CN110972142A - 5G network weak coverage area locking and adjusting method based on GNG mode - Google Patents
5G network weak coverage area locking and adjusting method based on GNG mode Download PDFInfo
- Publication number
- CN110972142A CN110972142A CN201911337251.8A CN201911337251A CN110972142A CN 110972142 A CN110972142 A CN 110972142A CN 201911337251 A CN201911337251 A CN 201911337251A CN 110972142 A CN110972142 A CN 110972142A
- Authority
- CN
- China
- Prior art keywords
- node
- gng
- nodes
- weak coverage
- age
- 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.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/02—Resource partitioning among network components, e.g. reuse partitioning
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention provides a method for locking and adjusting a weak coverage area of a 5G network based on a GNG mode, which comprises the following steps: step 1, initializing an input space N; step 2, initializing a network A; step 3, updating the input space N; and 4, executing a GNG algorithm, gradually adding new nodes, adjusting the positions of the original nodes in the network A, and training a stable network A. The invention can avoid the loss of the weak coverage area caused by unreasonable setting of the DBSCAN threshold.
Description
Technical Field
The invention belongs to the technical field of mobile communication, and particularly relates to a method for locking and adjusting a weak coverage area of a 5G network based on a GNG mode.
Background
In 2019, with the distribution of 5G license plates, the 5G construction in China enters a scale deployment stage, and the single-station coverage capability of 5G is weaker than that of 4G due to the higher frequency band. In the initial stage of 5G network deployment, the base station and the 4G station are constructed in a 1:1 co-site mode, so that a weak coverage area is easy to appear, and a more accurate weak coverage area positioning method is needed to be adopted to ensure that an operator creates a 5G fine network with excellent experience.
In the existing 4G weak coverage area positioning, an mr (measurement report) big data positioning mode is adopted, and a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used to perform cluster calculation on a difference grid, so as to plan a new site Based on the difference grid, and improve the construction accuracy and the blind-repairing efficiency. The DBSCAN clustering algorithm utilizes the idea of density-based clustering, which defines clusters as the largest set of density-connected points, can divide areas with sufficiently high density into clusters, and can find clusters of arbitrary shape in a spatial database with "noise". DBSCAN includes 3 input data in total:
1. data set D: scene type (e.g., dense urban, general urban, suburban, rural);
2. the minimum neighborhood point number, MinPts, for a given point to become a core object within the neighborhood: the minimum number of weakly covered grids that exist within the neighborhood radius;
3. neighborhood radius Eps: and taking the number of grids as the radius of the search neighborhood.
Wherein Eps and MinPts need to be set manually according to specific applications.
As illustrated in fig. 1: this takes 1 grid as the search radius, and the minimum number of weakly covered grids existing within the neighborhood radius is 3 (2 is the core point, 1 is the edge point, 0 is the noise point). By adopting the algorithm, the clustering speed is high, the noise points can be effectively processed, the spatial clusters with any shapes can be found, and the number of clusters to be divided does not need to be input. In the wireless network construction, the method plays an important role in providing station construction basis in a staged manner.
The prior art has the following defects:
(1) when the density of spatial clustering is not uniform and the difference of clustering intervals is large, the parameters MinPts and Eps are difficult to select, and the clustering effect is not ideal: if the value of MinPts is too small, the result in the sparse cluster is regarded as a boundary point and is not used for further extension of the class because the density is less than MinPts; if the value is too large, two neighboring clusters with a greater density may be merged into the same cluster. If MinPts is not changed, the acquired value of Eps is too large, most points can be gathered in the same cluster, and if Eps is too small, one cluster can be split; if Eps is not changed, the value of MinPts gets too large, resulting in the same cluster midpoint being marked as an outlier, while MinPts is too small, resulting in a large number of core points being found.
(2) The method only reflects the stage weak coverage problem, is easily influenced by short-term factors such as seasons, festivals and holidays, is difficult to observe the general trend, and has certain limitation on guidance construction: in suburb agricultural areas, the dense degree of leaves has great influence on network coverage, signal propagation attenuation is up to 10dB due to tree and forest shielding in summer, data are extracted in summer months, weak coverage areas are obviously more than in winter, and the clustered station building needs more; in a campus area, telephone bills are hardly generated in summer holidays and cold holidays, and the coverage weak area cannot be exposed in MR data in 7-8 months or 1 month by adopting a DBSCAN clustering algorithm, so that a reasonable construction suggestion is formed.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the defects of the prior art, the invention provides a 5G Network weak coverage area locking and observing method based on a GNG (Growing Neural Gas Network) mode, which comprises the following steps:
step 3, updating the input space N;
and 4, executing a GNG algorithm, gradually adding new nodes, adjusting the positions of the original nodes in the network A, and training a stable network A.
The step 1 comprises the following steps:
initializing input space N, obtaining probability p of randomly generating an input signal ξ (ξ):
the step 2 comprises the following steps:
step 2-1, initializing the network a, arbitrarily placing the Node positions of 2 nodes c1 and c2 as initial points, setting the positions of 2 nodes c1 and c2 as the initial 2 weak coverage occurrence positions, and then:
A={c1,c2};(2)
step 2-2, randomly initializing position vectors (directional line segments with the origin of coordinates as a starting point and the position of the moving particle as an end point at a certain moment) of 2 nodes c1 and c2 by using the probability p (ξ), and initializing an adjacency matrix Is empty, C is 0.
The step 3 comprises the following steps:
step 3-1, updating an input space N to obtain the probability p (ξ) of randomly generating an input signal ξ as 1/N, and randomly generating an input signal ξ with the probability p (ξ) in the input space;
step 3-2, winning nodes s1 and s2 are calculated.
In step 3-2, winning nodes s1 and s2 are calculated according to the following formula:
s1=argm inc∈A||ξ-wc|| (3)
s2=argm inc∈A\{s1}||ξ-wc|| (4)
wherein, argmincThe variable value when the target function takes the minimum value is referred to; the M represents the value of the determinant of the matrix M, being an area, volume or super volume; s1, s2 ∈ A. w is acRepresenting the weight vector of node c.
Step 4 comprises the following steps:
step 4-1, adding a weak coverage area signal; (the weak coverage area is made up of randomly converging points of weak coverage in the field, requiring recalculation)
Step 4-2, calculating the distance between a newly added weak coverage area signal and the existing Node, wherein the Node with the closest distance is marked as s1 and is called as a winning Node Winner (), and the Node with the second-most distance is marked as s 2; (the nearest and second nearest points are only labeled as s1 and s2, and are used as before and only for identification purposes.)
Step 4-3, if there is no connection between s1 and s2, then a connecting edge C is created:
C=C∪{(s1,s2)}; (5)
step 4-4, adjusting the accumulated error Δ E of the winning node s1s1;
Step 4-5, correcting the position of the existing Node;
step 4-6, if the number of times the input signal is generated (randomly generated) is an integer multiple of λ (in mathematics, the ratio which is often used to represent the constant fraction in the constant fraction point of the vector), then a new node is randomly inserted;
step 4-7, calculating the time interval delta t of the operation:
Δt=tk-tk-1(10)
wherein, tk-1Represents the time to complete k-1 times;
if Δ T ≧ TGNGIf the stopping condition is not met and no weak coverage exists, turning to step 3, TGNGIs GNG operation period;
if Δ t<TGNGAnd if the stop condition is not satisfied, the operation is suspended (T)GNGTime Δ t) and then to step 3.
In step 4-3, the age (s1, s2) of the connecting edge C is set to 0.
In step 4-4, the accumulated error Δ E of the winning node s1 is adjusted according to the following formulas1:
ΔEs1=||ξ–ws1||2(6)
Wherein, ws1Representing the nearest cell in the input space.
The steps 4-5 comprise:
step 4-5-1, determining whether the target function can converge to a local minimum value and when the target function converges to the local minimum value according to a learning rate (important hyper-parameters in supervised learning and deep learning)bAnd εnAdjusting winning nodess1 and the position vector of the node connected thereto;
Δws1=εb(ξ–ws1) (7)
where Δ is the greek letter delta, indicating a change, e.g., t1-t 0; w is as1A weight vector representing node s1, wi represents a weight vector for node i; in the formula, i represents that N is contained ins1Any value within; epsilon1Representing only a positive number infinitely close to 0, no specific value, mathematically expressed as ε1→0;Ns1Representing the neighbor set of node s 1.
Step 4-5-2, adjusting the age of the connecting edge of the winning node s 1:
wherein, age (s1, i) represents the age difference of the connecting edge of the winning node;
if age (i, j)>amaxDeleting edges (i, j) and nodes not connected to edges, age (i, j) representing the age of the edges (i, j), amaxRepresenting the maximum age that results in a point with no diverging edges, i, j is the connected edge vector.
Has the advantages that: the invention innovatively adopts a weak coverage area locking and adjusting method based on a GNG mode to be applied to wireless network MR difference grid clustering. The GNG has the characteristics of clustering, dimensionality reduction and self-learning, and has the advantages of high convergence speed, small cost error, stable convergence and the like. A dynamic change mechanism is established by utilizing a GNG algorithm, and the change of the weak coverage area is observed quantitatively by dotting at regular intervals, so that analysis concepts such as tide and structural change are established, and the loss of the weak coverage area caused by unreasonable setting of a DBSCAN threshold can be avoided.
Drawings
The foregoing and/or other advantages of the invention will become further apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings.
Fig. 1 is a schematic diagram of 1 grid as a search radius.
Fig. 2 is a discrete MR weak coverage difference grid.
Fig. 3 is a schematic diagram of an initial node.
FIG. 4 is a schematic diagram of creating a connecting edge.
Fig. 5a is a schematic diagram of an input signal.
FIG. 5b is a diagram illustrating adjusting weights.
FIG. 5c is a schematic diagram of deleting aged edges.
Fig. 5d is a schematic diagram of adding a new node.
FIG. 6 is a dynamic view.
Detailed Description
Different from traditional network optimization means such as drive test and fixed point test, a measurement report mr (measurement report) is a report that a mobile terminal periodically reports the downlink signal strength and quality of a cell to a base station at fixed time intervals on a traffic channel through a control channel, so as to comprehensively and efficiently reflect the network coverage condition and accurately position a coverage hole. But the MR difference grid raw data is discrete as shown in fig. 2. Fig. 6 is a schematic diagram showing a front-back change of a weak coverage area in the GNG mode.
And (3) locking the discrete weak coverage areas by adopting a GNG mode, and establishing a quantitative observation mechanism:
as shown in fig. 3, the Node positions (two large dots in fig. 3) of 2 nodes are arbitrarily placed as initial points, the initial 2 Node positions are the initial 2 weak coverage positions, and as the weak coverage positions are more and more, new nodes are gradually added and the original Node positions are adjusted to gradually train a stable Node network.
The nodes of the GNG are parameters (e.g., set to 250 in DEMO, i.e., up to 250 nodes) that reflect the area of weak coverage distribution. The node network forms a range outline of weak coverage, and the positions of the nodes represent areas with frequent MR of the weak coverage, similar to weak coverage core points in DBSCAN.
A={c1,c2}; (2)
randomly initializing the position vector of 2 nodes with probability p (ξ), initializing the adjacency matrixIs empty set, C is 0;
and 3, updating the input space N to obtain the probability p (ξ) of randomly generating an input signal ξ as 1/N, randomly generating an input signal ξ by the probability p (ξ) in the input space, and calculating winning nodes s1 and s2, s1, s2 epsilon A, as shown in the figure 5 a.
s1=argm inc∈A||ξ-wc|| (3)
s2=argm inc∈A\{s1}||ξ-wc|| (4)
argmincThe variable value when the target function takes the minimum value is referred to; and M represents the value of the determinant of the matrix M, and is an area, a volume or a super volume.
Under the existing 2 nodes, a weak coverage area signal (the weak coverage area is formed by randomly converging on-site weak coverage points and needs to be recalculated) is added (fig. 4 triangular area), and the following operations are performed according to the GNG algorithm:
1. the distance between this signal and the existing Node is calculated, the nearest Node is labeled as s1, called Winner (Node on the upper right of fig. 4), the next farthest is labeled as s2 (only the nearest and next nearest points are labeled as s1 and s2, which are different from the previous meaning and are used for the same purpose, and only identification is done) (Node on the lower left of fig. 4).
Step 4, if there is no connection between s1 and s2, then a connecting edge C is created:
C=C∪{(s1,s2)} (5)
setting the age (s1, s2) of the connecting edge C to be 0;
adjustment ofAccumulated error Δ E of winning node s1s1:
ΔEs1=||ξ–ws1||2(6)
2. And correcting the position of the existing Node:
determining whether and when the objective function can converge to a local minimum at a learning rate (important hyper-parameters in supervised learning and deep learning)bAnd εnThe position vector of the winning node s1 and the nodes connected thereto is adjusted as shown in fig. 5 b.
Δws1=εb(ξ–ws1) (7)
Age of connecting edge of adjusted winning node s 1:
step 5, if age (i, j)>amaxThe edges (i, j) are deleted, while nodes not connecting edges are deleted, as shown in fig. 5 c.
If the number of times the input signal is generated is an integer multiple of λ (in mathematics, the ratio usually used to represent the constant fraction in the constant fraction points of the vector), a new node is inserted, as shown in fig. 5 d.
Step 7, calculating the time interval delta t of the operation:
Δt=tk-tk-1(10)
if Δ T ≧ TGNG(TGNGGNG operation cycle) and the stop condition is not met, go to step 3.
If Δ t<TGNGAnd the stop condition is not satisfied (no weak coverage point), the process is suspended (T)GNGTime Δ t) and then to step 3.
Interpretation and application of the results that the GNG algorithm tends to stabilize:
1. nodes connected with the line form a cluster (overlong lines are removed), the extracted envelope can be directly applied to the presentation of weak coverage, and the size of the cluster is quantized;
2. the trimming of EDGE and nodes in GNG is realized according to set Age and other thresholds, the sampling period and the thresholds are reasonably set, and the nodes and EDGE in the area where weak coverage does not appear again can disappear for 1 month, so that a dynamic weak coverage observation mechanism is established;
3. the locked nodes are the areas which are most worth of being put into the base station, a reasonable station building process is established, and the density degrees of the nodes reflect the frequency and the concentration of weak coverage;
4. snapshot dotting is performed on the GNG results of morning and evening/day/week, and then: tidal phenomena with poor coverage issues (such as changes as users migrate through holidays); the newly appeared weak coverage area of the network or the change trend of the original weak coverage area is visually presented in a GIS mode.
The present invention provides a method for locking and adjusting a weak coverage area of a 5G network based on a GNG method, and a plurality of methods and approaches for implementing the technical solution, and the above description is only a preferred embodiment of the present invention, and it should be noted that, for a person skilled in the art, a plurality of improvements and modifications may be made without departing from the principle of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. All the components not specified in the present embodiment can be realized by the prior art.
Claims (9)
1. The method for locking and adjusting the weak coverage area of the 5G network based on the GNG mode is characterized by comprising the following steps:
step 1, initializing an input space N;
step 2, initializing a network A;
step 3, updating the input space N;
and 4, executing a GNG algorithm, gradually adding new nodes, adjusting the positions of the original nodes in the network A, and training a stable network A.
3. the method of claim 2, wherein step 2 comprises:
step 2-1, initializing the network a, arbitrarily placing the Node positions of 2 nodes c1 and c2 as initial points, setting the positions of 2 nodes c1 and c2 as the initial 2 weak coverage occurrence positions, and then:
A={c1,c2}; (2)
4. The method of claim 3, wherein step 3 comprises:
step 3-1, updating an input space N to obtain a probability p (ξ) ═ N of randomly generating an input signal ξ, and randomly generating an input signal ξ with the probability p (ξ) in the input space;
step 3-2, winning nodes s1 and s2 are calculated.
5. The method of claim 4, wherein in step 3-2, winning nodes s1 and s2 are calculated according to the following formula:
s1=argm inc∈A||ξ-wc|| (3)
s2=argm inc∈A\{s1}||ξ-wc|| (4)
wherein, argmincThe variable value of the target function when the target function takes the minimum value, | | M | | | represents the value of the determinant of the matrix M, s1, s2 belongs to A; w is acRepresenting the weight vector of node c.
6. The method of claim 5, wherein step 4 comprises:
step 4-1, adding a weak coverage area signal; the weak coverage area is formed by randomly converging on-site weak coverage points;
step 4-2, calculating the distance between a newly added weak coverage area signal and the existing Node, wherein the Node with the closest distance is marked as s1 and is called as a winning Node Winner, and the Node with the second-most distance is marked as s 2;
step 4-3, if there is no connection between s1 and s2, then a connecting edge C is created:
C=C∪{(s1,s2)} ; (5)
step 4-4, adjusting the accumulated error Δ E of the winning node s1s1;
Step 4-5, correcting the position of the existing Node;
step 4-6, if the times of generating the input signals are integral multiples of lambda, inserting new nodes randomly;
step 4-7, calculating the time interval delta t of the operation:
Δt=tk-tk-1(10)
wherein, tk-1Represents the time to complete k-1 times;
if Δ T ≧ TGNGIf the stopping condition is not met and no weak coverage exists, turning to step 3, TGNGIs GNG operation period;
if Δ t<TGNGAnd if the stop condition is not satisfied, the operation is suspended (T)GNGTime Δ t) and then to step 3.
7. The method according to claim 6, wherein in step 4-3, the age (s1, s2) of the connecting edge C is set to 0.
8. The method of claim 7,in step 4-4, the accumulated error Δ E of the winning node s1 is adjusted according to the following formulas1:
ΔEs1=||ξ–ws1=||2(6)
Wherein, ws1Representing the nearest cell in the input space.
9. The method of claim 8, wherein steps 4-5 comprise:
step 4-5-1, learning ratebAnd εnAdjusting the position vector of the winning node s1 and the nodes connected thereto;
Δws1=εb(ξ–ws1) (7)
where Δ is the greek letter delta, representing a change value; w is as1A weight vector representing node s1, wi represents a weight vector for node i; in the formula, i represents that N is contained ins1Any value within;1represents a positive number; n is a radical ofs1A neighbor set representing node s 1;
step 4-5-2, adjusting the age of the connecting edge of the winning node s 1:
wherein, age (s1, i) represents the age difference of the connecting edge of the winning node;
if age (i, j)>amaxDeleting edges (i, j) and nodes not connected to edges, age (i, j) representing the age of the edges (i, j), amaxRepresenting the maximum age that results in a point with no diverging edges, i, j is the connected edge vector.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911337251.8A CN110972142A (en) | 2019-12-23 | 2019-12-23 | 5G network weak coverage area locking and adjusting method based on GNG mode |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911337251.8A CN110972142A (en) | 2019-12-23 | 2019-12-23 | 5G network weak coverage area locking and adjusting method based on GNG mode |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110972142A true CN110972142A (en) | 2020-04-07 |
Family
ID=70035771
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911337251.8A Pending CN110972142A (en) | 2019-12-23 | 2019-12-23 | 5G network weak coverage area locking and adjusting method based on GNG mode |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110972142A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112423337A (en) * | 2020-11-18 | 2021-02-26 | 中国联合网络通信集团有限公司 | Network coverage diagnosis model construction method, network coverage diagnosis method and device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104796916A (en) * | 2015-05-11 | 2015-07-22 | 苏州惠通天地信息科技有限公司 | Establishment method of wireless sensor network high barrier coverage for target detection |
CN104822152A (en) * | 2015-05-11 | 2015-08-05 | 苏州惠通天地信息科技有限公司 | Target-detection-oriented construction method for weak barrier coverage of wireless sensor network |
CN106878999A (en) * | 2017-03-20 | 2017-06-20 | 中国联合网络通信集团有限公司 | A kind of method and device for determining weak overlay area |
CN107231649A (en) * | 2016-03-25 | 2017-10-03 | 中国移动通信有限公司研究院 | A kind of determination method and device of indoor weak covering |
CN110602741A (en) * | 2019-09-23 | 2019-12-20 | 中国移动通信集团内蒙古有限公司 | Network weak coverage identification method, device, equipment and storage medium |
-
2019
- 2019-12-23 CN CN201911337251.8A patent/CN110972142A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104796916A (en) * | 2015-05-11 | 2015-07-22 | 苏州惠通天地信息科技有限公司 | Establishment method of wireless sensor network high barrier coverage for target detection |
CN104822152A (en) * | 2015-05-11 | 2015-08-05 | 苏州惠通天地信息科技有限公司 | Target-detection-oriented construction method for weak barrier coverage of wireless sensor network |
CN107231649A (en) * | 2016-03-25 | 2017-10-03 | 中国移动通信有限公司研究院 | A kind of determination method and device of indoor weak covering |
CN106878999A (en) * | 2017-03-20 | 2017-06-20 | 中国联合网络通信集团有限公司 | A kind of method and device for determining weak overlay area |
CN110602741A (en) * | 2019-09-23 | 2019-12-20 | 中国移动通信集团内蒙古有限公司 | Network weak coverage identification method, device, equipment and storage medium |
Non-Patent Citations (5)
Title |
---|
仲朝亮: "受生物启发的移动机器人空间认知及其导航", 《中国博士学位论文电子期刊网 信息科技辑》 * |
仲朝亮等: "动态环境下基于GNG网络的拓扑地图创建", 《华东理工大学学报(自然科学版)》 * |
仲朝亮等: "未知环境下基于GNG网络的环境地图自主认知", 《第二十九届中国控制会议论文集》 * |
庄思敏: "MR技术应用研究", 《中国新通信》 * |
王东署等: "未知环境中自主机器人环境探索与地图构建", 《郑州大学学报(理学版)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112423337A (en) * | 2020-11-18 | 2021-02-26 | 中国联合网络通信集团有限公司 | Network coverage diagnosis model construction method, network coverage diagnosis method and device |
CN112423337B (en) * | 2020-11-18 | 2023-06-09 | 中国联合网络通信集团有限公司 | Network coverage diagnosis model construction method, network coverage diagnosis method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108399748B (en) | Road travel time prediction method based on random forest and clustering algorithm | |
CN106780089B (en) | Permanent basic farmland planning method based on neural network cellular automaton model | |
CN107688906B (en) | Multi-method fused transmission line meteorological element downscaling analysis system and method | |
CN105657726B (en) | A kind of PCI planning base station selection method | |
CN107038167A (en) | Big data excavating analysis system and its analysis method based on model evaluation | |
CN104112062B (en) | The acquisition methods of wind-resources distribution based on interpolation method | |
KR102655903B1 (en) | Processing method and processing device for saving energy in a base station | |
CN104731916A (en) | Optimizing initial center K-means clustering method based on density in data mining | |
CN109887297A (en) | The division methods of urban traffic control sub-district based on quick global K-means spectral clustering | |
CN110430578B (en) | Method for realizing cell azimuth prediction based on mobile terminal data | |
CN104933483A (en) | Wind power forecasting method dividing based on weather process | |
CN107886160B (en) | BP neural network interval water demand prediction method | |
CN113422695A (en) | Optimization method for improving robustness of topological structure of Internet of things | |
CN109543939A (en) | A kind of method of green building productions certification risk evaluation model building | |
CN110134907B (en) | Rainfall missing data filling method and system and electronic equipment | |
CN110444022A (en) | The construction method and device of traffic flow data analysis model | |
CN113435630B (en) | Basin hydrological forecasting method and system with self-adaptive runoff yield mode | |
CN106980874B (en) | A kind of multi-time Scales dimension data fusion method towards distribution big data | |
Tabatabaei et al. | Optimization of sediment rating curve coefficients using evolutionary algorithms and unsupervised artificial neural network | |
CN115689051A (en) | Method for automatically calibrating SWMM model parameters based on GA algorithm coupling Morris and GLUE | |
CN106844757A (en) | A kind of antenna azimuth Outlier mining method | |
CN113743659B (en) | Urban layout prediction method based on component method and Markov cellular automaton and application | |
CN110972142A (en) | 5G network weak coverage area locking and adjusting method based on GNG mode | |
CN113537596A (en) | Short-time passenger flow prediction method for new line station of urban rail transit | |
CN107577896B (en) | Wind power plant multi-machine aggregation equivalent method based on hybrid Copula theory |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200407 |