7.2.
APPENDIX ON NON-TECHNICAL LOSSES
7.2.1.
MORE
DETAIL AROUND NTL TYPES
AND EXTENT
Types of
NTL
Electricity theft is listed
by several sources
as a significant contributor to NTL5. The estimated
total global annual losses
due to electricity theft are more than $25 billion,
of which $4.5 billion
occur in lndia
[Depuru 2011]6.
Measurement, accounting and information errors are listed
in several other references7.
Unmetered supplies (including unmetered auxiliary services and unmetered own consumption) are also listed
frequently as sources of NTL8.
Time difference between meter
readings and period of calculation refers to energy usage estimations that have
to be made when losses are calculated for a defined period (e.g. a month or a
year) but the meter readings for the end date (e.g. 31st Dec 24h00) or for the start date (e.g. 1st Jan 00h00) are not available.
Size and types of losses in Europe
ln Europe, excluding the UK, NTL include theft, non-registered
consumption, own consumption, non-
metered supplies such as public lighting and errors in metering, billing and
data processing (including time lags between meter readings and statistical
calculation) [ERGEG 2008].
Total distribution losses range
from 2.3% (Sweden) to 11.8% (Poland), with Romania being an outlier at 13.5%
[ERGEG 1]. 19% of energy used in Turkey is illegal [Cavdar 2004]9.
Some new Member States have much
higher losses at the distribution level than the other Member States; possible
reasons include the condition of the networks
and higher-than-average amount
of NTL, üo$oe. g. from unmetered consumption, metering errors and theft
[ERGEG 2008].
References quoted in [Smith 2004]
state that a utility in Budapest estimates annual losses due to theft to be
6.5%.
ln Spain 35% to 45% of NTL are estimated to be due to
fraud [Monedero 2011].
One village in Romania
experienced 84% of total supplied energy
not being billed;
after implementation of
mitigation measures this was reduced to 26.1% (9.7% of which were TL)
[Harabagiu 2005].
One utility in the UK [SPEN
2014] lists the types of NTL it has encountered as illegal connections, meter tampering/bypassing (by a minority of customers), unmetered supplies that are
incorrectly or inaccurately estimated and billing errors due to incorrect
recording of consumption. Total
losses in its networks were estimated to be 5.8% to 6.0% in 2009-10.
5 Examples are [Antman 2009], [ERGEG 2008], [SPEN 2014],
[ENW 2015], [SSEPD 2015], [WPD 2015],
[Monedero 2011], [Nagi 2009],
[Bastos 2009], [Beutel 2015], [News 2016-03] and [News 2016-08].
6 Referenced from an lndian government
website that is no longer available.
7 [Antman 2009], [ERGEG 2008], [SPEN
2014], [ENW 2015], [SSEPD 2015], [Nagi 2009], [Bastos 2009] and [Agha
2016], amongst others.
8 [Antman 2009], [ERGEG 2008], [SPEN
2014], [ENW 2015], [SSEPD 2015], [WPD 2015], [Monedero 2011] and
[Nagi 2009], amongst others.
9 Referenced in [Pasdar 2007].
Another UK utility [ENW 2015] mentions bypassing or otherwise
tampering with meters to avoid paying the high electricity costs associated
with the heat lamps used to produce cannabis as a concern for DSOs in the UK.
A
third UK utility [SSEPD 2015] identifies the case where there is no registered supplier at a connection
point as an additional source of NTL. Total losses on their distribution
systems are about 6% to 9%.
Western Power Distribution in the UK [WPD 2015] reports that about 6000
cases per year of "illegal
abstraction" (meter tampering, bypassing etc.) occur,
of which about 1000 relate to cannabis production.
The regulator in the UK [OFGEM
2015] estimates that total T&D losses in the UK were 7.2% in 2013, of which
most (about 73%) were technical losses on the distribution system. NTL on the
distribution system were about 4% of the total losses (or 0.27% of energy
supplied to the transmission system).
Size and types of losses in Central
and North America
References quoted in [Smith 2004]
state that electricity theft by illegal connections to the network are widely
prevalent in parts of Mexico, with
losses reported to be $475 million annually. This figure is confirmed by [News 2002]. Marijuana use also drives electricity theft in Canada
[News 2010-10] and the
USA [Depuru 2011].
A study in Arizona [Culwell 2001]10 found a probable meter tamper rate of 1% and
annual losses to the utility of $7,967,279, with the majority of losses
occurring from commercial customers. Electricity theft and theft of equipment
from the electricity network costs about $6 billion per year in the USA [News
2013-02].
Bureaucratic processes are reported as preventing people
from applying for legal connections in Mexico [News 2002].
Size and types of losses in South
America
The following is known about Brazil:
•
ln 2006, the Brazilian
electric sector lost 15.3% of its internal energy
supply; NTL alone may surpass 25% in some cases [Bastos 2009]11.
•
Another estimate is
that 7.3% of the energy supplied to
distribution systems is lost to NTL, at a cost of about $ 1.76 billion annually
[Barioni 2015].
•
The biggest
components of NTL obtained
from a case study of one area are (starting
with the highest) are fraud, errors in metering and
faulty equipment, illegal reconnection by ex-consumers.
Size and types of losses in Asia
China and Vietnam generally have
relatively low levels of losses [Antman 2009], but no details are available.
Urban cooperatives in the Philippines do not operate
as efficiently as they could,
but the private
utilities operate reasonably well [Antman 2009]. Malaysia's privatised electricity T&D system and
Thailand's public system both have total T&D losses of about 11% [Smith 2004].
10 Quoted by [Smith 2004].
11 From another document referenced in
[Bastos 2009].
Total losses exceed 30% in
Pakistan [Antman 2009]. Total losses exceed 20% in Bangladesh [Antman 2009].
[Transparency 1999]12 estimates a loss of 14%
due to NTL in Bangladesh in 1998-99. A utility in lndonesia estimated theft
losses to be 7% [Smith 2004].
Power utilities PEA of Thailand
and TNB in Malaysia consider electricity theft to be the major source of NTLs,
the majority of which involve meter tampering or vandalism or illegal
connections [Nagi 2009]. NTLs are estimated to be around 15% on the Malaysian
peninsular [Nagi 2009].
[Nagi 2009] also reports that:
•
Tampering is by
breaking the meter's housing seal for
tampering with the components inside. Once this is accomplished the spinning
disc is mechanically obstructed or the numbers that the meter readers use are
turned back (this method is obviously not applicable for digital display meters).
•
lllegal connections to LV power lines (220V single-phase systems) occur mainly
in shanty towns and
other residential districts and by small businesses and street vendors
•
Other types of NTL include inaccurate, inadequate or
faulty metering, inaccurate billing, other types of fraud such as collusion
with utility personnel, non-reporting of faulty
meters and inaccurate estimation of non-metered supplies.
Electricity usage is not charged for in
certain countries to certain pre-allocated groups of consumers,
e.g. the residence of the president or prime minister or
employees of the electricity utility
such as the 32,000 employees of the Electricity Generating Authority of Thailand
(EGAT) who receive free electricity
worth 1.5 billion baht (around $42.8 million) [Smith 2004].
ln 2004, TNB in Malaysia estimated a loss of $229 million
per year due to NTL, not only theft [Chauhan 2015].
The following is known about lndia:
•
Average total
transmission and distribution losses have been officially stated
as 23% of the electricity generated, but has been
estimated to be up to 58% in at least one case [Monedero 2011].
•
More specifically,
total transmission and distribution were approximately 15% up to 1966-67, but increased gradually to 28.36 % by
2011-12 [Navani 2012].
•
The two private
utilities supplying Mumbai have total losses of 11-12%, most of the state
utilities have losses of over 30% [Antman 2009].
• The effects
of restructuring have not always been positive
when it comes to losses [Monedero
2011]:
–
Before restructuring Orissa reported
23% total T&D losses, after restructuring they were 51%.
–
ln Andhra Pradesh these losses were about 25% before restructuring about
45% afterwards.
–
Haryana's losses went from 32% to 40%.
–
Rajasthan's losses increased from 26%
to 43%.
•
Types of NTL are
illegal connections; meter tampering/bypassing and errors in equipment;
unmetered supplies to agricultural pumps with larger loads than originally paid
for; single point connections to
small low-income domestic consumers; estimation of consumption where no
meters are installed [Monedero 2011].
•
Total distribution
and peak distribution loss collected from a field survey in a mixed urban and
rural area in Ghaziabad district were found
to be 27.45 % and 34.2 % respectively [Navani
2012].
The former Soviet Union has high total
losses and poor revenue collection rates, due to weak metering,
billing and payment collection in the 1980s;
this has not changed in Russia and some other former Soviet countries
[Antman 2009].
12 Quoted in [Smith 2004].
Size and types of losses in Africa and the Middle East
ln Sub-Saharan Africa only 50% of electricity is paid for due to, amongst
other reasons, a large amount of
consumed energy not being billed [Antman 2009].
Botswana's stated-owned utility is the best performing utility in the region, with
system losses of 10% [Antman 2009].
The following is known about NTL in South
Africa:
•
As much as 7% of electrical energy produced in South Africa is lost to theft
[News 2014-12].
•
Types of NTL include
meter bypassing or tampering and illegal
connections to the network [Beutel 2015]. The latter
have resulted in deaths, often children, due to the dangerous manner in which they
are carried out [News 2015-01], [News 2015-02], [News 2015-03].
•
The reliance on human
resources for reading meters means that when
there are staff shortages, electricity
cannot be billed.
As an example, meters could not be read in King William's Town for about 6
months at the time of writing,
resulting in a revenue loss of more than R2.4 million [News 2016-05].
•
52% of business
customers of Eskom, the country's largest utility, have been found to bypass
their meters [News 2016-03].
Senegal [Guymard 2013]:
• 21%
of the produced energy is lost without being sold.
•
The distribution between TL and NTL
(the latter mostly fraud) is unknown in
the distribution grid.
•
The spread of
distribution technical losses between the 30 kV, 6.6 kV and 400V grids is also
unknown.
• NTL
represent 40 billion FCFA (= $70 million), presumably per annum.
•
Meters are in short
supply, making it impossible to control new
connections and to replace defective
meters.
•
Own consumption is not known in some buildings since they are not metered.
Uganda loses almost $29.6 million annually due to
electricity theft [News 2016-08].
Types of NTL in Jordan are CT and VT accuracy errors, meter accuracy
errors, meter failure, billing errors (shifting period errors differences in
meter reading dates and human errors),
energy theft [Agha 2016]. Total
distribution losses in NEPCO (Jordan)
in 2014 were 13.79%, an increase
from 12.29% in 2011. Total T&D losses were
14.33% in 2014 [NEPCO 2014].
An lranian paper states that theft
is the largest component of NTL, e.g. illegal connections onto the network,
meter bypassing or tampering and evading payment. The former is the most common
form of NTL [Hossein 2015].
7.2.2.
MITIGATION
OF NTL
lntroduction
This appendix first lists
experiences on mitigation of NTL per region, then detail measures or
experiences not linked to particular regions or countries, with a focus on
central check/observer meters (a commonly proposed measure also
known as "energy balancing") and data mining/analysis.
Mitigation of NTL in the UK
Methods presently used to reduce NTLs in the UK include13:
•
Targeted inspections
and audits, including auditing of unmetered
supplies and updating records to ensure accuracy of the estimates.
•
Perform
investigations systematically e.g. investigate parties who applied for connections but didn't complete the process, or to ensure that energy theft isn't repeated where it was previously identified.
•
Responding to leads from stakeholders.
•
lmplementing a team of staff to identify parts of the network that produce the highest technical
or non-technical losses and implement mitigation measures.
• Actively
performing risk assessments and determine trends and hotspots.
• ldentify
tampering or bypassing when installing
smart meters.
•
Use
network (especially at secondary substations) and smart customer metering to identify
areas with high levels of NTLs and/or
individual customers with suspicious
consumption behaviour.
• lmprove
accuracy of records for unmetered supplies.
•
Normalizing or repairing tampered
installations and other equipment.
•
Addressing unrecorded
energy by updating information systems with more accurate consumption data.
•
Cooperating with enforcement agencies,
pursuing prosecution, social services and storing evidence.
• Liaison
with other stakeholders in the industry.
• Providing
appropriate training and awareness.
•
Updating records to reduce billing errors.
• External
engagement around best practice.
Revenue protection is seen as a key aspect of the strategy, e.g.
updating of records; stakeholder
engagement and knowledge
sharing to improve
processes and systems;
development and use of a theft risk assessment tool; engages and
assists with respect to prosecution of offenders; conducts internal and
external awareness programs [SPEN 2015].
lnaccuracies around estimation of
NTL are to be addressed via increased or improved measurements on and modelling
of the network, e.g. using smart meters and monitoring of secondary substations
[SPEN 2014].
There are limitations with the
present losses estimation method since it is not time-based, but smart meters
(which are expected to be installed by the end 2020) should increase the
accuracy to +/- 1.5% [OFGEM 2015].
The UK has a maximum sentence of 5 years for electricity
theft [UKRPA 2015].
Future industry drivers in the UK include more accurate measurement of losses (e.g. via smart meters),
hence stricter incentive schemes for reduction of losses like in some other countries [ENW 2015]. Secondary substation monitoring (for technical loss
reduction but could also be used for
detection of NTLs) and detection of the presence of cannabis
heat lamps are also being
investigated [OFGEM 2015].
13 Obtained from [SPEN 2014], [SPEN 2015], [ENW 2015], [OFGEM
2015], [SSEPD 2015].
One UK DSO is actively
investigating smart meters coupled with substation metering to detect areas
with high losses [SSEPD 2015].
Mitigation
of NTL in the remainder of Europe
To date there have been no formal
attempts to harmonize the treatment of network losses at a pan- European level.
The majority of European countries
broadly account for physical losses, thefts
and metering errors in their regulation. The objectives of the regulatory treatment
of losses are to protect the interest of customers and to promote the efficiency of the network
[ERGEG 2008]. lt is considered that
improvements in metering will improve the evaluation of losses in distribution
networks; it is therefore suggested that more metering points are implemented
and that losses are taken into account in the cost/benefit analysis of new metering equipment. However,
measuring non-technical losses is very expensive, often more expensive than the
value of the lost energy [ERGEG 2009].
Methods for analyzing losses
should be simple, transparent, predictable and reasonably cost reflective and
should allow for losses to be monitored so that
comparisons (improvements or otherwise) can be made over time. Some respondents
suggest that specific non-technical losses that can be estimated (like public
lighting) should be isolated and treated accordingly. TL are difficult to
reduce in the short term due to the long life of plant, but NTL may be more amenable
to reduction by increased attention to theft prevention and to the data acquisition procedures. NTL
can be reduced by incentives designed to reduce theft, improve metering
and reduce unmetered
supplies. However, reducing
total losses may be
a better option in practice
due to the difficulties in separately measuring
TL and NTL. The most effective
NTL reduction measures may well vary between member states as the conditions
vary between these countries [ERGEG 2009].
A Spanish study that involves
detecting anomalous drops in consumed energy (reportedly the most frequent sign
of a theft or tampering by a customer) by using windowed analysis and the
Pearson coefficient is reported in [Monedero 2011] and [Guerrero 2013]. Usually
mitigation is presently via physical inspections of customers chosen from consumption studies (labour intensive)
or customers with zero consumption over a certain period (this only catches
very obvious theft). The algorithms developed in this study use only consumption data from the
previous two years and have been tested with real (large) customers, and the
system has been put into operation. 38% of the
customers flagged via these methods
were found to have NTL, which is better
than the 10% obtained by routine inspections. An automated system was
developed for detection of NTLs on company databases; it uses all information available, not just
consumption data, e.g. data mining, statistical techniques, text mining, neural networks
and expert systems,
in order to classify a customer problem
as accurately as possible.
40.66% of customers flagged using the automated system had NTLs, at 90% less time spent than traditional methods. So far
this method has only been used for
large customers.
The Spanish DSO lberdrola (as a Prime Alliance member)
has implemented power line carrier
(PLC) technology between smart customer meters and secondary
substations which have supervisory meters on the secondary side of each MV/LV transformer. The customer and supervisory
hourly values of their meters are sent from the secondary substations to a central system with
regular communication. This allows balancing between customer energy usage and energy supplied by the secondary
substations to be performed, which in turn allows secondary substations with
high losses to be identified for inspection and determination of the cause of the losses.
These kinds of inspection directed
at areas with high losses have in most cases
uncovered high rates of fraud due to
illegal connections to the grid (buried underground or behind walls).
Refer to section
7.8 for further
information on energy balancing.
lberdrola is also working on an innovation project with
advanced supervisory meters on each line of the
secondary substation with Ariadna lnstruments S.L. These meters have higher supervisory capabilities than normal,
such as recording the load curve with per-second
resolution. This advanced supervision allows the identification of connectivity errors (differences between
the connection of a customer in the field and data base information), meter tampering detection
and power quality reports.
This kind of report compares the advanced supervision load per second
with the highest
load values of the customers. The
result of this advanced supervision shows a
100% success rate in identifying meter tampering and connectivity problems.
[Guerrero 2013] also reports that many
advanced techniques have been used by other researchers for detecting NTL,
examples are listed below (these methods are not necessarily used only in
Europe):
• Detection
rules where a series of data mining
tasks are compounded.
• A
system based on a non-supervised
artificial neural network.
• Use of Na1ve Bayesian or Decision Tree algorithms.
•
Support Vector
Machines which can be combined
with other computational intelligence such as fuzzy
inference systems or genetic algorithms.
• Deferential
evolution algorithms.
• Rough sets.
• Feature selection.
• Statistical analysis.
• Fuzzy
clustering.
• Euclidean distance.
• Various
methods of profiling customer loads using smart meter data or otherwise using
smart grids.
Most countries from the former Soviet Union who joined
the EU have successfully privatized their DSOs
– the implication is that their losses have reduced, but
this is not explicitly stated [Antman 2009].
The measures applied in the Romanian village [Harabagiu
2005] included:
•
Reduction of the average number of consumers per transformer point to
reduce the number of disconnected
consumers for unpaid bills or other reasons.
• Reducing
the length of LV feeders, i.e.
lengthening MV feeders.
•
Meters located in the distribution box
of the transformer point.
• Remote
meter reading system.
•
Replacement of transformers with lower power
ratings, each equipped
with automatic safety
switches for each consumer and a 3-phase circuit breaker with overload
and short-circuit protection.
Some utilities in Eastern Europe provide
incentives for staff for collection of revenue [Suriyamongkol 2002].
Mitigation of NTL in Malaysia [Nagi
2009]
ln Malaysia, TNB has adopted 3 main measures
to minimize and work towards preventing NTLs:
•
lnstallation of a Remote Meter Reading (RMR) service to
provide power consumption statistics
and online billed data for HV clients.
•
lnstallation of a prepayment metering system and physical protection of metering installations for high voltage
high risk customers (HRCs).
•
Setting up of a "Special Enforcement Against
Losses" (SEAL) team to investigate problems by conducting onsite customer
meter installation inspections on LV commercial domestic and light industry
customers and hence to reduce and minimize NTL problems faced by TNB. The SEAL
team's activities include improving metering and billing processes, ensuring
metering is accurate, and reducing the theft of electricity. The team was set up in 2004 and by 2005 distribution losses had
been reduced by about 1%.
Further information from TNB is as follows:
•
The electricity theft
rate for TNB in 2005 was reduced by almost 50%, where the theft rate for large customers (LPCs) was reduced from
3% to 1.5% and for ordinary customers (OPCs) the theft rate was reduced from
4.1% to 2%.
•
ln 2006 additional
engineers, technicians and meter readers were specifically trained to spot billing consumption irregularities.
New equipment was also procured and
transportation was provided, in
order to pursue suspected cases of power theft,
which were overlooked in the
previous years.
•
The SEAL team
aggressively carried out various activities including improvement of the customer billing process, conducting physical meter
inspections, testing and rectification of the
metering systems for LPCs and certain
numbers of the OPCs; these efforts
resulted in substantial amounts of back billing and collections.
•
Additionally, the
SEAL team installed secure meter boxes for HRCs and (ii) Expanded Metal
Protection Doors (EMPDs) for OPCs, in order to prevent from meter tampering.
TNB also expanded their "Enhanced Customer
lnformation Billing System" (e-ClBS) in 2006, in order to
identify HRCs for better security against power theft; the e-ClBS provides
the SEAL team with accurate analysis and consumption
reports, which can identify consumption patterns of repeated power theft.
ln 2007 and 2008, large inspection
campaigns were carried out by the SEAL team including meter checking and
premise inspection, reporting on irregularities and monitoring of unbilled
accounts, meter reading and sales. However, this had little success due to
newer and improved methods of electricity theft, which are difficult to
identify, and customer installation being inspections carried out without any
specific focus or direction most inspections are carried out at random, while
some targeted raids are undertaken based on information reported by the public
or meter readers.
The study presented by [Nagi 2009] proposes a method to overcome such limitations by monitoring and detecting deviations in customers' load profiles (i.e. fraud), as an alternative to complement the ongoing
existing actions enforced by power utilities
to reduce NTLs. The fundamental
approaches are:
•
An unsupervised
approach for determining outliers with no prior knowledge of the data using
unsupervised clustering.
•
Semi-supervised
modeling only normality, or in a few cases modeling abnormality, using semi-
supervised recognition or detection.
•
Supervised modeling of both normality and abnormality using
supervised classification with pre- labeled
data.
The three broad machine learning
approaches mentioned use the following major outlier detection methods:
statistical-based methods, distance-based methods, density-based methods,
clustering- based methods, deviation-based methods. The method was found to
result in an improvement in the detection of fraud, e.g. in 2005:
•
Without the use of a fraud detection system the average hit-rate was 3-5%.
•
With the use of the effective fraud detection system the hit-rate
increased 38%.
Comparison of models:
• Standard
Support Vector Classification (SVC): 32% hit-rate.
•
SVC and Fuzzy lnference System (FlS)
model: 40%.
•
ln some Malaysian
cities where the combined SVC and FlS
system was tested a hit-rate of up to 48% was achieved.
Mitigation of NTL in lndia
Some state-owned utilities that were privatized reduced losses
significantly through business efficiency improvements such as the introduction
of "state-of-the-art management
and information technology tools" [Antman 2009].
However, privatization is not necessary to reduce losses. The Andhra Pradesh State Electricity Board (APSEB) restructured into separate generation, transmission and distribution units, while keeping state ownership, resulted in transmission
and distribution losses being reduced from about 38% in 1999 to 26% in 2003 and less than 20% in 2008. A major portion
of this resulted from reduction of theft
(including illegal connections to the power
system and tampering with or bypassing meters, often with the assistance
of utility staff) by [Antman 2009]:
•
Acknowledging the problem;
• Estimating
its size via energy audits;
• Enacting
strict laws against electricity theft;
• Establishing
dedicated enforcement and prosecution mechanisms;
•
lmproving
efficiencies including advanced and tamper-proof meters, remote meter reading
and lT systems;
• Enclosing tranformers;
•
Normalizing 2.25 million illegal connections;
• A
comprehensive communication program and close monitoring of progress.
North Delhi Power Limited (NDPL) reduced total losses
of 53% when it was founded in 2002 as a public/private partnership to 15% in 2009, by implementation of advanced
metering infrastructure (AMl) for the customers responsible for
most of the revenue (considered the
most successful measure), use of MV
networks in theft-prone areas, replacement
of meters, energy audits, improved
enforcement, involvement of the
community, education around safety and appropriate regulatory incentives
[Antman 2009].
Awareness programs in lndia
resulted in drop in total transmission and distribution losses from 38.86% in
2001-2002 to 34.54% in 2005-2006 [Depuru 2011]14.
Other methods used to reduce NTLs include [Navani 2012]:
•
Upgrading of electricity meters to
meet standard accuracy (including statistical
analysis);
• Smart
card technology (to minimize the theft of energy);
• lntegrated
billing system and prepaid energy meters;
•
Technical training to
the operating personnel (and enhance employees loyalty to eliminate pilferage
in the distribution system);
•
Statistical monitoring of energy consumption per sector, class and geographical setup (and statistical evaluation of meter readings).
Mitigation of NTL in Latin America
ln Latin America losses were often
larger than 30% in the 1980s. Utilities that privatized in the 1980s and 1990s
produced substantial reductions in TL and NTL through improved efficiency.
Measures included electrification of previously unelectrified areas, but this
aspect was not always successful (in Chile, for example, it was successful).
Chile also assisted utilities in other countries in the region to successfully
privatize. The utility Chilectra Metropolitana has losses of about 5%, reduced
from 22% [Antman 2009].
Enersis (various utilities owned
by the same company) achieved a significant reduction in losses by
normalization of illegal and unmetered customers, improved efficiencies,
stepped-up maintenance, customer communication around payment and safety,
tamper-proofing meters, training and monitoring of contractors, enforcement and
prosecution. Over half of the non-technical losses addressed resulted in
reduced demand. There was also effective but fair regulation [Antman 2009].
ln the 1990s El Salvador,
Guatemala, Panama and Nicaragua created a completely new regulatory framework
for their power industries, which included unbundling, open transmission access
and privatization of the distribution business. The experience of the El
Salvadorian utility DELSUR in improving customer care and response and
implementing advanced lT systems resulted in the total losses more than halving
in 5 years (2002-2007) [Antman 2009].
Brazil privatized almost half of
their DSOs in the second half of the 1990s, almost all of them took advantage
of the regulated loss targets. The Brazilian utility CEMlG is an example of a
state-owned utility that has performed similarly well. ln Buenos Aires 655,000
low-income users were successfully
14 Referenced from the same lndian government website that is
no longer available.
normalized by the two local
private utilities and appropriately incentivized by the government. AMl has
been very effective in detecting
and discouraging bypassing, tampering, collusion
and other unmetered consumption in developing
countries such as the Dominican Republic, Honduras and Brazil [Antman 2009].
MV distribution with shielded
connections and split meters is used in parts of Rio de Janeiro
state where access is difficult. Each customer must have a dedicated low capacity MV-LV transformer (ideally)
or at least a direct
connection to the transformer. This together with having the meter in a sealed
container at the transformer makes undetected illegal connections very
difficult. lnstalling meters at transformers supplying more than one customer
is recommended for fast detection of theft
[Antman 2009].
A
small number of LPCs, often supplied at medium
and high voltage and usually
representing less than 1% of
the total number of customers,
represent more than 30% of a DSO's
revenue. These customers should be fully billed, monitored and, if necessary,
normalized on a continuous basis. Many utilities in Latin America have achieved
this by sound business efficiency
and use of appropriate lT systems. A
dedicated customer service
department for dealing
with LPCs, including speedy resolution of problems, is critical as it reduced the incentive to bypass or tamper. ln some parts of Brazil the utility is not allowed access to its metering
equipment on the customer's premises, a second meter has been installed outside the premises [Antman 2009].
[Antman 2009] further states that
"comprehensive experience in almost all Latin American reforming countries
show that consumer discipline is achieved in quite a short time if those
irregular consumers become aware that the utility is able to make fast detection
and take corrective action on fraud and theft".
ELEVAL (a utility in Valencia,
Venezuela) has developed a tool for detecting and locating concealed illegal
connections to underground cables [Parra 2006], once their presence has been
established via use of a check meter. The amount of NTL established in this way
is then used to prioritize networks for inspection. The tool involves locating
unauthorized underground connections via high frequency injection, reducing the
amount of digging required time by more than 50% and cost by more than 80%.
Split meters with the meters in boxes with tamper detection
are also used in Brazil [Ribeiro
2012]. Utilities in Brazil may
by law retrospectively charge customers for stolen energy up to 6 months by
estimating the energy that was used
[Barioni 2015].
Bayesian networks are explored in [Bastos
2009] using real Brazilian data.
A
method based on the application of unbalanced weighted least squares
state estimation and anomaly
detection is used to estimate and identify TL and NTL is suggested
by [Rossoni 2015].
The method was successfully implemented in a
laboratory environment where the
behaviour of typical Brazilian
consumers was modelled. The method was also applied on an actual
Brazilian distribution network;
with promising results.
Mitigation of NTL in Africa
Split meters are being rolled out
in parts of South Africa [News 2016-07]. This makes it more difficult to
bypass, or otherwise tamper with, the meter, since it is no longer located
inside the consumer's premises.
According to radio reports
on the 9th and 10th of May 2016, Johannesburg's municipal electricity provider,
City Power, has appointed contractors to check every electricity meter for
illegal connections. A new dedicated
unit will also be established within the organisation to deal with management of meters.
Other initiatives deployed
successfully in South Africa are removal of illegal connections, conducting
meter audits and correcting or replacing faulty or vandalised meters [Eskom
2016].
Another (extreme) tactic is to
deal with illegal connections is to disconnect power in an entire affected area
[News 2015-03] (for safety reasons in this case), but this can result in anger
from those disconnected.
Methods to reduce NTL in Senegal
are recommended to include industrial meter safety checks and meter roll-out
outside of residential premises, measuring device roll-out on
each LV feeder, comparison between the network
energy flow measurements and the sum of clients
billing statements of the supplied
zone would easily flag zones where fraud
is most widespread, targeting controls on these zones for a better efficiency.
The installation and controlled reading of meters
is an important part of this [Guymard
2013]. Whether any of these have been
implemented and, if so, the success rate is not known.
A ClRED paper from Egypt [Emara
2013] reports on an approach based on a genetic algorithm with multi-objective
optimization analysis, management of electricity
distribution network planning includes, in addition to load flow and voltage
profile analyses, power delivery quality assessments, operation reliability analyses and, above all, economic evaluation of all investments. The method does not appear
to have been implemented at the
time of publishing of this paper.
Comments
not related to specific countries or regions
General
comments
When users who previously did not pay for electricity then have to pay, their demand
usually decreases, resulting
in a decrease also in technical losses [Antman
2009].
NTL also result in increased TL due to the
additional current that is drawn [Suriyamongkol 2002].
Privatized distribution utilities
show 11% lower distribution losses, 32% more electricity sold per worker and 45% greater bill collection rate than state-owned utilities (achieved within
five or more years of being privatized). However, the regulatory framework must be present to support privatized utilities, especially
setting realistic loss targets where utilities
keep the profit is they beat targets but suffer a loss if they don't (this
worked well in several Latin American countries) [Antman 2009].
A disproportionately high portion of losses tend to emanate from users that
require the largest amount of electricity, so tackling
these first has been successful in emerging economies; these losses have been
known to include collusion by
utility employees. "Naming and shaming" such large customers, usually
well-known companies, has also been successful, as has successfully and visibly prosecuting politically- connected energy thieves [Antman 2009].
Since most criminals are not aware of
the fraud detection methods that have been successful in the past, they will adopt
strategies which will more likely
lead to identifiable frauds; therefore to detect fraud earlier detection tools need to be
applied as well as the latest developments [Nagi 2009].
Factors that influence electricity
theft include belief that stealing from utilities is not wrong or illegal,
difficult economic times, lack of education, lack of law enforcement,
corruption of utility employees (by deliberately reading lower consumption values
for example), certain areas not electrified and high (or believed to be high)
energy prices [Depuru 2011].
A
mitigation measure proposed
by [Depuru 2011]
is to charge a lower tariff for low income customers to make theft less attractive. Other
methods cited by the same reference from various sources are smart meters
(which are stated as being difficult to tamper with), power line impedance measurements, use of shunts detecting equipment to detect unauthorized
connections and use of check meters.
Mitigation methods listed by
[Smith 2004] include technical/engineering methods, e.g. tamer-proof meters,
managerial methods, e.g. regular inspection and/or monitoring focussing on
users with the
highest usage (cf. poor areas that use
comparatively little power), tackling of
corrupt employees and system change, e.g. public v. private utilities, regulation, who pays for losses
etc. Privatisation does not automatically or necessarily result in
a more efficient business, and a multi-method approach is recommended.
Software has been developed
that has shown good results in producing optimised strategies for dealing with electricity theft by
deployment of AMR and manual
inspection [Ribeiro 2012].
Another type of NTL is the willful hacking and
modification of smart meter usage data via cyber-attack, which is becoming more and more of a risk due to the data-intensive nature of smart grids [Leite 2016]. Cyber-attacks may be targeted
at any part of the system,
e.g. at the meter itself
or at the communication
system or at the data values themselves [Jokar 2016]. Privacy of customer information should be
considered in any method that uses customer
usage data [Jokar 2016]. An advanced
method for detecting and locating such attacks is also given in [Leite 2016],
but the method has not been implemented in the field as yet.
Comments
related to specific technologies
Prepaid AMl also has advantages of much lower hardware costs and the ability to permanently monitor consumption over card-based
prepaid systems [Antman 2009].
Time domain reflectometry (TDR)
may be used to localize sources of NTL [Trupinic 2005] the premise is that the
reflected pulse shape is different if an illegal or tampered connection is
connected to if only meters are connected to a feeder. This method can be used
for very precise local research once suspicion is established. The method has
been tested in a limited way, but does not appear to have been implemented at
the time of publishing of this paper.
TDR is also proposed for use in detecting NTL by [Hossein 2015], to
complement analysis of data. The
method works by sending a pulse down a cable to detect any changes
in cable impedance. Events such as change
of cable type, broken cable or fault will result in a change in impedance can
in principle be detected, but tests with a commercially available
TDR instrument showed that the results
require careful examination.
[Depuru 2011] proposes a smart
system where illegal customers
(those with >5% NTL) are identified, paying customers are then disconnected,
the system is then subjected to a voltage high in harmonics that destroys
appliances connected to the system and finally paying customers
are re-connected again. This system has several practical hurdles, including protecting system
components from damage due to the
harmonics, public resistance, cost of roll-out,
and the efficiency of detection of legal and illegal customers. Also, the
legal framework may not be in place in all countries.
Minimising the length of the
LV network is proposed for Turkey by [Yorukoglu 2016] (the original idea is from one of the references in that document). lt is also proposed that
meters are installed at the top of poles
(out of reach). Only supplying power to agricultural areas during certain
times of the day is also proposed (the idea for this is from
one of the references citing the
practice in lndia). This would require separation of agricultural networks from
other networks in the areas under consideration.
Using armoured cables makes connecting
illegally onto them more difficult [Ribeiro 2012].
A proposed method of reducing illegal connections to the network is to raise the voltage to a
level that would damage normal
electrical equipment (say 350 V on a 230 V system)
and step it down to its normal
level at each consumer [Babu 2013]. This would also reduce technical losses on
the LV network, but the method has
not, to the author's knowledge, been trialed in the field.
Placing all meters in an area
(presumably a relatively small area) in the same enclosure on the street makes
detecting of bypassing by inspection much easier [Bandim 2003].
The use of unmanned aerial
vehicles is suggested for assistance with detecting electricity theft once
smart grid technologies have been used to determine the general location of
such theft [Rengaraju 2014].
Central
check/observer/supervisory meters
This is a method
that has been suggested by several sources
as a way of detecting NTL, and is therefore
covered in a separate section here. lt is also
known as "energy balancing" and involves some form of checking that the energy that is
recorded as being consumed by the customers on a network is the same as the
total energy that is supplied to that network, i.e. checking that there is are
not unmetered or undetected loads connected to the network.
The following details are amongst those
available:
•
The Spanish DSO
lberdrola (as Prime Alliance member) has implemented power line carrier (PLC) technology between smart customer
meters and secondary substations which have supervisory meters on the secondary
side of each MV/LV transformer. lberdrola is also working on an innovation project with advanced
supervisory meters on each line of the secondary substation. These meters have
higher supervisory capabilities than normal, such as recording the load curve
with per-second resolution. Both projects have resulted in significant
successes, further details may be found in Section 7.3. Mobile check meters,
left in place for at least a month,
are proposed as a way of doing this with lower costs real
customer data was used to verify
this [Doorduin 2004].
•
This method can be used to determine any type of NTL, not just theft [Bandim 2003]. This ref uses mathematical methods to identify
customers with problems and/or a high probability of bypassing, focusing inspections to only those
premises. This has not, to the author's
knowledge, been trialed
or otherwise implemented in the field.
•
An extended version of this is proposed whereby the current is
measured at every pole on an overhead LV feeder, and the values compared
[Chauhan 2015]. This has not, to the author's knowledge, been trialed or
otherwise implemented in the field.
•
Another variation of the method where customer consumption is compared to a central observer
meter is covered in [Jokar 2016]. The
method was tested on real customer data.
•
The theft detection
rate at ENEL (ltaly) increased from
5% to 50% after check meters were installed
at 5% of supply
points of delivery to compare with the power usage
of the customer meters [Lu 2013].
•
A method for
detecting NTL whereby the current is
measured at various points on the network is proposed by [Beutel 2015], but has
as yet been implemented. Smart customer meters are included, as well as a meter at the secondary
substation. This is similar to the proposals
of [Bandim 2003] and [Santos 2015].
•
A method for
determining which customer is connected illegally, once the presence of an illegal connection has been established, using power line carrier
is proposed in [Pasdar 2007].
The method has not been
implemented, to the best of the author's knowledge.
•
A variation without
smart meters is proposed in [llo 2003], but does not appear to have been
implemented at the time of publishing
of this paper.
•
The principle is also stated in
[Navani 2012] and [Emara 2013].
•
[Berrisford 2013]
found a value of C$732 million in
rolling out smart meters at BC Hydro for revenue protection alone. The main part of
this strategy is energy balancing. The method is described and includes correlation between and
non-linear optimisation of the
customer voltages (and energy usage). The method is also able to detect high
resistance joints on the service connection, as well as network data or
topology errors. lnitial results are promising.
The use of central check or observer
meters would likely need to be part of a wider "smart DSO" system,
such as that discussed in Europe
[EDSO 2016], and would allow more frequent and localized checking of energy
balance. This would allow for faster detection and more accurate location of NTL than in a "traditional DSO" and with less manual intervention. A "smart DSO"
would also involve
greater business efficiency
generally, theoretically resulting in reduced risk of metering or billing errors.
Data mining/analysis
According to [Gemignani 2009] it
is possible to statistically pre-identify fraudsters through load and demand factor comparison. The concept revolves
around having a typical load/demand factor profile for each profile of consumer. Assuming this profile is regularly updated, it
becomes possible to identify the behaviour of some meters
against these profiles. The calculations of the load and demand
factors have been carried
through for the residential, commercial and
industrial classes, but the
method does not appear to have been implemented at the time of publishing of this paper.
Detection of physical tampering already exists in smart meters, e.g. sensors
for change in inclination (bumping) and removal
of the cover, but can give false positives. An advanced data-analysis method of detecting energy theft, while minimizing
false positives, is presented. lt looks promising but has not yet been rolled
out [McLaughlin 2013].
Mathematical methods for detecting
NTL (specifically theft) can be state-based, or classification-based or use
game theory several references of such methods are given in [Jokar 2016].
Some statistical/data techniques for detecting
theft are referenced in [Chauhan 2013].
A state estimation method using smart meter
data is proposed for detection of NTL [Lu 2013].
An example of a data-analysis method using consumption profiles is given in [Angelos
2011], the method was validated on real data but has not as yet been implemented.
A method for speeding up consumption data
analysis is given in [Depuru 2013]. Another method for using analysis of
consumption data is given in [Mashima 2012].
