Intelligent alarm method of hydropower station computer monitoring system

Abstract

The invention relates to the technical field of hydropower, and discloses an intelligent alarm method of a hydropower station computer monitoring system, comprising the following steps: S100, preprocessing historical operation data of hydropower station equipment to obtain a first data set; s200, establishing a data model based on the first data set; s300, acquiring first associated information based on physical structures of various devices of the hydropower station; s400, acquiring second associated information based on the operation data of various devices of the hydropower station; s500, constructing an equipment association model based on the first association information, the second association information and the data model; s600, monitoring operation data of the target equipment in real time, and dynamically acquiring operation state information of the target equipment based on the equipment association model; and when the operation state information is abnormal, triggering an alarm based on a preset alarm strategy. The problems of high false alarm rate and the like in the prior art are solved.

Description

Technical Field

[0001] The invention relates to the technical field of hydropower, in particular to an intelligent alarm method for a computer monitoring system of a hydropower station. Background Art

[0002] The monitoring system of a hydropower station is an important means to ensure the safe, stable and economical operation of the power station. The traditional alarm method is mainly based on the threshold comparison method, which has problems such as large amount of alarm information, low accuracy and poor readability, which brings a great burden to the on-duty personnel and affects the safe operation of the power station.

[0003] The alarm method in the traditional hydropower station monitoring system mainly relies on the threshold comparison method. Although this method is simple and direct, it has the following major problems in practical application: 1) Large amount of alarm information: The threshold comparison method will generate a large amount of alarm information, because any parameter change beyond the preset threshold will trigger an alarm, which will cause the on-duty personnel to be troubled by a large amount of irrelevant alarm information; 2) Low accuracy: Since the threshold is usually static and cannot adapt to the complex changes in actual operation, many false alarms or missed alarms will be generated. False alarms will waste the time and energy of the on-duty personnel, while missed alarms will cause the real problems to be ignored; 3) Poor readability: Traditional alarm information often lacks context and detailed explanations, making it difficult for on-duty personnel to quickly understand the severity and cause of the alarm; 4) Lack of intelligent analysis: The threshold comparison method usually does not contain intelligent analysis. 5) It is difficult to adapt to complex situations: When faced with complex operating environments and changing operating conditions, fixed thresholds cannot accurately reflect the true state of the system, resulting in the alarm system's response being inflexible; 6) Reliance on manual judgment: Due to the large amount and complexity of alarm information, on-duty personnel need to spend a lot of time analyzing and judging which alarms need to be handled immediately and which can be temporarily ignored; 7) Difficult to update and maintain: As power plant equipment and operating conditions change, thresholds may need to be updated frequently. However, manually adjusting thresholds is time-consuming and can result in inaccuracy; 8) Lack of predictive capabilities: Traditional threshold alarm systems usually do not have the ability to predict failures. They can only provide alarms after a problem occurs, rather than warnings before the problem occurs. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the present invention provides an intelligent alarm method for a hydropower station computer monitoring system, which solves the problems of high false alarm rate and the like in the prior art.

[0005] The technical solution adopted by the present invention to solve the above problems is:

[0006] An intelligent alarm method for a hydropower station computer monitoring system comprises the following steps:

[0007] S100, preprocessing historical operation data of the hydropower station equipment to obtain a first data set;

[0008] S200, establishing a data model based on the first data set;

[0009] S300, acquiring first associated information based on the physical structure of various equipment in the hydropower station;

[0010] S400, acquiring second related information based on operation data of various equipment in the hydropower station;

[0011] S500, constructing a device association model based on the first association information, the second association information, and the data model;

[0012] S600, monitor the operating data of the target device in real time, and dynamically obtain the operating status information of the target device based on the device association model; when there is an abnormality in the operating status information, trigger an alarm based on a preset alarm strategy.

[0013] As a preferred technical solution, step S200 includes the following steps:

[0014] S211, extracting feature data from the first data set; wherein the feature data includes one or more of mean, variance, peak-to-peak value, and slope;

[0015] S212, performing feature selection processing on the feature data to obtain a second data set; constructing a sample set based on the second data set, typical normal samples and typical fault samples; wherein the sample set includes a training set and a test set;

[0016] S213, construct a supervised learning model, train the supervised learning model through the training set, test the supervised learning model that meets the training conditions on the test set according to preset indicators, dynamically tune the supervised learning model based on the test results, and record the dynamically tuned supervised learning model as the data model.

[0017] As a preferred technical solution, step S300 includes the following steps:

[0018] S310, extracting the physical connection relationship between devices according to the system topology structure in the hydropower station;

[0019] S320, acquiring fault prediction propagation path information based on the physical connection relationship;

[0020] S330, acquiring the first associated information based on the fault prediction propagation path information, historical fault information and corresponding handling measures.

[0021] As a preferred technical solution, step 400 includes the following steps:

[0022] S410, calculating the statistical correlation between operating parameters of different devices;

[0023] S420, obtaining frequent co-occurrence pattern information of device status based on a data mining algorithm; wherein the data mining algorithm includes one or more of an Apriori algorithm and an FP-growth algorithm;

[0024] S430, based on the frequent co-occurrence pattern information and the expert database information, the second association information is obtained; wherein the second association information includes strong association pattern information with preset value screened out from the massive association rules of the hydropower station equipment.

[0025] As a preferred technical solution, step 420 includes the following steps:

[0026] S421, determining an input data set for a data mining algorithm based on statistical correlation;

[0027] S422, obtaining a target frequent itemset based on the input data set;

[0028] S423, obtaining frequent co-occurrence pattern information based on the target frequent item set.

[0029] As a preferred technical solution, step 421 includes the following steps:

[0030] A100, select the running status parameters of the equipment as the transaction set for association analysis;

[0031] A200, discretizes the running status parameters, converting the running status parameters into Boolean values ​​or enumeration values;

[0032] A300,consists of all device states at each time sampling point into a transaction record, and several transaction records constitute the input data set of the data mining algorithm.

[0033] As a preferred technical solution, step 422 includes the following steps:

[0034] B100, set a minimum support threshold; based on the input data set, obtain the occurrence frequency of each state; remove the item sets with support lower than the minimum support threshold in the input data set to obtain the first frequent item set; wherein the first frequent item set is a frequent set of a single item;

[0035] B200, performing a self-join operation on the first frequent item set to generate a first candidate item set containing two items; scanning the input data set to obtain the actual support of the first candidate item set; filtering out the first candidate item set whose actual support is lower than the minimum support threshold to obtain a second frequent item set; wherein the actual support is the actual number of occurrences of each first candidate item set;

[0036] B300, perform a self-join operation on the Pth frequent item set to generate a Pth candidate item set containing two items; scan the input data set to obtain the actual support of the Pth candidate item set, each actual support is the actual number of occurrences of each type of the Pth candidate item set; filter out the Pth candidate item set whose actual support is lower than the minimum support threshold to obtain the P+1th frequent item set; when the actual support of the Pth candidate item set is lower than the minimum support threshold, record the Pth frequent item set as the target frequent item set; where P≥2.

[0037] As a preferred technical solution, step 423 includes the following steps:

[0038] C100, enumerate all non-empty subsets of the target frequent itemset; each non-empty subset and its complement constitute a candidate association rule;

[0039] C200, calculate the confidence of each candidate association rule;

[0040] C300, filter out candidate association rules whose confidence is less than the minimum confidence threshold to obtain frequent co-occurrence pattern information.

[0041] As a preferred technical solution, step 430 includes the following steps:

[0042] S431, sorting the rules in the frequent co-occurrence pattern information according to support and confidence;

[0043] S432, filtering out the representational rules in the strong association rule set based on the expert database information to obtain second association information.

[0044] As a preferred technical solution, in step S600, dynamically acquiring the operating status information of the target device based on the device association model includes the following steps:

[0045] D100, a connectivity indicator for obtaining operation data based on a device association model;

[0046] D200, obtains the centrality index of each node based on the device association model; each node corresponds to a device;

[0047] D300,uses community discovery algorithm to analyze connectivity index and centrality index,and obtains the device subgroups whose actual closeness exceeds the,closeness threshold;

[0048] D400, analyze the device subgroup to obtain operation status information; wherein the operation status information includes potential fault nodes and corresponding fault impacts and abnormal future expansion paths.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention provides a model including comprehensive association information between devices through a device association model constructed according to first association information, second association information and a data model, and the model can help quickly locate the root cause of the fault and improve the efficiency of fault diagnosis; through the device association model, a number of operating data of the monitored target device can be analyzed in real time, and by deeply analyzing the device data and relationships, the hidden association relationships that may exist between the devices can be discovered from a comprehensive perspective of the devices, so that possible faults and damages can be better predicted and prevented, thereby improving the safety and reliability of the hydropower station; through predictive maintenance and optimized maintenance plans, maintenance costs and unplanned downtime can be reduced; with the help of triggering early warning strategies, preventive maintenance can be effectively implemented, maintenance costs caused by equipment failures can be reduced, and the long-term stable operation of the hydropower station equipment can be guaranteed to the greatest extent, and intelligent monitoring and alarming make operation management more efficient and accurate, which helps to improve the overall operation efficiency and production capacity of the hydropower station; this comprehensive intelligent alarm method can not only ensure the normal operation of the hydropower station, but also provide solid data support and basis for future operation and maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flow chart of an intelligent alarm method for a hydropower station computer monitoring system provided by an embodiment of the present disclosure;

[0052] Figure 2 A flowchart of a method for establishing a data model provided in an embodiment of the present disclosure;

[0053] Figure 3 A schematic diagram of a flow chart of a method for acquiring first associated information provided by an embodiment of the present disclosure;

[0054] Figure 4 A schematic diagram of a flow chart of a method for acquiring second associated information provided by an embodiment of the present disclosure;

[0055] Figure 5 A flowchart of a method for acquiring frequent co-occurrence pattern information provided by an embodiment of the present disclosure;

[0056] Figure 6 A schematic diagram of a flow chart of a method for determining an input data set provided in an embodiment of the present disclosure;

[0057] Figure 7 A flowchart of a method for obtaining a target frequent item set provided by an embodiment of the present disclosure;

[0058] Figure 8 A flowchart of a method for obtaining frequent co-occurrence pattern information provided by an embodiment of the present disclosure;

[0059] Fig. 9A flowchart of a method for acquiring second association information based on frequent co-occurrence pattern information and expert database information provided in an embodiment of the present disclosure;

[0060] Fig.10 A schematic diagram of a process flow of operating status information of a target device provided by an embodiment of the present disclosure;

[0061] Fig.11 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0062] The present invention will be further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0063] Example 1

[0064] like Figures 1 to 11 As shown, refer to Figure 1 The first aspect of the present application discloses an intelligent alarm method for a hydropower station computer monitoring system, the method specifically comprising the following steps:

[0065] S100, pre-processing the historical operation data of various equipment in the hydropower station to obtain a first data set; the operation data includes at least one of a state quantity, a telemetry quantity, and a telesignaling quantity.

[0066] Specifically, preprocessing includes cleaning, normalization and other processing to remove outliers and noise and ensure data quality and availability.

[0067] These operating data may include state quantities (such as temperature, pressure), remote measurement quantities (such as flow, voltage), remote signal quantities (such as switch status) and other data.

[0068] S200: Establish a data model based on the first data set.

[0069] In this step, the data model established is a data model that reflects the operating characteristics of the equipment; specifically, the data model established may be a regression model, a time series model, and the like.

[0070] S300, obtaining first associated information based on the physical structure of various types of equipment in the hydropower station; the first associated information includes the physical connection relationship between the equipment, the cause-effect chain model and the fault knowledge base.

[0071] In this step, the first correlation information helps understand the dependencies and impact links between devices, which helps to more accurately analyze the root causes of device failures; through the causal chain model, possible fault propagation paths can be quickly located and identified, thereby improving the efficiency of fault diagnosis.

[0072] S400, obtaining second associated information based on operation data of various equipment in the hydropower station; the second associated information includes statistical correlation and frequent co-occurrence pattern information.

[0073] In this step, the second association information reveals the statistical correlation and co-occurrence pattern between devices, helping to discover possible but not obvious device associations; the statistical correlation and co-occurrence pattern information provides a basis for comprehensive analysis of the complex relationships between devices, helping to improve the reliability and stability of the overall system.

[0074] S500, constructing a device association model based on the first association information, the second association information, and the data model.

[0075] In this step, the device association model comprehensively considers the physical structure, statistical relationship and prediction model, and provides a comprehensive analysis capability of the operating status of the entire system; the model can support the system to make intelligent decisions and responses in complex environments, and improve the real-time and adaptability of the system.

[0076] S600, real-time monitoring of several operating data of the target device, and dynamically obtaining operating status information of the target device based on the device association model;

[0077] When there is an abnormality in the running status information, an alarm is triggered based on the preset alarm strategy.

[0078] In this step, the intelligent monitoring and alarm system can detect changes in equipment status in real time and respond quickly to reduce potential system losses and risks; timely alarms can help implement preventive maintenance measures and reduce the impact of equipment failures on system operations.

[0079] The intelligent alarm method of the hydropower station computer monitoring system disclosed in the present application provides a model containing comprehensive association information between devices through an equipment association model constructed according to the first association information, the second association information, and the data model. The model can help quickly locate the root cause of the fault and improve the efficiency of fault diagnosis; through the equipment association model, a number of operating data of the monitored target equipment can be analyzed in real time, and by in-depth analysis of the equipment data and relationships, the hidden association relationships that may exist between the equipment can be discovered from a comprehensive perspective of the equipment, which can better predict and prevent possible failures and damages, thereby improving the safety and reliability of the hydropower station; through predictive maintenance and optimized maintenance plans, maintenance costs and unplanned downtime can be reduced; with the help of triggering early warning strategies, preventive maintenance can be effectively implemented, maintenance costs caused by equipment failures can be reduced, and the long-term stable operation of hydropower station equipment can be guaranteed to the greatest extent. Intelligent monitoring and alarm make operation management more efficient and accurate, which helps to improve the overall operation efficiency and production capacity of the hydropower station; this comprehensive intelligent alarm method can not only ensure the normal operation of the hydropower station, but also provide solid data support and basis for future operation and maintenance decisions.

[0080] The method disclosed in this application is highly intelligent, automated, and user-friendly. It can adjust its monitoring and alarm strategies according to different equipment and environmental conditions. It has strong adaptability and flexibility, can improve the accuracy and efficiency of alarms, reduce the burden on duty personnel, and ensure the safe and stable operation of power stations. It can also improve the safety management level of hydropower stations, and help improve operational efficiency and economic benefits. It is an important direction for the modernization and intelligent development of hydropower stations.

[0081] In this embodiment, the method for obtaining the preset alarm strategy includes: determining the alarm classification standard according to the importance of the equipment and the fault consequence information, the alarm classification standard including different levels of accident levels, different levels of fault levels and different levels of warning levels; for typical working conditions, setting the trigger conditions, confirmation mechanism, processing flow, etc. of various alarms, and establishing a standard alarm strategy library.

[0082] Specifically, the preset alarm strategies may include object-oriented alarm screening strategies, hierarchical alarm strategies, conditional alarm strategies, delayed alarm strategies, and related quantity alarm strategies.

[0083] By setting different levels of accident levels, fault levels and warning levels, the priority of alarms can be clearly divided. This helps operation and maintenance personnel to quickly determine which alarms need to be handled first when receiving multiple alarms, thereby reasonably allocating resources and ensuring the stable operation of important equipment.

[0084] Setting alarm trigger conditions for typical working conditions can enable the system to automatically trigger alarms under specific circumstances, reducing the burden of manual monitoring while improving the timeliness and accuracy of responses. Operation and maintenance personnel can respond quickly according to preset alarm strategies, effectively shortening fault handling time; establishing a standard alarm strategy library means that each alarm has a clear processing flow and confirmation mechanism, which helps the operation and maintenance team to form unified operating specifications, improve the standardization of fault handling, and reduce processing delays or errors caused by human factors.

[0085] Through the preset alarm strategy, real-time monitoring and early warning of equipment status can be achieved, so as to promptly discover and deal with potential problems and avoid huge losses caused by equipment damage or system downtime. This can not only reduce maintenance costs, but also reduce production losses caused by equipment failure. Through reasonable alarm strategies and timely processing mechanisms, the reliability of the hydropower station system can be significantly improved. The preset alarm strategy can promptly remind the operation and maintenance personnel when the equipment is abnormal, thereby avoiding equipment failures from causing serious impact on system operation.

[0086] Reference Figure 2 , the method for establishing the data model specifically includes the following steps:

[0087] S210, extracting feature data from the first data set;

[0088] Performing feature selection processing on the feature data to obtain a second data set;

[0089] The characteristic data includes one or more of mean, variance, peak-to-peak value, and slope.

[0090] In this step, the characteristic data is a characteristic quantity that is meaningful for judging the device status, such as statistical quantities such as mean, variance, peak-to-peak value, slope, etc.

[0091] Among them, feature selection processing includes: removing redundant and noise features and reducing model complexity. Commonly used feature selection methods include filtering, wrapping, embedding, etc.

[0092] By performing feature selection on feature data, the complexity and noise impact of model training can be reduced, while the generalization ability of the model can be improved.

[0093] When necessary, transform the features to improve their expressiveness, such as performing polynomial expansion and logarithmic transformation on nonlinear relationships.

[0094] S220, constructing a sample set based on the second data set, typical normal samples and typical fault samples, where the sample set includes a training set and a test set.

[0095] Among them, the training set is used for model learning, and the test set is used for performance evaluation.

[0096] S230, training the constructed supervised learning model through the training set, testing the supervised learning model that meets the training conditions on the test set according to preset indicators, dynamically tuning the supervised learning model based on the test results, and recording the dynamically tuned supervised learning model as the data model.

[0097] The supervised learning model constructed can be a logistic regression, decision tree, support vector machine, neural network and other models.

[0098] Among them, the preset indicators can be accuracy, recall rate, F1-score, etc.; the dynamic tuning methods can include adjusting model hyperparameters, feature selection strategies, data preprocessing methods, etc.

[0099] Through dynamic tuning, the model can be more adaptable to data changes and failure conditions in actual scenarios, improving the model's prediction accuracy and stability; through repeated testing and adjustment, the model's performance and generalization capabilities can be optimized to ensure its effective application on new data in the future.

[0100] Furthermore, methods such as oversampling and undersampling can be used to balance the data set to improve the generalization ability of the model.

[0101] In this step, the model hyperparameters can be set reasonably to control the complexity and learning ability of the model; the optimal parameter combination can be selected through grid search, random search and other methods.

[0102] The model is learned on the training set, and the model parameters are updated iteratively until the preset stopping conditions are reached, such as the convergence of the loss function and the number of training rounds.

[0103] When evaluating model performance on the test set, common indicators such as accuracy, recall, F1 value, ROC curve, etc. can be used to comprehensively examine the classification effect of the model.

[0104] Specifically, error analysis can be performed to find samples that are misclassified, analyze the causes, and if necessary, add representative samples, correct features, etc. When tuning the model, the model structure can be adjusted, regularization terms can be introduced, and ensemble learning can be tried to further improve performance.

[0105] The method for establishing a data model disclosed in this embodiment can achieve timely prediction and diagnosis of equipment failures through feature extraction and model training, thereby improving equipment reliability and production efficiency; accurate fault prediction can help enterprises take preventive maintenance measures before equipment failures occur, thereby reducing maintenance costs and production downtime; and through the application of data models, production plans and resource allocation can be optimized, thereby improving production efficiency and product quality.

[0106] Through the above steps, we have established a data model that is highly correlated with the equipment operating status. The model can automatically learn the fault symptom characteristics from the equipment operation data, quickly identify the current situation, and achieve the intelligent monitoring goal of early detection, early warning, and early processing of equipment failures.

[0107] Compared with the traditional alarm method based on manual experience, this modeling method fully taps the value of massive historical data, extracts more comprehensive and rich feature indicators, learns the multi-dimensional operation rules of equipment, and greatly improves the coverage and accuracy of alarms. On the one hand, it reduces the false alarm and missed alarm rates, and on the other hand, it also reduces the workload of on-duty personnel, freeing them from heavy monitoring tasks and focusing on the analysis and handling of abnormal situations.

[0108] For example, for hydroelectric generator sets, we selected more than ten monitoring quantities such as generator temperature, current, voltage, turbine pressure, flow, vibration, etc. as features, and established a support vector machine classification model with multi-monitoring fusion. In practical applications, the model can provide more than 30 minutes of advance warning for several typical faults such as generator winding inter-turn short circuit and water guide shoe temperature rise, which saves valuable time for the safe control and maintenance of the generator set.

[0109] Reference Figure 3 The method for obtaining the first associated information specifically includes the following steps:

[0110] S310, extracting the physical connection relationship between various types of equipment according to the system topology structure in the hydropower station.

[0111] Among them, the physical connection relationship between devices such as common bus, tie switch, etc.

[0112] For example, for the connection between the turbine and the generator, the physical connection method can be confirmed by checking the engineering design drawings and the on-site layout; the input and output interfaces of each device and the connected sensors or actuators can be determined.

[0113] In this step, extracting physical connection relationships can help understand the information flow path between devices, laying the foundation for subsequent fault prediction and propagation path analysis; effective physical connection information can reduce misunderstandings and incorrect inferences, and improve the accuracy and credibility of predictions.

[0114] S320, acquiring fault prediction propagation path information based on the physical connection relationship.

[0115] Specifically, the possible propagation path of the fault in the hydropower station can be determined by combining the physical connection relationship and equipment operation information; the data transmission, signal transmission and energy transfer paths between devices can be considered. Obtaining propagation path information helps to identify potential fault propagation methods and provide early warning of potentially affected equipment and system components; after a fault occurs, the source of the fault can be quickly located, the speed of troubleshooting can be accelerated, and the production interruption time can be reduced.

[0116] For example, for turbine fault prediction, the signal transmission path between it and the turbine control system, turbine speed governor and power dispatching system can be analyzed; and the possible fault propagation mechanisms on each path can be determined, such as abnormal signal transmission, data loss, etc.

[0117] S330, obtaining first associated information based on the fault prediction propagation path information, historical fault information and corresponding handling measures.

[0118] Specifically, the propagation path information of the fault prediction is compared and analyzed with the historical fault data to determine the first associated device or system of the fault prediction; the experience and strategy of fault handling are comprehensively considered to formulate targeted handling measures. Obtaining the first associated information helps to quickly respond to possible fault events, take preventive or emergency measures, and reduce the impact of faults on production; improve the accuracy and timeliness of fault prediction, optimize maintenance strategies, and reduce operating costs and maintenance expenses.

[0119] For example, if a problem is found in the abnormal signal transmission path during fault prediction of a turbine control system, the performance of the path in past faults can be analyzed in combination with historical data; based on fault handling records and technical experience, recommendations can be made to optimize or replace related sensors or signal processing equipment.

[0120] Through the method for obtaining the first associated information disclosed in this embodiment, through accurate physical connection analysis and fault propagation path prediction, potential fault risks can be discovered in a timely manner, and preventive measures can be taken to ensure the stable operation of equipment and systems; based on historical data and the first associated information, maintenance strategies and budget allocations can be optimized to ensure effective use of maintenance resources and cost control; by reducing unexpected downtime and maintenance time, the equipment utilization rate and energy production efficiency of the hydropower station can be improved, thereby enhancing market competitiveness.

[0121] Reference Figure 4 The method for obtaining the second associated information specifically includes the following steps:

[0122] S410, calculating the statistical correlation between operating parameters of different devices.

[0123] The statistical correlation may be a Pearson correlation coefficient, mutual information, or the like.

[0124] Specifically, for each pair of equipment operating parameters, such as turbine speed and generator output power, the correlation coefficient between them is calculated, such as the Pearson correlation coefficient or the Spearman correlation coefficient; statistical correlation can reflect the strength of the linear or nonlinear relationship between equipment operating parameters, which helps to understand the mutual influence and linkage between equipment.

[0125] For example, for the operating data of turbines and generators, the correlation coefficients between them can be calculated through statistical methods. The correlations between other equipment parameters such as pressure and temperature can also be analyzed to fully grasp the operating status of the equipment.

[0126] By determining the correlation between equipment operating parameters, it helps to identify possible common failure modes or impact paths, improve the accuracy of fault prediction, optimize equipment operation control strategies, make the equipment operate in the best condition, and extend equipment life and reliability.

[0127] S420, obtaining frequent co-occurrence pattern information of device status based on a data mining algorithm; the data mining algorithm includes one or more of an Apriori algorithm and an FP-growth algorithm.

[0128] Specifically, the Apriori algorithm or FP-growth algorithm is applied to mine frequently occurring patterns or rules in the equipment status data. These patterns indicate common combinations between equipment operating states and may indicate potential common faults or operating modes.

[0129] For example, the status data of equipment such as turbines, turbine governors, and generators can be mined to find out the frequent co-occurrence patterns between them, such as the operating modes under specific working conditions. The algorithm will automatically discover these patterns and generate specific frequent item sets or association rules.

[0130] In this embodiment, frequent co-occurrence patterns provide important clues to changes in equipment status, which helps to predict possible abnormal operating conditions or fault conditions of the equipment. It can optimize the equipment operation monitoring and early warning system, take timely measures to prevent faults from occurring, and improve the stability and availability of the equipment.

[0131] S430, acquiring second related information based on the frequent co-occurrence pattern information and the expert database information;

[0132] The second association information includes a number of strong association pattern information with preset values ​​screened out from the massive association rules of various types of equipment in the hydropower station.

[0133] Specifically, by combining frequent co-occurrence patterns and expert experience, strong correlation patterns with preset values ​​are screened out; strong correlation patterns may indicate important state combinations in equipment operation or failure tendencies under specific conditions.

[0134] By analyzing the frequent co-occurrence patterns obtained through mining and combining them with expert knowledge to judge their importance and application value in actual scenarios, we can screen out association patterns that are of practical significance for equipment operation status monitoring and fault prediction.

[0135] The solution disclosed in this embodiment, obtaining the second associated information helps to optimize fault prediction and maintenance management strategies, improve the accuracy and efficiency of prediction, and can help the operation team quickly identify potential equipment problems, take timely maintenance measures, and reduce the impact of faults on production.

[0136] The method for obtaining the second associated information disclosed in this embodiment enhances the understanding and prediction capabilities of the equipment operating status through statistical correlation analysis and frequent co-occurrence pattern mining, which helps to discover potential problems in advance and intervene; formulates more accurate maintenance plans and resource allocation strategies based on the second associated information to improve maintenance efficiency and cost control; integrates statistical analysis and data mining results to provide decision support for management and optimize operation and maintenance decisions.

[0137] Reference Figure 5 , the method for obtaining frequent co-occurrence pattern information specifically includes the following steps:

[0138] S421, determining an input data set of a data mining algorithm based on statistical correlation.

[0139] Specifically, according to the results of the previous statistical correlation analysis, equipment operating parameters with significant correlation are selected as input features of the data mining algorithm. For example, parameters such as turbine speed, water level, generator output power, etc. are selected as features of the input data set.

[0140] Assume that statistical analysis shows that there is a significant linear correlation between turbine speed and generator output power; use turbine speed and generator output power as input features of the data mining algorithm to construct an input data set.

[0141] Through this step, it can be determined that the appropriate input data set can improve the efficiency and accuracy of the data mining algorithm and avoid the interference of invalid features; the features selected by statistical correlation are more likely to reflect the key information of the equipment operation status, which is helpful for the subsequent mining of frequent co-occurrence patterns.

[0142] S422, obtaining a target frequent itemset based on the input data set.

[0143] Specifically, the Apriori algorithm or the FP-growth algorithm is applied to mine frequent itemsets in the data set, which reflect the patterns that often co-occur in the device status data; wherein, a frequent itemset refers to a set of features or attributes that often co-occur in the data set.

[0144] For the selected input data set (such as turbine speed, water level, generator output power), the Apriori algorithm is used to discover frequent item sets of these parameters. The algorithm will identify the patterns in which these parameter combinations frequently appear under different operating conditions.

[0145] This embodiment discloses a method for obtaining target frequent item sets, which helps to understand the common patterns and conditions of the equipment operating status and provide clues for predicting possible equipment status changes or failure tendencies. These frequent item sets can serve as the basis for subsequent analysis to help determine key patterns and behaviors in equipment operation.

[0146] S423, obtaining frequent co-occurrence pattern information based on the target frequent item set.

[0147] Specifically, based on the target frequent item sets, the device state data is further analyzed to identify frequently co-occurring patterns therein, which may include common change patterns or event sequences of device states under specific conditions.

[0148] Assume that frequent item sets of turbine speed and water level under different working conditions are obtained through data mining. These item sets are further analyzed to identify the frequent co-occurrence pattern that the water level decreases when the turbine speed increases.

[0149] In this step, obtaining frequent co-occurrence pattern information helps to understand the patterns and laws of changes in the equipment's operating status, providing a basis for fault prediction and preventive maintenance. It can detect possible problems with the equipment in advance and take measures to avoid or reduce the occurrence of faults.

[0150] The method for obtaining frequent co-occurrence pattern information disclosed in this embodiment, through statistical correlation analysis and data mining, obtains frequent co-occurrence pattern information, which can more comprehensively understand the changes and correlations in the operating status of equipment and improve the depth and accuracy of analysis; based on the frequent co-occurrence pattern information, an intelligent fault prediction model can be established to provide early warning of possible equipment failures or abnormalities and reduce operational risks; through data-driven equipment status analysis, equipment operation and maintenance strategies and resource allocation can be optimized, operation and maintenance efficiency and equipment availability can be improved, and maintenance costs can be reduced.

[0151] Reference Figure 6 , the method for determining the input data set specifically includes the following steps:

[0152] A100, select the operating status parameters of several devices as the transaction set for correlation analysis.

[0153] Specifically, operating status parameters of several devices, such as switch values, alarm signals, and over-limit signs, are selected as transaction sets for association analysis.

[0154] Specifically, according to the equipment type and importance of the hydropower station, representative operating status parameters are selected as the objects of analysis. These parameters may include turbine speed, water level, generator output power, etc., to ensure that the selected parameters can effectively reflect the operating status and performance changes of the equipment.

[0155] For example, the speed, output power and control valve opening of the turbine, generator and turbine governor are selected as the transaction set for association analysis. These parameters cover the core indicators of equipment operation and help to understand the mutual influence and linkage between equipment.

[0156] In this step, selecting appropriate operating status parameters can improve the effectiveness and accuracy of the association analysis, avoid irrelevant or unimportant parameters interfering with the data mining results, and fully grasp the operating conditions of the equipment, providing a basis for subsequent data processing and model building.

[0157] A200, reasonably discretize the operating status parameters and convert them into Boolean values ​​or enumeration values.

[0158] For example, the temperature over-limit alarm is changed to two values: "Yes" and "No".

[0159] Specifically, according to the specific operating status parameter characteristics, determine the appropriate discretization method, such as threshold division, classification standards, etc., to convert continuous numerical parameters into discrete Boolean values ​​(such as normal / abnormal) or enumeration values ​​(such as low / medium / high).

[0160] For example, for the turbine speed, discrete states such as low speed, medium speed, and high speed can be set; for the generator output power, discrete states such as low load, medium load, and high load can be set.

[0161] In this step, discretization can simplify the processing complexity of the data mining algorithm, reduce the dimension of the parameters, and make the association analysis more effective and efficient; after conversion into discrete values, the changes in the operating status of the equipment can be understood more intuitively, which facilitates subsequent pattern mining and the discovery of association rules.

[0162] A300, all device states at each time sampling point constitute a transaction record, and several transaction records constitute the input data set of the data mining algorithm.

[0163] Specifically, at each time point, all selected equipment status parameters in the hydropower station are organized into a transaction record according to the discretized values, so that each transaction record contains the status information of each device at the same time point and is the basic unit for association analysis.

[0164] For example, assuming that sampling is performed once per minute, the equipment status data per minute (such as the discretized state of turbine speed and generator output power) constitutes a transaction record. The transaction records at multiple time points are combined into a complete input data set for subsequent association rule mining.

[0165] In this step, each transaction record contains the status information of multiple devices at the same time point, which helps to discover the time correlation and co-occurrence patterns between devices. Data mining algorithms can use these transaction records to identify frequently co-occurring patterns, thereby helping to predict possible linkage failures or abnormal states of devices.

[0166] The method for determining the input data set disclosed in this embodiment can fully understand the operating status and changes of various equipment in the hydropower station by selecting appropriate operating status parameters and effective discretization processing; constructing a data set based on transaction records helps the data mining algorithm to accurately mine frequent co-occurrence patterns between equipment states and improve prediction accuracy; the method of integrating discretization processing and transaction record construction can provide optimized operation and maintenance management and fault prediction decision support for hydropower stations.

[0167] Reference Figure 7 , the method for obtaining the target frequent itemset specifically includes the following steps:

[0168] B100, setting the minimum support threshold;

[0169] Based on the input data set, obtain the occurrence frequency of each state;

[0170] Eliminate the item sets whose support is lower than the minimum support threshold in the input data set to obtain the first frequent item set; the first frequent item set is the frequent set of a single item.

[0171] Specifically, a minimum support threshold is set, such as 10%. The input data set is scanned, the occurrence frequency of each state is counted, and a 1-frequent item set (ie, the first frequent item set) is generated.

[0172] The minimum support is a pre-set threshold used to define the minimum occurrence frequency or probability of frequent itemsets. Only itemsets that meet the minimum support can be considered frequent, otherwise they will be excluded from the mining results.

[0173] Assume that the minimum support threshold is set to 0.1, that is, the frequency of an item set in the data set is at least 10%, which means that only those item sets with a frequency higher than 10% can be considered frequent and used for subsequent frequent item set mining. Setting the minimum support threshold can filter out item sets with low frequency in the data set, improve the accuracy and effectiveness of the mining results, and focus on analyzing those frequent item sets that are truly important and trendy, reducing unnecessary calculation and analysis workload.

[0174] B200, performing a self-join operation on the first frequent item set to generate a number of first candidate item sets containing two items;

[0175] Scan the input data set to obtain the actual support of the first candidate item set, which is the actual number of occurrences of each first candidate item set;

[0176] The first candidate item set whose actual support is lower than the minimum support threshold is filtered out to obtain the second frequent item set, and B300 is executed cyclically based on the second frequent item set.

[0177] Specifically, a self-join operation is performed on the 1-frequent itemsets to generate a 2-candidate itemset (i.e., the first candidate itemset). The input data set is scanned, the support of the 2-candidate itemsets is counted, and the 2-frequent itemsets (i.e., the second frequent itemsets) are filtered out.

[0178] In this step, self-connection refers to the process of generating candidate item sets from known frequent item sets, which contain more items but still remain frequent; for the Apriori algorithm, the self-connection step is to generate new candidate item sets by connecting two frequent item sets.

[0179] For example, suppose the first frequent item set contains single items such as {turbine speed} and {generator output power}. Through the self-join operation, a candidate item set containing two items can be generated, such as {turbine speed, generator output power}.

[0180] In this step, the self-join operation expands the combination possibilities of frequent itemsets, which helps to discover more complex frequent itemset patterns. In the process of continuous iteration, it can generate new candidate itemsets that meet the preset minimum support, laying the foundation for the next step of frequent itemset mining.

[0181] B300, perform self-join operation on the Pth frequent item set to generate several Pth candidate item sets containing two items; P ≥ 2;

[0182] Scan the input data set to obtain several actual supports of the Pth candidate item set, each of which is the actual number of occurrences of each type of the Pth candidate item set;

[0183] Filter out the Pth candidate item set whose actual support is lower than the minimum support threshold, and obtain the P+1th frequent item set;

[0184] When the actual support of the Pth candidate item set is lower than the minimum support threshold, the Pth frequent item set is recorded as the target frequent item set.

[0185] Specifically, step B300 is repeated to generate k-candidate item sets and k-frequent item sets layer by layer until no new frequent item sets are generated.

[0186] Repeating step B300 means that, when P is 2, a self-join operation is performed on the second frequent item set to generate a number of second candidate item sets containing two items, and the input data set is scanned to obtain a number of actual support degrees of the second candidate item sets, where each actual support degree is the actual number of occurrences of each type of the second candidate item set;

[0187] When the actual support of the second candidate item set is lower than the minimum support threshold, the second frequent item set is recorded as the target frequent item set, that is, the process stops when no new frequent item set is generated.

[0188] For the known Pth frequent item set, a new candidate item set containing P+1 items is generated through self-join operation. This step is the key step in the Apriori algorithm, which generates a higher-order candidate item set by combining frequent item sets.

[0189] For example, suppose it is known that the second frequent itemset contains itemsets such as {turbine speed, generator output power}, {turbine speed, turbine governor status}, etc. Through self-joining, a new candidate itemset containing three items can be generated, such as {turbine speed, generator output power, turbine governor status}.

[0190] In this step, the recursive nature of the self-join operation can gradually expand the complexity of frequent itemsets and discover longer frequent itemset patterns. It can ensure that the newly generated candidate itemsets meet the preset minimum support requirements at each step, thereby ensuring the accuracy and practicality of the mining results.

[0191] The method for obtaining the target frequent item set disclosed in this embodiment can efficiently discover and screen out qualified frequent item sets through step-by-step self-connection operations; by setting the minimum support threshold and self-connection operations, it is ensured that the mining results have a certain degree of accuracy and interpretability; the implementation of the overall solution can provide a basis for decision support and optimization of the equipment status of the hydropower station, thereby improving the efficiency and reliability of equipment operation.

[0192] Reference Figure 8 , the method for obtaining frequent co-occurrence pattern information specifically includes the following steps:

[0193] C100, enumerate all non-empty subsets of the target frequent itemset; each non-empty subset and its complement constitute a candidate association rule.

[0194] Specifically, for each frequent item set, all its non-empty subsets are enumerated, where a non-empty subset refers to a subset that selects at least one item from the frequent item set.

[0195] For example, assuming that the target frequent item set is {turbine speed, generator output power}, the non-empty subsets that can be generated are {turbine speed}, {generator output power}, and {turbine speed, generator output power}.

[0196] In this step, enumerating non-empty subsets can generate multiple candidate association rules. Each rule represents a combination relationship in the frequent item set. The generation of these rules is the basis for the subsequent calculation of confidence and helps to discover the correlation and patterns between different items.

[0197] C200, calculate the confidence of each candidate association rule,

[0198] In this embodiment, the confidence is the quotient of the support of the candidate association rule divided by the support of the antecedent subset, and the confidence reflects the credibility of the rule.

[0199] Specifically, confidence refers to the conditional probability of the rule "If X appears, then Y appears", and the calculation formula is: confidence (X>Y) = support (X∪Y) / support (X), where support (X∪Y) represents the frequency of simultaneous occurrence of X and Y, and support (X) represents the frequency of X appearing alone.

[0200] For example, for the rule {turbine speed}->{generator output power}, calculate its confidence: support({turbine speed, generator output power}) / support({turbine speed}).

[0201] In this step, confidence evaluation helps to identify and quantify the degree of association between rules. Rules with high confidence usually have stronger associations. Rules with confidence below the threshold can be filtered out, reducing unnecessary information and improving the effectiveness of association rule mining.

[0202] C300, filter out candidate association rules whose confidence is less than the minimum confidence threshold to obtain frequent co-occurrence pattern information.

[0203] The minimum confidence threshold may be set to 80%, and the rules whose confidence does not meet the requirement are filtered out to obtain frequent co-occurrence pattern information. In this embodiment, the frequent co-occurrence pattern information is the strong association rule set information.

[0204] Specifically, all calculated candidate association rules are compared with a set minimum confidence threshold, and rules with confidence greater than or equal to the threshold are retained as frequent co-occurrence pattern information.

[0205] For example, assuming that the minimum confidence threshold is set to 0.6, that is, only association rules with a confidence not less than 0.6 are retained, the filtered rule may be {turbine speed}>{generator output power}, which represents a strong association between the two.

[0206] In this step, screening out high-confidence association rules helps to understand the common association patterns between equipment status parameters. These frequent co-occurrence pattern information can serve as the basis for equipment operation status analysis, fault prediction, and optimization decisions.

[0207] The method for obtaining frequent co-occurrence pattern information disclosed in this embodiment can deeply analyze and understand the relationship and pattern between equipment status parameters through association rule mining; the discovered frequent co-occurrence pattern information can help predict possible linkage failures or abnormal conditions of equipment and perform preventive maintenance in advance; it provides a scientific basis and data support for operation and maintenance management and equipment optimization, and improves the operating efficiency and reliability of hydropower station equipment.

[0208] Specifically, for each frequent item set, enumerate all its non-empty subsets. Each subset and its complement constitute a candidate association rule. For example, if a frequent item set contains {A, B, C}, then its non-empty subsets are {A}, {B}, {C}, {A, B}, {A, C}, {B, C}.

[0209] Calculate the confidence of each candidate rule. The confidence is defined as the support of the rule divided by the support of the antecedent subset. For example, for the rule {A, B}->{C}, its confidence is calculated as support({A, B, C}) / support({A, B}).

[0210] Set a minimum confidence threshold, such as 80%. Filter out rules with confidence below the threshold, and only keep those with confidence above or equal to the threshold. These retained rules constitute a strong association rule set, which is considered to be a relatively credible association relationship.

[0211] By calculating all possible association rules of frequent item sets and filtering out rules that do not meet the set confidence threshold, strong association rules in the data set can be found. These association rules can help understand the association phenomena in the data or be used to develop recommendation systems, market analysis and other applications.

[0212] Reference Fig. 9 Based on the frequent co-occurrence pattern information and the expert database information, the method for obtaining the second related information specifically includes the following steps:

[0213] S431, sorting the rules in the frequent co-occurrence pattern information according to support and confidence.

[0214] It should be noted that other evaluation indicators such as Lift and Chi-square can also be used to comprehensively examine the importance of the rules.

[0215] Among them, support measures the frequency of occurrence of the rule in the data set, that is, the probability that the union of the left and right item sets of the rule appears in the data set. Confidence evaluates the reliability of the rule, that is, the probability that the conclusion part also appears when the condition part of the rule appears.

[0216] For example, for the rules {turbine speed}>{generator output power} and {water level}>{turbine governor status} in the frequent co-occurrence pattern, their support and confidence are calculated respectively; the rules are sorted according to the level of support and confidence, and the rules with high support and confidence are given priority.

[0217] In this step, sorting can help experts and decision makers quickly understand and identify the most influential and relevant rules in frequent co-occurrence patterns, and can give priority to rules with high support and confidence, thereby improving analysis and decision-making efficiency.

[0218] S432, filtering out the representational rules in the strong association rule set based on the expert database information to obtain second association information.

[0219] The purpose of filtering out the representational rules in the strong association rule set is to retain valuable novel rules, that is, to filter out valuable strong association patterns from massive association rules with the help of expert experience in the expert database information and reveal the potential relationship between devices.

[0220] Among them, superficial rules refer to rules that appear to be relevant or highly confident on the surface, but in fact have no real practical meaning or causal relationship; expert database information can provide an understanding of the actual relationship behind the rules and help identify and exclude such superficial rules.

[0221] For example, suppose there is a high-confidence rule {water level}>{turbine governor status} in the frequent co-occurrence pattern, but experts believe that this is only an accidental co-occurrence and has no actual causal relationship; based on expert opinion, such superficial rules can be screened out to obtain more realistic correlation information.

[0222] In this step, screening out representational rules helps to improve the accuracy and practicality of association rule mining, ensure the scientificity and interpretability of the mining results, and avoid misunderstandings and unnecessary decision-making deviations caused by the interference of representational rules.

[0223] The method disclosed in this embodiment for obtaining the second related information based on frequent co-occurrence pattern information and expert library information can obtain deeper and more realistic related information through sorting and expert screening, which helps to fully understand the relationship between equipment status and performance; provide more accurate and reliable decision support for operation and maintenance management and equipment optimization, and help improve equipment operation efficiency and reliability; and comprehensively utilize data mining technology and the knowledge of domain experts to achieve a higher level of equipment status analysis and fault prediction.

[0224] In summary, through the implementation and integration of the above steps, a complete association rule mining and analysis process can be established to provide a scientific basis and technical support for hydropower station equipment management and operation and maintenance decision-making.

[0225] The following is a practical application case to illustrate the specific analysis process when the data mining algorithm is the Apriori algorithm.

[0226] The one-year historical data of six units of a hydropower station includes more than 300 telemetered and telesignaled quantities, with a data volume of more than 10 million. We selected 20 representative parameters such as unit temperature, pressure, flow, and electrical quantity as the analysis objects, and used the Apriori algorithm to deeply explore the correlation rules contained therein.

[0227] Through frequent item set analysis, we found some typical combination patterns of equipment status. For example, the combination of {generator stator winding temperature is too high, generator U-phase current is too large, and excitation current is insufficient} appears many times in historical data, forming a frequent alarm pattern (that is, the information contained in the frequent co-occurrence pattern information).

[0228] By further generating association rules, we obtain several high-confidence judgment rules, such as: insufficient excitation current & excessive current of generator U phase → excessive temperature of generator stator winding (support 12%, confidence 85%).

[0229] This rule shows that when the excitation is insufficient and the single-phase current of the generator is too large, there is an 85% probability that the stator winding will overheat. Analyzing its physical mechanism, it may be due to the weakening of the magnetic field caused by low excitation current, the aggravation of stator current imbalance, and the heating of local windings causing temperature rise.

[0230] Equipment experts further explained that this frequent alarm pattern reflects the problem of unbalanced excitation of the unit. Once the combined alarm of abnormal excitation and unbalanced current appears, it is a harbinger of stator winding problems, and it is necessary to notify the relevant person in charge in advance to conduct an investigation, which provides an important basis for the subsequent improvement of the correlation model and the reasonable configuration of alarm items.

[0231] Similarly, we also found a number of predictive association rules, such as {the low-voltage side circuit breaker of the No. 2 main transformer tripped → the No. 2 unit was shut down}, {the cooling water pressure of the No. 1 unit was low & the vibration of the No. 1 unit increased → the bearing temperature of the No. 1 unit was high}, which greatly expanded the analysis vision of the operators and deepened their understanding of the internal connections of the equipment.

[0232] It can be seen that the Apriori algorithm is good at automatically discovering frequent co-occurrence patterns from large-scale discrete events, and using intuitive logical expressions to describe the associations and causal relationships between things. This data-driven association analysis method can make up for the blind spots of human experience and reveal some laws that are easily overlooked but objectively exist, which is a useful supplement to other analysis methods.

[0233] Of course, frequent co-occurrence does not necessarily mean a definite causal relationship. The association rules found by Apriori need to be judged and interpreted in combination with physical mechanisms to identify the rationality and value of the association and combine it with actual business to truly form an operational equipment association model to guide monitoring and early warning work.

[0234] In this embodiment, the constructed device association model can be a graph model. For example, for device topology and operation association, the first association information, the second association information, and the data model are embodied in the form of directed graphs, undirected graphs, weighted graphs, and the like.

[0235] Specifically, in the figure, each node represents a device object, such as a generator, a transformer, etc., in units of devices.

[0236] If two devices are physically or logically associated, an edge is connected between the corresponding nodes; the association can come from multiple dimensions such as device topology connection, signal configuration, and operating data correlation.

[0237] Then, a weight value can be assigned to each edge based on factors such as association strength and association type. The weight can be static or dynamically adjusted.

[0238] Reference Fig.10 The method for obtaining the running status information of the target device specifically includes the following steps:

[0239] D100, obtains connectivity indicators of several operating data based on the device association model.

[0240] This step aims to evaluate the complexity of the system's associations. The connectivity index can reflect the closeness of the associations between devices, and fragmented connectivity indicates a loose system.

[0241] Specifically, methods in graph theory (such as the shortest path, network traffic, etc.) can be used to calculate the connection strength or path density between devices. Among them, connectivity indicators can reflect the effectiveness and speed of information transmission between devices, as well as the possible synergy between devices.

[0242] For example, assuming that in a device association model, there is a direct connection relationship between devices A, B, and C, the shortest path length or the number of paths from device A to device B can be calculated as part of the connectivity index.

[0243] In this step, connectivity indicators can help identify the close connections between devices, help understand the information transfer and interdependencies during device operation, and provide a data basis for subsequent device status analysis, especially when identifying device subgroups and possible fault nodes.

[0244] D200, obtains the centrality index of each node based on the device association model; each node corresponds to a device.

[0245] Among them, centrality indicators include degree centrality, betweenness centrality, closeness centrality, etc.; nodes with high centrality are more important in association analysis.

[0246] Specifically, centrality indicators (such as betweenness centrality, degree centrality, etc.) are used to measure the central position and influence of device nodes in the network. These indicators can help identify key devices, that is, those nodes that have an important position in the device relationship network.

[0247] For example, assuming that in the device association model, device A has a high degree centrality because it is directly connected to multiple other devices, the degree or other centrality indicators of device A can be calculated to quantify its importance in the network.

[0248] In this step, the centrality index helps to identify and focus on key nodes in the device network, which have an important impact on the stability and operating status of the overall network. It can provide a basis for subsequent device subgroup discovery and operating status analysis, especially in determining possible hidden fault nodes.

[0249] D300,uses a community discovery algorithm to analyze the connectivity and centrality,indicators to obtain the device subgroups whose actual closeness exceeds the,closeness threshold.

[0250] These device subgroups often share common risk factors, either physically or logically.

[0251] Among them, the community discovery algorithm can identify subgroups with high density and close internal connections in the network. These subgroups can be understood as a collection of devices with high interaction and dependency in the device network.

[0252] For example, suppose that through the community discovery algorithm, a subgroup consisting of devices A, B, and C is identified. This subgroup has high internal connectivity and centrality, which may indicate that they have high synergy in operation or are subject to similar external influences.

[0253] In this step, community discovery can help identify functional modules or integrated units in the device network, help understand the operating status and interaction patterns of devices at the group level, and provide more accurate and detailed information support for subsequent equipment status analysis and identification of potential fault nodes.

[0254] D400 analyzes the equipment subgroups to obtain operating status information; the operating status information includes potential fault nodes and the corresponding fault impact and abnormal future expansion path.

[0255] Further fault diagnosis and status analysis are performed on the identified equipment subgroups to identify potential fault nodes.

[0256] Specifically, the interactions and impacts between device subgroups are analyzed to predict possible abnormal expansion paths and impact ranges.

[0257] For example, if device A is identified as a potential hidden fault node in the identified device subgroup, it may have a significant impact on the operating status of the entire subgroup. The failure impact and possible abnormal propagation path of device A can be further analyzed to provide detailed information for operation and maintenance decisions.

[0258] In this step, through in-depth analysis of equipment subgroups, possible hidden faults can be discovered in advance, preventive maintenance measures can be taken, and the adverse effects of equipment failures on system operation can be avoided. A comprehensive assessment and prediction of equipment status can be provided to help optimize equipment operation strategies and maintenance plans.

[0259] The method for obtaining the operating status information of the target device disclosed in this embodiment can deeply analyze the location and role of the device in the network and fully understand the operating status of the device through connectivity indicators, centrality indicators and community discovery algorithms; take measures in advance to prevent equipment failures and optimize equipment operation and maintenance strategies by identifying potential fault nodes and abnormal expansion paths; integrate data analysis and expert knowledge to provide scientific basis and actionable suggestions for equipment management and operation and maintenance decisions. Through the implementation and integration of the above steps, an effective equipment association model analysis process can be established to provide support and guidance for the comprehensive evaluation and optimization of the operating status of hydropower station equipment.

[0260] Specifically, we can use graph theories and algorithms, such as the shortest path and connectivity, to analyze the importance and impact paths of device nodes, determine the nodes that may be the origin of system failures, and accurately locate the faulty nodes by analyzing the abnormal propagation paths. We can perform reasoning and prediction on the graph, such as locating faulty nodes and evaluating the impact of faults, to simulate the transmission and diffusion of device abnormalities, that is, to evaluate the diffusion path and impact range of faults in the system, and understand how abnormalities are transmitted and diffused between devices. We can also perform simulations to simulate the propagation process of abnormal device events, thereby predicting possible system responses and impacts.

[0261] The following takes the vibration fault scenario of a hydropower station unit as an example for a detailed description.

[0262] The main components of the unit are used as nodes, including the spiral casing, top cover, water guide mechanism, volute, tailwater pipe, main shaft, impeller, generator stator, rotor, etc.

[0263] According to the unit structure, the physical connection relationship between the component nodes is established, such as the impeller is connected to the main shaft, the stator and the rotor are matched, the water guide mechanism is connected to the tailwater pipe, etc.

[0264] According to the arrangement of vibration sensors of each component, the corresponding relationship between signal nodes and equipment nodes is established to characterize the logical association between monitoring points and measured objects.

[0265] Based on the correlation analysis of historical vibration data, weights are added to some physical connection edges and logical corresponding edges to dynamically reflect the strength of the association.

[0266] Calculations show that the graph has a small number of connected subgraphs and a high average node degree, indicating that the components of the unit are highly correlated and need to be considered as a whole.

[0267] Degree centrality and betweenness centrality show that the main axis node is at the center of the graph and is a key component that connects the upstream and downstream. Main axis failures can easily spread to upstream hydraulic components and downstream power generation components.

[0268] The community discovery algorithm extracts three closely related component subgroups, corresponding to the turbine flow path component group, the main shaft component group, and the generator component group, which is consistent with the physical division of the unit.

[0269] When a rotor vibration signal frequently alarms, the shortest path analysis shows that the rotor is directly connected to the main shaft and has a clearance fit with the stator. The initial guess is that the rotor vibration is likely caused by the main shaft vibration or the stator vibration.

[0270] Further analysis of the vibration trends of the main shaft and stator revealed that the main shaft vibration amplitude was large and the frequency was close to the rotor vibration, so it was determined that the main shaft was the root cause of the fault.

[0271] After tracing the source to the main shaft node, reverse analysis was performed and it was found that the main shaft problem may affect the impeller fixation and water guide mechanism sealing, thereby worsening the vibration state of the turbine flow path components.

[0272] The vibration spectrum analysis shows that the main shaft is unbalanced and needs to be corrected. At the same time, it is recommended to inspect the impeller fixing bolts, water guide mechanism sealing ring and other affected parts to prevent the problem from expanding.

[0273] Compared with isolated monitoring of a single device, this solution is supported by graph theory algorithms and analyzes the cross-influence between devices from the perspective of combined topology, making the correlation analysis more systematic and standardized. Especially in large systems, graph models can quickly identify risk areas, narrow the scope of investigation, and reduce manual blindness.

[0274] Furthermore, for core equipment, detailed mechanism models can be established, such as the dynamic model of the turbine speed control system.

[0275] Specifically, mathematical models of components such as turbines, generators, speed governors, and PID controllers are established, such as the torque equation of the turbine and the actuator equation of the speed governor.

[0276] According to the physical connection relationship of the components, such as the direct connection between the turbine output shaft and the generator shaft, the governor output signal as the PID controller input, etc., the component models are connected in series and parallel to form a complete system model.

[0277] Define the input and output interfaces of the model, with inputs such as water head, volute opening, etc., and outputs such as unit speed, power, etc., corresponding one-to-one with the actual monitored quantities.

[0278] Typical speed regulation process test data are selected, including the timing curves of speed, opening, power, etc. under conditions such as unit load remote regulation and load shedding.

[0279] The monitoring data is imported into the mechanism model, and the model parameters are fitted using optimization methods such as the least squares method and genetic algorithm to make the simulation curve match the measured curve to the greatest extent.

[0280] Verify the model accuracy under different working conditions, fine-tune the model structure and parameters, and ensure that the model can reflect the true dynamic characteristics of the speed control system.

[0281] Based on the mechanism model, the relationship between component parameters and system response is analyzed and sensitivity analysis is performed.

[0282] Through simulation experiments, the manifestation characteristics and transmission laws of equipment failures are studied to guide the selection of alarm quantities.

[0283] Taking the speed control system of a hydropower station as an example, the following demonstrates a case analysis of a speed control system failure.

[0284] The torque balance equation and flow continuity equation of the turbine are established to describe the nonlinear coupling relationship between the unit speed and the water head, opening and output.

[0285] A PID controller model is established to characterize the logical relationship between the frequency deviation signal of the power supply system and the action of the speed regulator.

[0286] The dynamic model of the speed governor actuator and servo system is established to reflect the response characteristics of the speed governor after receiving the control command.

[0287] The input and output of the turbine, controller and actuator are connected to build a closed-loop speed control system simulation model.

[0288] The full-load rejection test data of the unit are used to identify the dynamic characteristic parameters of the turbine, the proportional, integral and differential coefficients of the governor, the time constant of the actuator, etc. in the model.

[0289] By fixing other parameters and changing the PID parameters of the speed regulator, it is found that the larger the proportional coefficient, the faster the system response, but it will cause overshoot; the larger the integral coefficient, the smaller the system steady-state error, but it will cause low-frequency oscillation.

[0290] By fixing other parameters and changing the actuator time constant, it is found that the larger the time constant, the greater the transmission delay from the governor to the turbine and the worse the system dynamic performance.

[0291] When the proportional coefficient of the speed regulator is set too large, the simulation finds that when the unit is cutting the load, the speed overshoots significantly and high-frequency jitter occurs, and multiple adjustments are required to restore stability, which is consistent with the instability phenomenon in actual operation.

[0292] When the actuator is stuck, the actuator time constant is changed, and the simulation shows that the governor output signal is decoupled from the guide vane opening. The actual guide vane opening cannot reach the governor set value, and the system control accuracy decreases.

[0293] It can be seen that improper speed governor parameters and actuator failures will directly cause abnormal speed and output responses of the unit. By real-time monitoring of the distortion degree of the speed and output curves and constructing reasonable alarm logic, potential problems in the speed control system can be discovered in a timely manner.

[0294] This case uses the mechanism modeling method to analyze the correlation between various links in the equipment from the perspective of physical mechanism, revealing the impact of local defects on overall performance. Unlike pure data-driven correlation analysis, the mechanism model integrates domain knowledge and can more accurately explain the internal logic of equipment failures. The relationship between failure manifestations and root causes is more convincing.

[0295] Mechanism modeling simulates the dynamic behavior of equipment in different dimensions such as time domain and frequency domain, which can provide rich characteristic information and facilitate intelligent diagnosis. The mechanism model can also run online in real time, and through residual analysis and other means, it can timely discover hidden dangers such as system parameter drift and performance degradation, providing a reliable basis for equipment health management.

[0296] In addition, the causal relationship of the mechanism model is clear, the model parameters often have physical meaning, and are highly interpretable, making them easier for front-line personnel to understand and accept. Through mechanism simulation training, operation and maintenance personnel can deepen their understanding of complex equipment and strengthen their ability to judge equipment status and handle faults based on mechanisms.

[0297] By conducting multi-dimensional and multi-angle correlation analysis, we fully explore the intricate internal connections between equipment from data and knowledge. This correlation analysis is from the shallow to the deep, from the surface to the inside, taking into account the prior information such as physical structure and measurement point layout, and making full use of the statistical laws contained in the massive operation data. At the same time, it uses expert experience and mechanism models for evaluation and optimization, and finally forms a device correlation model with both breadth and depth.

[0298] Compared with monitoring a single device in isolation, the device association model depicts the internal mechanism of the system from a global perspective, enhancing the systematic and targeted nature of monitoring. On the one hand, the alarm of abnormal equipment can quickly locate the root cause and trace the upstream and downstream devices of the fault propagation. On the other hand, through the combined alarm of associated devices, the development trend of the situation can be predicted, practical control suggestions can be given, and intelligent alarm and intelligent decision-making can be achieved.

[0299] For example, when a transformer has an internal short circuit, the concentration of dissolved gas in the oil rises sharply, and parameters such as oil temperature and oil level also become abnormal. At the same time, the impact of the short circuit current will also cause a series of chain reactions such as a sudden drop in bus voltage and overcurrent tripping of the connecting switch. Traditional alarms only list the limit violations of each monitored quantity one by one, while correlation analysis reveals the entire process of fault propagation through a graphical model, quickly locks the fault location, reduces identification time, and provides a reliable basis for emergency repairs.

[0300] For example, when the generator excitation system is short-circuited, the correlation analysis can not only detect abnormal changes in the excitation current, but also predict subsequent effects such as frequency fluctuations. At the same time, the correlation model also gives the best excitation control measures to guide the on-duty personnel to adjust the excitation mode or remove the faulty excitation machine in time to minimize the impact of the accident. It can be seen that equipment correlation modeling makes the alarm more comprehensive and systematic, and can actively reveal some hidden dangers that are difficult to observe directly but are extremely risky, truly preventing them before they happen.

[0301] Furthermore, this application also includes: intelligent alarm generation and display; specifically, real-time monitoring of equipment operation data, combined with alarm models and strategies, to dynamically evaluate the health status of equipment. When an abnormality is detected, the corresponding alarm is activated; the corresponding alarm message is generated according to the alarm level, including attribute information such as the alarm object, alarm reason, and alarm time, and displayed on the monitoring interface; the alarm message is displayed in intelligent sorting according to the screening conditions and priorities, and attracts the attention of the on-duty personnel in various ways such as pop-up windows, sound and light.

[0302] When an abnormality is detected, the corresponding alarm is activated.

[0303] Furthermore, the present application also includes optimization of alarm strategies based on security reinforcement learning. Specifically, a Markov decision process (MDP) model of device health status and alarm status is established; wherein the state space S represents the health status of the device, the action space A represents whether to alarm, and the reward function R represents the utility brought by the alarm.

[0304] The reduced-order MDP method is used to map the high-dimensional state space to the low-dimensional feature space to reduce the search complexity of the state space; commonly used dimensionality reduction methods include principal component analysis and independent component analysis.

[0305] Safety constraints are introduced to impose penalties on alarm states that endanger system safety, and to reward alarm states that are beneficial to system safety. The reward function of the original MDP model is modified to ensure that the reinforcement learning process meets safety requirements.

[0306] Adopt reinforcement learning algorithms such as Q-learning and SARSA to learn the optimal warning strategy through continuous interaction with the environment. Among them, the state transition probability and reward function can be set according to actual data statistics or expert experience.

[0307] Apply the learned alarm strategy online and select the optimal alarm action based on the real-time health status of the device. At the same time, continuously optimize and update the strategy to adapt to changes in device status.

[0308] The corresponding alarm message is generated according to the alarm level, including attribute information such as alarm object, alarm reason, alarm time, etc., and displayed on the monitoring interface.

[0309] Alarm messages are displayed in intelligent order according to screening conditions and priorities, and attract the attention of on-duty personnel through pop-ups, sounds and lights, and other methods.

[0310] The following is a detailed explanation with an application case.

[0311] The safety alarms for the six units of a power station face two major challenges: first, there are thousands of unit status parameters with complex nonlinear correlations, which makes it difficult to model traditional rule-based or simple threshold-based alarm methods; second, blind and frequent alarms will interfere with normal operation, and untimely alarms may lead to accidents. How to optimize the alarm strategy while ensuring safety has become the key to intelligence.

[0312] To address this problem, we tried to use a safety reinforcement learning algorithm based on reduced-order MDP to optimize the alarm logic: First, we selected 200 representative parameters of the unit, such as vibration, temperature, and current, to construct a high-dimensional equipment state space. At the same time, we divided the alarm actions into "alarm" and "no alarm", and based on historical data and expert experience, we preliminarily set the safety utility values ​​of different state-action combinations.

[0313] Then, principal component analysis was used to reduce the 200-dimensional state to 20 dimensions, which greatly compressed the state space. In the MDP model after dimensionality reduction, we further introduced safety constraints, imposed severe penalties on alarm actions that caused unit tripping, and gave additional rewards to alarm actions that successfully avoided faults, guiding the direction of strategy search.

[0314] Next, we use the Q-learning algorithm for policy learning. Q-learning gradually improves the long-term utility of the policy by constantly trying different state-action combinations and updating the action value function Q(s, a) based on the benefits of environmental feedback. During the learning process, we balance exploration and exploitation in an ε-greedy way to avoid falling into a local optimum too early.

[0315] After 10,000 rounds of training, the algorithm converged to a safe and efficient alarm strategy. This strategy can adaptively adjust the alarm sensitivity according to the device status. When the device is in a minor abnormality, the strategy tends to wait and see to avoid excessive alarms; when the device has multiple indicators that are abnormal continuously, the strategy will decisively issue an alarm and attach a severity level prompt.

[0316] In actual applications, the intelligent alarm strategy has reduced the missed alarm rate by 50% and the false alarm rate by 80% compared with traditional methods, greatly improving the reliability and safety of alarms. The strategy will continue to optimize with the dynamic changes of the equipment, so that the operator can grasp the health status of the equipment in time and respond to unexpected risks calmly.

[0317] This case demonstrates the advantages of safety reinforcement learning in equipment alarm optimization: the equipment status is efficiently modeled through reduced-order MDP, the inherent rules of state-action mapping are extracted through reinforcement learning, and domain knowledge is embedded through safety constraints, making the alarm strategy more intelligent, safe, and explainable. This algorithm provides a new idea for improving the automation and intelligence level of equipment alarms.

[0318] Furthermore, this application also includes the query and management of alarm information; specifically, current and historical alarms are classified and counted, and trend analysis is performed to generate alarm tables to provide data support for equipment maintenance; on-duty personnel at different positions are granted corresponding alarm query and processing permissions. Major alarms must be approved before they can be eliminated; alarm strategies and knowledge bases support online editing and self-learning optimization, forming an expert system to continuously improve the intelligence level of the system.

[0319] Example 1 (object-oriented alarm screening): In the interactive topology diagram of the distribution automation master station system, each device object has an independent alarm status light. The operator can quickly screen abnormal devices according to voltage level, device type, etc. When the bus voltage exceeds the limit, the bus object flashes red to alarm. Select the object to view key information such as bus voltage and connection switch, which is convenient for timely disposal.

[0320] Example 2 (conditional alarm): For the main transformer oil temperature alarm, it is generally necessary to verify the transformer load at the same time. The conditional statement can be expressed as: when the transformer load is greater than 80% of the rated value for 30 consecutive minutes, and the oil temperature exceeds 85 degrees for 5 minutes, the overtemperature alarm is triggered. This greatly reduces unnecessary frequent alarms.

[0321] Example 3 (related quantity alarm): When the generator stator winding temperature exceeds the limit alarm, the system automatically displays the generator stator current, cooling water temperature, unit active power and other related quantities. The duty officer can quickly analyze the cause of the temperature rise without searching and take effective measures such as controlling the machine or adjusting the cooling water. The alarm response time is shortened from minutes to seconds.

[0322] This intelligent alarm method makes full use of big data and artificial intelligence technology to realize the automation and intelligence of hydropower station alarms, and improve the pertinence, reliability and operability of alarms. The system can be widely used in the centralized control operation of large and medium-sized hydropower stations, bringing a new management model for equipment status monitoring, fault warning and auxiliary decision-making.

[0323] In a second aspect, the embodiment of the present disclosure further provides an intelligent alarm system for a hydropower station computer monitoring system, comprising:

[0324] A preprocessing module is used to preprocess the historical operation data of various equipment in the hydropower station to obtain a first data set; the operation data includes at least one of a state quantity, a telemetry quantity, and a telesignal quantity;

[0325] A building module, used for building a data model based on the first data set;

[0326] A first associated information acquisition module is used to acquire first associated information based on the physical structure of various types of equipment in the hydropower station; the first associated information includes the physical connection relationship between the equipment, the causal chain model and the fault knowledge base;

[0327] A second associated information acquisition module is used to acquire second associated information based on the operation data of various equipment in the hydropower station; the second associated information includes statistical correlation and frequent co-occurrence pattern information;

[0328] A construction module, used to construct a device association model based on the first association information, the second association information, and the data model;

[0329] The alarm module is used to monitor certain operating data of the target device in real time and dynamically obtain the operating status information of the target device based on the device association model;

[0330] When there is an abnormality in the running status information, an alarm is triggered based on the preset alarm strategy.

[0331] It should be noted that the intelligent alarm method of the hydropower station computer monitoring system disclosed in the first aspect of this application is applicable to the intelligent alarm system of the hydropower station computer monitoring system disclosed in the second aspect of this application, so it will not be described in detail here.

[0332] The computer device according to the embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0333] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory, so that the computer device executes all or part of the steps of the intelligent alarm method of the hydropower station computer monitoring system of each embodiment of the present disclosure.

[0334] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.

[0335] Fig.11 A schematic diagram of the structure of a computer device suitable for implementing the embodiments of the present disclosure is shown. Fig.11 The computer device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0336] like Fig.11 As shown, the computer device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage device into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer device are also stored. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0337] Typically, the following devices can be connected to the I / O interface: input devices such as sensors or visual information acquisition devices; output devices such as display screens; storage devices such as tapes, hard disks, etc.; and communication devices. The communication device can allow the computer device to communicate with other devices (such as edge computing devices) wirelessly or by wire to exchange data. Fig.11A computer device having various devices is shown, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0338] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, all or part of the steps of the intelligent alarm method of the hydropower station computer monitoring system of the embodiment of the present disclosure are executed.

[0339] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0340] The computer-readable storage medium according to the embodiment of the present disclosure stores non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by the processor, all or part of the steps of the intelligent alarm method of the hydropower station computer monitoring system of each embodiment of the present disclosure are executed.

[0341] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).

[0342] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0343] As described above, the present invention can be preferably implemented.

[0344] All features disclosed in all embodiments in this specification, or steps in all methods or processes implicitly disclosed, except for mutually exclusive features and / or steps, can be combined and / or expanded or replaced in any manner.

[0345] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. According to the technical essence of the present invention, within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent alarm method for a hydropower station computer monitoring system, characterized in that: The following steps are involved: S100, preprocessing historical operation data of the hydropower station equipment to obtain a first data set; S200, establishing a data model based on the first data set; S300, acquiring first associated information based on the physical structure of various equipment in the hydropower station; S400, acquiring second related information based on operation data of various equipment in the hydropower station; S500, constructing a device association model based on the first association information, the second association information, and the data model; S600, monitor the operating data of the target device in real time, and dynamically obtain the operating status information of the target device based on the device association model; when there is an abnormality in the operating status information, trigger an alarm based on a preset alarm strategy.

2. The intelligent alarm method of the hydropower station computer monitoring system according to claim 1 is characterized in that: Step S200 includes the following steps: S211, extracting feature data from the first data set; wherein the feature data includes one or more of mean, variance, peak-to-peak value, and slope; S212, performing feature selection processing on the feature data to obtain a second data set; constructing a sample set based on the second data set, typical normal samples and typical fault samples; wherein the sample set includes a training set and a test set; S213, construct a supervised learning model, train the supervised learning model through the training set, test the supervised learning model that meets the training conditions on the test set according to preset indicators, dynamically tune the supervised learning model based on the test results, and record the dynamically tuned supervised learning model as the data model.

3. The intelligent alarm method of the hydropower station computer monitoring system according to claim 1 is characterized in that: Step S300 includes the following steps: S310, extracting the physical connection relationship between devices according to the system topology structure in the hydropower station; S320, acquiring fault prediction propagation path information based on the physical connection relationship; S330, acquiring the first associated information based on the fault prediction propagation path information, historical fault information and corresponding handling measures.

4. The intelligent alarm method of the hydropower station computer monitoring system according to claim 3 is characterized in that: Step 400 includes the following steps: S410, calculating the statistical correlation between operating parameters of different devices; S420, obtaining frequent co-occurrence pattern information of device status based on a data mining algorithm; wherein the data mining algorithm includes one or more of an Apriori algorithm and an FP-growth algorithm; S430, based on the frequent co-occurrence pattern information and the expert database information, the second association information is obtained; wherein the second association information includes strong association pattern information with preset value screened out from the massive association rules of the hydropower station equipment.

5. The intelligent alarm method of the hydropower station computer monitoring system according to claim 4 is characterized in that: Step 420 includes the following steps: S421, determining an input data set for a data mining algorithm based on statistical correlation; S422, obtaining a target frequent itemset based on the input data set; S423, obtaining frequent co-occurrence pattern information based on the target frequent item set.

6. The intelligent alarm method of the hydropower station computer monitoring system according to claim 5 is characterized in that: Step 421 includes the following steps: A100, select the running status parameters of the equipment as the transaction set for association analysis; A200, discretizes the running status parameters, converting the running status parameters into Boolean values ​​or enumeration values; A300,consists of all device states at each time sampling point into a transaction record, and several transaction records constitute the input data set of the data mining algorithm.

7. The intelligent alarm method of the hydropower station computer monitoring system according to claim 6 is characterized in that: Step 422 includes the following steps: B100, set a minimum support threshold; based on the input data set, obtain the occurrence frequency of each state; remove the item sets with support lower than the minimum support threshold in the input data set to obtain the first frequent item set; wherein the first frequent item set is a frequent set of a single item; B200, performing a self-join operation on the first frequent item set to generate a first candidate item set containing two items; scanning the input data set to obtain the actual support of the first candidate item set; filtering out the first candidate item set whose actual support is lower than the minimum support threshold to obtain a second frequent item set; wherein the actual support is the actual number of occurrences of each first candidate item set; B300, perform a self-join operation on the Pth frequent item set to generate a Pth candidate item set containing two items; scan the input data set to obtain the actual support of the Pth candidate item set, each actual support is the actual number of occurrences of each type of the Pth candidate item set; filter out the Pth candidate item set whose actual support is lower than the minimum support threshold to obtain the P+1th frequent item set; when the actual support of the Pth candidate item set is lower than the minimum support threshold, record the Pth frequent item set as the target frequent item set; where P≥2.

8. The intelligent alarm method of the hydropower station computer monitoring system according to claim 7 is characterized in that: Step 423 includes the following steps: C100, enumerate all non-empty subsets of the target frequent itemset; each non-empty subset and its complement constitute a candidate association rule; C200, calculate the confidence of each candidate association rule; C300, filter out candidate association rules whose confidence is less than the minimum confidence threshold to obtain frequent co-occurrence pattern information.

9. The intelligent alarm method of the hydropower station computer monitoring system according to claim 8 is characterized in that: Step 430 includes the following steps: S431, sorting the rules in the frequent co-occurrence pattern information according to support and confidence; S432, filtering out the representational rules in the strong association rule set based on the expert database information to obtain second association information.

10. The intelligent alarm method of a hydropower station computer monitoring system according to any one of claims 1 to 9, characterized in that: In step S600, dynamically acquiring the operating status information of the target device based on the device association model includes the following steps: D100, a connectivity indicator for obtaining operation data based on a device association model; D200, obtains the centrality index of each node based on the device association model; each node corresponds to a device; D300,uses community discovery algorithm to analyze connectivity index and centrality index,and obtains the device subgroups whose actual closeness exceeds the,closeness threshold; D400, analyze the device subgroup to obtain operation status information; wherein the operation status information includes potential fault nodes and corresponding fault impacts and abnormal future expansion paths.

Images