[summary of the invention]
The objective of the invention is to obtain problems such as difficulty, frequency optimization DeGrain, frequency optimization scheme search efficiency on road be low, proposed a kind of frequency optimization method based on frequency sweep and propagation model coverage prediction in order to have overcome TD-SCDMA mobile phone measurement report data.
For achieving the above object, design a kind of frequency optimization method, it is characterized in that this method may further comprise the steps based on frequency sweep and propagation model coverage prediction:
A, test at main roads by sweep generator, to obtain the frequency sweep data:
By the drive test mobile phone main roads are tested, to obtain drive test data, utilize drive test data to carry out propagation model revision again, and then carry out coverage prediction, can obtain the path loss matrix of each grid point to related cell by prediction and calculation based on the propagation model after proofreading and correct;
B, set up interference model: the frequency sweep data rasterizing based on frequency sweep and coverage prediction, the grid block that map is divided into several N*N, N is the size of grid block, unit is a rice, be assigned in each grid block according to the number of test points certificate of longitude and latitude position again frequency sweep, consider the practical significance of data, and do preliminary treatment;
Frequency sweep data rasterizing path loss matrix merges with coverage prediction path loss matrix: frequency sweep data rasterizing path loss data are relative more accurate, but only are present on the road, and it is also imperfect, is the subclass of coverage prediction path loss matrix;
Set up interference matrix: interference matrix is divided into co-channel interference matrix and adjacent interference matrix two parts frequently;
C, make up the dynamic evaluation function based on interference model:
The dynamic evaluation function comprises simplifies evaluation function and emulation of coverage capability evaluation function, but and the C/I interference threshold dynamic-configuration of evaluation function;
D, according to dynamic evaluation function search rate prioritization scheme: comprise the condition of obligating and self-adapted search method;
E, continuous repeating step d, the frequency optimization scheme after being optimized.
Network planning software NeSTTD-SCDMA software is used in the correction of described drive test data.
Described frequency sweep data processing step is:
A, delete in each sweep check point the RSCP value less than the frequency and the scrambler data of thresholding;
B, filter out the inadequate grid block of data, the frequency sweep data point number that promptly wherein comprises is less than the grid block of thresholding.By because itself there is certain sudden and inaccuracy in the frequency sweep data,, can cause certain influence to the accuracy of final path loss data if the frequency sweep data that comprise in the grid block are insufficient;
C, average for the identical frequency of all frequency sweep data points in the same grid block and a plurality of RSCP of scrambler;
D, get union for the different frequent points of all frequency sweep data points in the same grid block and the data of scrambler, and with it all as the valid data of this grid block;
E, be cell ID with frequency in each grid block and scrambler data map, the mode of mapping is: get apart from the cell ID of the nearest same frequency of this grid block with scrambler;
F, for each sub-district in each grid block, it is as follows to calculate path loss:
Path loss=transmitting power+antenna gain-RSCP
Described frequency sweep data rasterizing path loss matrix and coverage prediction path loss matrix merge the direct replacement method of employing, and for the grid block that exists in frequency sweep data rasterizing path loss data, its path loss data are used the data in frequency sweep data rasterizing path loss matrix; Data in the path loss The data coverage prediction path loss matrix of other grid block.
Described frequency sweep data rasterizing path loss matrix and coverage prediction path loss matrix merge the employing extension method of average, and in the grid block that exists in frequency sweep data rasterizing path loss data, its path loss data are used the data in frequency sweep data rasterizing path loss matrix; For the adjacent cells piece of the grid block that exists in the frequency sweep data rasterizing path loss data, the mean value of its path loss The data frequency sweep rasterizing path loss and coverage prediction path loss; Data in the path loss The data coverage prediction path loss matrix of other grid block.
Described interference matrix calculation procedure is as follows:
The minizone is less than the relevant total grid block number in sub-district of the grid block number of C/I thresholding T/in twos in twos
1) co-channel interference matrix: C/I thresholding T is the interference value of-3 o'clock minizone in twos;
2) adjacent interference matrix frequently: C/I thresholding T is-21 o'clock interference values of minizone in twos.
Described simplification evaluation function is as follows:
K1*∑(∑CoI
i,j+∑AdjI
i,k)+K2*∑∑(∑∑CoI
l,ll,m,mm+∑∑AdjI
l,ll,n,nm)
Described emulation of coverage capability evaluation function is as follows:
K1* main carrier C/I is less than the grid block number of the auxilliary carrier wave C/I of grid block number+K2* of T1 less than T2
It is described that to obligate condition as follows:
Can not be with the station cell main carrier with frequently; Can not be with cell carrier with frequently.
Described adaptive searching method determination methods is as follows:
(Avg-C)/Avg>0.9
The present invention compares with prior art, made full use of this measured data consistent of frequency sweep with user's perception, also taken all factors into consideration simultaneously the propagation model coverage prediction, set up interference model by the two fusion, and make up the dynamic evaluation function, thereby search the frequency optimization scheme that meets the demands with performance efficiently based on this accurate model.
[embodiment]
Below in conjunction with accompanying drawing, the present invention will be further described, and this process is clearly to this professional personage.
Referring to Fig. 1, technical step of the present invention is as follows:
1, obtains frequency sweep and coverage prediction data
Test at main roads by sweep generator, to obtain the frequency sweep data.
By the drive test mobile phone main roads are tested, to obtain drive test data, utilize drive test data to carry out propagation model revision again, and then carry out coverage prediction, can obtain the path loss matrix of each grid point to related cell by prediction and calculation based on the propagation model after proofreading and correct.Drive test is proofreaied and correct propagation model and is the network planning software's basic function based on the coverage prediction of propagation model, can use network planning software such as NeST TD-SCDMA software to finish.
2, set up interference model based on frequency sweep and coverage prediction
(1) frequency sweep data rasterizing is referring to Fig. 2
Map is divided into the grid block of several N*N, and N is the size (unit is rice) of grid block, again according to the longitude and latitude position with the number of test points of frequency sweep according to being assigned in each grid block.Consider the practical significance of data, rasterizing precision (being the size of grid block) should not be higher than the accuracy of map.For improving follow-up computational efficiency, data need be done following preliminary treatment:
1) deletes in each sweep check point the RSCP value less than the frequency and the scrambler data of thresholding.
2) filter out the inadequate grid block of data, the frequency sweep data point number that promptly wherein comprises is less than the grid block of thresholding.By because itself there is certain sudden and inaccuracy in the frequency sweep data,, can cause certain influence to the accuracy of final path loss data if the frequency sweep data that comprise in the grid block are insufficient.
After the data preliminary treatment is finished, handle as follows:
1) averages for the identical frequency of all frequency sweep data points in the same grid block and a plurality of RSCP of scrambler.
2) get union for the different frequent points of all frequency sweep data points in the same grid block and the data of scrambler, and with it all as the valid data of this grid block.
3) be cell ID with frequency in each grid block and scrambler data map, the mode of mapping is: get apart from the cell ID of the nearest same frequency of this grid block with scrambler.
4) for each sub-district in each grid block, it is as follows to calculate path loss:
Path loss=transmitting power+antenna gain-RSCP
So far, obtained path loss matrix based on the frequency sweep data rasterizing.
(2) frequency sweep data rasterizing path loss matrix and coverage prediction path loss matrix merge, and be more accurate relatively referring to Fig. 3 frequency sweep data rasterizing path loss data, but only be present on the road, and it is also imperfect, is the subclass of coverage prediction path loss matrix.Need merge the two, adopt following two kinds of methods:
1) directly replaces method.For the grid block that exists in frequency sweep data rasterizing path loss data, its path loss data are used the data in the frequency sweep data rasterizing path loss matrix; The path loss data of other grid block are adopted
With the data in the coverage prediction path loss matrix.
2) the extension method of average.For the grid block that exists in frequency sweep data rasterizing path loss data, its path loss data are used the data in the frequency sweep data rasterizing path loss matrix; For the adjacent cells piece of the grid block that exists in the frequency sweep data rasterizing path loss data, the mean value of its path loss The data frequency sweep rasterizing path loss and coverage prediction path loss; Data in the path loss The data coverage prediction path loss matrix of other grid block.
(3) set up interference matrix
Interference matrix is divided into co-channel interference matrix and adjacent interference matrix two parts frequently.
Obtain the interference matrix between the sub-district in twos according to the path loss matrix, as follows:
The minizone is less than the relevant total grid block number in sub-district of the grid block number of C/I thresholding T/in twos in twos
1) co-channel interference matrix (To be improved)
C/I thresholding T is the interference value of-3 o'clock minizone in twos
2) adjacent interference matrix (To be improved) frequently
C/I thresholding T is-21 o'clock interference values of minizone in twos
3, make up the dynamic evaluation function based on interference model
The dynamic evaluation function comprises simplifies evaluation function and emulation of coverage capability evaluation function, but and the C/I interference threshold dynamic-configuration of evaluation function.Common Control Channel is bigger to the systematic function influence, and Common Control Channel is configured on the main carrier frequency, therefore carries when making up evaluation function, needs to consider farthest to reduce the interference that main carrier frequency is subjected to.
(1) simplifies evaluation function
It is as follows to simplify evaluation function:
K1*∑(∑CoIi,j+∑AdjIi,k)+K2*∑∑(∑∑CoIl,ll,m,mm+∑∑AdjIl,ll,n,nn)
Illustrate:
K1: the interference weight of main carrier and main carrier
K2: the interference weight of auxilliary carrier wave/main carrier and auxilliary carrier wave
CoIi, j: main carrier is with cell i frequently and the co-channel interference value of sub-district j
AdjIi, k: the adjacent interference value frequently of the cell i of the adjacent frequency of main carrier and sub-district j
CoIl, ll, m, mm: under the same situation frequently of the ll carrier wave of sub-district l and the mm carrier wave of sub-district m, the co-channel interference value of sub-district l and sub-district m
AdjIl, ll, n, nn: under the adjacent situation frequently of the ll carrier wave of sub-district l and the nn carrier wave of sub-district n, the adjacent interference value frequently of sub-district l and sub-district n
The K1 default value is 0.8, and the K2 default value is 0.2;
CoIi, j etc. the same frequency between the sub-district or adjacent interference value frequently in twos obtain from interference matrix.
(2) emulation of coverage capability evaluation function
With based on the emulation of coverage capability index C/I of path loss matrix computations gained for mainly considering target, simplify evaluation function relatively, C/I is more accurate.But the C/I index is calculated more complicated, and search performance is had certain influence.The emulation of coverage capability evaluation function is as follows:
K1* main carrier C/I is less than the grid block number of the auxilliary carrier wave C/I of grid block number+K2* of T1 less than T2
Illustrate:
K1: main carrier weight
K2: auxilliary carrier wave weight
T1: main carrier C/I thresholding
T2: auxilliary carrier wave C/I thresholding
Usually main carrier should be paid the utmost attention to, so the K1 acquiescence is made as 0.8, and the K2 acquiescence is made as 0.2;
The reasonability of T1 and T2C/I threshold setting is bigger to evaluation of estimate influence, for addressing this problem, and can be by a plurality of segment thresholds being set and corresponding segment threshold weight solves.
4, according to dynamic evaluation function search rate prioritization scheme
(1) obligates condition
Need consider to obligate condition according to the TD-SCDMA network requirement in the frequency optimization scheme search procedure, as follows:
1) can not be with the station cell main carrier with frequently
2) can not be with cell carrier with frequently
(2) adaptive searching method
No matter searching algorithm uses genetic algorithm or other algorithm, evaluation function is most crucial part all the time.And the selection key of evaluation function is the good search performance of balance and optimize effect.Therefore, simplify evaluation function raising search performance early stage use of search, the search later stage uses the emulation of coverage capability evaluation function in the hope of obtaining the better optimize effect.The selection of evaluation function can realize by adaptive mode.Below provide a kind of determination methods:
(Avg-C)/Avg>0.9
Illustrate:
C: the evaluation function value of current frequency optimization scheme
Avg: the average ratings functional value of the frequency optimization scheme that N search recently obtains, N is defaulted as 10 times.
When above condition is set up, think that search has been tending towards convergence state, evaluation function can be used meticulous relatively emulation of coverage capability evaluation function instead.