WO2021057327A1 - 天馈接反检测方法和装置 - Google Patents

天馈接反检测方法和装置 Download PDF

Info

Publication number
WO2021057327A1
WO2021057327A1 PCT/CN2020/109225 CN2020109225W WO2021057327A1 WO 2021057327 A1 WO2021057327 A1 WO 2021057327A1 CN 2020109225 W CN2020109225 W CN 2020109225W WO 2021057327 A1 WO2021057327 A1 WO 2021057327A1
Authority
WO
WIPO (PCT)
Prior art keywords
training
data
detection
antenna
port
Prior art date
Application number
PCT/CN2020/109225
Other languages
English (en)
French (fr)
Inventor
任牧青
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2021057327A1 publication Critical patent/WO2021057327A1/zh

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W24/00Supervisory, monitoring or testing arrangements
    • H04W24/04Arrangements for maintaining operational condition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B17/00Monitoring; Testing
    • H04B17/10Monitoring; Testing of transmitters
    • H04B17/15Performance testing
    • H04B17/17Detection of non-compliance or faulty performance, e.g. response deviations
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B17/00Monitoring; Testing
    • H04B17/20Monitoring; Testing of receivers
    • H04B17/29Performance testing
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W24/00Supervisory, monitoring or testing arrangements
    • H04W24/08Testing, supervising or monitoring using real traffic

Definitions

  • This application relates to the field of communications, in particular to a method and device for detecting antenna feeder reverse.
  • antenna feeder reversal is a common type of fault caused by hardware installation errors. This type of fault usually has a relatively large impact on network performance indicators, and it is generally not easy to troubleshoot.
  • the existing automatic detection method of antenna feedback connection is mainly realized by using the correlation between the received signals of different antennas and a preset threshold value. If the correlation coefficient exceeds the preset threshold value, it is considered that the connection is reversed.
  • This detection method has found obvious defects through practical applications. The detection accuracy rate is relatively low, and there are many false detections and missed detections.
  • the above-mentioned defects are mainly caused by improper threshold setting. If the threshold is set loosely, As a result, there will be a lot of misjudgments. If the threshold is set to be stricter, it is possible to miss a lot of cases where the actual connection is reversed. Based on the above-mentioned shortcomings, the above-mentioned automatic detection method still cannot meet the conditions for commercial deployment.
  • the embodiments of the present application provide an antenna feedback detection method, a model training method, an electronic device, and a computer-readable storage medium, which can improve the accuracy of antenna feedback detection.
  • an embodiment of the present application provides an antenna feeder reverse detection method, including: acquiring detection data of feeder ports of m sectors under the same base station, where m is an integer greater than 1, and the detection data includes the first Detected data; Obtain port combinations in the case of all the connections of the m sectors; According to the first detected data, calculate the correlation coefficient of each of the port combinations, according to the correlation coefficient from large to small in order from the all Select m port combinations from the port combinations in the case of connection, and the selected m port combinations do not intersect each other; construct a first set according to the m selected port combinations; extract the first set The judging feature set; the judging feature set is input into the antenna feed reverse detection model, and the antenna feed reverse detection model determines whether there is a reverse antenna feed based on the judgment feature set, and outputs the corresponding detection result.
  • an embodiment of the present application provides a model training method for obtaining an antenna feed reverse detection model.
  • the method includes: obtaining a training set; using the training set for training; outputting a model file;
  • the obtaining of the training set specifically includes: obtaining training sample data of the feeder ports of r sectors of the same base station, where r is an integer greater than 1, and the training sample data includes the first training data; and obtaining the r sectors Port combinations in all connection cases; calculate the correlation coefficient of each port combination according to the first training data, and select r from the port combinations in all connection cases in descending order of the correlation coefficient.
  • Port combinations, and the selected r port combinations do not intersect each other; construct a third set according to the selected r port combinations; extract the training feature set of the third set; corresponding to the training The feature set is to mark whether there is a reversed result to obtain a training target; the training feature set and the training target are merged and added to the training set.
  • an embodiment of the present application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the computer program executes the antenna feeder connection when the computer program is running.
  • Anti-detection method or the described model training method are examples of the present application.
  • an embodiment of the present application provides a computer-readable storage medium storing computer-executable instructions for executing the antenna feed reverse detection method or the model training method .
  • Figure 1A is a schematic diagram of an antenna feed reverse scenario
  • Figure 1B is a schematic diagram of another antenna feed reverse scenario
  • FIG. 2 is a flowchart of a method for detecting reverse antenna feeder connection provided by an embodiment of the present application
  • FIG. 3 is a schematic diagram of the detection result of an embodiment of the present application.
  • FIG. 4 is a flowchart of another antenna feed reverse detection method provided by an embodiment of the present application.
  • FIG. 5 is a flowchart of a model training method provided by an embodiment of the present application.
  • Fig. 6 is a flowchart of a method for obtaining a training set provided by an embodiment of the present application
  • FIG. 7 is a flowchart of another method for obtaining a training set provided by an embodiment of the present application.
  • FIG. 8 is a flowchart of another model training method provided by an embodiment of the present application.
  • FIG. 9 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • antenna feeder reversal is a common type of fault caused by hardware installation errors. This type of fault usually has a relatively large impact on network performance indicators, and it is generally not easy to troubleshoot.
  • the existing automatic detection method of antenna feedback is mainly realized by calculating the correlation between the received signals of different antennas, such as using RTWP (Received Total Wideband Power) correlation or RSSI (Received Signal Strength Indication) , detect whether the antenna is reversed.
  • RTWP Receiveived Total Wideband Power
  • RSSI Receiveived Signal Strength Indication
  • This type of automatic detection method essentially relies on the correlation coefficient threshold between the signals received by different antennas for detection, and if it exceeds the preset threshold, it is considered to be reversed.
  • the embodiments of the present application provide an antenna feed reverse detection method, a model training method, an electronic device, and a computer-readable storage medium, which can improve the accuracy of antenna feed reverse detection.
  • a base station 100 includes antennas, feeders, and sectors. Specifically, a base station 100 is usually provided with multiple sectors, and each sector is provided with at least two feeder ports 110. The antenna in charge of each sector is connected to the feeder port 110 of the corresponding sector through the feeder 120 to realize the pairing.
  • the correct connection method of the antenna is: the two feeders 120 connected to the antenna 1 should be connected to the two feeder ports 110 of the sector 1 to form the signal coverage of the sector 1, and the antenna 2
  • the two connected feeders 120 should be connected to the two feeder ports 110 of the sector 2 correspondingly to form the signal coverage of the sector 2.
  • the solution provided by the embodiment of the present application is mainly used to detect whether the above two types of reverse connection problems exist in the antenna feeders 120 of all sectors under the same base station 100.
  • an embodiment of the present application provides an antenna feed reverse detection method. Please refer to FIG. 2.
  • the method includes but is not limited to the following steps:
  • S1001 Obtain detection data of feeder ports of m sectors under the same base station, where m is an integer greater than 1, and generally, m represents the total number of sectors included in the base station.
  • the detection data should include at least the first detection data.
  • the first detection data may be RTWP data, where RTWP is the total broadband received power, which can reflect the power of the sector feeder port signal.
  • the first detection data may be RSSI data, and the RSSI is a received signal strength indicator value, which can reflect the strength of the received signal at the sector feeder port.
  • the first detection data may also be other data that can reflect the energy/intensity of the signal, and this application does not impose excessive restrictions on this.
  • the detection data further includes second detection data, and the second detection data may specifically be balance count data.
  • the number of balances reflects the degree of balance of the signal strength of any two feeder ports in the sector. Specifically, the energy received by these two feeder ports can be measured periodically to determine whether the energy difference between the two feeder ports is greater than the preset threshold. If yes, add one to the balance times of the feeder port with high energy; if not, add one to the balance times of the two feeder ports at the same time. The smaller the difference between the balance times of the two feeder ports, the more balanced the received signal strength of the two feeder ports, and vice versa, the more unbalanced the received signal strength.
  • the detection data further includes third detection data, and the third detection data may specifically be MIMO activation ratio data.
  • the MIMO activation ratio is the ratio of the time the user activates the MIMO to the total time the user occupies the service (the total time includes the time when MIMO is activated and the time when MIMO is not activated). It is only possible to activate MIMO when the user can receive the two signals from the same sector. In the case of reverse antenna feeder, the user generally can only receive one signal from the same sector, so it is difficult to activate MIMO, resulting in a relatively low proportion of MIMO activation.
  • the aforementioned RTWP data, RSSI data, balance number data, and MIMO activation ratio data can all be directly obtained from the performance data output by the base station to the data center.
  • S1003 Calculate the correlation coefficient of each port combination according to the first detection data, and select m port combinations from all the port combinations under connection conditions in descending order of the correlation coefficient, and select the elements of the m port combinations Disjoint each other.
  • the correlation coefficient of each port combination is calculated. Specifically, for each possible combination, the correlation coefficients of any two ports in the combination can be calculated and summed.
  • ⁇ (1,2,3,4) ⁇ (1,2)+ ⁇ (1,3)+ ⁇ (1,4)+ ⁇ (2,3)+ ⁇ (2,4)+ ⁇ (3, 4);
  • the correlation coefficient of any two ports can be calculated by the following formula:
  • the array X and the array Y refer to the first detection data of the two ports, for example, RSTW data or RSSI data.
  • m port combinations are selected from all the port combinations under the connection conditions in order of correlation coefficients from large to small, and the elements of the selected m port combinations do not intersect each other, namely: In the process of selecting m port combinations from large to small according to the correlation coefficient, for each port combination selected, the elements contained in the currently selected port combination cannot be included in the previously selected port combination.
  • the combination with the largest correlation coefficient is (1_1, 1_2, 1_3, 1_4), so this combination is selected as the first combination, and the combination with the second largest correlation coefficient is ( 1_1, 2_2, 2_3, 2_4), here because port 1_1 has appeared in the first combination above, the combination with the second largest correlation coefficient cannot be selected.
  • m port combinations are selected in turn, and m ports The combined elements do not intersect each other.
  • step S1004 Construct a first set according to the selected m port combinations. Based on the fact that these port combinations selected in step S1003 are selected in descending order of correlation coefficients, the high correlation of the port combinations indicates that the ports in the combination are likely to be connected to the same antenna device, so the first set is to a large extent Reflects the set of port combinations that are consistent with the current actual wiring situation, and subsequently can determine whether the antenna feeder is connected reversely according to the characteristics of the first set.
  • each sector has n antennas, and the first set is P;
  • the correlation coefficient of each combination in P is: ⁇ p1, ⁇ p2, ..., ⁇ pm;
  • the value of the first detection data (which can be RTWP/RSSI data or similar related data) corresponding to P is: ⁇ (rs_p11,rs_p12,...,rs_p1n),(rs_p21,rs_p22,...,rs_p2n),...,(rs_pm1 ,rs_pm2,...,rs_pmn) ⁇ ;
  • the value of the second detection data corresponding to P (here is the balance number data): ⁇ (bl_p11,bl_p12,...,bl_p1n),(bl_p21,bl_p22,...,bl_p2n),...,(bl_pm1,bl_pm2,...,bl_pmn ) ⁇ ;
  • the value of the third detection data corresponding to P (here is the Mimo activation ratio data): ⁇ (mi_p11,mi_p12,...,mi_p1n),(mi_p21,mi_p22,...,mi_p2n),...,(mi_pm1,mi_pm2,..., mi_pmn) ⁇ ;
  • the correlation coefficient of each combination in A is: ⁇ a1, ⁇ a2, ..., ⁇ am;
  • the value of the first detection data (which can be RTWP data, RSSI data or similar related data) corresponding to A is: ⁇ (rs_a11,rs_a12,...,rs_a1n),(rs_a21,rs_a22,...,rs_a2n),...,( rs_am1,rs_am2,...,rs_amn) ⁇ ;
  • the value of the second detection data corresponding to A (here, the number of balance data) is ⁇ (bl_a11,bl_a12,...,bl_a1n),(bl_a21,bl_a22,...,bl_a2n),...,(bl_am1,bl_am2,...,bl_amn) ⁇ ;
  • the value of the third detection data corresponding to A (here is the Mimo activation ratio data) is ⁇ (mi_a11,mi_a12,...,mi_a1n),(mi_a21,mi_a22,...,mi_a2n),...,(mi_am1,mi_am2,...,mi_amn ) ⁇ .
  • judgment features are extracted based on the above data, and a judgment feature set is constructed:
  • the absolute value of the difference between the first detected data of the first set P is:
  • rs_p abs(max(rs_p11,rs_p12,...,rs_p1n)-min(rs_p11,rs_p12,...,rs_p1n))+abs(max(rs_p21,rs_p22,...,rs_p2n)-min(rs_p21,rs_p22,...,rs_p2n) )+...+abs(max(rs_pm1,rs_pm2,...,rs_pmn)-min(rs_pm1,rs_pm2,...,rs_pmn));
  • the absolute value of the difference between the first detection data of the second set A is:
  • rs_a abs(max(rs_a11,rs_a12,...,rs_a1n)-min(rs_a11,rs_a12,...,rs_a1n))+abs(max(rs_a21,rs_a22,..., rs_a2n)-min(rs_a21,rs_a22,...,rs_a2n) )+...+abs(max(rs_am1,rs_am2,...,rs_amn)-min(rs_am1,rs_am2,...,rs_amn));
  • the absolute value growth rate of the first detected data difference between the first set P and the second set A is:
  • relation_cross_min min( ⁇ p1, ⁇ p2, ..., ⁇ pm).
  • the minimum value of the sample point after deduplication in the first set P is:
  • range_cross_num min(min(rs_nump11,rs_nump12,...,,rs_nump1n),min(rs_nump21,rs_nump22,...,rs_nump2n),...,min(rs_numpm1,rs_numpm2,...,rs_numpmn));
  • the minimum value of the fluctuation range of the first detection data of the first set P is:
  • range_cross_min min(min(rs_rangep11, rs_rangep12,...,, rs_rangep1n), min(rs_rangep21, rs_rangep22,..., rs_rangep2n)..., min(rs_rangepm1, rs_rangepm2,..., rs_rangepmn)).
  • the maximum value of the fluctuation range of the first detection data of the first set P is:
  • range_cross_max max(max(rs_rangep11,rs_rangep12,...rs_rangep1n), max(rs_rangep21,rs_rangep22,...,rs_rangep2n),...,max(rs_rangepm1,rs_rangepm2,...,rs_rangepmn)).
  • range_cross_rate range_cross_max/range_cross_min.
  • bl_p abs(max(bl_p11,bl_p12,...,bl_p1n)-min(bl_p11,bl_p12,...,bl_p1n))+abs(max(bl_p21,bl_p22,...,bl_p2n) -min(bl_p21,bl_p22,...,bl_p2n))+...+abs(max(bl_pm1,bl_pm2,...,bl_pmn)-min(bl_pm1,bl_pm2,...,bl_pmn));
  • bl_a abs(max(bl_a11,bl_a12,...,bl_a1n)-min(bl_a11,bl_a12,...,bl_a1n))+abs(max(bl_a21,bl_a22,...,bl_a2n) -min(bl_a21,bl_a22,...,bl_a2n))+...+abs(max(bl_am1,bl_am2,...,bl_amn)-min(bl_am1,bl_am2,...,bl_amn));
  • mimo_rate min(min(mi_meanp11,mi_meanp12,...mi_meanp1n),min(mi_meanp21,mi_meanp22,...,mi_meanp2n)...,min(mi_meanpm1,mi_meanpm2,...mi_meanpmn)).
  • judgment feature set can include any one or more of the aforementioned judgment features (1) to (9), and this application does not make too many restrictions.
  • test results can include one or more of the following:
  • the site ID or site name, cell frequency point information, and sector number for which antenna connection is present can be obtained at the same time as the detection data is obtained in step S1001; the port number or port number pair with antenna connection can be obtained according to the One set is obtained. Specifically, the first set is compared with the second set. The port combinations of the first set and the second set are different, that is, the port pair that has the anti-correspondence of the antenna connection, and the corresponding port can be obtained according to the port pair. number.
  • the output result of the antenna connection reverse detection model can be displayed in a table, which contains the site-carrier frequency information with the antenna connection reverse situation, for example, the standard ("Product” in the figure) Column), site name (the “Site Name” column in the figure), the frequency band that is connected reversely (the “Band” column in the figure), whether there is a reverse connection (the "Antenna cross result” column in the figure), and the fan that is connected reversely Zone number and corresponding feeder port number ("Detail” column in the figure).
  • Fig. 4 shows a method for detecting antenna feed reverse provided by another embodiment of the present application.
  • the method includes:
  • step S2001 Obtain detection data of feeder ports of m sectors under the same base station, where m is an integer greater than 1.
  • m is an integer greater than 1.
  • step S1001 the specific implementation of step S2001 can refer to the related description of step S1001, which will not be repeated here.
  • abnormal data can include but not limited to the following data: persistent abnormal data (all the same value); occasionally abnormal data (the same value for a period of time); data with too few sampling points; too little fluctuation (no business volume) ) Data; abnormal data for individual points.
  • step S2003 Obtain port combinations in all connection cases of m sectors.
  • step S2003 can refer to the related description of step S1002, which will not be repeated here.
  • step S2004 Calculate the correlation coefficient of each port combination according to the first detection data, and select m port combinations from all the port combinations under connection conditions in descending order of the correlation coefficient, and the selected m port combinations are not mutually exclusive intersect.
  • step S2004 can refer to the related description of step 1003, which will not be repeated here.
  • step S2005 the first set is constructed according to the selected m port combinations.
  • step S2005 can refer to the related description of step S1004, which will not be repeated here.
  • S2007 Calculate the growth rate of the correlation coefficient between the first set and the second set.
  • the detection result can be directly output: the antenna feeder is not connected; when the growth rate is greater than the preset threshold At the time, it is determined that there may be a reverse connection, and proceed to the next step.
  • the correlation coefficients of the first set of P and the second set of A are ⁇ (P) and ⁇ (A) respectively, and the preset threshold is Threshold, if ( ⁇ (P)- ⁇ (A))/ ⁇ (A) ⁇ Threshold, then output the test result: no antenna feeder is reversed; if ( ⁇ (P)- ⁇ (A))/ ⁇ (A)>Threshold, it is considered that there may be a reversed antenna feeder, and proceed to the next step.
  • the threshold here should be set relatively loosely, for example, in a range close to 0 (for example, the threshold is set in the range of 0.01 to 0.03) to ensure that possible reverse connections are not missed.
  • step S2008 extract the judgment feature set of the first set.
  • the specific implementation of step S2008 can refer to the related description of step S1005, which will not be repeated here.
  • step S2009 the judgment feature set is input into the antenna connection reverse detection model, and the antenna connection reverse detection model judges whether there is an antenna connection reverse based on the judgment feature set, and outputs the corresponding detection result.
  • the specific implementation of step S2009 can refer to the related description of step S1006, which will not be repeated here.
  • the embodiment shown in Fig. 4 realizes the pre-detection of antenna reversal through step S2006 and/or step S2007. Firstly, the obvious no reversal is excluded. For the possible reversal, the subsequent steps are used for further detection. Judging, this can not only increase the detection speed, but also ensure the detection accuracy.
  • the process of realizing the antenna feed reverse detection method may only include any one or two of the above-mentioned steps S2002, S2006, and S2007.
  • the order of step S2006 and step S2007 may be exchanged.
  • m port combinations are selected according to the correlation coefficient in descending order to construct the first set, and the first set thus obtained can reflect to a large extent A collection of port combinations that are consistent with the current actual wiring situation, so based on the first set, connection reverse judgment detection can be performed and the current connection reverse combination can be obtained.
  • connection reverse judgment detection can be performed and the current connection reverse combination can be obtained.
  • several features of the first set are extracted to construct a judgment feature set, and the judgment feature set is input into the antenna connection detection model for processing, and the detection result is output. In some cases, only RSSI/ The RTWP correlation is used as the basis for connection reverse judgment, and the detection accuracy of the solution of this application is higher.
  • FIG. 5 shows a model training method provided by an embodiment of the present application.
  • the model training method is used to obtain an antenna connection reverse detection model, and the judgment feature set can be processed through the antenna connection reverse detection model. And get the test results.
  • the model training methods include:
  • obtaining the training set may specifically include the following steps:
  • S3101 Obtain training sample data of feeder ports of r sectors under the same base station, where r is an integer greater than 1, and generally, m represents the total number of sectors included in the base station.
  • the training sample data should include at least the first training data.
  • the first training data may be RTWP data or RSSI data.
  • the first training data may also be other data that can reflect the energy/intensity of the signal, and this application does not make too many restrictions on this.
  • the training sample data further includes second training data, and the second training data may specifically be balance number data.
  • the training sample data further includes third training data, and the third training data may specifically be MIMO activation ratio data.
  • the aforementioned RTWP data, RSSI data, balance number data, and MIMO activation ratio data can all be directly obtained from the performance data output from the base station to the data center.
  • step S3102 Obtain port combinations in all connections of r sectors.
  • step S3102 is similar to the above-mentioned step S1002, and the relevant description of the above-mentioned step S1002 can be referred to, which will not be repeated here.
  • S3103 Calculate the correlation coefficients of each port combination according to the first training data, and select r port combinations from all the port combinations under connection conditions in descending order of the correlation coefficients, and the selected r port combinations are not mutually exclusive intersect.
  • the correlation coefficient of each port combination is calculated. Specifically, for each possible combination, the correlation coefficients of any two ports in the combination can be calculated and summed.
  • step S3103 is similar to the above-mentioned step S1003, and the relevant description of the above-mentioned step S1003 can be referred to, which will not be repeated here.
  • S3104 Construct a third set according to the selected r port combinations.
  • the port combinations selected in step S3103 are selected in descending order of correlation coefficients, and the high correlation of the port combinations indicates that the ports in the combination are likely to be connected to the same antenna device, so the third set reflects to a large extent A collection of port combinations that are consistent with the current actual wiring situation, and subsequently can be trained to detect whether the antenna feeder is connected reversely according to the characteristics of the third set.
  • the training feature set may include any one or more of the following:
  • the training feature set can be numbered, and a graph (for example, RSSI/RTWP graph) can be drawn according to the obtained first training data, and the result can be annotated according to the RSSI/RTWP graph, for example, it can be represented by "0" If the connection is not reversed, use "1" to indicate reverse connection.
  • a graph for example, RSSI/RTWP graph
  • S3107 Combine the training feature set and the training target, and add them to the training set. Use the obtained training set to train the antenna feed reverse detection model.
  • the process of obtaining the training set further includes an optional step S3108.
  • step S3108 it is detected whether the training sample data contains abnormal data, and if it contains abnormal data, the abnormal data is eliminated to improve the accuracy of the training sample data and avoid the adverse effect of abnormal values on the accuracy of the model.
  • abnormal data can include but not limited to the following data: persistent abnormal data (all the same value); occasionally abnormal data (the same value for a period of time); data with too few sampling points; too little fluctuation (no business volume) ) Data; abnormal data for individual points.
  • the process of obtaining the training set further includes an optional step S3109.
  • step S3109 the fourth set is constructed according to the port combination of the feeder ports of the m sectors when the feeder ports are correctly connected; when the third set is equal to the fourth set, the training sample data is removed, and when the third set and the fourth set are equal When the sets are not equal, go to the next step.
  • S3110 Calculate the growth rate of the correlation coefficient of the third set relative to the fourth set.
  • the correlation coefficients of the third set of P and the fourth set of A are ⁇ (P) and ⁇ (A) respectively, and the preset threshold is Threshold, if ( ⁇ (P)- ⁇ (A))/ ⁇ (A) ⁇ Threshold, then output the test result: no antenna feeder is reversed; if ( ⁇ (P)- ⁇ (A))/ ⁇ (A)>Threshold, it is considered that there may be a reversed antenna feeder, and proceed to the next step.
  • the threshold here should be set relatively loosely, for example, set in the range close to 0, to ensure that the possibility of reverse connection will not be missed.
  • the antenna feeder is pre-judged whether the antenna feeder is connected reversely, and the training sample data that can be obviously obtained without reverse connection is eliminated. For the situation where there may be reverse connection , And further perform training feature extraction on the third set. In this way, on the one hand, the amount of annotations in the training set can be reduced, and on the other hand, it is beneficial to obtain a training set with higher training value, which in turn makes the detection results output by the antenna feed reverse detection model more accurate.
  • the process of obtaining the training set may only include any one or two of the foregoing steps S3108, S3109, and S3110.
  • the order of step S3109 and step S3110 can be exchanged.
  • step S3200 training is performed using the training set, which may specifically be training through a machine learning algorithm.
  • the machine learning algorithm may be any one of logistic regression, random forest, support vector machine, and deep learning, which is not limited in the embodiment of the application.
  • step S3400 when the model training method is implemented, an optional step S3400 can be included.
  • the weight of each feature in the training feature set is adjusted to reduce the accuracy and recall rate. Control within the acceptable range, and finally output the model file.
  • the third set thus obtained can reflect the current situation to a large extent.
  • FIG. 9 shows that an embodiment of the present application provides the electronic device 200 provided by the embodiment of the present application.
  • the electronic device 200 includes: a memory 220, a processor 210, and a computer program stored on the memory 220 and capable of running on the processor 210.
  • the computer program is used to perform any one of the antenna reverse detection described in the first aspect when the computer program is running. Method or any of the model training methods described in the second aspect.
  • the processor 210 and the memory 220 may be connected by a bus or in other ways.
  • the memory 220 can be used to store non-transitory software programs and non-transitory computer-executable programs, as described in the first aspect of the embodiments of this application.
  • the processor 210 runs the non-transitory software programs and instructions stored in the memory 220 to implement any one of the day feed reverse detection methods described in the first aspect or any one of the model training methods described in the second aspect.
  • the memory 220 may include a storage program area and a storage data area, where the storage program area can store an operating system and an application program required by at least one function; the storage data area can store any one of the day feeds described in the first aspect.
  • the memory 220 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices.
  • the memory 220 may optionally include memories remotely provided with respect to the processor 210, and these remote memories may be connected to the terminal through a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
  • the non-transient software programs and instructions required to implement any one of the antenna reverse detection methods described in the first aspect or any of the model training methods described in the second aspect are stored in the memory 220, and when used by one or more When the processor 210 executes, it executes any one of the antenna feed reverse detection methods described in the first aspect or any of the model training methods described in the second aspect, for example, executes steps S1001 to S1006 of the method described in FIG. 2, as shown in FIG.
  • the method steps S2001 to S2009 described in 4 the method steps S3100 to S3300 described in Fig. 5, the method steps S3101 to S3107 described in Fig. 6, the method steps S3101 to S3110 described in Fig. 7, the method steps described in Fig. 8 S3100 to S3300.
  • the embodiments of the present application also provide a computer-readable storage medium that stores computer-executable instructions, and the computer-executable instructions are used to execute any one of the antenna feed reverse detection methods described in the first aspect or any one of the methods described in the second aspect. Item model training method.
  • the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by one or more control processors 210, for example, executed by one processor 210 in the aforementioned electronic device 200,
  • the above-mentioned one or more processors 210 can be made to execute any one of the antenna reverse detection methods described in the first aspect or any one of the model training methods described in the second aspect, for example, execute the method step S1001 described in FIG. 2 To S1006, the method steps S2001 to S2009 described in Fig. 4, the method steps S3100 to S3300 described in Fig. 5, the method steps S3101 to S3107 described in Fig. 6, the method steps S3101 to S3110 described in Fig. 7, in Fig. 8 The described method steps S3100 to S3300.
  • the embodiment of the application includes: obtaining detection data of feeder ports of m sectors under the same base station, where m is an integer greater than 1, and the detection data includes the first detection data; obtaining all connection conditions of the m sectors According to the first detection data, calculate the correlation coefficient of each of the port combinations, and select m port combinations from the port combinations under all connection conditions in descending order according to the correlation coefficient , And the selected m port combinations do not intersect each other; construct a first set according to the m selected port combinations; extract the judgment feature set of the first set; input the judgment feature set into days
  • a feed reverse detection model, the antenna feed reverse detection model determines whether there is a reverse antenna feed according to the judgment feature set, and outputs a corresponding detection result.
  • the antenna feed reverse detection accuracy rate is relatively high, and the misjudgment and missed judgment rate are extremely low.
  • the device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • computer storage medium includes volatile and non-volatile data implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data).
  • Information such as computer-readable instructions, data structures, program modules, or other data.
  • Computer storage media include but are not limited to RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or Any other medium used to store desired information and that can be accessed by a computer.
  • communication media usually contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as carrier waves or other transmission mechanisms, and may include any information delivery media. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Radio Transmission System (AREA)

Abstract

一种天馈接反检测方法、模型训练方法、电子装置和计算机可读存储介质。其中,天馈接反检测方法包括:获取同一基站下m个扇区的馈线端口的检测数据S(1001);按照所述相关系数由大到小依次从所述所有连接情况下的端口组合中选出m个端口组合S(1003);根据选出的所述m个端口组合,构建第一集合S(1004);提取所述第一集合的判断特征集S(1005);将所述判断特征集输入天馈接反检测模型,并输出相应的检测结果S(1006)。模型训练方法包括:获取训练集S(3100);使用所述训练集进行训练S(3200);输出模型文件S(3300)。

Description

天馈接反检测方法和装置
相关申请的交叉引用
本申请基于申请号为201910900620.3、申请日为2019年9月23日的中国专利申请提出,并要求该中国专利申请的优先权,该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请涉及通信领域,特别是涉及天馈接反检测方法和装置。
背景技术
在无线通信网络中,天馈接反是一类常见的硬件安装错误导致的故障,这类故障通常会给网络性能指标带来比较大的影响,且一般不太容易排查。现有的天馈接反的自动检测方法,主要利用不同天线的接收信号间的相关性和预设得门限值来实现,相关系数超过预设门限值即认为接反。这种检测方法通过实际应用发现存在比较明显的缺陷,检测准确率比较低,误检和漏检情况较多,上述缺陷主要是门限值设置不恰当造成,如果把门限值设置得比较宽松,结果中就会存在大量误判,如果把门限设置得比较严格,就有可能漏掉很多实际确实接反的情况。基于上述缺陷,上述自动检测方法还无法满足商用部署的条件。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
一方面,本申请实施例提供了天馈接反检测方法、模型训练方法、电子装置和计算机可读存储介质,可以提高天馈接反检测的准确率。
另一方面,本申请实施例提供了天馈接反检测方法,包括:获取同一基站下m个扇区的馈线端口的检测数据,其中,m为大于1的整数,所述检测数据包含第一检测数据;获取所述m个扇区所有连接情况下的端口组合;根据所述第一检测数据,计算各个所述端口组合的相关系数,按照所述相关系数由大到小依次从所述所有连接情况下的端口组合中选出m个端口组合,且选出的所述m个端口组合互不相交;根据选出的所述m个端口组合,构建第一集合;提取所述第一集合的判断特征集;将所述判断特征集输入天馈接反检测模型,所述天馈接反检测模型根据所述判断特征集,判断是否存在天馈接反情况,并输出相应的检测结果。
另一方面,本申请实施例提供了一种模型训练方法,用于获取天馈接反检测模型,所述方法包括:获取训练集;使用所述训练集进行训练;输出模型文件;其中,所述获取训练集具体包括:获取同一基站下r个扇区的馈线端口的训练样本数据,其中,r为大于1的整数,所述训练样本数据包含第一训练数据;获取所述r个扇区所有连接情况下的端口组合;根据所述第一训练数据,计算各个所述端口组合的相关系数,按照所述相关系数由大到小依次从所述所有连接情况下的端口组合中选出r个端口组合,且选出的所述r个端口组合互不相交;根据选出的所述r个端口组合,构建第三集合;提取所述第三集合的训练特征集;对应于所述训练特征集,对是否存在接反的结果进行标注,得到训练目标;将 所述训练特征集和所述训练目标进行合并,加入训练集中。
另一方面,本申请实施例提供了一种电子装置,包括:存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序,所述计算机程序运行时执行所述的天馈接反检测方法或者所述的模型训练方法。
再一方面,本申请实施例提供了一种计算机可读存储介质,存储有计算机可执行指令,所述计算机可执行指令用于执行所述的天馈接反检测方法或者所述的模型训练方法。
本申请的其它特征和优点将在随后的说明书中阐述,并且,部分地从说明书中变得显而易见,或者通过实施本申请而了解。本申请的目的和其他优点可通过在说明书、权利要求书以及附图中所特别指出的结构来实现和获得。
附图说明
附图用来提供对本申请技术方案的进一步理解,并且构成说明书的一部分,与本申请的实施例一起用于解释本申请的技术方案,并不构成对本申请技术方案的限制。
图1A是一种天馈接反场景示意图;
图1B是另一种天馈接反场景示意图;
图2是本申请实施例提供的一种天馈接反检测方法的流程图;
图3是本申请实施例的检测结果的示意图;
图4是本申请实施例提供的另一种天馈接反检测方法的流程图;
图5是本申请实施例提供的一种模型训练方法的流程图;
图6是本申请实施例提供的一种获取训练集方法的流程图;
图7是本申请实施例提供的另一种获取训练集方法的流程图;
图8是本申请实施例提供的另一种模型训练方法的流程图;
图9是本申请实施例提供的一种电子装置的结构示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅用以解释本申请,并不用于限定本申请。
需要说明的是,在本申请实施例的描述中,若干的含义是一个或者多个,多个(或多项)的含义是两个以上,大于、小于、超过等理解为不包括本数,以上、以下、以内等理解为包括本数。如果有描述到“第一”、“第二”、“第三”、“第四”等(如果存在)只是用于区分技术特征为目的,而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量或者隐含指明所指示的技术特征的先后关系。
本申请实施例的描述中,除非另有明确的限定,设置、安装、连接等词语应做广义理解,所属技术领域技术人员可以结合技术方案的具体内容合理确定上述词语在本申请实施例中的具体含义。
在无线通信网络中,天馈接反是一类常见的硬件安装错误导致的故障,这类故障通常会给网络性能指标带来比较大的影响,且一般不太容易排查。现有的天馈接反的自动检测方法,主要通过计算不同天线的接收信号间的相关性来实现,例如利用RTWP(Received Total Wideband Power,宽带接收总功率)相关性或者RSSI(Received Signal Strength  Indication,接收的信号强度)相关性,检测判断天线是否接反。这类自动检测方法本质上都依赖于不同天线接收信号间的相关系数门限进行检测,超过预设门限值即认为接反。由于RTWP/RSSI数据波动的剧烈程度和复杂程度,如果把门限值设置的比较宽松,结果中就会存在大量误判,如果把门限设置的比较严格,就有可能漏掉很多实际确实接反的情况,从而无法满足商用部署的条件。
因此,本申请实施例提供了天馈接反检测方法、模型训练方法、电子装置和计算机可读存储介质,可以提高天馈接反检测的准确率。
下面对本申请实施例所适用的场景进行介绍:
通常情况下,一个基站100包括天线、馈线、扇区。具体地,一个基站100下通常设有多个扇区,每个扇区设有至少两个馈线端口110,负责各扇区的天线通过馈线120连接到对应扇区的馈线端口110上,实现对预设扇区方向角的信号覆盖。例如,如图1A和图1B中,天线正确的接法是:天线1连接的两根馈线120应当对应地与扇区1的两个馈线端口110连接,形成扇区1的信号覆盖,天线2连接的两根馈线120应当对应地与扇区2的两个馈线端口110连接,形成扇区2的信号覆盖。但是,图1A中扇区1和扇区2的天线馈线120完全接反(俗称大鸳鸯),图1B中扇区1和扇区2的部分天线馈线120交叉(俗称小鸳鸯)。本申请实施例提供的方案主要用于对同一基站100下所有扇区的天线馈线120是否存在上述两类接反问题进行检测。
下面结合附图,对本申请实施例作进一步阐述。
第一方面,本申请的一个实施例提供了一种天馈接反检测方法,请参照图2,该方法包括但不限于如下步骤:
S1001,获取同一基站下m个扇区的馈线端口的检测数据,其中m为大于1的整数,一般性地,m代表该基站包含的扇区总数。
本实施例中,检测数据应当至少包含第一检测数据。一个示例中,第一检测数据可以是RTWP数据,RTWP为宽带接收总功率,能够反映扇区馈线端口信号的功率。另一个示例中,第一检测数据可以是RSSI数据,RSSI为接收信号强度指示值,能够反映扇区馈线端口接收信号的强度。当然,第一检测数据还可以是其他能够反映信号能量/强度的数据,本申请对此不作过多的限制。
在一些实施例中,检测数据还包含第二检测数据,第二检测数据具体可以是平衡次数数据。平衡次数反映扇区的任意两个馈线端口信号强度的平衡程度,具体可以通过周期性地测量这两个馈线端口接收的能量,判断两个馈线端口的能量差是否大于预设的门限值,若是,则对能量大的馈线端口平衡次数加一;若否,则两个馈线端口的平衡次数同时加一。两个馈线端口的平衡次数相差越小说明两个馈线端口接收信号强度越平衡,反之说明接收信号强度越不平衡。
在一些实施例中,检测数据还包含第三检测数据,第三检测数据具体可以是MIMO激活比例数据。MIMO激活比例为用户激活MIMO的时长相对用户占用业务的总时长(总时长包括激活MIMO的时长和未激活MIMO的时长)的比率。只有当用户能够接收到同一个扇区发出的两路信号时才有可能激活MIMO,在天馈接反的情况下,用户一般只能接收到同一个扇区发出的一路信号,因此很难激活MIMO,导致MIMO激活比例比较低。
具体实现时,上述的RTWP数据、RSSI数据、平衡次数数据和MIMO激活比例数据均可以直接从基站输出至数据中心的性能数据中获取。
S1002,获取m个扇区所有连接情况下的端口组合。
示例性地,假设某一基站站点共设有m个扇区,每个扇区中有n个馈线端口,那么把n*m个馈线端口分成m组的所有可能的组合数为:(C **C **…*C)÷A。例如,一种常见的情况是,一个基站下设有3个扇区,每个扇区设有4个馈线端口,那么一共有:
Figure PCTCN2020109225-appb-000001
种端口组合。
S1003,根据第一检测数据,计算各个端口组合的相关系数,按照相关系数由大到小依次从所有连接情况下的端口组合中选出m个端口组合,且选出的m个端口组合的元素互不相交。
示例性地,计算各个端口组合的相关系数,具体可以对每一种可能的组合,计算组合内任意两两端口的相关系数并求和。
例如,某个组合包含端口1,2,3,4,该组合的相关系数之和为:
ρ(1,2,3,4)=ρ(1,2)+ρ(1,3)+ρ(1,4)+ρ(2,3)+ρ(2,4)+ρ(3,4);
其中,任意两两端口的相关系数可以通过如下公式计算得出:
Figure PCTCN2020109225-appb-000002
其中,
Figure PCTCN2020109225-appb-000003
Figure PCTCN2020109225-appb-000004
其中,
Figure PCTCN2020109225-appb-000005
分别为数组X和数组Y的平均数,i=1,2,…,N。
这里数组X和数组Y是指两个端口的第一检测数据,例如,RSTW数据或者RSSI数据。
当得到各个端口组合的相关系数后,按照相关系数由大到小依次从所有连接情况下的端口组合中选出m个端口组合,且选出的m个端口组合的元素互不相交,即:按照相关系数由大到小选出m个端口组合的过程中,每选出一个端口组合,当前选出的端口组合包含的元素均不能包含在之前选出的端口组合中。举例来说,假设步骤S1002得到的所有组合中,相关系数最大的组合为(1_1,1_2,1_3,1_4),因此将该组合作为第一个组合选出,相关系数第二大的组合为(1_1,2_2,2_3,2_4),这里由于端口1_1已在上述第一个组合出现,所以该相关系数第二大的组合不能选,基于此规则,依次选出m个端口组合,并且m个端口组合的元素互不相交。
S1004,根据选出的m个端口组合,构建第一集合。基于步骤S1003中选出这些端口组合是按照相关系数由大到小依次选出的,而端口组合的相关性高说明组合中的端口很可 能接在同一天线设备上,因此第一集合很大程度反映与当前实际接线情况相符的端口组合的集合,后续可以根据第一集合的特征确定天馈有无接反。
S1005,提取第一集合的判断特征集。
以下为对步骤S1005的进一步示例性说明。
假设有m个扇区,每个扇区中有n个天线,并设第一集合为P;
P中每个组合的相关系数为:ρp1,ρp2,…,ρpm;
P的相关系数为各组合相关系数之和:ρ(P)=ρp1+ρp2+…+ρpm;
P所对应的第一检测数据(可以是RTWP/RSSI数据或者类似的相关数据)的值为:{(rs_p11,rs_p12,…,rs_p1n),(rs_p21,rs_p22,…,rs_p2n),…,(rs_pm1,rs_pm2,…,rs_pmn)};
P所对应的第二检测数据(这里是平衡次数数据)的值为:{(bl_p11,bl_p12,…,bl_p1n),(bl_p21,bl_p22,…,bl_p2n),…,(bl_pm1,bl_pm2,…,bl_pmn)};
P所对应的第三检测数据(这里是Mimo激活比例数据)的值为:{(mi_p11,mi_p12,…,mi_p1n),(mi_p21,mi_p22,…,mi_p2n),…,(mi_pm1,mi_pm2,…,mi_pmn)};
设第二集合为A;
A中每个组合的相关系数为:ρa1,ρa2,…,ρam;
A的相关系数为各组合相关系数之和:为ρ(A)=ρa1+ρa2+…+ρam;
A所对应的第一检测数据(可以是RTWP数据、RSSI数据或者类似的相关数据)的值为:{(rs_a11,rs_a12,…,rs_a1n),(rs_a21,rs_a22,…,rs_a2n),…,(rs_am1,rs_am2,…,rs_amn)};
A所对应的第二检测数据(这里是平衡次数数据)的值为{(bl_a11,bl_a12,…,bl_a1n),(bl_a21,bl_a22,…,bl_a2n),…,(bl_am1,bl_am2,…,bl_amn)};
A所对应的第三检测数据(这里是Mimo激活比例数据)的值为{(mi_a11,mi_a12,…,mi_a1n),(mi_a21,mi_a22,…,mi_a2n),…,(mi_am1,mi_am2,…,mi_amn)}。
本实施例基于以上数据提取如下判断特征,并构建判断特征集:
(1)第一集合相对第二集合的相关系数增长率:
relation_increase_rate=(ρ(P)-ρ(A))/ρ(A)。
(2)第一集合相对第二集合的第一检测数据差值绝对值增长率;
第一集合P的第一检测数据差值绝对值为:
rs_p=abs(max(rs_p11,rs_p12,…,rs_p1n)-min(rs_p11,rs_p12,…,rs_p1n))+abs(max(rs_p21,rs_p22,…,rs_p2n)-min(rs_p21,rs_p22,…,rs_p2n))+…+abs(max(rs_pm1,rs_pm2,…,rs_pmn)-min(rs_pm1,rs_pm2,…,rs_pmn));
第二集合A的第一检测数据差值绝对值为:
rs_a=abs(max(rs_a11,rs_a12,…,rs_a1n)-min(rs_a11,rs_a12,…,rs_a1n))+abs(max(rs_a21,rs_a22,…,rs_a2n)-min(rs_a21,rs_a22,…,rs_a2n))+…+abs(max(rs_am1,rs_am2,…,rs_amn)-min(rs_am1,rs_am2,…,rs_amn));
则,第一集合P相对第二集合A的第一检测数据差值绝对值增长率为:
average_increase_rate=(rs_p–rs_a)/rs_a*100%。
(3)第一集合的相关系数的最小值:
relation_cross_min=min(ρp1,ρp2,…,ρpm)。
(4)第一集合中第一检测数据去重后样本点的最小值;
设第一集合P中某个端口的第一检测数据为rs_pij,其去重后样本点为rs_numpij,则,第一集合P去重后样本点的最小值为:
range_cross_num=min(min(rs_nump11,rs_nump12,…,,rs_nump1n),min(rs_nump21,rs_nump22,…,rs_nump2n),…,min(rs_numpm1,rs_numpm2,…,rs_numpmn));
(5)第一集合的第一检测数据波动范围的最小值;
设第一集合P中某个端口的第一检测数据为rs_pij,其波动范围rs_rangepij=max(rs_pij)-min(rs_pij);
则,第一集合P的第一检测数据波动范围的最小值为:
range_cross_min=min(min(rs_rangep11,rs_rangep12,…,,rs_rangep1n),min(rs_rangep21,rs_rangep22,…,rs_rangep2n)…,min(rs_rangepm1,rs_rangepm2,…,rs_rangepmn))。
(6)第一集合的第一检测数据波动范围的最大值;
设第一集合P中某个端口的第一检测数据为rs_pij,其波动范围rs_rangepij=max(rs_pij)-min(rs_pij);
则,第一集合P的第一检测数据波动范围的最大值为:
range_cross_max=max(max(rs_rangep11,rs_rangep12,…rs_rangep1n),max(rs_rangep21,rs_rangep22,…,rs_rangep2n),…,max(rs_rangepm1,rs_rangepm2,…,rs_rangepmn))。
(7)第一集合的第一检测数据波动范围的最大值与最小值的比值:range_cross_rate=range_cross_max/range_cross_min。
(8)第一集合相对第二集合的平衡次数差值绝对值的增长率;
第一集合P平衡次数差值绝对值:bl_p=abs(max(bl_p11,bl_p12,…,bl_p1n)-min(bl_p11,bl_p12,…,bl_p1n))+abs(max(bl_p21,bl_p22,…,bl_p2n)-min(bl_p21,bl_p22,…,bl_p2n))+…+abs(max(bl_pm1,bl_pm2,…,bl_pmn)-min(bl_pm1,bl_pm2,…,bl_pmn));
第二集合A平衡次数差值绝对值:bl_a=abs(max(bl_a11,bl_a12,…,bl_a1n)-min(bl_a11,bl_a12,…,bl_a1n))+abs(max(bl_a21,bl_a22,…,bl_a2n)-min(bl_a21,bl_a22,…,bl_a2n))+…+abs(max(bl_am1,bl_am2,…,bl_amn)-min(bl_am1,bl_am2,…,bl_amn));
则,第一集合P相对第二集合A的平衡次数差值绝对值增长率为:average_increase_rate_balance=(bl_p–bl_a)/bl_a*100%。
(9)第一集合的MIMO激活比例最小值;
设第一集合P中某个端口的MIMO的数据为mi_pij,其均值mi_meanpij=mean(mi_pij);则,第一集合P的MIMO激活比例最小值:
mimo_rate=min(min(mi_meanp11,mi_meanp12,…mi_meanp1n),min(mi_meanp21,mi_meanp22,…,mi_meanp2n)…,min(mi_meanpm1,mi_meanpm2,…mi_meanpmn))。
需了解,判断特征集可以包含上述判断特征(1)至(9)中的任一项或多项,本申请 不作过多限制。
S1006,将判断特征集输入天馈接反检测模型,天馈接反检测模型根据判断特征集,判断是否存在天馈接反情况,并输出相应的检测结果。采用天馈接反检测模型对步骤S1005得到的判断特征集进行处理,并输出相应的检测结果。
其中,检测结果中可以包括如下一项或多项:
(1)存在天馈接反的站点ID或站点名称;
(2)存在天馈接反对应的小区频点信息;
(3)存在天馈接反对应的扇区编号;
(4)存在天馈接反对应的端口号或端口号对。
这里,存在天馈接反的站点ID或站点名称、小区频点信息、扇区编号均可以在步骤S1001获取检测数据时同时获取;存在天馈接反对应的端口号或端口号对可以根据第一集合得到,具体是将第一集合与第二集合进行比较,第一集合与第二集合不相同的端口组合,即为存在天馈接反对应的端口对,根据端口对可获得对应的端口号。
示例性地,如图3所示,天馈接反检测模型输出的结果可以通过表格展示,该表格包含存在天馈接反情况的站点-载频信息,例如,制式(图中的“Product”列),站点名称(图中的“Site Name”列),接反的频段(图中的“Band”列),是否存在接反(图中的“Antenna cross result”列),接反的扇区编号以及相应的馈线端口号(图中的“Detail”列)。
图4示出了本申请另一实施例提供的天馈接反检测方法,在该实施例中,方法包括:
S2001,获取同一基站下m个扇区的馈线端口的检测数据,其中m为大于1的整数。这里,步骤S2001的具体实现方式可以参照上述步骤S1001的相关描述,此处不再赘述。
S2002,检测检测数据是否包含异常数据,在包含异常数据的情况下,对异常数据进行剔除,以提高检测准确度。其中,异常数据可以包括但不限于如下数据:持续异常的数据(全是相同值);偶尔异常的数据(一段时间都是相同值);采样点数过少的数据;波动过小(没有业务量)的数据;个别点异常的数据。
S2003,获取m个扇区所有连接情况下的端口组合。这里,步骤S2003的具体实现方式可以参照上述步骤S1002的相关描述,此处不再赘述。
S2004,根据第一检测数据,计算各个端口组合的相关系数,按照相关系数由大到小依次从所有连接情况下的端口组合中选出m个端口组合,且选出的m个端口组合互不相交。这里,步骤S2004的具体实现方式可以参照上述步骤1003的相关描述,此处不再赘述。
S2005,根据选出的m个端口组合,构建第一集合。这里,步骤S2005的具体实现方式可以参照上述步骤S1004的相关描述,此处不再赘述。
S2006,根据m个扇区的馈线端口在正确连接情况下的端口组合,构建第二集合;当第一集合与第二集合相等时,可以直接输出检测结果:天馈无接反,当第一集合与第二集合不相等时,执行下一步骤。例如:设第一集合为P,第二集合为A,当P=A(例如,P=A={(1_1,1_2),(2_1,2_2)})时,则输出检测结果:天馈无接反;如果P≠A(例如,P={(1_1,2_1),(1_2,2_2)},A={(1_1,1_2),(2_1,2_2)}),则判定为可能存在接反,继续执行下一步骤。
S2007,计算第一集合相对第二集合的相关系数的增长率,当增长率不大于预设门限值时,可以直接输出检测结果:天馈无接反;当增长率大于预设门限值时,则判定为可能存在接反,继续执行下一步骤。例如,第一集合为P和第二集合为A的相关系数分别为ρ(P)和ρ(A),预设门限值为Threshold,若(ρ(P)-ρ(A))/ρ(A)≤Threshold,则输出检测结果:天馈无接反;若(ρ(P)-ρ(A))/ρ(A)>Threshold,则认为可能存在天馈接反情况,执行下一步骤。需说明的是,这里门限值应当设置得比较宽松,例如设置在接近0的范围(例如将门限值设置在0.01至0.03的范围)内,确保不会漏掉可能接反的情况。
S2008,提取第一集合的判断特征集。这里,步骤S2008的具体实现方式可以参照上述步骤S1005的相关描述,此处不再赘述。
S2009,将判断特征集输入天馈接反检测模型,天馈接反检测模型根据判断特征集,判断是否存在天馈接反情况,并输出相应的检测结果。这里,步骤S2009的具体实现方式可以参照上述步骤S1006的相关描述,此处不再赘述。
图4所示的实施例,通过步骤S2006和/或步骤S2007,实现对天馈接反的预检测,先排除明显无接反的情况,对于可能存在接反的情况,通过后续的步骤进一步检测判断,这样既能提高检测速度,又能保证检测准确性。
需了解,有一些实施例,实现天馈接反检测方法的过程可以仅包含上述步骤S2002、步骤S2006、步骤S2007中的任一项或任两项。另外,步骤S2006和步骤S2007可以交换顺序。
根据本申请实施例的上述技术方案,通过计算各个端口组合的相关系数,按照相关系数由大到小选出m个端口组合来构建第一集合,由此得到的第一集合可以很大程度反映与当前实际接线情况相符的端口组合的集合,所以基于第一集合可以进行接反判断检测以及得到当前接反的组合。为确保天馈接反检测的准确性,提取第一集合的若干特征构建判断特征集,将判断特征集输入天馈接反检测模型进行处理,输出检测结果,相较一些情形中仅将RSSI/RTWP相关性作为接反判断依据,本申请方案的检测准确性更高。
第二方面,图5示出了本申请实施例提供的一种模型训练方法,该模型训练方法用于获取天馈接反检测模型,通过天馈接反检测模型能够对判断特征集进行处理,并获得检测结果。该模型训练方法包括:
S3100,获取训练集;
S3200,使用训练集进行训练;
S3300,输出模型文件。
如图6所示,获取训练集具体可以包括如下步骤:
S3101,获取同一基站下r个扇区的馈线端口的训练样本数据,其中,r为大于1的整数,一般性地,m代表该基站包含的扇区总数。
训练样本数据应当至少包含第一训练数据。第一训练数据可以是RTWP数据或者RSSI数据。当然,第一训练数据还可以是其他能够反映信号能量/强度的数据,本申请对此不作过多的限制。
在一些实施例中,训练样本数据还包含第二训练数据,第二训练数据具体可以是平衡次数数据。
在一些实施例中,训练样本数据还包含第三训练数据,第三训练数据具体可以是MIMO激活比例数据。
具体实现时,上述的RTWP数据、RSSI数据、平衡次数数据和MIMO激活比例数据均可以直接从基站输出至数据中心的性能数据中获取。
S3102,获取r个扇区所有连接情况下的端口组合。这里,步骤S3102的具体实现方式与上述的步骤S1002类似,可以参照上述的步骤S1002的相关描述,此处不再赘述。
S3103,根据第一训练数据,计算各个端口组合的相关系数,按照相关系数由大到小依次从所有连接情况下的端口组合中选出r个端口组合,且选出的r个端口组合互不相交。示例性地,计算各个端口组合的相关系数,具体可以对每一种可能的组合,计算组合内任意两两端口的相关系数并求和。当得到各个端口组合的相关系数后,按照相关系数由大到小依次从所有连接情况下的端口组合中选出m个端口组合,且选出的m个端口组合的元素互不相交,即:按照相关系数由大到小选出m个端口组合的过程中,每选出一个端口组合,当前选出的端口组合包含的元素均不能包含在之前选出的端口组合中。本步骤S3103的具体实现方式与上述的步骤S1003类似,可以参照上述的步骤S1003的相关描述,此处不再赘述。
S3104,根据选出的r个端口组合,构建第三集合。步骤S3103中选出的端口组合是按照相关系数由大到小依次选出的,而端口组合的相关性高说明组合中的端口很可能接在同一天线设备上,因此第三集合很大程度反映与当前实际接线情况相符的端口组合的集合,后续可以根据第三集合的特征训练检测天馈有无接反。
S3105,提取第三集合的训练特征集。其中,训练特征集可以包括以下任一项或多项:
(1)第三集合相对第四集合的相关系数增长率;
(2)第三集合相对第四集合的第一检测数据差值绝对值增长率;
(3)第三集合的相关系数的最小值;
(4)第一集合中第一检测数据去重后样本点的最小值;
(5)第三集合的第一检测数据波动范围的最小值;
(6)第三集合的第一检测数据波动范围的最大值;
(7)第三集合的第一检测数据波动范围的最大值与最小值的比值;
(8)第三集合相对第四集合的平衡次数差值绝对值的增长率;
(9)第三集合的MIMO激活比例最小值。
以上各项特征的具体获取过程与上述的步骤S1005类似,可以参照上述的步骤S1005的相关描述,此处不再赘述。
S3106,对应于训练特征集,对是否存在接反的结果进行标注,得到训练目标。示例性地,可以对训练特征集进行编号,并根据获得的第一训练数据进行画图(例如,RSSI/RTWP图形),根据RSSI/RTWP图形,对结果进行标注,例如,可以用“0”表示未接反,用“1”表示接反。
S3107,将训练特征集和训练目标进行合并,加入训练集中。利用获得的训练集,对天馈接反检测模型进行训练。
如图7所示,在另一实施例中,获取训练集的过程还包括可选的步骤S3108。在步骤 S3108中,检测训练样本数据是否包含异常数据,在包含异常数据的情况下,对异常数据进行剔除,以提高训练样本数据的精度,避免异常值对于模型准确率的不利影响。其中,异常数据可以包括但不限于如下数据:持续异常的数据(全是相同值);偶尔异常的数据(一段时间都是相同值);采样点数过少的数据;波动过小(没有业务量)的数据;个别点异常的数据。
在另一实施例中,获取训练集的过程还包括可选的步骤S3109。在步骤S3109中,根据m个扇区的馈线端口在正确连接情况下的端口组合,构建第四集合;当第三集合与第四集合相等时,剔除训练样本数据,当第三集合与第四集合不相等时,执行下一步骤。例如:设第三集合为P,第四集合为A,当P=A(例如,P=A={(1_1,1_2),(2_1,2_2)})时,则输出检测结果:天馈无接反;如果P≠A(例如,P={(1_1,2_1),(1_2,2_2)},A={(1_1,1_2),(2_1,2_2)}),则执行下一步骤。
S3110,计算第三集合相对第四集合的相关系数的增长率,当增长率不大于预设门限值时,剔除训练样本数据;当增长率大于预设门限值时,执行下一步骤。例如,第三集合为P和第四集合为A的相关系数分别为ρ(P)和ρ(A),预设门限值为Threshold,若(ρ(P)-ρ(A))/ρ(A)≤Threshold,则输出检测结果:天馈无接反;若(ρ(P)-ρ(A))/ρ(A)>Threshold,则认为可能存在天馈接反情况,执行下一步骤。需说明的是,这里门限值应当设置得比较宽松,例如设置在接近0的范围内,确保不会漏掉可能接反的情况。
图6所示的实施例,通过步骤S3109和/或步骤S3110,对天馈是否接反进行预判,对可明显得出无接反情况的训练样本数据进行剔除,对于可能存在接反的情况,进一步对第三集合进行训练特征提取。这样,一方面可以减少训练集的标注量,另一方面有利于获得训练价值度更高的训练集,进而使得天馈接反检测模型输出的检测结果也更为准确。
需了解,在一些实施例中,获取训练集的过程可以仅包含上述步骤S3108、步骤S3109、步骤S3110中的任一项或任两项。另外,步骤S3109和步骤S3110可以交换顺序。
在步骤S3200中,使用训练集进行训练,具体可以是通过机器学习算法进行训练。机器学习算法可以是逻辑回归、随机森林、支持向量机、深度学习中的任一项,本申请实施例对此不作限制。
如图8所示,实现模型训练方法时,还可以包括可选的步骤S3400,在训练特征集包含多项特征的情况下,调整训练特征集中各项特征的权重,以将准确率和召回率控制在可接受的范围内,并最终输出模型文件。
根据本申请实施例的方案,通过计算各个端口组合的相关系数,按照相关系数由大到小选出r个端口组合来构建第三集合,由此得到的第三集合可以很大程度反映与当前实际接线情况相符的端口组合的集合,因此根据第三集合提取的训练特征集也较能表达天馈接反情况,进而基于上述训练特征集训练得到的天馈接反检测模型输出的结果具有更高的准确性。
第三方面,图9示出了本申请实施例提供了本申请实施例提供的电子装置200。电子装置200包括:存储器220、处理器210及存储在存储器220上并可在处理器210上运行的计算机程序,计算机程序运行时用于执行上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法。
处理器210和存储器220可以通过总线或者其他方式连接。
存储器220作为一种非暂态计算机可读存储介质,可用于存储非暂态软件程序以及非暂态性计算机可执行程序,如本申请实施例中第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法。处理器210通过运行存储在存储器220中的非暂态软件程序以及指令,从而实现上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法。
存储器220可以包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需要的应用程序;存储数据区可存储执行上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法。此外,存储器220可以包括高速随机存取存储器,还可以包括非暂态存储器,例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施方式中,存储器220可选包括相对于处理器210远程设置的存储器,这些远程存储器可以通过网络连接至该终端。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
实现上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法所需的非暂态软件程序以及指令存储在存储器220中,当被一个或者多个处理器210执行时,执行上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法,例如,执行图2中描述的方法步骤S1001至S1006,图4中描述的方法步骤S2001至S2009,图5中描述的方法步骤S3100至S3300,图6中描述的方法步骤S3101至S3107,图7中描述的方法步骤S3101至S3110,图8中描述的方法步骤S3100至S3300。
本申请实施例还提供了计算机可读存储介质,存储有计算机可执行指令,计算机可执行指令用于执行上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法。
在一实施例中,该计算机可读存储介质存储有计算机可执行指令,该计算机可执行指令被一个或多个控制处理器210执行,例如,被上述电子装置200中的一个处理器210执行,可使得上述一个或多个处理器210执行上述第一方面描述的任一项天馈接反检测方法或者第二方面描述的任一项模型训练方法,例如,执行图2中描述的方法步骤S1001至S1006,图4中描述的方法步骤S2001至S2009,图5中描述的方法步骤S3100至S3300,图6中描述的方法步骤S3101至S3107,图7中描述的方法步骤S3101至S3110,图8中描述的方法步骤S3100至S3300。
本申请实施例包括:获取同一基站下m个扇区的馈线端口的检测数据,其中,m为大于1的整数,所述检测数据包含第一检测数据;获取所述m个扇区所有连接情况下的端口组合;根据所述第一检测数据,计算各个所述端口组合的相关系数,按照所述相关系数由大到小依次从所述所有连接情况下的端口组合中选出m个端口组合,且选出的所述m个端口组合互不相交;根据选出的所述m个端口组合,构建第一集合;提取所述第一集合的判断特征集;将所述判断特征集输入天馈接反检测模型,所述天馈接反检测模型根据所述判断特征集,判断是否存在天馈接反情况,并输出相应的检测结果。本申请实施例的技术方案,天馈接反检测准确率较高,误判、漏判率极低。
以上所描述的装置实施例仅仅是示意性的,其中作为分离部件说明的单元可以是或者 也可以不是物理上分开的,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
本领域普通技术人员可以理解,上文中所公开方法中的全部或某些步骤、系统可以被实施为软件、固件、硬件及其适当的组合。某些物理组件或所有物理组件可以被实施为由处理器,如中央处理器、数字信号处理器或微处理器执行的软件,或者被实施为硬件,或者被实施为集成电路,如专用集成电路。这样的软件可以分布在计算机可读介质上,计算机可读介质可以包括计算机存储介质(或非暂时性介质)和通信介质(或暂时性介质)。如本领域普通技术人员公知的,术语计算机存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储装置、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。此外,本领域普通技术人员公知的是,通信介质通常包含计算机可读指令、数据结构、程序模块或者诸如载波或其他传输机制之类的调制数据信号中的其他数据,并且可包括任何信息递送介质。
以上是对本申请的较佳实施进行了具体说明,但本申请并不局限于上述实施方式,熟悉本领域的技术人员在不违背本申请精神的共享条件下还可作出种种的等同变形或替换,这些等同的变形或替换均包含在本申请权利要求所限定的范围内。

Claims (21)

  1. 天馈接反检测方法,包括:
    获取同一基站下m个扇区的馈线端口的检测数据,其中,m为大于1的整数,所述检测数据包含第一检测数据;
    获取所述m个扇区所有连接情况下的端口组合;
    根据所述第一检测数据,计算各个所述端口组合的相关系数,按照所述相关系数由大到小依次从所述所有连接情况下的端口组合中选出m个端口组合,且选出的所述m个端口组合互不相交;
    根据选出的所述m个端口组合,构建第一集合;
    提取所述第一集合的判断特征集;
    将所述判断特征集输入天馈接反检测模型,所述天馈接反检测模型根据所述判断特征集,判断是否存在天馈接反情况,并输出相应的检测结果。
  2. 根据权利要求1所述的天馈接反检测方法,还包括:根据所述m个扇区的馈线端口在正确连接情况下的端口组合,构建第二集合;在所述第一集合与所述第二集合不相等的情况下,执行提取所述第一集合的判断特征集。
  3. 根据权利要求1所述的天馈接反检测方法,还包括:根据所述m个扇区的馈线端口在正确连接情况下的端口组合,构建第二集合;计算所述第一集合相对所述第二集合的相关系数的增长率,在所述增长率大于预设门限值的情况下,执行提取所述第一集合的判断特征集。
  4. 根据权利要求1所述的天馈接反检测方法,其中,所述第一检测数据为RTWP数据或者RSSI数据。
  5. 根据权利要求1或4所述的天馈接反检测方法,其中,所述提取所述第一集合的判断特征集,包括:根据所述第一检测数据,提取以下任一项或多项判断特征加入至判断特征集:
    所述第一集合相对第二集合的相关系数增长率,其中,所述第二集合是同一基站下m个扇区的馈线端口在正确连接情况下的端口组合的集合;
    所述第一集合相对第二集合的第一检测数据差值绝对值增长率,其中,所述第二集合是同一基站下m个扇区的馈线端口在正确连接情况下的端口组合的集合;
    所述第一集合的相关系数的最小值;
    所述第一集合中第一检测数据去重后样本点的最小值;
    所述第一集合的第一检测数据波动范围的最小值;
    所述第一集合的第一检测数据波动范围的最大值;
    所述第一集合的第一检测数据波动范围的最大值与最小值的比值。
  6. 根据权利要求1或4所述的天馈接反检测方法,其中,所述检测数据还包含第二检测数据,所述第二检测数据为平衡次数数据;
    对应地,所述提取所述第一集合的判断特征集,包括:根据所述第二检测数据,提取以下判断特征加入至判断特征集:
    所述第一集合相对第二集合的平衡次数差值绝对值增长率,其中,所述第二集合是同 一基站下m个扇区的馈线端口在正确连接情况下的端口组合的集合。
  7. 根据权利要求1或4所述的天馈接反检测方法,其中,所述检测数据还包含第三检测数据,所述第三检测数据为MIMO激活比例数据;
    对应地,所述提取所述第一集合的判断特征集,包括:根据所述第三检测数据,提取以下判断特征加入至判断特征集:
    所述第一集合的MIMO激活比例最小值。
  8. 根据权利要求1所述的天馈接反检测方法,其中,当获取同一基站下m个扇区的馈线端口的检测数据,还检测所述检测数据是否包含异常数据,在包含异常数据的情况下,对所述异常数据进行剔除。
  9. 根据权利要求1所述的天馈接反检测方法,其中,所述检测结果包括以下任一项或多项:
    存在天馈接反的站点ID或站点名称;
    存在天馈接反对应的小区频点信息;
    存在天馈接反对应的扇区编号;
    存在天馈接反对应的端口号或端口号对。
  10. 根据权利要求1所述的天馈接反检测方法,其中,所述天馈接反检测模型通过训练集进行训练得到,所述训练集包括训练特征集和训练目标,其中,所述训练特征集来自同一基站下扇区的馈线端口的训练样本数据;训练目标为对应于所述训练特征集,对是否存在接反的结果进行的标注。
  11. 一种模型训练方法,用于获取天馈接反检测模型,所述方法包括:
    获取训练集;使用所述训练集进行训练;输出模型文件;
    其中,所述获取训练集具体包括:
    获取同一基站下r个扇区的馈线端口的训练样本数据,其中,r为大于1的整数,所述训练样本数据包含第一训练数据;
    获取所述r个扇区所有连接情况下的端口组合;
    根据所述第一训练数据,计算各个所述端口组合的相关系数,按照所述相关系数由大到小依次从所述所有连接情况下的端口组合中选出r个端口组合,且选出的所述r个端口组合互不相交;
    根据选出的所述r个端口组合,构建第三集合;
    提取所述第三集合的训练特征集;
    对应于所述训练特征集,对是否存在接反的结果进行标注,得到训练目标;
    将所述训练特征集和所述训练目标进行合并,加入训练集中。
  12. 根据权利要求11所述的模型训练方法,还包括:根据同一基站下r个扇区的馈线端口在正确连接情况下的端口组合,构建第四集合;在所述第三集合与所述第四集合不相等的情况下,执行提取所述第三集合的训练特征集。
  13. 根据权利要求11或12所述的模型训练方法,还包括:根据同一基站下n个扇区的馈线端口在正确连接情况下的端口组合,构建第四集合;计算所述第三集合相对第四集合的相关系数的增长率,在所述增长率大于预设门限值的情况下,执行提取所述第三集合 的训练特征集。
  14. 根据权利要求11所述的模型训练方法,其中,所述第一训练数据为RTWP数据或者RSSI数据。
  15. 根据权利要求11或14所述的模型训练方法,其中,所述提取所述第三集合的训练特征集,包括:根据所述第一训练数据,提取以下任一项或多项训练特征加入至训练特征集:
    所述第三集合相对第四集合的相关系数增长率,其中,所述第四集合是同一基站下m个扇区的馈线端口在正确连接情况下的端口组合的集合;
    所述第三集合相对第四集合的第一训练数据差值绝对值增长率,其中,所述第四集合是同一基站下m个扇区的馈线端口在正确连接情况下的端口组合的集合;
    所述第三集合的相关系数的最小值;
    所述第三集合中第一训练数据去重后样本点的最小值;
    所述第三集合的第一训练数据波动范围的最小值;
    所述第三集合的第一训练数据波动范围的最大值;
    所述第三集合的第一训练数据波动范围的最大值与最小值的比值。
  16. 根据权利要求11所述的模型训练方法,其中,所述训练样本数据还包含第二训练数据,所述第二训练数据为平衡次数数据;
    对应地,所述提取所述第三集合的训练特征集,包括:根据所述第二训练数据,提取以下训练特征加入至训练特征集:
    所述第三集合相对第四集合的平衡次数差值绝对值的增长率,其中,所述第四集合是同一基站下n个扇区的馈线端口在正确连接情况下的端口组合的集合。
  17. 根据权利要求11所述的模型训练方法,其中,所述训练样本数据还包含第三训练数据,所述第三训练数据为MIMO激活比例数据;
    对应地,所述提取所述第三集合的训练特征集,包括:根据所述第三训练数据,提取以下训练特征加入至训练特征集:
    所述第三集合的MIMO激活比例最小值。
  18. 根据权利要求11所述的模型训练方法,其中,当使用所述训练集进行训练,在所述训练特征集包含多项特征的情况下,还调整所述训练特征集中各项特征的权重。
  19. 根据权利要求11所述的模型训练方法,其中,当获取同一基站下m个扇区的天线的训练样本数据,还检测所述训练样本数据是否包含异常数据,在包含异常数据的情况下,对所述异常数据进行剔除。
  20. 电子装置,包括:存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序,其中,所述计算机程序运行时执行:
    如权利要求1至10任一项所述的天馈接反检测方法;或者
    如权利要求11至19任一项所述的模型训练方法。
  21. 计算机可读存储介质,存储有计算机可执行指令,其中,所述计算机可执行指令用于执行:
    权利要求1至10中任一项所述的天馈接反检测方法;或者
    权利要求11至19中任一项所述的模型训练方法。
PCT/CN2020/109225 2019-09-23 2020-08-14 天馈接反检测方法和装置 WO2021057327A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910900620.3 2019-09-23
CN201910900620.3A CN112543069B (zh) 2019-09-23 2019-09-23 天馈接反检测方法和装置

Publications (1)

Publication Number Publication Date
WO2021057327A1 true WO2021057327A1 (zh) 2021-04-01

Family

ID=75012848

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/109225 WO2021057327A1 (zh) 2019-09-23 2020-08-14 天馈接反检测方法和装置

Country Status (2)

Country Link
CN (1) CN112543069B (zh)
WO (1) WO2021057327A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101505489A (zh) * 2008-12-19 2009-08-12 华为技术有限公司 一种检测天馈设备接反的小区的方法和装置
US20120087450A1 (en) * 2009-06-08 2012-04-12 Telefonaktiebolaget L M Ericsson (Publ) Wireless communication node connections
CN103096359A (zh) * 2011-11-04 2013-05-08 华为技术有限公司 诊断馈线接错的方法及装置
CN106792741A (zh) * 2016-11-28 2017-05-31 北京市天元网络技术股份有限公司 一种判断基站扇区间天馈接反的方法和系统
CN109257125A (zh) * 2018-11-22 2019-01-22 中国联合网络通信集团有限公司 一种基站天馈接反的检测方法和装置

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104639263B (zh) * 2013-11-13 2017-12-05 中国电信股份有限公司 检测基站天馈安装问题的方法与装置
JP2016092709A (ja) * 2014-11-10 2016-05-23 株式会社日立製作所 分散アンテナシステムのアンテナ系統異常検知方法、システム、及びそれに用いる中継装置
CN105721072B (zh) * 2014-12-04 2021-01-26 中兴通讯股份有限公司 一种判断天线故障的方法、装置及终端
CN106856417B (zh) * 2015-12-09 2020-06-02 中国电信股份有限公司 用于检测天馈安装问题的方法和装置
CN109583904B (zh) * 2018-11-30 2023-04-07 深圳市腾讯计算机系统有限公司 异常操作检测模型的训练方法、异常操作检测方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101505489A (zh) * 2008-12-19 2009-08-12 华为技术有限公司 一种检测天馈设备接反的小区的方法和装置
US20120087450A1 (en) * 2009-06-08 2012-04-12 Telefonaktiebolaget L M Ericsson (Publ) Wireless communication node connections
CN103096359A (zh) * 2011-11-04 2013-05-08 华为技术有限公司 诊断馈线接错的方法及装置
CN106792741A (zh) * 2016-11-28 2017-05-31 北京市天元网络技术股份有限公司 一种判断基站扇区间天馈接反的方法和系统
CN109257125A (zh) * 2018-11-22 2019-01-22 中国联合网络通信集团有限公司 一种基站天馈接反的检测方法和装置

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
CHEN XIUMIN: "Algorithm for Detecting Reversed Connection of Antenna Feeds in CDMA Base Stations", MOBILE COMMUNICATIONS, vol. 35, no. 8, 18 December 2011 (2011-12-18), pages 18 - 22, XP055794221 *
TAN YI, ZHANG CHUN, WANG DONG-JU: "Application of the correlation analysis in the CDMAwireless system feeder inspection", TELECOM ENGINEERING TECHNICS AND STANDARDIZATION, no. 2,, 15 February 2013 (2013-02-15), pages 42 - 44, XP055794224, ISSN: 1008-5599, DOI: 10.13992/j.cnki.tetas.2013.02.012 *

Also Published As

Publication number Publication date
CN112543069A (zh) 2021-03-23
CN112543069B (zh) 2023-01-06

Similar Documents

Publication Publication Date Title
CN103916820B (zh) 基于接入点稳定度的无线室内定位方法
US9913092B2 (en) Mitigating signal noise for fingerprint-based indoor localization
CN108228722B (zh) 破碎化区域采样点的地理空间分布均匀度检测方法
JP6681617B2 (ja) 波源位置推定装置、コンピュータに実行させるためのプログラム、およびプログラムを記録したコンピュータ読み取り可能な記録媒体
CN105430732A (zh) 一种wifi发射功率调节方法、终端以及系统
CN104349456B (zh) WiFi定位方法和WiFi定位平台
CN111654870A (zh) 调整小区覆盖区的控制方法、装置、设备及存储介质
CN107395301A (zh) 一种基于k均值算法的频谱感知方法
CN109257125B (zh) 一种基站天馈接反的检测方法和装置
CN106407052A (zh) 一种检测磁盘的方法及装置
WO2013136128A1 (en) Generating radio channel models parameter values
CN104427505A (zh) 一种小区场景划分的方法及装置
CN110362492A (zh) 人工智能算法测试方法、装置、服务器、终端及存储介质
CN107171981B (zh) 通道校正方法及装置
WO2021057327A1 (zh) 天馈接反检测方法和装置
CN109788432B (zh) 室内定位方法、装置、设备及存储介质
CN101893707A (zh) 非视距传播识别方法、装置与基站
CN112448775B (zh) 蓝牙辐射性能测试方法、装置及存储介质
CN101902264B (zh) 802.11无线通信中智能天线的控制方法
CN104270778A (zh) 天馈系统的下行检测方法及装置
CN113015180A (zh) 网络参数更新方法、装置及存储介质、电子设备
CN103138856A (zh) 一种检测干扰的方法及装置
CN113543006B (zh) 耳机测试方法、装置、电子设备及介质
CN105184198B (zh) 检测保护方法及移动终端
CN106856417B (zh) 用于检测天馈安装问题的方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20867132

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20867132

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 20/02/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20867132

Country of ref document: EP

Kind code of ref document: A1