• Nie Znaleziono Wyników

PD knowledge rules for insulation condition assessment of distribution power cables

N/A
N/A
Protected

Academic year: 2021

Share "PD knowledge rules for insulation condition assessment of distribution power cables"

Copied!
17
0
0

Pełen tekst

(1)

PD Knowledge Rules for Insulation Condition Assessment

of Distribution Power Cables

Edward Gulski, Johan J. Smit

Delft University of Technology Faculty of Technology and Management Mekelweg 4, 2628 CD Delft, The Netherlands

and

Frank J. Wester

Nuon InfraCore, Alkmaar, The Netherlands

ABSTRACT

This paper discusses practical experiences in the Netherlands with insulation con -dition assessment of distribution power cable networks by partial discharge diag -nosis and database support. In particular, practical steps are discussed in collect -ing, analysing and processing the diagnostic data for decision support of utility as -set management.

1

GENERAL

With regard to distribution power cables the technical Ž .

information needed for asset management AM decision support is not only dependent on diagnostics applied to the cable section but also it is dependent on several asset-related effects such as network operation and cable

Ž .

construction Figure 1 . The cable network is built up us-ing different types of components, with use different con-struction and insulation materials. Therefore, due to ferent service conditions the components may age in dif-ferent ways. In order to achieve a good diagnostic end result, it is necessary to have a number of information sources, concerning the test object, available to support the operator in making correct interpretations of the

mea-Ž . sured diagnostic signals e.g. partial discharge PD phe-nomena. Furthermore, the diagnostic information has to be combined with the knowledge rules to assess the asset condition. Finally this information can be used to support

w x the AM decision processes 1᎐15 .

Index Terms — Partial discharges, power cables,

on-site diagnosis, condition assessment, decision support.

2

INSULATION CONDITION

ASSESSMENT

Ž . Many breakdowns in the medium voltage MV power

w cables are caused by damages due to digging activities 12,

x

13 . But still, more than half of the breakdowns in the

Manuscript recei®ed on 15 April 2002, in final form 17 March 2004.

cable network are caused by internal fault in the

insula-Ž .

tion systems of the cable network Table 1 .

Visual inspection of the disturbed components gives in-sight in the different types of breakdown related insula-tion defects. Based on these visual inspecinsula-tions, a list of defects in different components of cable network is made, as reflected in Table 1. For many years, withstand testing Žac and dc was the only testing method applied in the. MV power cables network, but nowadays also PD testing has become an accepted method. As shown in Table 2, PDs are sensitive and till now they are the best symptoms

Ž

of discharging weak spots insulation defects, degradation

. w x

processes in the HV insulation 12 .

Due to the fact that only after the transition from wa-ter-tree to electrical-tree PD phenomena may occur for water-tree aged cables PD detection is not always an ef-fective diagnostics tool.

Figure 1. Different categories of technical information to support

(2)

Table 1. Typical insulation defects for different types of cable

com-w x ponents 1 .

Cable type Accessories Insulation Low oil level Damage outer sheet Sharp edges on connectors Tracking

PILC Moisture penetration

Airrgas bubbles Internal damage Žas a result of bending. Bad hardened resin

Sharp edges on connectors Moisture penetration Airrgas bubbles

XLPE Field grading movement Damage outer sheet Bad hardened resin

Interface problems Remaining semicon

3

SELECTING ADVANCED PD

DIAGNOSIS ON-SITE

Due to the physical character of discharge occurrence, such as the PD inception voltage and the PD pulse magni-tudes, PD patterns and mappings for a utility interested in applying PD diagnostics for condition assessment of its distribution power cable networks, a number of technical and economical aspects is of importance:

.

a Voltage type: equivalence in PD inception pro-cesses among different voltage stresses for solid insulating materials.

.

b Non-destructiveness: non-destructiveness of voltage stress during the diagnosis.

.

c IEC 60270 conformity: in the case of measuring the w x PD quantity apparent charge of PD pulses in pC and wnC the PD detection methods applied has to fulfil thex recommendation of IEC 60270.

.

d Sensitivity: immunity for on-site interferences and the level of system background noise.

.

e Analysis: possibility to generate a broad spectrum of PD diagnostic information to support diagnostic knowl-edge rules.

.

f Efficiency: investment costs, maintenance costs, transportability and operation of the method in different field circumstances.

Nowadays, a number of powerful diagnostics has found approval in the field. In Table 3 an overview of energizing methods for PD diagnosis is given.

Table 3. Standard off-line methods for PD diagnostic testing of

distribution power cables.

Voltage type Voltage Source

AC 50r60 Hz Inductively tuned resonant circuits

AC 15᎐00 Hz Frequency tuned resonant circuit

VLF 0.1 Hz 0.1Hz sine wave generator and amplifier

DAC Damped AC voltages 50᎐00 Hz Damped oscillating voltage wave excitation sources

Taking into account the specific utility situation, the search among different available on-site methods for best and optimal solution comparison analysis has to be done w10 . For example, evaluating the above mentioned param-x eters could be helpful in making proper selection among different methods. In particular, those methods showing the largest and centrally located curve are suitable to meet the defined requirements. Based on the Nuon utility ap-proach, the example in Figure 2 shows that in this particu-lar case, the DAC 50᎐500 Hz methods show the best fit-ting to aspects as defined above. Evaluafit-ting technical and economical aspects of a method may support the utility w x decision process in setting up diagnostic facilities 10, 13 .

4

DAC 50

᎐500 Hz: OSCILLATING

WAVE TESTS SYSTEM

To generate DAC voltages with a duration of a few tens of cycles of ac voltage at frequencies up to a few hundreds

w x

of Hz a system has been developed 4, 15, 20᎐22, 24 and

Ž .

in practical use for four years, Figure 3 . This method is used to energize, measure and locate on-site partial dis-charges in power cables in accordance with IEC 60270 recommendations. The system consists of a digitally con-trolled flexible power supply to charge capacitive load of power cables with lengths up to 10 km. With this method, the cable under test is charged during tchar g es Uma x

Ccabl erIl o ad with increasing voltage over a period of just a few seconds to the usual service voltage. Then a specially designed solid state switch connects an air-core inductor to the cable sample in a closure time of -1 ␮s. Now a series of voltage cycles starts oscillating with the resonant

Ž .

frequency of the circuit fD ACs 1r6 L Ccabl e , where L Ž . represents the fixed inductance of the air core 0.8 H and

w x

Table 2. Typical insulation degradation processes of the cable insulation 1 .

interface problems ™ PD™ tracking; Accessories bad hardening ™ cracking ™ PD;

conductors problems ™ overheating ™ cracking ™ PD; local field concentrations ™ PD

water trees ™electrical trees ™ PD

Extruded Insulation insulation voids ™delamination ™electrical trees ™ PD local field concentrations ™ PD

oil leaks ™dry regions ™ overheating ™ PD PaperrOil Insulation water ingress ™ load effects ™ overheating ™ PD

(3)

Figure 2. Example of an overall evaluation of different PD

diagnos-tics for distribution power cables as applied at Nuon utility.

Ž .

Figure 3. Proposal of a circuit to generate damped ac DAC

volt-age stresses.

Ccablerepresents the capacitance of the cable sample. The air core inductor has a low loss factor, so that the reso-nant frequency lies close to the range of power frequency of the service voltage: 50 Hz to several hundred Hz. As a

Ž

result, a slowly decaying oscillating waveform decay time .

up to 0.3 s of test voltage is applied to energize the cable sample. The quality factor QC of the resonant circuit, which is responsible for the attenuation of the oscillations,

Ž Ž U 2.

can be expressed as: QC s6 Lr C R . Here R is theA A equivalent circuit resistance. The quality factor Q of the

resonant circuit remains high depending upon the cable Ž30 to more than 100 , as a result of the relative low dissi-.

Figure 4. Partial discharge occurrence during DAC voltages.

pation factor of power cables. In the case of 1rQ is higher

than the dielectric losses of the test object the total decay of the damped ac wave may be less uncertain. During tens of power frequency cycles the PD signals are initiated in a

Ž .

way similar to 50 or 60 Hz inception conditions, see Fig-ure 4. The laboratory and the field research has shown, that applying DAC voltages with frequencies up to 500 Hz

Ž shows no significant differences in the PD occurrence PD

. w x Ž . inception voltage, PD levels 15 Figure 5 .

In general, when switching on or increasing a dc volt-age, the dielectric is stressed as if it were an ac voltage. As a result, capacitive current ic flows which is i s Cc

ŽdUrdt. w23 . With regard to the charging period of thex

Ž .

cable sample as proposed in the DAC method Figure 3 , when raising the voltage the field distribution is capacitive and the ⑀ ’s of the insulating materials are directing the

w x

field 23 . Moreover, applying to the cable section under test a continuously increasing voltage stress: duration 1᎐2

Ž .

s 0.5 Hz ᎐ 1 Hz repetition frequency , directly followed by a switching and damped ac stresses in the range of tens of Hz no ‘‘steady state’’ dc stresses occur in the test

sam-Ž .

Figure 5. Example of PD mapping as made for a 10 kV paper insulated lead cable PILC power cable tested at two different DAC voltage

(4)

w x

ple 23 . As a result of this continuous ac stress, in the cable insulation no charge accumulation will occur.

Due to the fact that the decay of the DAC voltage de-pends on the dielectric losses of the cables section under test, different time constants may be observed. This effect can also be used to evaluate e.g. compare the dielectric

y3Ž .

losses in Nⴢ10 e.g. by calculating the tan ␦ parameter as obtained on periodic inspections during the service life of the same type of cable insulation.

5

PD QUANTITIES AND

INTERPRETATION

All relevant information from a PD cable measurement should be collected. Moreover, this diagnostic information as given by PD advanced quantities has to be combined

Figure 6. Basic knowledge elements of condition assessment: PD

Ž .

diagnostic information upper part ; power cable technical

informa-Ž .

tion lower part .

(5)

Ž with technical data of the particular cable section Figure

. 6 .

To describe the PD process in the cable section under investigation OWTS diagnostics method generates several PD quantities, which can be divided into two groups:

w x

1. Basic quantities: PD level in pC or nC , PDIV and w x

PDEV in kV .

2. Derived quantities: e.g. q-V curve, phase-resolved PD Ž

pattern, PD magnituderintensity local frequency of PD pulses in function of the power cable length: PD map-pings.

This PD information which can be determined at differ-ent voltage levels e.g. up to 2U is collected in a so-called0

PD ‘fingerprint’ of a power cable section. An example of a complete PD ‘fingerprint’ is shown in Figure 7.

The characteristic of these quantities measured on dif-ferent cable sections may vary in their dependence on

fac-tors like type, age, service history, and location of the ele-ments used. Linked to the different parameters from the PD fingerprint, interpretation rules are of importance as

w x

reflected in 12, 13, 15 .

In order to provide more diagnostic information about the cable tested in the field the PD measuring data have to be processed and compared to those obtained on simi-lar cable sections in the past.

In particular, using special databases containing

impor-Ž .

tant information see Figure 6 the decision support can w x

be made for asset management purposes 17᎐19 . The stochastic character of PD phenomena on the one hand, as well as a number of parameters like test voltage level and measurement duration on the other hand may influence the quality of the diagnostic information used for database evaluation. Therefore, to obtain maximum of benefit on diagnostic inspections standardized database

Figure 8. PD mapping and database protocol made automatically after the 3 phase of the 600 m long distribution power cable are tested in

accordance with the test protocol as shown in Figure 10: the mapping of PD amplitudes r intensities as observed at test voltages between PDIV

Ž .

and 2 U0 upper part , database report of important PD quantities as observed for the whole cable section and particular insulation parts, joints

Ž .

(6)

protocol has been developed to extract in similar way the necessary information, see Figure 8.

w x

In 5 the data mining approach show the use of this diagnostic information to provide decision support for in-sulation condition assessment and maintenance purposes.

6

ON-SITE DIAGNOSIS AND

EVALUATION

6.1

TESTING PROCEDURE

Due to the fact that from the network operation point of view, the network unavailability time of a cable section Žfor maintenance and diagnosis purposes should be as. short as possible, the on-site PD diagnosis should be done in accordance to a certain time regime. Moreover, follow-ing the same way of applyfollow-ing the test voltage, and collect-ing the measurcollect-ing data is of importance for the data qual-ity.

In Figure 9 an example of such field routine procedure is shown. It follows from this example that within 1 h the relevant PD measuring data can be collected and

evalu-Figure 9. Example of a test time regime as developed for testing

distribution power cables using DAC method; to obtain an overview of PD location in function of power cable length the Time Dominie

Ž .

Reflectometry TDR tools are used.

Figure 10. Example of a database record as automatically generated for a power cable. Based on the measuring data as observed for at

different voltage levels and analyzed for particular cable parts, accessories of all three phases a database record can be made consisting for condition assessment relevant information.

(7)

Ž . Ž .

Figure 11. PD measuring result 10kV XLPE power cable where PDIV is just above U and PDEV just below U0 0 8.2 kV . In case of an Ž .

over-voltage during operation, the PD source will be ignited and discharges will stay during service U .0

ated in a test report for a 3-phase power cable. In addi-tion, an in-depth analysis is possible after the power cable has returned into operation. In Figure 10 an example of a database protocol is shown which can be automatically generated after the on-site testing has been finished. Ob-taining detailed information of PD activity in all parts of the power cable and accessories in function of test voltage is relevant for condition assessment as discussed in Sec-tions 7᎐13.

6.2

DISTRIBUTION POWER CABLES:

XLPE INSULATION

PD testing of XLPE insulated power cables is in use as an after-laying test to search for poor workmanship of in-stallationrrepair work or during the service life to search for insulation degradation processes. In both cases no PD will be accepted in the cable insulation.

On the basis of different parameters of the PD ‘fingerprint’, evaluations of the insulation condition of a cable section can be made with the results of one mea-surement session. Actually, the most important parameter of the fingerprint is the inceptionrextinction voltage of the partial discharges. If the partial discharge inception

volt-Ž .

age PDIV is below the operation voltage of the power cable, this means that PD process is continuously active

during operation of the circuit. However, testing a power cable up to the operation voltage is not adequate. Even if the PDIV is higher than the operation voltage, the

extinc-Ž .

tion voltage PDEV can be lower, Figure 11. In case of just a small over-voltage, as a result of e.g. switching, the PD source will ignite and stay active during service. This will accelerate the aging of the cable insulation located at the PD source. Therefore, power cables should always be tested with higher stresses than nominal. Of course, more parameters are of importance to the evaluation of insula-tion condiinsula-tions.

Figure 12 shows the PD occurrence observed for the three phases of a 10 kV XLPE cable. It follows from this figure that only phase L1 shows PD at test voltage

operation voltage PDIV and PDEV of phase L3 are .

higher than U , with PD magnitudes0 )100 pC. Using

Ž .

time domain reflectometry TDR analysis the PD source was located in a termination. In Table 4, based on two

Table 4. PD occurrence in XLPE distribution power cables.

PD presence PD location 10% UPD IV1.3 U0 17% in splices 27% UPD IVs1.3 U0

18% UPD IVsU0 83% in terminations 45% PD free 0% in cable insulation

(8)

years systematic field measurements in the Netherlands on new and service aged XLPE power cables, results of PD occurrence are shown. Moreover, based on the

practi-Ž .

cal experiences Nuon, utility , a cable specimen under test is considered to have passed the PD-free requirements when PDIV) 1.3 U and PDEV ) 1.1 U . On the con-0 0

trary, based on two years of practical experiences in the dependence on the type of the joint termination, certain

Ž .

levels of PD e.g. - 500 pC can be accepted in the cable accessories.

6.3

DISTRIBUTION POWER CABLES:

PILC INSULATION

Ž . The PD diagnosis of paper-insulated lead cables PILC has been introduced a few years ago to localize insulation

w x

weak-spots in aged power cables 10, 15 . Due to the fact that these cables are not PD-free for both the cable insu-lation as well as cable accessories depending on the type

of PD patternsrPD mappings, certain levels of PD can be accepted.

Figure 13 shows the measuring results on a 3-phase PIL cable section with two aged joints. The PD pattern of phase L3 differs from L1 and L2; in L3 PD occurs in the negative as well as the positive side of the oscillating volt-age. As known, this refers to different PD source types, which is true in this particular case. The PD mapping of the cable section in Figure 11, shows that the detected PD is originating from two different types of joints at 1260 and 1300 m from the measuring side. Due to the fact that for PILC insulation a certain PD levelrintensity is allowed it is of importance to determine the critical limits for PD activity. As a result, knowing the type of the discharging splice makes it possible to evaluate the measured values with acceptance threshold values of this particular type.

In Table 5 based on four years systematic field mea-surements in the Netherlands on new and service aged

Figure 12. PD test results obtained from three phases of a XLPE 10 kV power cable. Phase L1 shows the presence of a PD source, located in

(9)

Figure 13. PD diagnosis of a 2860m long 10 PILC cable showing discharge activity in different types of splices: splice L3 located in the joint at

1268m and splices L1 and L2 located in the joint at 1300m.

Table 5. PD occurrence in PILC distribution power cables.

PD presence PD location 29% UPD IVs1.3 U0 51% in splices

61% UPD IVsU0 33% in terminations 10% UPD IV-U0 16% cable insulation

PILC power cable results of PD occurrence are shown. Moreover based on the practical experiences for most fre-quently occurring PILC accessories types acceptance threshold values have been determined, see Table 8 w12᎐13, 15 .x

7

HV COMPONENT IDENTIFICATION:

POWER CABLE SYSTEM

Due to the fact that the construction and the operation of a HV asset is of importance for systematic condition assessment, the relevant elements have to be identified. For example, a distribution power cable is used to inter-connect small substations. These cable sections may

con-Ž .

sist of three different types of components Figure 14 . For practical reasons, a cable section may be constructed from multiple parts of cable, which are connected to each other with cable joints. At both ends of the cable, a

termi-Figure 14. Construction of a cable-section from different types of

components the age of components may also vary.

nation is mounted to connect the cable to HV installa-tions. Due to repairs or changes in topology, cable sec-tions often consist of various types of joints, cable parts and terminations. In addition, the age of components may vary due to repairs over the years. Moreover, the different topological and operational conditions of a particular ca-ble system influence the service life of a caca-ble system enormously.

(10)

Table 6. Database domains of power cables.

Component consisting of specific insulation characteristics domain of the cable structure; e.g. cable insulation,

accessories;

Diagnostic consisting of the type of diagnostics applied; domain e.g. general condition assessment, assessment of

weak spots;

Measuring data defining specific diagnostic quantities used domain during condition assessment.

For assessment of the insulation condition of power ca-bles, some linkage between the following information has to be obtained, then combined and used to generate knowledge adequate for decision making, see Table 6.

8

DATA BASE

To support condition assessment of distribution power cables, generic database systems are needed to manage the data from measurements and inspections. The insula-tion condiinsula-tion assessment based on suitable diagnostic pa-rameters requires so called knowledge rules and trigger Žor norm levels.. In addition, the systems for manage-ment of measuremanage-ment data can be used to perform statis-tical analysis, from which knowledge rules with trigger lev-els can be set. Furthermore these systems can also be used to trigger inspection, maintenance or replacement actions. Such a system at least consists of the following compo-nents, see Figure 15:

. Ž

a Module to input or define the cable data structure, .

component types . .

b Module to input measurements. .

c Module for the analysis of the data. .

d Database to store the data.

Separating the functionality of the system in these four blocks has certain advantages. Extra modules can be added to store data in or to use data from the database. Also this design anticipates on a future multi-user environment. For example, multiple users could use multiple copies of the modules, while all the data resides only in one big central database.

9

ACCEPTANCE LEVELS FOR

CONDITION BASED MAINTENANCE

(

CBM

)

Statistical analysis of historic data can be used to estab-lish criteria for condition assessment. In particular, by

us-Ž

ing statistical analysis overall behaviour average values, .

95% significance level as norm is taken into account. The frequency distribution can be used to determine norm lev-els. To set a norm level, a decision has to be made about which risk is acceptable, which is mainly a comparison be-tween cost and reliability.

From this decision follows, which part of the population of components should be regarded as dangerous. The

ac-Figure 15. Schematic structure of a diagnostics database to support

condition assessment of power cables.

quired data are compared with the criteria by the norm

Ž .

system software . The norm levels are used to trigger cer-tain alerts for inspection, maintenance or replacement ac-tions. Moreover the criteria of two or more diagnostic parameters can be combined. However, one should take caution because several pitfalls exist. It has to be kept in mind that the statistical analysis should be performed on components and not on measured values. If a component has been measured multiple times, it will have a larger influence on the frequency distribution from measured values of a certain quantity than another component that has not been multiplied. Secondly, it may be questioned how representative the measured samples or components are for the total population. For example, experienced personnel of power utilities tend to start measurements or inspections at those components where they expect to find problems, so a representative sample set is only con-structed after a large number of components have been measured or inspected.

At last, due to noise and disturbances, or operator omission, the distribution may not be completely found in the measured data. For example, a measurement operator may find small PD not that important, and thus omits it in hisrher analysis.

10

DATA MINING FOR CBM

With regard to the definition of data mining in general the following characteristics can be given:

The data mining involves the earlier stages of data

selection and data transformation and the subsequent w x

stages of validation and interpretation 15 .

Furthermore data mining aims to provide an

alterna-tive to the traditional scientific method, where data analy-sis is largely directed by hypotheanaly-sis and theory.

䢇 The aim of data mining is to find intelligible patterns

w x

which are not predicted by established theory 6 . Format-ting the output data in a visual form that human intelli-gence can interpret is important.

䢇 In particular evaluation of existing experience in data

mining applications for condition monitoring, including w x validity and relevance of results is of importance 17 .

In the next sections, use of partial discharge data of power cables is discussed in the scope of these aspects.

(11)

Partial discharges, because of their wide variation with equipment type, and with electrical configuration and ma-terial type, provide a very suitable vehicle for detailed data mining applications. In addition to the example given re-lating to cables, partial discharges can be used as the basis for data mining applications.

The full potential of condition and maintenance infor-mation cannot always be realized by using traditional techniques of data handling and analysis. There are often underlying trends or features of the data that are not evi-dent from the usual analysis techniques. Such detail and trends can be important for the assessment of the equip-ment operation and there are increasing demands from operators and asset managers to fully exploit the capabili-ties of this data in order to optimize the utilization of high voltage electrical plant. The method of extracting full value from such extensive databases, using new analysis

tech-w x

niques, is commonly called data mining 6 . In its basic essence, data mining is the application of relatively novel data-driven approaches to find patterns in data obtained from electrical equipment. The data mining techniques are then used to relate these patterns to the operational con-dition of the equipment and to provide new knowledge about aging mechanisms, norms, and required mainte-nance activities. In order to support the data mining pro-cess utility may apply the knowledge expert database in which all the measured data and failure information is stored and analysed. The general features of this ap-proach are shown in Figure 16.

Ž .

This process provides three outcomes A, B and C of Ž .

the data mining approach. Outcome A refers to new knowledge about aging mechanisms of power cable

com-Figure 16. Schematic structure of data-mining process for

condi-tion-based maintenance of HV assets, e.g. power cables.

Ž .

ponents and outcome, B to recommended maintenance activities on power cable components, resulting from the database analysis. Due to the large amount of measure-ment data stored in the software data system, operating norms and criteria are continuously updated and fed back

Ž . to workers in the field as determined by result C .

11

DUTCH UTILITY PRACTICAL

EXPERIENCES: INSULATION

CONDITION KNOWLEDGE RULES

In Table 7 schematic example of interpretation rules for PD measurements on power cables are shown. These are rules of thumbs supporting the analysis of the mea-surement results from a cable section. Different aspects of the ‘fingerprint’ are used in these interpretation rules.

After measuring and analysing PD activity in a cable section, the second step is to make a decision on the insu-lation condition of the tested cable sample. Using the Ž . measured PD quantities and their interpretation rules 3 three condition classes can be derived from the analysis:

.

a Cable section NOT OK: weak spot in the cable sec-tion should be replaced.

. Ž

b Cable section NOT OK: periodic inspections

trend-. Ž

ing on the cable component is required e.g. 1 year, 3 .

years . .

c Cable section OK: no weak spots in the cable section, cable section is OK.

Figure 17 shows the decision diagram for power cables. Analysing the derived measurement data through the dia-gram, the cable systems insulation condition can be

deter-Figure 17. Decision diagram for diagnostic data analysis to

catego-rize the insulation condition in one of the three condition classes.

Table 7. Interpretation rules for PD diagnostics on power cables.

Inception & Extinction )cable operating voltage -cable operating voltage

Ž . Ž .

PD magnitude -typical values e.g.-4000 pC )typical values e.g.-4000 pC

Ž . Ž .

PD intensity Low e.g.-5 pulsesrperiod High e.g.-5 pulsesrperiod PD pattern harmful fault type less harmful fault type PD location cable insulation cable accessories PD mapping PD concentration scattered PD location

(12)

mined in one of the four classes. The decision diagram is based on all PD quantities for measurements on power cables. Not locally concentrated PD in the PILC cable in-sulation is not an indication of aging of the cable insula-tion materials.

For PILC related insulation, in the dependence on the insulation type, certain levels of PD can be accepted in

Ž .

the cable insulation e.g. less 8000 pC , joint and termina-Ž

tion, depending on the design of the component between . w x

500 pC and 5.000 pC 2 . Also, XLPE related accessories Ž

could stand certain PD levels generally lower pC levels .

compared to PILC . However, XLPE cable insulation is

Ž .

required to be PD free PDIV)1.3U , PDEV)1.1U ,0 0

where the background noise during an on-site measure-ment in dependence on the background noise and distur-bances level is not allowed to exceed 10᎐20 pC. The typi-cal PD values for the different components can be derived from the statistical analysis of all required measurement

data in a database. As a result for particular components threshold values can be derived to support the condition

Ž .

based maintenance CBM , see Table 8.

Table 8. Indication of some typical PD levels for different types of

cable insulation and accessories based on experiences from The Netherlands.

Cable element Type Trend values insulation PILC 10.000pC

XLPE -20 pC

splices Oil-joint 5.000pC

Ž Ž .

Type 1 resin insulation 500pC asymmetric Žoil insulation. )10.000 pC Žoilrresin insulation. 5.000 pC

Ž .

Type 2 resin insulation 4.000 pC terminations Oil-termination 6.000 pC Dry termination 3.500 pC Type 3 termination 250 pC

Figure 18. PD levels as observed at U for PILC database population of two different types of cable splices. a, splice type resin insulated with0

(13)

12

DUTCH UTILITY PRACTICAL

EXPERIENCES

12.1

DATA MINING APPROACH AND

NORMS

Statistical evaluation of diagnostic parameters e.g. the PD level at the service voltage or PD inception voltage ŽPDIV are powerful means to obtain for specific compo-. nents e.g. certain types of splices, terminations or cable insulation types characteristic values and the statistical significance limits. Based on these parameters for those components norms can be estimated. Based on a large population of measuring data as obtained on one type of cable accessory the average PD level and the 95% signifi-cant limits can be calculated.

In Figure 18 for two different types of splice insulation the population of PD level at U is shown. It follows from0

these examples that for resin insulated splices the average PD level is 2.4 nC and for oil filed splices is 3.6 nC. Re-spectively, the significance limits are 4.7nC for resin-in-sulated splices and 8.2 nC for oil filed splices.

As a result of the above mentioned examples it can be concluded that using large population of measuring data, database selection of different insulation types result in sensitive distinction in threshold values for decision sup-port.

12.2

SEPARATION OF CONSTRUCTION

EFFECTS AND AGING

Using PD diagnosis as applied to large populations of PILC power cables provides support in separating the ag-ing processes and effects of different power cable insula-tion types. As is well known, a PILC power cable is not always PD free and a certain level in the range of nC, e.g. 10 nC of PD activity is accepted. Figure 19 shows an ex-ample of PD amplitudes as detected at U on a large pop-0

ulation of PILC cables. It follows from this figure that at

Ž . Ž .

U0 service voltage the typical average -PD amplitude in PILC power cable insulation is in the range of 4.7 nC. As a result, cable section with PD larger as 12.9 nC are signif-icantly different from the cable population.

Figure 19. Database selection of all PILC power cables showing PD activity at U level. a, definition of the database filter; b, population of0

(14)

It has been observed that in contrast to mineral oil im-pregnated insulation in the case of synthetic oil

impreg-Ž .

nated cables manufactured after 1980 due to tempera-Ž . Ž

turerpressure effects TrP load changes or switching off

. w x

the cable section result in additional PD activities 10 . Large discharge amplitudes characterize these PD pro-cesses and they are of temporary art and they do not indi-cate insulation degradation.

This situation forces to take during data evaluating and condition assessment into account the construction as-pects, e.g. type of insulation impregnation. As a result, selecting in the database the cables influenced by TrP ef-fects shows that the typical parameters like the average PD and the significance levels are changing.

Figure 20 shows an example of the database selection of PD amplitudes as detected at U on a large population0

of PILC cable only where the TrP effects have been ob-Ž served. It follows from this analysis that at U0 service

. Ž .

voltage the typical average PD amplitude in PILC power cable insulation is in the range of 9.3 nC. As a result, cable section with PD larger as 18 nC are significantly different from the cable population. In order to obtain typical PD amplitudes in relation to insulation degrada-tion, those cables where no TP effect is measured should be obtained.

Figure 21 shows an example of the database selection of PD amplitudes as detected atU on a large population0

of PILC cable, where the cable sections the TrP effects are left out the observation. It follows from this analysis

Ž . Ž .

that atU service voltage the typical average PD ampli-0

tude in a PILC power cable insulation is in the range of 3.7 nC. As a result, cable section with PD larger as 11 nC are significantly different from the cable population. From these examples, it follows that as a result of TrP effects, the norm level is influenced. If all PILC insulation were used for norm calculations, the norm would be 12 nC.

Figure 20. Database selection of those PILC power cables showing PD activity at U level, where the temperaturerpressure effects have been0

(15)

Figure 21. Database selection of those power cables showing PD activity at U level where the temperaturerpressure effects are not observed.0

a, definition of the database filter; b, population of cable sections and the PD levels; norm 11 nC.

The right norm level should be 11nC, obtained by filter-ing only the degradation effects. This example shows, that results of statistical analysis are sensitive to different con-struction parameters. As a result, condition assessment and decision support need data-mining support, see Fig-ure 16.

12.3

DECISION SUPPORT

Combining the component data with the condition data as obtained using diagnostic inspections can be used to develop decision support for asset management.

In Figure 22 two examples are shown of such decision w x

support. Based on the following information 6,7 : 1. PD parameters as detected for the particular cable type.

Ž .

2. Significance limits norms .

Ž .

These cable sections can be selected marked in red from whole cable population where statistical deviating

condition is observed. Conform the Figures 1 and 2 of this paper this technical information now can be further com-bined to economic and strategic decision processes as de-fined by the asset management of the power utility.

13

CONCLUSIONS

N this paper, several aspects condition

assess-I

ment for MV power cables are discussed and con-cluded.

1. Condition assessment is an important element of As-set Management maintenance approach.

2. Condition assessment means combination of ad-vanced diagnostics and technical information of the actual network situation.

3. With regard to distribution power cables, PD detec-tion is a good method to get insight into discharging insu-lation defects.

4. Selecting proper advanced PD diagnosis needs evalu-ation of technical and economical aspects.

(16)

Figure 22. Example of using knowledge rules and database to support the decision process of asset managers in making distinction between

cable networks with different insulation conditions; the reed marked cable section shows increased insulation degradation as defined by the knowledge rules.

5. Using PD diagnosis at DAC voltages PD discharging defects in XLPE and PILC cable insulation can be de-tected. To obtain on-site full diagnostics information re-quires standard test procedures.

6. To use diagnostic information for AM decision sup-port standardized database protocols are necessary.

7. Database support in combination with data mining are important elements in obtaining relevant information for condition assessment of assets e.g. powers cables. By using dedicated non-destructive diagnostic techniques and storage of the quantities in a database, analysis can be performed to extract additional information not shown by the measurements only. Based on this information, deci-sions can be made about levels for updating or mainte-nance. When these levels are applied on the database as

norm levels, components are easily selected for inspec-tion, maintenance or replacement actions.

REFERENCES

w x1 E. Lemke, R. Roding and W. Wei¨ ␤enberg, ‘‘On-site Testing of Extruded Power Cables by PD Measurements at SI Voltages’’, CIGRE Symp, Vienna, Paper 1020-02, 1987.

w x2 F. Farneti, F. Ombello, E. Bertani and W. Mosca, ‘‘Generation of Oscillating Waves for After-laying Test of HV Extruded Ca-ble Links, CIGRE, Paris, Paper 21᎐10, 1990.

w x3 R. Bach and W. Kalkner, ‘‘Comparative Study on Alternative Test Voltages for Laid Medium Voltage Cables’’, 7th ISH, Dres-den, Germany, CIGRE, Paper 4.3, 1991.

w x4 R. Plath,Oscillating ®oltages als Prufspannung zur Vor-Ort Prufung¨ ¨

und TE-Messung kunstoffisolierter Kabel, Doctoral Thesis, TU,

Berlin, Germany, 1994.

w x5 E. Lemke and P. Schmiegel, ‘‘Complex Discharge Analysing ŽCDA - An Alternative Procedure for Diagnosis Tests of HV.

(17)

Power Apparatus of Extremely High Capacity’’, 9th ISH, Graz, Austria, Paper 5617, 1995.

w x6 W. Boone, G.C. Damstra, W.J.L. Jansen and C. de Ligt, ‘‘A Very Low Frequency High Voltage Generator for Testing Ca-bles After Laying’’, 5th ISH, Braunschweig, Germany, Paper 1.28, 1987.

w x7 E. Gulski, J.J. Smit, P.N. Seitz and M. Tuner, ‘‘On-site PD Di-agnostics of Power Cables Using Oscillating Wave Test System’’, 12th ISH, London, Paper 5.112᎐5.115, 1999.

w x8 E.R.S. Groot, Ph. Wester, E. Gulski, F.J. Wester, C.G.N. de Jong, M. van Riet, D.M. Harmsmen, J. Pellis and H. Geene, ‘‘Tools for Quality Assessment of Distribution Cable Networks’’,

Ž .

Int. Conf. Electricity Distribution CIRED , Paper 3.17, 2001. w x9 R. Plath, W. Kalkner and I. Krage, ‘‘Vergleich von

Diagnosesys-temen zur Beurteilung des Alterungszustandes PErVPE-iso-lierter Mittelspannungskabel’’, Elektr. Wirtschaft, Vol. 96, pp. 1130᎐1140, 1997.

w10 V. Colloca, A. Fara, M. de Nigris and G. Rizzi, ‘‘Comparisonx Among Different Diagnostic Systems for Medium Voltage Cable

Ž .

Lines.’’ Int. Conf. Electricity Distribution CIRED , paper 1.53, 2001.

w11 Y. Muramoto, E. Gulski, F.J. Wester and J.J. Smit, ‘‘PD Incep-x tion Conditions and PD Pattern of Defects in PILC’’, IEEE 7th ICSD, pp. 369᎐372, 2001.

w12 E. Gulski, F.J. Wester, W. Boone, N. van Schaik, E.F. Steennis,x E.R.S. Groot, J. Pellis and B.J. Grotenhuis, ‘‘Knowledge Rules Support for CBM of Power Cable Circuits’’, CIGRE, Paris, SC 15 Paper 104, 2002.

w13 F.J. Wester, E. Gulski and J.J. Smit, ‘‘CBM of MV Power Cablex Systems on the Base of Advanced PD Diagnosis’’, 16th Int. Conf.

Ž .

Electricity Distribution CIRED , Amsterdam, The Netherlands, Paper No. 1.29, 2001.

w14 E. Lemke, P. Schmiegel, H. Elze and D. Rux ␤ wurm, ‘‘Procedure for Evaluation of Dielectric Properties Based on Complex

Dis-Ž .

charge Analyzing CDA ’’, IEEE Intern. Symp. Electr. Insul. ŽISEI , Montreal, Canada, pp. 385. ᎐388, 1996.

w15 E. Gulski, F.J. Wester, J.J. Smit, P.N. Seitz and M. Turner,x ‘‘Advanced PD Diagnosis of MV Power Cables using Oscillating Wave Test System’’, IEEE Electr. Insul. Magazine, Vol. 16, No. 2, pp. 17᎐25, 2000.

w16 P. Groenefeld and R.v. Ohlhausen, ‘‘A Very Low Frequency 200x kV Generator as a Precondition for Testing Insulating Materials with 0.1 Hz AC Voltage’’, 4th ISH Symposium, Athens, Greece, pp. 7᎐15, 1983.

w17 B. Quak, E. Gulski and J.J. Smit, ‘‘Datamining and Informationx Processing to Support the AM Decision Making Process’’, ERA Conf. Engineering Asset Management, London, UK, 2002. w18 J. McGrail, E. Gulski, E.R.S. Groot, D. Allan, D. Birtwhistlex

and T.R. Blackburn, ‘‘Datamining Techniques to Assess the Condition of High Voltage Electrical Plant’’, Cigre, Paris, Paper´

WG15.11, 2002.

w19 Wester Ph. and R.D. Damstra, ‘‘The Impact of A Conditionx Based Management Strategy On NetworkSystem Operators’’, 38th Cigre Session, Paris, Paper 1´ ᎐13, 2000.

w20 F. Farneti, F. Ombello, E. Bertani and W. Mosca, ‘‘Generationx of Oscillating Waves for After-laying Test of HV Extruded Ca-ble Links’’, CIGRE Session, Paris, Paper 21-10, 1990.

w21 Lefevre, W. Legros and W. Salvador, ‘‘Dielectric Test with Os-x ` cillating Discharge on Synthetic Insulation Cables, Int. Conf.

Ž .

Electricity Distribution CIRED , Paris, France, pp. 270᎐273, 1989.

w22 S. Brettschneider, E. Lemke, J.L. Hinkle and M. Schneider,x ‘‘Recent Field Experiments in PD Assessment of Power Cables, Using Oscillating Voltage Waveforms’’, IEEE ISEI, pp. 546᎐552, 2002.

w23 F.H. Kreuger,x Industrial High DC Voltage, Delft University Press,

1995.

w24 CIGRE WG D1.33, ‘‘Experiences in Partial Discharge Detec-x tion of Distribution Power Cable Systems’’, Electra, No. 2008, pp. 34᎐41, 2003.

Cytaty

Powiązane dokumenty

Fakt, iż historia nowożytnej podmiotowości jawi się nam dziś nie jako otwarte pole możliwości, lecz raczej pole ograniczone, którego bieguny określają możliwe postaci

Spośród sześciu realizowanych przez PROW na lata 2014-2020 prio- rytetów wyznaczonych dla unijnej polityki rozwoju obszarów wiejskich na lata 2014-2020 najbardziej istotny

c Figure A.4: Statistical overview of the registered outages related to the different components insulation / construction in the distribution cable network: a cable, categorised

Eor the present study, all the surface data are represented using the Initial Graphics Exchange Specification (IGES) format. Then, the data were used for multiblock grid

The second rule states that no functional element with a weighting coefficient of three or more shall have a condition – determined by matching the level of defect on the scale used

enhancement. presence of internal cavities on the outer conductor in MV cables design has lower breakdown impact for the same cavity which is close to inner cable conductor. In

The origin of discharging defects can be related to poor workmanship (new or repaired cables) or to service related insulation degradation effects. In all these cases, due to

De globale methode voor het toetsen en ontwerpen van steenzettingen, ook wel black- box model genoemd, is in de afgelopen jaren steeds verder verbeterd.. Getracht is om de