A RELIABILITY AND COST ASSESSMENT METHODOLOGY OF MEDIUM VOLTAGE FEEDERS

In this work, we introduce a methodology which quantifies medium voltage (MV) feeder reliability in the form of commonly used indices. The contribution of the proposed methodology focuses on the modelling not only of the grid itself, but also of the detailed actual operational practices of the Distribution System Operator (DSO), taking into consideration the specific parameters of the examined network. By doing so, the influence of diverse reliability enhancement measures on different feeders can be evaluated and assessed. The method is tested using real (MV) feeders, considering as measures the installation of protection units and remote control capabilities to the switches. Furthermore, by considering the cost of protection and remotely controlled units, the proposed methodology performs techno-economic analyses to assess the trade-offs between enhancing grid reliability and increasing costs, serving as a decisionmaking tool for modern DSOs.


INTRODUCTION
Distribution grids are experiencing many changes lately caused by the introduction of renewable energy resources and new types of load in medium voltage (MV) and low voltage (LV) grids.Such changes pose new challenges to distribution system operators (DSOs) which need to consider not only technical grid constraints, but also regulatory aspects.Concerning the latter, in many European countries there is pressure to reduce grid costs, e.g. through incentive-based regulation, without compromising the reliability of supply.Thus, there is an increasing necessity for DSO-tools, capable of quantifying the system's reliability as well as modeling current practices to restore safe grid operation after faults.Reliability assessment of electric power systems has been studied extensively in the literature, e.g.[1][2], using various techniques ranging from Markov modelling [1,3], state enumeration [1], or Monte Carlo simulations [4].Furthermore, although standard power system analysis software, e.g.NEPLAN [5] or SINCAL [6], offer tools for reliability studies, they are inflexible in terms of modeling specific DSO procedures that lead to the calculation of realistic reliability indices.Other recent works, e.g., [7] focus on reliability studies of an urban MV grid, using historical average reliability data.However, especially for DSOs with rural supply areas, considering the specific local parameters is crucial.The contribution of the proposed methodology is twofold.First, by considering the actual operational practices of the DSO and the specific parameters of the examined network, the impact of various reliability enhancement measures on different feeders can be evaluated and the most effective one can be identified based on a newly introduced key number.Second, by performing a technoeconomic analysis, the trade-offs between enhancing grid reliability and increasing costs can be assessed, making the tool suitable to assist DSOs in their decision-making process.

STATE ENUMERATION
In this paper, the state enumeration principle is chosen as the modelling technique of the developed tool, mainly for two reasons.On the one hand, it shows good tractability features with increasing grid size and on the other hand, it offers the desired level of modelling detail with reasonable computational burden for a typical DSO planning department.State enumeration is an analytical simulation technique that models the system as a collection of (semistationary) states.Each state represents a certain contingency or a certain combination of concurrent failures.The impact of each of these states and the triggered system's response can be easily simulated, for example, by a stepwise process.Each failure state S i represents a failure of a component on the evaluated feeder with a given probability of occurrence λ i .For each failure state, both the system's reactions as well as the restoration process are modelled.The resulting duration of the contingency at each node in the grid, as well as the connected number of customers are then used to determine the impact of each failure state on the reliability.Figure 1 shows a schematic overview of the developed tool based on state enumeration, as implemented using the programming language R [8].The proposed methodology considers sustained interruptions of components on real MV grid feeders in order to calculate the reliability of the grid.Starting from input data, describing for example the network's topology or equipment cost, the restoration process is modelled in detail, starting by calculating the actual driving time to all locations of interest, using Google Maps [9].Then, by following a systematic approach to simulate the fault, localize it, isolate it and finally restore the whole grid, realistic reliability indices are determined.Different measures to improve reliability of supply can be evaluated within the proposed methodology, using commonly used reliability indices, such as System Average Interruption Duration Index (SAIDI) and System Average Interruption Frequency Index (SAIFI).

ANALYSIS OF THE RELIABILITY AND COST ASSESSMENT METHODOLOGY
The main benefit of the proposed method lies on the detailed modelling of the DSO's operational practices during the localization of a failure and the restoration of the system taking into consideration specific parameters of the examined network.The simulation tool requires good knowledge of the practices and the investigated network.

Input data
Three types of input data are needed, as seen in Figure 1.These data refer to the detailed modelling of the grid, the utility's operational practices, and finally to the specific hardware and operational costs of the available measures to increase reliability.a. Network Data The first type of input data describes the network topology using the geographical coordinates of all elements, type and lengths of overhead lines and underground cables, protection units' information (e.g.remote control capability) and alternate supply and sectionalizing points.Figure 2 shows how the network data are displayed for a test network, part of a real MV grid, and consisting of overhead lines (shown with straight lines), cables (dashed lines), transformer stations (TS), alternate supply points and switches equipped with remote control or protection.
b. Reliability Data The second data type contains all information related to the reliability calculation.Reliability data can be split into failure and operational data.The first category mainly includes historical failure rates for overhead lines, underground cables, transformer stations on poles and ground.The latter category refers to the DSO-specific procedures once a failure is detected.The timings of all necessary actions from the identification of the fault to the final re-energization of the whole grid, e.g. of the switching actions, are an important part of this section.They define to a great extent, how long it takes to detect a failure and re-energize the customers and are considered operational data.In contrast to other available reliability tools, e.g.[7], the calculation of these timings is not based on average non-availability values of the network elements but on an actual simulation of the restoration process.More specifically, the time needed for the dispatched crew to drive to the switches plays a crucial role in the calculation of the final reliability indices.This parameter is not given as input, but instead, it is calculated using Google Maps [9] for each needed route, in order to represent realistic timings, considering actual traffic at the time of the fault, alternative driving routes, road maintenance works, etc.

Considered Failure States
For the calculations, SAIDI is chosen as the main index to reflect reliability.Thus, in general, all failures that have an impact on SAIDI have to be considered in the calculations.Failures that do not cause sustained interruptions are therefore neglected.Planned outages are not considered because the evaluated measures (remote control and protection devices) do not influence the number or duration of planned outages.

Step-wise simulation of system restoration
The main core of the proposed analytical simulation method is the step-wise simulation of the restoration process according to the utility's operational procedures.First, the reaction of the installed protection devices on the feeder is modelled.Since the considered grid consists of radial or at least radially operated feeders, a simple model for protection systems can be implemented.It is expected, that always the right protection device is triggered first, assuming perfect selectivity of the examined grid.
The general principle of fault localization applied in grids without full coverage of fault detectors is basically a trial and error process based on sectionalizing.The potential network area where the failure might have occurred is sectionalized by opening a switch along the grid and a try is given to re-energize the upstream section by reclosing the tripped circuit breaker.If the circuit breaker trips again, the failure occurred in the grid section upstream of the opened switch, otherwise it is downstream and the customers upstream from the opened switch are temporarily re-energized.The considered area for the fault search can then be reduced accordingly to one section and the same procedure is repeated.Hence, it is straightforward to model the fault localization process in a step-wise fashion.Each restoration step consists either of one set of sectionalizing operations (select and open switch, reclose circuit breaker) or a partial service restoration.Partial service restoration (PSR) refers to the situation where first some customers downstream from a switch recently opened for sectionalizing are resupplied through an alternate supply point, before the fault is finally restored.
For the sectionalizing operations, the algorithm determines the switch that best halves the remaining area where the fault potentially lies, in order to reduce the considered area in the most effective manner.The time needed for the switching action is thereby considered and remotely controlled switches are, therefore, prioritized.At each step there are two alternatives.Either, the search can be continued with the next set of sectionalizing operations or the supply can be restored in a part of the network.The decision between the two alternatives has a significant impact on SAIDI.The proposed methodology compares two metrics to determine if it is more effective to first resupply a part of the customers or continue with the fault search.The two metrics are defined as (2 where i is the step number, n c the number of customers that can be resupplied per step, and t the time needed for the necessary operations.p s is the probability that sectionalizing is successful and, hence, the customers in the upstream section (between the supply point the switch opened for sectionalizing) can be resupplied.The metrics take into account the possible next two steps and consider the number of customers (n c in Eq. ( 1) and ( 2)) that can be resupplied either through PSR or sectionalizingweighted by the probability that a secure resupply will be successfuland the time it takes to resupply them.The option with the higher metric is then executed.Once there are no more switches to be used in the sectionalizing process or the remaining area to consider for fault search is smaller than a predefined threshold, the searching procedure for the exact location is done visually by the field crew, e.g., by driving or walking along an overhead line to identify possible failures.The time needed for the fault localization via visual search is modelled as a linear function of the remaining line length.Figure 3 shows a graphical representation of the stepwise fault localization process for one fault on the example feeder (introduced in Figure 2), where the red areas are de-energized.In step one the yellow area is the one considered for the fault searchthe region between the primary substation (UST) and the next protection devices (CB1 and CB2).The green areas correspond to regions that have been successfully re-energized, either from the UST (darker green) or from an alternate supply point (AS, light green).

Calculation of results
In order to be able to rate different investigated topologies not only for the same feeder but for all feeders in the grid in a simple manner, a new key number is introduced.The key number is calculated as follows: where LCC corresponds to the relevant life cycle costs of the investigated (top) and reference (ref) feeders, and  to the reliability indices of the compared configurations.
This key number specifies by how many minutes/year SAIDI can be improved due to the investigated topology, assuming a reference value   of additional life cycle cost.The selected reference cost value is 100'000 Swiss Francs.The introduced key number is, hence, applicable also if the analysed feeders differ or different measures are to be examined since it compares the reliability improvement and additional cost to a given reference.

Analysis of one medium voltage feeder
First, the impact of an increased number of protection devices and remotely controlled switches on total costs and reliability of one MV feeder is investigated.Therefore, nine different protection and remote control concepts have been developed.Topologies T1 to T3 have an increasing number of switches that are each equipped with protection and remote control capabilities.In topologies T4 to T6 the number of remotely controlled switches is increased without installing any additional protection devices.Topologies T7 to T9 consist of combinations of topologies T1 to T6. Additionally to these nine topologies, a basic topology that has no protection relays or remotely controlled switches except for the switch of the primary substation is used as a reference.
The results for the key number 'k' as defined in Eq. ( 3) relative to the topology with the highest results are shown in Figure 4. Topology T1 achieves the best result in terms of the key number and is therefore chosen as the reference in this plot.The results show two main findings.First, it can be concluded that installing protection relays at the right places leads to a more cost efficient SAIDI reduction than adding only remote control capabilities.Secondly, it is observed that when the topologies are compared to the basic topology, it is the topology with the fewest pieces of equipment that achieves the highest cost efficient key number.As expected, the devices that are installed first have the largest impact on SAIDI and lower costs than the topologies with more devices already installed beforehand.

Analysis of different feeder types
In order to make general statements, the impact of an increased number of protection devices and an increased number of remotely controlled switches on total costs and reliability is investigated for different MV feeders.Nine real MV feeders are selected for this case study.They can be divided into three different types that vary mainly in terms of total length and cabling rate.The main characteristics of the three feeder types are summarized in Table 1.For each feeder, nine concepts are created and compared to a basic topology analogous to the first case study.For all topologies the key number, as defined in Eq. ( 3) is calculated in reference to the basic topology.a.Comparison to a basic reference topology First, only the topology with the highest key number 'k' from each feeder is considered to analyze the maximum improvement relative to the needed investment cost for each feeder type.Figure 5 shows the achieved key number results for the feeder types, as well as the SAIDI improvement potential relative to the total SAIDI of the complete network without any protection or remotely controlled switches and the necessary investment costs relative to the total switchgear costs in the complete network with many components.As can be observed, long feeders have the highest improvement potential as indicated by the high k values, assuming that they have no protection devices or remote control installed initially.Long feeders also have the largest influence on SAIDI, and hence, the highest potential for SAIDI improvement.However, since the measures on long feeders also contain more devices, the investment costs are increased more than for measures on shorter feeders.

b. Comparison to a today's topology
This part aims at identifying also the potential for SAIDI reduction compared to today's grid.In this analysis, only the topologies with a lower SAIDI value than today's topology were considered for each feeder.Among the examined topologies, the ones with the highest key number are evaluated.Figure 6 shows the results relative to the total SAIDI and switchgear cost of today's complete network.As can be observed, medium feeders in terms of length have the highest SAIDI improvement potential.This makes sense since previous measures were mostly focused on the long feeders which have the largest influence on the reliability of supply.In today's grid, these long feeders are already quite well equipped and further measures on these feeders have high costs with a relative lower impact.Medium length feeders, however, have not been the main point of attention so far and, therefore, have the largest potential to improve reliability by reducing SAIDI.

CONCLUSION
In this paper, a systematic methodology is presented, to assess the trade-offs between reliability and cost of different configurations of medium voltage (MV) feeders.The proposed tool simulates the realistic operational responses to failures within radially operated MV feeders, considering short circuit failures on overhead lines, underground cables and transformer station bus bars that lead to a sustained interruption.The developed tool allows comparing different feeder configurations in terms of costs and reliability and can be used in the planning phase by modern Distribution System Operators.Using today's topology as a reference, results show that it is most efficient to focus on reliability enhancement measures of medium length feeders.However, in a large network consisting of many components, measures on only one feeder do not have much of an influence on the total reliability and cost values, no matter the feeder type.The evaluated measures consider additional installation of protection devices or remote control.Future research will focus on developing a tractable optimization framework to derive the optimal configuration of a given set of reliability improvement measure combinations.Such measures could include for example replacing overhead lines with cables, increasing the resupply alternatives with new lines or also optimizing the resupply process by increasing for example field crew on call.

Figure 1 .
Figure 1.General overview of the reliability analysis algorithm

Figure 2 .
Figure 2. Schematic of the example network c.Cost Data Finally, the last input data type corresponds to cost data.The considered data consist of actual switchgear cost, such as switches, protection devices and remote control features, as well as average maintenance and operational

Figure 3 .
Figure 3. Stepwise simulation of the operational steps to simulate BKW`s restoration process after a fault on the example feeder.

Figure 4 :
Figure 4: Key number results for the example feeder

Figure 5 .
Figure 5. SAIDI improvement potential relative to the basic topology and necessary investment costs for each feeder type.

Figure 6 .
Figure 6.SAIDI improvement potential relative to today's topology and necessary investment costs for each feeder type.

Table 1 .
Characteristics of the investigated feeder types