CN115175306A - Electric power Internet of things indoor positioning method based on convolutional neural network - Google Patents

Electric power Internet of things indoor positioning method based on convolutional neural network Download PDF

Info

Publication number
CN115175306A
CN115175306A CN202210729213.2A CN202210729213A CN115175306A CN 115175306 A CN115175306 A CN 115175306A CN 202210729213 A CN202210729213 A CN 202210729213A CN 115175306 A CN115175306 A CN 115175306A
Authority
CN
China
Prior art keywords
rssi
neural network
convolutional neural
sample
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202210729213.2A
Other languages
Chinese (zh)
Other versions
CN115175306B (en
Inventor
宋文卓
郭夫然
张亮
宋景博
汪赟
姚晗
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Corp of China SGCC
Economic and Technological Research Institute of State Grid Henan Electric Power Co Ltd
Original Assignee
State Grid Corp of China SGCC
Economic and Technological Research Institute of State Grid Henan Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by State Grid Corp of China SGCC, Economic and Technological Research Institute of State Grid Henan Electric Power Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN202210729213.2A priority Critical patent/CN115175306B/en
Publication of CN115175306A publication Critical patent/CN115175306A/en
Application granted granted Critical
Publication of CN115175306B publication Critical patent/CN115175306B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • H04W64/006—Locating users or terminals or network equipment for network management purposes, e.g. mobility management with additional information processing, e.g. for direction or speed determination
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y10/00—Economic sectors
    • G16Y10/35—Utilities, e.g. electricity, gas or water
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y40/00—IoT characterised by the purpose of the information processing
    • G16Y40/60—Positioning; Navigation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/309—Measuring or estimating channel quality parameters
    • H04B17/318—Received signal strength
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings

Abstract

The invention discloses an electric power Internet of things indoor positioning method based on a convolutional neural network, which comprises the following steps: s1, in an off-line fingerprint library construction stage, generating a probability density function graph for each acquisition position by using a kernel density estimation method for constructing an RSSI statistical sample fingerprint library; s2, in a training stage, training the convolutional neural network by using a statistical sample fingerprint library; and S3, in an online positioning stage, predicting the position of the equipment based on a convolutional neural network algorithm according to the RSSI sample acquired by the equipment needing to be positioned. The spatial statistical distribution characteristic of the signal intensity and the spatial feature extraction advantage of the convolutional neural network are fully utilized, and an accurate and stable indoor positioning technology can be provided in the application scene of the power internet of things.

Description

Electric power Internet of things indoor positioning method based on convolutional neural network
Technical Field
The invention relates to the technical field of wireless sensor networks, in particular to an indoor positioning method of an electric power internet of things based on a convolutional neural network.
Background
With the continuous progress of technologies such as wireless communication and microelectronics, the application of the internet of things in a power system is more and more extensive. Various wireless communication technologies are currently applied to power systems, such as LoRa, NB-IoT, OFDM, etc. In an electric power system, such as a power plant, a substation, etc., many devices are indoors, and position information of the devices is needed when the system is operated. However, the traditionally used outdoor positioning technology, such as GPS, is affected by obstacles such as walls and is difficult to apply to indoor environments, so indoor positioning in the power internet of things becomes a problem to be solved urgently.
Indoor positioning is subject to more and more domestic and foreign researches. RSSI is the most commonly used indicator in indoor positioning because it can be easily obtained in many wireless communication devices (e.g., loRa, NB-IoT, zigBee, OFDM). Although there are various indoor positioning methods based on RSSI, the accuracy, stability and the like of these methods need to be improved, and in a power system, there are a lot of problems of interference, noise, shielding and the like, so that a high-accuracy and stable indoor positioning method oriented to the power internet of things is urgently needed.
Disclosure of Invention
In order to overcome the defects, the invention aims to provide the electric power internet of things indoor positioning method based on the convolutional neural network, which makes full use of the spatial statistical distribution characteristic of the signal intensity and the spatial feature extraction advantage of the convolutional neural network, and can provide an accurate and stable indoor positioning technology in the electric power internet of things application scene.
In order to realize the purpose, the invention adopts the following technical scheme:
an electric power internet of things indoor positioning method based on a convolutional neural network comprises the following steps:
s1, in an off-line fingerprint library construction stage, generating a probability density function graph for each acquisition position by using a kernel density estimation method for constructing an RSSI statistical sample fingerprint library;
s2, in a training stage, training the convolutional neural network by using a statistical sample fingerprint library;
and S3, in an online positioning stage, predicting the position of the equipment based on a convolutional neural network algorithm according to the RSSI sample acquired by the equipment needing to be positioned.
Optionally, the step of the offline fingerprint database construction phase includes:
s11, let L = (L) 1 ,…,l n ) Position acquisition for n RSSI's, where l i ＝(x i ,y i ,z i ) Is a position l i Three-dimensional coordinates of (a); let A = (a) 1 ,…,a h ) To locate a set of APs within a region, wherein a i Is the ith AP; at any RSSI acquisition location l i Collecting m RSSI samples and recording the RSSI sample set as
Figure BDA0003712258430000021
Wherein
Figure BDA0003712258430000022
For the jth RSSI sample,
Figure BDA0003712258430000023
Obtaining the RSSI value of the kth AP at the acquisition position in the jth sample; after the RSSI sample is collected for each position to be collected, the construction is finishedRaw RSSI sample library
Figure BDA0003712258430000024
S12, giving a raw RSSI sample library R, and acquiring samples R of positions aiming at any RSSI i (i is more than or equal to 1 and less than or equal to n) training a kernel density estimator:
s121, given
Figure BDA0003712258430000025
Order to
Figure BDA0003712258430000026
Needs to acquire location l at RSSI for each AP i Training a kernel density estimator, wherein the kernel density estimator f of the kth AP i,k (r) is given by:
Figure BDA0003712258430000027
wherein K (-) is a kernel function and u is a smoothing parameter; estimating by adopting a Gaussian kernel, and determining a smoothing parameter by adopting a maximum likelihood estimation method; thus, at RSSI acquisition location l i There are h kernel density estimators noted as F i ＝(f i,1 (r),…,f i,h (r))；
S13, order r min And r max Setting the minimum value and the maximum value of the RSSI which can be received, and setting lambda as a step length parameter; then a horizontal axis interval sample point X = (r) of RSSI probability density distribution can be created min ,r min +λ,r min +2λ,…,r max ) (ii) a For any RSSI acquisition position l i The following operations are performed:
s131, for any AP a k (k is more than or equal to 1 and less than or equal to h), utilizing a kernel density estimator f i,k (r) generating respective probabilities Y of all sample points in X i,k ＝(f i,k (r min ),f i,k (r min +λ),f i,k (r min +2λ),…,f i,k (r max ) ); then, let X = (r) min ,r min +λ,r min +2λ,…,r max ) Is a horizontal axis coordinate, Y i,k ＝(f i,k (r min ),f i,k (r min +λ),f i,k (r min +2λ),…,f i,k (r max ) ) is a vertical axis coordinate, a line graph G is generated i,k (ii) a Obviously, any RSSI acquisition position l i There are h broken line graphs, which are recorded as G i ＝(G i,1 ,…,G i,h )；
S132, after the operation of S131 is performed on all RSSI acquisition positions, a final offline RSSI fingerprint database G = (G) can be obtained 1 ,…,G n )。
Optionally, the step of the training phase includes:
s21, constructing a convolutional neural network training set by utilizing an offline RSSI fingerprint database G, and acquiring a position l for any RSSI i Taking i as a label, transversely splicing h fold lines to form an image and recording the image as P i As characteristic data, the training data at the acquisition position is recorded as D i ＝(i,P i ). Finally obtaining a training set D = (D) 1 ,…D n )；
S22, building a convolutional neural network n classification by using the existing architecture, and recording the classifier as f;
and S23, training the classifier f on the training set D.
Optionally, the online positioning stage includes:
s31, continuously collecting m RSSI samples for a certain device d to be positioned, and recording the set of the RSSI samples as
Figure BDA0003712258430000031
Wherein
Figure BDA0003712258430000032
For the jth RSSI sample,
Figure BDA0003712258430000033
Obtaining the RSSI value of the kth AP at the acquisition position in the jth sample;
s32, needing to locate each AP to be detectedTraining a kernel density estimator on device d, wherein kernel density estimator f of kth AP d,k (r) is given by:
Figure BDA0003712258430000041
wherein K (-) is a kernel function and u is a smoothing parameter; considering that the signal intensity presents Gaussian distribution, the method adopts a Gaussian kernel to estimate, and the smooth parameter is determined by adopting a maximum likelihood estimation method; thus, there are h kernel density estimators, denoted as F, on the device d to be positioned d ＝(f d,1 (r),…,f d,h (r))；
S33, for any AP a k (k is more than or equal to 1 and less than or equal to h), utilizing a nuclear density estimator f d,k (r) generating respective probabilities Y of all sample points in X d,k ＝(f d,k (r min ),f d,k (r min +λ),f d,k (r min +2λ),…,f d,k (r max ) ); then, let X = (r) min ,r min +λ,r min +2λ,…,r max ) Is a horizontal axis coordinate, Y d,k ＝(f d,k (r min ),f d,k (r min +λ),f d,k (r min +2λ),…,f d,k (r max ) ) is a vertical axis coordinate, a line graph G is generated d,k (ii) a H broken line graphs are arranged on the equipment d to be positioned, and the broken line graphs are recorded as G d ＝(G d,1 ,…,G d,h )；
And S34, positioning by using the trained convolutional neural network f.
Optionally, the step of performing positioning by using the trained convolutional neural network f includes:
s341, adding G d The h broken line graphs in the middle are transversely spliced to form an image which is marked as P d As characteristic data;
s342, adding P d F is transmitted in for classification, K positions with the maximum f output layer softmax function estimation value are taken, and the K positions with the maximum probability are recorded as
Figure BDA0003712258430000042
Wherein
Figure BDA0003712258430000043
Figure BDA0003712258430000044
Are respectively as
Figure BDA0003712258430000045
Three-dimensional coordinates of (a);
s343, the final positioning result of the equipment d to be positioned is
Figure BDA0003712258430000046
The invention has the following positive beneficial effects:
the method aims at the characteristics of complex radio frequency environment, serious interference and the like of the power internet of things, utilizes the convolutional neural network to extract the spatial distribution characteristics of the signal intensity, and is used for accurate and robust indoor positioning of the power equipment. The method comprises the following steps: the method comprises an off-line fingerprint database construction stage, a training stage and an on-line positioning stage. The spatial statistical distribution characteristic of the signal intensity and the spatial feature extraction advantage of the convolutional neural network are fully utilized, and an accurate and stable indoor positioning technology can be provided in the application scene of the power internet of things.
Drawings
Fig. 1 is a schematic block diagram of an indoor positioning method of an electric power internet of things based on a convolutional neural network according to embodiment 1 of the present invention;
fig. 2 is a schematic diagram of acquiring an original RSSI provided in step S11 in embodiment 1 of the present invention;
fig. 3 is a schematic diagram of line graph generation provided in step S131 in embodiment 1 of the present invention;
fig. 4 is a schematic diagram of training set generation provided in step S21 in embodiment 1 of the present invention;
fig. 5 is a schematic diagram of generating a final positioning of an output layer in step S342 according to embodiment 1 of the present invention.
Detailed Description
The invention will be further described with reference to some specific embodiments.
Example 1
As shown in fig. 1 to 5, an electric power internet of things indoor positioning method based on a convolutional neural network includes the steps:
s1, in an off-line fingerprint library construction stage, generating a probability density function graph for each acquisition position by using a kernel density estimation method for constructing an RSSI statistical sample fingerprint library;
s2, in a training stage, training the convolutional neural network by using a statistical sample fingerprint library;
and S3, in an online positioning stage, predicting the position of the equipment based on a convolutional neural network algorithm according to the RSSI sample acquired by the equipment needing to be positioned.
According to the method, aiming at the characteristics of complex radio frequency environment, serious interference and the like of the power internet of things, the spatial distribution characteristics of the signal intensity are extracted by using the convolutional neural network, the method is used for accurate and robust indoor positioning of the power equipment, the spatial statistical distribution characteristics of the signal intensity and the spatial characteristic extraction advantages of the convolutional neural network are fully utilized, and accurate and stable indoor positioning technology can be provided in the application scene of the power internet of things.
The step of the off-line fingerprint database construction phase comprises the following steps:
s11, let L = (L) 1 ,…,l n ) Position acquisition for n RSSI's, where l i ＝(x i ,y i ,z i ) Is a position l i Three-dimensional coordinates of (a); let A = (a) 1 ,…,a h ) To locate a set of APs within a region, wherein a i Is the ith AP; at any RSSI acquisition location l i Collecting m RSSI samples and recording the RSSI sample set as
Figure BDA0003712258430000061
Wherein
Figure BDA0003712258430000062
For the jth RSSI sample,
Figure BDA0003712258430000063
Obtaining the RSSI value of the kth AP at the acquisition position in the jth sample; after the RSSI sample is collected for each position to be collected, an original RSSI sample library is constructed
Figure BDA0003712258430000064
S12, giving a raw RSSI sample library R, and acquiring samples R of positions aiming at any RSSI i (i is more than or equal to 1 and less than or equal to n) training a kernel density estimator:
s121, given
Figure BDA0003712258430000065
Order to
Figure BDA0003712258430000066
Needs to acquire location l at RSSI for each AP i Last train a kernel density estimator, wherein kernel density estimator f of kth AP i,k (r) is given by:
Figure BDA0003712258430000067
wherein K (-) is a kernel function and u is a smoothing parameter; considering that the signal intensity presents Gaussian distribution, a Gaussian kernel is adopted for estimation, and a maximum likelihood estimation method is adopted for determining a smoothing parameter; thus, at RSSI acquisition location l i There are h kernel density estimators noted as F i ＝(f i,1 (r),…,f i,h (r))；
S13, order r min And r max Taking the minimum value and the maximum value of the RSSI which can be received, and taking lambda as a step length parameter; then a horizontal axis interval sample point X = (r) of RSSI probability density distribution can be created min ,r min +λ,r min +2λ,…,r max ) (ii) a For any RSSI acquisition position l i Performing the following operations:
s131, for any AP a k (k is more than or equal to 1 and less than or equal to h), utilizing a nuclear density estimator f i,k (r) generating respective probabilities Y of all sample points in X i,k ＝(f i,k (r min ),f i,k (r min +λ),f i,k (r min +2λ),…,f i,k (r max ) ); then, let X = (r) min ,r min +λ,r min +2λ,…,r max ) Is a horizontal axis coordinate, Y i,k ＝(f i,k (r min ),f i,k (r min +λ),f i,k (r min +2λ),…,f i,k (r max ) ) is a vertical axis coordinate, a line graph G is generated i,k (ii) a Obviously, any RSSI acquisition location/ i There are h broken line graphs, which are recorded as G i ＝(G i,1 ,…,G i,h )；
S132, after the operation of S131 is performed on all RSSI acquisition positions, a final offline RSSI fingerprint database G = (G) can be obtained 1 ,…,G n )。
Most of the existing indoor positioning algorithms are used for positioning based on mean values, and statistics and spatial distribution characteristics of RSSI samples are ignored, so that the statistical characteristics of original samples are extracted by utilizing a kernel density estimation method aiming at each RSSI acquisition position of spatial distribution.
The step of the training phase comprises:
s21, constructing a convolutional neural network training set by utilizing an offline RSSI fingerprint database G, specifically: for any RSSI acquisition position l i Taking i as a label, transversely splicing h fold lines to form an image and recording the image as P i As characteristic data, the training data at the acquisition position is recorded as D i ＝(i,P i ). Finally obtaining a training set D = (D) 1 ,…D n )；
S22, building a convolutional neural network n classification by using an existing architecture (such as AlexNet, VGGNet and the like), and recording the classifier as f;
and S23, training the classifier f on the training set D.
The step of the online positioning phase comprises:
s31, continuously collecting m RSSI samples for a certain device d to be positioned, and recording the set of the RSSI samples as
Figure BDA0003712258430000081
Wherein
Figure BDA0003712258430000082
For the jth RSSI sample,
Figure BDA0003712258430000083
Obtaining the RSSI value of the kth AP at the acquisition position in the jth sample;
s32, training a kernel density estimator on the device d to be positioned for each AP, wherein the kernel density estimator f of the kth AP d,k (r) is given by:
Figure BDA0003712258430000084
wherein K (-) is a kernel function and u is a smoothing parameter; considering that the signal intensity presents Gaussian distribution, the method adopts a Gaussian kernel to estimate, and the smooth parameter is determined by adopting a maximum likelihood estimation method; thus, there are h kernel density estimators, denoted as F, on the device d to be positioned d ＝(f d,1 (r),…,f d,h (r))；
S33, for any AP a k (k is more than or equal to 1 and less than or equal to h), utilizing a nuclear density estimator f d,k (r) generating respective probabilities Y of all sample points in X d,k ＝(f d,k (r min ),f d,k (r min +λ),f d,k (r min +2λ),…,f d,k (r max ) ); then, let X = (r) min ,r min +λ,r min +2λ,…,r max ) Is a horizontal axis coordinate, Y d,k ＝(f d,k (r min ),f d,k (r min +λ),f d,k (r min +2λ),…,f d,k (r max ) ) is a vertical axis coordinate, a line graph G is generated d,k (ii) a H broken line graphs are arranged on the equipment d to be positioned, and the broken line graphs are recorded as G d ＝(G d,1 ,…,G d,h )；
And S34, positioning by using the trained convolutional neural network f.
The step of positioning by using the trained convolutional neural network f comprises the following steps:
s341, adding G d The h fold lines in the middle are transversely spliced to form an image marked as P d As characteristic data;
s342, adding P d F is transmitted in for classification, the classification result is not taken as an estimation position, but K positions with the maximum f output layer softmax function estimation value are taken, and the K positions with the maximum probability are respectively recorded as
Figure BDA0003712258430000085
Wherein
Figure BDA0003712258430000086
Figure BDA0003712258430000087
Are respectively as
Figure BDA0003712258430000088
Three-dimensional coordinates of (a);
s343, the final positioning result of the equipment d to be positioned is
Figure BDA0003712258430000091
The method aims at the characteristics of complex radio frequency environment, serious interference and the like of the power internet of things, utilizes the convolutional neural network to extract the spatial distribution characteristics of the signal intensity, and is used for accurate and robust indoor positioning of the power equipment. The spatial statistical distribution characteristic of the signal intensity and the spatial feature extraction advantage of the convolutional neural network are fully utilized, and an accurate and stable indoor positioning technology can be provided in the application scene of the power internet of things.
The concrete expression is as follows:
(1) Different from the existing method which directly adopts the mean value for estimation, the method effectively extracts the probability statistical characteristics of the signal intensity at each sampling point in the space by utilizing the kernel density estimation method, and reserves more characteristics for positioning as much as possible in an image mode for a trained probability statistical model.
(2) In addition, the method does not directly use the result of the convolutional neural network for positioning, but uses the average value of K positions with the maximum output layer softmax function estimation value to position, thereby improving the stability of the positioning system.
Finally, the above embodiments are only used for illustrating the technical solutions of the present invention and not for limiting, and other modifications or equivalent substitutions made by the technical solutions of the present invention by those of ordinary skill in the art should be covered within the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solutions of the present invention.

Claims (5)

1. An electric power Internet of things indoor positioning method based on a convolutional neural network is characterized by comprising the following steps:
s1, in an off-line fingerprint library construction stage, generating a probability density function graph for each acquisition position by using a kernel density estimation method for constructing an RSSI statistical sample fingerprint library;
s2, in a training stage, training the convolutional neural network by using a statistical sample fingerprint library;
and S3, in an online positioning stage, predicting the position of the equipment based on a convolutional neural network algorithm according to the RSSI sample acquired by the equipment needing to be positioned.
2. The electric power internet of things indoor positioning method based on the convolutional neural network as claimed in claim 1, wherein the step of the offline fingerprint database construction stage comprises:
s11, let L = (L) 1 ,…,l n ) Position acquisition for n RSSI's, where l i ＝(x i ,y i ,z i ) Is a position l i Three-dimensional coordinates of (a); let A = (a) 1 ,…,a h ) To locate a set of APs within a region, wherein a i Is the ith AP; at any RSSI acquisition location l i Collecting m RSSI samples and recording the RSSI sample set as
Figure FDA0003712258420000011
Wherein
Figure FDA0003712258420000012
For the jth RSSI sample,
Figure FDA0003712258420000013
Obtaining the RSSI value of the kth AP at the acquisition position in the jth sample; after the RSSI sample is collected for each position to be collected, an original RSSI sample library is constructed
Figure FDA0003712258420000014
S12, giving a raw RSSI sample library R, and acquiring samples R of positions aiming at any RSSI i (i is more than or equal to 1 and less than or equal to n) training a kernel density estimator:
s121, given
Figure FDA0003712258420000015
Order to
Figure FDA0003712258420000016
Needs to acquire location l at RSSI for each AP i Training a kernel density estimator, wherein the kernel density estimator f of the kth AP i,k (r) is given by:
Figure FDA0003712258420000021
wherein K (-) is a kernel function and u is a smoothing parameter; estimating by adopting a Gaussian kernel, and determining a smoothing parameter by adopting a maximum likelihood estimation method; thus, at RSSI acquisition location l i There are h kernel density estimators noted as F i ＝(f i,1 (r),…,f i,h (r))；
S13, order r min And r max Taking the minimum value and the maximum value of the RSSI which can be received, and taking lambda as a step length parameter; then a horizontal axis interval sample point X = (r) of RSSI probability density distribution can be created min ,r min +λ,r min +2λ,…,r max ) (ii) a For any RSSI acquisition position l i Performing the following operations:
s131 for any AP a k (k is more than or equal to 1 and less than or equal to h), utilizing a nuclear density estimator f i,k (r) generating respective probabilities Y of all sample points in X i,k ＝(f i,k (r min ),f i,k (r min +λ),f i,k (r min +2λ),…,f i,k (r max ) ); then, let X = (r) min ,r min +λ,r min +2λ,…,r max ) Is a horizontal axis coordinate, Y i,k ＝(f i,k (r min ),f i,k (r min +λ),f i,k (r min +2λ),…,f i,k (r max ) ) is a vertical axis coordinate, a line graph G is generated i,k (ii) a Obviously, any RSSI acquisition location/ i There are h broken line graphs, which are recorded as G i ＝(G i,1 ,…,G i,h )；
S132, after the operation of S131 is performed on all RSSI collecting positions, a final offline RSSI fingerprint database G = (G) may be obtained 1 ,…,G n )。
3. The electric power internet of things indoor positioning method based on the convolutional neural network as claimed in claim 1, wherein the training stage comprises the steps of:
s21, constructing a convolutional neural network training set by utilizing an offline RSSI fingerprint database G, and acquiring a position l for any RSSI i Using i as its label, and transversely splicing h fold lines to form an image P i As characteristic data, the training data at the acquisition position is recorded as D i ＝(i,P i ). Finally obtaining a training set D = (D) 1 ,…D n )；
S22, building a convolutional neural network n classification by using the existing architecture, and recording the classifier as f;
and S23, training the classifier f on the training set D.
4. The indoor positioning method for the power internet of things based on the convolutional neural network as claimed in claim 1, wherein the step of the online positioning stage comprises:
s31, continuously collecting m RSSI samples for a certain device d to be positioned, and recording the set of the RSSI samples as
Figure FDA0003712258420000031
Wherein
Figure FDA0003712258420000032
For the jth RSSI sample,
Figure FDA0003712258420000033
Obtaining the RSSI value of the kth AP at the acquisition position in the jth sample;
s32, training a kernel density estimator on the device d to be positioned for each AP, wherein the kernel density estimator f of the kth AP d,k (r) is given by:
Figure FDA0003712258420000034
wherein K (-) is a kernel function and u is a smoothing parameter; considering that the signal intensity presents Gaussian distribution, the method adopts a Gaussian kernel to estimate, and the smooth parameter is determined by adopting a maximum likelihood estimation method; thus, there are h kernel density estimators, denoted as F, on the device d to be positioned d ＝(f d,1 (r),…,f d,h (r))；
S33, for any AP a k (k is more than or equal to 1 and less than or equal to h), utilizing a nuclear density estimator f d,k (r) generating respective probabilities Y of all sample points in X d,k ＝(f d,k (r min ),f d,k (r min +λ),f d,k (r min +2λ),…,f d,k (r max ) ); then, let X = (r) min ,r min +λ,r min +2λ,…,r max ) Is a horizontal axis coordinate, Y d,k ＝(f d,k (r min ),f d,k (r min +λ),f d,k (r min +2λ),…,f d,k (r max ) ) is a vertical axis coordinate, a line graph G is generated d,k (ii) a H broken line graphs are arranged on the equipment d to be positioned, and the broken line graphs are recorded as G d ＝(G d,1 ,…,G d,h )；
And S34, positioning by using the trained convolutional neural network f.
5. The electric power internet of things indoor positioning method based on the convolutional neural network as claimed in claim 4, wherein the step of positioning by using the trained convolutional neural network f comprises the following steps:
s341, adding G d The h broken line graphs in the middle are transversely spliced to form an image which is marked as P d As characteristic data;
s342, adding P d F is transmitted in for classification, K positions with the maximum estimation value of the softmax function of the f output layer are taken, and the K positions with the maximum probability are recorded as
Figure FDA0003712258420000035
Wherein
Figure FDA0003712258420000036
Are respectively as
Figure FDA0003712258420000041
Three-dimensional coordinates of (a);
s343, the final positioning result of the equipment d to be positioned is
Figure FDA0003712258420000042
CN202210729213.2A 2022-06-24 2022-06-24 Indoor positioning method of electric power Internet of things based on convolutional neural network Active CN115175306B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210729213.2A CN115175306B (en) 2022-06-24 2022-06-24 Indoor positioning method of electric power Internet of things based on convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210729213.2A CN115175306B (en) 2022-06-24 2022-06-24 Indoor positioning method of electric power Internet of things based on convolutional neural network

Publications (2)

Publication Number Publication Date
CN115175306A true CN115175306A (en) 2022-10-11
CN115175306B CN115175306B (en) 2024-05-07

Family

ID=83486988

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210729213.2A Active CN115175306B (en) 2022-06-24 2022-06-24 Indoor positioning method of electric power Internet of things based on convolutional neural network

Country Status (1)

Country Link
CN (1) CN115175306B (en)

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108282743A (en) * 2018-03-05 2018-07-13 桂林理工大学 Indoor orientation method, apparatus and system
CN108924737A (en) * 2018-06-20 2018-11-30 北京三快在线科技有限公司 Localization method, device, equipment and computer readable storage medium
US10620290B1 (en) * 2019-02-19 2020-04-14 Ideal Industries Lighting Llc Indoor positioning system
US10716089B1 (en) * 2019-06-03 2020-07-14 Mapsted Corp. Deployment of trained neural network based RSS fingerprint dataset
US20200285464A1 (en) * 2017-06-05 2020-09-10 Umajin Inc. Location tracking system and methods
CN111669820A (en) * 2020-05-27 2020-09-15 南京邮电大学 Density peak value abnormity detection method and intelligent passive indoor positioning method
CN112153736A (en) * 2020-09-14 2020-12-29 南京邮电大学 Personnel action identification and position estimation method based on channel state information
CN112329685A (en) * 2020-11-16 2021-02-05 常州大学 Method for detecting crowd abnormal behaviors through fusion type convolutional neural network
US20210092611A1 (en) * 2019-09-19 2021-03-25 Colorado State University Research Foundation Security-enhanced Deep Learning Fingerprint-Based Indoor Localization
CN112712557A (en) * 2020-12-17 2021-04-27 上海交通大学 Super-resolution CIR indoor fingerprint positioning method based on convolutional neural network
WO2021103027A1 (en) * 2019-11-30 2021-06-03 Beijing Didi Infinity Technology And Development Co., Ltd. Base station positioning based on convolutional neural networks
CN113316080A (en) * 2021-04-19 2021-08-27 北京工业大学 Indoor positioning method based on Wi-Fi and image fusion fingerprint
CN113453148A (en) * 2021-06-25 2021-09-28 南通大学 Indoor position fingerprint positioning method combining deep learning and weighted K-neighbor algorithm
US20220014933A1 (en) * 2020-07-09 2022-01-13 Industry Foundation Of Chonnam National University Deep learning-based beamforming communication system and method
CN114449452A (en) * 2021-12-29 2022-05-06 华南师范大学 Indoor positioning algorithm for heterogeneous equipment

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200285464A1 (en) * 2017-06-05 2020-09-10 Umajin Inc. Location tracking system and methods
CN108282743A (en) * 2018-03-05 2018-07-13 桂林理工大学 Indoor orientation method, apparatus and system
CN108924737A (en) * 2018-06-20 2018-11-30 北京三快在线科技有限公司 Localization method, device, equipment and computer readable storage medium
US10620290B1 (en) * 2019-02-19 2020-04-14 Ideal Industries Lighting Llc Indoor positioning system
US10716089B1 (en) * 2019-06-03 2020-07-14 Mapsted Corp. Deployment of trained neural network based RSS fingerprint dataset
US20210092611A1 (en) * 2019-09-19 2021-03-25 Colorado State University Research Foundation Security-enhanced Deep Learning Fingerprint-Based Indoor Localization
WO2021103027A1 (en) * 2019-11-30 2021-06-03 Beijing Didi Infinity Technology And Development Co., Ltd. Base station positioning based on convolutional neural networks
CN111669820A (en) * 2020-05-27 2020-09-15 南京邮电大学 Density peak value abnormity detection method and intelligent passive indoor positioning method
US20220014933A1 (en) * 2020-07-09 2022-01-13 Industry Foundation Of Chonnam National University Deep learning-based beamforming communication system and method
CN112153736A (en) * 2020-09-14 2020-12-29 南京邮电大学 Personnel action identification and position estimation method based on channel state information
CN112329685A (en) * 2020-11-16 2021-02-05 常州大学 Method for detecting crowd abnormal behaviors through fusion type convolutional neural network
CN112712557A (en) * 2020-12-17 2021-04-27 上海交通大学 Super-resolution CIR indoor fingerprint positioning method based on convolutional neural network
CN113316080A (en) * 2021-04-19 2021-08-27 北京工业大学 Indoor positioning method based on Wi-Fi and image fusion fingerprint
CN113453148A (en) * 2021-06-25 2021-09-28 南通大学 Indoor position fingerprint positioning method combining deep learning and weighted K-neighbor algorithm
CN114449452A (en) * 2021-12-29 2022-05-06 华南师范大学 Indoor positioning algorithm for heterogeneous equipment

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
SHIXUN WU; WENXIAN HUANG; MIN LI; KAI XU: ""A Novel RSSI Fingerprint Positioning Method Based on Virtual AP and Convolutional Neural Network"", 《IEEE SENSORS JOURNAL》, 22 February 2022 (2022-02-22) *
YU LIN; YAO TONG; QINKUN ZHONG; RUIPENG GAO; BUYI YIN; LEI LIU; LI MA;: ""DCCP: Deep Convolutional Neural Networks for Cellular Network Positioning"", 《2021 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM)》, 2 February 2022 (2022-02-02) *
刘帅;王旭东;吴楠: ""一种基于卷积神经网络的CSI指纹室内定位方法"", 《工程科学学报》, 12 October 2021 (2021-10-12) *
张亮;郭彤旭;王佐强;郭庆楠;雷洛: ""一种单目视觉的水下目标定位与跟踪技术"", 《大众标准化》, 23 May 2021 (2021-05-23) *
曹建荣;张旭;武欣莹;吕俊杰;杨红娟: ""结合CNN和WiFi指纹库的室内定位算法"", 《计算机系统应用》, 15 July 2020 (2020-07-15) *

Also Published As

Publication number Publication date
CN115175306B (en) 2024-05-07

Similar Documents

Publication Publication Date Title
CN112533149B (en) Moving target positioning algorithm based on UWB mobile node
CN108174343B (en) Wireless positioning method and system for power indoor communication operation and maintenance scene
CN109640269B (en) Fingerprint positioning method based on CSI and time domain fusion algorithm
CN114449452B (en) Wi-Fi indoor positioning method based on CNN-RNN
CN112881979B (en) Initial state self-adaptive fusion positioning method based on EKF filtering
CN110388926B (en) Indoor positioning method based on mobile phone geomagnetism and scene image
CN113610243A (en) Atmospheric pollutant tracing method based on coupled machine learning and correlation analysis
CN112040405A (en) Indoor positioning method based on kernel limit learning machine and particle filtering
CN105657653B (en) Indoor positioning method based on fingerprint data compression
CN109005390B (en) Method and system for establishing personnel distribution model based on signal intensity and video
CN109121081B (en) Indoor positioning method based on position candidate set and EM algorithm
CN114236480A (en) Airborne platform sensor system error registration algorithm
CN115175306A (en) Electric power Internet of things indoor positioning method based on convolutional neural network
CN109803234B (en) Unsupervised fusion positioning method based on weight importance constraint
CN107656905A (en) A kind of air quality data real-time calibration method using error propagation
CN108519579B (en) WiFi fingerprint positioning method for analyzing optimal AP based on interval overlapping degree
CN114710744B (en) Indoor positioning method integrating WiFi ranging and PDR calculation in depth
CN115905793A (en) Track reproduction method based on Internet of things and curve fitting
CN115908051A (en) Method for determining energy storage capacity of power system
CN112040408B (en) Multi-target accurate intelligent positioning and tracking method suitable for supervision places
CN113271542B (en) Indoor mobile terminal positioning method based on Bluetooth and visible light
CN112887904B (en) WiFi positioning method suitable for intelligent stadium
CN114279311A (en) GNSS deformation monitoring method and system based on inertia
CN114584230A (en) Predictive channel modeling method based on countermeasure network and long-short term memory network
CN114286282A (en) Fingerprint positioning method based on WiFi RSS data of mobile phone and dimension reduction algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant