Deterioration analysis method and system for generator sets in hydropower stations

Abstract

The embodiments of the present invention relate to the field of data analysis technology, and specifically, to a method and system for analyzing the degradation of a hydropower station generator set. The degradation state analysis network is trained by using historical degradation state monitoring data, and the network neuron weights are locked to ensure the stability and accuracy of the model. Furthermore, by inputting the target state monitoring data and target operating state element representation array of the non-degraded hydropower station generator set, potential degradation state monitoring data can be predicted, thereby providing timely early warning information for the operation and maintenance team. In this way, not only the monitoring level of the operating state of the hydropower station generator set is improved, but also strong technical support is provided for preventing equipment degradation and failure, significantly improving the safety and operation efficiency of the hydropower station.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a method and system for analyzing the deterioration of a generator set in a hydropower station. Background Art

[0002] In the operation and maintenance of hydropower station generators, real-time monitoring of their operating status and prediction of potential deterioration are crucial. Traditional monitoring methods often rely on regular manual inspections and simple threshold alarm systems. Although these methods can detect problems to a certain extent, they often have lags and are difficult to accurately predict and identify the deterioration trend of generators in a timely manner. Summary of the invention

[0003] In order to at least overcome the above-mentioned deficiencies in the prior art, one of the objects of the present invention is to provide a method and system for analyzing the deterioration of a generator set in a hydropower station.

[0004] An embodiment of the present invention provides a method for analyzing the deterioration of a generator set in a hydropower station, which is applied to a data analysis system. The method includes:

[0005] Obtaining a target operating state element representation array, wherein the target operating state element representation array is obtained by entering historical deterioration state monitoring data into a pre-debugged deterioration state analysis network, and updating the operating state element representation array after parameter reset on the basis of locking the neuron weights of the deterioration state analysis network, and the target operating state element representation array reflects the operating state monitoring of the deterioration hydropower station generator set thermal network;

[0006] The target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element representation array are entered into the deteriorated state analysis network to obtain the potential deteriorated state monitoring data corresponding to the target state monitoring data.

[0007] Optionally, before obtaining the target running state element representation array, the method further includes:

[0008] The operating state element representation array that has completed parameter reset and the historical deterioration state monitoring data are entered into the pre-debugged deterioration state analysis network, and the operating state element representation array is updated based on locking the neuron weights of the deterioration state analysis network.

[0009] Optionally, the inferior state analysis network includes a structural hidden danger identification branch and a device loss identification branch;

[0010] The step of inputting the operating state element characterization array and the historical deterioration state monitoring data of the completed parameter reset into the pre-debugged deterioration state analysis network includes:

[0011] Entering the historical degradation state monitoring data into the feature mining model of the structural hidden danger identification branch to map the historical degradation state monitoring data into a historical structural hidden danger state vector, wherein the historical structural hidden danger state vector is used to obtain a corresponding adversarial state vector through the adversarial branch processing of the device loss identification branch;

[0012] The anti-confrontation branch in the device loss identification branch performs disturbance cleaning on the confrontation state vector based on the running state element representation array after parameter reset, to obtain the result output by the anti-confrontation branch, wherein the running state element representation array after parameter reset is projected to the anti-confrontation branch by the mutual focus strategy;

[0013] The method further comprises:

[0014] Determine a first AI algorithm indicator based on a result output by the anti-adversarial branch;

[0015] Based on the first AI algorithm indicator, an array representing target operating status elements is obtained.

[0016] Optionally, the anti-adversarial branch includes X depth residual layers corresponding to X perturbation cleanings, where X is a positive integer; and the method further includes:

[0017] Determine a d-th state identification feature based on the running state element representation array of the completed parameter reset and the first configuration factor of the d-th depth residual layer, where d is a positive integer less than or equal to X;

[0018] Determine a d-th state attribute feature based on the running state element representation array of the completed parameter reset and the second configuration factor of the d-th depth residual layer;

[0019] Based on the d-th state identification feature and the d-th state attribute feature, the running state element representation array after completing parameter reset is projected to the d-th depth residual layer.

[0020] Optionally, determining the first AI algorithm indicator based on the result output by the anti-confrontation branch includes: determining the first AI algorithm indicator based on the result output by the anti-confrontation branch and a random disturbance feature;

[0021] The obtaining of a target operating state element representation array based on the first AI algorithm indicator includes: obtaining a target operating state element representation array by determining a minimum value of the first AI algorithm indicator.

[0022] Optionally, the operation state element characterization array for which parameter resetting is completed and the historical deterioration state monitoring data are input into a pre-debugged deterioration state analysis network, and the operation state element characterization array is updated on the basis of locking the neuron weights of the deterioration state analysis network, including:

[0023] Enter the historical deterioration state monitoring data and its corresponding state monitoring annotation data into the pre-debugged deterioration state analysis network;

[0024] On the basis of locking the neuron weights of the inferior state analysis network, determining a second AI algorithm indicator based on information in a key area of ​​the annotation set in the historical inferior state monitoring data;

[0025] Obtaining a target operating state element representation array based on the second AI algorithm indicator;

[0026] Among them, the key area of ​​the annotation set corresponds to the hidden danger data set in the historical degradation status monitoring data.

[0027] Optionally, the determining of the second AI algorithm indicator based on locking the neuron weights of the inferior state analysis network and based on information of a key area in the annotation set in the historical inferior state monitoring data includes:

[0028] Determining a comparison result of the random perturbation characteristics and the result output by the anti-adversarial branch in the inferior state analysis network;

[0029] Determining a second AI algorithm indicator based on the comparison result and information of a key area in the annotation set in the historical degradation state monitoring data;

[0030] Obtaining a target operating state element representation array based on the second AI algorithm indicator includes:

[0031] By determining the minimum value of the second AI algorithm indicator, an array representing the target operating state elements is obtained.

[0032] Optionally, the target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element characterization array are entered into the deteriorated state analysis network to obtain potential deteriorated state monitoring data corresponding to the target state monitoring data, including:

[0033] Entering the target state monitoring data of the non-deteriorated hydropower station generator set, the classification label corresponding to the target state monitoring data, and the target operating state element representation array into the deteriorated state analysis network to obtain the potential deteriorated state monitoring data corresponding to the target state monitoring data;

[0034] Among them, the potential degradation status monitoring data includes a hidden danger data set that matches the classification label.

[0035] Optionally, the target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element characterization array are entered into the deteriorated state analysis network to obtain potential deteriorated state monitoring data corresponding to the target state monitoring data, including:

[0036] Entering the target state monitoring data of the non-inferior hydropower station generator set into the inferior state analysis network, performing X-order feature confrontation on the target state monitoring data through the confrontation branch of the inferior state analysis network, and obtaining a confrontation state vector of the target state monitoring data, wherein the confrontation state vector includes a v-order confrontation state vector, X is a positive integer, and v+1 is a positive integer not greater than X;

[0037] The adversarial state vector of the target state monitoring data is subjected to X-order perturbation cleaning by mapping the anti-adversarial branch having the target running state element representation array; wherein the potential degraded state vector obtained by the v-order perturbation cleaning includes: a first subvector of the secondary label matching the classification label in the v-order input vector, and a second subvector feature of the secondary label matching the classification label in the perturbation cleaning vector corresponding to the v+1-order input vector;

[0038] The potential degradation state vector obtained after X-order disturbance cleaning is processed through the data generation branch of the structural hidden danger identification branch in the degradation state analysis network to obtain the potential degradation state monitoring data corresponding to the target state monitoring data.

[0039] An embodiment of the present invention also provides a data analysis system, comprising a processor and a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-mentioned hydropower station generator set degradation analysis method.

[0040] The embodiment of the present invention further provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, the above-mentioned method for analyzing the degradation of a generator set in a hydropower station is implemented.

[0041] Beneficial Effects

[0042] The embodiment of the present invention proposes a state monitoring and early warning system based on machine learning technology. By using historical degradation state monitoring data to train the degradation state analysis network and locking the network neuron weights, the stability and accuracy of the model are ensured. Furthermore, by inputting the target state monitoring data and target operating state element representation array of the non-degraded hydropower station generator set, potential degradation state monitoring data can be predicted, thereby providing timely early warning information for the operation and maintenance team. In this way, not only the monitoring level of the operating status of the hydropower station generator set is improved, but also strong technical support is provided for preventing equipment degradation and failure, significantly improving the safety and operation efficiency of the hydropower station. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 The present invention provides a flowchart of a method for analyzing the deterioration of a generator set in a hydropower station according to an embodiment of the present invention.

[0045] Figure 2 A block diagram of a data analysis system provided by an embodiment of the present invention.

[0046] icon:

[0047] 100-Data analysis system;

[0048] 101 - processor; 102 - memory; 103 - bus. DETAILED DESCRIPTION

[0049] The exemplary embodiments disclosed in the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0050] In order to better understand the above technical scheme, the technical scheme of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical scheme of the present invention, rather than limitations on the technical scheme of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0051] Figure 1 This is a flow chart of a method for analyzing the deterioration of a generator set in a hydropower station according to an embodiment of the present invention, which is applied to a data analysis system and includes steps 110 to 120.

[0052] Step 110, obtaining a target operating state element characterization array, wherein the target operating state element characterization array is obtained by entering historical degradation state monitoring data into a pre-debugged degradation state analysis network, and updating the operating state element characterization array after parameter reset on the basis of locking the neuron weights of the degradation state analysis network, and the target operating state element characterization array reflects the operating state monitoring thermal network of the degradation hydropower station generator set.

[0053] In the embodiment of the present invention, the target operating state element characterization array is a specific data set used to accurately describe the operating state of the hydropower station generator set at a specific time point or time period. This array contains multiple key parameters, such as temperature, pressure, speed, vibration frequency, etc. These parameters are important factors in characterizing the operating state of the generator set. By monitoring and analyzing these parameters, the real-time working state of the generator set can be understood and abnormal conditions can be discovered in time.

[0054] Historical degradation status monitoring data These data are the operating data of generator sets in degradation status recorded in the past. These data include parameters collected by various sensors, such as abnormal temperature, pressure fluctuation, excessive vibration, etc., which reflect the status of the generator set when performance degradation or failure occurs. These data are crucial for building and training the degradation status analysis network because they provide actual cases and patterns of generator set degradation.

[0055] The pre-debugging deterioration state analysis network is a network model built based on a machine learning algorithm. It has undergone preliminary training and debugging and can identify and process the state data of the generator set. By learning and analyzing historical deterioration state monitoring data, this network can establish a complex relationship model between the deterioration state of the generator set and various operating parameters. Once new state data is input, this network can predict whether the generator set is at risk of deterioration.

[0056] In the degradation state analysis network, the connection weights between neurons represent the importance of different input parameters on the network output. These weights are gradually adjusted and determined by the optimization algorithm during the network training process to ensure that the network's prediction results match the actual degradation state. Locking the neuron weights means that these weights will remain unchanged during the use of the network to ensure the stability and reliability of the network model.

[0057] The parameter reset completed operating status element characterization array refers to a new array after the parameters are reset on the basis of the operating status element characterization array. Parameter reset may include updating the values ​​of certain parameters, adding new parameters, or deleting parameters that are no longer needed. This new array will more accurately reflect the current operating status of the generator set and provide a more accurate data basis for subsequent deterioration status analysis.

[0058] Degraded hydropower station generator sets refer to those with degraded or faulty performance. Degraded performance may manifest as reduced efficiency, increased energy consumption, increased noise and vibration, etc. Timely detection and treatment of degraded status is crucial to ensure the safe and stable operation of hydropower stations.

[0059] The operation status monitoring thermal network is a visual monitoring network used to display the operation status of the hydropower station generator set in real time. Through the form of thermal maps, key parameters such as temperature distribution and pressure changes in various parts of the generator set can be intuitively seen. This thermal network helps staff quickly identify abnormal areas and potential problems, so that they can take corresponding maintenance measures.

[0060] In detail, in step 110, the data analysis system first obtains a large amount of historical degradation state monitoring data, which is the basis for building a degradation state analysis network. These data record in detail the various operating parameters of the generator set in the degradation state, such as temperature, pressure, speed, etc., and the changes of these parameters over time. Next, the data analysis system enters these historical degradation state monitoring data into a pre-debugging degradation state analysis network. This network has been trained and debugged a lot before, and has the ability to identify and process the generator set state data. By entering historical data into the network, the system can further optimize the performance and accuracy of the network. After entering the data, the data analysis system locks the neuron weights of the degradation state analysis network. This means that the connection weights between neurons will remain unchanged during the network processing data. This is done to ensure the stability and reliability of the network model and avoid deviations in the prediction results due to changes in weights. Then, the data analysis system updates the array of operating state element representations that completes parameter reset. This array originally contains various key parameters of the generator set operating state, but it may need to reset or adjust the parameters due to some reasons (such as equipment updates, environmental changes, etc.). By updating this array, the system can more accurately reflect the current operating status of the generator set. Finally, after processing and analyzing the above steps, the data analysis system obtains an array of target operating status element representations. This array not only contains the current operating parameters of the generator set, but also reflects the operating status of the generator set of the degraded hydropower station and the situation of the thermal network. Through this array, key information such as temperature distribution and pressure changes in various parts of the generator set can be intuitively understood, potential deterioration risks can be discovered in time, and corresponding maintenance measures can be taken.

[0061] Step 120: Enter the target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element representation array into the deteriorated state analysis network to obtain potential deteriorated state monitoring data corresponding to the target state monitoring data.

[0062] In the embodiment of the present invention, non-deteriorated hydropower station generator sets refer to those generator sets that are currently in normal working condition and have not yet experienced performance degradation or failure. These units show stable performance in daily operation, and all operating parameters are within the normal range, without obvious signs of degradation. However, even if the unit is currently in good condition, continuous status monitoring is required to promptly detect and prevent potential degradation problems.

[0063] Target status monitoring data refers to the operating status data collected in real time from the generator sets of non-inferior hydropower stations. These data include key operating parameters such as temperature, pressure, vibration, and speed of the unit, which can reflect the working status of the unit at a certain moment. By continuously monitoring these data, the health status of the unit can be evaluated, abnormal conditions can be discovered in time, and corresponding preventive and maintenance measures can be taken.

[0064] Potential degradation status monitoring data refers to the possible degradation status data of the unit predicted by analyzing the target status monitoring data and combining the historical degradation status data and the operating status element characterization array. These data reveal the performance degradation or failure trend that may occur in the generator set in the future, provide early warning and decision support for operation and maintenance personnel, and help to discover and deal with potential degradation problems in advance.

[0065] In detail, step 120 is a key step in the data analysis system, which involves entering the target state monitoring data and the target operating state element representation array of the non-inferior hydropower station generator set into the inferior state analysis network to obtain the potential inferior state monitoring data corresponding to the target state monitoring data. First, the data analysis system obtains the target state monitoring data of the non-inferior hydropower station generator set. These data are collected in real time and reflect the current operating state of the generator set. These data include but are not limited to key parameters such as temperature readings, pressure values, vibration frequencies and speeds. Through precise sensors and data acquisition systems, these data are accurately captured and transmitted to the data analysis system. Next, the data analysis system enters the target state monitoring data and the previously acquired target operating state element representation array into the inferior state analysis network. This network has been trained and optimized to identify and process the operating state data of the generator set. After entering the data, the network will use its learned knowledge and patterns to perform in-depth analysis and processing of the current data. During the processing of the inferior state analysis network, it will compare the similarity between the target state monitoring data and the historical inferior state data, and at the same time combine the information in the target operating state element representation array to perform complex calculations and analysis. The above process aims to identify the signs or trends of degradation that may be hidden in the current data. Finally, the degradation state analysis network will output the potential degradation state monitoring data corresponding to the target state monitoring data. These data are predictions of the possible degradation state of the generator set in the future, including possible failure points, performance degradation trends, etc. These data are of great reference value to operation and maintenance personnel, helping them to promptly discover and deal with potential degradation problems and ensure the stable operation of the generator set.

[0066] In combination with the above content, in an exemplary complete application scenario, the data analysis system first performs step 110, that is, obtains the target operating state element representation array. The background of this step is that in order to effectively monitor the operating state of the hydropower station generator set and prevent potential deterioration, the data analysis system needs to first build a monitoring model that can reflect the operating state of the deteriorated hydropower station generator set. The data analysis system first collects a large amount of historical deterioration state monitoring data, which contains the operating parameters of the generator set under various deterioration conditions. These data are entered into a pre-debugged deterioration state analysis network, which is trained and optimized by machine learning methods so that it can identify the deterioration state. After locking the neuron weights of the deterioration state analysis network, the data analysis system resets the parameters of the operating state element representation array, which contains various key parameters of the generator set operating state. By continuously updating this array, the system can generate a target operating state element representation array, which can accurately reflect the operating state of the deteriorated hydropower station generator set to monitor the thermal network.

[0067] Next, the data analysis system executes step 120, that is, the target state monitoring data and the target operating state element representation array of the non-degraded hydropower station generator set are entered into the degraded state analysis network. The purpose of this step is to predict and identify whether the generator set currently operating normally has a potential degraded risk. The system inputs the real-time target state monitoring data, such as temperature, pressure, vibration and other parameters, together with the target operating state element representation array into the degraded state analysis network. The network will process and analyze the current data through the degraded mode it has learned, and output the corresponding potential degraded state monitoring data. These data can help the staff to timely discover the abnormal conditions of the generator set, such as abnormal wear of a certain component, excessive temperature, etc., so as to take corresponding maintenance measures to prevent the unit from entering the degraded state and ensure its stable operation. Throughout the process, the data analysis system uses machine learning technology and big data analysis to realize real-time monitoring of the operating status of the hydropower station generator set and early warning of potential degraded states, greatly improving the safety and operation efficiency of the hydropower station.

[0068] In summary, the embodiment of the present invention obtains the target operating state element characterization array, which is obtained by training the inferior state analysis network using historical inferior state monitoring data, and updating the operating state element characterization array after locking the network neuron weights. This method not only deepens the understanding of the operating state of the inferior hydropower station generator set, but also ensures the stability and accuracy of the model prediction by locking the neuron weights. In addition, by inputting the target state monitoring data and the target operating state element characterization array of the non-inferior hydropower station generator set into the network, the potential inferior state monitoring data can be predicted. This predictive capability enables the operation and maintenance team to take necessary preventive and maintenance measures before the equipment deteriorates significantly, thereby significantly reducing the equipment failure rate and improving the overall operating efficiency and safety of the hydropower station. Therefore, the implementation of the embodiment of the present invention can not only realize the real-time monitoring of the operating state of the hydropower station generator set, but also provide an effective early warning system, providing solid technical support for the long-term stable operation of the hydropower station.

[0069] In some optional embodiments, before obtaining the target operating status element representation array, the method further includes: entering the operating status element representation array that has completed parameter reset and historical degradation status monitoring data into a pre-debugged degradation status analysis network, and updating the operating status element representation array based on locking the neuron weights of the degradation status analysis network.

[0070] Based on this embodiment, before obtaining the target operating state element characterization array, the data analysis system will perform a series of operations to ensure the accuracy of the data and the reliability of the model. The above process begins by entering the operating state element characterization array that has completed parameter reset and the historical deterioration state monitoring data into the pre-debugged deterioration state analysis network.

[0071] The "operating state element characterization array that has completed parameter reset" in the embodiment of the present invention refers to an array whose parameters have been adjusted according to the latest situation, which can more accurately reflect the current operating state of the hydropower station generator set. The "historical deterioration state monitoring data" includes various data when the generator set has deteriorated in the past, which is crucial for training and optimizing the deterioration state analysis network.

[0072] The deterioration state analysis network is a machine learning-based model that can predict the deterioration state of the generator set by analyzing the input data. In the above process, "pre-commissioning" refers to a series of adjustments and tests performed on the network before it is officially used to ensure its performance is optimal.

[0073] After the data analysis system enters the parameter reset operating state element characterization array and historical degradation state monitoring data into the degradation state analysis network, the next key step is to lock the network's neuron weights. Neuron weights play a vital role in the network, as they determine how input data affects the network's output. Locking weights means that these weights will not be changed during training, which helps maintain the stability and consistency of the model.

[0074] After the weights are locked, the data analysis system starts to update the operating status element representation array. This update process is based on the analysis results of the operating status element representation array of the historical degradation state monitoring data and the parameter reset of the degradation state analysis network. In this way, the system can ensure that the operating status element representation array is always consistent with the actual operating status of the generator set.

[0075] After completing these steps, the data analysis system will obtain an optimized and updated array of target operating status factor representations. This array not only reflects the current operating status of the generator set, but also incorporates information on historical degradation status, allowing the system to more accurately predict and prevent potential degradation problems.

[0076] In summary, through this series of operations, the data analysis system can more effectively monitor and evaluate the operating status of the hydropower station's generator sets, and promptly detect and deal with potential deterioration problems. This not only improves the safety and operating efficiency of the hydropower station, but also reduces maintenance costs and downtime, providing strong technical support for the long-term stable operation of the hydropower station.

[0077] In some examples, the inferior state analysis network includes a structural hidden danger identification branch and a device loss identification branch; the operation state element characterization array after parameter reset and the historical deterioration state monitoring data are entered into the pre-debugged inferior state analysis network, including: entering the historical deterioration state monitoring data into the feature mining model of the structural hidden danger identification branch to map the historical deterioration state monitoring data into a historical structural hidden danger state vector, wherein the historical structural hidden danger state vector is used to obtain a corresponding adversarial state vector through the adversarial branch processing of the device loss identification branch; the anti-adversarial branch in the device loss identification branch performs perturbation cleaning on the adversarial state vector based on the operation state element characterization array after parameter reset to obtain the result output by the anti-adversarial branch, wherein the operation state element characterization array after parameter reset is projected to the anti-adversarial branch with the mutual focus strategy. And the method also includes: determining a first AI algorithm indicator based on the result output by the anti-adversarial branch; obtaining a target operation state element characterization array based on the first AI algorithm indicator.

[0078] Based on this example, the poor condition analysis network used by the data analysis system is a composite network, which includes two main branches: structural hidden danger identification branch and device loss identification branch. These two branches each undertake different tasks and work together to achieve a comprehensive analysis of the status of the hydropower station generator set.

[0079] First, the data analysis system inputs the historical degradation state monitoring data into the feature mining model of the structural hidden danger identification branch. This model is specially designed to map the complex historical degradation state monitoring data into a more concise and representative historical structural hidden danger state vector. This vector not only compresses the dimension of the original data, but also highlights the possible structural hidden danger information of the generator set.

[0080] Next, this historical structural hidden danger state vector will be transmitted to the adversarial branch of the device loss identification branch for processing. The role of the adversarial branch is to process the vector in a special way to generate the corresponding adversarial state vector. This adversarial state vector simulates the data performance of the generator set under device loss to some extent, which helps the system to more comprehensively understand the actual operating status of the generator set.

[0081] Then, the data analysis system uses the operating state element characterization array that has completed parameter reset to further clean and process the adversarial state vector. The above process is completed in the anti-adversarial branch of the device loss identification branch. It is particularly worth mentioning that this operating state element characterization array is projected to the anti-adversarial branch using the mutual focus strategy. This strategy ensures that the information in the array can be used more effectively, thereby improving the analysis accuracy of the entire system.

[0082] In the anti-adversarial branch, the system perturbs and cleans the adversarial state vector to remove noise and interference factors, and obtain purer and more accurate data results. The above process is crucial to improving the accuracy and reliability of subsequent analysis.

[0083] After completing the above steps, the data analysis system determines the first AI algorithm index based on the output of the anti-opposition branch. This index is a comprehensive evaluation of the operating status of the generator set by the system. It integrates information such as structural hazards and device losses, and can more comprehensively reflect the actual operating status of the generator set.

[0084] Finally, the system obtains the target operating status element representation array based on this first AI algorithm indicator. This array is the system's accurate description of the current operating status of the generator set, which will serve as the basic data for subsequent analysis and prediction.

[0085] Through this series of operations and analyses, the data analysis system can more comprehensively and accurately understand and evaluate the operating status of the hydropower station's generator sets. This not only helps to promptly discover and prevent potential deterioration problems, but also provides strong data support for subsequent maintenance and management work. Therefore, this technical solution significantly improves the safety and operating efficiency of the hydropower station, reduces maintenance costs and downtime, and provides a solid technical guarantee for the long-term stable operation of the hydropower station.

[0086] In some other embodiments, the anti-counterattack branch includes X deep residual layers corresponding to X disturbance cleanings, where X is a positive integer; the method also includes: determining the d-th state identification feature based on the running state element representation array of the completed parameter reset and the first configuration factor of the d-th deep residual layer, where d is a positive integer less than or equal to X; determining the d-th state attribute feature based on the running state element representation array of the completed parameter reset and the second configuration factor of the d-th deep residual layer; and projecting the running state element representation array of the completed parameter reset to the d-th deep residual layer based on the d-th state identification feature and the d-th state attribute feature.

[0087] In this embodiment, the anti-adversarial branch in the device loss identification branch of the data analysis system is designed to be more sophisticated and complex. This anti-adversarial branch consists of X deep residual layers, each of which is responsible for performing a perturbation cleaning to gradually remove interference factors in the adversarial state vector, and X is a positive integer representing the number of deep residual layers.

[0088] In the above process, the data analysis system uses the running state element characterization array that completes parameter reset and the specific configuration factors of each deep residual layer to determine the state identification feature and state attribute feature. For example, for the dth deep residual layer (d is a positive integer less than or equal to X), the system first determines the dth state identification feature based on the running state element characterization array that completes parameter reset and the first configuration factor of the dth deep residual layer. This state identification feature can be regarded as a "label" of the generator set in a certain state, which helps the system identify and classify different states.

[0089] Next, the system uses the same operating state element characterization array and the second configuration factor of the dth deep residual layer to determine the dth state attribute feature. Different from the state identification feature, the state attribute feature focuses more on describing the specific properties or characteristics of the generator set in a certain state, such as vibration frequency, temperature distribution, etc.

[0090] Once the d-th state identification feature and the d-th state attribute feature are determined, the data analysis system will use these two features to project the running state element representation array that completes the parameter reset to the d-th deep residual layer. The above process can be regarded as the conversion of data between different feature spaces, which helps the system to more accurately capture and analyze the state information of the generator set.

[0091] Through this series of operations, the data analysis system can clean and analyze the adversarial state vector layer by layer, thereby obtaining a purer and more accurate data representation. This not only improves the system's ability to perceive the status of the generator set, but also provides more reliable data support for subsequent fault warnings and preventive maintenance.

[0092] It can be seen that by introducing multiple deep residual layers and corresponding state identification and attribute feature extraction methods, the data analysis system can more accurately identify and evaluate the state of the hydropower station generator set. This technical solution not only improves the accuracy and sensitivity of state monitoring, but also provides a strong technical guarantee for timely detection and prevention of potential faults. Therefore, the beneficial effect of this embodiment is to enhance the data analysis system's comprehensive understanding and control of the generator set state, providing a more solid support for the safe and stable operation of the hydropower station.

[0093] In the next step, determining the first AI algorithm index based on the result of the anti-confrontation branch output includes: determining the first AI algorithm index based on the result of the anti-confrontation branch output and the random disturbance feature. Then obtaining the target operating state element representation array based on the first AI algorithm index includes: obtaining the target operating state element representation array by determining the minimum value of the first AI algorithm index.

[0094] Specifically, the data analysis system determines the first AI algorithm indicator based on the output of the anti-adversarial branch. The key to this step is that the system not only considers the direct output of the anti-adversarial branch, but also introduces a random perturbation feature. This random perturbation feature is often used as a regularization method in machine learning, which helps to enhance the generalization ability of the model, that is, the model's ability to predict unseen data.

[0095] For example, the data analysis system combines the results of the anti-adversarial branch output with the random disturbance characteristics, and calculates a value through a certain algorithm. This value is the first AI algorithm indicator. This indicator can be regarded as a comprehensive evaluation of the current data status by the system, which reflects the overall quality of the data after a series of processing.

[0096] Next, the system obtains the target operating state element representation array based on this first AI algorithm indicator. The "target operating state element representation array" in the embodiment of the present invention is the data representation that the system expects to obtain that best represents the current operating state of the generator set. In order to obtain this array, the system attempts to find the set of data that minimizes the first AI algorithm indicator. In other words, the system continuously adjusts and optimizes the data until a state is found that minimizes the first AI algorithm indicator in this state.

[0097] The above process is usually implemented through an optimization algorithm, such as gradient descent or its variants. The system continuously iterates and adjusts the data, and each step moves in the direction of reducing the first AI algorithm indicator until a local minimum or global minimum is reached.

[0098] Finally, when the system finds the data state that minimizes the first AI algorithm indicator, it outputs the data in this state as the target operating state element representation array. This array not only contains the current operating state information of the generator set, but also has been optimized by the system, so it can more accurately reflect the actual state of the generator set.

[0099] Through this technical solution, the data analysis system can more accurately identify and evaluate the status of the generator set of the hydropower station. The introduction of random disturbance characteristics and the minimization process of the first AI algorithm indicator enhance the system's precision and accuracy in data processing. This optimization method not only improves the sensitivity of condition monitoring, but also provides a more reliable data basis for subsequent fault warning and preventive maintenance. Overall, the beneficial effect of this implementation method is to improve the accuracy and reliability of the data analysis system's assessment of the status of the generator set, and provide a more solid guarantee for the safe and stable operation of the hydropower station.

[0100] In other embodiments, the operating state element characterization array for completing parameter reset and the historical degradation state monitoring data are entered into a pre-debugged degradation state analysis network, and the operating state element characterization array is updated based on locking the neuron weights of the degradation state analysis network, including: entering the historical degradation state monitoring data and its corresponding state monitoring annotation data into the pre-debugged degradation state analysis network; determining a second AI algorithm indicator based on information in a key area of ​​an annotation set in the historical degradation state monitoring data based on locking the neuron weights of the degradation state analysis network; obtaining a target operating state element characterization array based on the second AI algorithm indicator; wherein the key area of ​​the annotation set corresponds to a hidden danger data set in the historical degradation state monitoring data.

[0101] Based on this embodiment, the data analysis system adopts a more sophisticated method when processing the running state element characterization array and historical degradation state monitoring data after parameter reset. This method involves entering the historical degradation state monitoring data and its corresponding state monitoring annotation data into the pre-debugged degradation state analysis network. The state monitoring annotation data is an interpretation and annotation of the historical degradation state monitoring data, which provides key information and features about the data, helping the system to understand and analyze the data more accurately.

[0102] Based on locking the neuron weights of the deterioration state analysis network, the data analysis system further utilizes the information in the key areas of the annotation set in the historical deterioration state monitoring data. The "key areas of the annotation set" here refer to the areas specifically pointed out in the annotation data that correspond to the hidden danger data sets in the historical deterioration state monitoring data. The hidden danger data set is a data set containing potential problems or anomalies. The analysis of these data helps the system identify and prevent potential problems.

[0103] Based on the information of these key areas, the data analysis system determines a second AI algorithm indicator. This indicator is a quantitative representation of the hidden danger characteristics in the data, which can help the system locate and evaluate potential problems more accurately. Through this indicator, the system can understand the gap between the current data status and the ideal state, and make corresponding adjustments and optimizations.

[0104] Subsequently, the data analysis system obtains the target operating status element representation array based on the second AI algorithm indicator. This array is the data representation obtained after the system has been optimized, and it can more accurately reflect the actual operating status of the generator set. The system continuously adjusts and optimizes the data to optimize the second AI algorithm indicator, thereby obtaining the target operating status element representation array.

[0105] In this way, hidden dangers and problems in the status of generator sets can be located and evaluated more accurately. By combining the information of condition monitoring annotation data and hidden danger data sets, the system can more comprehensively understand the characteristics and potential problems of the data, thereby making more accurate predictions and decisions. This not only improves the sensitivity and accuracy of condition monitoring, but also provides more reliable data support for subsequent fault warnings and preventive maintenance. In this way, the accuracy and comprehensiveness of the data analysis system's evaluation of the status of generator sets are improved, providing a more solid guarantee for the safe and stable operation of hydropower stations.

[0106] In some preferred design ideas, based on locking the neuron weights of the inferior state analysis network, the second AI algorithm index is determined based on the information in the key area of ​​the annotation set in the historical degradation state monitoring data, including: determining the comparison result of the random disturbance feature and the result of the anti-confrontation branch output in the inferior state analysis network; determining the second AI algorithm index based on the comparison result and the information in the key area of ​​the annotation set in the historical degradation state monitoring data. Then, obtaining the target operating state element representation array based on the second AI algorithm index includes: obtaining the target operating state element representation array by determining the minimum value of the second AI algorithm index.

[0107] Based on this design idea, the data analysis system pays special attention to the key area information in the annotation set when processing historical degradation status monitoring data. This information is crucial for accurately evaluating the operating status of the generator set. In order to make full use of this information, the system first locks the neuron weights of the degradation status analysis network to ensure the stability of the network during the processing.

[0108] Next, the data analysis system determines the comparison between the random perturbation feature and the output of the anti-adversarial branch in the inferior state analysis network. The random perturbation feature here is a means of testing the robustness of the network by adding small random perturbations to the network input and observing the changes in the network output to evaluate the stability of the network. The anti-adversarial branch is a special part of the network, which is designed to enhance the network's resistance to potential interference.

[0109] The system compares the random perturbation characteristics and the output of the anti-adversarial branch to obtain a comparison result. This comparison result reflects the stability of the network in the face of small perturbations. If this comparison result shows that the network's resistance to perturbations is weak, the system will adjust the calculation method of the second AI algorithm indicator accordingly.

[0110] The data analysis system then determines the second AI algorithm indicator based on this comparison result and the information in the key areas of the annotation set in the historical degradation status monitoring data. This indicator not only takes into account the stability performance of the network, but also incorporates the information of the key areas, allowing the system to more comprehensively evaluate the operating status of the generator set.

[0111] In order to obtain the target operating state element representation array, the data analysis system attempts to determine the minimum value of the second AI algorithm indicator. By continuously adjusting and optimizing the data input, the system can find a data state that minimizes the second AI algorithm indicator. The data in this state is the target operating state element representation array pursued by the system, because it can most accurately reflect the actual operating state of the generator set.

[0112] In this way, various factors can be considered more comprehensively, resulting in more accurate evaluation results. By combining the random disturbance characteristics, the output of the anti-opposition branch, and the information of the key areas, the system can more accurately identify and evaluate the status of the generator set. This optimization method not only improves the accuracy and stability of condition monitoring, but also provides a more reliable data basis for subsequent fault warning and preventive maintenance. In this way, the accuracy and comprehensiveness of the data analysis system's evaluation of the generator set status are improved, further ensuring the safe and stable operation of the hydropower station.

[0113] Under some other preferred design ideas, the target state monitoring data of the non-inferior hydropower station generator set and the target operating state element representation array are entered into the inferior state analysis network to obtain the potential inferior state monitoring data corresponding to the target state monitoring data, including: entering the target state monitoring data of the non-inferior hydropower station generator set, the classification label corresponding to the target state monitoring data and the target operating state element representation array into the inferior state analysis network to obtain the potential inferior state monitoring data corresponding to the target state monitoring data; wherein, the hidden danger data set in the potential inferior state monitoring data matches the classification label.

[0114] In this design, the data analysis system further uses the inferior state analysis network to process the target state monitoring data of the non-inferior hydropower station generator units. This process not only involves entering the target state monitoring data and the target operating state element representation array into the network, but also includes considering the classification labels corresponding to the target state monitoring data.

[0115] For example, the data analysis system first enters the target state monitoring data of the non-inferior hydropower station generator set, the corresponding classification labels of these data, and the previously acquired target operating state element representation array into the inferior state analysis network. The classification label here is a kind of annotation of the target state monitoring data, which indicates the category or characteristics of the data, and helps the system understand and process the data more accurately.

[0116] By entering this data and information into the network, the deterioration state analysis network can conduct in-depth analysis and processing. The complex algorithms and models within the network will comprehensively consider various factors, such as data characteristics, classification labels, and operating state element representations, to generate potential deterioration state monitoring data corresponding to the target state monitoring data.

[0117] Importantly, among these potential degradation status monitoring data, the system pays special attention to those hidden danger data sets that match the classification labels. Hidden danger data sets refer to data sets that may indicate that the generator set has potential problems or is about to enter a degradation state. By identifying and analyzing these data, the data analysis system can detect potential problems of the generator set in advance, providing an important basis for subsequent maintenance and repair work.

[0118] In this way, the condition monitoring data of the hydropower station generator set can be mined and processed more deeply. By comprehensively considering the characteristics of the data, classification labels, and the characterization of the operating status elements, the system can more accurately identify and predict the potential deterioration of the generator set. This not only helps to improve the safety and reliability of the operation of the hydropower station, but also provides more effective decision-making support for operation and maintenance personnel. In this way, the accuracy and predictability of the data analysis system for the condition monitoring of the hydropower station generator set are enhanced, providing a strong guarantee for the long-term stable operation of the hydropower station.

[0119] In some optional embodiments, the target state monitoring data of the non-inferior hydropower station generator set and the target operating state element representation array are entered into the inferior state analysis network to obtain potential inferior state monitoring data corresponding to the target state monitoring data, including: entering the target state monitoring data of the non-inferior hydropower station generator set into the inferior state analysis network, performing X-order feature confrontation on the target state monitoring data through the confrontation branch of the inferior state analysis network, and obtaining the confrontation state vector of the target state monitoring data, wherein the confrontation state vector includes a v-order confrontation state vector, X is a positive integer, and v+1 is a positive integer not greater than X; by mapping The anti-antagonistic branch mapped with the target operating state element representation array performs X-order perturbation cleaning on the antagonistic state vector of the target state monitoring data; wherein the potential degraded state vector obtained by the v-order perturbation cleaning includes: a first sub-vector of the secondary label matching the classification label in the v-order input vector, and a second sub-vector characteristic of the secondary label matching the classification label in the perturbation cleaning vector corresponding to the v+1-order input vector; the potential degraded state vector obtained after the X-order perturbation cleaning is processed by the data generation branch of the structural hidden danger identification branch in the inferior state analysis network to obtain the potential degraded state monitoring data corresponding to the target state monitoring data.

[0120] Based on this embodiment, when the data analysis system processes the target state monitoring data of the non-inferior hydropower station generator set, it will enter it into the inferior state analysis network together with the target operating state element representation array. This process involves multiple complex steps to ensure the accuracy and validity of the data.

[0121] First, the data analysis system enters the target state monitoring data of the non-inferior hydropower station generator sets into the inferior state analysis network. This step is the basis of data processing and provides raw data for subsequent feature extraction and adversarial analysis.

[0122] Next, the adversarial branch of the inferior state analysis network performs X-order feature adversarial on the target state monitoring data. Feature adversarial is a technique that tests the robustness of a model by introducing small perturbations, which can reveal the performance of the model in the face of slight changes. In this process, the system generates adversarial state vectors of the target state monitoring data, which contain adversarial features at different orders. In particular, the adversarial state vectors include adversarial state vectors of order v, where v+1 is a positive integer not greater than X, which means that the system performs multi-order feature adversarial to more comprehensively capture the potential features of the data.

[0123] The data analysis system then uses the anti-adversarial branch that maps the array of representations of the target operating state elements to perform X-order perturbation cleaning on the adversarial state vector. The purpose of this step is to remove the noise and interference that may be introduced in the adversarial state, thereby obtaining a purer data representation. During the v-order perturbation cleaning process, the system focuses on the first sub-vector of the secondary label that matches the classification label, and the second sub-vector features of the secondary label that matches the classification label in the perturbation cleaning vector corresponding to the v+1-order input vector. These vector features are the key to data analysis, and they can reflect the stability characteristics of the data under different orders of perturbations.

[0124] Finally, the data generation branch of the structural hidden danger identification branch in the inferior state analysis network processes the potential inferior state vector obtained after cleaning the X-order disturbance. This step is the key to data conversion and presentation. It converts the cleaned vector into potential inferior state monitoring data with practical significance. These data not only reflect the current operating status of the generator set, but also predict possible hidden dangers and problems, providing important decision-making basis for operation and maintenance personnel.

[0125] It can be seen that through multi-order feature confrontation and disturbance cleaning, as well as targeted data processing and conversion, the data analysis system can more deeply explore and understand the status monitoring data of the hydropower station generator set. This not only improves the accuracy and efficiency of data processing, but also provides more accurate and reliable data support for subsequent fault warning and preventive maintenance. In this way, the depth and comprehensiveness of the data analysis system's status monitoring of the hydropower station generator set are improved, providing a more solid guarantee for the safe and stable operation of the hydropower station. An embodiment of the present invention provides a computer-readable storage medium on which a program is stored, and when the program is executed by a processor, the method for analyzing the deterioration of the hydropower station generator set is implemented.

[0126] An embodiment of the present invention provides a processor, which is used to run a program, wherein the method for analyzing the degradation of a generator set in a hydropower station is executed when the program is run.

[0127] In the embodiment of the present invention, Figure 2 As shown, the data analysis system 100 includes at least one processor 101, and at least one memory 102 and a bus 103 connected to the processor 101; wherein the processor 101 and the memory 102 communicate with each other through the bus 103; the processor 101 is used to call the program instructions in the memory 102 to execute the above-mentioned hydropower station generator set degradation analysis method.

[0128] The present invention is described with reference to flowcharts and / or block diagrams of methods, data analysis systems (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0129] In a typical configuration, the data analysis system includes one or more processors (CPU), memory, and bus. The data analysis system may also include input / output interfaces, network interfaces, and the like.

[0130] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip. The memory is an example of a computer-readable medium.

[0131] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage computer readable storage media or any other non-transmission media that can be used to store information that can be accessed by a data analysis system. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0132] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or computer-readable storage medium that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such process, method, commodity or computer-readable storage medium. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the presence of other identical elements in the process, method, commodity or computer-readable storage medium that includes the elements.

[0133] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for analyzing the deterioration of a hydropower station generator set, characterized in that: Applied to a data analysis system, the method comprises: Obtaining a target operating state element representation array, wherein the target operating state element representation array is obtained by entering historical deterioration state monitoring data into a pre-debugged deterioration state analysis network, and updating the operating state element representation array after parameter reset on the basis of locking the neuron weights of the deterioration state analysis network, and the target operating state element representation array reflects the operating state monitoring of the deterioration hydropower station generator set thermal network; Entering the target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element representation array into the deteriorated state analysis network to obtain potential deteriorated state monitoring data corresponding to the target state monitoring data; Before obtaining the target operating state element representation array, the method further includes: Entering the running state element representation array after parameter reset and the historical deterioration state monitoring data into the pre-debugged deterioration state analysis network, and updating the running state element representation array on the basis of locking the neuron weights of the deterioration state analysis network; The inferior state analysis network includes a structural hidden danger identification branch and a device loss identification branch; The step of inputting the operating state element characterization array and the historical deterioration state monitoring data of the completed parameter reset into the pre-debugged deterioration state analysis network includes: Entering the historical degradation state monitoring data into the feature mining model of the structural hidden danger identification branch to map the historical degradation state monitoring data into a historical structural hidden danger state vector, wherein the historical structural hidden danger state vector is used to obtain a corresponding adversarial state vector through the adversarial branch processing of the device loss identification branch; The anti-confrontation branch in the device loss identification branch performs disturbance cleaning on the confrontation state vector based on the running state element representation array after parameter reset, to obtain the result output by the anti-confrontation branch, wherein the running state element representation array after parameter reset is projected to the anti-confrontation branch by the mutual focus strategy; The method further comprises: Determine a first AI algorithm indicator based on a result output by the anti-adversarial branch; Based on the first AI algorithm indicator, an array representing target operating status elements is obtained.

2. The method according to claim 1, characterized in that The anti-adversarial branch includes X depth residual layers corresponding to X perturbation cleanings, where X is a positive integer; the method further includes: Determine a d-th state identification feature based on the running state element representation array of the completed parameter reset and the first configuration factor of the d-th depth residual layer, where d is a positive integer less than or equal to X; Determine a d-th state attribute feature based on the running state element representation array of the completed parameter reset and the second configuration factor of the d-th depth residual layer; Based on the d-th state identification feature and the d-th state attribute feature, the running state element representation array after completing parameter reset is projected to the d-th depth residual layer.

3. The method according to claim 2, characterized in that The determining the first AI algorithm index based on the result output by the anti-confrontation branch includes: determining the first AI algorithm index based on the result output by the anti-confrontation branch and a random disturbance feature; The obtaining of a target operating state element representation array based on the first AI algorithm indicator includes: obtaining a target operating state element representation array by determining a minimum value of the first AI algorithm indicator.

4. The method according to claim 1, characterized in that: The operation state element characterization array after parameter reset and the historical deterioration state monitoring data are input into the pre-debugged deterioration state analysis network, and the operation state element characterization array is updated on the basis of locking the neuron weights of the deterioration state analysis network, including: Enter the historical deterioration state monitoring data and its corresponding state monitoring annotation data into the pre-debugged deterioration state analysis network; On the basis of locking the neuron weights of the inferior state analysis network, determining a second AI algorithm indicator based on information in a key area of ​​the annotation set in the historical inferior state monitoring data; Obtaining a target operating state element representation array based on the second AI algorithm indicator; Among them, the key area of ​​the annotation set corresponds to the hidden danger data set in the historical degradation status monitoring data.

5. The method according to claim 4, characterized in that The determining of the second AI algorithm indicator based on the information of the key area of ​​the annotation set in the historical degradation state monitoring data on the basis of locking the neuron weights of the degradation state analysis network includes: Determining a comparison result of the random perturbation characteristics and the result output by the anti-adversarial branch in the inferior state analysis network; Determining a second AI algorithm indicator based on the comparison result and information of a key area in the annotation set in the historical degradation state monitoring data; Obtaining a target operating state element representation array based on the second AI algorithm indicator includes: By determining the minimum value of the second AI algorithm indicator, an array representing the target operating state elements is obtained.

6. The method according to claim 4 or 5, characterized in that: The target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element characterization array are entered into the deteriorated state analysis network to obtain the potential deteriorated state monitoring data corresponding to the target state monitoring data, including: Entering the target state monitoring data of the non-deteriorated hydropower station generator set, the classification label corresponding to the target state monitoring data, and the target operating state element representation array into the deteriorated state analysis network to obtain the potential deteriorated state monitoring data corresponding to the target state monitoring data; The potential degradation status monitoring data matches the hidden danger data set of the classification label.

7. The method according to claim 4 or 5, characterized in that: The target state monitoring data of the non-deteriorated hydropower station generator set and the target operating state element characterization array are entered into the deteriorated state analysis network to obtain the potential deteriorated state monitoring data corresponding to the target state monitoring data, including: Entering the target state monitoring data of the non-inferior hydropower station generator set into the inferior state analysis network, performing X-order feature confrontation on the target state monitoring data through the confrontation branch of the inferior state analysis network, and obtaining a confrontation state vector of the target state monitoring data, wherein the confrontation state vector includes a v-order confrontation state vector, X is a positive integer, and v+1 is a positive integer not greater than X; Performing X-order perturbation cleaning on the adversarial state vector of the target state monitoring data through an anti-adversarial branch mapped with the target running state element representation array; wherein the potential degraded state vector obtained by the v-order perturbation cleaning includes: a first subvector of a secondary label matching a classification label in a v-order input vector, and a second subvector of a secondary label matching the classification label in a perturbation cleaning vector corresponding to a v+1-order input vector; The potential degradation state vector obtained after X-order disturbance cleaning is processed through the data generation branch of the structural hidden danger identification branch in the degradation state analysis network to obtain the potential degradation state monitoring data corresponding to the target state monitoring data.

8. A data analysis system, characterized in that: It comprises a processor, a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other via the bus; the processor is used to call program instructions in the memory to execute the method for analyzing the deterioration of a generator set in a hydropower station as described in any one of claims 1 to 7.

Images