.bloch@epfl.ch (L. Bloch), christophe.ballif@epfl.ch (C. Ballif),
nicolas.wyrsch@epfl.ch (N. Wyrsch).

https://doi.org/10.1016/j.epsr.2020.106763
Received 1 October 2019; Received in revised form 17 April 2020; Accepted 2 August 2020
Available online 10 August 2020
0378-7796/ © 2020 Elsevier B.V. All rights reserved.
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

which specially focuses on the profitability of PV and battery system for

owners. This analysis describes how capacity tariffs can lower the
revenue of the owners because PV generation cannot efficiently reduce
peak demand. The previously mentioned studies are missing a few
considerations, such as accounting for more than a single building, al-
lowing the PV and battery size to adapt to the scenarios, or integrating
their assessment of building performance and network impact. There is
thus a lack of literature considering the synergies between mitigating
the network impact of PV systems and enabling profitable investment in
PV-battery systems. Fig. 2. Optimization inputs and outputs.
In this work, we investigate the effect of five different tariff sce-
narios, on network operation when optimizing both the design and
such that the revenue of the DSO stays as close as possible to the rev-
operation of all buildings connected to that network, using a metho-
enue of the reference case.
dology developed in [10]. The scenarios consider pure volumetric
The following briefly summarizes the buildings’ optimization pro-
electricity tariffs, a mix of volumetric and capacity-based tariffs, or a
blem, defines the performance metrics for the design and operation, and
block rate tariff. The optimization is run for a set of buildings in a sub-
presents the performance metrics from the grid perspective.
network of Rolle (Switzerland). The buildings’ characteristics are
known from a geographical information system. The resulting loads and
generations at each injection point allow for solving the power flow 2.1. PV-battery optimal sizing and operation
equations over a full year to extract the distribution of the voltage level
and line loading. Finally, these distributions are compared with a re- The PV and battery sizing and operation are optimized for each
ference case to assess the effect of the selected tariffs. building to minimize the total cost of ownership, i.e. the sum of the
annualized investment, maintenance and operational cost, as described
in (1), subject to a power balance constraint (2). By definition, the
2. Methodology
optimization problem relies on the assumption that an exact forecast of
both the PV generation and the electrical load is provided. The impact
The methodology to assess the impact of distributed PV on a low-
of forecast errors and energy managers’ performance is outside the
voltage grid consists of a two steps process, which is sequentially re-
scope of this study. The optimization model’s input and output is gra-
peated for each tariff scenario. First, the optimal design and operation
phically summarized in Fig. 2. The formulation of the optimization is
of each building energy system are individually optimized using a
written as:
mixed-integer linear programming (MILP) formulation under a parti-
cular tariff scenario. After each individual building optimization, the pv
minimizecapex (Pcap , E batc ap) + opex (Ptimp, Ptexp) (1)
performance of the considered building is assessed. Once all building
have been optimized, the load flow problem is solved for a distribution subject to P impt − P expt − P chat + P dist − P curt + P pvt
grid using the building-grid power exchanges resulting from the first = P loadt ∀t∈T (2)
step. Finally, the impact on the grid is assessed by evaluating dedicated
grid performance metrics. The workflow of this process is illustrated in where capex the annualized investment and maintenance costs, are a
pv cap
Fig. 1. It can be noted that the first scenario (s = 1) serves as reference function of the installed PV capacity Pcap and the battery size Ebat opex
scenario. The tariff parameters of the other scenarios are calibrated are the operating costs and are a function of the grid exchange power
Ptimp, exp . P cha, dist are the battery charging and discharging power re-
spectively, P curt is the power actively curtailed from the PV generation
P pvt , and P loadt is the uncontrollable electric load of the building.
The operational costs are simply the cost of exchanging energy or
power with the grid (3a). Three kinds of grid exchange costs are
modeled: Volumetric (3b) in which costs are proportional to the ex-
changed energy, capacity-based (3c) in which the cost is proportional to
the maximum monthly exchanged power, and a block rate (3d) in
which the marginal cost of buying/selling energy depends on the power
level at which it is exchanged. Mathematically, this can be written as:

Op. costopex = ∑ ox p
p ∈ vol, cap, block (3a)
T
Volumetric ox vol = ∑ [P impt ·t impt − P expt ·t expt ]·tst
t=1 (3b)
M
Capacityox cap = ∑ Pmaxm ·t max
m=1 (3c)
T
Block rate oxblock = ∑ k =max (P impt ·aki mp ·tst + bki mp)
1⋯K
t=1
T
− ∑ min (P expt ·ake xp ·tst + bke xp)
k=1⋯K
t=1 (3d)
e
where t impt , t xpt
are the volumetric import and export tariff respectively
(in CHF/kWh), tst is the simulation time-step, the maximum power for
Fig. 1. Process workflow. month m, P maxm , is calculated by requiring that both the import and the

2
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

Fig. 3. A map of the considered low-voltage grid (black lines). The circle diameters are scaled according to the PV potential capacity (orange) and annual energy
demand (violet). The relative circle diameter gives an indication on the potential PV penetration considering an annual irradiation of 1000 h.

export power be smaller than this variable, the pairs of coefficients price of one unit of energy that is consumed under the optimized
aki mp , bki mp and ake xp , bke xp are the component of the piece-wise linear system. These indicators are defined for each building as follows (the
cost functions of buying, respectively selling energy, a representing the use of a subscript b has been omitted for better readability):
slope, or marginal cost,b the corresponding intercept term and simply pv pv
PV host = Pcap / Pcap ,max (4a)
ensure the continuity of the cost function (see the buy cost and sell cost
curve in Fig. 4). All these variables are parameters of the building op-
timization problems. Only P pvcap , E batc ap, Ptimp and Ptexp are the opti-
PV penetration = ∑ Ptp v/ ∑ Ptl oad
t t (4b)
mization decision variables. Variables P chat , P dist , P curt , P pvt and P max m
are intermediate decision variables constraint by the decision variables E batc ap
BAT auto =
and parameters. The reader should refer to [10] for the modeling de- mean daily consumption (4c)
tails of this optimization problem.
PV cur = ∑ P curt / ∑ P pvt
t t (4d)
2.2. Performance metrics
GU IMP, EXP = max(P imp, expt )/max(P load t)
t t (4e)
These metrics aim to assess the system design’s reaction, in terms of
equipment size and operation, and the network’s reaction, in terms of NPV = ∑ cft /(1 + r )t
voltage profile and line loading, when changing the electricity pricing t (4f)
structure. From a building design perspective, the PV hosting ratio, PV T
host (4a), is the ratio between the installed PV capacity and the max- ∑t = 1 cft − opex t0
DPP = T such as =0
imum PV potential capacity of the building. The PV penetration ratio, (1 + r )t (4g)
PV penetration (4b), compares the energy generated by the PV arrays
npv
with the annual energy demand. The battery autonomy ratio, BAT auto LCOE = l
L (·∑t P oadt ·TSt )
(4c), corresponds to the ratio between the battery capacity and the ∑i (4h)
(1 + r )i
mean daily energy demand of the building. This metric can be under-
stood as the fraction of a day that can be covered by the battery in the ∑t min(Ptl oad, Pte xp + Ptl oad − Pti mp)
event of a blackout. From an operation perspective, the PV curtailement SS =
∑t Ptl oad (4i)
ratio, PV cur (4d), is the fraction of the energy that is curtailed from the
PV generation. The self-sufficiency SS (4i) is the fraction of the energy where cft is the net cash flow (investment + maintenance cost +
demand that is self-covered by the PV-battery system. The definition of operational cost, including battery replacement) at time t and opex t0 is
(4i) is derived from [11]. To assess how the buildings interact with the the original operating cost without the investment in the PV and bat-
grid, we defined in [10] a grid usage ratio, GU IMP,EXP (4e), as the tery.
ratio between the maximum withdrawn/injected power and the max- The resolution of the power-flow equations allows for extracting the
imum load. Finally, from an economic perspective, the discounted voltage (in per unit, pu) at every node of the network and the current
payback period, dpp (4g), (time to recover the investment) is of crucial flowing through every line. Given the lines’ properties, in particular the
interest to evaluate the profitability of the proposed economic frame- maximum allowable current, a representative metric for grid conges-
work. The levelized cost of energy demand (lcoe )(4h) can be compared tion is the line loading level. We consider separately the situation when
with the import tariffs of each scenario and provide an indication of the the bus voltage at an injection point is above 1 pu and when it is below

3
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

Fig. 4. Scenario preview for a typical day with PV generation.

1 pu, and use the 95th percentile of the bus voltage deviation. This (from similar buildings) and allocate smart-meter time-series to each
allows us to distinguish when there is a local excess of energy from meter. The roof’s surface area, azimuth, and tilt are known for each
when there is a local deficit of energy. Finally, one of the key issues for building. With this information, the maximum potential PV hosting
the high penetration of distributed stochastic generators is the reverse- capacity of each roof and then of each building could be calculated.
power flow occurring at the medium- to low-voltage transformer. For Fig. 3 shows this maximum PV capacity and the building’s annual en-
this reason, the load duration curve enables assessing the requirement ergy demand.
in terms of power that has to flow into and out of the low-voltage grid. Five tariff scenarios, and their corresponding coefficient from (3b),
(3c), (3d), are defined in Table 1. The first three scenarios consider
volumetric tariffs in which the costs/revenues are proportional to the
3. Case study
exchanged energy according to a given energy tariff. The first scenario
is a reference tariff that is constant throughout the day. The second
The methodology is applied to a case study in Rolle (Switzerland)
scenario, called ”solar”, promotes consumption when solar irradiance is
where a low-voltage distribution grid has been modeled (Fig. 3). The
higher by setting a low energy rate during the time period 11 h-15 h.
grid hosts 41 buildings. The properties of each building are known
thanks to publicly available geographical information system1. The
annual energy demand for each meter in the grid was provided, al-
1
lowing us to match the meters with real smart-meter measurements https://www.asitvd.ch/

4
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

Table 1 In order to not unintentionally incentive or penalize PV installation,

Tariff scenarios or change the value of a unit of electric energy, the tariffs (except the
Scenario Description Tariff (CHF cts/kWh) reference one) have been calibrated before running the building opti-
mization (reminding Fig. 1). The calibration aims here at keeping the
reference t impt : 21.02 DSO revenues (from selling energy minus cost of buying exported
t expt : 8.16 power) of the network remains as close as possible to the net profit of
solar t impt ∈ 11h :15h : 14.68
the reference scenario, under the hypothesis that all buildings will be-
t expt ∈ 11h :15h : 7.07
have exactly the same way as under the reference scenario. With this in
t impt ∉ 11h :15h : 23.17
mind, the values of the coefficient in Table 1 have been found em-
t expt ∉ 11h :15h : 11.12
spot market t impt : EPEX*3.9468 pirically.
t expt : EPEX*1.604 The PV and battery price levels have been set according to the re-
capacity t impt : 15.91 ference year 2025. The price projections have been extracted from the
t expt : 12.09 IRENA report [12] and calibrated for the Swiss price levels provided in
t max : 5.02 CHF/kW/month a recent market study [13] using the empirical model presented in [10].
block rate Power (kW) akimp akexp The resulting PV and battery cost function is decomposed into a fixed
0 to 1 13.72 13.07 and a linear component, CFpv, bat and C pv, bat respectively. These linear cost
1 to 2 15.06 11.73
functions are annualized and used along with the PV yearly main-
2 to 4 16.80 9.99
4 to 6 19.07 7.73 tenance cost and and battery lifetime in 1. All parameters of the pro-
6 to 8 22.01 4.79 blem, except the tariffs’ parameters, are provided in Table 2.
8 to 10 25.83 0.96
4. Results and discussion

Table 2 The optimizations of the 41 buildings were performed on an Intel(R)

Parameters
Xeon(R) CPU E5-2630 v3 @ 2.40 GHz processor with 8 Cores and 32 GB
Parameter Value Description of RAM using GUROBI [14] to solve the mixed-integer-linear problem.
The load flow problems were then solved for each time step with a
GENERAL T 35040 number of time steps
resolution of 15 min using PANDAPOWER [15]. The computation time for
M 12 number of months
ts 900 s time steps the building optimizations depends very much on the scenario as re-
L 25 years system lifetime ported in Table 3. In this table, the total column is the sum of the op-
r 3% discount rate timization across the 41 buildings’ optimizations. The computation time
pv a
maximum building PV capacity
PV Pcap ,max, b required to solve the load flow problem is negligible compared to the
CFpv 10049 CHF PV fixed cost b
total optimization time.
C pv 1.05 CHF/W PV specific costs b
b
BAT C bat 229 CHF/kWh battery specific cost
4.1. Design and operation of the PV-battery energy systems
CFbat 0 CHF battery fixed cost b

a
The resulting designs are pictured in Fig. 5. In all scenarios, except
data from the geographical information system
b the block rate tariff, almost all the roofs are covered with PV leading to
same value for all buildings
a PV hosting value close to one. The block rate scenario however limits
the penetration of PV with a PV hosting of ca. 0.5. The reason is the
Table 3
nature of the block rate that favor smaller PV installation, because the
Computation time statistics
marginal export revenue decreases with increasing exported power.
Scenario Building optimization Load flow Regarding battery size, investment in such technology is driven by
economic opportunities, namely by variations in the electricity price
min max median total
[min] [min] [min] [h] [min] (solar and spot market tariff scenarios) or by a strong incentive to limit
the exchanged power (capacity tariff scenario). Although this last as-
reference 3.7 57.3 13.6 11.8 6.6 pect is also present in the block rate scenario, the incentive is lower,
solar 2.4 41.6 11.3 8.8 6.6
leading to lower relative battery size. In terms of grid usage, only the
spot market 2.4 43.6 13.1 10.2 6.8
capacity 31.8 147.2 90.4 60.6 6.6 capacity tariff and block rate tariff scenarios provide a clear stimulus to
block rate 43.5 321.2 103.9 81.2 6.6 reduce the maximum power exchanged with the grid. The spot market
and solar tariff scenarios, due to the volatility of electricity prices, tend
to increase the power injected or withdrawn from the grid. On the
The third scenario mirrors the spot market price (specifically the in- economic side, the discounted payback periods are similar, ranging
traday continous price from the EPEX market data2). The fourth sce- from 14 to 23 years for all scenarios. The median is, however, higher for
nario is a mix of a volumetric and capacity-based tariffs, for which the the block rate tariff, 21 years against 19 years for the other scenarios.
cost is proportional to the monthly maximum power exchanged with The relative size of the battery does not scale linearly with PV pe-
the grid and the energy consumed from the grid, while the revenue is netration, as depicted in Fig. 6a, except for the dynamic volumetric
proportional to the energy injected into the grid. The fifth scenario tariffs (spot market and solar). For the capacity and block rate tariffs,
considers a block rate tariff, in which the cost/revenue is proportional low PV penetration, underlying a small PV production, i.e. small PV
to the energy exchanged with the grid, but the energy cost depends on capacity, compared to the energy demand, tends to increase the battery
the power level at which the energy is exchanged. For each scenario, autonomy ratio in order to limit the imported power. Conversely, at
the coefficients of (3b), (3c) and (3d) that do not appear in Table 1 are high PV penetration, the battery autonomy tends to decrease for the
equal to zero. capacity and block rate tariffs. For the first case, the role of the battery
to cut injection peaks is replaced by the curtailment of the PV gen-
eration (Fig. 7). As curtailing is free (it is not imposed but it results from
2
EPEX price for 2018 https://www.epexspot.com/en/market-data/ the optimization of the system operation), there is no need to invest in
intradaycontinuous/intraday-table/-/CH batteries for this purpose. For the block rate scenario, high injection is

5
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

Fig. 5. PV hosting ratio, battery autonomy, grid usage ratio (import/export) and discounted pay back period for all scenarios. Metrics are defined in (4). Solid lines
are the median, dashed lines are the 75th percentile, and dotted lines are the 25th percentile.

Fig. 7. Ratio of energy curtailed and PV penetration. In the inset, the bars are
the weekly ratio of energy curtailed (left axis) and lines the weekly minimum
t exp (right axis).

Fig. 6. Upper (a), battery autonomy versus PV penetration, lower (b) self-suf-
ficiency level against the battery autonomy ratio.

not penalized; the marginal revenue is just decreased. Thus, it limits the Fig. 8. Export vs importgrid usage ratio.
profitability of having a high PV capacity compared to its energy de-
mand level, but it requires neither curtailment of the PV energy nor technologies since they provide economic opportunities to generate
investing in storage technologies. The fraction of energy curtailed is profit for the building owners. The capacity tariff promotes investment
zero for all scenarios except for the capacity and spot market tariffs. For in storage which main action is to reduce consumption peaks. The block
the latter, the small fraction of curtailed energy is due to negative spot rate tariff promotes smaller PV penetration (thus, PV capacity) and
market prices as shown in the inset of Fig. 7. As a general trend, a larger battery capacity but achieves a self-sufficiency level similar to the re-
battery size (relative to the building energy demand) increases the self- ference case. As pictured in Fig. 5, these considerations have an impact
sufficiency ratio as shown in Fig. 6b. This trend is very pronounced for on grid usage behavior. In particular, the grid usage ratios are higher
the spot market, solar and capacity tariffs, although a saturation ap- (regardless when importing or exporting) for both the spot market and
pears for the latter for large battery autonomy. solar tariff. It is especially pronounced for the spot market case. Con-
In summary, compared to the reference scenario, dynamic volu- versely, capacity tariffs significantly reduce the grid usage ratio for
metric tariffs (solar and spot market) promote investments in storage import, while the block rate tariff lowers both. Fig. 8 illustrates these

6
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

Fig. 9. Load duration curve at the transformer. Dots on the vertical axis indicate
the total installed PV capacity per scenario. The nominal transformer capacity is
400 kW. Negative values indicate power flow from the high-voltage toward the
low-voltage side.

Fig. 11. Voltage deviation distribution.

th
Fig. 10. 95 percentile of the line loading level.

observations. This figure allows us to distinguish between three types of

grid user: the exporters (grid usage ratio for export above 1) and even
reduce their import grid usage by covering their own consumption; the
energy traders, who buy or sell energy to maximize their profit (import
and export grid usage above 1); and the low grid users, who interact less
with the network. Almost all buildings fall in this category for the block
rate scenario. The following will show how these local design and op-
eration adaptations affect the network operational metrics.

4.2. Low-voltage grid impact

In order to have a complete overview of the grid reaction to the

different tariffs (with PV installations and batteries always present),
two additional scenarios are added. The first one considers only the Fig. 12. Levelized cost of energy demand per scenario.
original load without any investment in PV or batteries; the second
considers that all roofs are covered with PV (regardless of the profit-
maximum nominal current of the line). Fig. 10 shows that line loading
ability of such a decision) but with no investment in batteries. These
level is significantly higher for all scenarios including PV, with the most
two scenarios gives reference values for the grid impact metrics.
extreme values attained under the spot market and full PV scenarios.
The load duration curve in Fig. 9 highlights the violation of the
The block rate tariff helps to significantly reduce the loading level of the
transformer power capacity for reverse power flow. All scenarios, ex-
lines. In this case, even the most loaded lines are less congested than in
cept the load-only case, experience a maximum power flowing from the
the load-only scenario.
low-voltage side to the high-voltage side above 400 kW. The block rate
One of the main concerns of grid operators regarding high PV pe-
tariff, with a significantly lower installed PV capacity, has the lowest
netration is to keep voltage levels to a value as close as possible to 1 pu.
maximum reverse power, but a significant number of hours are above
As a matter of fact, all scenarios, except the spot market one, fulfill the
400 kW. The most-demanding scenario is the spot market scenario
criteria EN50160, meaning that the voltage levels fall within +-10% of
which exhibits the highest power demand and the highest injection
the nominal voltage for 95% of each week. When considering only the
power. The solar tariff also shows a significant increase in power de-
case when the voltage deviations exceed 1 pu, Fig. 11a demonstrates
mand compared to the other scenarios. This has direct consequences for
the effectiveness of the capacity and block rate tariffs in terms of load
the level of loading of the lines (ratio between current and the

7
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

5. Conclusions

In this study, the optimization of 41 buildings under five different

tariff scenarios, including three volumetric tariffs, one combination of
volumetric and capacity tariff and one block rate tariff, was performed.
The resulting systems vary in term of installed PV capacity and battery
capacity. The highest PV penetration is achieved with the capacity tariff
while this scenario significantly reduces the voltage deviation and the
line loading level. Volumetric tariffs, with high price volatility such as
in the spot market scenario, lead to more investment and profitability of
the batteries but also increase stress on the network. Although the block
rate scenario promotes smaller PV installation, it achieves the smallest
median cost of providing energy for the end-users (from 18 cts/kWh in
the reference case down to 14 cts/kWh). It reduces the 95th percentile
of the positive voltage deviation of the bus with the largest deviation
Fig. 13. Net revenue for the distribution system operator. from 6% to 3.6% and reduces the maximum reverse power from
1060 kW to 550 kW, while remaining above the 400 kW nominal power
management, as both lower the deviation of the most sensible bus of the transformer. Further studies should elaborate on the design of
compared to the reference and full PV case. Note that the load-only block rate tariffs to mitigate network impact and incentivize high pe-
scenario is not displayed in this figure because the voltage levels never netration of PV. Indeed, the design and operation of PV-battery systems
exceed 1 pu. Alternatively, when the load level is below 1 pu (Fig. 11b), strongly depends on the tariffs parameters. The approach chosen to
only the spot market case significantly increases voltage deviations. calibrate the tariffs has very significant impact on the resulting system.
These observations highlight that advanced tariff structures can Hence, a deeper analysis on how to design and calibrate such tariffs is
have two competing impacts. On the one hand, they may help miti- needed. The effect of a growing penetration of electric vehicles and heat
gating the grid impact of distributed generation by promoting either pumps should also be considered.
small PV installations or moderate grid exchange. On the other hand,
they can bring economic opportunities for significant investments in Declaration of Competing Interest
batteries and PV capacities but may increase the stress on the grid. The
most-concerning aspect is the transformer capacity to bring the excess The authors declare that they have no known competing financial
power from the low-voltage to medium-voltage side. Over/under-vol- interests or personal relationships that could have appeared to influ-
tage and overloading of the lines are in all cases far less concerning. ence the work reported in this paper.
Although each scenario has been carefully calibrated to globally keep
similar DSO revenues (with respect to the reference case), it is worth Acknowledgement
investigating a posteriori these revenues and the levelized cost of the
energy demand (LCOE) in order to make sure that the calibrations do This research is part of the activities of the Swiss Centre for
not lead to a net increase in the energy price. Competence in Energy Research on the Future Swiss Electrical
Infrastructure (SCCER-FURIES), which is financially supported by the
Swiss Innovation Agency (Innosuisse - SCCER program).
4.3. Economic aspects
References
In the load-only scenario, the LCOE is 21 CHF cts/kWh (corre-
sponding to the import tariff of the reference scenario, see Fig. 12), [1] R. Tonkoski, D. Turcotte, T.H.M. El-Fouly, Impact of High PV Penetration on
while the LCOE can become higher in the full PV scenario and ex- Voltage Profiles in Residential Neighborhoods, IEEE Transactions on Sustainable
Energy 3 (3) (2012) 518–527, https://doi.org/10.1109/TSTE.2012.2191425.
plaining why, in the reference case, some roofs are not covered with PV. [2] P. Arboleya, M. Huerta, B. Mohamed, C. Gonzalez-Moran, X. Dominguez, Assessing
Additionally, in the solar and spot market scenarios, only a small the effect of nearly-zero energy buildings on distribution systems by means of quasi-
fraction of the buildings exhibits an LCOE exceeding 21 cts/kWh, while static time series power flow simulations, 2017 IEEE Power & Energy Society
General Meeting, IEEE, 2017, pp. 1–5, https://doi.org/10.1109/PESGM.2017.
the large majority would gain from switching to these tariff structures. 8274228.
For the capacity tariff scenario, a significant fraction of the buildings [3] D.I. Hidalgo-Rodriguez, J. Myrzik, Optimal operation of interconnected home-mi-
has an LCOE higher than the reference value, showing that, despite its crogrids with flexible thermal loads: A comparison of decentralized, centralized,
and hierarchical-distributed model predictive control, 20th Power Systems
positive impact on grid operation, such a tariff comes with a price for
Computation Conference, PSCC 2018 (2018), https://doi.org/10.23919/PSCC.
some building owners. The block rate scenario, though less prone for PV 2018.8442807.
and battery investment, presents an LCOE lower than the reference [4] N. Hashemipour, T. Niknam, J. Aghaei, H. Farahmand, M. Korpas, M. Shafie-Khah,
values for all buildings. LCOE, however doesn’t give any indication of G.J. Osorio, J.P. Catalao, A linear multi-objective operation model for smart dis-
tribution systems coordinating tap-changers, photovoltaics and battery energy
the DSO revenues as shown in Fig. 13. Recalling that each tariff has storage, 20th Power Systems Computation Conference, PSCC 2018 (309048)
been calibrated to keep the revenue for the DSO approximately equal to (2018), https://doi.org/10.23919/PSCC.2018.8442849.
the one of the reference scenario, Fig. 13 highlights that after the op- [5] A. Sani Hassan, L. Cipcigan, N. Jenkins, Impact of optimised distributed energy
resources on local grid constraints, Energy 142 (2018) 878–895, https://doi.org/
timization phase, the revenues change significantly due to the fact that 10.1016/J.ENERGY.2017.10.074.
each building optimize its operation in order to increase its own profit. [6] H. Wang, N. Good, E.A. Cesena, P. Mancarella, Co-optimization of a multi-energy
The only presence of PV generators significantly lowers the DSO rev- microgrid considering multiple services, 20th Power Systems Computation
Conference, PSCC 2018 (2018), https://doi.org/10.23919/PSCC.2018.8442772.
enues as shown by comparing the only loads scenario with the full PV [7] M. Schreiber, M.E. Wainstein, P. Hochloff, R. Dargaville, Flexible electricity tariffs:
one. The block rate scenario actually slightly increases the revenue Power and energy price signals designed for a smarter grid, Energy 93 (2015)
because the total installed PV capacity is smaller. The design of such a 2568–2581, https://doi.org/10.1016/J.ENERGY.2015.10.067.
[8] T.A. Deetjen, J.S. Vitter, A.S. Reimers, M.E. Webber, Optimal dispatch and equip-
tariff is a matter of compromise between promoting distributed re- ment sizing of a residential central utility plant for improving rooftop solar in-
newable energy and mitigating grid impact. There is a lack of literature tegration, Energy 147 (2018) 1044–1059, https://doi.org/10.1016/J.ENERGY.
for designing such block rate tariffs. 2018.01.110.
[9] Z. Ren, G. Grozev, A. Higgins, Modelling impact of PV battery systems on energy

8
J. Holweger, et al. Electric Power Systems Research 189 (2020) 106763

consumption and bill savings of Australian houses under alternative tariff struc- [12] IRENA, Electricity Storage and Renewables: Costs and Markets To 2030, (2017).
tures, Renewable Energy 89 (2016) 317–330, https://doi.org/10.1016/J.RENENE. [13] L. Deschaintre, F. Jacqmin, Rapport ý Observation du marché photovoltaïque 2018
2015.12.021. é, Technical Report, (2019).
[10] L. Bloch, J. Holweger, C. Ballif, N. Wyrsch, Impact of advanced electricity tariff [14] L. Gurobi Optimization, Gurobi optimizer 8.1, reference manual, 2019.
structures on the optimal design, operation and profitability of a grid-connected PV [15] L. Thurner, A. Scheidler, F. Schäfer, J. Menke, J. Dollichon, F. Meier, S. Meinecke,
system with energy storage, Energy Informatics 2 (16) (2019), https://doi.org/10. M. Braun, pandapower – an open-source python tool for convenient modeling,
1186/s42162-019-0085-z. analysis, and optimization of electric power systems, IEEE Transactions on Power
[11] R. Luthander, J. Widén, D. Nilsson, J. Palm, Photovoltaic self-consumption in Systems 33 (6) (2018) 6510–6521, https://doi.org/10.1109/TPWRS.2018.
buildings: A review, Applied Energy 142 (2015) 80–94, https://doi.org/10.1016/J. 2829021.
APENERGY.2014.12.028.