Privacy Preserving P2P technique

8/12/2019 Privacy Preserving P2P technique

1/19

1

Chapter 1

Introduction

Energy theft has been a notorious problem in traditional power systems. The

Generation, Transmission and Distribution (T&D) of electricity involve huge

operational losses. The magnitude of these losses is rising at an alarming rate in

several countries. In order to identify illegal consumers of electricity in the view

of enhancing the economy of utilities, efficiency and security of the grid, a new

method of analyzing electricity consumption patterns of customers and

identifying illegal consumers is proposed and realized.

In this chapter, Section 1.1 discusses the motivation and objectives behind the

development of a technique for detection of illegal consumers in a power grid,

followed by Section 1.2, which presents the organization of the report.

1.1 Motivation

Losses that occur during generation can be technically defined, but T&D lossescannot be quantified completely from the sending-end information. Distribution

losses in several countries have been reported to be over 30%. Substantial

quantity of losses proves the involvement of Non-Technical Losses (NTL) in

distribution. Total losses during T&D can be evaluated from the information

like total load and the total energy billed, using established standards and

formulae. In general, NTL are caused by the factors external to the power

system. Electricity theft constitutes a major chunk of the NTL. Major forms of

electricity theft include bypassing (illegal tapping of electricity from the feeder),tampering the energy meter, and physical methods to evade payment. Electricity

theft can be defined as, using electricity from the utility without a contract or

valid obligation to alter its measurement. Worldwide T&D losses are more than

the total installed generation capacity of countries such as Germany, the UK, or

France. It is estimated that utilities (worldwide) lose more than $25 Billion

every year due to illegal consumption of electricity. In Pennsylvania, PPL, a

utility reports an increase in electricity theft by 16% compared to 2008. It has

also been identified that the illegal consumption of electricity by local businesssector is increasing.


2/19


3/19

3

Chapter 2

Methods of illegal electricity usage

In general, electricity consumers may be generalized as genuine customers,

partial illegal consumers, and illegal consumers. This chapter presents several

simple and sophisticated methods used in pilfering electricity, discusses factors

that influence illegal consumers to steal electricity.

2.1 Methods of Stealing Electricity

The most common and simplest way of pilfering electricity is tapping energy

directly from an overhead distribution feeder as shown in Figure 2-1. The nextmost prominent method of electricity theft is the manipulation of energy meters

that are used for recording and billing industrial, commercial and household

energy consumption. Though there are many techniques for tampering with

such meters, some of these may include:

Exposing meters to strong magnetic fields to wipe out the memory. Inserting a film or depositing high viscous fluid to disturb the

rotation of disc.

Implementation of sophisticated technologies like remote sensingdevices.

Tampering the crystal frequency of integrated circuits. Creating a link between the breaking control wires in an energy

meter would divert the current reading in the meter reflecting zero

reading at all times.

In the case of electronic meters, Radio Frequency (RF) devices aremounted to affect the accuracy of the meter.

A shunt is installed between the incoming and outgoing meterterminals.

Inter-changing the incoming and outgoing terminals of the meter. Damaging the pressure coil of the meter. Resetting the meter reading. Introducing unwanted harmonics. Exposing the meter to mechanical shock.


4/19

4

Voltage is regulated from the meter terminals, making it read lesserquantity then the original consumption.

Figure 2-1: Tapping electricity directly from a distribution feeder - bypassing

the meter.

Other engineered methods of tampering with the meter without damaging its

terminals are illustrated below. Two-watt hour meters (employed for measuring

the energy consumed by large loads with three phase electric supply) are

tampered according to the following process: Damage the terminal seal; connect

one of the load terminals to the ground; and open the ground wire from the

energy meter. In the case of three phase meters, phases are shifted to lower the

power consumption reading by the energy meter. Another popular way of

lowering the energy meter reading without directly tampering with the meter is

shown in Figure 2-2. Here, supply voltage is regulated to manipulate the meter

reading. Illegal consumers accomplish this by using one of the three phases;

disconnect neutral from the distribution feeder, and using a separate neutral for

the return path. Therefore, the energy meter assumes that the voltage between

the connected phase and this new neutral is zero, implying that the total energy

consumed is zero. Another way of stealing electricity is by isolating neutral and

disturbing the electronic reference point by physically damaging the meter. The


5/19

5

voltage to be read by energy meter can then be manipulated by controlling the

neutral.

In general, illegal consumption of electricity will be predominant only at desired

hours of the day - when the customers demand is high i.e. using legalelectricity for small household loads and illegally tapped electricity for heavy

loads. This kind of theft (partial illegal consumption) is very difficult to

measure, as the energy consumption pattern is uneven over a period of time. In

addition, corrupt employees are often responsible for billing irregularities; they

record an amount of consumption that is lower than the original consumption.

On the other hand, improper calibration and illegal de-calibration (during

manufacturing) of energy meters can also cause NTL. In most of the meter

tampered locations, damaged meter terminals and/or illegal practices may not bevisible during inspection.

Figure 2-2: Technique used by illegal consumers to regulate the supply

voltage and manipulate the energy meter reading.


6/19

6

Chapter 3

Methods of Theft Detection

3.1 Historical development in Theft Detection Method

Nizar et al. employ a data mining technique known as Extreme Learning

Machine (ELM) to classify users electricity consumption patterns or loadprofiles. By comparing the results to a database of users load profiles, the

proposed algorithm yields a list of users who could be stealing energy, which

we call energy thieves.

Nagi et al. propose a similar approach but choose to use genetic algorithms andSupport Vector Machine (SVM) instead of ELM.

Depuru et al. develop another data mining based scheme utilizing SVM as well.

Unfortunately, these techniques cannot sort out the energy thieves with absolute

certainty.

In contrast, Bandim et al. propose a central observer to measure the total energy

consumption of a small number of users, and are able to identify all the energy

thieves by comparing the total energy consumption with the reported energyconsumption from the users.

Notice that in all the above works, the UC has to know some of users privateinformation, e.g., users load profiles or meter readings at certain times, in orderto find the energy thieves. However, the disclosure of such information would

violate users privacy and raise concerns about privacy, safety, etc.

3.2 Load profile and Privacy

The disclosure of such information would violate users privacy and raiseconcerns about privacy, safety, etc. In particular, users private information may

be sold to interested third-parties. Insurance companies may buy load-profiles

from the UC to make premium adjustments on the users policies. For example,they could find electricity consumption patterns that increase the risk of fire in a

property and increase insurance premiums accordingly. Marketing companies

may also be interested in this data to identify potential customers. Moreover,

criminals may utilize such private information to commit crimes. For instance,

robbers may analyze the energy consumption pattern of potential victims to


7/19

7

deduce their daily behavior. They can even know if a robbery alarm has been set

at their target location.

Many researchers, such as Quinn, have realized how high resolution electricity

usage information can be used to reconstruct many intimate details of aconsumers daily life and invade his/her privacy, and thus call for statelegislators and public utility commissions to address this new privacy threat.

Table 3.1.1 shows some of the outcomes of the research done by Quinn


8/19

8

Concern Type Related Question Answered by Detailed uses

Nefarious Uses When are you usually away from home? Is your household protected with an electronic

alarm system? If so, how often do you arm it?

And the Psycho question: when do you usuallyshower (and so prompt a long draw from yourwater heater)?

Insurance How often do you arrive home around the timethe bars close?

How often do you get a full nights sleep v.drive sleep deprived?

How often are you late to work, or rushing toget there on time?

Does the time it takes you to get from yourhome to your workplace require that you breakthe speed limit to get there?

Do you have a propensity for leaving appliancesturned on and leaving the house, say, a curlingiron or a stove range?

Banking/NBFCs

(For Loans)

An individuals loadprofile gives a rough idea ofhis willingness to save.

Is he/she going to job regularly/punctually i.e. isthere any threat to persons job?

How much time does a person gives to sleep i.e.laziness.

Table 3.1.1Concern Type Related Question Answered by Detailed uses

As per researchers, if these information will be sold by utilities then people will

behave like if they are watched by cameras in their personal life which is very

serious threat to the privacy of any individual.


9/19

9

3.3 Privacy Preserving P2P Computing Technique

It is a technique of theft detection without losing privacy. In section 3.3.1

Network model is given. Section 3.3.2 gives mathematical model of our

Problem and in Section 3.3.3, LUD algorithm is shown.

3.3.1 Network Model

Fig.3.1A typical architecture of Neighborhood Area Network (NAN).

Typical network architecture is depicted in Fig. 3.1. In a serviced area, the

users SMs together with the collector form a Neighborhood Area Network(NAN). The communications among SMs and between SMs and the collector

are carried out wirelessly due to SMs communication capability, while the

communications among the CC, the DS, and the collector are conducted viawired medium.

In the smart grid, communications and electricity networks overlay each other.

Utility companies (UCs) deploy control centers (CCs) to monitor their

distribution stations (DSs), distribution networks and deploy SMs at userspremises to measure their individual energy consumption. Since a CC is usually

physically far away from users, a communication entity that can facilitate the

communication between users and the CC is necessary. To this end, an access

point, called the collector, is placed at each of the serviced areas. One SM is


10/19

10

installed at each collector to measure the total energy consumed by the serviced

area.

3.3.2 Mathematical Model

In our mathematical model for energy theft detection, we assume that an SM isinstalled at the collector such that the collector can know the total energy

consumption of the users in the service area. We also assume that the UC

installs an SM at each of the users premises, which is capable of recordingenergy consumption at any time instant.

Consider a NAN of n users. We define asampling period denoted by SP. Then,

after every sampling period, all the n+1 SMs will record their energy

consumption in the past sampling period. We denote such energy consumption

recorded by user j (1 j n) and by the collector at time ti, by pti,j and ti,

respectively. We further define an honesty coefficient, denoted by kjwhere kj>

0, for each userj. Thus, kj* pti,jgives the real energy consumption of userj from

time instant ti SP to time instant ti. Since the sum of all users real energyconsumption in the past sampling period must be equal to the total energy

consumption of the neighborhood measured at the collector at time ti, we have

k1pti,1+ k2pti,2+ ... + knpti,n= ti(1)

Our objective is to find all the kjs. Obviously,

1) if kj= 1, then userj is honest and did not steal energy;

2) if kj > 1, then user j recorded less energy than what he/she consumes and

hence is an energy thief; and

3) if 0 < kj< 1, then user j recorded more than what he/she consumes, which

means that his/her smart meter may be malfunctioning.

In particular, with n linear equations, we can have a linear system of equations

(LSE) as follows:

k1pt1,1+ k2pt1,2+ ... + knpt1,n= t1


11/19

11

.

.

.

k1ptn,1+ k2ptn,2+ ... + knptn,n= tn

which can also be formulated in matrix form:

Pk = . (2)

The jth column of P represents the data recorded and stored by user j or SMj,

while the ith

row of P represents the data recorded by all the users at ti. The

collector can choose n time instances when ti s all have different values. Inthis case, it is highly likely that the LSE is independent and the rows of P are

independent as well, especially when n is large. Thus, the above LSE only has a

single unique solution, i.e., the feasible solution kj=ti,j/pti,jwherepti,jis the real

energy consumption of user j from time instant tiSP to time instant ti. Note

that our model does not take into account energy dissipation, or technical losses(TLs), in the power system, which are mainly caused by the intrinsic

inefficiencies in transformers and low voltage power lines. However, TLs can

be calculated without using consumers energy measurements.

For example, Oliveira et al. describe how to calculate TLs using measurements

at the distribution station and the knowledge of the distribution network which

does not need to compromise users privacy. Thus, once the technical losses arecalculated by the collector, the collector can adjust the model by subtracting the

TLs from vector P. Besides, note that finding the honesty coefficient vector, k,

is delay tolerant. In other words, k is not required to be found and transmitted to

the collector in real time. This gives priority to other real time traffic in the

NAN, such as electricity pricing, incentive-based load reduction signals, and

emergency load reduction signals.

3.3.3 LUD algorithm

It is based on distributed LU decomposition. The LU decomposition is to

factorize the energy consumption data matrix P into two triangular matrices: alower triangular matrix L and an upper triangular matrix U, i.e.,


12/19

12

P = LU.

The elements of upper triangular matrix U can be calculated as follows:

ui,j= 0, i>j

u1,j=pt1,j, j= 1, 2, ..., n

ur,j=ptr,j r,kuk,j, r= 2, ..., n,j = r, ..., n (3)

where pti,j is the ith element of column j in matrix P. Besides, the elements of

lower triangular matrix L can be obtained by

li,j= 0, i


13/19

13

is only known to the SM itself and the collector. The collector also transmits

tj+1to each SMjto allow the SMs to collaboratively compute L , U, and y. Wedenote the smart meter at the collector as SM0. All SMs start running Procedure

1 when the collector requests them to by sending a control message.

Procedure 1: Distributed LU Decomposition

Input:j SMj,tj+1SMj

1: ifj = 0 or SMjreceives packets from SMj1then

2: ifj = 0 then

3: Computey1using (7)

4: Transmity1only to SM1

5: end if

6: if 1 j n 1 then

7: for q = 1 toj do

8: Compute uq,jusing (3)

9: end for

10: for q =j + 1 to n do

11: Compute lq,jusing equation (4)

12: end for

13: Computeyj+1using (7)

14: Transmit columns 1,2,...,j of L and all known

elements ofy1, ..., yj+1only to SMj+1

15: end if

16: ifj = n then

17: Notify the collector that L , U, and y are available


14/19

14

18: end if

19: end if

Specifically, SM0first calculatesy1=Pt1, and then transmits it to SM1. SM0does

not need to compute any element of L or U. After SM1receivesy1, it computes

column 1 of U, column 1 of L , andy2. Then, SM1transmits column 1 of L ,y1,

andy2to SM2. SMj(1 < j < n), receivesy1throughyjand columns 1 throughj1

of L from SMj1, based on which it calculates columnj of U, columnj of L and

yj+1. After that, SMj transmits columns 1 through j of L and y1 through yj+1 to

SMj+1. Finally, SMn receives yn and columns 1 through n1 from SMn1,

calculates column n of U and column n of L , and notifies the collector that the

Back Substitution procedure, i.e., Procedure 2, can start. Notice that each SMj(j> 0) is responsible for computing columnj of U, columnj of L , andyj+1(1 j

Documents

Privacy Preserving P2P technique