Impact of Voltage dips monitored in the MV networks on aggregated customers

.


Introduction
Voltage dip is an important power quality (PQ) issue that can be caused by short-circuit faults or connection of heavy loads and/or power transformers in the electricity network.As defined in several standards [1][2][3], a voltage dip is the temporary reduction in the RMS voltage below a specified threshold followed by its quick recovery.Because of its propagation from one part of a network to another, the frequency of occurrence of voltage dips and its economic impact on a wide range of industrial customers is larger than other PQ disturbances [4][5][6].For these reasons, there has been a noticeable increase of PQ monitoring in the power systems in recent years in which detection, classification and characterization of voltage dips are crucial requirements for quantifying the monitored data.
Most PQ standards consider the lowest-magnitude and totalduration of voltage dips when evaluating the frequency of occurrence over a certain period of time and presenting statistical measurement results for regulatory purposes.According to [7], the standards can characterize single-phase and three-phase balanced voltage dips effectively but fail to characterize unbalanced voltage dips without loss of information.Considering the types of faults, types of winding connections of transformers between load and fault points, complex phase-voltages are considered in [6,8] to classify voltage dips caused by short-circuit faults into seven (A-G) types.Although this method is suitable for describing the propagation of voltage dips through transformers and for equipment testing against voltage dips, it is not recommended for the characterization of voltage dips from recorded waveforms and it is inapplicable to distinguish the type of unbalanced dips due to large dynamic loads [9].Based on symmetrical component of voltages, another method of characterization and classification of voltage dips is also introduced in [8].This method is highly recommended for characterizing voltage dips from recorded waveforms [9] and also suitable for determining the fault location [10].Based on threephase voltage magnitudes and their phase-angles, a method that is less restricted to large phase-angle jumps or low phase voltage drops than the symmetrical component and the six-phase algorithms is presented in [11].
The performance of an industrial process, which is usually composed of several equipment, to voltage dips depends on the interaction and sensitivity of individual components in the process [4,12].Several studies [13][14][15][16][17][18] have addressed that the magnitude and the duration of voltage dips are not the only parameters that affect the immunity of devices.The behavior of AC contactors [13] can be significantly affected by the point-on-wave dip initiation and phase-shift while the behavior of adjustable speed drives [14,15] is highly influenced by the type of dip and loading conditions, but not by the point-on-wave and phase-shift during the dip.It is reported in [16][17][18] that the drop in speed of induction machines also depends on the starting time of the voltage dip, type of the dip and loading condition of the machine.Considering point-on-wave and phase-angle as additional parameters, advanced methods of voltage dips characterization are discussed in [19,20], and different methods of calculating the phase-angle jump for measured voltage dips in a three-phase system are proposed and compared in [21].Although it will complicate the reporting of the quality of supply, an event can be represented very accurately when a higher number of characteristics are considered.
By comparing the sensitivity of process equipment to voltage dips with the performance of the supply system, a compatibility analysis is generally required to evaluate the effects of voltage dips on an industrial plant [22].The performance of the supply system is often represented by the magnitude and duration of voltage dips whereas the equipment sensitivity is represented by the voltage-tolerance curves.Industrial standards, like SEMI F47 [23] and ITIC [24], and international standards IEC 61000-4-11/34 [25,26] are available providing guidelines to manufacturers and users for developing products with specific requirements, and for performing and evaluating product compliance tests.For manufacturers, these guidelines are not mandatory requirements and a wide range of equipment exhibiting different sensitivity against voltage dips are used in industries [13,14,22,27,28].Moreover, different brands of the same equipment type, and even different models of the same equipment brand can have different sensitivity against voltage dips.In order to simplify the selection and ordering process of equipment, and to ease the performance tests on limited number of test points, five immunity classes are proposed in [29] for equipment against balanced and unbalanced voltage dips.
Monitoring voltage dips in the MV distribution systems continuously can provide network operators and customers with sufficient information about the quality of supply voltage.A wide range of customer processes, each composed of different devices, are connected to the network and the effect of voltage dips on the combined customers is more complicated than what is explained in the standards.Despite all efforts on voltage dips characterization, and recommendations for testing and classifying the equipment dip immunity, the relationship between voltage dip characteristics and its impact on combined customers is lacking in the previous researches and existing standards.This paper presents a methodology for estimating the impact of voltage dips on aggregated customers and describes the correlation between the parameters for various types of voltage dips with their impact on aggregated customers.The proposed approach is also applied to obtain system average dip severity indices (SADSI) that can be used for estimating the economic impact of voltage dips and for setting voltage dip limits for regulatory purposes.

Methodology
In this paper, a dataset of voltage events continuously monitored for four years and collected from six substations is used to assess the frequency, type and severity of voltage dips.A new method of evaluating the voltage dips impact on aggregated customers is presented and the correlation between the voltage dips severity and their impact on customers is described.

Network description and monitoring data
The Dutch MV-networks use underground cables mostly operating at 10 kV.Depending on the size of the substation, the MV-networks considered in this paper consist of between 11 and 32 radially-operating feeders connected to the primary substation.A schematic of a generic network, with the types of measured-data that are used for the study, is shown in Fig. 1.The feeders are numbered as F 1 , F 2 , . .., F n and each feeder consists of a primary protection (CB 1 ), and a secondary protection (CB 2 , if a feeder has branches) to clear faults depending on the fault location.In this paper, customers of each feeder are aggregated into one big-customer connected to the point of common coupling (PCC) of the primary substation.Based on the type of HV/MV transformers, the six substations are classified into two categories.As can be referred to Table 1, category-I includes four 10 kV networks which have Yd types of transformers in the primary substation.With such types of transformers, the MV-networks are isolated and the zero  With regard to the monitoring tools installed in the primary substations, two groups of data recorded by the monitors called SASensors are used in the analysis.The first part includes digitalized waveforms of phase-to-ground voltages sampled at a sampling frequency of 4 kHz.The other group of data includes half-cycle RMS values of per-phase power for each feeder connected to the main substation.

Voltage dip impact evaluation
In order to estimate the impact of voltage dips on all customers connected to the MV-network, an approach based on the amount of change in load/generation is proposed in this work.The approach was introduced in [30] but several important points which were not considered in the previous paper are addressed here in detail.This includes-estimating the loss of power starting at a feeder level, finding the relative origin of the dip to distinguish the change of power due to interruption or due to connection of additional loads from the losses related to a voltage dip impact, treating the loss for combined loads and DGs per feeder separately, analyzing more data from several substations to study the effect of size and types of customers as well as the capacity of HV/MV substations during the occurrence times of events for correlating voltage dip parameters with their impact on customers based on the absolute or relative loss of power, and proposing system average dip severity indices for estimating the economic impact of voltage dips relative to a process interruption.
The flowchart in Fig. 2 gives an overview of procedures for evaluating the impact of voltage dips on aggregated customers.For each voltage event monitored at the PCC of a substation that consists of n feeders, the evaluation procedure includes: (i) characterization of the voltage event (set of voltage waveforms with high time resolution) and checking if it qualifies for a voltage dip; (ii) evaluation of active-power for aggregated customers in each feeder before, during and after the event when condition (i) is satisfied; (iii) checking if the relative origin of the dip is in the upstream of the feeder under consideration; (iv) estimating the loss of power for aggregated customers in each feeder that satisfies the condition in (iii) and categorizing it as absolute loss of loads or DGs; (v) evaluating site loss of loads and DGs by combining the losses in all feeders calculated in step (iv); and (vi) estimating the relative loss of power for customers in the entire substation.

Voltage dip severity assessment
As described earlier, the waveforms of phase-to-ground voltages are monitored at the PCC of the MV-networks while end-users are mostly connected to the points of connections (POCs) at the busbars of the MV-networks through Dy n transformers.The phase-phase voltages (p.u.) propagate into phase-to-neutral voltages (p.u.) in the LV-networks and end-users are essentially affected by the phase-phase dips in the MV-networks [31,32].For this reason, phase-dips and phase-phase dips associated with phaseto-ground voltages and phase-phase voltages in the MV-networks are analyzed and treated separately in this work.
In order to detect and characterize voltage dips, samples of the three phase-phase voltages are obtained from waveforms of three phase-ground voltages in the MV-network.As recommended by Fig. 2. Flowchart of a voltage dip impact evaluation procedure.
the IEC 61000-4-30 standard [33], the method of half-cycle sliding window is used to calculate the characteristics RMS voltages as a function of time of each event from the recorded waveforms.From the characteristic phase-ground and phase-phase voltages, the magnitude, duration and type of voltage dips with residual voltage magnitudes ranging from 1% to 90% of nominal voltage lasting between half-cycle to 1 min [3] are evaluated at the PCCs.For dip events that involved multi-phases, concept of phase and time aggregation, which is based on the lowest-magnitude and totalduration method, is applied to prevent voltage dips due to the same root cause from being counted several times.The same aggregation method is also used with multiple-dip events in order to consider the cumulative effect of individual dips on several devices of the aggregated customers in the entire MV-network.Using the procedure recommended by the IEEE Std 1564 [34], single event indices are aggregated to constitute the annual site indices at the PCC of each substation.

Relative origin of voltage dip event
Before and after the occurrence a dip event, a feeder may have large-positive, very small (negligible) or large-negative values of power.In this paper, the feeder power is considered positive if the feeder draws power from the substation, and negative when the feeder generates power towards the main busbar.With regard to the pre-dip and post-dip powers of the combined customers in each feeder, the different situations associated with the status of the feeder during a dip event at the PCC are categorized as: • interruption of customers when the post-dip power of a feeder is reduced due to a fault in the same feeder; • partial loss of loads when a feeder shows less load-power or more generation-power after the dip than before the dip event and the disturbance is not in the same feeder; • partial loss of distributed generations (DGs) if the post-dip power of a feeder shows an increase in load-power or a decrease in generation-power relative to the pre-dip power, and the dip is not caused by the connection of additional loads in the same feeder; • more loads (less DGs) when the post-dip power of a feeder is larger than the pre-dip power and the dip is due to the switching of additional loads in the same feeder; or • no change in power when the power for aggregated customers of a feeder is not affected by the dip event.
For aggregated customers of a feeder, a voltage dip that is monitored at the PCC of an MV-network may originate in downstream (within the same feeder) or upstream (in the HV-network or other neighboring feeders).Using the power characteristics of aggregated customers, the Upstream/Downstream criteria [35] is used to detect if the dip event is originating in the upstream or downstream of the feeder.According to this criteria, the dip originates in the downstream of the feeder under consideration when the duringdip power is much higher than the pre-dip power.In this work, the difference between the pre-dip and post-dip powers of each feeder is considered for estimating the impact of voltage dips on the customers only if the dip originates in the upstream of the feeder under consideration.

Estimating losses due to voltage dips
The proposed method of assessing the impact of voltage dips makes use of active-powers of aggregated customers in each feeder, connected to the main substation, measured before and after the dip events.The estimation procedure follows two steps-estimating the losses in each feeder of a substation and then evaluating the losses in the entire substation.
In the first step, the amount of disconnected loads or DGs in a feeder f due to a dip event k is estimated from the change of active-power ( P) using (1), where P f,pre and P f,post are the pre-dip and post-dip powers of the feeder following a voltage dip event.In the analysis, the change of power is not considered for estimating the impact of voltage dips on customers if the dip originates in downstream of the same feeder while the changes of powers in the other feeders are considered as the loss of loads or loss of DGs.Fig. 3 illustrates the six possible situations of a feeder which may be considered for the loss of loads and the loss of generations during the analysis.Pre-dip and postdip powers having the same direction of flow leading to the loss of loads are indicated by 'a 1 ' and 'a 3 ' while 'a 2 ' involves a change in direction from positive to negative.In a similar fashion, the postdip power has the same direction as the pre-dip power in 'b 1 ' and 'b 3 ' when evaluating the loss of DGs whereas the direction reverses from negative to positive in case of 'b 2 '.It is possible that a feeder may include a mixture of loads and DGs but it is difficult to know the exact amount of loads and DGs from the available data.The second step deals with evaluating the amount of disconnected customers in the entire substation.After identifying the feeders affected by the dip originating in the upstream, the loss of loads and DGs of the substation only due to voltage dips are treated separately.For a substation that consists of n feeders connected to the PCC of an MV-network, the absolute impact of each voltage dip event k in the loss of loads and DGs in the entire substation is calculated using (2), where P f and P sub refer to the change of power in each feeder (f) and in the whole substation (sub) during a voltage dip event (k).The absolute loss of power depends on the size and type of customers as well as the occurrence time of faults, and the loss with similar dips may significantly vary from place to place and time to time.In this paper, the severity of voltage dips are compared with the relative loss of loads and DGs separately obtained using (3).

Evaluation of voltage dip severity indices
The main concern of voltage dips is its high economic loss for industrial and big commercial customers, and direct evaluation of the economic impact of voltage dips is often difficult.Here, the proposed approach of estimating the impact of voltage dips on aggregated customers is extended to obtain the system average dip severity indices (SADSI).The SADSI can help customers or DSOs to estimate the anticipated annual economic losses caused by the expected number of voltage dips.Employing the loss of loads and loss of DGs evaluated by (2), the percentage SADSI of dips in each cell of the voltage dip profile with various magnitude and duration ranges can be obtained from the weighted average of the relative losses.This can be estimated using (4), Nr,c i=1 ( P sub i ,load + P sub i ,gen where N r,c denotes the total number of voltage dips for the cell with residual voltage magnitude in the r's row and duration in the c's column, and P sub,pre represents the sum of absolute pre-dip power of all feeders for each of the N r,c events.Applying (4) for the monitored dips, the percentage SADSI densities can be evaluated for various types of voltage dips whose frequencies of occurrence are represented by the voltage dip tables in the standard EN 50160.By combining the annual voltage dip densities with the respective SADSI densities, the annual economic loss of voltage dips can be evaluated relative to the cost of a process interruption.

Result analysis
Using the proposed method, the impact of various types voltage dips on group of customers and the relationship with the dip characteristic are described in this section.

Voltage dip profiles
From a total of 2226 voltage events collected from six substations over four years, only 176 of the events are found to be voltage dips.Fig. 4(a) shows the profile of total phase-dips from all monitoring locations.On average, about 7.33 dips per year are monitored in each substation, and majority of the dips (72%) have residual voltage greater than 40% of the nominal voltage while about 75% of the dips have duration t ≤ 1000 ms.It is found that around 62% of the dips are one-phase dips, 24% are two-phase dips and 14% are three-phase dips.Some of the phase-dips have relatively long duration and this is because of the aggregation method used with multi-dip events.
Considering the transfer of voltage dips, the profiles of phasephase dips in the MV-networks propagating to phase-dips in the LV-networks are shown in Fig. 4(b).It can be observed that the frequency of dips at the customer terminals is not only significantly reduced to an average of around 4 dips per year but also most (85%) of the dips have residual voltage above 40% of the nominal voltage while 89% of the dips have duration t ≤ 1000 ms.The reason is that one-phase dips of category-I substations, unlike category-II substations, are not seen as phase-phase dips in the MVnetworks and the magnitude of phase-phase dips due to one-phase and two-phase faults become shallower than phase-dips.Industrial equipment that comply with the SEMI F47 power acceptability curve are immune to 64% of the dips which fall above SEMI F47 curve.Fig. 5 shows the stochastic and random nature of voltage dips both in time and place that is expected depending on the type of networks and various activities.

Impact of voltage dips
In Fig. 6, examples of voltage dips and their absolute impact on aggregated loads and DGs per feeder are shown to get insight into the six situations considered during the analysis.The first three (a-c) are for the loss of loads and the other three (d-e) are accountable for the loss of DGs.During the analysis, it is also noticed that one-phase fault cleared by the secondary protection of a feeder can cause outage of big customers while the feeder is not entirely interrupted by the fault.With category-I substations, one-phase   dips are not experienced by most end-users and Fig. 7(a) addresses that around 95% of the total loss of loads (43.5%) in the substation is from one feeder while others share an insignificant amount.In such substations, one-phase dips have little overall impact on the total customers in the entire network and such outages of customers are not treated here.With category-II substations, however, one-phase dips are transferred and are experienced by other customers in different feeders.Fig. 7(b) demonstrates that the loss of loads due to a similar situation (one-phase fault) and the total loss of loads (26%) is shared among several feeders.Fig. 8 illustrates a voltage dip due to a two-phase fault, characterized by magnitude of 0.28 p.u. and duration of 320 ms, and caused the loss of about 1.56 MW and 7.05 MW of aggregated loads in one feeder and in the entire substation respectively.In the same way, the losses associated with customers in each feeder and in the entire substation are evaluated for each voltage dip at the PCC of the MV-networks.

Voltage dip severity and impact correlation
Due to the fact that most end-users experience voltage dips having the same properties as phase-phase dips at the PCC of the MV-networks, the correlation between severity of voltage dips and their impact on aggregated customers is further studied for the phase-phase dips in the MV-networks.Three dip parameters are used to characterize the severity of phase-phase dips-magnitude, duration and type.As can be seen from Fig. 4(b), cluster of phasephase dips whose magnitudes are varying from shallow to deeper  have certain interval of times that have to do with the protections in the HV-and MV-networks.From this point of view, dips with shortduration ( t ≤ 200 ms), medium-duration (200 < t ≤ 1000 ms) and long-duration ( t > 1000 ms) are grouped while the type of dips are defined as one-line (L 001 ), two-line (L 011 ) and three-line (L 111 ) depending on the number of line-to-line voltages being affected by the dip.

Loss of aggregated loads
Fig. 9 shows the relative loss of aggregated loads due to various types of phase-phase dips in the ZHV1 substation, which have a wide range of dip parameter variations.It can be seen that shortdips, mostly originated from the HV-networks, are experienced by end-users as relatively shallow dips with residual voltage magnitudes u ≥ 0.4 p.u.; whereas medium-and long-dips have retained voltages ranging from deeper to shallow magnitudes.Depending on the magnitude of residual voltages, short-duration dips of L 001 , L 011 and L 111 types resulted in the relative loss of less than 1%, 1-11% and 7-50% of aggregated loads respectively (Fig. 9(a)).Medium-duration dips in three-lines caused to the loss of 26-68% of aggregated loads when the magnitude varies from 0.77 to 0.09 p.u. (Fig. 9(b)); whereas long-duration dips of L 011 and L 111 showed 20-36% and 27-78% loss of aggregated loads (Fig. 9(c)).Hence, for dips within each duration category, the more phase-phase voltages are affected by the dip, the higher the relative loss of loads is.Moreover, the deeper the magnitude of each type of dips, the higher the relative loss of loads and thus the bigger the expected impact on combined customers is.In each substation, it is observed that the relative loss of loads increases when the severity of voltage dips, which depends on the type, magnitude and duration, increases.
The relative loss of loads due to phase-phase dips from the monitors of all substations are shown in Fig. 10.In general, clusters of dips in each time interval show higher loss of loads when more phase-phase voltages are affected by the dips and the impact due to each type of dips worsens when the dips get deeper and longer.The uneven trend in the relative loss of loads of few dips having similar properties might be because of variations in the size and sensitivity of customers, and the occurrence-time of dips in different substations (for instance, voltage dips have higher impact on big customers with more sensitive processes than customers with less sensitive processes; and more customers are likely to be affected during a day-time than during a night-time).Table 2 summarizes the variations in the relative loss of aggregated loads for various types of dips with short-, medium-and long-duration.On average, the relative loss of loads due to shortduration dips of L 001 -L 011 -L 111 dips varied in the range of 2-8-32% respectively, and losses increased with medium-and long-duration dips to 18-23-45% and 28-49% respectively.The 95-percentile is introduced to avoid effects of extreme values to the average, and results for the 95% situations show that the relative loss of combined customers continuously increase with the type and duration of the dips.

Loss of aggregated generations
The fault-ride through capabilities of DG-units of wind turbine or CHP-plant connected to the busbars of the MV-networks depend on the phase-phase voltages at the connection points of DG-units [36].Due to the fact that the dip-profiles are available at the PCC of the substations while most of the DGs are often connected to the other busbars, far away from the PCC, it is hard to exactly know if   the dips disconnect the DGs or not.Fig. 11 shows the relative loss of DGs for different types of dips in the ZHV1 MV-network.Unlike the impact of voltage dips on aggregated loads in the same substation (shown in Fig. 9), no general conclusion can be drawn from Fig. 11 regarding the correlation between the severity of voltage dips and their relative impact on aggregated DGs.At the times of voltage dip occurrences, the amount of power generated from the feeders with net DG-power significantly varies from time to time.Some transformers supply farms with CHP and most have loads whereas moment of just loads and just generations are assumed during the analysis for each feeder connected to the PCC.Hence, voltage dips having similar properties do not always show similar effect on the relative loss of DGs in the MV-networks.

System average dip severity indices
Considering the weighted average of the relative loss of power for aggregated customers in the MV-networks associated with the phase-to-phase dips collected from the six substations, system average dip severity indices (SADSI) for the three types of voltage dips are shown in Table 3.When completing the values of SADSI that correspond to dips in each cell of the EN 50160 voltage dip table, missing values are interpolated.The values in each cell represent the severity of voltage dips relative to that of a complete interruption.For instance, an L 111 dip in the cell X 55 is, on average, as severe as about 0.79 times that of an interruption.An important feature is that the values of SADSI increase when voltage dips get deeper and longer.For voltage dips of the same magnitude and duration, the SADSI values are higher for poly-phase dips.

Discussion on validation
As discussed in the introduction, substantial studies have used simulation and/or experimental approaches to investigate the effect of voltage dips on individual devices.In practice, many types of customer categories with various processes composed of several devices having different sensitivities to voltage dips are connected to the POCs of the MV-networks.The impact of voltage dips on the  customers in the entire network not only depends on the sensitivity of individual devices but also on the composition, interconnection and interaction of equipment and processes in the network.For these reasons, it is difficult to use simulation or experimental approaches for assessing the impact of voltage dips on aggregated customers and validation will be needed by more field measurements.With continuous voltage dips monitoring schemes, the proposed approach considers real measurement data of power in each feeder that comprises various types of end-users' equipment and processes and the approach is very useful in estimating the impact of voltage dips on aggregated customers in the entire network.
Because of the growing interest in environmental issues together with the advancement of technologies to connect renewable sources to the grid and the liberalization the energy market, the share of distributed generation systems is increasing nowadays.When a feeder consists of a mixture of loads and DGs, a voltage dip at the PCC may lead to the loss of loads, loss of DGs or both.The proposed method only considers the change in power of aggregated customers between the pre-dip and post-dip event and it is difficult to determine the exact amount disconnected loads and DGs which might lead to some uncertainty.Such uncertainty may be avoided or reduced if the exact amount of loads and DGs per each feeder are known before and after the dip event.

Conclusion
Voltage dip indices including the frequency, type and severity are assessed at the PCC of the MV distribution networks to evaluate the quality of the networks; and the dips that can be seen at the customer terminals are determined.It is observed that the type and severity of voltage dips at the monitoring points, and the type and number of transformers can highly influence the number and severity of voltage dips propagating to the end-user terminals.
Based on the changes in load or generation, the proposed methodology helps to estimate the absolute and relative loss of power during voltage dips.The approach provides better insight into the impact of various types of voltage dips on aggregated customers (in a substation or a feeder), and assists to correlate the severity of dips with their impact on the customers.It is noticed that the relative loss of aggregated customers highly depends on the magnitude, duration and type of phase-phase voltage dips in the MV-networks.Depending on the severity of dips, on average dips in one phase-phase voltage of the MV-networks resulted in the relative losses ranging from 2% to 18% of aggregated loads while dips in two and three phase-phase voltages caused losses in the range of 8-28% and 32-49% respectively.The method shows a strong correlation between the severity of voltage dips and their impact on aggregated customers.That is, the relative loss of power and thus the impact on combined customers increases with longer and deeper dips; and even gets worse when more phase-phase voltages of the MV-networks are affected by the dips.The estimation method is employed to obtain system average dip severity indices for various types of voltage dips.This can be very helpful for estimating the economic loss of voltage dips and for setting voltage dip limits in the MV distribution networks for regulatory purposes.

Fig. 3 .
Fig. 3. Conditions considered for estimating the loss of (a) loads and (b) generations.

Fig. 4 .
Fig. 4. Scatter plot of voltage dip profiles at the PCC of substations: (a) phase-ground and (b) phase-phase.

Fig. 5 .
Fig. 5. Variation of voltage dip frequency in time and place (Line = phase-phase).

Fig. 7 .
Fig. 7. Share of feeders to losses due to a one-phase dip in substations of: (a) category I and (b) category II.

Fig. 8 .
Fig. 8. Power of aggregated customers in a feeder and a substation during a voltage dip.

Fig. 11 .
Fig. 11.Relative loss of DGs in ZVH1 substation due to different types of phase-phase dips with: (a) short-duration, (b) medium-duration and (c) long-duration.

Table 1
Type and capacity of substations.

Table 2
Effect of dip parameters as compared to their impact on aggregated customers.