Harmonic Allocation Allowing for Uncertainty of

Distributed Generation in MV/LV Power Systems

Jason David∗ , Vic Gosbell, Duane A. Robinson, Sarath Perera, Sean Elphick
Australian Power Quality & Reliability Centre
University of Wollongong, Australia
∗ [email protected]

Abstract—A key component to management of harmonic volt- Where -

age levels in MV and LV distribution systems is the allocation of
harmonic emissions to MV and LV customers. The principals of GhM V is the total global impact that MV loads are to have of
harmonic emission allocations outlined in IEC Technical Report the distortion at the Point of Common Coupling (PCC).
61000-3-6 are relatively straight forward to apply when both Si is the agreed power of the connecting installation (load or
the capacity of the network and the extent of all future load generation).
connections are known. However, allocations are made more St is the total supply capacity of the system accounting for
complex when distributed generation, typically power electronic
based, are added to the mix of customer connections. This future connections and diversity.
added complexity is due to distributed generation potentially α is an exponent to allow for diversity, suggested values are
impacting the hosting capacity of the network, i.e. enabling more given in [1].
customer connections, and the uncertainty for network planners
in quantifying the likelihood of future distributed generation In radial network where distributed generation (DG) is
connections, given the prolific growth of renewables. This paper not present, the IEC approach is relatively simple, provided
introduces a tapered probabilistic approach for network planners present and future installations are well known. The reality
to undertake harmonic emission allocations for MV and LV is however that the relationship between the total distorting
networks which are likely to host significant levels of renewable load connected to a system and St is becoming increasingly
based distribution generation. The method is demonstrated using
a simplified network model. tenuous due to the expansion of DG. Whilst DG does not
operate as a network load at the fundamental frequency, in
Index Terms—Harmonic Allocation, Renewables, Distributed terms of harmonic emissions, it is equivalent to a conven-
Generation, Distribution Systems, IEC Standards
tional energy consuming installation and must be provided a
harmonic allocation. There remains a lack of agreement in
I. I NTRODUCTION determining the DG hosting capacity due to the number of
Harmonic emissions allocation is a task which is nor- system and technological variables involved [2], [3]. However,
mally carried out by network operators to manage harmonic even defining this value for a network does not guarantee
voltage levels within the power system. There are a num- that the corresponding level of DG penetration will ever
ber of approaches used worldwide, one of the more pop- be connected. Hence, a robust methodology is required for
ular being the methodology described in Technical Report harmonic allocation which allows for the uncertainty inher-
IEC TR 61000.3.6:1996 [1]. Whilst there are updated versions ently present with regard to DG connections. Any alterations
of this technical report, countries such as Australia still refer considered to the present IEC approach should also reduce
to this version to define a standard harmonic allocation proce- the dependency on system supply capability, St , and replace
dure. Given that the superseded version of the technical report it with a more suitable term to account for all future distorting
is still applied, it will be used in this paper to highlight the loads i.e. both consuming and generating installations. Failure
difficulties in applying the process in the presence of uncertain to address these shortcomings in the methodology is likely to
levels of Distributed Generation (DG) penetration. The process result in restricting DG integration levels to networks where
given in Stage 2 of [1] calculates the harmonic emission (a) significant under-utilisation of the harmonic absorption
allocation to a new installation based on its agreed power, capability, or (b) exceedance of stipulated harmonic planning
Si , in relation to the total agreed power (present and future) levels may result.
of the system, St , i.e. the apparent power of the system at the A review of the present IEC methodology is presented in
time that it is fully loaded. This is given the name of Supply this paper for radial LV and MV power systems. Two proposed
Capability in [1]. The resulting allocated harmonic level is alterations to the IEC method to deal with the uncertainty of
calculated using (1) (explained in detail in later sections). future penetration levels of DG have been developed and are
evaluated in this paper. The proposed approaches attempt to be
 (1/α)
Si pragmatic by both reducing the complexity of the allocation
EU hi = GhM V (1) process and minimising the impact on existing installations.
St
978-1-7281-3697-4/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: Monash University. Downloaded on April 13,2023 at 00:46:00 UTC from IEEE Xplore. Restrictions apply.
 The paper will first introduce a simple system on which Scenario (iii) - All future DG connections are uncertain.
all allocations will be performed and compared. Example This is the worst-case scenario and as such, is the more
allocations will then be presented using firstly the IEC method, difficult scenario for the network planner to account for.
a number of updates to this method are then compared with the
III. H ARMONIC A LLOCATION U SING THE IEC M ETHOD
ideal scenario in which all present and future connections are
known and allocated as such. The impact of each method to the The IEC harmonic emission allocation method presented
allocated harmonic voltage to each customer is also compared, in [1] is a staged approach giving allocations related to the
the paper is then concluded with reference to future work to agreed power of the installation (Si ) and the supply capability
be completed. of the network (St ). When no DG is considered, it is generally
accepted that the supply capability St can be established
using the net thermal capacity of the system, i.e. the lesser
II. M ETHODOLOGY
of the transformer, busbar or cable/line ratings, in the hope
The examples provided in this paper are for a simple that it is representative of all future connections. In principle
distribution system as shown in Fig. 1 for a harmonic order however, St is the summation of all future capacity allocations
in which α = 1.4 (as per [1], exponent to account for connected downstream, with diversity considered, including
diversity). The system under study is a relatively ideal radial DG units where applicable [4]. These two approaches may
network i.e. no long feeder lengths, capacitor banks, etc. All result in values for St which differ significantly, especially
existing connections can be modelled directly connected to where DG units are installed, and some pragmatism is required
the MV busbar, LV loads are lumped into MV loads and from the network operator. As an example, the system in
future connections are assumed to have negligible impact on Fig. 1 is connecting a new installation (consuming load or
the harmonic resonance of the system. Where - generator) and must allocate a harmonic emission limit prior
to connection. To do this, only Si , St and the planning level
at the PCC (Gh,M V ) need to be known, for this example these
are given as:
St = 10 M V A
Stx Si = 1 M V A
Gh,M V = 1 pu

St It should be noted that Gh,M V is set to 1 pu with

Vbase = Vh,planninglevel , the resulting allocation is then a
pu allocation of the planning level, simplifying the calculation
process. The method then uses (1) to calculate the resulting
harmonic voltage allocation for the installation:
Si Sexisting  1/α
Si
EU hi = Gh,M V ·
St
Fig. 1. Simple Network for Study
 1/1.4
1 MV A
= 1 pu ·
10 M V A
Stx is the rated MVA power of the local substation trans- = 0.193 pu
former.
Sexisting is the aggregate power of existing connections on If this process is followed for all connections, and DG is
the network, load or generation. treated identically to consuming loads, the allocated harmonic
levels, using diversified summation as per (2), for the network
The proposed alterations to existing methods in this paper
in Fig. 1, would result in an allocated harmonic voltage profile
attempt to introduce more flexibility in the allocation process
as shown in Fig. 2.
to address uncertainty related to DG penetration levels. As
α α α 1/α
such, a number of network scenarios are used to compare Vh (n) = (Vh1 + Vh2 + ... + Vhn ) (2)
efficacy of each method, these are as below:
Difficulty in the presence of DG occurs due to the unknown
Scenario (i) - All future connections, both generating and level of penetration expected for a system. In the scenario that
consuming are known and certain. This is the ideal case and DG was not expected, the supply capability defines the total
simplest to plan for, although very unlikely to occur in reality. expected load on the network. Defining a similar limit for DG
Scenario (ii) - Some future DG connections are certain, on the same network or including it within the St term does
beyond which point the network is less certain of further not guarantee that the specified level of DG penetration will
DG penetration. This is considered a likely scenario for many eventuate due to many influencing factors such as financial
networks as DG connections become more prominent. viability, public opinion of energy generation, governmental
initiatives and support, among many others.

Authorized licensed use limited to: Monash University. Downloaded on April 13,2023 at 00:46:00 UTC from IEEE Xplore. Restrictions apply.
 planning levels at the point that the aggregate of the agreed
power of all connections (load and generation) is equal to St ,
shown in Fig. 3 (blue). However for less certain scenarios, such
as Scenario (ii) and Scenario (iii), this could result in under-
utilisation, or exceedance of the network harmonic distortion
planning levels.
As an example, for Scenario (ii); if the network operator had
3 MVA of DG to be connected to the network and assumed
that another 2 MVA would be connected in the future but
this never eventuated, the resulting allocated harmonic levels
would reach approximately 0.9 pu of planning levels, this is
also shown in Fig. 3 (red).
Similarly, if the network estimated that no more DG was
likely to be connected after the initial 3 MVA, allocated
harmonic levels would reach planning levels once all certain
distorting loads were connected. In this case the planning lev-
els are met before total potential DG is connected (compared to
Fig. 2. Allocation example using IEC method
Scenario (i)). Any proposed alteration considered must reach
as close to planning levels as possible at the point that all
A simple approach would be for the network operator to loads and generation are connected, effectively reducing the
make an educated guess in relation to the expected level of impact that uncertainty has on the allocation process.
future DG penetration. This does not however, fully address
the issue of uncertainty and an incorrect guess is capable
of leading to harmonic levels exceeding predefined planning
levels or under-utilisation of the harmonic absorption capa-
bility of the system. A robust methodology must therefore be
developed that pragmatically mitigates the uncertainty inherent
in the presence of DG. This is explored further in following
sections.

IV. T OTAL D ISTORTING L OAD

The definition of St in [1] has since been updated in
subsequent versions of the technical report. The 2012 iteration
of the report [4], states that St may include the contribution
from DG however does not specifically detail how this could
be implemented.
It is proposed therefore that the definition be updated to
Total Distorting Load and include the total forecast agreed
power for consuming loads and some level of agreed power Fig. 3. Comparison of allocated harmonic levels using (3) for Sce-
for generating installations as both are sources of harmonics nario (i) (blue), (ii) - Overestimation of DG (red), (iii) - Underestimation
and must be accounted for in the allocation process. The first, of DG (green)
and simplest approach to implementing this is shown in (3).
It is clear based on these outcomes that the allocation
St = Sload + SDG (3) method must be more robust and immune to the impacts of
uncertainty related to the penetration of DG. One approach is
Where - to define the level of DG with a probability curve, reducing the
Sload is the total forecast agreed power of standard consuming impact that uncertain DG connections have on the allocation to
loads certain installations. This is explored further in the following
SDG is the total forecast agreed power of all present and section.
future DG connections on the network. In-lieu of certainty
an estimate would be required for this variable V. D EFINING A P ROBABILITY C URVE FOR DG
For Scenario (i) (defined in Section II), assume a network C ONNECTIONS
operator has an additional 5 MVA of DG to be connected to In place of defining uncertain limits on the level of future
the network. If St were calculated using (3) with all future DG penetration for a network, it may be more pragmatic to
distorting loads known and certain, the allocated harmonic define the level of DG that is certain for the network and
levels, calculated for each installation using (1), would reach then apply a scaling factor to the probability of further DG

Authorized licensed use limited to: Monash University. Downloaded on April 13,2023 at 00:46:00 UTC from IEEE Xplore. Restrictions apply.
penetration that is less certain. Fig. 4 provides an example A. Example Allocation using the Probability Curve
probability curve in which the lightly shaded area accounts for Scenario (ii) is again used as a comparison to determine the
installations that are certain (P = 1) and the darker shaded efficacy of this approach, Scenario (iii) has also been intro-
area progressively reduces the probability of future DG beyond duced to determine the worst-case-scenario. For this example,
the connections that are certain. the network operator has identified the following:

TABLE I
PARAMETERS DEFINED BY NETWORK OPERATOR

Value
Parameter
Scenario (ii) Scenario (iii)
St 10 MVA
SDG,max 10 MVA
SDG,certain 5 MVA 0 MVA

Further, the probability curve shown in Fig. 4 will be used

with values from Table I used to define key points on the
curve and all consuming installations are connected prior
to the connection of generating installations. The resulting
comparative allocated voltage levels are shown in Fig. 5. It
Fig. 4. Example of a probability curve for SDG of a network can be seen that the resulting allocated emission levels for
Scenario (ii) result in 0.98 pu of the planning levels being
Using this method, a network planning engineer would utilised. Whilst the final allocated emission levels for Scenario
be required to define a maximum level of DG penetration (iii) reach 0.93 pu, provided the level of uncertainty in these
(SDG,max ) that the network is capable of hosting and also examples, the outcomes provide acceptable results.
define any level of DG that is certain to be connected. As
previously mentioned, the process of defining a DG hosting
capacity for a system currently lacks an industry accepted
process. In the meantime, a reasonable estimate can be made
and the method attempts to minimise this impact of the
required estimate by allocating more to all installations that are
certain, and reducing the allocation to uncertain installations
so that; if the connections never eventuate a greater portion
of the planning levels would already be allocated, minimising
under-utilisation of the absorption capability of the network.
Using this approach, SDG in (3) would then be defined as
in (4).

SDG,max − SDG,connected
SDG = SDG,connected + (4)
P
Where -
SDG is the total forecast agreed power of DG installations at
Fig. 5. Comparison of allocated harmonic levels using probability curve for
the present time. Due to uncertainty, this value changes Scenario (i) (Blue), Scenario (ii) (Red) and Scenario (iii) (Green)
as more connections are made and is influenced by the
probability curve defined for the network. Further, it should be noted that when allocating for Scenario
SDG,connected is the current day level of DG penetration. (ii), the resulting harmonic distortion reaches ≈ 0.81 pu at
SDG,max is the predefined maximum level of DG penetration the point that SDG,Connected = SDG,Certain and 0.79 pu for
able to be hosted on the network. Scenario (iii) at the same level of DG penetration suggesting
P is the probability of the connection of future installations, the sensitivity of defining SDG,Certain is relatively low. This
derived from the probability curve. example also shows that the approach increases the use of
The definition of the probability curve beyond SDG,certain available planning levels when there is a higher level of
would be defined by the network planning engineer and intro- certainty in the future loading of the network, a desirable
duces further uncertainty. A brief study suggests however that outcome.
the process is not particularly sensitive to this definition and Whilst the outcomes of the above examples provide a
the curve shown in Fig. 4 provides a reasonable assumption positive influence on the resulting allocated emissions levels
in-lieu of more detailed information. in the presence of uncertainty in DG penetration levels, the

Authorized licensed use limited to: Monash University. Downloaded on April 13,2023 at 00:46:00 UTC from IEEE Xplore. Restrictions apply.
reality and logistics of implementing such an approach may voltage emissions as the approach provides a larger allocation
not be practical. Issues may arise related to keeping track to the installations that are certain. The results for both
of the position of a site on the probability curve (as the scenarios are shown in Table II.
level of connected DG increases over time). Further, the
required definitions of SDG,max and SDG,certain may render TABLE II
R ESULTS OF Scenarios (ii) AND (iii) USING THE MAINTAINED RESERVE
this approach somewhat difficult to apply in a real system. METHOD
Based on these arguments, it would be fair to suggest
that a simpler approach with similar outcomes would be Value
Parameter
Scenario (ii) Scenario (iii)
preferred. This is presented in the following section in which Sconnected 20 M V A 16 M V A
a maintained reserve of future distorting load is applied to St Vh 0.97 pu 1 pu
suggesting that future distorting loads are always possible.
VI. M AINTAINED R ESERVE M ETHODOLOGY For senario (ii) in which a level of certain DG exists,
the total amount of DG penetration able to be connected to
To further simplify the process yet maintain the observed the network is 4 M V A greater than scenario (iii) in which
impact of the probabilistic approach in Section V, another there is no certainty and planning levels are reached when
method is presented that reduces the allocation to every Sconnected = 16 M V A.
subsequent connection past SDG,certain by increasing the It can be seen in Fig. 6 that there is only a small difference in
total distorting load as the system becomes more loaded by the final allocated voltage emission level between scenarios (i)
both consuming and generating harmonic loads. A simple and (ii). This suggests that the use of the maintained reserve
description of the methodology is applying a buffer to St , methodology provides the desired outcome for the levels of
effectively increasing the value and reducing the allocation harmonic distortion on the network whilst being simple for
for connections that occur past St + SDG,Certain , practically the network operators to implement without requiring detailed
implemented by applying (5) to calculate St . studies or site-specific information related to the level of
harmonic loading present at the time of study. Further, a
St = max((St +SDG,Certain ), (Si +Spresent ))+Sbuf f er (5) greater degree of certainty in the level of DG penetration
increases the final use of planning levels once full loading
This results in a value for the total distorting load that is on the network is reached, approaching closer to the ‘ideal’
never achieved by the network and instead, over time, reduces outcome.
the allocation as more distorting loads are connected.
This equation effectively decreases the allocated limit by VII. C OMPARISON OF A LLOCATED VOLTAGES
adding a reserve (or buffer) to the St term. Scenarios (ii) and The practical impact of each of the approaches presented
(iii) are compared with the use of (5), using the values in in this paper will result in altered harmonic emission limits
Table I and an Sbuf f er of 5 M V A. The comparative allocated for connecting customers. A reduced allocation could lead to
harmonic emission levels can be seen in Fig. 6. harmonic mitigation being required which can be costly and
should be avoided where possible. An investigation therefore
of the impact that each method has on the allocation for each
customer is inspected. Note, for all methods presented in this
paper Scenario (i) results in the same allocation, this is because
all loads are already known and certain.
A. Allocated Voltages Using Tapered Probabilistic Approach
If a network were to implement the tapered probabilistic
approach identified in Section V, connecting customers would
be allocated a reduced harmonic voltage as DG begins to
connect. The degree of impact that this has on the resulting
allocations is dependent on the amount of DG identified as
certain when developing the probability curve such as that
shown in Fig. 4. The comparisons shown in Fig. 7 provides
the resulting allocated harmonic voltages for each customer as
network connections take place for each scenario considered.
It can be seen that reduced allocations begin at the point that
SDG,connected > SDG,Certain . In Fig. 7 the final allocation
Fig. 6. Comparison of buffer method for Scenario (i) (Blue), Scenario (ii)
(Red), and Scenario (iii) (Green) is returned to the Scenario(i) determination as it assumed
that the planning engineer recognises that it will be the
It can be seen that for Scenario (ii), in which there is some last allocation, however equally it could also be assumed
certainty of DG penetration, results in lower total harmonic to follow (extend) the existing curve. In the terms of the

Authorized licensed use limited to: Monash University. Downloaded on April 13,2023 at 00:46:00 UTC from IEEE Xplore. Restrictions apply.
 allow for uncertainty by allocating more to the certain installa-
tions, leaving less available planning level for the installations
that are uncertain. This can be readily seen in Fig. 8 in which
the green curve (Scenario (iii)) allocates a considerably larger
voltage to earlier installations, once the system reaches the
point SConnected = SCertain future DG installations are now
uncertain and receive a sliding reduction in voltage allocation.
This change in allocation can also be seen for Scenario (ii)
in which a greater amount of the planning levels are reserved
for certain allocations, again once the network has connected
all certain installations, a sliding reduction in the allocated
voltage is noticed.
VIII. C ONCLUSION
The increase in the number of DG installations currently
occurring is posing a risk to network operators when consid-
ering harmonic distortion management. Further, the level of
Fig. 7. Comparison of allocated levels for installations using probability uncertainty around these types of installations creates difficul-
method - Scenario (i) (Blue), Scenario (ii) (Red), Scenario (iii) (Green) ties when allocating harmonic levels to such installations.
This paper provides some background related to these diffi-
culties and proposes some simplified approaches that may be
probability method, including any certain DG installations into
implemented in order to address this uncertainty. In practical
the calculation results in these customers receiving greater
terms, a trade-off between maintaining harmonic distortion
allocations and potentially reducing the level of mitigation
below planning levels and placing fair emission limits on
required or removing the requirement altogether. The reader
customers attempting to utilise the absorption capacity of the
is also reminded that when a known level of DG is included
network must be carefully balanced. Two methods in particular
into this allocation method, the final allocated harmonic level
have been reviewed briefly, the first a probabilistic approach
is closer to 1 pu compared to when all DG is considered
related to uncertainty of DG penetration levels and a second
uncertain.
approach that applies a buffer value once installations that are
B. Allocated Voltages Using Maintained Reserve Method certain have been connected. Both methods reduce the total
The resulting harmonic voltage allocations for the main- allocation to uncertain installations and both result in reduced
tained reserve method presented in Section VI, shown in error (compared to an ideal scenario, i.e. no uncertainty) when
Fig. 8. a level of certain DG is accounted for.
Future work is required to understand the practicality of
each approach and ensure that the reviewed alterations are
likely introduce more flexibility in the harmonic management
process that are amenable to the industry. Further alterations
and updates to the IEC method should also be considered that
provide a simple and robust approach to harmonic allocation
and management for the network operators.
R EFERENCES
[1] International Electrotechnical Commission, “Electromagnetic compat-
ibility (EMC) - Part 3: Limits - Section 6: Assessment of emission
limits for distorting loads in MV and HV power systems - Basic EMC
publication,” IEC/TR 61000-3-6 Ed. 1.0b: 1996.
[2] Carlos A. N. Pereira, Joao A. Pecas Lopes, Manuel A. C. C.
Matos, “Assessment of the distributed generation hosting capacity
incorporating harmonic distortion limits,” Inter. Conf. on Smart En-
ergy Systems and Technologies (SEST), 10-12 September 2018. DOI:
10.1109/SEST.2018.8495843
[3] Namhun Cho, Hyungchan Lee, Rajatha Bhat, Kangjoon Heo, “Analysis
of Harmonic Hosting Capacity of IEEE Std. 519 with IEC 61000-3-
6 in Distribution Systems,” IEEE PES GTD Grand Inter. Conf. and
Exposition Asia (GTD Asia), 19-23 March 2019. DOI: 10.1109/GTDA-
Fig. 8. Comparison of allocated levels for installations using maintained sia.2019.8715918
reserve method - Scenario (i) (Blue), Scenario (ii) (Red), Scenario (iii) (Green) [4] International Electrotechnical Commission, “Technical Report: Electro-
magnetic Compatibility (EMC) Part 3.6: Limits-Assessment of emission
Similar to Fig. 7, a clear decrease in allocated voltages oc- limits for the connection of distorting installations to MV, HV and EHV
curs once DG begins to connect to the network. As previously power systems,” IEC TR 61000.3.6:2012, 2012.
mentioned, the methods presented in this paper attempt to

Authorized licensed use limited to: Monash University. Downloaded on April 13,2023 at 00:46:00 UTC from IEEE Xplore. Restrictions apply.