Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD FOR DETECTING LOW VOLTAGE CONNECTIVITY IN PART OF AN ELECTRICITY GRID
Document Type and Number:
WIPO Patent Application WO/2012/113936
Kind Code:
A1
Abstract:
Method for detecting low voltage connectivity in part of an electricity grid Method for detecting low voltage connectivity in part of an electricity grid (1), the electricity grid (1) comprising a distribution substation (2) being electrically connected to at least one connection by a feeder (3), characterized in that the method comprises the steps of selecting a number of possible connections (6, 7) connected to the feeder (3), statistically determining, based on the retrieved energy records the second smart meters (5, 8), the probable contribution of the selected connections to the energy recorded by a first smart meter (4) and based on the probable contribution of the selected connections (6, 7) to the energy recorded by the first smart meter (4) determining the probability that each of the selected connections (6, 7) is connected to the feeder (3).

Inventors:
PYCKE TOM (BE)
Application Number:
PCT/EP2012/053219
Publication Date:
August 30, 2012
Filing Date:
February 24, 2012
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
EANDIS (BE)
PYCKE TOM (BE)
International Classes:
G01D4/00
Domestic Patent References:
WO2009061291A12009-05-14
Foreign References:
US20070005277A12007-01-04
Other References:
None
Attorney, Agent or Firm:
DUYVER, Jurgen et al. (Holidaystraat 5, Diegem, BE)
Download PDF:
Claims:
CLAIMS:

Method for detecting low voltage connectivity in part of an electricity grid, the electricity grid (1) comprising a distribution substation (2) being electrically connected to at least one connection by a feeder (3), a respective smart meter (4, 5) being provided between the feeder (3) and the connection (6, 7) and/or distribution substation (2) for measuring the energy consumption of the connections (6, 7) and the energy delivered by the distribution substation (2) to the feeder (3) by recording the energy going through a first smart meter (4) from the distribution substation (2) to the feeder (3) and by recording the energy going through a second smart meter (5, 8) from the feeder (3) to the connection (6, 7), the smart meters (4, 5, 8) measuring the energy going through it and recording the energy going through it at a predetermined recording rate in energy records, characterized in that the method comprises the steps of selecting a period having a predetermined number of energy records depending on the recording rate, selecting a number of possible connections (6, 7) connected to the feeder (3), retrieving the energy records going through the first and second smart meters (4, 5, 8) during the selected period, statistically determining, based on the retrieved energy records by the second smart meters (5, 8), the probable contribution of the selected connections to the energy recorded by the first smart meter (4) and based on the probable contribution of the selected connections (6, 7) to the energy recorded by the first smart meter (4) determining the probability that each of the selected connections (6, 7) is connected to the feeder (3).

Method according to claim 1 , characterized in that the statistically determining, based on the retrieved energy records by the second smart meters, the probable contribution of the selected connections to the energy recorded by the first meter is done by determining the values for β, in the equation MEC(f) =∑MEC(cl) - l wherein MEC(f) is the energy recorded by the first smart meter of the feeder f and MECfc) is the energy by the at least one second smart meter of the respective at least one selected connection c„ wherein β, is the probable contribution of the selected connections c, to the energy recorded by the first meter of the feeder f and wherein MEC(f) is determined by making the sum of the respective products of MECfc) with /3,for all selected connections c,.

3. Method according to claim 2, characterized in that when the predetermined number of energy records is n for every connection c„ the equation is solved by solving the equation:

MEC(f) MEC(Cl )a - MEC(cm),A ( β

MEC(f)„ MEC(Cl )m ··■ MEC(cm )tl β.

wherein the number of connections c, is m, MEC(f)t with j running from 1 to n, being the jth energy recorded by the first smart meter of the feeder f within the selected period, MEC(Ci)tj, with j running from 1 to n and /' running from 1 to m, being the jth energy recorded by the second smart meter of the ith connection within the selected period.

4. Method according to any one of claims 1 - 3, characterized in that the β, values are be calculated by using a QR- decomposition or an iterative least squares method with constraints on the resulting coefficients.

5. Method according to any one of claims 1 -4, characterized in that the statistical determination is done by using a least-squares method.

6. Method according to any one of claims 1 -5 at least in combination with claim 2, characterized in that the calculated β, values determining the probability that each of the selected connections is connected to the feeder.

7. Method according to claim 6, characterized in that a threshold is determined and that the obtained calculated β, values determining the probability that each of the selected connections is connected to the feeder is compared to the threshold, wherein values /3/< of the calculated values β, which are higher than or equal to the threshold are considered to indicate that connections ck are connected to the feeder and wherein values /3,, different than the values /¾, of the calculated values β, ^ which are lower than the threshold are considered to indicate that connections c, are not connected to the feeder.

Method according to claim 7, characterized in that the threshold value is determined such that MEC(f) subtracted with the sum of all MEC(Ck) having a corresponding fik which is higher than the determined threshold is positive and substantially the smallest with respect to MEC(f) subtracted with a different sum of MECfc) not necessarily having a corresponding β, which is higher than the determined threshold.

Method according to claim 8, characterized in that the threshold value is determined using a scoring function.

0. Method according to any one of claims 1 - 9, characterized in that the distribution substation comprises more than one feeder (3, 9) and that the method is repeated for every feeder connected to the distribution substation.

1. Method according to claim 10 at least in combination with claim 2 and preferably in combination with claim 8, characterized in that a postprocessing step is performed after repeating the method for every feeder in a selected area, the step comprising the following:

first all connections deemed connected to a single feeder are considered to be connected to those respective feeders,

subsequently the connections ck deemed connected to more than one feeder are considered to be connected to that feeder, if present, having the largest maximum of the absolute value of 0.5- the corresponding fik and the lowest MEC(f) subtracted with the sum of all MECic^,

subsequently the connections ck deemed connected to more than one feeder which are connected to a PLC gateway on a feeder and which are deemed connected to that feeder are considered to be connected to those feeders.

12. Method according to any one of claims 1 - 11 , characterized in that the selection of the probable connections to the feeder is done using a GIS system and the probable connections are connections within a predetermined distance, preferably a predetermined distance from the feeder.

13. Method according to any one of claims 1 - 12, characterized in that the number of energy records is at least the number of possible connections connected to the feeder selected.

14. Method according to any one of claims 1 - 13, characterized in that the method is used for detecting the phase(s) the smart meters are connected to.

15. Method according to any one of claims 1 - 14, characterized in that the smart meters record the energy going through the respective smart meters by measuring the current going through the respective smart meters.

16. Method according to any one of claims 1- 15, characterized in that the method is performed by a computer.

17. Method for detecting non-technical losses on a feeder according to any one of claims 1 - 16 at least in combination with 8, characterized in that MEC(f) subtracted with the sum of all MEC(Ck) having a corresponding /3/< which is higher than the determined threshold is compared to zero, values for MEC(f) subtracted with the sum of all MEC(Ck) having a value greater than zero indicating a non-technical loss on the feeder, values for MEC(f) subtracted with the sum of all MEC(Ck) having a value of substantially zero, or even zero, indicating substantially no, or even no, non-technical losses.

18. Computerprogram in a computer readable format comprising software code parts for executing the method according to any one of the claims 1-17.

Description:
Method for detecting low voltage connectivity in part of an electricity grid

Method for detecting low voltage connectivity in part of an electricity grid, the electricity grid comprising a distribution substation being electrically connected to at least one connection by a feeder, a respective smart meter being provided between the feeder and the connection and/or distribution substation for measuring the energy consumption of the connections and the energy delivered by the distribution substation to the feeder by recording the energy going through a first smart meter from the distribution substation to the feeder and by recording the energy going through a second smart meter from the feeder to the connection, the smart meters measuring the energy going through it and recording the energy going through it at a predetermined recording rate in energy records, according to the preamble of the first claim.

The term low voltage, LV, connectivity refers to (a) which customers, in the context of this application generally referred to as connections, are connected to which feeder, in this case a low voltage feeder or LV feeder or cable, and (b) which feeder is connected to which distribution substation. GIS (Geographic Information System) systems typically try to determine relation 'a 1 , while the most interesting relation is a combination of ' and 'b: to which distribution substation is a customer connected.

For example, before a planned outage of the LV- electricity grid, customers are informed about the upcoming maintenance. Knowing the exact connectivity avoids informing a customer that will not be affected. It also avoids not informing an affected customer, which could be worse as in some countries, DGOs (Distribution Grid Operators) have to pay a fee to customers when they suffered an unplanned electricity downtime. This requires the DGO to know the exact LV-connectivity.

Moreover, synergies between smart metering and smart grid initiatives allow demand side management for residential customers. Bad connectivity data decreases the efficiency of the used algorithms. Also, smart metering projects have a positive business case thanks to possibility to detect unmeasured electricity consumption. Bad connectivity data will cause a lot of false positive detections. The overall minimum detectable fraud will rise in case of unreliable connectivity data.

Also, knowing the exact connectivity of a distribution grid allows the DGO to use this grid more optimal in regard to balancing and preventive maintenance.

Until now, the only way to know the real connectivity of a customer was to use special equipment. With this equipment and a lot of manual effort, the exact connectivity could be determined. However, the known methods were very time-consuming and expensive and therefore the connectivity for a large area could not be easily determined. As an alternative to this method, utilities often use geographic data and a simple heuristic to determine the LV-connectivity on a larger scale. The cable segment that lies geographically closest to the customer or connection in question, in such a method is considered the connected cable. However, in various situations, this heuristic fails since streets can have more than 1 LV-cable, near a junction of 2 or even more feeders, near distribution substations, with parallel streets close to the connection, etc. In urban areas, these situations also occur more often than in rural regions. Moreover, the geographic location of old infrastructure is not always known and is hard to localize with cables running underground.

It is therefore an object of the current invention to allow for an improved method for determining low voltage connectivity in at least part of an electric grid.

This is achieved according to the method for detecting low voltage connectivity in part of an electricity grid according to the characterizing part of the first claim.

Thereto, the method comprises the steps of selecting a period having a predetermined number of energy records depending on the recording rate, selecting a number of possible connections connected to the feeder, retrieving the energy records going through the first and second smart meters during the selected period, statistically determining, based on the retrieved energy records by the second smart meters, the probable contribution of the selected connections to the energy recorded by the first smart meter and based on the probable contribution of the selected connections to the energy recorded by the first smart meter determining the probability that each of the selected connections is connected to the feeder.

Surprisingly, such a method allows in a relative simple and cheap way to determine the probability that certain connections are connected to a certain feeder leaving a distribution substation.

Moreover, it has been found that the method according to the invention can be used in situations where several feeders are interconnected to several distribution substations and interconnect these several distribution substations to several connections. In such situations in a first step a specific feeder connected to a specific distribution substation is selected after which the above described method is further applied.

Moreover, as more smart meters will be installed for more connections, the amount of data will rise such that the accuracy of the method will also increase. According to preferred embodiments of the current invention, the statistically determining, based on the retrieved energy records by the second smart meters, the probable contribution of the selected connections to the energy recorded by the first meter is done by determining the values for β, in the equation

MEC(f) =∑MEC(c l ) - l wherein MEC(f) is the energy recorded by the first smart meter of the feeder f and MECfc) is the energy by the at least one second smart meter of the respective at least one selected connection c„ wherein β, is the probable contribution of the selected connections c, to the energy recorded by the first meter of the feeder f and wherein MEC(f) is determined by making the sum of the respective products of MECfc) with /3, for all selected connections c,. Such an equation has been found to be easy to solve using, especially using a computer. Moreover, it has been found that the determined values for β, are a good representation of the probability that each of the selected connections is connected to the feeder. The mathematical basis for the equation will be explained later on. According to more preferred embodiments of the current invention, when the predetermined number of energy records is n for every connection c„ the equation is solved by solving the equation:

MEC(f) MEC(c 1 ) n ■■■ MEC(c m ) n

MEC(f)„ MEC( Cl ) m ··■ MEC(c m )„ β.

wherein the number of connections c, is m, MEC(f) t j, with j running from 1 to n, being the j th energy recorded by the first smart meter of the feeder f within the selected period, MEC(c) t j, with j running from 1 to n and / ' running from 1 to m, being the j th energy recorded by the second smart meter of the i th connection within the selected period. It has been found that such a matrix equation can be easily solved by a computer.

According to more preferred embodiments of the current invention, the β, ^ values are be calculated by using a QR-decomposition or an iterative least squares method with constraints on the resulting coefficients as such method can be easily performed by a computer, even for large values for n and m.

According to preferred embodiments of the current invention, the statistical determination is done by using a least-squares method such as for example using a QR-decomposition or an iterative least squares method with constraints on the results. It has been found that such a least- squares method can be done relative easily.

According to preferred embodiments of the current invention, the calculated β, values determine the probability that each of the selected connections is connected to the feeder as it has been found that the determined values for β, are a good representation of the probability that each of the selected connections is connected to the feeder.

According to more preferred embodiments of the current invention, a threshold is determined and the obtained calculated β, values determining the probability that each of the selected connections is connected to the feeder is compared to the threshold, wherein values fi k of the calculated values β, which are higher than or equal to the threshold are considered to indicate that connections c k are connected to the feeder and wherein values /3,, different than the values jS fc , of the calculated values β, ^ which are lower than the threshold are considered to indicate that connections c, are not connected to the feeder. It has been found that such a threshold allows an easy way of deciding whether a selected connection is connected to the feeder or not.

According to more preferred embodiments of the current invention, the threshold value is determined such that MEC(f) subtracted with the sum of all MEC(Ck) having a corresponding fi k which is higher than the determined threshold is substantially closest to zero, preferably closest to zero, more preferably positive and/or substantially the smallest, with respect to MEC(f) subtracted with a different sum of MECfc) not necessarily having a corresponding β, ^ which is higher than the determined threshold. It has been found that such a threshold allows to establish which connections are connected to the feeder with sufficient accuracy.

According to more preferred embodiments of the current invention, the threshold value is determined using a scoring function. According to preferred embodiments of the current invention, the distribution substation comprises more than one feeder and the method is repeated for every feeder connected to the distribution substation as such a method allows to determine which connections are connected to which feeder, for example when several feeders are present in a single street and the different connections connect in a random way to one of the feeders.

According to more preferred embodiments of the method according to the present invention a post-processing step is performed after repeating the method for every feeder in a selected area, for example every feeder connected to the distribution substation, the step comprising the following:

first all connections deemed connected to a single feeder are considered to be connected to those respective feeders,

subsequently the connections c k deemed connected to more than one feeder are considered to be connected to that feeder, if present, having the largest maximum of the absolute value of 0.5- the corresponding jS fo or for example 0.5 - the corresponding fi k squared, and the lowest MEC(f) subtracted with the sum of all MEC(Ck),

subsequently the connections c k deemed connected to more than one feeder which are connected to a PLC gateway on a feeder and which are deemed connected to that feeder are considered to be connected to those feeders.

It has been found that the connections c k still left over after this post-processing step are connections having a very low electricity consumption and that the obtained connectivity of the different connections to the different feeders is substantially correct and sufficiently accurate.

According to preferred embodiments of the current invention, the selection of the probable connections to the feeder is done using a GIS system and the probable connections are connections within a predetermined distance, preferably a predetermined distance from the feeder. Although such a system is not at all perfect, as described above, it allows performing a first selection of possible connections which can then be further evaluated.

According to preferred embodiments of the current invention, the recording rate is every quarter of an hour such that the energy going through the smart meters during every 15 minutes is being recorded. It has been found that such recording rate, 96 times a day, provides sufficient results for performing the method according to the current invention with sufficient accuracy in a period chosen within an interval of days. It has been found that good results can be obtained in a period of three days. The theoretical minimum number of time intervals preferably equals the number of possible connections connected to the feeder selected. However, it has been found that, independent of the length of the time interval, the number of time intervals preferably is larger in order to obtain good results. This is because time intervals with a low general energy consumption don't add any value to the end result of the coefficients. According to preferred embodiments of the current invention, the method is used for detecting the phase(s) the smart meters are connected to.

According to preferred embodiments of the current invention, the smart meters record the energy going through the respective smart meters by measuring the current going through the respective smart meters. By measuring the current going through them, a more precise measurement is obtained as less losses are included in the measurement.

According to preferred embodiments of the current invention the method is performed by a computer.

According to preferred embodiments of the current invention the method is related to detecting non-technical losses, for example fraud, MEC(f) subtracted with the sum of all MEC(Ck) having a corresponding /3 /< which is higher than the determined threshold is compared to zero, values for MEC(f) subtracted with the sum of all MEC(Ck) having a value greater than zero indicating a non-technical loss on the feeder, values for MEC(f) subtracted with the sum of all MEC(Ck) having a value of substantially zero, or even zero, indicating substantially no, or even no, non-technical losses.

The invention also relates to a computerprogram in a computer readable format comprising software code parts for executing the method according to the invention.

The invention will be further elucidated by means of the following description and the appended figures.

Figure 1 shows an overview of an example of an electricity grid 1 , the electricity grid 1 comprising a distribution substation 2 being electrically connected to at least one connection by a feeder 3.

Figure 2 shows a detail of figure 1 in which a respective smart meter 4, 5 is provided between the feeder 3 and the connection 6, 7 and/or distribution substation 2 for measuring the energy consumption of the connections 6, 7 and the energy delivered by the distribution substation 2 to the feeder 3 by recording the energy going through a first smart meter 4 from the distribution substation 2 to the feeder 3 and by recording the energy going through a second smart meter 5, 8 from the feeder 3 to the connection 6, 7. Figure 3 shows the recorded energy at the first meter 4, the sum of the energies recorded at the second smart meters 5, 8 and the difference between those two.

Figure 4 shows a result of a least-square method.

Figure 5 shows the recorded energy at the first meter 4, the sum of the energies recorded at the second smart meters 5, 8 and the difference between those two with different connections selected than in figure 3.

In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention and how it may be practiced in particular embodiments. However, it will be understood that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures and techniques have not been described in detail, so as not to obscure the present invention. While the present invention will be described with respect to particular embodiments and with reference to certain drawings, the invention is not limited hereto. The drawings included and described herein are schematic and are not limiting the scope of the invention. It is also noted that in the drawings, the size of some elements may be exaggerated and, therefore, not drawn to scale for illustrative purposes.

The present invention will be described with respect to particular embodiments and with reference to certain drawings but the invention is not limited thereto but only by the claims. The drawings described are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes. The dimensions and the relative dimensions do not necessarily correspond to actual reductions to practice of the invention.

Furthermore, the terms first, second, third and the like in the description and in the claims, are used for distinguishing between similar elements and not necessarily for describing a sequential or chronological order. The terms are interchangeable under appropriate circumstances and the embodiments of the invention can operate in other sequences than described or illustrated herein. Moreover, the terms top, bottom, over, under and the like in the description and the claims are used for descriptive purposes and not necessarily for describing relative positions. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments of the invention described herein are capable of operation in other orientations than described or illustrated herein.

The term "comprising;', used in the claims, should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. It needs to be interpreted as specifying the presence of the stated features, integers, steps or components as referred to, but does not preclude the presence or addition of one or more other features, integers, steps or components, or groups thereof. Thus, the scope of the expression "a device comprising means A and B' should not be limited to devices consisting only of components A and B.

Because of a research on fraud detection using the concept of energy balance, every distribution feeder in the substations has a smart meter connected to it using a current transformer. This allows us to measure the energy going through every feeder. Every connection on this feeder is also measured.

The unit of measurement usually is kWh (energy). This unfortunately means that copper/cable losses on the distribution feeders are included in the energy measurements. The energy going through the smart meters could also be indicated by measuring the current going through the smart meters increasing the precision of the algorithm by avoiding measurement of losses.

The recording rate used is 1 energy measurement every 15 minutes. However, as a statistical analysis is being performed in the proposed algorithm, the number of energy records and not the recording rate is critical for the invention.

A typical example of an overview of an electricity grid is shown in figure 1. The electricity grid 1 comprises a distribution substation 2 which is electrically connected to at least one connection by a feeder 3. In addition a second distribution substation is shown connected through switches and/or jumpers with the first distribution power station.

It has been found that a mathematical model can be constructed representing based on which low voltage connectivity in part of an electricity grid can be detected.

Assume:

Cj\ connections 6, 7 on distribution feeder f

EC(x): Actual energy consumption

MEC(x): Measured energy consumption

Every smart meter 4, 5, 8 has a measurement error, which is considered constant over a longer recording interval.

Then

We assume that every connection has an impact on the cable losses that is directly proportional with its consumption:

CableLosses =* ^oc, EC(c t )

This gives us:

MEC ( / )

~ I ^(1 + oC j ) · ECic · MeasurementError(f)

„ (1 + a, ) · MEC (c ) · MeasurementError (/)

i MeasurementError^^

: ^ £C( C ,) . ,

Per connection q we need to find a constant β, which contains the measurement error and the impact of this connection on the cable losses.

Because of the approximation of the cable losses and other side effects such as small deviations on the smart meter ' s system clocks, we need several energy records to determine β,.

The most straightforward method to calculate β, would be a least-squares method: MEC(f) =∑MEC(c l ) - l

Assuming we have for every connection a load profile of n energy records, this becomes in a matrix representation:

The β, values can be calculated by using a QR- decomposition or an iterative least squares method with constraints on the resulting coefficients.

In a perfect situation, β, should be between 1 and 1.1 , assuming a maximum cable loss of 10%.

Now assuming:

Of , connections on distribution feeder f

Cgj. connections not on distribution feeder f

Ci = { Cf , Cgj}

In other words, we extend our collection of connections with some extra ones which are not on feeder f.

The resulting coefficients >8 will be:

β(¾,β = 0

The least squares algorithm will assign a small, or even zero coefficient to the load profiles not connected to feeder f.

The resulting coefficients of the least squares fitting algorithm in other words determine which load profiles belong to the feeder and which don't. Assuming our fitting algorithm is optimized to return a coefficient between 1.1 and 0 (iterative variant with boundaries), we need a threshold for determining which load profiles belong to the feeder and which don't. Load profiles with coefficients above this threshold will be selected, load profiles with values below the threshold will be rejected. In order to achieve a better performance, we need another scoring mechanism and a way to calculate the optimum threshold value.

We assume that the Error relating to a feeder equals Load Profile (feed-meter) - Sum(Load Profile(selected meters)) should preferably always be substantially above zero because of energy losses on the cable. This for example means an error below zero should be extra penalized, for example by making the error 10 times larger when below zero. Possible scoring-functions are:

111111— € * - Λ· - £#;Γ pi ί ' -vi J J n s ch that

wherein x represents the load profiles, C the coefficients and d the measured values.

Another and more simple option is to make the error 10 times larger when below zero.

The most optimal threshold value will depend from feeder to feeder and can be determined by trying every value between 1 and 0.

The complete algorithm can work on 3 different datasets (timespans):

Training: to calculate the coefficients,

Validation: to calculate the minimum threshold based on the calculated coefficients

Test: to calculate the error

Using the information given above, we can design an algorithm to determine the connectivity:

1. Select a feeder f 3.

2. Choose a period p which, depending on the recording rate, contains enough energy records.

3. Select a collection of connections 6, 7 which are good candidates to be connected to feeder f 3. Good selection criteria can be street names, geographical coordinates, additional information from PL-communication (PLC or Power Line Communication), an area around the feeder such as for example defined by a distance from the feeder, the distance being for example 100 meters - 200 meters, for example 150 meters, ...

4. Retrieve the load profiles for period p, for both the feeder 3 and the collection of connections 6, 7.

5. Build the corresponding matrix representation and feed them into the least squares algorithm.

6. The load profiles with a corresponding coefficient β with a value around 1 , are connected to feeder f 3. The threshold for this determination preferably is determined using the method described above.

The same approach can be used for determining the phase(s) the meter is connected to. To support this, the tri-phase smart meters (in both the distribution cabinet and at our customers' premises) will need to measure the energy or current going through every phase. No additional features are required for the mono-phase meters.

In the algorithm every phase will need to be treated as if it was a separate feeder in the original algorithm. All load profiles will need to be added to the "collection of connections" of each phase.

A typical residential LV-distribution feeder contains both mono-phase and tri-phase connections. For mono-phase connections this means the load profile will be excluded on 2 phases and selected on 1 phase. For tri-phase connections there will be 3 load profiles. One will be selected per phase, and the other 2 will be excluded. Because we introduce more unknown coefficients into the model than we add measurements, the precision of the phase-identification algorithm will be lower than the connectivity determination algorithm. Therefore, more measurements are preferably added to increase the precision.

Determining the electrical connectivity of several feeders 3, 9 is a global optimization problem. For this kind of problem preferably the calculation is split in 2 steps in order to make it computationally more efficient: first applying the proposed algorithm on every feeder separately. After this is finished for all feeders 3, 9, we apply a post-processing step on the results of the local optimization algorithm.

When applying the algorithm to several feeders, we will see these results:

some load profiles assigned to 1 feeder,

some load profiles assigned to several feeders (mostly zero or low consumption) and

some load profiles are not assigned to any feeders For load profiles assigned to several feeders a probability based on the following criteria can be used:

PLC connectivity: connected to a concentrator/gateway on this feeder

The geographically closest feeder (distance-based, perpendicular)

Within a distance of x meter from this feeder (geographically in the neighborhood)

Resulting coefficient /3(c): Around 0 or 1 is better than 0.5: for example the function (/3(c)-0.5) A 2 will result in an improved polarization around 0 and 1 - Variance and mean of the Error relating to the feeder: A lower variance and mean is better

Bus-stops, billboards, xDSL amplifiers, ... are so called 'Unmetered points" because they are not metered, but billed using a fixed rate. When not taken into account, these load profiles show up as an "erroi" in the residue. Because these profiles are typically constant or block load profiles, they can be detected and eliminated.

According to a first example of the post-processing step:

first all connections deemed connected to a single feeder are considered to be connected to those respective feeders,

subsequently the connections c k deemed connected to more than one feeder are considered to be connected to that feeder, if present, having the largest maximum of the absolute value of 0.5- the corresponding fi k and the lowest MEC(f) subtracted with the sum of all MEC(Ck),

- subsequently the connections c k deemed connected to more than one feeder which are connected to a PLC gateway on a feeder and which are connected to that feeder are considered to be connected to those feeders.

It has been found that the connections c k still left over after this post-processing step are connections having a very low electricity consumption and that the obtained connectivity of the different connections to the different feeders is substantially correct.

Because every day doesn't contain too much time- intervals (96 in case of 15 minute readings), we can apply an exhaustive search algorithm to detect block profiles. The algorithm preferably is as follows:

Take a load profile of a feeder, consisting of several days of data

Consider every day to be a separate load profile from Oh to 24h

o For every timestamp (96 times in the case of 15 minute readings)

create a block load profile starting on this interval, with optimum energy and optimum duration as to minimize the error of all load profiles from Oh to 24h

accept the resulting block profile if the duration is longer than 1 hour and the energy is above a certain noise level (eg. 400W).

o Repeat

Only keep the best (minimum total error) block profiles, depending on how much you expect (typically 1 or 2)

EXAMPLE

Figures 3 - 5 show some experimental results.

The experimental results are based on the smart metering data collected in 2010.

Because our smart metering infrastructure is not configured to measure each phase separately, we will only illustrate the connectivity determination algorithm by providing a real-life example.

For illustration purposes, we will use a small distribution feeder of 9 connections. Figure 3 shows the load profile of the distribution feeder (the upper one), the sum of the load profiles of the connections on the feeder (the middle one) and a curve showing the difference between these two (the lower one). The positive difference indicates one or more missing connections. As can be seen the load profile of the distribution feeder is generally higher than the sum of the load profiles of the connections on the feeder due to cable losses.

To find a collection of "candidate connections" for the selected distribution feeder all connections in a radius of 100 meters around the distribution feeder (according to our GIS and asset data) were selected.

After retrieving the new load profiles from our Meter Data Management system, we fed them into an iterative least squares algorithm in Matlab and received the coefficients shown in figure 4. The algorithm clearly shows that c1 - c5 and c7 are also connected to our distribution feeder. The resulting load profiles can be seen in figure 5 and shows a perfect match between the energy that goes through the distribution feeder, and the energy that is consumed by the connections.

Experiments with larger distribution feeders showed similar results.

Smart meters measuring the energy consumption or current per phase will allow us to calculate the LV-connectivity per phase instead of per connection. This will bring advantages to people responsible for net-balancing, especially in areas with a lot of solar panels.