1. Introduction
The efficient and safe operation of a power system is related to the balance of social
power supply <note: This part is ambiguous; clarification by the author is needed>
and demand. Power quality monitoring, power fault and equipment fault diagnosis, and
other types of power system monitoring are extremely important. In order to ensure
the safety of a power system, relevant departments <note: ambiguous> have introduced
communication technology to optimize a power system [1].
<New paragraph> As a distributed sensor network, a wireless sensor network (WSN) has
the characteristics of low cost, flexibility, and self-organization and can provide
important information for power system optimization [2]. Compared with traditional communication technologies [3], a WSN's collaboration and context awareness performance are superior, which makes
it have better fault tolerance, higher accuracy, larger coverage and local feature
extraction, as rapid and direct deployment, large scope, low installation and maintenance
costs, and easy replacement and upgrading. However, like all types of wireless networks,
WSNs are vulnerable to an open communication environment and constantly change with
the change of the environment. Therefore, it is impossible to apply advanced and complex
security mechanisms to a WSN, which makes its protection measures more complex and
makes a WSN more vulnerable to external attacks. This brings certain risks to the
system [4].
<New paragraph> At the same time, another disadvantage of long-distance use of a WSN
is the large delay of information transmission. The delay of information transmission
is related to the routing algorithm used, the distance of data transmission, and the
time when the last message is finally sent to the base station. In order to achieve
efficient communication, it is very important to minimize the delay of information
transmission [5].
A smart meter is an important part of a smart grid, and its measurement results are
related to grid safety. Therefore, it is of great significance to judge the bad operation
state of a smart meter [6]. With the expansion of power grid scale, the number of metering points of smart meters
continues to increase, and more than 500 million meters have been operated in China.
In order to ensure the accuracy of measurement, it is necessary to change smart meters
from regular replacement to status replacement <note: ambiguous> and it is imperative
to explore an efficient and accurate real-time remote estimation method for smart
meter operation error [7].
<New paragraph> At present, a traditional verification method is used for the accuracy
verification of smart meters in power enterprises. With the increase of power grid
scale, it is increasingly difficult to continue using the verification method [8]. The main reasons are reflected in two aspects. First, the calibration work is intensive
and heavy, which requires much manpower and vehicles to regularly calibrate each measurement
point. Secondly, the verification cycle is long, and the defects and abnormalities
of an electric energy meter (EEM) device between detection cycles cannot be found
and handled in time, resulting in low management efficiency [9].
<New paragraph> With the comprehensive development of the power Internet of things,
the collection and control of power data are also facing challenges. In order to effectively
control the data collection errors and equipment failures of smart meters, it is necessary
to analyze remote data and solve the problems of inaccurate remote online recording
of meter data and wrong judgment of meter failure types [10]. It is also of great significance to study a remote recognition method for bad operation
of smart meters.
Nowadays, the power department <note: ambiguous> has developed a variety of remote
online error analysis and diagnosis methods based on traditional error diagnosis methods
and combined with emerging science and technology. Kong et al. proposed a remote estimation
method of the error of smart meters, which makes the meter data and the error analysis
more accurate. However, this method can only be used for low-power data, and the analysis
of meter data in complex power grids is too complex and time-consuming [11].
<New paragraph> Farokhi applied a non-invasive load-monitoring algorithm to smart
meters to analyze and compare data beyond the error range and improve the reliability
of the results [12]. In order to avoid the risk of data eavesdropping caused by third-party attacks and
realize the two-way authentication of data collection and meter data, Dong's research
team studied an encryption algorithm to encrypt the data collection and storage process,
ensuring the integrity of the data transmission process [13]. However, the large-area damaged data of smart meters cannot be repaired in time,
which easily causes misdiagnosis and affects the transmission of remote control follow-up
instructions. Applying a WSN to smart meters can effectively be used to monitor abnormal
conditions and give warnings.
<New paragraph> the contributions of this paper are as follows:
1. A data acquisition method for a smart-city smart meter based on a WSN is proposed,
which realizes the efficient transmission of data information, reduces information
delay, and enhances the reliability of data information.
2. A WSN-based remote recognition method for bad operation of smart meters in smart
cities is proposed. The limited memory recursive least squares (FMRLS) algorithm was
used to recognize the errors of smart meters remotely, which effectively improves
the running speed of meters and reduces the bad operation of meters. In order to verify
the application effect of the proposed WSN communication technology in smart meters,
a certain area was studied as an experimental point for testing, and relevant experiments
were set up to prove it.
2. Materials and Methods
2.1 Data Acquisition from a Smart Meter in a Smart City based on a WSN
ZigBee technology was used as a communication technology for WSNs to collect smart
meter data in a smart city. When a WSN collects smart meter data in a smart city,
ZigBee technology divides the transmission process of the WSN into a master node and
slave node. The master node is the network coordinator, which mainly exchanges data
and information with the data processing module [14]. The slave node is the routing and terminal node, which connects to the terminal
server for information exchange, but they have roughly the same hardware design and
wireless transceiver. The structure of ZigBee technology is shown in Fig. 1.
ZigBee technology can provide effective and reliable data transmission to remote devices
with little power supply. The main features of ZigBee technology are as follows:
(1) High performance and low cost
The indoor transmission distance can reach 300 feet (100 m), the outdoor visual distance
can reach 1 mile (1.6 km), the transmission power output is 100 MW, and the RF speed
can reach 260 kbps.
(2) Low power consumption
The transmitting current is 300 mA, the receiving current is 50 mA, and the shutdown
current is < 1 uA.
(3) Advanced network and security
ZigBee technology has a confirmation and retransmission mechanism and supports point-to-point,
point to multipoint, peer-to-peer topology, automatic healing, and a fault-tolerant
mesh network.
(4) Ease of use
Without external RF communication, RF module parameters can be configured in AT or
API command mode, and XBee serial port communication can be configured with a free
test and software [15].
Fig. 1. Diagram of ZigBee technology structure.
2.2 Routing Algorithm for WSNs based on LEACH Protocol
When a WSN uses ZigBee technology to collect smart meter data in smart cities, bandwidth
resources are limited, which greatly affects the life cycle of sensor networks and
the quality of network information collected. The Low Energy Adaptive Clustering Hierarchy
(LEACH) protocol can evenly share data information to the cluster center point and
send it from the lower level to the higher-level layer by layer. It can effectively
reduce the traffic of terminal nodes, maintain the low energy consumption of sensor
nodes, and achieve the purpose of extending the network life cycle [16]. Therefore, the LEACH protocol was selected as the communication protocol of the
WSN to improve the performance of sensors and reduce the network delay of information
collection. The WSN routing algorithm based on LEACH protocol mainly includes the
following parts.
(1) Collection and transmission of node status information
Time Division Multiple Access (TDMA) can divide time into periodic frames, and each
frame is further divided into multiple time slots for signal transmission to ensure
that all signals do not interfere with each other [17]. In a stable state, each common node in the cluster starts the sending device in
its own working time slot for data transmission and turns off the sending device at
a time that is not its own time to enter the sleep state. In order to reduce unnecessary
energy overhead, the steady-state phase includes multiple TDMA frames for a long time.
In each TDMA frame, the cluster head node (CHN) can receive the data sent by other
nodes, compress the received data [18] and send it to the Sink node at the last moment of the TDMA frame to complete the
transmission of a frame of data. The node reports the current state information at
the same time in the last TDMA frame of each round. The Sink node will use the received
node state information as the basis for the next round of clustering.
After collecting their own state information, nodes also need to carry out the final
data transmission: ordinary nodes should send their own state information and their
perception data in the last TDMA frame of this round to the CHN, and the CHN should
send its own state information and the fused cluster information to the sink <note:
Did you mean ``Sink''?> node. This process must consume energy [17]. In order to collect the current energy state information of nodes more accurately,
a residual energy prediction mechanism is given, and the algorithm is expressed as
follows:
In Eq. (1), $k$ denotes the number of clusters, $\varepsilon $ is the adjustment coefficient,
$d$ denotes the distance from the cluster member node to the CHN, $E_{r}$ is the final
residual energy of the node after the end of this round of work, $E_{C}$ denotes the
current residual energy of the node, and $E_{Tx}$ denotes the energy that the node
will consume in this data transmission.
(2) Inter-cluster routing based on multi-hop
The communication between a CHN and cluster members still adopts the single hop communication
mode of the LEACH protocol, while the inter-cluster communication mode is changed
to the multi-hop forwarding mode after the first round [19]. An inter-cluster routing mechanism was designed based on the consumption of minimum
transmission energy supported by a Sink node:
Assuming that there are $k$ CHNs in the network in the next round, after running the
simulated annealing algorithm to determine which nodes are the CHNs of $k$ optimal
clusters, the $k$ nodes are arranged in ascending order according to their distance
from the node. That is, if there are CHNs $CH_{1},CH_{2}\cdots CH_{k-1},CH_{k}$, the
expression is as follows:
Some existing minimum energy routing protocols only consider the power dissipation
of data transmission and ignore the power consumption of data reception. The total
energy consumption of the two nodes must be considered as small as possible. The total
energy consumption of the two nodes must be considered as follows:
where $i$ represents the number of data forwarding. It can be seen that while reducing
the data transmission distance, we should also reduce the forwarding times. We stipulate
that the CHN only selects the nearest adjacent CHN with one-hop reachability and a
serial number greater than its own for data forwarding.
Assuming that $CH_{A},CH_{B},CH_{C}$ <note: There should be an ``and'' here> are three
CHNs and satisfy the relationships of $A<B<C$ and $d_{AB}<d_{AC}<d_{BC}$, $CH_{A}$
forwards data to $CH_{B}$, and $CH_{B}$ forwards data to $CH_{C}$.
(3) Route maintenance
A route maintenance mechanism based on local maintenance was designed to repair the
route breakpoints caused by node failure. Node failure is divided into cluster member
failure and CHN failure, and different strategies are adopted according to their respective
characteristics. If the CHN does not receive the data packet of a cluster member within
a certain time, it is considered that the cluster member is invalid. When the CHN
reports the status information of the nodes in the cluster to the sink <note: Did
you mean ``Sink''?> node, its residual energy is zero by default. In the whole WSN,
the failure of a cluster member will not have a great impact on the whole network,
so we do not deal with this situation. If the Sink node does not receive the data
packet of a CHN within $T_{r}/2$, it is considered that the CHN has temporarily failed
for some reason. The Sink node notifies the standby CHN to start working, and other
CHNs update the changed CHN in the routing table.
<New paragraph> After receiving the instruction, the standby CHN only broadcasts its
identity to the nodes in the cluster and does not divide the time slots in the cluster
again [17]. After receiving the broadcast, the nodes in the cluster still communicate with the
standby cluster head according to the original time slot divided by the Sink node.
If the original CHN can receive the broadcast of the standby cluster head, it will
transmit data to the new CHN in the time slot in which the original standby cluster
head communicates with itself.
2.3 Theoretical Model for Remote Identification of Operation Error of Smart Meter
in Smart City
The distribution diagram of power distribution stations in each station area of the
smart city is shown in Fig. 2. <note: Generally, paragraphs should be longer than this> A total smart meter (check
meter) is installed under the distribution transformer in each station area of a smart
city, and a total of m smart meters (user meter) are connected under it. To obtain
the error value of each user's smart meter, the absolute error of at least one meter
needs to be known in the station area.
A smart meter is usually set as a standard meter, and its reading is considered to
have no error. There is line loss in the network topology of the distribution station
area. Based on the law of energy conservation, in any measurement period, the reading
of the EEM in the station area is equal to the sum of the actual value of each user's
EEM plus the total loss of the station line in the upper station area in that period
[21]. For any $t$-th unit measurement period, we have:
where $y_{0}\left(t\right)$ is the reading of the total EEMs, $z_{i}\left(t\right)$
represents the reading increment of the $i$-th user energy meter, $\xi _{i}\left(t\right)$
is the error of the $i$-th user energy meter, $\left[1+\xi _{i}\left(t\right)\right]$
is the electric energy value actually consumed by the $i$-th user EEM, $w_{loss}\left(t\right)$
is the power loss of all lines in the station area, and $m$ is the total number of
user energy meters.
Since the line loss $w_{loss}\left(t\right)$ in the distribution station area cannot
be read directly by the measuring equipment, it needs to be obtained by a power flow
analysis and calculation in the station area. A BP neural network model optimized
by the Levenberg Marquardt (LM) algorithm was used to calculate the line loss rate
in the substation area [22]. The total line loss in the substation area was calculated from the calculated line
loss rate and the reading of the total EEM in the substation area.
Let $y\left(t\right)=y_{0}\left(t\right)-w_{loss}\left(t\right)$ represent the value
of the total EEM reading in the station area minus the line loss in any $t$-th measurement
period, and we use $\theta _{i}\left(t\right)$ to represent item $1+\xi _{i}\left(t\right)$.
A set of measurement data sequences is formed by each unit measurement period $z_{i}\left(t\right)$
and $y\left(t\right)$ obtained by solution. Eq. (4) can be transformed into matrix form. According to the measurement data of a smart
meter collected by the WSN, the remote identification model of operation error of
a smart meter in a smart city can be obtained:
where $Z\left(t\right)=\left[z_{1}\left(t\right),z_{2}\left(t\right),\cdots ,z_{m}\left(t\right)\right]$
is the measurement data matrix of each user's EEM in $t$ periods, $\hat{\otimes }\left(t\right)=\left[\theta
_{1}\left(t\right),\theta _{2}\left(t\right),\cdots ,\theta _{m}\left(t\right)\right]^{T}$
represents the error parameter matrix to be identified by each user's EEM in the $t$-th
measurement period, $\theta _{1}\left(t\right)$ is the operation error parameter of
the smart meter $\mathrm{i}$ to be obtained in the $t$-th measurement period, and
$\hat{\xi }\left(t\right)=\left[\hat{\xi }_{1}\left(t\right),\hat{\xi }_{2}\left(t\right),\cdots
,\hat{\xi }_{m}\left(t\right)\right]$ is the remote identification value of the operation
error of the EEM in $t$ measurement periods.
Fig. 2. Typical topology diagram of distribution station area.
2.4 Remote Identification of Smart Meter Error based on LMRLS Algorithm
The limited memory recursive least squares (LMRLS) algorithm was used to identify
the error of a smart meter remotely. The least square (LS) algorithm is one of the
most widely used online parameter identification algorithms at present. The calculation
is simple, and the established model has good dynamic characteristics [23]. The LS algorithm calculates the input data by minimizing the square of the sum of
the residuals between the output signal and the expected signal. For a smart meter
system:
where $Y_{k}$is a $k$-dimensional output vector satisfying $Y_{k}=\left[y_{1}y_{2}\enspace
\enspace \enspace \cdots \enspace \enspace \enspace y_{k}\right]^{T},\theta _{k}=\left[\theta
_{1}\theta _{2}\enspace \enspace \enspace \cdots \enspace \enspace \enspace \theta
_{n}\right]^{T}$, and $\Phi _{k}$ is a $k\times n$-dimensional input matrix: $\Phi
_{k}=\left[\begin{array}{l}
\varphi _{1}^{T}\\
\varphi _{2}^{T}\\
\vdots \\
\varphi _{k}^{T}
\end{array}\right]=\left[\begin{array}{llll}
\varphi _{1}\left(1\right) & \varphi _{1}\left(2\right) & \cdots & \varphi _{1}\left(n\right)\\
\varphi _{2}\left(1\right) & \varphi _{2}\left(2\right) & \cdots & \varphi _{2}\left(n\right)\\
\vdots & \vdots & \ddots & \vdots \\
\varphi _{k}\left(1\right) & \varphi _{k}\left(2\right) & \cdots & \varphi _{k}\left(n\right)
\end{array}\right]$. $\theta _{k}$ is the parameter to be identified meeting
$k>n$.
The least squares identification of $\theta _{k}$ is:
Since the online system will constantly obtain new data, if all data are retained
and the LS algorithm is used continuously, it will consume considerable memory and
computing resources [24]. Therefore, it is necessary to adopt a recursive form to keep the online update <note:
ambiguous> of the parameters to be identified. When obtaining the measurement data
of the $k+1$-th group, we have:
In Eq. (8), $\left\{\begin{array}{l}
K_{k+1}=P_{k+1}\varphi _{k+1}\\
P_{k+1}^{-1}=P_{k}^{-1}+\varphi _{k+1}\varphi _{k+1}^{T}\\
P_{k}^{-1}=\Phi _{k}^{T}\Phi _{k}
\end{array}\right.$ is required.
According to the Sherman-Morrison-Woodbury equation [24], the inverse of rank $k$ correction of multiple matrices can be obtained by calculating
the inverse of rank $k$ correction of the original matrix. Assuming $A\in R^{n\times
n}$, $U\in R^{n\times k}$, $C\in R^{k\times k}$, $C\in R^{k\times n}$, $A$, $C$, and
$UCV$ are nonsingular matrices (i.e., inverse existence), then:
The special cases are:
Therefore, the satisfaction condition of Eq. (8) is changed to:
The final recursive least squares expression is as follows:
Theoretically, the identification accuracy of the relative error of a smart meter
will improve with the increase of measured data. However, a significant disadvantage
of the RLS algorithm is that with the increase of measured data, the influence of
new data on identification parameters becomes smaller and smaller [25]. That is, data saturation occurs. As a result, when the parameters change, the RLS
algorithm cannot track the change in real time, which reduces the recognition accuracy
or even fails.
<New paragraph> Therefore, a forgetting factor was added to the old data to weaken
the impact of the old data in the iterative process, improve the data saturation phenomenon,
and accelerate the convergence speed of the iteration. After the forgetting factor
is introduced, the equation of fading memory recursive least square method is as follows:
In Eq. (13), $\delta $ is the forgetting factor, and the value range is 0 - 1.
In this paper, the LMRLS algorithm described in Eq. (13) was introduced to solve the remote identification model of operation error of smart
meter shown in Eq. (5). Firstly, the initial value of the algorithm is given. After sufficient smart meter
data in a smart city are obtained by the WSN (the number of data groups is larger
than or equal to the number of parameters to be identified), it can be substituted
into the theoretical equation of remote identification of operation error of the smart
meter in the smart city, and the LMRLS algorithm is used to solve the error of each
sub-meter. After obtaining the new measurement data, the identification value is corrected
by an iterative algorithm to realize the remote identification of the operation error
of a smart meter in a smart city in a WSN.
3. Results
To validate the feasibility of the remote identification method for operation error
of a smart meter in a smart city based on a WSN, the research method was applied to
a smart meter in a certain area of a city to verify the error remote identification
performance of the proposed method. The ZigBee sensor node of the WSN was set in the
smart meter to collect the operation data of the smart meter. The initial energy of
the sensor was 10 J. The sensor node was set with continuous transmission mode, and
each node transmitted 300-byte sensing data in each time slot. When counting the number
of different CHNs, the total energy consumed by the WSN when collecting smart meters
in smart cities changes. The specific results are shown in Fig. 3.
It can be seen in Fig. 3 that the energy consumption required for the WSN to collect smart-city smart-meter
data decreases first and then increases with the increase of the number of CHNs. When
the number of CHNs is 6, the energy consumption required for the WSN to collect smart-city
smart-meter data is the lowest. This is because when the CHN data are greater than
6, the information processing efficiency slows down, and the required energy consumption
gradually increases in order to obtain better results. To obtain the best communication
performance of WSNs, the number of CHNs is set to 6 when collecting smart-meter data
in a smart city using a WSN.
In the WSN used in the proposed method, the LEACH protocol was used as the routing
algorithm. In WSNs, it is stipulated that once a node runs out of its own energy,
the node will die and can no longer receive and transmit data. The static energy consumption
is ignored in the simulation, and the energy consumed by the node sensing data is
not considered.
Energy efficiency is a very important factor to be considered in the routing design
of WSNs. Fig. 4 shows the relationship between the network running time and the current number of
surviving nodes in the network in a network life cycle of the LEACH protocol used
in the proposed method. This method was compared with methods the references [11,12].
As shown in Fig. 4, the death time of the last protocol surviving node of one method [11] is about 1200 s, the death time of the last protocol surviving node in another method
[12] is about 700 s, and the death time of the last node of the LEACH protocol used in
this study is about 1400 s. Compared with the other two methods, the time is 300 s
and 700 s later, which indicates that the number of surviving nodes in the proposed
protocol is always higher than that in the other two methods, and the LEACH protocol
used in the proposed method can avoid excessive energy consumption.
As the total amount of data increases, the number of surviving nodes in the network
also changes. Table 1 shows the relationship between the total amount of data and the number of currently
surviving nodes in the network during the network life cycle of the LEACH protocol
used in the proposed method. The method was compared with methods in the references
again [11,12].
It can be seen in Table 1 that with the increase of the total amount of data, the number of surviving nodes
in the network shows a downward trend. When the first node dies, the total amount
of data received is about 4${\times}$10$^{4}$, and the total amount of data received
in one literature method [11] is about 3.5${\times}$10$^{4}$. The amount of data received in the other literature
method [12] is only 3${\times}$10$^{4}$. When the number of network nodes is the same, the total
amount of data received by our method is always higher than that of the comparison
methods, indicating that the energy efficiency of the proposed method is better than
that of those methods. The statistical method proposed in the study uses a WSN to
collect smart-meter data in smart cities and detects the performance of a WSN method
by analyzing the network throughput change when collecting different numbers of smart
meters. We compared the throughput-change result of this method with the references
[11,12].
As shown in Fig. 5, in one method [11], the network throughput fluctuates between 50 and 200 when collecting data, and in
the other method [12], the network throughput fluctuates between 100 and 220. The throughput of the proposed
method for data collection is stable in the range of 300-350. Compared with other
methods, the network throughput of our method is significantly higher, which verifies
that using this method to remotely identify bad operation of smart meters in smart
cities has higher network throughput and good communication performance.
A program-controlled load simulation system in the low-voltage distribution station
area was used to simulate an actual distribution situation. According to the collected
operation data, the loss rate of the distribution line in the study area was obtained.
The results are shown in Fig. 6.
Table 1. Relationship between total data and surviving nodes.
|
Total data/*104
|
Number of surviving nodes
|
|
Paper method
|
Reference [11] method
|
Reference [12] method
|
|
3
|
/
|
/
|
400
|
|
3.5
|
/
|
400
|
276
|
|
4
|
400
|
337
|
135
|
|
4.5
|
289
|
261
|
57
|
|
5
|
193
|
127
|
0
|
|
5.5
|
106
|
37
|
0
|
|
6
|
42
|
0
|
0
|
|
6.5
|
0
|
0
|
0
|
Table 2. Results of remote identification of operating errors of smart meters.
|
Meter serial number
|
Electric meter account number
|
Number of meters/kw·h
|
running error/kw·h
|
|
1
|
6254846152
|
5684
|
1.52
|
|
2
|
3254814521
|
3845
|
2.16
|
|
3
|
5421597484
|
2754
|
0.85
|
|
4
|
8545612845
|
1856
|
0.45
|
|
5
|
6485174593
|
3485
|
1.62
|
|
6
|
1584759623
|
9452
|
2.41
|
|
7
|
1858477512
|
12358
|
3.52
|
|
8
|
6284518762
|
3452
|
0.84
|
|
9
|
1235741965
|
4895
|
0.48
|
|
10
|
8564785941
|
3741
|
0.64
|
|
11
|
3542896415
|
6854
|
0.71
|
|
12
|
2310506485
|
8451
|
0.65
|
|
13
|
6028415225
|
9452
|
0.74
|
|
14
|
9600547152
|
3584
|
1.58
|
|
15
|
3684084065
|
2954
|
1.62
|
Fig. 6 show that the accuracy of the loss rate of the distribution line obtained by the
proposed method is high. When the identification accuracy of the loss rate of the
distribution line is within 0.1%, it has little impact on the operation error identification
accuracy of a smart meter, which can ensure the accuracy of operation error identification
of a smart meter. The method in this paper was used to remotely identify the operation
error of smart meters in a smart city, and 15 meters were randomly selected.
As shown in Table 1, the proposed method can effectively collect the operation data of smart meters in
smart cities and realize the remote identification of operation errors of smart meters
according to the collected operation data from the smart meters. The operation errors
are below 2 <note: ambiguous> which shows that the research method can effectively
control the bad operation of smart meters.
To further validate the remote identification performance of operation error of a
smart meter by using the proposed method, the actual operation state of special transformer
users in a typical distribution station area was fully simulated. This simulation
system was used to simulate the power consumption and line loss of four types of stations
of typical large-scale industrial special transformer users, medium-sized industrial
special transformer users, small-scale industrial special transformer users, and large-scale
commercial special transformer users to obtain the real error of a smart meter. The
MAPE and RMSE were used as an evaluation basis. In the remote identification process
of smart meter error, the smaller the MAPE value and RMSE values are, the higher the
accuracy of the identified error parameters is.
MAPE and RMSE can be expressed as:
where $q$ is the total number of users studied, and $\hat{\xi }\left(i\right)$ and
$\xi \left(i\right)$ respectively represent the identified value and actual value
of the operation error of the smart meter. According to the statistics, the method
in this paper was used to remotely identify the MAPE and RMSE of operation error of
a smart meter in a smart city. The MAPE and RMSE of operation error were compared
with those of other methods [11,12].
According to the experimental results in Figs. 7 and 8, the MAPE of remote identification
of smart-city smart-meter operation error by the research method is less than 1%,
and the RMSE fluctuates within 0.5%, which significantly reduces the error compared
with other methods. It was verified that the proposed method has ideal remote identification
performance of operation error of a smart meter in a smart city. This method can effectively
identify the operation error and improve the operation reliability of a smart meter.
<New paragraph> With the increase of remote recognition times, the MAPE and RMSE curves
of the other two methods fluctuate greatly, and the index values of MAPE and RMSE
are significantly higher than those of the method in this paper. Using the method
in this paper, the MAPE and RMSE of operation error of a smart meter are less than
1%. Compared with the other two methods, the proposed method has higher remote recognition
accuracy of operation error of a smart meter in a smart city and can effectively reduce
the operation error of the smart meter.
In order to illustrate the effectiveness of the method used in the study, the ROC
curve was used for performance comparison with the other methods. The results are
shown in Fig. 9. It can be seen from the ROC curve in Fig. 9 that the method used in this study has the best recognition effect and the highest
recall rate and accuracy. We calculated the area included in the curve and obtained
AUC values of 0.936, 0.842, and 0.795 for the research method, the first reference
method [11], and the other reference method [12], respectively. The calculation results also show that the remote identification method
of an intelligent meter proposed by this research performs best and has great practical
application value.
Fig. 3. Number of CHNs and energy consumption.
Fig. 4. The relationship between running time and surviving nodes.
Fig. 5. Comparison of network throughput changes.
Fig. 6. Calculation result of line loss rate.
Fig. 7. Mean absolute percent error.
4. Discussion
We studied a remote identification method for operation error of a smart meter in
a smart city based on a WSN, used a WSN to collect smart-meter data from a smart city,
and realized the remote identification of operation error of a smart meter in a smart
city according to the collected data. The main reasons for the errors of smart meters
in a smart city are as follows:
(1) Nonstandard installation
From a macro perspective, the most basic error problem of an EEM device is reflected
in the installation stage. The staff fail to consider the synthetic error of the transformer,
resulting in subsequent detection problems and interfering with the work of the power
system.
(2) The on-site inspection is not rigorous
After the installation of the EEM, the staff does not repeatedly check it, and there
are few checks on the product performance, operation state, and load condition. The
wiring and internal capacity are not considered, so a damaged EEM is still used for
metering, resulting in the failure of the power system.
(3) Impact of user load
China is a country with a large population, so the scale of power consumption is huge.
Especially in the peak stage of power consumption, the regional power consumption
rises sharply, resulting in a sharp rise in power flow and a full load state. The
regional current is unstable, and some load currents stay below 30% of the rated current,
resulting in measurement error.
(4) Error analysis of EEM
The first error is manufacturing process error. There are some errors in the EEM when
it leaves the factory. Factors such as materials, parts, and process level will affect
the quality of the EEM and interfere with the measurement accuracy. The second error
is improper use error. The normal operation of the EEM is ensured by the internal
line <note: ambiguous>. If the line is irregular or the use method is wrong, it will
cause measurement error. Although the error seems to be kept at a low level, considering
the current multiple <note: ambiguous> it will produce large error.
(5) Line conditions and grid environmental impact
In addition to the internal factors of the power system, external environmental factors
also interfere with the metering work of the EEM, such as temperature, humidity, magnetic
field, etc., which comprehensively affect the metering error. Even for the staff,
it is difficult to obtain the actual working condition data, and they cannot evaluate
the operation condition of the EEM.
In view of the possible causes of smart meter error in a smart city above, the following
strategies to improve smart-city smart-meter error are proposed:
(1) Standardization of the installation of EEM devices
Since installation work is basic work and is greatly affected by human beings, it
is very necessary to standardize the installation process. The staff are required
to master sufficient professional knowledge, investigate the working environment in
advance, take the design standard as a working basis, clean up the surrounding garbage
and dust, debug the parameters of the EEM in advance, and test it again after installation.
If abnormal operation is found, targeted countermeasures are taken according to the
problems to correct them in time to ensure the normal operation of the EEM.
(2) Determination of the transformation ratio of current transformer
The staff needs to investigate the power consumption of users in advance, including
the power consumption scale, residential type, etc., to predict the load condition
in advance, to predict the variation range of electric energy, and to adjust the internal
factors of transformer. Selecting an appropriate transformer and ensuring the transformation
ratio are important to avoid measurement problems under abnormal conditions and to
protect the normal life of the masses and the economic interests of power enterprises.
(3) Improvement of the level of on-site inspection
When an electric energy device starts to operate, the staff should regularly investigate
the device status, adjust the internal wiring according to the changes of users' needs,
periodically evaluate the metering performance, strictly abide by relevant regulations,
and record and store the work flow in a database for future reference and accountability.
(4) Reasonable reconstruction of the line<note: ambiguous>
The line reconstruction <note: ambiguous> should comply with relevant national regulations,
which should not be too far from the original line <note: ambiguous> and affect other
infrastructure. It should also ensure the stable output of equipment. Staff should
firstly understand the actual detection range of voltage from a macro perspective
and then improve the power measurement and statistics method from a micro perspective
to ensure that the EEM still operates normally under various power conditions.
5. Conclusion
The LEACH protocol was selected as a communication protocol of a WSN. The WSN was
used to collect the operation data of a smart meter in a smart city, and the remote
identification of operation error of smart meter in a smart city was realized based
on the FMRLS algorithm. Experiments showed that the LEACH protocol used in this method
can effectively improve the energy consumption characteristics of the network, prolong
the life cycle of the network, improve the network throughput, and enhance the scalability
and reliability of the network. Using this method to remotely identify the operation
error of smart meters in smart cities can improve the accuracy and real-time remote
error identification of smart meters and help to realize their transformation from
regular replacement to state replacement<note: ambiguous>. This can let us find suspected
abnormal metering points from technical means<note: ambiguous> overcome the bottleneck
of heavy manual troubleshooting and lack of pertinence, and provide support for efficient
power consumption inspection in a timely manner.
REFERENCES
Cheng L., Xue M., Wang Y., Bi Y., 2021, A robust tracking algorithm based on modified
generalized probability data as-sociation for wireless sensor network, IEEE Transactions
on Industrial Electronics, pp. 1-8
Priyanka E. B., Devi T. K., Sakthivel P., Sagayaraj A. S., 2022, Gate diffusion input
(gdi) codes involved viterbi decoders in wireless sensor network for enhancing qos
service, Analog Integrated Circuits and Signal Processing, Vol. 111, No. 2, pp. 287-298
Nancy P., Muthurajkumar S., Ganapathy S., Kumar S., Selvi M., Arputharaj K., 2020,
Intrusion detection using dynamic feature selection and fuzzy temporal decision tree
classification for wireless sensor network s, IET Communications, 2020, Vol. 14, No.
5, pp. 888-895
Qiu S., Zhu Y. H., Tian X., Chi K., 2020, Goodput-maximised data delivery scheme for
battery-free wireless sensor network, IET Communications, Vol. 14, No. 4, pp. 665-673
Abbasikesbi R., Nikfarjam A., Nemati M., 2020, Developed wireless sensor network to
supervise the essential parameters in greenhouses for internet of things applications,
IET Circuits Devices & Systems, Vol. 14, No. 8, pp. 1258-1264
Wang Y., Li Y. R., 2020, Simulation of Clustering Scheduling Method for Multi-Concurrent
Tasks in Internet of Things, Computer Simulation, Vol. 37, No. 3, pp. 475-479
Liu F., Liang C., He Q., 2020, Remote malfunctional smart meter detection in edge
computing environment, IEEE Access, Vol. 99, pp. 1-7
Taso Y. C., Ho C. W., Chen R. S., 2020, The impact of problem awareness and biospheric
values on the intention to use a smart meter, Energy Policy, Vol. 147, No. 1, pp.
111873
Yusoff N. S., Kaman Z. K., Zahari A. R., Wan H., Abdullah A. B., 2021, Examining smart
meter users' experience on continuance intention in adopting smart meter in malaysia
- result from a pilot study, Asia Proceedings of Social Sciences, Vol. 7, No. 2, pp.
110-113
Wang Q., Dai Y., Wu Y., Liu Y., Wang T., Dong X., 2020, Data model for smart electricity
meter comprehensive verification based on bp neural network, Journal of Physics: Conference
Series, Vol. 1486, No. 2, pp. 022028(9pp)
Kong X., Zhang X., Bai L., 2021, A remote estimation method of smart meter errors
based on nnf and gdrls, IEEE Transactions on Industrial Informatics, Vol. 99, pp.
1-6
Farokhi F., 2020, A fundamental bound on performance of non-intrusive load monitoring
algorithms with application to smart-meter privacy, IFAC-PapersOnLine, Vol. 53, No.
2, pp. 2280-2285
Dong W., Wang Y., Zhou L., et al. , 2019, An Anti-Eavesdropping Method in Data Collection
of Smart Meter, Computer and communication 2019, Vol. 7, No. 9, pp. 38-49
Karimi-Bidhendi S., Guo J., Jafarkhani H., 2020, Energy-efficient node deployment
in heterogeneous two-tier wireless sensor network s with limited communication range,
IEEE Transactions on Wireless Communications, Vol. 99, pp. 1-7
Tokhy M., 2022, Capacity analysis of cognitive radio wireless sensor network under
optimal power allocation in imperfect channel, Wireless Networks, Vol. 28, No. 4,
pp. 1805-1826
Refaee E. A., Shamsudheen S., 2021, Trust- and energy-aware cluster head selection
in a uav-based wireless sensor network using fit-fcm, The Journal of Supercomputing,
Vol. 78, No. 4, pp. 5610-5625
Kanwar V., Kumar A., 2021, Dv-hop localization methods for displaced sensor nodes
in wireless sensor network using pso, Wireless Networks, Vol. 27, No. 1, pp. 91-102
Umashankar M. L., 2021, An efficient hybrid model for cluster head selection to optimize
wireless sensor network using simulated annealing algorithm, Indian Journal of Science
and Technology, Vol. 14, No. 3, pp. 270-288
Vineeth V. V., Ambrish V., Haricharann D. V., Harshini V., Abilash C., 2021, Power
theft recognition and data security in smart meter reading of a smart grid, Journal
of Physics: Conference Series, Vol. 1916, No. 1, pp. 012216 (3pp)
Heidari-Akhijahani A., Safdarian A., Aminifar F., 2021, Phase identification of single-phase
customers and pv panels via smart meter data, IEEE Transactions on Smart Grid, Vol.
12, No. 5, pp. 4543-4552
Alubodi A., Al-Mashhadani I., Mahdi D., 2021, Design and implementation of a zigbee,
bluetooth, and gsm-based smart meter smart grid, IOP Conference Series: Materials
Science and Engineering, Vol. 1067, No. 1, pp. 012130 (10)
Qiao S., Qian X., Zhou M., Wang Y., 2021, Analysis on the maintainability of smart
meter software for full lifetime, Journal of Physics Conference Series, Vol. pp. 1748,
No. 4, pp. 042025
Lorenz Pascal, Schott Rene, Staples G., etal. , 2019, New Path Centrality Based on
Operator Calculus Approach for Wireless Sensor Network Deployment, IEEE Transactions
on Emerging Topics in Computing, Vol. 7, No. 1, pp. 162-173
Ajmi N., Helali A., Lorenz P., et al. , 2021, MWCSGA-Multi Weight Chicken Swarm Based
Genetic Algorithm for Energy Efficient Clustered Wireless Sensor Network, Sensors,
Vol. 3, pp. 791-780
Prabu P., Senthilnathan T., 2020, Secured and flexible user authentication protocol
for wireless sensor network, International Journal of Intelligent Unmanned Systems,
Vol. 8, No. 4, pp. 253-265
Author
Wenyu Liu graduated from the School of Business Administration of Northeastern
University in 2012 with a bachelor's degree. In 2016, he graduated from the School
of Software Engineering of Northeastern Univer-sity with a master's degree. At present,
He working in the Marketing Service Center of State Grid Liaoning Electric Power Co.,
LTD., engaged in production plan management, verification data verification, inspection
and supervision of energy meter disassembly and back inspection, energy meter operation
error detection, energy meter operation data collection, orderly power consumption
guarantee and other work. He has published several papers in Chinese core journals
and EI conference papers. His interests include big data algorithms, high frequency
data acquisition and data information security.
Haishan Zhou received his Bachelor's degree in engineering from North China Electric
Power University, China in 2011. He received his master's degree in Electrical engi-neering
from North China Electric Power University, China. In 2019. Currently, he is working
in State Grid liaoning Marketing Service Center, State Grid Corporation of china,
responsible for metering asset management in the whole province. His areas of interest
include electricity metering, electricity meters related fields.
Muxin Zhang received a bachelor's degree in engineering from Northeast University
of China in 2015. Presently, he is working on electric energy measurement in the marketing
service center of State Grid Liaoning Electric Power Co., Ltd. His field of interest
is power metering and data mining.
Ying Shang graduated from Northeast Electric Power University with a bachelor's
degree in engineering in 2011. In 2019, She obtained a master's degree in North China
Electric Power University. At present, she works in the marketing service center of
Liaoning Electric Power Co., Ltd. in electric energy metering, electric energy information
collection, anti-stealing monitoring, etc. She has published many papers during the
Chinese Core Journaland EI conference papers. Her areas of interest include big data
algorithms, high-frequency data acquisition, and data information security.
Yudan Liu received her BE in Electrical Engineering from Shenyang Institute of
Engineering, 2016. Presently, she works for State Grid Liaoning Electrich Power Supply
Com. Ltd. Her areas of interest include electrical metrology and published articles
in 3 international reputed peer reviewed journals and conferences proceedings.