ADVANCED FAILURE RATE AND DISTRIBUTION NETWORK RELIABILITY MODELLING AS PART OF NETWORK PLANNING SOFTWARE

Transcription

1 ADVANCED FAILURE RATE AND DISTRIBUTION NETWORK RELIABILITY MODELLING AS PART OF NETWORK PLANNING SOFTWARE Jouni PYLVÄNÄINEN, Pekka VERHO, Jussi JÄRVINEN Tampere University of Technology - Finland jouni.pylvanainen@tut.fi Susannna KUNTTU, Janne SARSAMA VTT Industrial Systems - Finland INTRODUCTION Due to regulation power quality and reliability issues are now one of the major concerns in network management business. Many utilities endeavour to rationalize their network operations and optimize total life cycle costs of their network components without endangering the reliability and safety of the network. New type of modelling methods which utilise collected network information more effectively can help utilities in managing this kind of asset management problems. One type of enhanced asset management tool can be based on distribution network reliability analysis. Failure rate models used in traditional reliability analysis are based on constant failure rate without any consideration of operational stresses and environmental factors of the network components. In the paper an advanced failure rate modelling is depicted. As a basis for the modelling, a large amount of failure data with accurate environmental information from various network utilities has been collected and analysed. Failure rate model together with the distribution network reliability and cost analysis has been implemented as a part of the distribution network planning software, which is also shortly depicted in this paper. BACKGROUND The operational basis of many electricity distribution utilities has changed quite radically during recent years. Many municipality owned utilities has been privatized and the new owners are usually capital investors looking for new profitable investments. This sets new type of challenges to control regional network monopolies. To avoid power quality level reduction in distribution networks, various types of regulation models are issued. Regulation models enforce one way or another network utilities (i.e. both public utilities and privately owned companies) to optimise their operations without endangering the reliability and safety of the network. Because of this, e.g. economical, environmental and reliability related aspects should be taken, besides the electrotechnical conditions, more carefully into consideration when making decisions on future network development actions. In many countries huge renovation process is waiting in a near future and for this reason, modern tools for optimal asset management are needed. A basis for enhanced asset management application is possible to adapt from traditional reliability analysis. One version of reliability analysis has been developed in the Tampere University of Technology (TUT) already in 1980 s. Although the developed reliability analysis is quite comprehensive and can be utilised e.g. to evaluate optimal disconnector locations, there are a lot of possibilities to develop it further. E.g. in the analysis, failure rates are constant for similar components (6 failures per year per 100 km for all aerial lines, 2 failures per year per 1000 transformers, etc.) without taking into account any other aspects influencing to the reliability of components. In reality, e.g. mechanical, environmental and electrical stresses with component type affect to the failure rate of the component. The enhanced reliability based network modelling methods and algorithms have been developed as part of LuoVa-project (reliability based network analysis -project). The project has been carried out in co-operation with TUT, VTT, various software houses specialised in distribution systems and several network utilities operating in Finland. In this way the best knowledge on Finnish distribution networks has been available for advanced reliability modelling development. The different parts of the developed modern reliability based network analysis are depicted in Figure 1. In this paper, emphasis is in component modelling but also other parts of the entity are depicted shortly in following sections. λ 1= (w1w2 wn)λ1 λ 2= (w1w2 wn)λ2. λ n= (w1w2 wn)λn Component modelling λ k = Σ λ 1 n λk1. λkm Radial network reliability analysis Customer information λ U Cost Model Figure 1: Modern reliability based network analysis COMPONENT FAILURE MODELLING Existing component failure models C, etc. In literature, component failure models used in reliability calculations are traditionally based on exponent distribution and failure rates are considered as constants [1, 2, 3]. However, the constant failure rate is inadequate approach if effects of the component type and surroundings are wanted to be analysed. In some cases, Monte Carlo simulation is utilised to estimate component failure rates with exponent distribution e.g. to take into account effects of the surroundings [4,5]. In literature, there are also some enhanced component failure models developed based on constant component failure rates, e.g. to take into account the effects of the weather with separate weather dependent component failure rates [6]. Enhanced methods can be utilised when evaluating environmental and Ccrit Mcrit Ecrit

2 component related aspects in reliability analysis. Another common component modelling approach is to utilise several different reliability functions (e.g. Weibull and/or exponent distribution functions) to model age dependent component failure rate as a bathtub curve [1]. In a case of distribution network modelling, the necessary aging information of the components is not extensively available from existing failure statistics. Therefore the approach is not commonly used in component modelling of the reliability analysis. Component failure modelling can be done also with proportional hazard method. In the proportional hazard method, e.g. following equation can be utilised: h ( x1 β 1 + x 2 β 2... x n ) ( t ) h0 ( t ) exp + + n β = x (1) where t is functional time, h(t x) is probability to fail in short time interval (t+ t), h 0 (t) is average failure rate, x 1,x 2 and x n are background variables like effects of surroundings and β 1,β 2 and β 3 are weight factors for background variables. Proportional hazard method can take into account age and various additional information (like weather and the information about the surroundings of the component) in component failure modelling. Proportional hazard models have not commonly been used in component failure modelling in a field of electricity distribution. The reason could be that the models require lots of data to find essential dependencies affecting to component probability to fail. Proportional hazard modelling is studied generally e.g. in a field of system analysis [7]. Sometimes Markov Models are also used in distribution network modelling. In Markov models, the component failure modelling is done by estimating the effects of the component faults for the system [8]. In large systems like distribution network the use of Markov models is possible as depicted in [9] but usually complex system models are needed because there is huge amount of possible transitions between system states. In the method, separate failure rates are needed for each component e.g. for different weather conditions. Conclusions from review The literature review revealed that studies in the component failure modelling of the distribution network is concentrated on using constant component failure rates. In literature it is also presented improved component failure modelling techniques, which can take into account various factors affecting to the probability of failures, like the weather effects, aging and maintenance time period, etc. Anyway those models are mainly useful to model the effect of a one factor at a time or the modelling needs much more information than is available. However, if the necessary component related data is available, all the studied advanced component failure modelling methods or their combination can be utilised for enhanced network reliability calculations. Advanced component failure model Any of the models presented in literature were not straightforwardly applicable for the aims of the LuoVa-project. Main demand for the method was to have estimates of failure rates which take into consideration main stress factors affecting the failure rate. demand was the possibility to have first estimates from incomplete data and update the values when more and better data is available. Distribution network consist of several components whose failure rate is dependent on different factors. Therefore components of a network must be modelled separately. Because developed component modelling will be used with reliability analysis mainly to give guidelines for strategic decisions, each pole is not needed to be modelled individually. Therefore components are modelled as certain entities in the network. In this study distribution network has been divided into five main components; aerial lines, CC-lines, cables, transformers and switches. For each component it has been determined the main reasons for permanent faults and for auto reclosings (see first column of table 1). Separate failure rates for each component type are formed straight from main failure reasons. E.g. for transformer the overall failure rate for permanent faults is a sum of separate failure rates due to lightning and animals and due to other fault causes. For all defined reasons, it has been determined the main stress factors which affect the failure rate (see columns 2 and 3 in table 1). For example the separate permanent failure rate due to wind and/or snow for aerial lines is dependent on the location of the line (forest, road, field). Table 1: Modelled parameters separated based on failure causes Component: permanent fault /auto reclosing reason Stress factors of the permanent fault Stress factors of the auto reclosing events Aerial line Wind/Snow Location (forest, Neutral earthing, road, field) location (forest, road, field) Lightning Feeder over voltage type ratio Condition, selected Condition line type CC-line Wind/Snow Location (forest, Location (forest, road,field) road,field) Condition Condition Cable Structure Cable type Environment Environment (vulnerability for digging accidents) Condition, number of joints Transformer Lightning Over voltage protection type and location Animal Animal protection, complexity of the connection Switch Condition, complexity of the connection, loading level Animal protection, condition Animal protection, neutral earthing, complexity of the connection Animal condition protection, All the stress factors are classified into appropriate classes. For example classes of location can be forest, near to road and

3 field. For all classes a weight has been defined. Weights represent the effect of certain class to failure rate. Total failure rate for permanent and temporary faults can be calculated with equation 2: λ TOT... + w = w 1_ n 1_1 n _ n n _ 1 λ λ + w n 1 1_ 2 n _ 2 λ + where λ 1,λ 2,λ n, are partial failure rates of the component,w 1-i, w 2-i,... are weights for i:th class of partial failure rate and λ TOT is total failure rate of a component. Practical approach in component modelling includes the idea that parameters used in failure rate modelling should be possible to be affected with selected planning strategies or component selection. For this reason effects of the weather are not included as an own parameter to the failure rate modelling even if the effects of the certain natural phenomenon (storm, etc.) are extensive when happening. The effects of the weather are included in condition information of the component (affect to the stress tolerance). Also, the effects of the aging have been studied and it seemed to have an influence on the failure rate, but in more particular analysis the increase in failure rate in older structures can usually be explained with other component related factors (e.g. planning strategies at construction time and maintenance work done). Therefore, age factor is included in condition weight information. However, if the correlation between age and failure rate can be proven later on, the age can be included component failure models e.g. as independent stress factor. Voltage dip analysis From the customer point of view the harm caused by a single deep dip is nearly the same as caused by a short interruption due to fast reclosing but the number of dips is typically much greater than the number of interruptions. Thus, also voltage dip analysis is modelled as part of the developed entity. Voltage dip rates for each component are defined based on the permanent and temporary short circuit failures. Defined component related dip rates are used to define number and depth of dips in existing network. Component failure rate modelling (equation 2) can not be used in voltage dip analysis because it contains both earth faults and short circuit faults. When analyzing isolated or coil earthed distribution network, earth faults can be neglected because they don t cause voltage dips that can be seen on customer side of distribution transformer [10]. Adding information of total short circuit ratio to every separate failure rate (as depicted in equation 2), dip rate of modelled component can be calculated as follows: DIP ( λ1 w1 e1 + λ2 w2 e2 + + n wn en ) (3) λ =... λ where λ i is the separate failure rate, w i is the weight for separate failure rate λ i and e n is ratio of short circuit to total amount of failures. It s important to notice that both permanent faults and faults cleared with auto-reclosing are included to equation. 2 (2) Statistics for component failure modelling Before utilizing developed failure rate modelling methods for network reliability calculations, failure rate parameters has to be determined. The statistics for determination has been collected from Finnish network companies. Used statistics on population and outages are shown in Table 2. On temporary faults data has been collected only from outgoing feeder, time and fault type (fast or delayed auto reclosing event). Population Fault Table 2: Collected statistics of population and faults General Aerial line Cable Transformer Isolated or compensated Length, Length, Rated power, Age neutral Envi- Protection earthing, ron- against over Number ment voltage and of disconnectors, animals, Earthing type Number of joints, Major changes Feeder, Time, Reason, Duration, Weather, Fault type Age Line joints Protection against over voltage and animals, Rated power, Avg. power, Location, (indoors, outdoors), Network connection, Shielding Location The analysed data consists of about 2400 faults; over 60 % of those were faults of aerial lines. The population covers about km of cables and aerial lines together and about transformers for several years time period. Weights and failure rates depicted in Table 1 were calculated by utilising analysed fault information and engineering judgement. General failure rate of components were calculated as a weighted mean from failure rates of separate companies. The following equation is used: λa = t n i fa / y i i t ti where t i is population of the company i, t is total population, n fa is the number of faults in the company i for the component a, y i is the number of years the company i has derived data and λ a is failure rate of the component a. To get weights for equation 2 the idea is to calculate separate failure rates for defined parameter groups (e.g. type or location groups) by using equation 3 and then weights are determined to be equivalent to those results. More information (3)

4 about statistical analysis can be found e.g. in [11]. RADIAL NETWORK ANALYSIS Basic input data for the reliability analysis is the component information (type, failure rate, etc.) and the network topology. However, also some other information is needed from the network which are affecting to the results of the analysis, e.g. repair times and automation devices installed. In the enhanced radial reliability analysis, network is analyzed feeder by feeder and zone by zone as traditionally. A zone refers to a part of feeder, which can be isolated by one or more switches from the rest of the feeder. In the enhanced reliability analysis, expected amount of permanent and temporary failures and voltage dips in a zone is calculated as a sum of the individual network component failures/dips. Determination of repair time in permanent fault case is done by analysing the possibilities to isolate load points from the faulted component and then restore the load points with disconnectors. Following outage times are used depending on the used switch types or faulted component type (example numerical value of given outage time is presented in parenthesis): - remote fault isolation restoration (10 min) - manual fault isolation, remote restoration (45 min) - manual restoration (60 min) - common repair time (120 min) - transformer change time (150 min) - cable replacement with temporary backup cable or aggregate (240 min) In a case of temporary faults, the whole feeder is experiencing the same short interruption. All three phase faults cause also voltage dips of some level for other feeders supplied by same power transformer. In enhanced reliability analysis, experienced permanent and temporary faults and voltage dips are defined for each load point. More information about enhanced reliability analysis is found e.g. in [12,13]. COST MODELLING Defined cost information is based on the total interruption times in a certain area and permanent and temporary fault and voltage dip occurrences defined with the radial network reliability analysis. Outage costs are defined in the analysis as the inconvenience both to utility and to customers. The evaluation of the utility outage costs is based on the value of nondistributed energy and fault repair costs. Besides nondistributed energy, other costs like losses in production are taken into account in defining inconvenience costs for customer point of view. In Finland some studies have been carried out to determine inconvenience value for different customer groups. In used cost model, the customer dependent cost parameters depicted in [14] are taken into use. The expected permanent outage annual costs (C) caused by a fault in the zone under study is defined by using the equation 4: C = λ ( a + b t ) P n j J i I zone i i j ij However, the results are only as accurate as used input parameters. Even if the failure rate and reliability analysis paij (4) where J is set of load points to which fault in zone causes interruption, I is set of customer groups, λ ZONE is sum of the individual network component failures per year in the zone, a i is constant outage cost parameter of customer group i [ /kw], b i is time dependent outage cost parameter of customer group i [ /kwh], t j is expected outage duration of load point j [h], n ij is number of customers of group i in load point j, P ij is power of average customer of group i in load point j [kw]. The expected outage duration (t j ) depends on the relation of fault studied and load point. Similar equation and parameters are used to define costs caused by temporary faults and voltage dips. More information from customer inconvenience modelling is given e.g. in [15]. SOFTWARE IMPLEMENTATION Depicted reliability based network analysis is quite comprehensive and it requires quite a lot implementing if the application development has to be started with a clean sheet. On that account the developed reliability based network analysis has been integrated with existing systems. The foundation of the developed reliability based network analysis is provided by NIS. NIS is used because it offers easy interface for network component information and topology. NIS also contains interface for Customer Information System (CIS) and in this way makes possible to access customer structure information. If reliability based network analysis will be developed further, also SCADA can be utilised to provide operational data as an input for the reliability calculation. More information about software implementation is depicted e.g. in [13]. System studies Basic function of the developed tool is to analyze existing network for one-year period to find the most important components from the network. Based on the achieved network reliability information e.g. near future maintenance and/or renovation decisions can be done. Another key functionality of the application is to study alternative network development strategies by running the analysis in planning mode. In this type strategic planning the study period may be for example 20-years. This type of studies has been made and are presented e.g. in [16, 17] DISCUSSION The developed reliability based network analysis tool can be utilised for advanced network planning processes and the modelling results have been promising. Main goals were also achieved. Model is quite simple even thought it contains large number of factors. The results can be represented as a number of failures per year which is useful and easily interpreted value. Failure rate and weights are easy to calculate and update with latest data. First estimates for partial failure rates can be made without stress factors if needed data is not available, then weights are simply 1 for all factors.

5 rameters have been adjusted with statistics and best available knowledge, further information is still needed for more accurate failure rate modelling. Network utilities can adjust the parameters with their own network information but also more extensive failure databases are needed. Fortunately, there are several projects going on in Nordic countries to define necessary information from the component failures, so later on the existing parameters and developed component failure rate model can be accommodated more precise with reality. REFERENCES [1] A. Høyland, M. Rausand, 1994, System reliability theory: models and statistical methods. John Wiley and Sons, New York. 518 s. [2] D.O. Koval, H.K. Kua, J.P. Ratusz, 1988, Power system network reliability analysis based on Weibull analysis. International Reliability, Availability, Maintainability Conference for Electric Power Industry, p [3] W. Li, 2004, Evaluating mean life of power system equipment with limited end-of-life failure data, Power Systems, IEEE Transactions, Vol. 19, pp [4] R. Billington, A. Sankarakrishnan, 1994, A comparison of Monte Carlo simulation techniques for composite power system reliability assessment, Communications, Power, and Computing. Conference Proceedings. IEEE, Volume: 1, pp [5] R. Billinton, W. Peng, 1999, Teaching distribution system reliability evaluation using Monte Carlo simulation, Power Systems, IEEE Transactions, Vol14, pp [6] M. Bollen, 2001, Effects of adverse weather and aging on power system reliability, Industry Applications, IEEE Transactions, Vol. 37, pp [7] D.W. Coit, J.R. English, 1999, System reliability Modeling Considering the Dependence of Component Environmental Influences. Proceedings Annual Reliability and Maintainability Symposium, pp [8] R.E. Brown, S. Gupta, R.D Christie, S.S. Venkata, R. Fletcher, 1999, Distribution system reliability assessment using hierarchical Markov modeling, Power Delivery, IEEE Transactions, Vol. 11 pp [9] R. Billinton, C. Wu, 2001, Predictive reliability assessment of distribution systems including extreme adverse weather, Canadian Conference on Electrical and Computer Engineering, Vol 2. s [10] M. Bollen, Understanding Power Quality Problems Voltage Sags and Interruptions, vol. 1. New York: IEEE Press, 2000, 543 p [11] J. Pylvänäinen, J. Järvinen, P. Verho S. Kunttu, J. Sarsama, 2004, Advanced Relibility Analysis for Distribution Network, IEEE International Conference on Electric Utility, DPRT, Hong Kong, p 6. [12] J. Järvinen, J. Pylvänäinen, P. Verho, 2004, Voltage Dip Analysis as Part of Distribution Network Planning Software. Nordic Distribution and Asset Management Conference 2004, Finland, 11 p. [13] P. Verho, P. Järventausta, A. Mäkinen, K. Nousiainen, M. Pouttu, P. Juuti, 2003, Information system solution for reliability based analysis and development of distribution networks" in Proc. 2003, 17 th International Conference on Electricity Distribution, CIRED, 5 p. [14] B. Lemström, M. Lehtonen, 1994, Kostnader för elavbrott. Nordiska ministerrådets, serie TemaNord, 1994:627, 165 p. (in Swedish). [15] A. Mäkinen, J. Partanen J., E. Lakervi,1990, A Practical Approach for Estimating Future Outage Costs in Power Distribution Networks. IEEE Transactions on Power Delivery, Vol. 5, pp [16] J. Pylvänäinen, J. Järvinen, M. Oravasaari, P. Verho, 2005, Software tool for reliability based distribution network analysis, Proposed to Reliability of Transmission and Distribution Networks, RTDN2005, UK, 5 pp. [17] M. Oravasaari, J. Pylvänäinen, P. Verho, 2005, Evaluation of Alternative MV Distribution Network Development Plans from Reliability Point of View, CIRED 2005, Italy. 5 p.