CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

POWER QUALITY FACTORS IN EFFICIENCY BENCHMARKING

Lassila J 1)., Honkapuro S., Viljainen S., Tahvanainen K., Partanen J. Lappeenranta University of Technology
Kivikko K., Antila S., Mäkinen A., Järventausta P. Tampere University of Technology.
Finland, 1) Jukka.Lassila@lut.fi

Abstract - In this paper methods to take power quality factors
(e.g. number and duration of planned and unplanned
interruptions, number of auto-reclosing operations) account
into efficiency benchmarking as outage costs is presented.
The effects of outage costs in efficiency benchmarking are
evaluated by means of sensitivity analyses. Effects of power
quality factors in different kind of regulation models in actual
network planning task are estimated.
INTRODUCTION
The importance of power quality (PQ) – including both
voltage quality and interruptions – is growing in regulation
process and distribution business. In many countries, e.g. in
Norway [1] and in Finland [2], PQ is already part of
efficiency benchmarking. Although PQ has been monitored
auto-reclosing operations and voltage levels can be included
to efficiency benchmarking are active. Some of this data is
already collected from network companies annually.
Regulator can calculate outage costs for each company by
using this data. This is possible when unit prices of each
power quality factors are defined. In Finland e.g. price of
unplanned (unexpected) interruption for residential customer
is 0,068 €/kW and 0,61 €/kWh [3]. Unit prices are defined for
five customer types and four interruption types. PQ factor
can be formed as a sum of separated PQ cost components. In
efficiency benchmarking it is possible to add outage costs to
company’s operational costs as presented in Fig. 1.
Efficiency benchmarking (DEA)
Interruption statistics
u1 ⋅ Energy + u2 ⋅ Network + u3 ⋅ Customers + c
Service customer Max h 0 =
Public customer v1 ⋅ (Operational costs + Outage costs)
Industry customer
Unit costs
Agriculture customer
Service customer

+
Residential customer
rather detailed (e.g. number and duration of planned and - planned interruptions:
- unplanned interruptions:
1,5 h
2,5 h
Public customer
Industry customer Outage costs
- high speed auto-reclosing: 45
unplanned interruptions, number of auto-reclosing operations, - delayed auto-reclosing: 13
Agriculture customer
Residential customer

voltage levels etc.) many years, PQ is taken into account quite - planned interruptions:
- unplanned interruptions:
- high speed auto-reclosing:
0,034 €/kW and 0,30 €/kWh
0,068 €/kW and 0,61 €/kWh
0,068 €/kW

simplified way in efficiency benchmarking. Usually only - delayed auto-reclosing: 0,088 €/kW

duration of interruptions is notified in efficiency
benchmarking. The main reason is that quality of data of PQ
statistics is not good enough to use them in benchmarking.
Other reason is that use of more than one PQ factor in
benchmarking requires sophisticated efficiency benchmarking
methods.
In Data Envelopment Analysis (DEA) it is possible to use
factors, which are not commensurate with each other’s. This
means that the factors with various units (time, number,
currency etc.) can be used in benchmarking at the same time.
DEA weights benchmarking factors so that efficiency score is
as high as possible for every company. E.g. if company is best
as a sense of one factor (e.g. PQ), DEA weights this property
of company more than the factors which are not beneficial for
the company in efficiency benchmarking. In practice this may
cause problem in sense of insignificant factors. If PQ is
included in DEA as four separated factors (e.g. number and
duration of planned and unplanned interruptions), DEA may
weight one or more of these factors so that the factor does not
actually affect to the efficiency score at all. In this situation
company may improve or worse these properties without
affecting its efficiency score.
One way to avoid this problem of insignificant factors is to
combine PQ factors to one outage cost and add this cost to
operational costs. The basic idea of outage cost is to
determine unit costs for different kinds of outages for
different kinds of customer types. In Finland interruption time
of customers is one and only PQ factor in efficiency
benchmarking at the moment. Discussions that how the other
power quality components such as number of interruptions,
Fig. 1. Outage cost method and DEA-model.
BACKGROUND OF POWER QUALITY
REGULATION IN FINLAND
The Finnish Energy Market Authority presented efficiency
benchmarking in 1999. Efficiency scores were calculated by
using Data Envelopment Analysis (DEA) with five factors
[4]. Power quality was one of those factors. It was measured
as a total interruption time of customers. The reason why only
interruption time was included in to efficiency benchmarking
was insufficient PQ statistics. There was no reliable statistics
of other kind of interruption types like number of short
interruptions (e.g. auto-reclosing operations). The basic form
of efficiency benchmarking with DEA is presented in (1).
u1 ⋅ Energy + u2 ⋅ Network + u3 ⋅ Customers − v2 ⋅ Interrupti on time + c
Max h 0 =
v1 ⋅ Operationa l costs (1)
In (1) PQ is measured as a total interruption time of
customers. Theory and features of DEA is presented e.g. in
[5]. The problems considering insignificancy of certain
factors when DEA was used as a regulation tool was
presented in [2]. Actual outage cost for company
(€/customer,h) could be calculated because of link between
efficiency benchmarking and allowed rate of return [6] in
distribution business as presented in (2).
∆DEA (2)
Price Outage = ⋅ Operational cost
∆Interruption time
Where

CIRED2005

Session No 2
CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

Price Outage = marginal cost of interruption been collected earlier and they are on the list for PQ trend
∆ DEA = change in efficiency score monitoring reasons.
∆Interruption time = change in interruption time
OUTAGE COST MODELLING
As separated factor, interruption time was insignificant factor
for many companies and actual cost of outages varied from 0 The target of regulator is to define one outage cost for every
to 500 €/customer,h from company to another. Problem is company. This outage cost is combination of costs caused by
solved in [7] where efficiency benchmarking with outage factors 1-6 presented previously. Outage cost modelling
costs was presented. Basic idea was handle power quality as requires that unit costs for power quality factors are available.
outage costs and adds outage costs to operational costs as Power quality unit costs in Finland are presented in Table I.
presented in (3).
TABLE I. UNIT COSTS FOR POWER QUALITY FACTORS FOR CUSTOMER
GROUPS IN FINLAND. (AR = AUTO-RECLOSING) [3]
u1 ⋅ Energy + u2 ⋅ Network + u3 ⋅ Customers + c (3) Unplanned Planned High Delayed
Max h 0 =
v1 ⋅ ( Operationa l costs + Interrupti on costs) interruption interruption speed AR AR
[€/kW][€/kWh] [€/kW] [€/kWh] [€/kW] [€/kW]
In Fig. 2 the numbers of companies that have insignificant Residential 0,068 0,61 0,034 0,30 0,034 0,088
factors are presented before and after development work. It Agriculture 0,54 4,9 0,18 1,6 0,25 0,70
Industry 2,6 8,7 0,80 3,8 1,1 2,9
can be seen that power quality is insignificant factor for over Public 0,65 3,4 0,23 1,5 0,23 0,73
20 companies if power quality is measured as separated factor Service 1,9 11 0,80 7,2 0,95 2,1
(present DEA-model) compared to situation, where power
quality is measured as outage costs and it is added to
Unit costs are based mainly on a Nordic research work done
operation cost (developed DEA-model).
by 90’s. More detailed definition of these unit costs is
50
presented in [9] and [3]. There is going a new research work
to update these unit costs. This is done co-operation with
Distribution companies

40
Finnish Energy Market Authority, technical universities and
30 Present DEA-
distribution companies.
model
Developed DEA-
20 model
To reach for as accurate interruption data as possible in
10
economically, companies are obligated to collect interruptions
on secondary (distribution) substation-specific. In Fig. 4
0
Operational costs Power quality * Distributed energy Network length Customers example of distribution substation-specific outage cost
Fig. 2. The number of companies that have insignificant factors. [7] modelling is presented.
DS 1 DS 4
During first three-year long regulation period (2005-2007) 170 MWh 102 MWh
6 customers
12 customers
power quality has no specifically defined role in Finnish
DS 5
110/20 kV
distribution regulation. During this time, regulator collects PQ 65 MWh
7 customers
reference data from distribution companies. This data will be
DS 6
used as a reference during second regulation period. DS 2
437 MWh DS 3
332 MWh
45 customers
Regulator has decided to collect following eight power 23 customers 215 MWh
15 customers
quality factors now on [3], [8]. DS = distribution substation (20/0,4 kV)

1. Customer’s average annual interruption time that is Fig. 4. Medium voltage line and example distribution substation (DS 4).
caused by unexpected interruptions.
2. Customer’s average annual number of interruptions Distribution substation-specific outage cost modelling
that are caused by unexpected interruptions. requires information of customer types and energy
3. Customer’s average annual interruption time that is consumptions. In table II this kind of statistics for distribution
caused by planned interruptions. substation 4 (DS 4) is presented.
4. Customer’s average annual number of interruptions
TABLE II. STATISTICS OF DISTRIBUTION SUBSTATION 4 (DS 4).
that are caused by planned interruptions.
Customer Number of Energy consumption Average power
5. Customer’s average annual number of interruptions that group customers [MWh/a] [kW/customer]
are caused by delayed auto-reclosings. Residential 3 42 1,6
6. Customer’s average annual number of interruptions that Agriculture 2 50 2,9
are caused by high speed auto-reclosings. Public 1 10 1,1
7. Annual number of unexpected interruptions in low
voltage network. In table III interruption statistics for DS 4 is presented.
8. Annual number of unexpected interruptions in
medium voltage network.

PQ factors 1-6 are weighted with annual energy of the

secondary (distribution) substation. PQ factors 7 and 8 have

CIRED2005

Session No 2
CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

TABLE III. INTERRUPTION STATISTICS OF DS 4. POWER QUALITY IN REGULATION AND

Interruption type Number of interruptions Total interruption time EFFICIENCY BENCHMARKING
per year per year
Unplanned 6 3,5 h
Planned 1 1,0 h Next properties of different kind of regulation and efficiency
High speed AR 12 - benchmarking methods in sense of power quality and
Delayed AR 4 - investments are described.

It can be seen that number of unplanned interruptions is 6 and Allowed rate of return and DEA efficiency benchmarking
they have last 3,5 h together. Outage cost e.g. for duration of with interruption time
unplanned interruptions can be calculated as presented in (4).
This is model, which has been used in Finland few years.
Allowed rate of return is based on present value of the
C Unplan-time = cUnplan-time ⋅ ncustomer ⋅ Pcustomer ⋅ t Unplanned (4)
distribution network. The model encourages investing
because every investment increases value of the network and
Where by that way increases rate of return of company. Financing
cUnplan-time = unit cost for duration of unplanned interruption [€/kWh] costs of investment can be gathered from customers. In
ncustomer = number of customers in customer group efficiency benchmarking (1) power quality affects to
Pcustomer = average power of customer in group [kW/customer] efficiency score varying in companies as presented in Fig. 3.
tUnplanned = duration of unplanned interruptions in DS 4 [h]
Allowed rate of return and DEA efficiency benchmarking with
From table I, II and III outage costs for residential customers outage costs
caused by unplanned interruptions (Cunplan-time-res) are (5)
This is developed model [7] where power quality is measured
as outage costs and they are added to operational costs in
C Unplan-time-res = 0,61 €
kWh ⋅ 3 ⋅ 1,6 kWh ⋅ 3,5 h
a = 10 €
a
(5)
DEA-model (3). The model still encourages investing to
network because every investment increases value of the
In table IV outage cost for all customer types and interruption network. Even investments are not part of efficiency
types is calculated for distribution substation 4. benchmarking, efficiency score can be affected by investing
to targets, which improve power quality (outage costs) and by
TABLE IV. TOTAL INTERRUPTION COSTS FOR DS 4. (N = NUMBER, T =
that way improve efficiency score. If investment is beneficial
DURATION, HS AR = HIGH SPEED AUTO-RECLOSING,)
Unplanned Planned HS Delayed Total
as sense of better power quality, efficiency score grow up and
interruption interruption AR AR costs this brings additional benefit for company. In Fig. 7
N [€] T N [€] T [€] [€] [€] percentage values of outage costs for all companies are
Residential 2,0 10 0,2 1,4 2,0 1,7 17 presented.
Agriculture 18 98 1,0 9,1 17 16 159 100 %
Public 4,3 13 0,3 1,7 3,0 3,2 26 OUTAGE High speed
90 % auto-reclosing
Total costs 24 120 1,4 12 22 21 100 % COST 24,2 %
80 %

70 %

The bottom row shows total costs of each interruption type 60 % Interruption time
(unplanned)
and rightmost column shows interruption costs of each 50 %
46,0 %
40 % OPEX
customer type. Total interruption costs for this particular 30 %
Delayed
auto-reclosing
14,2 %
distribution substation are 200 €/a. 20 %
Interruption number
10 % Interruption time
(planned)
Interruption number
0,5 % (planned)
In Fig. 6 percentage values of outage cost components in 0%
OPEX + Outage cost
(unplanned)
10,8 %
4,3 %

every Finnish distribution companies are presented. At the a) b)

moment reliable and extensive enough distribution substation- Fig. 7. Average percentage values of a) total cost (outage cost and OPEX)
specific interruption statistics are not available. Outage costs and b) outage costs in Finland.
are based on average values of companies and nationwide
outage costs with common shares of customer groups. This is It can be seen from Fig. 7 a. that outage costs are on the
presented more detailed in [10]. average 17 % of total costs (outage costs and OPEX). Half of
100 %
outage costs are caused by interruption time as it can be seen
90 % High speed
80 %
auto-reclosing from Fig. 7 b.
Delayed
70 %
auto-reclosing
60 % Interruption number
Allowed rate of return and DEA efficiency benchmarking with
50 %
(planned) outage costs and investments
Interruption number
40 %
(unplanned) This is developed model [11] where investments are taken
30 %
Interruption time
(planned)
account in efficiency benchmarking and they have as
20 %

10 %
Interruption time importance role as outage costs and operational costs in
(unplanned)
0%
benchmarking. This model directs investments to those
Fig. 6. Percentage values of outage cost components in Finland. targets where minimum of total costs (operational costs,

CIRED2005

Session No 2
CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

outage costs and investments) can be achieved in the long Fault frequency : 0,05 faults/km,a
Operating time (manual disconnecor) : 1 h
run. The investments, which does not affect to operational Remote-controlled disconnector ? 3 km, 40 kW Operating time (remote-controlled) : 0,2 h
..and 0,6 h if reserve line is needed
kW
, 60
costs or power quality, decrease efficiency score of the 3 km
Repairing time : 3 h
Remote-controlled disconnector : 20 000 Eur
B
company. In Fig. 8 percentage values of outage costs, kW
(lifetime 25 years --> 1420 Eur/a, p = 5 %)
00
investments and operational costs in Finnish distribution PS
4 km, 100 kW 2
km
,1
6 km,
150 kW 3 km,
100 kW
Reser
companies are presented. ve line
A C
100 %

90 % Fig. 10. Medium voltage feeder and potential place for remote controlled
80 % disconnector. (PS = Primary substation 110/20 kV)
70 %

60 % Interruption c osts Investment of remote-controlled disconnector is 20 000 € and

50 %
Investments annual cost of investment is 1 420 € (p is 5 %). This
Operational c osts investment affects mostly on interruption time. Unit price for
40 %

30 %
unplanned interruption time is 4,3 €/kWh. This figure based
20 %
on unit costs (table I) and nationwide energy consumption
10 %
weighting in customer groups [10]. In table V changes in
0%
interruption times and savings in outage costs in different
Distribution companies (total number is 94) areas after investment is presented.
Fig. 8. The relative parts of operational costs, investments and interruption
costs in Finnish distribution companies. TABLE V. SAVINGS IN INTERRUPTION TIMES AND OUTAGE COSTS AFTER
REMOTE-CONTROLLED DISCOONECTOR INVESTMENT.

Revenue regulation and power quality Fault, Saving [h] Saving [€/a]
effect before–after
One example of network revenue regulation model with A B 1 – 0,6 = 0,4 (0,05 . 4 km) . 200 kW . 0,4 h . 4,3 €/kWh = 69 €/a
outage costs is presented in [1]. In this Norwegian model C 1 – 0,6 = 0,4 (0,05 . 4 km) . 250 kW . 0,4 h . 4,3 €/kWh = 86 €/a
B A 1 – 0,2 = 0,8 (0,05 . 8 km) . 100 kW . 0,8 h . 4,3 €/kWh = 138 €/a
outage costs are part of network revenue evaluating process.
C 1 – 0,2 = 0,8 (0,05 . 8 km) . 250 kW . 0,8 h . 4,3 €/kWh = 344 €/a
If outage costs are more that expected outage costs for certain C A 1 – 0,2 = 0,8 (0,05 . 9 km) . 100 kW . 0,8 h . 4,3 €/kWh = 155 €/a
time period, company are forced to diminish its revenue by B 1 – 0,2 = 0,8 (0,05 . 9 km) . 200 kW . 0,8 h . 4,3 €/kWh = 310 €/a
corresponding sum. If company can improve its power quality Total saving: 1 100 €/a
so that outage costs are less than expected outage costs,
company may increase its profit as presented in Fig. 9. It can be seen that outage costs decrease 1 100 € per year if
investment is done. Annual cost of investment is 1 420 €, so
Expected outage Outage costs
costs
Outage costs Outage costs investment cost are more than savings in outage cost. This
REVENUE
kind of investment affects also to operational costs but in this
Operational Additional profit if example, focus is on outage costs. Next it is presented, in
costs company can decrease
its outage costs which way decreased outage costs and disconnector
Payment if company
increases its outage costs
investment affects to company’s allowed rate of return and
Depreciations efficiency benchmarking results.

Return on
capital
Fig. 9. Cost structure of network revenue regulation with outage costs.
In Norway power quality is also taken account in efficiency
benchmarking done by DEA-model. Demand for improving
power quality is delivered for a time period.
EFFECTS OF REGULATION TO THE POWER
QUALITY INVESTMENTS
Profitability of power quality investment depends greatly on
chosen regulation model. In next example effects of
regulation model and efficiency benchmarking method to
power quality investment is presented. In Fig. 10 typical
medium voltage feeder is presented. In this case it is studied,
is it profitable to invest to remote-controlled disconnector as a
sense of smaller outage costs.
Allowed rate of return and DEA efficiency benchmarking with
interruption time
Investment is profitable in all cases because it increases
present technical value of the network and by that way
allowed rate of return. Because only interruption time is
noticed in efficiency benchmarking, benefit of investment
depends on savings in total interruption time and significance
of PQ factor (interruption time) of company in DEA-model.
If PQ factor is insignificant in DEA for the company (price of
interruption is 0 €/customer,h) investment does not bring
additional profit for the company. If price is >0, e.g. 50
€/customer,h, and annual change in interruption time is e.g.
100 customer,h/a, additional profit from DEA-model is 50 x
100 €/a = 5 000 €/a. The profit from DEA-model can be
remarkable high compared to investment cost and savings in
outage costs.
Allowed rate of return and DEA efficiency benchmarking with
outage costs
Investment is profitable in all cases because it increases

CIRED2005

Session No 2
CIRED 18th International Conference on Electricity Distribution
present technical value of the network and by that way
allowed rate of return. If interruption time is one of the PQ
factors in outage costs, efficiency score grows and investment
is profitable. Additional profit from efficiency benchmarking
is 1 100 €.
Allowed rate of return and DEA efficiency benchmarking with
outage costs and investments
Investment is increases present technical value of the network
and by that way allowed rate of return. Annual cost of
investment is more than annual saving in outage costs and
hence there is an additional 1 420 € – 1 100 € = 320 €/a
payment from efficiency benchmarking. Investment affects
also to operational costs but main effect is in PQ. Investment
is profitable only if it reduces operational costs at least 320
€/a.
Revenue regulation and power quality
Cost structure of network revenue regulation with outage
costs is presented in Fig. 9. Like in previous case annual cost
of investment is more than savings in outage costs.
Investment is profitable only if it reduces operational costs at
least 320 €/a.
Four different ways to estimate overall profitability of remote-
controlled disconnector investment in distribution business
were presented. Two last regulation methods meet
satisfactorily the requirements that are used in traditional
network planning and development. In these options
minimum of all cost components has to be founded before
profitableness of investment can be reached. In two first
options (Allowed rate of return and efficiency benchmarking
with interruption time/outage costs) there is possibility for
local optimization in sense of short term cost minimization.
Added to this, efficiency benchmarking contains a lot of
uncertainty when PQ (interruption time) is handled as
separated factor in DEA-model.
Similar case considering earth fault current compensation
equipment is presented in [10].
CONCLUSION
Power quality factors affects in different ways in different
kind of regulation and efficiency benchmarking methods.
This has to be taken account when unit costs of outages are
considered to be used in regulation process. In worst case
some of PQ factors do not affect at all to company’s
efficiency score and profit no matter how important and
necessary investment is in sense of economic network
developing. In DEA model power quality cannot be handled
as a separated factor because of unfair behavior of power
quality factor.
It is not enough to reach for as real outage unit costs for
different customer types and for different interruption types as
possible, but it is also necessary to understand in which way
power quality factors (and outage costs) affects to companies
CIRED2005
Session No 2
Turin, 6-9 June 2005
net revenue e.g. after efficiency benchmarking with power
quality in it. Because of new relation between power quality
and net revenue or efficiency benchmarking exists, actual
outage costs in companies’ perspective can be calculated. In
this paper, effects of PQ investment to companies allowed
rate of return in different kind of regulation models is
estimated.
REFERENCES
[1] Langset T., Trengereid F., The Norwegian Water Resources and
Energy Directorate. Samdal K., Heggset J., SINTEF Energy
Research, Norway, 2001. Quality dependent revenue caps - A model
for quality of supply regulation. Electricity Distribution. Part 1:
Contributions. CIRED. (IEE Conf. Publ No. 482), pages: 354-354.
[2] Lassila J., Viljainen S., Partanen J, 2002. Analysis of the
benchmarking results of the electricity distribution companies in
Finland. IEEE Postgraduate Conference on Electric Power Systems.
Budapest, Hungary. ISBN 963420725-1.
[3] Järventausta P., Mäkinen A., Nikander A., Kivikko K., Partanen J.,
Lassila J., Viljainen S., Honkapuro S, 2003. The Role of Power
Quality in Electricity Distribution Business Regulation, Publications
of Energy Market Authority 1/2003. (in Finnish), 171 p.
[4] Korhonen P., Syrjänen M., Tötterström M. Quantitive Methods Unit,
Helsinki School of Economics and Business Administration, Helsinki,
Finland. Assessment of Cost Efficiency in Finnish Electricity
Distribution Using DEA. (Summary in English) [3 Jan 2005]
www.energiamarkkinavirasto.fi/data.asp?articleid=254&pgid=133 (or
www.energiamarkkinavirasto.fi , English Æ Electricity market Æ
Supervision on network services)
[5] Charnes, A., Cooper, W. W., Lewin, A. Y.,1995. Data Envelopment
Analysis: Theory, Methodology and Applications. Boston, USA,
Kluwer Academic Publisher, ISBN 0-7923-9480-1, pages: 23-47.
[6] Lavaste K. Energy Market Authority. Consideration of efficiency in
the assessment of the reasonableness of electricity distribution
pricing.
www.energiamarkkinavirasto.fi/data.asp?articleid=253&pgid=133
(Summary) [3 Jan 2005]
[7] Lassila J., Viljainen S., Honkapuro S., Partanen J. 2003. Data
Envelopment Analysis in the benchmarking of electricity distribution
companies. CIRED 2003, International Conference on Electricity
Distribution. Barcelona, Spain.
[8] Kivikko K., Mäkinen A., Järventausta P., Tampere University of
Technology. Lassila J., Viljainen S., Partanen J., 2004. Methods for
Advanced Compilation of Interruption Statistics to Cater the
Renewing Regulation. Nordic Distribution and Asset Management
Conference, NORDAC 2004. Espoo, Finland.
[9] Kivikko K., Mäkinen A., Verho P., Järventausta P., Tampere
University of Technology. Lassila J., Viljainen S., Honkapuro S.,
Partanen J, 2004. Outage Cost Modeling for Reliability Based
Network Planning and Regulation of Distribution companies.
Proceedings of the DPSP 2004 Conference, Amsterdam, Netherlands,
4/2004. 4 p.
[10] Lassila J., Honkapuro S., Partanen J. Economic Analysis of Outage
Costs Parameters and Their Implications on Investment Decisions.
Accepted to be presented and published in IEEE PES general meeting
in San Francisco, California 6/2005.
[11] Lassila J., Viljainen S., Honkapuro S., Partanen J. Investments in the
Benchmarking of the Distribution Network Companies. The 2nd
International Conference on Electricity Utility Deregulation,
Restructuring and Power Technologies, IEEE, Hong Kong 4/2004.