CN109729540B - Base station parameter automatic optimization method based on neural network - Google Patents
Base station parameter automatic optimization method based on neural network Download PDFInfo
- Publication number
- CN109729540B CN109729540B CN201910048705.3A CN201910048705A CN109729540B CN 109729540 B CN109729540 B CN 109729540B CN 201910048705 A CN201910048705 A CN 201910048705A CN 109729540 B CN109729540 B CN 109729540B
- Authority
- CN
- China
- Prior art keywords
- base station
- ring
- configuration
- cell
- neuron
- 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.)
- Active
Links
Images
Abstract
The invention relates to a base station parameter automatic optimization method based on a neural network, which comprises the following steps: step S1, scanning the MRE data of the total base station, and respectively outputting the base station configuration with the ring switching problem and the base station configuration without the ring switching problem, and step S2, taking the base station configuration set without the ring switching problem as input, and adopting a self-organizing neural network Kohonen algorithm to carry out model training to obtain a trained neural network model; step S3, clustering the base stations with switching problems according to the trained neural network model; and step S4, finding the nearest base station configuration without the switching problem in the cluster, and replacing the base station configuration with the problem by taking the configuration as an optimal configuration scheme. The invention automatically generates optimized parameter configuration for the problem base station, solves the ring switching problem and improves the intelligent level of network optimization work.
Description
Technical Field
The invention relates to the field of mobile communication, in particular to a base station parameter automatic optimization method based on a neural network.
Background
When a mobile communication device (e.g., handset) is within the coverage area of multiple mobile base station cells, if the associated base station is configured with problems, it may cause the mobile device to frequently disconnect from one connected cell and switch to another cell. Even when the mobile device is still, the mobile device is switched among a plurality of cells endlessly to form a ring switch, which causes poor connection quality of the mobile device. As a mobile operator, it is necessary to optimize the setting of base station parameters, so as to minimize the occurrence of ring switching and improve the service quality of mobile communication.
In the prior art, the discovery of the ring switching problem mainly includes statistical analysis of the performance data of the drive test or the cell in a short time, and the solution of the ring switching problem mainly depends on the manual optimization configuration of a network optimization engineer.
Disclosure of Invention
In view of this, the present invention aims to provide a method for automatically optimizing parameters of a base station based on a neural network, which automatically generates an optimized parameter configuration for a problem base station, solves the problem of ring switching, and improves the intelligent level of network optimization work.
In order to achieve the purpose, the invention adopts the following technical scheme:
a base station parameter automatic optimization method based on a neural network is characterized by comprising the following steps:
step S1, scanning the MRE data of the total base station, and respectively outputting the base station configuration with the ring switching problem and the base station configuration without the ring switching problem;
step S2, taking a base station configuration set without the problem of ring switching as input, and carrying out model training by adopting a self-organizing neural network Kohonen algorithm to obtain a trained neural network model;
step S3, clustering the base stations with switching problems according to the trained neural network model;
and step S4, finding the nearest base station configuration without the switching problem in the cluster, and replacing the base station configuration with the problem by taking the configuration as an optimal configuration scheme.
Further, in step S1, the presence or absence of the ring switching problem is discriminated according to the conditions that can form the ring switching parameter problem specified in the TS36.331 measurement event protocol of 3 GPP.
Further, the step S2 is specifically:
step S21, zero-mean normalization is carried out on each parameter of the base station without the problem of ring switching one by one, and the conversion formula is as follows:
wherein
The mean value of one parameter of each base station is defined, and sigma is the standard deviation of one parameter value of each base station;
step S22, initializing a self-organizing neural network Kohonen model, each neuronkWeight of (1)kIs a set of arrays (w)1,w2,…,wn)Wherein n is the number of the base station parameters, and each initial value of w is a random decimal between 0 and 1;
step S23, a base station cell without ring switching problem is addednoringThe parameters of (2) are input, and the neuron is calculated and matched with each neuronkEuclidean distance d (cell)noring,neuronk)：
Wherein the neuronkThe number of the k-th neuron is,
the ith configuration parameter of the base station without the problem of ring switching is obtained, and n is the number of the parameters;
step S24, get d (cell)noring,neuronk) Minimal neuronal neuronkMarking the winning neuron of the base station and the cluster number of the base station as k;
step S25, updating neuronkWeight of (1)kA value of (w)1',w2',…,wn')：
Wherein l is a self-defined learning rate and can be set as a number between [0,1] according to needs.
Step S26, each base station cell without the ring switching problemnoringConfiguration input, repeating step S23-step S25;
step S27, saving each neuron finally generatedkWeight of (1)kAnd storing the trained neural network model.
Further, the step S3 is specifically:
step S31, each parameter of each base station is normalized one by zero-mean value, and the conversion formula is as follows:
wherein
For each base station a mean value of a parameter, σ isA standard deviation of a parameter value for each base station;
step S32, a base station cell with ring switching problem is addedringThe parameters of (2) are input, and the neuron is calculated and matched with each neuronkEuclidean distance d (cell)ring,neuronk)：
Wherein the neuronkThe number of the k-th neuron is,
the ith configuration parameter of the base station with the ring switching problem is obtained, and n is the number of the parameters;
step S33, get d (cell)ring,neuronk) Minimal neuronal neuronkMarking the winning neuron of the base station and the cluster number of the base station as k;
step S34, each base station cell with the ring switching problemringAnd (4) configuring input, repeating the step S32 and the step S33, and finishing clustering all base stations with the ring switching problem.
Further, the step S4 is specifically:
step S41, inputting a base station with ring switching problem, according to its cluster number k, calculating Euclidean distance d (cell) in all base stations without ring switching problem and with cluster number kring,cellnoring)：
Wherein the cellringFor base stations with ring switching problems, cellsringIn order for a base station to have a ring handover problem,
for a certain configuration parameter of the base station with the problem of ring switching,
a certain configuration parameter of a base station without the problem of ring switching, wherein n is the number of the parameters;
step S42, get d (cell)ring,cellnoring) Minimum base station configuration cell without ring switching problemnoringReplacing the cell as an optimized configurationnoringThe configuration of (1);
step S43, each base station cell with the ring switching problemringAnd (4) inputting configuration, repeating the step S42 and the step S43, and finishing the generation of the optimization scheme of all base stations with the ring switching problem.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention automatically generates a reasonable base station parameter optimization configuration scheme, and the optimization quality is more stable.
2. The invention can carry out rapid detection and output optimization scheme on the switching problem of the whole network ring, thereby greatly improving the efficiency.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention is further explained below with reference to the drawings and the embodiments.
Referring to fig. 1, the present invention provides a method for automatically optimizing parameters of a base station based on a neural network, which automatically generates an optimized parameter configuration for a problem base station, and specifically includes the following steps:
1. the MRE data (event type measurement report) of the total number of base stations is scanned, and the base station configuration with the ring switching problem and the base station configuration without the ring switching problem are output according to the conditions which can form the ring switching parameter problem and are specified in the TS36.331 measurement event protocol of 3 GPP.
2. And taking a base station configuration set without the ring switching problem as input, performing model training by adopting a self-organizing neural network Kohonen algorithm, and clustering the base stations. The method comprises the following specific steps:
2.1. each parameter of each base station is normalized zero-mean value (also called standard deviation normalization) one by one, and the conversion formula is as follows:
wherein
Is the mean value of one parameter for each base station, and σ is the standard deviation of one parameter value for each base station.
2.2. Initializing a self-organizing neural network Kohonen model, each neuronkWeight of (1)kIs a set of arrays (w)1,w2,…,wn)Wherein n is the number of the base station parameters, and each initial value of w is a random decimal between 0 and 1.
2.3. A base station cell without the ring switching problem is usednoringThe parameters of (2) are input, and the neuron is calculated and matched with each neuronkEuclidean distance d (cell)noring,neuronk)：
Wherein the neuronkThe number of the k-th neuron is,
the ith configuration parameter of the base station without the ring switching problem is obtained, and n is the number of the parameters.
2.4. Get d (cell)noring,neuronk) Minimal neuronal neuronkAnd marking the winning neuron of the base station, and marking the cluster to which the base station belongs to have the number of k.
2.5. Updating neuronal neuroneskWeight of (1)kA value of (w)1',w2',…,wn')：
Wherein l is a self-defined learning rate and can be set as a number between [0,1] according to needs.
2.6. Each base station cell without the ring switching problemnoringConfiguring input, and repeating the steps of 2.3-2.5.
2.7. Saving each neuron finally generatedkWeight of (1)kAnd storing the trained neural network model.
3. And clustering the base stations with the switching problem by using the trained model.
3.1. Each parameter of each base station is normalized zero-mean value (also called standard deviation normalization) one by one, and the conversion formula is as follows:
wherein
Is the mean value of one parameter for each base station, and σ is the standard deviation of one parameter value for each base station.
3.2. A base station cell with a ring switching problem is arrangedringThe parameters of (2) are input, and the neuron is calculated and matched with each neuronkEuclidean distance d (cell)ring,neuronk)：
Wherein the neuronkThe number of the k-th neuron is,
the ith configuration parameter of the base station with the ring switching problem is n, and n is the number of the parameters.
3.3. Get d (cell)ring,neuronk) Minimal neuronal neuronkAnd marking the winning neuron of the base station, and marking the cluster to which the base station belongs to have the number of k.
3.4. Each base station cell with the ring switching problemringAnd (4) configuring input, and repeating the steps of 3.2-3.3 to finish clustering all base stations with the ring switching problem.
4. And finding the closest base station configuration without the switching problem in the cluster, and replacing the base station configuration with the problem by taking the configuration as an optimal configuration scheme.
4.1. Inputting a base station with ring switching problem, according to its cluster number k, in all base stations without ring switching problem and with cluster number k, calculating Euclidean distance d (cell)ring,cellnoring)：
Wherein the cellringFor base stations with ring switching problems, cellsringIn order for a base station to have a ring handover problem,
for a certain configuration parameter of the base station with the problem of ring switching,
n is the number of parameters for a certain configuration parameter of the base station without the problem of ring switching.
4.2. Get d (cell)ring,cellnoring) Minimum base station configuration cell without ring switching problemnoringReplacing the cell as an optimized configurationnoringThe configuration of (2).
4.3. Each base station cell with the ring switching problemringAnd (4) configuring input, and repeating the steps of 4.1-4.2 to complete the generation of the optimization scheme of all base stations with the ring switching problem.
The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.
Claims (4)
1. A base station parameter automatic optimization method based on a neural network is characterized by comprising the following steps:
step S1, scanning the MRE data of the total base station, and respectively outputting the base station configuration with the ring switching problem and the base station configuration without the ring switching problem;
step S2, taking a base station configuration set without the problem of ring switching as input, and carrying out model training by adopting a self-organizing neural network Kohonen algorithm to obtain a trained neural network model;
the step S2 specifically includes:
step S21, zero-mean normalization is carried out on each parameter of the base station without the problem of ring switching one by one, and the conversion formula is as follows:
wherein
The mean value of one parameter of each base station is defined, and sigma is the standard deviation of one parameter value of each base station;
step S22, initializing a self-organizing neural network Kohonen model, wherein the weight of each neuron is an array set (w)1,w2,…,wn) Wherein n is the number of the base station parameters, and each initial value of w is a random decimal between 0 and 1;
step S23, a base station cell without ring switching problem is addednoringThe Euclidean distance d (cell) to each neuron is calculated by inputting each parameternoring,neuronk)：
Wherein the neuronkFor the number k of the neurons, the number k,
the ith configuration parameter of the base station without the problem of ring switching is obtained, and n is the number of the parameters;
step S24, get d (cell)noring,neuronk) The minimum neuron is a winning neuron of the base station, and the cluster number to which the base station belongs is marked as k;
step S25, updating neuronkWeight of (1)kA value of (w)1′,w2′,…,wn′)：
Wherein l is a self-defined learning rate and can be set as a number between [0,1] according to needs;
step S26, each base station cell without the ring switching problemnoringConfiguration input, repeating step S23-step S25;
step S27, storing the weight value of each neuron which is finally generated, namely storing the trained neural network model;
step S3, clustering the base stations with switching problems according to the trained neural network model;
and step S4, finding the nearest base station configuration without the switching problem in the cluster, and replacing the base station configuration with the problem by taking the configuration as an optimal configuration scheme.
2. The method of claim 1, wherein the method comprises the following steps: the step S1 distinguishes whether or not there is a ring switching problem according to the conditions that can form a ring switching parameter problem specified in the TS36.331 measurement event protocol of 3 GPP.
3. The method of claim 1, wherein the method comprises the following steps: the step S3 specifically includes:
step S31, each parameter of each base station is normalized one by zero-mean value, and the conversion formula is as follows:
wherein
The mean value of one parameter of each base station is defined, and sigma is the standard deviation of one parameter value of each base station;
step S32, a base station cell with ring switching problem is addedringThe Euclidean distance d (cell) to each neuron is calculated by inputting each parameterring,neuronk)：
Wherein
The ith configuration parameter of the base station with the ring switching problem is obtained, and n is the number of the parameters;
step S33, get d (cell)ring,neuronk) The minimum neuron is a winning neuron of the base station, and the cluster number to which the base station belongs is marked as k;
step S34, each base station cell with the ring switching problemringAnd (4) configuring input, repeating the step S32 and the step S33, and finishing clustering all base stations with the ring switching problem.
4. The method of claim 1, wherein the method comprises the following steps: the step S4 specifically includes:
step S41, inputting a base station with ring switching problem, according to its cluster number k, calculating Euclidean distance d (cell) in all base stations without ring switching problem and with cluster number kring,cellnoring)：
Wherein the cellringIn order for a base station to have a ring handover problem,
in order to have the configuration parameters of the base station with the problem of the ring switching,
configuring parameters of a base station without the problem of ring switching, wherein n is the number of the parameters;
step S42, get d (cell)ring,cellnoring) Minimum base station configuration cell without ring switching problemnoringReplacing the cell as an optimized configurationnoringThe configuration of (1);
step S43, each base station cell with the ring switching problemringAnd (4) inputting configuration, repeating the step S42 and the step S43, and finishing the generation of the optimization scheme of all base stations with the ring switching problem.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910048705.3A CN109729540B (en) | 2019-01-18 | 2019-01-18 | Base station parameter automatic optimization method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910048705.3A CN109729540B (en) | 2019-01-18 | 2019-01-18 | Base station parameter automatic optimization method based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109729540A CN109729540A (en) | 2019-05-07 |
| CN109729540B true CN109729540B (en) | 2022-05-17 |
Family
ID=66298494
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910048705.3A Active CN109729540B (en) | 2019-01-18 | 2019-01-18 | Base station parameter automatic optimization method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109729540B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111954206B (en) * | 2019-05-17 | 2024-04-09 | 株式会社Ntt都科摩 | Terminal and base station |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101635949A (en) * | 2008-07-25 | 2010-01-27 | 三星电子株式会社 | Switching method in wireless cellular network |
| CN102149114A (en) * | 2011-05-10 | 2011-08-10 | 北京邮电大学 | Femto base station network control system and method |
| CN102740335A (en) * | 2012-06-13 | 2012-10-17 | 北京邮电大学 | Self-optimizing method for switching femtocell network |
| CN103385022A (en) * | 2011-01-31 | 2013-11-06 | 诺基亚西门子通信公司 | Method and apparatus for managing neighbour relations |
| JPWO2016076293A1 (en) * | 2014-11-11 | 2017-08-24 | シャープ株式会社 | Base station apparatus, location management apparatus, terminal apparatus, and control method |
| CN107645393A (en) * | 2016-07-20 | 2018-01-30 | 中兴通讯股份有限公司 | Determine the method, apparatus and system of the black-box system input and output degree of association |
| CN108134979A (en) * | 2017-12-04 | 2018-06-08 | 东南大学 | Small base station switch control method based on deep neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9565576B2 (en) * | 2013-10-09 | 2017-02-07 | At&T Intellectual Property I, L.P. | Network operating system client architecture for mobile user equipment |
-
2019
- 2019-01-18 CN CN201910048705.3A patent/CN109729540B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101635949A (en) * | 2008-07-25 | 2010-01-27 | 三星电子株式会社 | Switching method in wireless cellular network |
| CN103385022A (en) * | 2011-01-31 | 2013-11-06 | 诺基亚西门子通信公司 | Method and apparatus for managing neighbour relations |
| CN102149114A (en) * | 2011-05-10 | 2011-08-10 | 北京邮电大学 | Femto base station network control system and method |
| CN102740335A (en) * | 2012-06-13 | 2012-10-17 | 北京邮电大学 | Self-optimizing method for switching femtocell network |
| JPWO2016076293A1 (en) * | 2014-11-11 | 2017-08-24 | シャープ株式会社 | Base station apparatus, location management apparatus, terminal apparatus, and control method |
| CN107645393A (en) * | 2016-07-20 | 2018-01-30 | 中兴通讯股份有限公司 | Determine the method, apparatus and system of the black-box system input and output degree of association |
| CN108134979A (en) * | 2017-12-04 | 2018-06-08 | 东南大学 | Small base station switch control method based on deep neural network |
Non-Patent Citations (1)
| Title |
|---|
| "The Restriction of Cell Reselection and Handover Parameters";ZTE;《3GPP TSG RAN WG3 #66bis R3-100183》;20100722;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109729540A (en) | 2019-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10911318B2 (en) | Future network condition predictor for network time series data utilizing a hidden Markov model for non-anomalous data and a gaussian mixture model for anomalous data | |
| CN108540330B (en) | Network fault diagnosis method based on deep learning in heterogeneous network environment | |
| CN107947878B (en) | Cognitive radio power distribution method based on energy efficiency and spectrum efficiency joint optimization | |
| CN107113635A (en) | Method and apparatus for determining cell status to adjust antenna configuration parameters | |
| CN111698728B (en) | Topology control system for dynamic network and control method thereof | |
| CN113242526A (en) | Cloud computing server room real-time monitoring system | |
| CN109729540B (en) | Base station parameter automatic optimization method based on neural network | |
| CN116545954B (en) | Communication gateway data transmission method and system based on internet of things | |
| Jaffry et al. | Cellular traffic prediction using recurrent neural networks | |
| CN108600999A (en) | FD-D2D is based on channel distribution and power control combined optimization method | |
| CN104981002A (en) | Position determining method of convergent node in wireless sensor network | |
| Chamola et al. | Outage estimation for solar powered cellular base stations | |
| Lee et al. | Affinity propagation for energy-efficient BS operations in green cellular networks | |
| CN113365323A (en) | Low-energy-consumption multipath transmission method for wireless sensor network | |
| CN106954268A (en) | Access network resource distribution method under a kind of SDN frameworks | |
| CN115314395B (en) | NR signal coverage drive test optimization method | |
| CN109819494B (en) | Energy consumption balanced WSN (wireless sensor network) hierarchical routing protocol implementation method | |
| CN114662624B (en) | Non-invasive load identification method and system based on retraining twin network | |
| CN113923745B (en) | Communication relay selection method for power communication system and communication method thereof | |
| CN110968075A (en) | Fault diagnosis method and system based on active learning self-organizing cellular network | |
| CN113950134B (en) | Dormancy prediction method, device, equipment and computer readable storage medium of base station | |
| El Morabit et al. | Small cell switch off using genetic algorithm | |
| CN114861993A (en) | Regional photovoltaic power generation prediction method based on federal learning and deep neural network | |
| CN114698016A (en) | Network fault detection method and device for poor-quality cell | |
| CN113660674A (en) | Method for intelligently realizing network construction and network optimization |
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 | ||
| GR01 | Patent grant |