Electric power facility predictive maintenance method and system

Through real-time sensing monitoring and prediction models, abnormal points in power facilities are identified and maintenance decisions are optimized, which solves the problems of untimely response and low efficiency of traditional maintenance methods, and improves the stability and safety of equipment.

CN119067265AInactive Publication Date: 2024-12-03WUXI XINENG REAL ESTATE MANAGEMENT CO LTD +1

Patent Information

Application Number
CN202411553840.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-04
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional power facility maintenance methods are not timely and inefficient, making it difficult to ensure stable operation of equipment under high load complex environments.

Method used

Through real-time sensing monitoring of power facilities operation data, a prediction model is built to predict state, a facility state prediction parameter is generated, override and identify abnormal points and generate fault signals, make maintenance decisions, and combine the prediction model to perform abnormal detection optimization maintenance decisions to achieve prediction Sexual maintenance.

Benefits of technology

It realizes early prevention and efficient maintenance, improves the stability and safety of equipment operation, and solves the problems of untimely response and low maintenance efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067265A_ABST
    Figure CN119067265A_ABST
Patent Text Reader

Abstract

The invention discloses an electric power facility predictive maintenance method and system, and relates to the technical field of electric power maintenance, and the method comprises the steps: obtaining a facility operation data set through real-time sensing monitoring; performing analysis based on the facility operation data set, constructing a prediction model, performing state prediction on the electric power facility, and obtaining a plurality of facility state prediction parameters; performing coverage traversal identification on the facility state prediction parameters to generate a plurality of labels to be detected; performing anomaly analysis based on the to-be-detected label, generating a fault signal, and making a first maintenance decision based on the facility state prediction parameter in combination with the fault signal; and performing anomaly detection on the power facility execution maintenance decision in combination with the prediction model, generating a detection feedback result, optimizing the first maintenance decision according to the detection feedback result, and obtaining a second maintenance decision to perform predictive maintenance response on the power facility. According to the invention, the technical problems of untimely reaction and low maintenance efficiency in the prior art are solved, and the technical effect of improving the stability and safety of equipment operation is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power maintenance, and in particular to a predictive maintenance method and system for power facilities. Background Art

[0002] With the continuous growth of electricity demand and the increasing complexity of power systems, the operation and maintenance of power facilities are particularly important in modern society. Power facilities are a vital component of the power system, and their operating status directly affects the transmission and supply of electricity. However, with the expansion of the scale of power facilities and the extension of operating time, how to ensure the stable operation of these equipment under high load and complex environments has become a difficult problem. Traditional power facility maintenance methods are mainly based on regular maintenance and post-repair, which have problems such as untimely response and low maintenance efficiency. Summary of the invention

[0003] The present application provides a predictive maintenance method and system for electric power facilities, which are used to solve the technical problems of untimely response and low maintenance efficiency in the prior art.

[0004] In view of the above problems, the present application provides a method and system for predictive maintenance of power facilities.

[0005] In a first aspect of the present application, a method for predictive maintenance of electric power facilities is provided, the method comprising:

[0006] By performing real-time sensor monitoring on power facilities, a facility operation data set is obtained; based on the facility operation data set, an analysis is performed to build a prediction model, and the state of the power facilities is predicted through the prediction model to obtain multiple facility state prediction parameters; the multiple facility state prediction parameters are covered and traversed, and the traversed node set is evaluated and identified to generate multiple labels to be detected; an anomaly analysis is performed based on the multiple labels to be detected to generate a fault signal, and a first maintenance decision is made based on the multiple facility state prediction parameters and the fault signal; the first maintenance decision is executed on the power facility and the prediction model is combined to perform anomaly detection, and a detection feedback result is generated; the first maintenance decision is optimized according to the detection feedback result, and a second maintenance decision is obtained to respond to the predictive maintenance of the power facility.

[0007] A second aspect of the present application provides a predictive maintenance system for electric power facilities, the system comprising:

[0008] A real-time monitoring module, wherein the real-time monitoring module obtains a facility operation data set by performing real-time sensor monitoring on the power facilities; a state prediction module, wherein the state prediction module performs analysis based on the facility operation data set, constructs a prediction model, predicts the state of the power facilities through the prediction model, and obtains multiple facility state prediction parameters; a coverage traversal module, wherein the coverage traversal module performs coverage traversal on the multiple facility state prediction parameters, evaluates the traversal node set for identification, and generates multiple labels to be detected; an anomaly analysis module, wherein the anomaly analysis module performs an anomaly analysis based on the multiple labels to be detected, generates a fault signal, and formulates a first maintenance decision based on the multiple facility state prediction parameters combined with the fault signal; an anomaly detection module, wherein the anomaly detection module performs anomaly detection on the power facilities by executing the first maintenance decision in combination with the prediction model, generates a detection feedback result, optimizes the first maintenance decision according to the detection feedback result, and obtains a second maintenance decision to perform predictive maintenance response on the power facilities.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The present application obtains a facility operation data set by real-time sensor monitoring of power facilities; analyzes the facility operation data set, builds a prediction model, predicts the state of the power facilities through the prediction model, and obtains multiple facility state prediction parameters; covers and traverses multiple facility state prediction parameters, evaluates and traverses the node set for identification, and generates multiple tags to be detected; performs abnormal analysis based on multiple tags to be detected, generates a fault signal, and makes a first maintenance decision based on multiple facility state prediction parameters combined with the fault signal; performs abnormal detection on the power facility execution maintenance decision combined with the prediction model, generates a detection feedback result, optimizes the first maintenance decision according to the detection feedback result, and obtains a second maintenance decision to perform predictive maintenance response on the power facility. The present invention solves the technical problems of untimely response and low maintenance efficiency in the prior art, monitors the operation data of power facilities through real-time sensor, builds a prediction model for state prediction, generates facility state prediction parameters, covers and traverses these parameters, identifies abnormal points and generates fault signals, thereby making maintenance decisions, and obtains a second maintenance decision by executing the maintenance decision combined with the prediction model for feedback optimization, so as to achieve early prevention and efficient maintenance, and achieve the technical effect of improving the stability and safety of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0012] Figure 1 A schematic diagram of a process flow of a predictive maintenance method for electric power facilities provided in an embodiment of the present application;

[0013] Figure 2 A schematic diagram of the structure of a predictive maintenance system for electric power facilities provided in an embodiment of the present application.

[0014] Explanation of the accompanying drawings: real-time monitoring module 11, state prediction module 12, coverage traversal module 13, anomaly analysis module 14, anomaly detection module 15. DETAILED DESCRIPTION

[0015] The present application provides a predictive maintenance method and system for power facilities, aiming to solve the technical problems of untimely response and low maintenance efficiency in the prior art. Through real-time sensing and monitoring of the operating data of power facilities, a prediction model is constructed to predict the state, and facility state prediction parameters are generated. These parameters are covered and traversed, abnormal points are identified and fault signals are generated, so as to make maintenance decisions. By executing maintenance decisions and combining them with the prediction model for feedback optimization, a second maintenance decision is obtained, thereby achieving early prevention and efficient maintenance, thereby achieving the technical effect of improving the stability and safety of equipment operation.

[0016] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0017] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0018] Embodiment 1, as Figure 1 As shown, the present application provides a method for predictive maintenance of electric power facilities, the method comprising:

[0019] Step S100: Obtain facility operation data set by performing real-time sensor monitoring on power facilities.

[0020] In the embodiment of the present application, firstly, the power facilities are monitored in real time by pre-installed temperature sensors, current sensors, voltage sensors, etc., to obtain a facility operation data set. The facility operation data set includes temperature data, current data, voltage data, etc.

[0021] Step S200: Analyze the facility operation data set to construct a prediction model, predict the state of the power facility through the prediction model, and obtain multiple facility state prediction parameters.

[0022] In an embodiment of the present application, a data trend graph is first drawn based on the facility operation data set, and then a correlation analysis is performed to generate autocorrelation data and partial autocorrelation data. The autocorrelation data is then smoothed using a sliding average method to generate a smoothing index, and the autocorrelation data and partial autocorrelation data are combined to perform a time series analysis and construct an analysis data list. Finally, based on the analysis data list, the autocorrelation and partial autocorrelation data are further divided to construct a prediction model. Once the model is constructed, the facility is predicted to be in a state and multiple facility state prediction parameters are generated. The facility state prediction parameters include the future operating state of the facility at different time points, such as future current fluctuations, temperature rise trends, etc.

[0023] Furthermore, in the method provided in the embodiment of the application, the analysis is performed based on the facility operation data set to construct a prediction model, and the method further includes:

[0024] Data is captured based on the facility operation data set, and a data trend graph is drawn; correlation analysis is performed on the facility operation data set based on the data trend graph to generate autocorrelation data and partial autocorrelation data; based on the partial autocorrelation data, the autocorrelation data is sliding averaged to generate a smoothing index; according to the smoothing index, time series analysis is performed in combination with the autocorrelation data and the partial autocorrelation data to construct an analysis data list; based on the analysis data list, the autocorrelation data and the partial autocorrelation data are divided to construct the prediction model.

[0025] In the embodiment of the present application, in order to build a prediction model, the facility operation data set is first analyzed, and the data is processed by data capture technology to extract the operation parameters, such as temperature, voltage, current, vibration, etc. Then, data visualization tools such as Matplotlib are used to display these parameters in a graphical form and draw a data trend chart.

[0026] Next, the data in the data trend chart are analyzed for autocorrelation using statistical methods such as the Pearson correlation coefficient or the Spearman correlation coefficient to obtain autocorrelation data. At the same time, partial autocorrelation analysis is performed to generate partial autocorrelation data by using PACF to eliminate the influence of the middle point of the time series on the correlation.

[0027] After obtaining the autocorrelation and partial autocorrelation data, the sliding average algorithm is applied to smooth the autocorrelation data. By calculating the average value within the rolling window in the time series, short-term fluctuations in the data are eliminated to obtain a smoothing index. Then, time series analysis techniques such as ARIMA are used to combine the smoothing index, autocorrelation data, and partial autocorrelation data for in-depth analysis to obtain an analysis data list, which contains prediction results at different time points, such as the future trend of key parameters such as voltage and current.

[0028] After constructing the analysis data list, the autocorrelation data and partial autocorrelation data are divided into training set and test set by random partitioning, with the training set accounting for 70% and the test set accounting for 30%. Then, machine learning algorithms such as random forest are used to train the model through the training set, learn the patterns in the data, adjust the model parameters, and improve the prediction accuracy. Finally, the divided test set is used to verify the model. If the performance of the model on the test set meets expectations, it is considered that the model has good generalization ability, the training is completed, and the prediction model is obtained.

[0029] Step S300: performing a coverage traversal on the plurality of facility status prediction parameters, evaluating the traversal node set for identification, and generating a plurality of labels to be detected.

[0030] In an embodiment of the present application, multiple facility status prediction parameters are covered and traversed, and each status node is checked using a depth-first search to generate a traversed node set, each node representing a specific facility status data point, such as temperature, voltage, or vibration status information at a certain moment. For each traversed node, a status evaluation is performed based on a pre-set threshold or normal operating range. For example, whether a temperature value exceeds the normal operating range, or whether a current value has abnormal fluctuations. In this way, nodes that are beyond the normal range are identified, and these potential abnormal points are recorded, generating multiple tags to be detected.

[0031] Step S400: performing an abnormality analysis based on the multiple tags to be detected, generating a fault signal, and formulating a first maintenance decision based on the multiple facility status prediction parameters and the fault signal.

[0032] In the embodiment of the present application, starting from extracting the parameters to be detected, the extracted facility status prediction parameters are analyzed based on multiple tags to be detected. First, a set of facility status indicators is set, and then the extracted parameters are evaluated for fluctuations, and their variation ranges are analyzed to generate parameter fluctuation results. Through these fluctuation results, the fluctuation extremes of each parameter are determined, and multiple possible abnormal status data points are identified.

[0033] Next, feature analysis is performed on the abnormal data points to extract the features of each abnormal state and compare them with the historical fault data. A set of similarity coefficients is generated through similarity determination. These coefficients are used to measure the degree of match between the current abnormal state and the historical faults. Through fault verification, a fault signal is generated based on the similarity to mark the confirmed fault state.

[0034] Then, all the facility status prediction parameters are traversed and matched with the verified abnormal status data points to determine a set of facility fault status prediction parameters. Based on these parameters and fault signals, a comprehensive evaluation is performed to generate a comprehensive score for facility faults. Finally, a source analysis is performed based on the fault score to determine the specific facility parts that need maintenance, and the facility maintenance cycle is set based on these data to finally generate the first maintenance decision.

[0035] Furthermore, in the method provided in the embodiment of the application, abnormal analysis is performed based on the multiple tags to be detected to generate a fault signal, and the method further includes:

[0036] Based on the multiple tags to be detected, the multiple facility status prediction parameters are extracted to obtain multiple parameters to be detected; a facility status index is set, and fluctuation evaluation is performed on the multiple parameters to be detected according to the facility status index to obtain parameter fluctuation results; parameter fluctuation extreme values ​​are determined according to the parameter fluctuation results, and multiple abnormal state data points are determined according to the parameter fluctuation extreme values; feature analysis is performed based on the multiple abnormal state data points to generate abnormal state features, and similarity judgment is performed on the abnormal state features and historical fault features to generate multiple similarity coefficients; fault verification is performed on the multiple abnormal state data points based on the multiple similarity coefficients, and the fault signal is determined according to the verification result.

[0037] In an embodiment of the present application, first, operating data related to the tag to be detected is extracted from multiple facility status prediction parameters to obtain multiple parameters to be detected, including temperature, current, voltage, vibration, etc. Next, technical experts set facility status indicators, which represent the normal working range or allowed operating threshold of the equipment. After that, the fluctuation of multiple parameters to be detected is evaluated according to the facility status indicators. Specifically, by using statistical analysis techniques such as standard deviation calculation to measure volatility, parameter fluctuation results are generated. According to the parameter fluctuation results, the fluctuation extremes of each parameter are further analyzed and determined. The fluctuation extremes represent the maximum or minimum values ​​that occur during the operation of the equipment, which may indicate an abnormal state. This step uses a threshold detection algorithm, such as a Z-Score algorithm based on statistical distribution, to calculate the Z-Score value of each data point, evaluate the gap between the data point and the mean, and if the Z-Score value exceeds the set threshold, the data point is marked as an abnormal state data point.

[0038] Then, feature analysis is performed on the abnormal state data points, and the key features of each abnormal data point are extracted using principal component analysis to generate abnormal state features. Next, the similarity between the extracted abnormal state features and the historical fault features is determined by Euclidean distance, and the distance between the current abnormal features and the historical fault features is calculated. The shorter the distance, the higher the similarity. In this way, multiple similarity coefficients are generated.

[0039] Based on multiple similarity coefficients, multiple abnormal state data points are verified for faults. At this time, the similarity threshold is used as the judgment standard to obtain a preset threshold, such as 0.05. When the Euclidean distance of a certain abnormal point is less than 0.05, it is considered that the abnormal state has a high similarity with the historical fault. If the distance is greater than 0.05, it is considered that the current abnormal point does not match the historical fault very well, which may be just an accidental event and does not need to be immediately regarded as a fault. Secondly, the frequency of occurrence of abnormal points is considered. In order to prevent one-time fluctuations or occasional abnormalities from being mistakenly marked as faults, a frequency threshold is set. For example, it is set that only when a certain abnormal point appears at least 3 times within 5 detection cycles, the abnormality is considered to be persistent. If the abnormal point only appears once in these 5 detections, it is judged that this may be an accidental situation and will not be marked as a fault signal. Through this verification rule combining the similarity threshold and the frequency of occurrence, occasional abnormalities and persistent faults are distinguished. Finally, if the abnormal point meets these preset conditions, it is regarded as a real fault and a fault signal is generated; otherwise, it is considered that the abnormality may not pose a threat and no fault signal is generated.

[0040] Furthermore, in the method provided in the embodiment of the application, formulating a first maintenance decision based on the plurality of facility status prediction parameters combined with the fault signal further includes:

[0041] Traversing the multiple facility status prediction parameters and matching them with the verified abnormal status data points, a plurality of facility fault status prediction parameters are determined; a comprehensive evaluation is performed based on the multiple facility fault status prediction parameters in combination with the fault signal to generate a comprehensive score for the facility fault; tracing the source according to the comprehensive score for the facility fault is performed to determine a plurality of data to be maintained; setting the first facility maintenance cycle for the plurality of data to be maintained according to maintenance rules, and constructing the first maintenance decision according to the first facility maintenance cycle.

[0042] In the embodiment of the present application, firstly, multiple facility status prediction parameters are traversed and matched with the verified abnormal status data points. Specifically, the prediction parameters of each component are matched with the abnormal points identified during the fault verification process. If the temperature abnormality of a component has passed the fault verification, the temperature prediction parameters of the component correspond to the abnormal data and the match is successful. Through this matching, multiple facility fault status prediction parameters are determined, including temperature, voltage, current, vibration and other prediction data related to specific faults.

[0043] Next, the severity of the parameters predicting the fault status of these facilities is further evaluated through comprehensive evaluation and fault scoring. This step uses the analytic hierarchy process and weighted scoring method. First, the factors affecting equipment failure are divided into multiple levels, such as fault severity, fault frequency, and impact range. Fault severity includes the magnitude of temperature excess and the range of current fluctuation. Fault frequency is the frequency of occurrence of abnormal data points, such as whether temperature anomalies of the same component occur repeatedly. The impact range is the potential impact of the fault on other components or equipment. Each level is scored according to the weight, for example, the fault severity may account for 50% of the weight, the fault frequency accounts for 30%, and the impact range accounts for 20%. Through the weighted scoring method, these scores are weighted and combined to generate a comprehensive score for facility failure.

[0044] After obtaining the comprehensive score of the facility failure, a traceability analysis is conducted to determine the root cause of the failure through cause-and-effect analysis combined with pattern recognition technology. The possible causes of the failure are found by tracing back the historical data of the equipment. For example, if the temperature failure score of a component is high, the historical operation data of the component will be traced back to analyze whether the component is prone to temperature abnormalities under specific conditions, such as high load. Through pattern recognition technology, historical failure patterns are compared to find patterns similar to the current failure, so as to determine the root cause of the current failure. These reasons may include high-load operation of the equipment, changes in environmental conditions, etc. Through traceability analysis, maintenance data is generated, which clearly points out the equipment components or areas that need to be inspected and maintained. For example, it may be necessary to check the cooling system of a component, or replace or maintain a sensor of the equipment.

[0045] Next, use condition-based maintenance technology to set the first maintenance cycle for each component to be maintained. CBM dynamically adjusts the maintenance cycle based on the current status of the equipment and historical fault records. Decide when to perform maintenance based on the health status of the equipment, fault score, and historical maintenance data. For components with high scores and frequent failures, set a shorter maintenance cycle, such as an inspection every 3 months; for components with low scores, you can set a longer maintenance cycle, such as an inspection every 12 months. The maintenance cycle will be adjusted according to the actual operating status of the equipment to ensure effective maintenance before a failure occurs.

[0046] Finally, the maintenance plan is constructed using the rule engine and sequential scheduling method. The rule engine directly generates the maintenance sequence based on the priority and maintenance cycle, and sequential scheduling ensures that the tasks are arranged reasonably on time. Specifically, the maintenance tasks are arranged in the order of the equipment with higher scores first and the equipment with lower scores last in the maintenance cycle of the first facility. Through this process, the first maintenance decision is obtained.

[0047] Step S500: Execute the first maintenance decision on the power facility and perform abnormality detection in combination with the prediction model to generate a detection feedback result, optimize the first maintenance decision according to the detection feedback result, and obtain a second maintenance decision to perform a predictive maintenance response on the power facility.

[0048] In an embodiment of the present application, firstly, simulated maintenance of the power facilities is performed according to the first maintenance decision to generate a simulated maintenance result of the facilities. Based on the previous prediction model and in combination with the simulated maintenance result of the facilities, the system performs a secondary prediction to generate a set of simulated maintenance status parameters of the facilities. The system traverses the set of simulated maintenance status parameters of the facilities and performs anomaly detection in combination with historical fault characteristics. If certain parameters are beyond the normal range, they are marked as abnormal and anomaly detection results are generated. If there are no abnormalities, it is proved that the previously generated first maintenance decision is comprehensive for fault maintenance, and the first maintenance decision can be used. If there are abnormalities, an impact analysis is performed on the detected anomalies, and the impact coefficient of each anomaly is calculated. The abnormal detection results are arranged in descending order according to the impact coefficients to generate an abnormal detection sequence, and the abnormal detection sequence is used as the detection feedback result.

[0049] Next, the self-inspection cycle is set according to the test feedback results, and the first facility maintenance cycle is updated to generate the second facility maintenance cycle. Next, the maintenance rules are optimized, and the maintenance data is analyzed in combination with the test feedback to generate the maintenance optimization effect. Finally, the first maintenance decision is adjusted according to the optimization effect, and the second maintenance decision is determined.

[0050] Furthermore, in the method provided in the embodiment of the application, executing the first maintenance decision on the power facility in combination with the prediction model to perform abnormality detection also includes:

[0051] Execute the first maintenance decision to perform simulated maintenance on the power facilities and generate a facility simulation maintenance result; perform secondary prediction based on the prediction model and the facility simulation maintenance result to generate a facility simulation maintenance state parameter set; traverse the facility simulation maintenance state parameter set and perform anomaly detection in combination with the historical fault characteristics to generate anomaly detection results; perform anomaly impact analysis on the anomaly detection results to obtain multiple impact coefficients; arrange the anomaly detection results in descending order according to the multiple impact coefficients to generate an anomaly detection sequence, and add the anomaly detection sequence to the detection feedback result.

[0052] In an embodiment of the present application, firstly, simulation modeling is used to simulate the execution of previously made maintenance decisions. The simulated operation includes the repair, adjustment and restoration of the operating status of the equipment. The simulation model is calculated based on the real parameters of the equipment to simulate the changes in various key operating parameters during the maintenance process, such as temperature, current, etc. This process generates a facility simulation maintenance result, which contains the operating data of the equipment after maintenance. Next, the ARIMA model is used to perform time series prediction on the simulated maintenance results of the equipment. The model predicts the future status of key parameters such as temperature and current based on the time series data in the maintenance results, and generates a facility simulation maintenance status parameter set.

[0053] Next, the threshold detection is performed on the parameter set of facility simulation maintenance status generated by the secondary prediction. Each key parameter, such as temperature and voltage, has a preset normal operating range, and parameters exceeding this range are marked as abnormal. For example, if the temperature exceeds 90°C and the current exceeds the rated range, it is judged as abnormal. The abnormal detection result is obtained through this process.

[0054] Next, fault tree analysis is used to analyze the impact of abnormalities on the overall operation of the equipment, and an impact coefficient is assigned to each abnormality through clear calculation steps. Specifically, a fault tree is first established to represent the possible fault logic of the equipment. The fault tree starts from the final equipment failure at the top, such as equipment shutdown and system failure, and expands downward layer by layer to trace the basic events that caused the failure, such as temperature abnormalities, current fluctuations, etc. Next, based on historical data and equipment operation records, the probability of occurrence of each basic event is defined. For example, the historical probability of temperature abnormality of a certain device can be set to 0.15, and the probability of voltage fluctuation is 0.05. The combined impact of these basic events is then calculated through AND and OR gate logic. Under AND gate logic, multiple abnormal events occurring simultaneously will cause equipment failure, and its impact coefficient is obtained by multiplying the probability of these events. For example, the probability of temperature and current abnormalities occurring simultaneously is the probability of temperature abnormality multiplied by the probability of current abnormality. Under OR gate logic, the occurrence of any one event may cause a failure, and the impact coefficient is calculated by the formula Calculation means that as long as one event occurs, the device may fail. , is the probability of a basic event occurring.

[0055] Through this combination of logic gates, calculations are performed layer by layer until the contribution of each basic abnormal event to the overall failure of the device is obtained, generating an impact coefficient. The impact coefficient indicates the threat of each abnormality to the final failure of the device. For example, the impact coefficient of temperature anomaly may be 0.8, while the impact coefficient of current fluctuation may be 0.2, and the former poses a greater risk to the device. According to the impact coefficient of each abnormality, a sorting algorithm is used to sort the abnormality detection results in descending order of the impact coefficient to generate an abnormality detection sequence, and finally the abnormality detection sequence is added to the detection feedback result.

[0056] Furthermore, in the method provided in the embodiment of the application, traversing the facility simulation maintenance status parameter set and combining the historical fault characteristics to perform anomaly detection and generate anomaly detection results also includes:

[0057] A facility fault threshold is set based on the historical fault characteristics; the facility simulation maintenance status parameter set is traversed and compared with the facility fault threshold to determine whether the facility simulation maintenance status parameter set is greater than or equal to the facility fault threshold; if there are facility simulation maintenance status parameters greater than or equal to the facility fault threshold in the facility simulation maintenance status parameter set, an abnormal identification is generated, and the abnormal point of the power facility is determined based on the facility simulation maintenance status parameters greater than or equal to the facility fault threshold, the abnormal identification is mapped with the abnormal point, and the abnormal detection result is generated.

[0058] In an embodiment of the present application, the fault threshold of the power facility is first set by historical data analysis and statistical methods. Specifically, fault records are extracted from the historical operation data of the power facility, and statistical methods such as mean value, standard deviation, etc. are used to analyze the failure mode of each key parameter, such as temperature, voltage, and current. By analyzing and identifying the conditions under which the equipment is prone to failure in past operation, for example, if the equipment has problems when the temperature exceeds 85°C, 85°C is set as the fault threshold of the temperature. Other key parameters such as voltage and current are also set with corresponding thresholds through similar analysis. Through this process, the setting of the facility fault threshold is completed.

[0059] Then, the data in the parameter set of the simulated maintenance status of the facility is traversed, and the parameter traversal algorithm is used to detect the real-time operating parameters of the power facility one by one. The simulated maintenance status parameter set contains the current status data of all key operating parameters of the equipment after maintenance. The traversal algorithm extracts each parameter in sequence to ensure that the key operating data of all equipment can be analyzed one by one and compared with the corresponding fault threshold. During the traversal process, each parameter is compared with the set fault threshold one by one using conditional judgment and comparison operations. For each key parameter, check whether the parameter value is greater than or equal to the fault threshold. If a parameter is detected to exceed the set threshold, the parameter is marked as abnormal. For example, if the current temperature of the power facility is 87°C and the temperature threshold is 85°C, the temperature is judged to be abnormal; similarly, when the voltage and current parameters exceed their respective set thresholds, they will also be marked as abnormal.

[0060] If there are facility simulation maintenance status parameters greater than or equal to the facility fault threshold in the facility simulation maintenance status parameter set, the corresponding abnormal identification is generated through identification generation and data mapping technology and the abnormal point of the equipment is determined. Specifically, when a key parameter exceeds the threshold, an abnormal identification is generated. For example, when the temperature exceeds the threshold, a "temperature abnormality" identification is generated. Similarly, for abnormal current or voltage conditions, identifications such as "current abnormality" or "voltage abnormality" are generated. At the same time, the point where the abnormality occurs is located by combining the operating data and location information of the equipment in the simulation state. Each simulated maintenance status parameter is associated with a specific location or component of the equipment. For example, during the simulation process, if a temperature parameter is marked as abnormal, the location of the corresponding temperature sensor is found through the sensor data of the equipment. In this way, the abnormal temperature value is associated with the specific sensor location to generate an abnormal point.

[0061] After determining the specific locations of all anomalies, anomaly detection results are generated through data mapping and integration technology.

[0062] Furthermore, in the method provided in the embodiment of the application, optimizing the first maintenance decision according to the detection feedback result further includes:

[0063] A facility self-inspection cycle is set for the power facility according to the detection feedback result; the first facility maintenance cycle is updated according to the facility self-inspection cycle to generate a second facility maintenance cycle; the maintenance rules are optimized based on the second facility maintenance cycle to obtain maintenance optimization rules; maintenance analysis is performed on the multiple data to be maintained in combination with the detection feedback result according to the maintenance optimization rules to generate a maintenance optimization effect; based on the maintenance optimization effect, the maintenance optimization rules are executed to optimize the first maintenance decision and determine the second maintenance decision.

[0064] In the embodiment of the present application, the current operating status of the device is first determined based on the detection feedback result of the device. If the detection feedback result shows that the device is operating normally, the self-test cycle is appropriately extended; if the feedback shows that the device has abnormal fluctuations, the preset step length self-test cycle is shortened and the preset step length self-test frequency is increased to monitor the device status more frequently. The setting of the facility self-test cycle is completed through this process.

[0065] After setting the self-inspection cycle, adjust the maintenance cycle based on the abnormal data in the detection feedback and the comprehensive score of the equipment's facility failure. A low comprehensive failure score means that there is a greater risk in the operation of the equipment, so the maintenance cycle needs to be shortened. Calculate the new maintenance cycle through the cycle adjustment algorithm. For example, if the basic maintenance cycle is 6 months, and the current comprehensive failure score and detection feedback results show that the risk of failure has increased, adjust the maintenance cycle to 3 months to ensure that the problem is solved in a short period of time and prevent the failure from worsening. The adjustment formula is ;in, It is the basic maintenance cycle. It is the adjustment factor calculated based on the comprehensive fault score, which is the ratio of the comprehensive fault score to the fault frequency.

[0066] Combined with the abnormal data identified in the detection feedback, the specific fault characteristics of the equipment are further analyzed. By analyzing the fluctuation data of key parameters such as temperature, voltage, and current, high-risk components and areas of failure are identified. For example, if the test results show that the temperature of the equipment frequently exceeds the standard, the temperature anomaly is identified as the main source of failure. In order to deal with these anomalies, the maintenance rules are optimized. For example, the inspection frequency of high-risk components will be increased, especially those with abnormal temperature and current. Additional maintenance steps may also be introduced, such as more frequent equipment status monitoring, or triggering automatic alarms and emergency maintenance processes when anomalies are detected. Based on the above analysis, new maintenance optimization rules are generated.

[0067] After optimizing the maintenance rules, the maintenance effect analysis is performed by combining multiple maintenance data and test feedback results. The execution effect of the maintenance optimization rules is evaluated by comparing the equipment operating status before and after maintenance. Specifically, data comparison technology is used to evaluate whether the health status, failure frequency, and operating stability of the equipment have improved. For example, analyze whether the failure frequency has decreased after maintenance and whether the key parameters are more stable. If the maintenance optimization rules successfully reduce the failure risk and equipment damage rate, these rules are applied to subsequent maintenance decisions. Through this analysis, the maintenance optimization effect is generated.

[0068] Based on the maintenance optimization effect, the original first maintenance decision is evaluated. If the number of failures decreases, the current maintenance strategy is retained; if the problem persists, the maintenance plan is adjusted. For equipment with a low comprehensive failure score, additional maintenance operations and detection measures are added, and a new second maintenance decision is finally generated through this to ensure that the equipment can be more comprehensively monitored and maintained in the next maintenance cycle, reducing the risk of failure.

[0069] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0070] The present application obtains a facility operation data set by real-time sensor monitoring of power facilities; analyzes the facility operation data set, builds a prediction model, predicts the state of the power facilities through the prediction model, and obtains multiple facility state prediction parameters; covers and traverses multiple facility state prediction parameters, evaluates and traverses the node set for identification, and generates multiple tags to be detected; performs abnormal analysis based on multiple tags to be detected, generates a fault signal, and makes a first maintenance decision based on multiple facility state prediction parameters combined with the fault signal; performs abnormal detection on the power facility execution maintenance decision combined with the prediction model, generates a detection feedback result, optimizes the first maintenance decision according to the detection feedback result, and obtains a second maintenance decision to perform predictive maintenance response on the power facility. The present invention solves the technical problems of untimely response and low maintenance efficiency in the prior art, monitors the operation data of power facilities through real-time sensor, builds a prediction model for state prediction, generates facility state prediction parameters, covers and traverses these parameters, identifies abnormal points and generates fault signals, thereby making maintenance decisions, and obtains a second maintenance decision by executing the maintenance decision combined with the prediction model for feedback optimization, so as to achieve early prevention and efficient maintenance, and achieve the technical effect of improving the stability and safety of equipment operation.

[0071] Embodiment 2, based on the same inventive concept as a method for predictive maintenance of electric power facilities in the above embodiment, Figure 2 As shown, the present application provides a predictive maintenance system for electric power facilities, and the system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0072] A real-time monitoring module 11, wherein the real-time monitoring module 11 obtains a facility operation data set by performing real-time sensor monitoring on the power facilities; a state prediction module 12, wherein the state prediction module 12 performs analysis based on the facility operation data set, constructs a prediction model, predicts the state of the power facilities through the prediction model, and obtains multiple facility state prediction parameters; a coverage traversal module 13, wherein the coverage traversal module 13 performs coverage traversal on the multiple facility state prediction parameters, evaluates the traversal node set for identification, and generates multiple labels to be detected; an anomaly analysis module 14, wherein the anomaly analysis module 14 performs an anomaly analysis based on the multiple labels to be detected, generates a fault signal, and formulates a first maintenance decision based on the multiple facility state prediction parameters combined with the fault signal; an anomaly detection module 15, wherein the anomaly detection module 15 performs anomaly detection on the power facilities by executing the first maintenance decision in combination with the prediction model, generates a detection feedback result, optimizes the first maintenance decision according to the detection feedback result, and obtains a second maintenance decision to perform predictive maintenance response on the power facilities.

[0073] Furthermore, the system is also used to implement the following functions:

[0074] Data is captured based on the facility operation data set, and a data trend graph is drawn; correlation analysis is performed on the facility operation data set based on the data trend graph to generate autocorrelation data and partial autocorrelation data; based on the partial autocorrelation data, the autocorrelation data is sliding averaged to generate a smoothing index; according to the smoothing index, time series analysis is performed in combination with the autocorrelation data and the partial autocorrelation data to construct an analysis data list; based on the analysis data list, the autocorrelation data and the partial autocorrelation data are divided to construct the prediction model.

[0075] Furthermore, the system is also used to implement the following functions:

[0076] Based on the multiple tags to be detected, the multiple facility status prediction parameters are extracted to obtain multiple parameters to be detected; a facility status index is set, and fluctuation evaluation is performed on the multiple parameters to be detected according to the facility status index to obtain parameter fluctuation results; parameter fluctuation extreme values ​​are determined according to the parameter fluctuation results, and multiple abnormal state data points are determined according to the parameter fluctuation extreme values; feature analysis is performed based on the multiple abnormal state data points to generate abnormal state features, and similarity judgment is performed on the abnormal state features and historical fault features to generate multiple similarity coefficients; fault verification is performed on the multiple abnormal state data points based on the multiple similarity coefficients, and the fault signal is determined according to the verification result.

[0077] Furthermore, the system is also used to implement the following functions:

[0078] Traversing the multiple facility status prediction parameters and matching them with the verified abnormal status data points, a plurality of facility fault status prediction parameters are determined; a comprehensive evaluation is performed based on the multiple facility fault status prediction parameters in combination with the fault signal to generate a comprehensive score for the facility fault; tracing the source according to the comprehensive score for the facility fault is performed to determine a plurality of data to be maintained; setting the first facility maintenance cycle for the plurality of data to be maintained according to maintenance rules, and constructing the first maintenance decision according to the first facility maintenance cycle.

[0079] Furthermore, the system is also used to implement the following functions:

[0080] Execute the first maintenance decision to perform simulated maintenance on the power facilities and generate a facility simulation maintenance result; perform secondary prediction based on the prediction model and the facility simulation maintenance result to generate a facility simulation maintenance state parameter set; traverse the facility simulation maintenance state parameter set and perform anomaly detection in combination with the historical fault characteristics to generate anomaly detection results; perform anomaly impact analysis on the anomaly detection results to obtain multiple impact coefficients; arrange the anomaly detection results in descending order according to the multiple impact coefficients to generate an anomaly detection sequence, and add the anomaly detection sequence to the detection feedback result.

[0081] Furthermore, the system is also used to implement the following functions:

[0082] A facility fault threshold is set based on the historical fault characteristics; the facility simulation maintenance status parameter set is traversed and compared with the facility fault threshold to determine whether the facility simulation maintenance status parameter set is greater than or equal to the facility fault threshold; if there are facility simulation maintenance status parameters greater than or equal to the facility fault threshold in the facility simulation maintenance status parameter set, an abnormal identification is generated, and the abnormal point of the power facility is determined based on the facility simulation maintenance status parameters greater than or equal to the facility fault threshold, the abnormal identification is mapped with the abnormal point, and the abnormal detection result is generated.

[0083] Furthermore, the system is also used to implement the following functions:

[0084] A facility self-inspection cycle is set for the power facility according to the detection feedback result; the first facility maintenance cycle is updated according to the facility self-inspection cycle to generate a second facility maintenance cycle; the maintenance rules are optimized based on the second facility maintenance cycle to obtain maintenance optimization rules; maintenance analysis is performed on the multiple data to be maintained in combination with the detection feedback result according to the maintenance optimization rules to generate a maintenance optimization effect; based on the maintenance optimization effect, the maintenance optimization rules are executed to optimize the first maintenance decision and determine the second maintenance decision.

[0085] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0087] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A predictive maintenance method for electric power facilities, characterized in that: The method comprises: Through real-time sensor monitoring of power facilities, the facility operation data set is obtained; Analyze the facility operation data set to construct a prediction model, predict the state of the power facility through the prediction model, and obtain multiple facility state prediction parameters; Performing a coverage traversal on the multiple facility status prediction parameters, evaluating the traversal node set for identification, and generating multiple labels to be detected; Performing anomaly analysis based on the multiple tags to be detected to generate a fault signal, and formulating a first maintenance decision based on the multiple facility status prediction parameters and the fault signal; Executing the first maintenance decision on the power facility in combination with the prediction model to perform abnormality detection, generating a detection feedback result, optimizing the first maintenance decision according to the detection feedback result, and obtaining a second maintenance decision to perform predictive maintenance response on the power facility; Capture data based on the facility operation data set and draw a data trend chart; Performing correlation analysis on the facility operation data set based on the data trend graph to generate autocorrelation data and partial autocorrelation data; According to the partial autocorrelation data, performing sliding averaging on the autocorrelation data to generate a smoothing index; According to the smoothing index, the autocorrelation data and the partial autocorrelation data are combined to perform time series analysis, and construct an analysis data list; The autocorrelation data and the partial autocorrelation data are divided based on the analysis data list to construct the prediction model.

2. A method for predictive maintenance of electric power facilities as claimed in claim 1, characterized in that: Based on the multiple tags to be detected, an abnormality analysis is performed to generate a fault signal, and the method includes: Extracting the plurality of facility status prediction parameters based on the plurality of tags to be detected to obtain a plurality of parameters to be detected; Setting a facility status indicator, performing fluctuation evaluation on the plurality of parameters to be detected according to the facility status indicator, and obtaining a parameter fluctuation result; Determine a parameter fluctuation extreme value according to the parameter fluctuation result, and determine a plurality of abnormal state data points according to the parameter fluctuation extreme value; Performing feature analysis based on the multiple abnormal state data points to generate abnormal state features, performing similarity determination between the abnormal state features and historical fault features to generate multiple similarity coefficients; Fault verification is performed on the multiple abnormal state data points based on the multiple similarity coefficients, and the fault signal is determined according to the verification result.

3. A method for predictive maintenance of electric power facilities as claimed in claim 2, characterized in that: Based on the plurality of facility status prediction parameters combined with the fault signal, a first maintenance decision is made, the method comprising: Traversing the plurality of facility status prediction parameters and matching them with the verified abnormal status data points to determine a plurality of facility failure status prediction parameters; Performing a comprehensive evaluation based on the plurality of facility fault state prediction parameters in combination with the fault signal to generate a comprehensive facility fault score; According to the comprehensive score of the facility failure, trace the source and determine multiple data to be maintained; The plurality of data to be maintained are set to a first facility maintenance cycle according to maintenance rules, and the first maintenance decision is constructed according to the first facility maintenance cycle.

4. A method for predictive maintenance of electric power facilities as claimed in claim 2, characterized in that: Executing the first maintenance decision on the power facility and performing anomaly detection in combination with the prediction model, the method comprising: Executing the first maintenance decision to perform simulated maintenance on the power facility and generating a simulated maintenance result of the facility; Perform secondary prediction based on the prediction model and the facility simulation maintenance result to generate a facility simulation maintenance state parameter set; Traversing the facility simulation maintenance status parameter set and combining the historical fault characteristics to perform anomaly detection and generate anomaly detection results; Performing abnormal impact analysis on the abnormal detection result to obtain multiple impact coefficients; The abnormality detection results are arranged in descending order according to the multiple influence coefficients to generate an abnormality detection sequence, and the abnormality detection sequence is added to the detection feedback result.

5. A method for predictive maintenance of electric power facilities as claimed in claim 4, characterized in that: Traversing the facility simulation maintenance state parameter set and combining the historical fault characteristics to perform anomaly detection and generate anomaly detection results, the method includes: setting a facility failure threshold based on the historical failure characteristics; Traversing the facility simulation maintenance state parameter set and comparing it with the facility failure threshold, and determining whether the facility simulation maintenance state parameter set is greater than or equal to the facility failure threshold; If there are facility simulation maintenance status parameters greater than or equal to the facility fault threshold in the facility simulation maintenance status parameter set, an abnormal identification is generated, and the abnormal point of the power facility is determined based on the facility simulation maintenance status parameters greater than or equal to the facility fault threshold, the abnormal identification is mapped with the abnormal point, and the abnormal detection result is generated.

6. A method for predictive maintenance of electric power facilities as claimed in claim 3, characterized in that: Optimizing the first maintenance decision according to the detection feedback result, the method includes: Setting a facility self-inspection cycle for the power facility according to the detection feedback results; Updating the first facility maintenance cycle according to the facility self-inspection cycle to generate a second facility maintenance cycle; Optimizing the maintenance rule based on the second facility maintenance cycle to obtain a maintenance optimization rule; Performing maintenance analysis on the plurality of data to be maintained in combination with the detection feedback results according to the maintenance optimization rules to generate a maintenance optimization effect; According to the maintenance optimization effect, the maintenance optimization rule is executed to optimize the first maintenance decision and determine the second maintenance decision.

7. A predictive maintenance system for electric power facilities, characterized in that: The system is used to execute a predictive maintenance method for electric power facilities according to any one of claims 1 to 6, and the system comprises: A real-time monitoring module, wherein the real-time monitoring module obtains a facility operation data set by performing real-time sensor monitoring on the power facilities; A state prediction module, wherein the state prediction module analyzes the facility operation data set, constructs a prediction model, predicts the state of the power facility through the prediction model, and obtains a plurality of facility state prediction parameters; A coverage traversal module, which performs coverage traversal on the multiple facility status prediction parameters, evaluates the traversal node set for identification, and generates multiple labels to be detected; an abnormality analysis module, wherein the abnormality analysis module performs abnormality analysis based on the multiple tags to be detected, generates a fault signal, and formulates a first maintenance decision based on the multiple facility status prediction parameters and the fault signal; An anomaly detection module, wherein the anomaly detection module performs anomaly detection on the power facility by combining the first maintenance decision with the prediction model, generates a detection feedback result, optimizes the first maintenance decision according to the detection feedback result, and obtains a second maintenance decision to perform a predictive maintenance response on the power facility.

Citation Information

Patent Citations

  • Cloud computing-based highway electromechanical equipment operation and maintenance decision processing method and system

    CN117592975A

  • Chemical safety automatic detection and monitoring system of cloud PLC (Programmable Logic Controller)

    CN117808166A

  • Wind turbine generator remote fault diagnosis method and system based on cloud platform

    CN117930815A

  • Cooling-water machine energy efficiency optimization system based on intelligent algorithm

    CN118089287A

  • Power transformation equipment temperature situation awareness auxiliary control method

    CN118152741A

Cited By

  • Predictive maintenance method and system for industrial equipment

    CN120087564A

  • Equipment anomaly detection method and system based on multi-source heterogeneous data

    CN120145206A

  • Factory affair equipment predictive maintenance method and system

    CN121457822A