Detection Method, Device, Electronic Device and Storage Medium for Faulty Base Station

Abstract

The present invention provides a detection method, apparatus, electronic device and storage medium for a faulty base station. Among them, the detection method for a faulty base station includes: obtaining operation data on the base station side and user data on the user side; performing time series decomposition and outlier detection on the operation data to obtain a suspected faulty base station situation; based on the suspected faulty base station situation, calculating the difference in signal quality distribution between the current user data and the historical user data within a square area centered on the suspected faulty base station; and determining the faulty base station based on the difference in signal quality distribution. By the above method, the present invention considers spatio-temporal factors and jointly processes the data on the base station side and the user side instead of single data, executes a fast detection algorithm on the small-scale data on the base station side, gives out abnormal information to screen the user side data, so that only a small amount of user side data needs to be deeply calculated, greatly reducing the average running time and reducing the consumption of computing resources.

Description

Technical Field

[0001] The present invention relates to the technical field of interruption detection and location, and particularly to a detection method, device, electronic device and storage medium for a faulty base station. Background Art

[0002] With the continuous development of wireless communication technology, compared with the 4G network, the next-generation wireless communication technology represented by 5G will also significantly improve its wireless coverage performance, transmission delay, system security and user experience. In the 5G network, due to the massive data traffic requirements of heterogeneous dynamic users and frequent communication overheads, it is extremely easy to lead to a reduction in user service quality and even QoE. The Cloud Radio Access Networks (C-RAN) is considered to be an architecture that can solve this problem in 5G. C-RAN virtualizes the Baseband Processing Unit (BBU) into a cloud resource pool. At the same time, the BBU and the Remote Radio Head (RRH) are connected through a high-speed fronthaul link, which can provide flexible distributed system capacity and reduce the overall network energy consumption, etc.

[0003] Meanwhile, due to the relatively high frequency band of 5G technology, the effective service distance of the base station has decreased significantly compared with the previous generation of networks. To meet the coverage requirements, the density of the base station and the complexity of the cellular network have inevitably increased, which brings a large amount of capital expenditure and operating expenditure (CAPEX and OPEX). To solve this problem, the solution proposed by researchers and the standardization organization 3GPP is to deploy a Self-Organizing Network (SON), which introduces self-organizing capabilities into the network to achieve autonomous operation of the network, thereby improving the network management efficiency and reducing labor costs. The self-organizing network is mainly divided into three key areas: autonomous configuration, autonomous optimization, and autonomous recovery. Due to the large number and density of base stations in 5G, the frequency of their failures is also increasing. Therefore, the autonomous recovery function of the network is very important. The autonomous recovery field can be subdivided into interruption detection, fault diagnosis, and interruption compensation.

[0004] A large number of studies have been conducted on the Call Drop Detection (COD) and Call Drop Compensation (COC) technologies for 4G LTE, and a large number of methods based on machine learning or intelligent algorithms have been applied to solve various SON-related use cases. However, most of the interruption detections are performed by collecting data from all user terminals, which results in excessive computational resource overhead when there are a large number of users and it is difficult to meet the rapid and accurate detection requirements of a large number of distributed RRH nodes in the future 5G C-RAN. Therefore, how to achieve active, fast and accurate diagnosis of interrupted RRH is one of the important problems that need to be solved. Summary of the Invention

[0005] The present invention provides a method, apparatus, electronic device and storage medium for detecting a faulty base station, aiming to solve the defect of excessive computational resource overhead in interruption detection in the prior art and implement a detection algorithm that is efficient, fast and has a small amount of computation.

[0006] The present invention provides a method for detecting a faulty base station, including: obtaining the operation data on the base station side and the user data on the user side; performing time series decomposition and outlier detection on the operation data to obtain a suspected faulty base station situation; wherein, the suspected faulty base station situation includes the base station serial number of the suspected faulty base station; based on the suspected faulty base station situation, calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station; and determining the faulty base station based on the signal quality distribution difference.

[0007] According to a method for detecting a faulty base station provided by the present invention, performing time series decomposition on the operation data includes: collecting the operation data on the base station side at equal intervals per hour; wherein, the operation data includes base station metrics, and the base station metrics include the number of user connections, power and total throughput; connecting and sorting the operation data within a preset number of days in chronological order to form a data set S ijk ; where S ijk represents the value of the j-th metric of the i-th base station at the k-th time point; performing STL time series decomposition on the time series data S ij of each base station and each metric respectively with a one-day period to obtain a periodic term Seasonal ij , a trend term Trend ij and a remainder term Rem ij ; obtaining a processed set S' ijk based on the data set S ij and the periodic term Seasonal ijk , where S' ijk = S ijk - Seasonal ijk .

[0008] According to a method for detecting a faulty base station provided by the present invention, performing time series decomposition and outlier detection on the operation data to obtain a suspected faulty base station situation includes: performing GESD outlier detection on the time series data S' ij ; if the j-th metric in the processed set S' ijk at the current time point k belongs to an outlier, obtaining the corresponding i-th base station as a suspected faulty base station.

[0009] A detection method for a faulty base station provided by the present invention, before calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station based on the situation of the suspected faulty base station, includes: determining a detection area centered on the suspected faulty base station and dividing the detection area into a number of grids; obtaining user data on the user side within the detection area and classifying the user data according to the grids based on the location information; wherein, the user data includes location information and signal metrics, and the signal metrics include the current cell signal, the maximum neighbor cell signal, and the signal-to-noise ratio of the current cell signal.

[0010] A detection method for a faulty base station provided by the present invention, calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station, includes: comparing the signal metric distribution at the current time point within each grid with the signal metric distribution in the same grid during the same period of the historical data to obtain the signal quality distribution difference of the signal metrics within the corresponding grid.

[0011] A detection method for a faulty base station provided by the present invention, comparing the signal metric distribution at the current time point within each grid with the signal metric distribution in the same grid during the same period of the historical data to obtain the signal quality distribution difference of the signal metrics within the corresponding grid, and determining the faulty base station based on the signal quality distribution difference, includes: calculating the Wasserstein distance between the signal metric distribution at the current time point within each grid and the signal metric distribution in the same grid during the same period of the historical data; wherein, the Wasserstein distance represents the distribution difference degree of the metrics within the corresponding grid area; converting the detection area into a pixel map of three layers through a mapping function f according to the distribution difference degree, where each layer represents a two-dimensional plane distribution of the signal metric difference degree; inputting the pixel map into a pre-trained convolutional neural network to obtain the fault probability of the base station corresponding to the pixel map; and determining the faulty base station based on the fault probability of the base station.

[0012] The present invention also provides a detection device for a faulty base station, including: a data acquisition module for obtaining the operation data on the base station side and the user data on the user side; a suspected faulty base station module for performing time series decomposition and outlier detection on the operation data to obtain the situation of the suspected faulty base station; wherein, the situation of the suspected faulty base station includes the base station serial number of the suspected faulty base station; a distribution difference module for calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station based on the situation of the suspected faulty base station; and a faulty base station module for determining the faulty base station based on the signal quality distribution difference.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the detection method of the faulty base station as described in any one of the above is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the detection method of the faulty base station as described in any one of the above is implemented.

[0015] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the detection method of the faulty base station as described in any one of the above is implemented.

[0016] The present invention provides a detection method, device, electronic device and storage medium for a faulty base station. By performing time series decomposition and outlier detection on the operation data of the base station side, the situation of suspected faulty base stations is obtained. According to the situation of suspected faulty base stations, the signal quality distribution difference between the current user data and the historical user data of the corresponding part of the user side is calculated, and finally the faulty base station is determined. Through the above method, the present invention combines the data on the base station side and the user side and jointly analyzes them with spatio-temporal information, without calculating all the user side data, greatly reducing the system's computational resource overhead and time overhead, and reducing the average running time. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to these drawings without creative efforts.

[0018] Figure 1 It is a flowchart of an embodiment of the detection method of the faulty base station of the present invention;

[0019] Figure 2 It is a flowchart of an embodiment of the base station side data processing in the detection method of the faulty base station of the present invention;

[0020] Figure 3 It is a flowchart of an embodiment of the user side data processing in the detection method of the faulty base station of the present invention;

[0021] Figure 4 It is a schematic structural diagram of an embodiment of the detection device of the faulty base station of the present invention;

[0022] Figure 5 It is a schematic structural diagram of an embodiment of the electronic device of the present invention. Detailed Embodiments

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] The present invention provides a detection method for a faulty base station, which can be applied to the interruption detection part in the field of autonomous recovery in a self-organizing network. Please refer to Figure 1 , Figure 1 which is a schematic flowchart of an embodiment of the detection method for a faulty base station of the present invention. In this embodiment, the detection method for a faulty base station may include steps S110 to S140, and the specific steps are as follows:

[0025] S110: Obtain the operation data on the base station side and the user data on the user side.

[0026] This embodiment is described by taking a 5G C-RAN architecture wireless cloud network cell as an example. According to the C-RAN architecture, it can be assumed that the detection device for the faulty base station is located in the SON (self-organizing network) entity in the cloud. Each base station collects the information of the users it serves, summarizes and sends it to the SON entity, and the detection device for the faulty base station collects and processes it.

[0027] The entire detection process is divided into two parts: preliminary detection on the base station side and in-depth detection on the user side. The base station side detection needs to process a large amount of operation data, and the primary goal is speed; the in-depth detection on the user side only needs to process part of the user data, and the primary goal is accuracy. The former decomposes the large amount of collected real-time data in time series to find the situation of suspected faulty base stations. The latter calculates the difference in the quality distribution of the data signals on the user side according to the situation of the suspected faulty base stations indicated by the former, and based on this, conducts in-depth detection and fault determination of the faulty base stations to determine the faulty base stations.

[0028] Optionally, obtaining the operation data on the base station side specifically includes:

[0029] Collect the operation data on the base station side at equal intervals every hour. Among them, the operation data includes base station metrics, and the base station metrics include the number of user connections, power, and total throughput.

[0030] Collect the user data of the user equipment near the suspected faulty base station at equal intervals every hour. The collected user data can be divided into current user data and historical user data, and the historical user data can be used as a sample for the difference in the quality distribution of signals.

[0031] S120: Perform time series decomposition and outlier detection on the operation data to obtain the situation of suspected faulty base stations.

[0032] Optionally, the operation data within a preset number of days are connected and sorted in chronological order respectively to form a data set S ijk ; where S ijk represents the value of the j-th indicator of the i-th base station at the k-th time point; the time-series data S ij of each indicator of each base station is subjected to STL time-series decomposition with a one-day period to obtain a periodic term Seasonal ij , a trend term Trend ij and a remainder term Rem ij ; based on the data set S ijk and the periodic term Seasonal ij a processed set S' ijk is obtained, where S' ijk =S ijk -Seasonal ijk .

[0033] After that, GESD outlier detection is performed on the time-series data S' ij ; if the j-th indicator in the processed set S' ijk at the current time point k belongs to an outlier, the corresponding i-th base station is obtained as a suspected faulty base station. Among them, the abnormal data is used to indicate the situation of the suspected faulty base station.

[0034] S130: Based on the situation of the suspected faulty base station, calculate the difference in the signal quality distribution between the current user data and the historical user data within a square area centered on the suspected faulty base station.

[0035] After the preliminary detection on the base station side, the situation of the suspected faulty base station is initially determined. Next, further fault detection and confirmation are required based on the user-side data.

[0036] Optionally, a detection area is determined with the suspected faulty base station as the center, and the detection area is divided into several grids; the user data on the user side within the detection area is obtained, and the user data is classified according to the grid according to the location information; among them, the user data includes location information and signal indicators, and the signal indicators include the current cell signal, the maximum neighbor cell signal, and the signal-to-noise ratio of the current cell signal. The signal indicator distribution at the current time point in each grid is compared with the signal indicator distribution in the same grid during the same period of the historical data to obtain the difference in the signal quality distribution of the signal indicators in the corresponding grid.

[0037] It should be noted that by adopting grid processing, a large square area around the base station is gridded into several small square areas, and user data in each small square area is statistically analyzed. At the same time, the area of the grid should not be too large or too small. If the area of each grid is too large, the corresponding position range of the statistically analyzed data will be too wide, and the calculated distribution data will tend to be unified. If the area of the grid is too small, the user data in each grid will be too few, resulting in inaccurate distribution statistics and poor anti-interference ability. Therefore, whether the grid is too large or too small will lead to a decrease in the accuracy rate of the subsequent detection process.

[0038] It should be noted that in this embodiment, the user data in the area near the suspected faulty base station is collected, rather than the user data served by the suspected faulty base station. The reasons are as follows: User devices in an area do not necessarily belong to one base station. Most user devices are served by the base station that is closer, and there are also many served by nearby base stations. After the base station fails, it is possible that the number of users served by this base station drops significantly, or there are no served users at all. In this area, users either switch to other base stations or have no service at all. This embodiment emphasizes the user data in this area, and the users corresponding to these data are served by different base stations, and only a part of them may be the users served by the faulty base station (it is also possible that the faulty base station has no served users at all). The reason for calculating the distribution difference in this embodiment is to take into account the users served by other base stations.

[0039] S140: Determine the faulty base station based on the distribution difference of signal quality.

[0040] In some embodiments, the step of determining the faulty base station based on the distribution difference of signal quality specifically includes:

[0041] Calculate the Wasserstein distance between the signal metric distribution at the current time point in each grid and the signal metric distribution in the same grid during the same period of historical data; where the Wasserstein distance represents the distribution difference degree of the metric in the corresponding grid area; transform the detection area into a pixel map of three layers through the mapping function f according to the distribution difference degree, where each layer represents a two-dimensional plane distribution of the signal metric difference degree; input the pixel map into a pre-trained convolutional neural network to obtain the failure probability of the base station corresponding to the pixel map; determine the faulty base station based on the failure probability of the base station. Optionally, when the failure probability of the base station is greater than 0.5, it is determined as a faulty base station.

[0042] In this embodiment, a convolutional neural network is used to calculate the base station failure probability information according to the pixel map converted from the distribution difference information, which is more valuable in practical applications compared with simple judgment information.

[0043] After the above steps, the Wasserstein distances of the three metrics in each grid within the detection area corresponding to all suspected faulty base stations are obtained. For each abnormally detected base station pointed out by the preliminary detection, a matrix with a shape of (3×ND×ND) is used to record the distribution of the distances.

[0044] If the base station is not abnormal, each value in the matrix should be small, indicating that there is no obvious difference in the distribution. If the base station has an interruption fault, then some values in the matrix will increase in a certain pattern because the distribution is abnormal.

[0045] In order to convert the difference situation of the distribution into the fault situation of the base station, in some embodiments of the present invention, a machine learning method can also be adopted to convert the detection area into a pixel map, and each layer of the pixel map can correspond to a signal metric. On this basis, the probability calculation of the base station fault can be realized by means of image recognition and it can be determined whether the base station is abnormal.

[0046] Since the value range of the Wasserstein distance is (0, +∞), while the value range of each layer in the pixel map is [0, 255], it is necessary to map the Wasserstein distance to convert the distance into a value in the range of [0, 255] before subsequent work can be carried out. This embodiment provides a method for detecting faulty base stations. By performing time series decomposition and outlier detection on the operation data on the base station side, the situation of suspected faulty base stations is obtained. According to the situation of suspected faulty base stations, the signal quality distribution difference between the current user data and the historical user data is calculated for the corresponding part of the user side, and finally the faulty base stations are determined. In the above manner, this embodiment takes time and space factors as important indicators, considers the periodicity of factors such as personnel distribution, and uses the coordinate information as the sole processing basis, with good results; jointly processes the base station side data and the user side data instead of single data, executes a fast detection algorithm for the small-scale data on the base station side, gives the preliminary abnormal base station information to screen the user side data, so that only a small amount of user side data needs to be deeply calculated, greatly reducing the average running time and reducing the consumption of computing resources. Making the algorithm feasible in reality.

[0047] Please refer to Figure 2 , Figure 2 which is a schematic flowchart of an embodiment of the base station side data processing in the method for detecting faulty base stations of the present invention. The following further describes the processing process of the operation data on the base station side in combination with Figure 2 :

[0048] For the detection device placed in the cloud SON entity, set N per hour t1A time slice, and report information of each base station is collected within each time slice. Since the distribution of users is periodic and the resource allocation of the base station is highly correlated with the user distribution, it is considered that the index data of the base station is also periodic. Based on this, anomaly detection of the data on the base station side is carried out.

[0049] 1) First, the detection device collects the regular report information of each base station, and samples N t1 times at equal intervals per hour. Indexes on the base station side are collected at each time point, including three indexes: the number of user connections, power, and total throughput. Connect and sort the data of different base stations and different types in chronological order within nearly N day1 days. A set S = {S ijk |i ∈ U BSS , j ∈ (1, 2, 3), k ∈ (1, …, N t1 *24*N day1 )} is formed. The data set S ijk represents the value of the j-th index of the i-th base station at the k-th time point.

[0050] 2) Then, the time series data of each index j of each base station i is decomposed by STL time series decomposition with a period of 1 day, and is divided into: a periodic term a trend term and a remainder term

[0051] The periodic terms are removed from the time series data of different types. The processed data set S' = {S' ijk |i ∈ U BSS , j ∈ (1, 2, 3), k ∈ (1, …, N t1 *24*N day1 )} is obtained, where the processed set S' ijk = S ijk - Seasonal ijk .

[0052] 3) After that, GESD outlier detection is performed on the processed time series data S' of each index j of each base station i ij . It is judged whether the data S' ijk at the current time point k belongs to an outlier. If any index data j belongs to an outlier, it is initially determined that the base station i may have a failure at this time, and further detection will be performed.

[0053] The specific algorithm is as follows:

[0054] Input: The operation data on the base station side; Output: Abnormal data (i.e., the set of abnormal base station numbers) used to indicate the situation of suspected faulty base stations.

[0055] Step 1: Receive the input, organize and arrange the data in chronological order to obtain S = {S ijk | i ∈ U BSS , j ∈ (1, 2, 3), k ∈ (1, …, N t1 *24*N day1 )}.

[0056] Step 2: Initialize the empty set R = {} of the detected abnormal base station numbers.

[0057] Step 3: Iterate cyclically with i ← 1:N BSS :[[]]

[0058] Step 3.1: Initialize the logical judgment flag Flag = False.

[0059] Step 3.2: Iterate cyclically with j ← 1:3:

[0060] Step 3.2.1: Perform STL time series decomposition, Seasonal ij , Trend ij , Rem ij = STL(S ij )

[0061] Step 3.2.2: Remove the periodic term from the original data to obtain S' = {S' ijk | i ∈ U BSS , j ∈ (1, 2, 3), k ∈ (1, …, N t1 *24*N day1 )}, where S' ijk = S ijk - Seasonal ijk .

[0062] Step 3.2.3: Perform GESD detection on to judge whether the data at the current time point is an outlier. If so, set Flag = True.

[0063] Step 3.3: If Flag == True, then R = R ∪ {i}

[0064] Step 4: Output R.

[0065] Explanation of some of the above mathematical symbols:

[0066] N BSS : Represents the number of base stations.

[0067] U BSS = {1, 2, …, N BSS}: Set of base station serial numbers.

[0068] N t1 = 12: Represents the number of time slices divided per hour in the preliminary anomaly detection.

[0069] R: Represents the set of base station serial numbers where anomalies are detected in the preliminary anomaly detection.

[0070] Please refer to Figure 3 , Figure 3 is a schematic flowchart of an embodiment of user-side data processing in the detection method of faulty base stations of the present invention. The following further describes the processing process of user data on the user side in combination with Figure 3 :

[0071] After the preliminary detection on the base station side, a set R composed of the initially determined faulty base station serial numbers is obtained. Next, further fault detection and confirmation are required based on the user-side data. For the user-side data, N t2 time slices are set per hour, and the coordinates, timestamps, and signal quality index data of the users served by the base station are periodically collected.

[0072] After receiving the abnormal data from the preliminary detection on the base station side at a certain moment, assuming that the detection device collects user-side data N t2 times per hour:

[0073] 1) First, for each base station in R, with its location as the center, a square area with a side length of (N d *L) is specified as the detection area, and logically, this area is divided into N d ×N d small square grids of L×L. 2) Then, for each piece of user data from the user side, including location information, current cell signal, maximum neighbor cell signal, and current cell signal-to-noise ratio, it is judged whether the area where it is located is within the detection area according to the location information, and it is judged which small square grid the data belongs to and classified and counted.

[0074] 3) After classifying the user data in the detection area, according to the statistical information, the distribution of signal index data in each small grid at the current time point is compared with the distribution of signal index data in the same small grid during the same time period in the past N day2 days, and the Wasserstein distance between the two distributions is calculated. This distance represents the difference in signal quality distribution of this index in this grid area. According to the signal quality distribution difference, the square detection area with a side length of (N d *L) is transformed into a pixel map with 3 layers through the mapping function f, and the shape is (3×N d ×N d) The matrix, where each layer represents the two-dimensional plane distribution of the difference degree of a signal index (current cell signal, maximum neighbor cell signal, signal-to-noise ratio of the current cell), and the value range is [0, 255]. Subsequently, the pixel map is used as input through a pre-trained convolutional neural network to obtain the failure probability of the central base station corresponding to the pixel map.

[0075] 4) Perform operations 2 - 3 for each base station in PT to obtain the failure probabilities of all base stations in R.

[0076] When the failure probability is greater than the preset value, the corresponding base station is considered a faulty base station.

[0077] The specific algorithm is as follows:

[0078] Input: User - side data D at the current time point now ={(x i , y i , rsrp i , rsrpn i , sinr i ) | i = 1, 2, …} (representing the abscissa, ordinate, current cell signal value, maximum neighbor cell signal value, and signal - to - noise ratio of the current cell respectively), user - side data D day2 in the same hour period within the past N past days={(x′ i , y′ i , rsrp′ i , rsrpn′ i , sinr′ i ) | i = 1, 2, …}, and the set R of abnormal base station numbers detected on the base - station side; Output: The failure probability P of each base station in R.

[0079] Step 1: Initialize the set C = {C ij | i ∈ N + , j ∈ N +}, where C ij ={} represents the information statistically collected in the detection small grid at the logical position of the i - th row and j - th column at the current time point. Initialize the set C′ = {C′ ij | i ∈ N + , j ∈ N +}, where C′ ij ={} represents the information statistically collected in the detection small grid at the logical position of the i - th row and j - th column in the same hour period within the past few days.

[0080] Step 2: Process the data at the current time point, and perform loop iteration i ← 1: |D now |:

[0081] Step 2.1: Check whether the data belongs to the detection area of a certain base station and count. Iterate cyclically j←1:|R|:

[0082] Step 2.2: Calculate the logical coordinates of the detection small grid corresponding to the position of the data where floor R→Z is the floor symbol.

[0083] Step 2.3: Determine whether the logical position (indexX, indexY) is within the detection area of base station R j If so, C indexX,indexY = C index,indexY ∪{(rsrp i , rsrpn i , sinr i )}.

[0084] Step 2.4 is executed until the j loop ends

[0085] Step 3: Process the data at the past time point. Iterate cyclically i←1:|D past |:

[0086] Step 3.1: Check whether the data belongs to the detection area of a certain base station and count. Iterate cyclically j←1:|R|:

[0087] Step 3.2: Calculate the logical coordinates of the detection small grid corresponding to the position of the data

[0088] Step 3.3: Determine whether the logical position (indexX, indexY) is within the detection area of base station R j If so, C′ indexX,indexY = C′ indexX,lndexY ∪{(rsrp′ i , rsrpn′ i , sinr′ i )}.

[0089] Step 3.4 is executed until the j loop ends

[0090] Step 4: Initialize R = {0,0,0…,0}, |R| = |R|, P j represents the failure probability of R j , represents the failure indication of base station R j .

[0091] Step 5: Iterate cyclically j←1:|R|:

[0092] Step 5.1: Initialize A pixel map representing the abnormality conversion of the base station detection area.

[0093] Step 5.2: Calculate the starting point of the logical coordinates of the base station detection area Where represents the coordinates of base station R j .

[0094] Step 5.3: Loop and iterate xbais←0:N d -1, ybais←0:N d -1:

[0095] Step 5.3.1: IMG xbias,ybias = Wasserstein(C minX+xbias.minY+ybias , C′ minX+xbias.minY+ybias ), where the Wasserstein function is used to calculate the Wasserstein distance of the distribution of three indicators.

[0096] Step 5.3.2: Convert the distance to a pixel value, IMG xbias,ybias = min(IMG xbias,ybias *5, 255).

[0097] Step 5.3.3: Calculate the failure probability p of R j = CNN(IMG).

[0098] Step 5.3.4: Update the obtained result, P j = p, if p>0.5 then

[0099] Step 6: Output the results P, P BOOL .

[0100] Explanation of some of the above mathematical symbols:

[0101] N t2 = 12: Represents the number of time slices divided per hour in the deep anomaly detection.

[0102] N d = 16: Represents the number of small square areas included in the rows / columns of the monitoring area in the deep anomaly detection.

[0103] L = 25(m): Represents the side length of the small square areas in the rows / columns of the monitoring area in the deep anomaly detection.

[0104] f R→R(x) = MIN(x * 5, 255): Represents the mapping function that converts the Wasserstein distance to pixel values in depth anomaly detection.

[0105] The detection device for faulty base stations provided by the present invention will be described below. The detection device for faulty base stations described below can be correspondingly referred to the detection method for faulty base stations described above.

[0106] The present invention also provides a detection device for faulty base stations. Please refer to Figure 4 , Figure 4 is a schematic structural diagram of an embodiment of the detection device for faulty base stations of the present invention. In this embodiment, the detection device 400 for faulty base stations may include an operation data module 410, an abnormal data module 420, a suspected faulty base station module 430, and a faulty base station module 440. Specifically:

[0107] The operation data module 410 is used to obtain the operation data on the base station side and the user data on the user side;

[0108] The suspected faulty base station 420 is used to perform time series decomposition and outlier detection on the operation data to obtain the situation of suspected faulty base stations; among them, the situation of suspected faulty base stations includes the base station serial numbers of suspected faulty base stations;

[0109] The distribution difference 430 is used to calculate the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station based on the situation of the suspected faulty base station;

[0110] The faulty base station module 440 is used to determine the faulty base station based on the signal quality distribution difference.

[0111] In some embodiments, the data acquisition module 410 is used for:

[0112] Collect the operation data on the base station side at equal intervals every hour; among them, the operation data includes base station metrics, and the base station metrics include the number of user connections, power, and total throughput; connect and sort the operation data within a preset number of days in chronological order to form a data set S ijk ; where S ijk represents the value of the j -th metric of the i -th base station at the k -th time point; perform STL time series decomposition on the time series data S ij of each base station and each metric with a one - day period respectively to obtain the periodic term Seasonal ij , the trend term Trend ij and the remainder term Rem ij ; based on the data set S ijk and the periodic term Seasonal ij obtain the processed set S′ ijk , where S′ijk = S ijk - Seasonal ijk 。

[0113] In some embodiments, the suspected faulty base station 420 is used for:

[0114] Performing GESD outlier detection on the timing data S'; if the j-index in the processing set S' at the current time point k ij belongs to an outlier, then the corresponding base station numbered i is obtained as the suspected faulty base station. ij

[0115] In some embodiments, the distribution difference module 430 is used for:

[0116] Determining a detection area centered on the suspected faulty base station, and dividing the detection area into several grids; obtaining user data on the user side within the detection area, and classifying the user data according to the grid based on the location information; wherein, the user data includes location information and signal metrics, and the signal metrics include the current cell signal, the maximum neighbor cell signal, and the signal-to-noise ratio of the current cell signal.

[0117] In some embodiments, the distribution difference module 430 is used for:

[0118] Comparing the signal metric distribution at the current time point in each grid with the signal metric distribution in the same grid during the same period of historical data, and obtaining the signal quality distribution difference of the signal metrics in the corresponding grid.

[0119] In some embodiments, the faulty base station module 440 is used for:

[0120] Calculating the Wasserstein distance between the signal metric distribution at the current time point in each grid and the signal metric distribution in the same grid during the same period of historical data; wherein, the Wasserstein distance represents the distribution difference degree of the metrics in the corresponding grid area; converting the detection area into a pixel map of three layers through a mapping function f according to the distribution difference degree, wherein each layer represents a two-dimensional plane distribution of the signal metric difference degree; inputting the pixel map into a pre-trained convolutional neural network to obtain the fault probability of the base station corresponding to the pixel map; and determining the faulty base station based on the fault probability of the base station.

[0121] ​This embodiment provides a detection device for a faulty base station. By performing time series decomposition and outlier detection on the operation data of the base station side, the situation of suspected faulty base stations is obtained. According to the situation of suspected faulty base stations, the signal quality distribution difference between the current user data and the historical user data of the corresponding part of the user side is calculated, and finally the faulty base station is determined. Through the above method, the faulty base station detection device of this embodiment considers the combined spatio-temporal information analysis of the base station side data and the user side data, and does not need to calculate all the user side data, which greatly reduces the computational resource overhead and time overhead of the detection method and reduces the average running time.

[0122] The present invention also provides an electronic device. Please refer to Figure 5 , Figure 5 which is a schematic structural diagram of an embodiment of the electronic device of the present invention. In this embodiment, the electronic device 500 may include a memory 510, a processor 520, and a computer program stored on the memory 520 and executable on the processor 510. When the processor 510 executes the program, it implements the detection method of the faulty base station provided by the above various methods.

[0123] Optionally, the electronic device 500 may further include a communication bus 530 and a communication interface 540. Among them, the processor 510, the communication interface 540, and the memory 520 complete mutual communication through the communication bus 530. The processor 510 may call the logical instructions in the memory 520 to execute the detection method of the faulty base station, and the method includes: obtaining the operation data of the base station side and the user data of the user side; performing time series decomposition and outlier detection on the operation data to obtain the situation of suspected faulty base stations; wherein, the situation of suspected faulty base stations includes the base station serial number of the suspected faulty base station; based on the situation of suspected faulty base stations, calculating the signal quality distribution difference between the current user data and the historical user data within the square area centered on the suspected faulty base station; and determining the faulty base station based on the signal quality distribution difference.

[0124] In addition, when the logical instructions in the above-mentioned memory 530 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0125] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the detection method of a faulty base station provided by the above-mentioned various methods. The steps and principles have been introduced in detail in the above methods and will not be repeated here.

[0126] On yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the detection method of a faulty base station provided by the above-mentioned various methods. The steps and principles have been introduced in detail in the above methods and will not be repeated here.

[0127] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0128] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A detection method for a faulty base station, characterized in that, Including: Obtaining the operation data on the base station side and the user data on the user side; The user data is the user data in the area near the suspected faulty base station; Performing time series decomposition and outlier detection on the operation data to obtain the situation of the suspected faulty base station; wherein, the situation of the suspected faulty base station includes the base station serial number of the suspected faulty base station; Based on the situation of the suspected faulty base station, calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station; Based on the signal quality distribution difference, determining the faulty base station; The performing time series decomposition on the operation data includes: Collecting the operation data on the base station side at equal intervals every hour; wherein, the operation data includes base station metrics, and the base station metrics include the number of user connections, power, and total throughput; Connect and sort the operation data within the preset number of days in chronological order respectively to form a data set S ijk ; where S ijk represents the value of the j-th indicator of the i-th base station at the k-th time point; The time series data S of each indicator of each base station ij Taking one day as a cycle, perform STL time series decomposition respectively to obtain the seasonal term Seasonal ij , the trend term Trend ij and the remainder term Rem ij ; Based on the data set S ijk and the seasonal term Seasonal ij obtain the processed set S′ ijk , where S′ ijk = S ijk - Seasonal ijk ; The performing time series decomposition and outlier detection on the operation data to obtain the situation of the suspected faulty base station includes: Perform GESD outlier detection on the time series data S′ ij ; If the j index in the processing set S' at the current time point k ijk belongs to an abnormal point, the corresponding base station numbered i is obtained as the suspected faulty base station; Before calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station based on the situation of the suspected faulty base station, it includes: Determining a detection area centered on the suspected faulty base station and dividing the detection area into several grids; Obtaining the user data on the user side within the detection area and classifying the user data according to the grid according to the location information; wherein, the user data includes the location information and signal metrics, and the signal metrics include the current cell signal, the maximum neighbor cell signal, and the current cell signal-to-noise ratio; The calculating the signal quality distribution difference between the current user data and the historical user data within a square area centered on the suspected faulty base station includes: Comparing the signal metric distribution at the current time point in each grid with the signal metric distribution in the same grid during the same period of historical data to obtain the signal quality distribution difference of the signal metrics in the corresponding grid; The comparing the signal metric distribution at the current time point in each grid with the signal metric distribution in the same grid during the same period of historical data to obtain the signal quality distribution difference of the signal metrics in the corresponding grid, and based on the signal quality distribution difference, determining the faulty base station includes: Calculating the Wasserstein distance between the signal metric distribution at the current time point in each grid and the signal metric distribution in the same grid during the same period of historical data; wherein, the Wasserstein distance represents the distribution difference degree of the metric in the corresponding grid area; Converting the detection area into a pixel map of three layers through a mapping function f according to the distribution difference degree, wherein each layer represents a two-dimensional plane distribution of the signal metric difference degree; Inputting the pixel map into a pre-trained convolutional neural network to obtain the fault probability of the base station corresponding to the pixel map; Based on the fault probability of the base station, determining the faulty base station.

2. A detection device for a faulty base station, characterized in that, Including: A data acquisition module for obtaining the operation data on the base station side and the user data on the user side; The user data is the user data in the area near the suspected faulty base station; The suspected fault base station module is used to perform time series decomposition and outlier detection on the operation data to obtain the situation of suspected fault base stations; wherein, the situation of suspected fault base stations includes the base station serial numbers of the suspected fault base stations. The distribution difference module is used to calculate the signal quality distribution difference between the current user data and the historical user data within the square area centered on the suspected fault base station based on the situation of the suspected fault base stations. The fault base station module is used to determine the fault base station based on the signal quality distribution difference. The data acquisition module is further configured to: collect the operation data of the base station side at equal intervals every hour; wherein, the operation data includes base station metrics, and the base station metrics include the number of user connections, power, and total throughput; connect and sort the operation data within a preset number of days in chronological order to form a data set S ijk ; wherein, S ijk represents the value of the j-th metric of the i-th base station at the k-th time point; for the time series data S of each base station and each metric ij Perform STL time series decomposition with a one-day period to obtain the periodic term Seasonal ij , the trend term Trend ij and the remainder term Rem ij ; Based on the data set S ijk and the periodic term Seasonal ij obtain the processed set S' ijk , wherein, S' ijk = S ijk - Seasonal ijk ; The suspected faulty base station is also used for: performing GESD outlier detection on the timing data S'; if the j index in the processing set S' at the current time point k ij belongs to an outlier, then obtain the corresponding base station numbered i as the suspected faulty base station; ijk ​ The distribution difference module is further used to: determine a detection area centered on the suspected fault base station and divide the detection area into several grids; obtain the user data on the user side within the detection area and classify the user data according to the grid according to the location information; wherein, the user data includes location information and signal indicators, and the signal indicators include the current cell signal, the maximum neighbor cell signal, and the current cell signal-to-noise ratio. The distribution difference module is further used to: compare the signal indicator distribution at the current time point within each grid with the signal indicator distribution within the same grid in the same time period of the historical data to obtain the signal quality distribution difference of the signal indicators within the corresponding grid. The fault base station module is further used to: calculate the Wasserstein distance between the signal indicator distribution at the current time point within each grid and the signal indicator distribution within the same grid in the same time period of the historical data; wherein, the Wasserstein distance represents the distribution difference degree of the indicator within the corresponding grid area; convert the detection area into a pixel map of three layers through a mapping function f according to the distribution difference degree, where each layer represents the two-dimensional plane distribution of a signal indicator difference degree; input the pixel map into a pre-trained convolutional neural network to obtain the fault probability of the base station corresponding to the pixel map; determine the fault base station based on the fault probability of the base station.

3. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the detection method of the fault base station as described in claim 1.

4. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the detection method of the fault base station as described in claim 1.

5. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the detection method of the fault base station as described in claim 1.

Images