A network equipment monitoring time series data processing method based on artificial intelligence

Abstract

The present invention provides a method for processing time series data for monitoring network equipment based on artificial intelligence, comprising: acquiring network topology structure data and historical load data, constructing a topology graph model including node importance indicators and connection strength parameters, and determining a candidate set of key equipment; calculating the dynamic importance score of each device in the current network structure according to real-time load data and the topology graph model, and generating a list of key equipment; extracting time series feature vectors of collaborative operation between devices from the time series data set based on the device dependency relationship of the topology graph model; when the fluctuation amplitude of the time series feature vector exceeds a preset threshold, determining it as a group collaborative abnormal state and generating a performance bottleneck warning signal; analyzing the high-frequency time series data of the key equipment to identify potential faulty equipment; updating the topology graph model according to the fault probability, adjusting the node weight parameters, and regenerating the candidate set of key equipment.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an artificial intelligence-based network equipment monitoring time series data processing method. Background Art

[0002] Processing time-series data from network device monitoring is a core area of efficient data center operation and fault prevention. Its importance lies in ensuring the stability and performance optimization of large-scale network systems. As data centers continue to expand, the surge in the variety and number of network devices has led to an explosive growth in the amount of monitoring data. Real-time processing and analysis of this time-series data is crucial to ensuring system reliability.

[0003] However, existing monitoring data processing methods have significant limitations when dealing with complex network environments. Traditional methods often rely on fixed sampling rates and preset rules, making them difficult to adapt to the dynamic changes in data center loads. They also cannot effectively identify the collaborative operation characteristics of devices, resulting in inefficient capture of key performance bottlenecks. Furthermore, when dealing with large-scale network topologies, existing methods often ignore the dependencies between devices. The extracted time series data lacks characteristics that reflect the collaborative state of the group, limiting the ability to optimize overall performance.

[0004] The key to processing time-series data for data center network equipment monitoring lies in accurately identifying critical paths and core devices within complex network topologies. The dynamic nature of network topologies and the heterogeneity of devices make it difficult to accurately capture the operating status of key devices using static rules. Unresolved identification issues further lead to inefficient monitoring sampling strategies. Peak load fluctuations exacerbate performance bottlenecks. Failure to dynamically adjust the sampling rate of key devices can lead to missing subtle signals that could signal a failure. Extracting time-series features based on inter-device dependencies presents a deeper challenge, as the coordinated state of a group of devices is highly dependent on the interaction between the topology and the operating load, making it difficult for traditional algorithms to model this complex relationship. These interrelated technical factors collectively constitute a core obstacle to data center monitoring data processing. Summary of the Invention

[0005] The present invention provides a method for processing network equipment monitoring time series data based on artificial intelligence, which mainly includes:

[0006] Acquire network topology data and historical load data, construct a topology model including node importance indicators and connection strength parameters, and determine the candidate set of key devices; calculate the dynamic importance score of each device in the current network structure based on the real-time load data and the topology model, and generate a list of key devices; set a dynamic sampling strategy based on the dynamic importance score, adopt a high-frequency sampling mode for key devices and a low-frequency sampling mode for non-key devices to collect time series data sets; based on the device dependency relationship of the topology model, extract the time series feature vector of the collaborative operation between devices from the time series data set; when the fluctuation amplitude of the time series feature vector exceeds the preset threshold, it is determined to be an abnormal state of group collaboration and a performance bottleneck warning signal is generated; analyze the high-frequency time series data of the key devices to identify potential faulty devices; update the topology model according to the failure probability, adjust the node weight parameters, and regenerate the candidate set of key devices.

[0007] Furthermore, the calculation formula of the node importance index is:

[0008] ;

[0009] in, is the node importance index, is the node betweenness centrality, is the bandwidth load factor, is the device health indicator, 、 、 is the weighted coefficient and satisfies .

[0010] Furthermore, the method for calculating the connection strength parameter includes: , real-time delay and packet loss rate , the connection strength is calculated using the following formula:

[0011] ;

[0012] in, A very small constant to prevent division by zero.

[0013] Furthermore, the dynamic importance score calculation formula is:

[0014] ;

[0015] in, Score the dynamic importance, For nodes The load fluctuation variance, is the average load, is the set of adjacent nodes.

[0016] Furthermore, the dynamic sampling strategy is set based on the dynamic importance score, including: calculating the mean and standard deviation according to the statistical distribution of the dynamic importance score, and dynamically adjusting the first preset threshold and the second preset threshold; if the dynamic importance score of the device exceeds the first preset threshold, setting the high-frequency sampling parameters; if the dynamic importance score of the device is lower than the second preset threshold, setting the low-frequency sampling parameters; and determining the sampling frequency of each device according to the high-frequency sampling parameters and the low-frequency sampling parameters.

[0017] Furthermore, the device dependency relationship based on the topological graph model extracts the timing feature vectors of the collaborative operation between devices from the timing data set, including: using a graph neural network with a spatiotemporal graph convolution structure to aggregate the timing features of adjacent nodes for the device dependency relationship in the topological graph model; and generating the timing feature vector of each device through the convolution operation of the graph neural network.

[0018] Furthermore, when the fluctuation amplitude of the time series characteristic vector exceeds a preset threshold, it is determined to be a group collaborative abnormal state, including: calculating the ratio of the fluctuation amplitude of the time series characteristic vector to the historical average value to obtain a fluctuation ratio; if the fluctuation ratio exceeds a third preset threshold, it is determined to be a group collaborative abnormal state.

[0019] Furthermore, the high-frequency time series data of the key equipment is analyzed to identify potential faulty equipment, including: using a multi-scale wavelet analysis method to decompose the high-frequency time series data of the key equipment to obtain multi-scale time series components; through abnormal pattern matching, the multi-scale time series components are compared with preset fault modes; if the match is successful, the corresponding equipment is determined to be a potential faulty equipment.

[0020] Furthermore, the updating of the topology model according to the failure probability and the adjustment of the node weight parameters include: adjusting the weight parameters of the corresponding nodes in the topology model according to the failure probability of the potential faulty equipment; recalculating the node importance index and the connection strength parameter through the updated weight parameters; reconstructing the topology model according to the updated node importance index and the connection strength parameter; and generating a new candidate set of key equipment based on the updated topology model.

[0021] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0022] This paper proposes an artificial intelligence-based method for processing time-series data for network device monitoring. This method constructs a topological graph model that incorporates node importance and connection strength, combines it with real-time load data to calculate dynamic device importance scores, and generates a list of critical devices. Based on these scores, a dynamic sampling strategy is set, with high-frequency sampling applied to critical devices and low-frequency sampling applied to non-critical devices, effectively balancing monitoring accuracy and resource consumption. A graph neural network is used to extract time-series features of collaborative operation between devices, enabling timely detection of abnormal group conditions. Furthermore, multi-scale wavelet analysis is performed on critical devices to identify potential faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The present invention is a flowchart of a method for processing network equipment monitoring time series data based on artificial intelligence. DETAILED DESCRIPTION

[0024] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] like Figure 1 In this embodiment, a method for processing network device monitoring time series data based on artificial intelligence may specifically include:

[0026] Step S101: Obtain network topology data and historical load data, construct a topology graph model including node importance indicators and connection strength parameters, and determine a candidate set of key devices;

[0027] Step S102. Calculate the dynamic importance score of each device in the current network structure based on the real-time load data and the topology model, and generate a list of key devices;

[0028] Step S103. Setting a dynamic sampling strategy based on the dynamic importance score, using a high-frequency sampling mode for key devices and a low-frequency sampling mode for non-key devices to collect time series data sets;

[0029] Step S104: Based on the device dependency relationship of the topology model, extract the time series feature vector of the collaborative operation between devices from the time series data set;

[0030] Step S105: When the fluctuation amplitude of the time series feature vector exceeds a preset threshold, it is determined to be an abnormal state of group collaboration and a performance bottleneck warning signal is generated;

[0031] Step S106: Analyze the high-frequency time series data of the key equipment to identify potential faulty equipment;

[0032] Step S107: Update the topology model according to the failure probability, adjust the node weight parameters, and regenerate the key device candidate set.

[0033] Specifically, network topology data and historical load data are acquired, and a topology graph model containing node importance indicators and connection strength parameters is constructed to identify a candidate set of key devices. Based on the real-time load data and the topology graph model, the dynamic importance score of each device in the current network structure is calculated to generate a list of key devices. Based on the statistical distribution of the dynamic importance score, the first and second preset thresholds are dynamically adjusted using the mean and standard deviation to determine the threshold ranges for high-frequency and low-frequency sampling. If the device's dynamic importance score exceeds the first preset threshold, the high-frequency sampling parameter is enabled; if it falls below the second preset threshold, the low-frequency sampling parameter is enabled. Based on the threshold judgment results, a dynamic sampling strategy is generated. A time series dataset is collected using the dynamic sampling strategy, with high-frequency sampling used for key devices and low-frequency sampling used for non-critical devices. Based on the device dependencies in the topology graph model, a graph neural network is applied to extract time series feature vectors that indicate inter-device collaborative operation from the time series dataset. If the fluctuation amplitude of the time series feature vector exceeds the third preset threshold, it is determined to be an abnormal state of group collaboration. Through abnormal state analysis, a performance bottleneck warning signal is generated. Based on performance bottleneck warning signals and time series feature vectors, a clustering algorithm is used to identify abnormal device clusters and determine their topological locations. Using the topological locations of abnormal device clusters and historical load data, the node importance indicators and connection strength parameters of the topology model are updated to produce an optimized topology model.

[0034] In step S101, the calculation formula of the node importance index is:

[0035] ;

[0036] in, is the node importance index, is the node betweenness centrality, is the bandwidth load factor, is the device health indicator, 、 、 is the weighted coefficient and satisfies .

[0037] Specifically, the adjacency data and device operation status data of each node in the network topology are obtained, and the Brandes algorithm is used to calculate the node betweenness centrality C(v). For example, the betweenness centrality value of node v is 0.85. Based on real-time network traffic data and bandwidth allocation data, the bandwidth load factor B(v) of each node is calculated using the formula B(v) = current traffic / maximum bandwidth. For example, the bandwidth load factor of node v is 0.72. Through multi-dimensional data fusion of device operation logs and health monitoring data, the device health index D(v) of each node is calculated. The calculation formula is:

[0038] ;

[0039] in, 、 and Reflect the weight coefficients of historical reliability, alarm severity and real-time performance status respectively, satisfying The actual value can be adjusted according to the device type, for example, 、 、 .

[0040] in: is the reliability factor, and the calculation formula is:

[0041] ;

[0042] The number of failures recorded by the device in the last 30 days. is the attenuation coefficient, which can be adjusted through historical data training.

[0043] For example, when the device When 3 failures occur,

[0044] is the alarm factor, and its calculation formula is:

[0045] ;

[0046] in, The number of currently active alarms, including critical alarms such as CPU overload and memory leak; , is the maximum normalized threshold;

[0047] For example, the device When there are 3 active alarms, .

[0048] is the performance factor, and its calculation formula is:

[0049] ;

[0050] in, is the CPU utilization, and the preset threshold is ; is the memory usage, and the preset threshold is ; is the disk I / O latency rate, with a preset threshold of 85%;

[0051] The node's device health indicator ,The node importance index uses the preset weighted coefficients α=0.4, β=0.3, and γ=0.3, satisfying α+β+γ=1, and the node importance index NI(v)=0.4×0.85+0.3×0.72+0.3×0.575 is calculated, and NI(v)=0.73 is obtained.

[0052] The calculation method of the connection strength parameter includes: , real-time delay and packet loss rate , the connection strength is calculated using the following formula:

[0053] ;

[0054] in, A very small constant to prevent division by zero.

[0055] In one embodiment, the historical data transmission volume between node A and node B is extracted from the network monitoring system. 1000MB, real-time delay 50ms, packet loss rate The original transmission characteristic data is obtained through log collection tools. Perform logarithmic transformation and calculate using log(1000+1) to obtain the normalized transmission characteristic value. The preset minimum constant =0.001, to avoid the denominator being 0, and to obtain the denominator of connection stability. =log(1000+1) / (50×0.02+0.001), calculate the connection strength =3.00.

[0056] In step S102, the dynamic importance score calculation formula is:

[0057] ;

[0058] in, Score the dynamic importance, For nodes The load fluctuation variance, is the average load, is the set of adjacent nodes.

[0059] Specifically, node information is extracted from the network topology model, combined with real-time load data, to calculate the historical load data sequence of each node. For example, the load data sequence of node A is [10, 15, 12, 18]. The load fluctuation variance is calculated using the variance formula =8.5, calculate the average load by arithmetic average method =13.75. According to the topology model, the adjacent node set N(v)={B, C, D} of node A is extracted to determine the connection relationship between node A and adjacent nodes B, C, and D. , real-time delay and packet loss rate , for example, node A and node B =1000, =5ms, =0.01, using the formula , the calculated connection strength CS(AB)=4.6. The Brandes algorithm is used to calculate the betweenness centrality of node A C(v)=0.3. The bandwidth load factor B(v)=0.8, the device health index D(v)=0.9, and the formula is used , where α=0.4, β=0.3, γ=0.3, the node importance index NI(v)=0.63 is calculated. Based on the load fluctuation variance σ(v)=8.5 and the average load μ(v)=13.75 of node A, the dynamic adjustment factor is calculated. =1.62, the load dynamic impact weight of node A is obtained. Through the node importance index NI(v)=0.63, dynamic adjustment factor 1.62 and connection strength =4.6, =3.8, =5.2, using the formula , the dynamic importance score of node A is calculated to be DIS(v)=0.63×1.62×(4.6+3.8+5.2)=13.8.

[0060] In step S103, the threshold setting of the dynamic sampling strategy is based on the statistical distribution of the dynamic importance score, and the high and low thresholds are dynamically adjusted through the mean and standard deviation, which also includes:

[0061] Obtain device operation data and calculate a dynamic importance score to obtain a score for each device. Based on the dynamic importance score, calculate the mean and standard deviation of all device scores to determine the statistical distribution parameters. If the mean and standard deviation are calculated, dynamically adjust the high-frequency sampling threshold and low-frequency sampling threshold by adding or subtracting multiples of the standard deviation from the mean to obtain the high and low threshold ranges. Obtain the device's current dynamic importance score and compare it with the high and low thresholds to determine whether it exceeds the first preset threshold or is below the second preset threshold. If the device score exceeds the first preset threshold, enable the high-frequency sampling parameters and generate a high-frequency sampling strategy. If it is below the second preset threshold, enable the low-frequency sampling parameters and generate a low-frequency sampling strategy to obtain the device sampling strategy.

[0062] Specifically, obtain the equipment operation data through the formula Calculate the dynamic importance score to obtain the score value of each device. According to the dynamic importance score, calculate the mean and standard deviation of the dynamic importance scores of all devices, for example, the mean is 75 and the standard deviation is 10, and determine the statistical distribution parameters. Dynamically adjust the high and low thresholds by adding or subtracting multiples of the standard deviation from the mean, for example, the high-frequency sampling threshold is the mean plus 1.5 times the standard deviation (90), and the low-frequency sampling threshold is the mean minus 1 times the standard deviation (65), to obtain the high and low threshold ranges. Obtain the current dynamic importance score of the device, for example, a device score is 92, and by comparing it with the high and low thresholds, determine whether it exceeds the first preset threshold (90). If the device score exceeds the first preset threshold, enable the high-frequency sampling parameter, for example, set the sampling interval to 1 second, and generate a high-frequency sampling strategy; if it is lower than the second preset threshold, enable the low-frequency sampling parameter, for example, set the sampling interval to 10 seconds, and generate a low-frequency sampling strategy, and obtain the device sampling strategy.

[0063] In step S104, extracting the time series feature vector of the collaborative operation between devices from the time series data set based on the device dependency relationship of the topology model further includes:

[0064] Obtain dependency data between devices and construct a node and edge structure based on a topological graph model. Extract the operational status features of each device from the time series dataset to generate an initial time series feature matrix. A graph neural network with a spatiotemporal graph convolutional structure aggregates the time series features of adjacent nodes in the topological graph. Through the convolution operation of the graph neural network, a collaborative time series feature vector for each device is generated.

[0065] Specifically, the dependency data between devices is obtained, and the node and edge structure based on the topological graph model is constructed. For example, the connection relationship between devices is extracted through device communication logs and operating status records to form a topological graph containing 80 nodes and 200 edges. The operating status characteristics of each device are extracted through the time series data set to generate an initial time series feature matrix. For example, parameters such as CPU occupancy, data transmission volume, and network load fluctuation rate are obtained to form a feature matrix of 1000 time steps. A graph neural network with a spatiotemporal graph convolution structure is used to aggregate the time series features of adjacent nodes in the topological graph. For example, the ST-GCN model is used to extract the features of the time dimension and the space dimension respectively through time convolution and space convolution. Through the convolution operation of the graph neural network, the collaborative time series feature vector of each device is generated. For example, the feature matrix after convolution is reduced in dimensionality to obtain a 128-dimensional feature vector for each device.

[0066] In step S105, the abnormal state of group collaboration is determined by calculating the ratio of the fluctuation amplitude of the characteristic vector to the historical average value, and triggering an early warning when it exceeds a set threshold, which also includes:

[0067] Obtain time series feature vector data, calculate the feature vector fluctuation amplitude, and perform statistical analysis on the sequence using the standard deviation formula to obtain the current fluctuation amplitude value. Obtain historical feature vector data and extract the feature vector sequence within a fixed time window in the past from the database to obtain a historical data set. Calculate the historical average value and perform mean calculation on the historical data set to obtain a baseline value for the historical fluctuation amplitude. Calculate the ratio of the fluctuation amplitude to the historical average value and divide the current fluctuation amplitude value by the historical average value to obtain the ratio result. If the ratio exceeds a third preset threshold, it is determined that the group coordination is abnormal and a performance bottleneck warning signal is generated.

[0068] Specifically, the time series feature vector data is obtained, and the sliding window standard deviation algorithm is used to calculate the fluctuation amplitude of the feature vector within the current window. The window size is 30 seconds and the step size is 5 seconds, resulting in a standard deviation value S = 2.45. The feature data of the same dimension within the past 24 hours is retrieved from the time series database TSDB. The historical series is aggregated at a granularity of 5 minutes, and the baseline value μ = 1.83 is calculated using the exponentially weighted moving average method. The current fluctuation amplitude is divided by the historical baseline value. When the ratio R = S / μ = 1.34 exceeds the preset threshold of 1.3, a collaborative anomaly determination is triggered, and the anomaly identification code A001 is generated. The anomaly identification code triggers an early warning signal, which is pushed to the monitoring center via the message queue.

[0069] In step S106, multi-scale wavelet analysis is performed on the high-frequency time series data of key equipment to identify potential faulty equipment through abnormal pattern matching, which also includes:

[0070] A dynamic sampling strategy is used to collect time series data sets, executing a high-frequency sampling pattern on key equipment to obtain high-frequency time series data. Multi-scale wavelet analysis is used to decompose the high-frequency time series data, generating time series characteristic coefficients at different scales. Pattern matching is performed on the time series characteristic coefficients based on predefined fault signatures in the abnormal pattern library to identify potential abnormal points. If an abnormal point is found, the abnormal point's time series feature vector is extracted to determine the abnormal pattern type.

[0071] Specifically, a dynamic sampling strategy is used to collect time series data sets. A high-frequency sampling mode is implemented for key equipment, with a sampling frequency set to 1000 times per second, to obtain high-frequency time series data. Multi-scale wavelet analysis is used to decompose the high-frequency time series data. A five-layer decomposition using Daubechies wavelet basis functions is performed to obtain time series characteristic coefficients at different scales. Pattern matching is performed on the time series characteristic coefficients based on predefined fault signatures in the abnormal pattern library. The Euclidean distance algorithm is used to calculate similarity and identify potential anomalies. If an anomaly is found, the time series feature vector of the anomaly is extracted, and the K-nearest neighbor algorithm is used to determine the type of abnormal pattern.

[0072] In step S107, when the topology model is updated, the node weight parameters are dynamically adjusted according to the failure probability, and the key device candidate set is regenerated, which also includes:

[0073] Network topology data and historical load data are obtained to construct a topology model that includes node importance indicators and connection strength parameters. The failure probability of each node is calculated based on the node's historical load data and fault records. Using the failure probability as input, node weights are calculated and the weight parameters of each node in the topology model are dynamically adjusted. If a node weight exceeds a preset threshold, the node importance indicator and connection strength parameters are recalculated. If a node importance indicator and connection strength parameters exceed the preset threshold, the node is added to the candidate set of key devices, resulting in an updated candidate set.

[0074] Specifically, the Poisson distribution model is used to calculate the failure probability of each node based on the node's historical load data and fault records. For example, if a node has experienced three failures in the past 30 days, its failure probability is 0.1. Using the failure probability as input, the node weight parameter is dynamically adjusted, and the weight formula is set as , W represents the node weight value, F represents the node failure probability, L represents the current node load rate, and The weight coefficients are 0.6 and 0.4 respectively. When the node weight value W exceeds the preset threshold of 0.45, the node importance index is recalculated. If the importance index of a node exceeds the threshold of 0.8 and the connection strength parameter exceeds the threshold of 4.5, it will be added to the key device candidate set.

[0075] The above is only a preferred embodiment of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and supplements without departing from the principles of the present invention. These improvements and supplements should also be regarded as the scope of protection of the present invention.

Claims

1. A network equipment monitoring time series data processing method based on artificial intelligence, characterized in that: include: Obtain network topology data and historical load data, build a topology model that includes node importance indicators and connection strength parameters, and determine the candidate set of key devices; Calculate the dynamic importance score of each device in the current network structure based on the real-time load data and the topology model, and generate a list of key devices; Setting a dynamic sampling strategy based on the dynamic importance score, adopting a high-frequency sampling mode for key devices and a low-frequency sampling mode for non-key devices to collect time series data sets; Extracting a time series feature vector of collaborative operation between devices from the time series data set based on the device dependency relationship of the topology model; When the fluctuation amplitude of the time series feature vector exceeds the preset threshold, it is determined to be an abnormal state of group collaboration and a performance bottleneck warning signal is generated; Analyze the high-frequency time series data of the key equipment to identify potential faulty equipment; The topology graph model is updated according to the failure probability, the node weight parameters are adjusted, and the key device candidate set is regenerated.

2. The method according to claim 1, wherein The calculation formula of the node importance index is: ； in, is the node importance index, is the node betweenness centrality, is the bandwidth load factor, is the device health indicator, is the weighted coefficient and satisfies .

3. The method according to claim 1, wherein The calculation method of the connection strength parameter includes: , real-time delay and packet loss rate , the connection strength is calculated using the following formula : ； in, A constant to prevent division by zero.

4. The method according to claim 1, wherein The calculation formula of the dynamic importance score is: ； in, Score the dynamic importance, For nodes The load fluctuation variance, is the average load, is the set of adjacent nodes, is the node importance index, For nodes and nodes The strength of the connection between them.

5. The method according to claim 1, wherein The setting of a dynamic sampling strategy based on the dynamic importance score includes: Calculating a mean and a standard deviation based on the statistical distribution of the dynamic importance score, and dynamically adjusting the first preset threshold and the second preset threshold; If the device dynamic importance score exceeds the first preset threshold, setting a high-frequency sampling parameter; If the device dynamic importance score is lower than the second preset threshold, setting the low-frequency sampling parameter; The sampling frequency of each device is determined according to the high-frequency sampling parameters and the low-frequency sampling parameters.

6. The method according to claim 1, wherein The extracting, from the time series data set, a time series feature vector of collaborative operation between devices based on the device dependency relationship of the topology graph model includes: A graph neural network with a spatiotemporal graph convolutional structure is used to aggregate the temporal features of adjacent nodes based on the device dependency in the topological graph model. Through the convolution operation of the graph neural network, a time series feature vector of each device is generated.

7. The method according to claim 6, wherein When the fluctuation amplitude of the time series feature vector exceeds a preset threshold, it is determined to be a group collaboration abnormal state, including: Calculating the ratio of the fluctuation amplitude of the time series characteristic vector to the historical average value to obtain a fluctuation ratio; If the fluctuation ratio exceeds a third preset threshold, it is determined to be a group coordination abnormal state.

8. The method according to claim 1, wherein The analysis of the high-frequency time series data of the key equipment to identify potential faulty equipment includes: Using a multi-scale wavelet analysis method, the high-frequency time series data of the key equipment is decomposed to obtain multi-scale time series components; Comparing the multi-scale time series components with a preset fault mode through abnormal pattern matching; If the match is successful, the corresponding device is determined to be a potential fault device.

9. The method according to claim 1, wherein The updating of the topology graph model according to the failure probability and adjusting the node weight parameters include: Adjusting the weight parameters of corresponding nodes in the topology graph model according to the failure probability of the potential faulty device; If the node weight value exceeds the preset threshold, the node importance index and connection strength parameter are recalculated; Reconstruct the topology graph model based on the updated node importance index and connection strength parameters; Based on the updated topology model, a new candidate set of key devices is generated.

Images