Equipment exception handling method and device, equipment, medium and product

Abstract

The embodiment of the invention discloses an equipment exception handling method and device, equipment, a medium and a product, and the method comprises the steps: obtaining the to-be-analyzed state sequence data of target power equipment, predicting an exception detection result based on a pre-trained exception detection model, and obtaining the to-be-analyzed state sequence data of the target power equipment; the anomaly detection model is obtained by training according to preset reference state sequence data in a normal operation state, determining an anomaly level prediction result according to an anomaly scoring result in the anomaly detection result, and determining a target anomaly processing scheme according to the anomaly level prediction result and a preset mapping relationship between an anomaly level and an anomaly processing scheme. According to the technical scheme, the problem that the abnormal condition of the power equipment cannot be automatically processed in time at present is solved, the abnormal condition of the power equipment and the corresponding abnormal processing scheme can be determined by fully utilizing the characteristics of the preset reference state sequence data in a machine learning mode, the abnormal processing measures are more flexible and effective, and the processing efficiency is improved. And stable operation of power equipment is ensured.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of computer technology, and in particular, to a method, apparatus, device, medium and product for handling device exceptions. Background Art

[0002] At present, in the actual application of power systems, although anomaly detection can provide abnormal warning, it is still a difficult problem to quickly take effective abnormal handling measures and restore the equipment to normal working state when an abnormality occurs in the equipment. Summary of the invention

[0003] The embodiments of the present invention provide a method, apparatus, device, medium and product for handling equipment abnormality. The methods can fully utilize the characteristics of preset reference state sequence data through machine learning to determine the abnormal conditions of power equipment and the corresponding abnormality handling solutions. The abnormality handling measures are more flexible and effective, thus ensuring the stable operation of power equipment.

[0004] In a first aspect, an embodiment of the present invention provides a method for handling device abnormalities, the method comprising:

[0005] Acquire state sequence data to be analyzed describing the operation state and running state of the target power equipment;

[0006] Inputting the state sequence data to be analyzed into a pre-trained anomaly detection model to obtain an anomaly detection result, wherein the pre-trained anomaly detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is state sequence data of the target power equipment in a normal operating state;

[0007] In the case where the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the anomaly duration is determined according to the time information of the abnormal state data in the anomaly detection result;

[0008] Determine whether exception handling is required based on the duration of the exception. If exception handling is required, determine a target exception handling solution based on the exception level prediction result and the mapping relationship between the preset exception level and the exception handling solution.

[0009] Perform abnormal processing on the target power equipment according to the target abnormal processing plan.

[0010] In a second aspect, an embodiment of the present invention provides a device exception handling device, the device comprising:

[0011] A state sequence data acquisition module, used to acquire state sequence data to be analyzed that describes the operation state and running state of the target power equipment;

[0012] An abnormality detection result determination module is used to input the state sequence data to be analyzed into a pre-trained abnormality detection model to obtain an abnormality detection result, wherein the pre-trained abnormality detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment in a normal operating state;

[0013] An abnormality level determination module is used to determine the abnormality level prediction result according to the abnormality scoring result in the abnormality detection result when the abnormality detection result is abnormal, and determine the abnormality duration according to the time information of the abnormal state data in the abnormality detection result;

[0014] An exception handling solution determination module is used to determine whether exception handling is required based on the exception duration. If exception handling is required, a target exception handling solution is determined based on the exception level prediction result and the mapping relationship between the preset exception level and the exception handling solution.

[0015] The exception handling module is used to perform exception handling on the target power equipment according to the target exception handling scheme.

[0016] In a third aspect, an embodiment of the present invention further provides a computer device, the computer device comprising:

[0017] one or more processors;

[0018] A memory for storing one or more programs;

[0019] When the one or more programs are executed by one or more processors, the one or more processors implement the device exception handling method provided by any embodiment of the present invention.

[0020] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a device exception handling method as provided in any embodiment of the present invention.

[0021] In a fifth aspect, an embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements a device exception handling method as provided in any embodiment of the present invention.

[0022] The embodiments of the above invention have the following advantages or beneficial effects:

[0023] In the embodiment of the present invention, by acquiring state sequence data to be analyzed that describes the operation state and running state of the target power equipment; inputting the state sequence data to be analyzed into the pre-trained anomaly detection model to obtain an anomaly detection result, wherein the pre-trained anomaly detection model is trained according to preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment in the normal operation state; when the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the abnormal duration is determined according to the time information of the abnormal state data in the anomaly detection result; whether an abnormality treatment is required is determined according to the abnormal duration, and when an abnormality treatment is required, the target abnormality treatment scheme is determined according to the abnormality level prediction result and the mapping relationship between the preset abnormality level and the abnormal treatment scheme; and the target power equipment is subjected to abnormal treatment according to the target abnormal treatment scheme. The technical solution of the embodiment of the present invention solves the problem that the abnormal condition of the power equipment cannot be handled automatically and timely at present, and can fully utilize the characteristics of the preset reference state sequence data to determine the abnormal condition of the power equipment and the corresponding abnormal treatment scheme through machine learning, so that the abnormal treatment measures are more flexible and effective, and the stable operation of the power equipment is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart of a device exception handling method provided by an embodiment of the present invention;

[0025] Figure 2 is a flow chart of a device exception handling method provided by an embodiment of the present invention;

[0026] Figure 3 is a schematic diagram of a prediction process of an anomaly detection model provided by an embodiment of the present invention;

[0027] Figure 4 It is a structural schematic diagram of a device exception handling apparatus provided by an embodiment of the present invention;

[0028] Figure 5 It is a structural schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0030] Figure 1The present invention provides a flowchart of a method for handling device exceptions. This embodiment is applicable to scenarios of power device exception handling. The method can be executed by a device exception handling device, which can be implemented by software and / or hardware and integrated into a computer device with application development capabilities.

[0031] like Figure 1 As shown, the device exception handling method of this embodiment includes the following steps:

[0032] S110 , obtaining state sequence data to be analyzed that describes the operation state and running state of the target power equipment.

[0033] The state sequence data to be analyzed may be sequence data arranged in chronological order. The state sequence data to be analyzed describing the operating state of the target power equipment may include voltage, current, power, temperature, etc., and the state sequence data to be analyzed describing the operating state of the target power equipment may include equipment start and stop data, state change data, and fault alarm data, etc.

[0034] The sequence data to be analyzed can be obtained by processing the collected raw data describing the operating status and running status of the target power equipment. In order to ensure the timeliness and accuracy of the data, the collection frequency of the raw data can be dynamically adjusted according to the equipment type and operating environment. Different types of raw data can adopt different collection frequencies. For example, for high-frequency raw data, such as current and voltage fluctuations, it can be collected once per second; for low-frequency raw data such as equipment start and stop, fault logs, etc., it can be collected every minute or longer.

[0035] The collected raw data can be transmitted to the edge platform or cloud platform for processing through the data transmission module based on communication protocols such as MQTT, HTTPS, etc. Considering that power equipment is usually widely distributed and there may be network delays or instability, the data transmission module must have breakpoint resumption and data caching functions to ensure that data is not lost.

[0036] After the raw data is transmitted to the edge or cloud platform, the platform receives, caches and stores the collected data. In order to efficiently manage and process massive data, the platform can use distributed storage systems such as Hadoop and cloud storage services for data storage, and use data stream processing technologies such as Apache Kafka and Apache Flink to process data streams in real time.

[0037] Time synchronization is performed on the data describing the running status and operation status of the equipment in the original data to ensure the consistency of timestamps between different devices and different data sources. A unified time synchronization protocol (such as NTP, PTP, etc.) is used to ensure the accuracy of the timestamps of each device. The synchronized data will be integrated into a multidimensional data model for subsequent analysis and modeling.

[0038] The data of all target power equipment are sorted and uniformly formatted by timestamp, and the data describing the running status and operating status of the equipment are associated to ensure that the data describing the running status and operating status of the same target power equipment at the same time point can correspond one to one to obtain the sequence data to be analyzed.

[0039] A multidimensional data model can also be generated based on the sequence data to be analyzed, where each dimension may include device status, device operating parameters, operation logs, environmental information, etc. This data model provides sufficient information support for subsequent machine learning models.

[0040] S120: Input the state sequence data to be analyzed into the pre-trained anomaly detection model to obtain anomaly detection results.

[0041] Among them, the pre-trained anomaly detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment under normal operating state.

[0042] The historical state sequence data of the target power equipment operation are collected in advance, and the state sequence data of the target power equipment in normal operation is obtained by screening the historical state sequence data. The anomaly detection model learns the characteristics of the preset reference state sequence data during the training phase, and can perform model prediction analysis on the input state sequence data to be analyzed based on the characteristics of the preset reference state sequence data to obtain anomaly detection results.

[0043] S130: When the abnormality detection result is abnormal, determine the abnormality level prediction result according to the abnormality scoring result in the abnormality detection result, and determine the abnormality duration according to the time information of the abnormal state data in the abnormality detection result.

[0044] The higher the anomaly score in the anomaly detection result, the more serious the anomaly. A mapping relationship between the anomaly score and the anomaly level can be preset, and the anomaly level prediction result can be determined based on the mapping relationship. The time information of the abnormal state data can be timestamp information. The duration of the anomaly can be determined based on the current time and the time information of the abnormal state data.

[0045] S140: Determine whether exception handling is required based on the exception duration. If exception handling is required, determine a target exception handling solution based on the exception level prediction result and a preset mapping relationship between the exception level and the exception handling solution.

[0046] Determine whether exception handling is required based on the duration of the exception, for example, by comparing the duration of the exception with a preset time threshold. The mapping relationship between the exception level and the exception handling solution can be determined in advance based on historical exception handling events. In the preset mapping relationship between the exception level and the exception handling solution, determine the exception handling solution corresponding to the exception level prediction result as the target exception handling solution.

[0047] The exception handling scheme may include a rollback strategy, etc. For example, the exception levels may be classified into minor, medium, and severe levels.

[0048] The exception handling solution for the minor level can be automatic system recovery, such as restarting, adjusting the working mode, or postponing the exception handling and continuing to monitor the device status. At this time, the system may execute a fast self-recovery process and adjust the device configuration, such as adjusting the load, restoring the current and voltage, to avoid affecting the normal operation of the device.

[0049] The corresponding abnormal handling plan for the medium level can be to first perform a partial rollback operation, such as temporarily switching to a backup device or restoring to the last stable state to reduce the impact of the equipment on the power system. On this basis, the system can notify the operation and maintenance personnel to conduct inspections and take appropriate repair measures.

[0050] The abnormal handling solution corresponding to the severity level can be to immediately perform a full rollback operation. The full rollback operation will restore the device to its previous normal operating state, which may include rolling back to the backup version of the device, restoring the hardware configuration, restarting the device, etc., to ensure the stable operation of the power system. Severity level abnormalities can include hardware failures and electrical short circuits.

[0051] S150: Perform abnormal processing on the target power equipment according to the target abnormal processing solution.

[0052] Execute the target abnormality handling scheme to handle the abnormality of the target power equipment. After executing the target abnormality handling scheme, the system restores the stable state of the equipment and performs hardware status, log and communication status checks to ensure that the equipment is fully restored to normal operation. The hardware status check can include self-checking the hardware of the equipment, including electrical status, temperature, operating load, etc., to ensure that there is no hardware damage or abnormality.

[0053] Log verification can include automatically checking the device's operation log, fault log, etc., to confirm whether the device's operation log is normal after the rollback, and compare it with the historical normal operation log and fault log to confirm whether the rollback is effective.

[0054] Ensure that the device's communication system is restored to normal, synchronize data with the cloud platform or edge platform in a timely manner, and ensure that the system continuously monitors the device status. If any check fails, the system will trigger an emergency recovery process or notify the operation and maintenance personnel to handle it on site.

[0055] Centrally manage the target exception handling plan and the corresponding state sequence data to be analyzed for operation and maintenance personnel to conduct root cause analysis. By analyzing the state sequence data to be analyzed, the root cause of the problem can be discovered and the equipment configuration can be optimized. Based on the state sequence data after the execution of the target exception handling plan, the anomaly detection model, the relevant thresholds in the anomaly detection model and the exception handling plan can be continuously optimized.

[0056] The optimization of the rollback strategy in the exception handling solution can be based on the rollback data and the fault type. For example, if it is found during the rollback process that some operations, such as restarting the device, have a poor effect on device recovery, the system can optimize the rollback strategy, such as adding a backup device switch, or adjusting the hardware recovery order.

[0057] The optimization of the rollback strategy can be achieved through the following rules:

[0058] Hardware recovery priority rule: For hardware-related failures, first try to restore the hardware device, and then perform software restart and other operations. Backup device priority rule: For minor failures, first try to restore system stability by switching to the backup device, and then perform device recovery operations.

[0059] For the anomaly detection model, the machine learning model is continuously optimized based on the state sequence data after the execution of the target anomaly handling solution. For example, the parameters of the machine learning model are continuously updated through incremental learning, transfer learning and other methods.

[0060] The technical solution of this embodiment obtains the state sequence data to be analyzed that describes the operating state and running state of the target power equipment; inputs the state sequence data to be analyzed into the pre-trained anomaly detection model to obtain the anomaly detection result, wherein the pre-trained anomaly detection model is trained according to the preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment under normal operating state; when the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the anomaly duration is determined according to the time information of the abnormal state data in the anomaly detection result; whether anomaly processing is required is determined according to the anomaly duration, and when anomaly processing is required, the target anomaly processing scheme is determined according to the anomaly level prediction result and the mapping relationship between the preset anomaly level and the anomaly processing scheme; and the target power equipment is processed according to the target anomaly processing scheme. The technical solution of the embodiment of the present invention solves the problem that the abnormal conditions of power equipment cannot be processed in time and automatically at present, and can fully utilize the characteristics of the preset reference state sequence data to determine the abnormal conditions of power equipment and the corresponding anomaly processing scheme through machine learning, so that the anomaly processing measures are more flexible and effective, and the stable operation of power equipment is guaranteed.

[0061] Figure 2 This is a flowchart of a device exception handling method provided by an embodiment of the present invention. This embodiment and the device exception handling method in the above embodiment belong to the same inventive concept, and further describes the process of determining the exception detection result. The method can be executed by a device exception handling device, which can be implemented by software and / or hardware and integrated into a computer device with application development function.

[0062] like Figure 2 As shown, the device exception handling method of this embodiment includes the following steps:

[0063] S210: Acquire state sequence data to be analyzed that describes the operation state and running state of the target power equipment.

[0064] S220, inputting the state sequence data to be analyzed into the temporal convolutional network of the pre-trained anomaly detection model to obtain the time series features of the state sequence data to be analyzed.

[0065] like Figure 3 As shown in the figure, the state sequence data to be analyzed is input into the temporal convolutional network of the pre-trained anomaly detection model. Each layer of the temporal convolutional network adopts one-dimensional convolution to capture the local features in the input data through the convolution kernel. In order to prevent information leakage, causal convolution is adopted to ensure that the convolution operation does not consider future information. In the process of generating the temporal features of the state sequence data to be analyzed, the output features of each layer can be expressed by the following formula:

[0066]

[0067] Where: represents the output of the i-th layer at time step t; represents the output of the previous layer at time step t-1, which is the output of the residual connection; W i is the weight of the convolution kernel of the i-th layer. After each convolution operation, the input will be added to the output of the previous layer to form a residual connection. After stacking multiple convolution layers, a time series feature representation is finally obtained. It is the timing characteristics of the state sequence data to be analyzed, and also the timing information representing the state of the device.

[0068] S230, inputting the time series features into the self-attention mechanism module of the pre-trained anomaly detection model to obtain a self-attention output result.

[0069] The self-attention mechanism module can help the anomaly detection model understand the important time periods of device state changes and give it higher attention. The self-attention mechanism module can generate a self-attention matrix group through a fully connected layer, determine the attention weight coefficient based on the self-attention matrix group, and obtain the self-attention output result.

[0070] In an optional embodiment, the timing features are input into the self-attention mechanism module of a pre-trained anomaly detection model to obtain a self-attention output result. The timing features can be input into the self-attention mechanism module of the pre-trained anomaly detection model, and the timing features are converted into a self-attention matrix group through the fully connected layer of the self-attention mechanism module, wherein the self-attention matrix group includes a search matrix, a key matrix and a value matrix; the attention weight coefficient is determined according to the search matrix and the key matrix, and the self-attention output result is obtained by performing a weighted operation on the attention weight coefficient and the value matrix.

[0071] For example, the temporal features output by the temporal convolutional network are converted into a self-attention matrix group through a fully connected layer. The self-attention matrix group includes Q, K, and V matrices, which are search matrix, key matrix, and value matrix respectively. The calculation process of search matrix, key matrix, and value matrix can be expressed by the following formula:

[0072]

[0073] Where: W q , W k , W v represents the learnable weight matrix; b q , b k , b v represents the bias term; Q, K, and V are Query, Key, and Value vectors, respectively.

[0074] The attention weight coefficient is calculated according to the search matrix and key matrix of the self-attention matrix group. The goal of the self-attention mechanism is to calculate the similarity between each time step and obtain the attention weight coefficient, which is used to weight the Value vector to form the final self-attention output. The calculation process of the self-attention output result can be expressed by the following formula:

[0075]

[0076] Where: K T represents the transpose of the K matrix; d k represents the dimension of the Key vector in the K matrix; Softmax(·) represents the attention weight coefficient obtained by normalization calculation, which is used to calculate the correlation between each time step. In this embodiment, an attention weight coefficient is assigned to the Value of each time step, so that the model pays attention to the time step that is critical to anomaly detection. The calculated attention output As the self-attention output result, it is used for subsequent anomaly detection and anomaly classification.

[0077] S240, inputting the state sequence data to be analyzed and the self-attention output result into the anomaly detection module of the pre-trained anomaly detection model, and performing an anomaly analysis on the state sequence data to be analyzed and the self-attention output result based on preset reference features through the anomaly detection module to obtain an anomaly detection result of the state sequence data to be analyzed.

[0078] The pre-trained anomaly detection model is trained based on preset reference state sequence data, where the preset reference state sequence data is state sequence data of the target power equipment in a normal operating state. The preset reference feature is a feature representation corresponding to the preset reference state sequence data.

[0079] The anomaly detection module learns the high-dimensional feature representation of the preset reference state sequence data, namely the preset reference feature, during the training phase. In the anomaly detection module, the state sequence data to be analyzed and the self-attention output results are analyzed according to the preset reference feature, such as calculating the reconstruction error and timing deviation, and obtaining the anomaly detection result of the state sequence data to be analyzed according to the reconstruction error and timing deviation.

[0080] In an optional implementation, the anomaly detection result of the state sequence data to be analyzed includes an anomaly scoring result, a data anomaly state, and an anomaly type. The state sequence data to be analyzed and the self-attention output result are input into an anomaly detection module of a pre-trained anomaly detection model. The anomaly detection module performs an anomaly analysis on the state sequence data to be analyzed and the self-attention output result based on preset reference features to obtain an anomaly detection result of the state sequence data to be analyzed, which may include the following steps A1-A3:

[0081] Step A1, input the state sequence data to be analyzed into the anomaly detection module of the pre-trained anomaly detection model, reconstruct the state sequence data to be analyzed based on preset reference features through the reconstruction error calculation module of the anomaly detection module, obtain the reconstructed data corresponding to the state sequence data to be analyzed, and calculate the reconstruction error corresponding to the state sequence data to be analyzed based on the state sequence data to be analyzed and the reconstructed data.

[0082] The state sequence data to be analyzed is reconstructed based on the preset reference features to obtain the reconstructed data. The state data of each time step in the state sequence data to be analyzed is compared with the reconstructed data to calculate the reconstruction error. The calculation process of the reconstruction error is as follows:

[0083]

[0084] Where: x t Represents the state data of the current time step; represents the reconstructed state data based on model reconstruction; ||·||2 represents the L2 norm.

[0085] Step A2: input the self-attention output result into the anomaly detection module of the pre-trained anomaly detection model, and predict the self-attention output result based on the preset reference feature through the timing deviation calculation module of the anomaly detection module to obtain the expected timing feature corresponding to the self-attention output result, and calculate the timing deviation corresponding to the state sequence data to be analyzed according to the self-attention output result and the expected timing feature.

[0086] Compare the self-attention output with the device's expected timing model, which is based on the preset reference features. The process of calculating the timing deviation is as follows:

[0087]

[0088] Where: f t Represents the self-attention feature of the current time step; Represents the expected time series characteristics predicted by the model.

[0089] Step A3: determine an abnormality scoring result according to the reconstruction error and the timing deviation, and determine the data abnormality state and abnormality type of the state sequence data to be analyzed according to the abnormality scoring result and the self-attention output result.

[0090] The reconstruction error and timing deviation are combined to obtain the final anomaly score result. The calculation process of the anomaly score result is as follows:

[0091] Anomaly Score(x t )=Reconstruction Error(x t)+λ·Temporal Bias(f t );

[0092] Where: λ represents the weight hyperparameter, which is used to adjust the relative importance of reconstruction error and timing deviation.

[0093] In an optional embodiment, the data abnormal state and abnormal type of the state sequence data to be analyzed are determined according to the abnormal scoring result and the self-attention output result. The abnormal scoring result can be compared with a preset abnormal judgment threshold. When the abnormal scoring result is greater than the preset abnormal judgment threshold, the data abnormal state is judged to be abnormal; when the data abnormal state is abnormal, the self-attention output result is predicted to be abnormal by the fully connected layer of the abnormal detection module, and the abnormal probability of each abnormal type corresponding to the self-attention output result is output; the abnormal probability with the highest probability value is determined as the target type probability, and the abnormal type corresponding to the target type probability is determined as the target abnormal type; the target type probability is compared with the preset abnormal type judgment threshold corresponding to the target abnormal type, and when the target type probability is greater than the preset abnormal type judgment threshold, the target abnormal type is determined to be the abnormal type of the state sequence data to be analyzed.

[0094] According to the preset reference state sequence data in the training set, a preset abnormality judgment threshold θ is pre-set to judge whether each time step is abnormal:

[0095]

[0096] When the anomaly score result Anomaly Score is greater than the preset anomaly judgment threshold θ, the data anomaly state is judged to be abnormal.

[0097] After detecting an anomaly, the type of anomaly is determined, and a multi-task learning framework is used to map the self-attention output results using a fully connected layer to output the predicted probability distribution of the anomaly type.

[0098] The predicted probability distribution of anomaly types can be expressed by the following formula:

[0099]

[0100] Where: y t represents the predicted probability distribution of abnormal types; W class represents the weight matrix of the fully connected layer; b class represents the bias term of the fully connected layer; Represents the self-attention output result;

[0101] According to the predicted abnormal type probability distribution, the abnormal probability with the highest probability value is determined as the target type probability, and the abnormal type corresponding to the target type probability is determined as the target abnormal type; the target type probability is compared with the preset abnormal type judgment threshold corresponding to the target abnormal type. When the target type probability is greater than the preset abnormal type judgment threshold, the target abnormal type is determined to be the abnormal type of the state sequence data to be analyzed.

[0102] S250: When the abnormality detection result is abnormal, determine the abnormality level prediction result according to the abnormality scoring result in the abnormality detection result, and determine the abnormality duration according to the time information of the abnormal state data in the abnormality detection result.

[0103] S260: Determine whether exception handling is required based on the exception duration. If exception handling is required, determine a target exception handling solution based on the exception level prediction result and a mapping relationship between preset exception levels and exception handling solutions.

[0104] In an optional implementation, determining whether exception handling is required based on the exception duration may be by comparing the exception duration with a preset exception handling execution time threshold, and determining that exception handling is required when the exception duration is greater than the preset exception handling execution time threshold.

[0105] The exception handling execution time threshold is used to avoid the impact of short-term interference and avoid unnecessary exception handling.

[0106] S270: Perform abnormal processing on the target power equipment according to the target abnormal processing solution.

[0107] When the anomaly detection result is abnormal, it is also possible to traverse whether the upstream device of the abnormal target power device output by the model has any abnormality. If there is an abnormality, the abnormal device is the affected device, and the detection result is output, the upgrade is interrupted, and the upgrade is carried out after the abnormal problem of the upstream device is resolved. If the abnormal device output by the model is the abnormal source, the target power equipment is handled abnormally.

[0108] Based on the state sequence data after the execution of the target exception handling solution, the model parameters, rollback strategy and anomaly detection model such as the preset anomaly judgment threshold are continuously optimized.

[0109] The optimization of the preset abnormal judgment threshold can be that the system will identify the optimal threshold of the target power equipment under different operating conditions by analyzing the fault data of the target power equipment. The fault data can be the state sequence data to be analyzed in the abnormal event associated with the target abnormal handling scheme. For example, according to the current change pattern of the equipment when the fault occurs, the current threshold of the equipment is adjusted to avoid overload. The temperature threshold of the equipment is optimized according to the trend of historical fault data to ensure that the equipment does not fail under extreme load.

[0110] The preset abnormal judgment threshold optimization formula is as follows:

[0111] θ new =θ old +α×(E threshold -θ old );

[0112] Where: θ new represents the optimized preset abnormality judgment threshold; θ old represents the original preset abnormal judgment threshold; α represents the learning rate, which controls the step size of the preset abnormal judgment threshold adjustment; E threshold It represents the expected preset abnormal judgment threshold, which is the optimal value obtained based on the analysis of the state sequence data to be analyzed.

[0113] The technical solution of this embodiment is to obtain state sequence data to be analyzed that describes the operating state and running state of the target power equipment; input the state sequence data to be analyzed into the temporal convolution network of the pre-trained anomaly detection model to obtain the time series features of the state sequence data to be analyzed; input the time series features into the self-attention mechanism module of the pre-trained anomaly detection model to obtain the self-attention output result; input the state sequence data to be analyzed and the self-attention output result into the anomaly detection module of the pre-trained anomaly detection model, and perform an anomaly analysis on the state sequence data to be analyzed and the self-attention output result based on the preset reference feature through the anomaly detection module to obtain the anomaly detection result of the state sequence data to be analyzed; the preset reference feature is the preset reference state sequence data. According to the corresponding feature representation; the pre-trained anomaly detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment under normal operating state; when the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the anomaly duration is determined according to the time information of the abnormal state data in the anomaly detection result; whether anomaly processing is required is determined according to the anomaly duration, and when anomaly processing is required, the target anomaly processing scheme is determined according to the anomaly level prediction result and the mapping relationship between the preset anomaly level and the anomaly processing scheme; the target power equipment is subjected to an anomaly processing according to the target anomaly processing scheme. The technical solution of the embodiment of the present invention solves the problem that the abnormal conditions of power equipment cannot be processed automatically and timely at present, and can fully utilize the characteristics of the preset reference state sequence data to determine the anomaly detection results and anomaly levels of power equipment through machine learning, and determine the corresponding anomaly processing scheme according to the anomaly level, so that anomaly detection is more accurate, and anomaly processing measures are more flexible and effective, thereby ensuring the stable operation of power equipment.

[0114] In a specific anomaly detection model training example of the present invention, the loss function of the model is as follows:

[0115] The data anomaly state determination and anomaly type classification of the joint training model are implemented. The loss function uses a joint loss function, including data anomaly state determination loss and classification loss.

[0116] The data anomaly state determination loss uses the binary cross entropy loss to measure the prediction error of the model. The data anomaly state determination loss can be:

[0117]

[0118] Where: y t Indicates the actual label, indicating whether the time step t is abnormal, 1 is normal, and 0 is abnormal; Represents the anomaly probability predicted by the model, which is calculated by the anomaly detection module;

[0119] The anomaly type classification loss is calculated by cross entropy loss, which is used to measure the difference between the anomaly type predicted by the model and the true label. The anomaly type classification loss can be:

[0120]

[0121] Where: y t,i Represents the true label of the i-th abnormal type; Represents the model's predicted probability for the i-th type of anomaly at time step t;

[0122] The total loss function can be:

[0123] L=L detect +L class ·L class ;

[0124] Where: class represents the weight hyperparameter.

[0125] Figure 4 The present invention provides a schematic diagram of a device abnormality handling device, which can be applied to power equipment abnormality handling scenarios. The device abnormality handling device can be implemented by software and / or hardware and integrated into a computer terminal device with application development function.

[0126] like Figure 4 As shown, the device exception handling apparatus includes: a state sequence data acquisition module 310 , an exception detection result determination module 320 , an exception level determination module 330 , an exception handling solution determination module 340 and an exception handling module 350 .

[0127] Among them, the state sequence data acquisition module 310 is used to obtain the state sequence data to be analyzed that describes the operating state and running state of the target power equipment; the abnormal detection result determination module 320 is used to input the state sequence data to be analyzed into the pre-trained abnormal detection model to obtain the abnormal detection result, wherein the pre-trained abnormal detection model is trained based on the preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment under normal operating conditions; the abnormal level determination module 330 is used to determine the abnormal level prediction result according to the abnormal scoring result in the abnormal detection result when the abnormal detection result is abnormal, and determine the abnormal duration according to the time information of the abnormal state data in the abnormal detection result; the abnormal handling scheme determination module 340 is used to determine whether abnormal handling is required according to the abnormal duration, and if abnormal handling is required, determine the target abnormal handling scheme according to the abnormal level prediction result and the mapping relationship between the preset abnormal level and the abnormal handling scheme; the abnormal handling module 350 is used to perform abnormal handling on the target power equipment according to the target abnormal handling scheme.

[0128] The technical solution of this embodiment obtains the state sequence data to be analyzed that describes the operating state and running state of the target power equipment; inputs the state sequence data to be analyzed into the pre-trained anomaly detection model to obtain the anomaly detection result, wherein the pre-trained anomaly detection model is trained according to the preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment under normal operating state; when the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the anomaly duration is determined according to the time information of the abnormal state data in the anomaly detection result; whether anomaly processing is required is determined according to the anomaly duration, and when anomaly processing is required, the target anomaly processing scheme is determined according to the anomaly level prediction result and the mapping relationship between the preset anomaly level and the anomaly processing scheme; and the target power equipment is processed according to the target anomaly processing scheme. The technical solution of the embodiment of the present invention solves the problem that the abnormal conditions of power equipment cannot be processed in time and automatically at present, and can fully utilize the characteristics of the preset reference state sequence data to determine the abnormal conditions of power equipment and the corresponding anomaly processing scheme through machine learning, so that the anomaly processing measures are more flexible and effective, and the stable operation of power equipment is guaranteed.

[0129] In an optional implementation, the abnormal detection result determination module 320 is specifically used to:

[0130] The state sequence data to be analyzed is input into the temporal convolutional network of the pre-trained anomaly detection model to obtain the time series features of the state sequence data to be analyzed; the time series features are input into the self-attention mechanism module of the pre-trained anomaly detection model to obtain the self-attention output results; the state sequence data to be analyzed and the self-attention output results are input into the anomaly detection module of the pre-trained anomaly detection model, and the anomaly detection module performs an anomaly analysis on the state sequence data to be analyzed and the self-attention output results based on preset reference features to obtain the anomaly detection results of the state sequence data to be analyzed, wherein the preset reference features are feature representations corresponding to the preset reference state sequence data.

[0131] In an optional implementation, the abnormal detection result determination module 320 is further configured to:

[0132] Input the state sequence data to be analyzed into the temporal convolutional network of the pre-trained anomaly detection model to obtain the time series features of the state sequence data to be analyzed;

[0133] Input the time series features into the self-attention mechanism module of the pre-trained anomaly detection model to obtain the self-attention output result;

[0134] The state sequence data to be analyzed and the self-attention output results are input into the anomaly detection module of the pre-trained anomaly detection model, and the anomaly detection module performs an anomaly analysis on the state sequence data to be analyzed and the self-attention output results based on preset reference features to obtain the anomaly detection results of the state sequence data to be analyzed, wherein the preset reference features are feature representations corresponding to the preset reference state sequence data.

[0135] In an optional implementation, the abnormal detection result determination module 320 is further configured to:

[0136] Input the time series features into the self-attention mechanism module of the pre-trained anomaly detection model, and convert the time series features into a self-attention matrix group through the fully connected layer of the self-attention mechanism module, where the self-attention matrix group includes a search matrix, a key matrix, and a value matrix;

[0137] The attention weight coefficient is determined according to the search matrix and the key matrix, and the self-attention output result is obtained by performing weighted operation according to the attention weight coefficient and the value matrix.

[0138] In an optional embodiment, the anomaly detection result of the state sequence data to be analyzed includes an anomaly scoring result, a data anomaly state and an anomaly type, and the anomaly detection result determination module 320 is further used to: input the state sequence data to be analyzed into the anomaly detection module of the pre-trained anomaly detection model, reconstruct the state sequence data to be analyzed based on preset reference features through the reconstruction error calculation module of the anomaly detection module, obtain reconstructed data corresponding to the state sequence data to be analyzed, and calculate the reconstruction error corresponding to the state sequence data to be analyzed according to the state sequence data to be analyzed and the reconstructed data; input the self-attention output result into the anomaly detection module of the pre-trained anomaly detection model, predict the self-attention output result based on the preset reference features through the timing deviation calculation module of the anomaly detection module, obtain the expected timing features corresponding to the self-attention output result, and calculate the timing deviation corresponding to the state sequence data to be analyzed according to the self-attention output result and the expected timing features; determine the anomaly scoring result according to the reconstruction error and the timing deviation, and determine the data anomaly state and anomaly type of the state sequence data to be analyzed according to the anomaly scoring result and the self-attention output result.

[0139] In an optional implementation, the abnormal detection result determination module 320 is further configured to:

[0140] The anomaly scoring result is compared with the preset anomaly judgment threshold. When the anomaly scoring result is greater than the preset anomaly judgment threshold, the data anomaly state is judged to be abnormal. When the data anomaly state is abnormal, the self-attention output result is predicted to be of abnormal type through the fully connected layer of the anomaly detection module, and the anomaly probability of each anomaly type corresponding to the self-attention output result is output. The anomaly probability with the highest probability value is determined as the target type probability, and the anomaly type corresponding to the target type probability is determined as the target anomaly type. The target type probability is compared with the preset anomaly type judgment threshold corresponding to the target anomaly type. When the target type probability is greater than the preset anomaly type judgment threshold, the target anomaly type is determined to be the anomaly type of the state sequence data to be analyzed.

[0141] In an optional implementation, the exception handling solution determination module 340 is specifically configured to:

[0142] The exception duration is compared with a preset exception handling execution time threshold, and when the exception duration is greater than the preset exception handling execution time threshold, it is determined that exception handling needs to be performed.

[0143] The device exception handling apparatus provided in the embodiment of the present invention can execute the device exception handling method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0144] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Figure 5 A block diagram of an exemplary computer device 12 suitable for use in implementing embodiments of the present invention is shown. Figure 5 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. The computer device 12 can be any terminal device with computing capabilities, such as an intelligent controller and server, a mobile phone and other terminal devices.

[0145] like Figure 5 As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0146] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0147] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0148] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, usually called a "hard drive"). Although Figure 5 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0149] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0150] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. Furthermore, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with the other modules of the computer device 12 via the bus 18. It should be understood that although Figure 5 Not shown, other hardware and / or software modules may be used in conjunction with computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RFID systems, tape drives, and data backup storage systems.

[0151] The processing unit 16 executes various functional applications and data processing by running the program stored in the system memory 28, for example, implementing the device exception handling method provided in the embodiment of the present invention, the method comprising:

[0152] Acquire state sequence data to be analyzed describing the operation state and running state of the target power equipment;

[0153] Inputting the state sequence data to be analyzed into a pre-trained anomaly detection model to obtain an anomaly detection result, wherein the pre-trained anomaly detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is state sequence data of the target power equipment in a normal operating state;

[0154] In the case where the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the anomaly duration is determined according to the time information of the abnormal state data in the anomaly detection result;

[0155] Determine whether exception handling is required based on the duration of the exception. If exception handling is required, determine a target exception handling solution based on the exception level prediction result and the mapping relationship between the preset exception level and the exception handling solution.

[0156] Perform abnormal processing on the target power equipment according to the target abnormal processing plan.

[0157] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a device exception handling method as provided in any embodiment of the present invention is implemented. The method includes:

[0158] Acquire state sequence data to be analyzed describing the operation state and running state of the target power equipment;

[0159] Inputting the state sequence data to be analyzed into a pre-trained anomaly detection model to obtain an anomaly detection result, wherein the pre-trained anomaly detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is state sequence data of the target power equipment in a normal operating state;

[0160] In the case where the anomaly detection result is abnormal, the anomaly level prediction result is determined according to the anomaly scoring result in the anomaly detection result, and the anomaly duration is determined according to the time information of the abnormal state data in the anomaly detection result;

[0161] Determine whether exception handling is required based on the duration of the exception. If exception handling is required, determine a target exception handling solution based on the exception level prediction result and the mapping relationship between the preset exception level and the exception handling solution.

[0162] Perform abnormal processing on the target power equipment according to the target abnormal processing plan.

[0163] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0164] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0165] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0166] Computer program code for performing the operation of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, Python, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0167] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements a device exception handling method as provided in any embodiment of the present application.

[0168] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, Python, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0169] It should be understood by those skilled in the art that the modules or steps of the present invention described above can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0170] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for handling device abnormalities, characterized in that: include: Acquire state sequence data to be analyzed describing the operation state and running state of the target power equipment; Inputting the state sequence data to be analyzed into a pre-trained anomaly detection model to obtain an anomaly detection result, wherein the pre-trained anomaly detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is state sequence data of the target power equipment in a normal operating state; In the case where the abnormality detection result is abnormal, determining the abnormality level prediction result according to the abnormality scoring result in the abnormality detection result, and determining the abnormality duration according to the time information of the abnormal state data in the abnormality detection result; Determining whether exception handling is required based on the exception duration, and if exception handling is required, determining a target exception handling solution based on the exception level prediction result and a preset mapping relationship between the exception level and the exception handling solution; The target power equipment is subjected to abnormal processing according to the target abnormal processing scheme.

2. The method according to claim 1, characterized in that The step of inputting the state sequence data to be analyzed into a pre-trained anomaly detection model to obtain an anomaly detection result includes: Inputting the state sequence data to be analyzed into a temporal convolutional network of a pre-trained anomaly detection model to obtain the temporal features of the state sequence data to be analyzed; Inputting the time series features into the self-attention mechanism module of the pre-trained anomaly detection model to obtain a self-attention output result; The state sequence data to be analyzed and the self-attention output result are input into the anomaly detection module of the pre-trained anomaly detection model, and the anomaly detection module performs an anomaly analysis on the state sequence data to be analyzed and the self-attention output result based on a preset reference feature to obtain an anomaly detection result of the state sequence data to be analyzed, wherein the preset reference feature is a feature representation corresponding to the preset reference state sequence data.

3. The method according to claim 2, characterized in that The step of inputting the time series feature into the self-attention mechanism module of the pre-trained anomaly detection model to obtain a self-attention output result includes: Inputting the time series features into the self-attention mechanism module of the pre-trained anomaly detection model, converting the time series features into a self-attention matrix group through the fully connected layer of the self-attention mechanism module, wherein the self-attention matrix group includes a search matrix, a key matrix, and a value matrix; The attention weight coefficient is determined according to the search matrix and the key matrix, and a weighted operation is performed according to the attention weight coefficient and the value matrix to obtain a self-attention output result.

4. The method according to claim 2, characterized in that: The anomaly detection result of the state sequence data to be analyzed includes an anomaly scoring result, a data anomaly state, and an anomaly type. The state sequence data to be analyzed and the self-attention output result are input into an anomaly detection module of the pre-trained anomaly detection model, and the state sequence data to be analyzed and the self-attention output result are analyzed by the anomaly detection module based on preset reference features to obtain an anomaly detection result of the state sequence data to be analyzed, including: Inputting the state sequence data to be analyzed into the anomaly detection module of the pre-trained anomaly detection model, reconstructing the state sequence data to be analyzed based on preset reference features through the reconstruction error calculation module of the anomaly detection module to obtain reconstructed data corresponding to the state sequence data to be analyzed, and calculating the reconstruction error corresponding to the state sequence data to be analyzed based on the state sequence data to be analyzed and the reconstructed data; Input the self-attention output result into the anomaly detection module of the pre-trained anomaly detection model, predict the self-attention output result based on the preset reference feature through the timing deviation calculation module of the anomaly detection module, obtain the expected timing feature corresponding to the self-attention output result, and calculate the timing deviation corresponding to the state sequence data to be analyzed according to the self-attention output result and the expected timing feature; An abnormality scoring result is determined according to the reconstruction error and the timing deviation, and a data abnormality state and an abnormality type of the state sequence data to be analyzed are determined according to the abnormality scoring result and the self-attention output result.

5. The method according to claim 4, characterized in that The determining the data abnormality state and abnormality type of the state sequence data to be analyzed according to the abnormality scoring result and the self-attention output result includes: Compare the abnormality scoring result with the preset abnormality judgment threshold, and when the abnormality scoring result is greater than the preset abnormality judgment threshold, judge the data abnormality state as abnormal; When the data abnormality state is abnormal, the self-attention output result is predicted to be an abnormal type through the fully connected layer of the abnormality detection module, and the abnormal probability of each abnormal type corresponding to the self-attention output result is output; Determine the abnormal probability with the highest probability value as the target type probability, and determine the abnormal type corresponding to the target type probability as the target abnormal type; The target type probability is compared with a preset abnormal type judgment threshold corresponding to the target abnormal type, and when the target type probability is greater than the preset abnormal type judgment threshold, the target abnormal type is determined to be the abnormal type of the state sequence data to be analyzed.

6. The method according to claim 1, characterized in that The determining whether to perform exception handling according to the exception duration includes: The exception duration is compared with a preset exception handling execution time threshold, and when the exception duration is greater than the preset exception handling execution time threshold, it is determined that exception handling needs to be performed.

7. A device for handling equipment abnormalities, characterized in that: include: A state sequence data acquisition module, used to acquire state sequence data to be analyzed that describes the operation state and running state of the target power equipment; An abnormality detection result determination module is used to input the state sequence data to be analyzed into a pre-trained abnormality detection model to obtain an abnormality detection result, wherein the pre-trained abnormality detection model is trained based on preset reference state sequence data, and the preset reference state sequence data is the state sequence data of the target power equipment in a normal operating state; an abnormality level determination module, used to determine an abnormality level prediction result according to an abnormality scoring result in the abnormality detection result when the abnormality detection result is abnormal, and determine an abnormality duration according to time information of abnormal state data in the abnormality detection result; An exception handling solution determination module is used to determine whether exception handling is required according to the exception duration, and if exception handling is required, determine a target exception handling solution according to the exception level prediction result and the mapping relationship between the preset exception level and the exception handling solution; The exception handling module is used to perform exception handling on the target power equipment according to the target exception handling solution.

8. A computer device, characterized in that: The computer device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the device exception handling method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the device exception handling method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the device exception handling method as described in any one of claims 1 to 6.

Images