Intelligent alarm method for unmanned hydropower station based on device model and historical self-learning

Abstract

The application provides an intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning, and relates to the field of intelligent alarm technology for hydropower stations.The steps of the application are as follows: S1, extracting multi-source data and performing standardization processing;S2, building a layered topological information model of all station equipment based on KKS coding;S3, extracting equipment operation characteristics and constructing a standardized feature library;S4, performing data self-learning operation and determining the upper and lower limits of the operation interval of the equipment operation characteristics;S5, performing fault determination and alarm according to the real-time value of the equipment operation characteristic data;S6, confirming the alarm result by the operation and maintenance personnel after operation and maintenance, and periodically self-learning the self-learning model in combination with the alarm result and the equipment operation characteristic data.The beneficial effects are as follows: the KKS coding can be bound to the equipment, the fault equipment can be quickly located, the self-learning model can ensure the real-time applicability of the fault determination threshold, and the false alarm rate can be reduced by setting the adjustment coefficient as adaptive adjustment.

Description

Technical Field

[0001] This invention relates to the field of intelligent alarm technology for hydropower stations, specifically to an intelligent alarm method for unmanned hydropower stations based on equipment models and historical self-learning. Background Technology

[0002] Current hydropower station monitoring systems generally adopt a single-point fixed threshold alarm mode, which only makes simple judgments on the over-limit of a single measuring point. The specific defects are as follows: it only supports single-point threshold judgment, and cannot identify abnormal equipment operation trends or abnormal multi-parameter correlations, resulting in high rates of missed and false alarms; it does not build equipment information models according to equipment hierarchy, resulting in data being disconnected from equipment objects, making fault location difficult and troubleshooting inefficient; and the alarm thresholds rely on manual setting and cannot be automatically optimized based on historical equipment operating data and changes in operating conditions, resulting in poor adaptability. Summary of the Invention

[0003] The main objective of this invention is to provide an intelligent alarm method for unmanned hydropower stations based on equipment models and historical self-learning, thereby solving the problems mentioned in the background art.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by this invention is: an intelligent alarm method for unmanned hydropower stations based on equipment models and historical self-learning, comprising the following steps: S1. Extract multi-source data and perform standardization processing; S2. Construct a hierarchical topology information model for all equipment in the station based on KKS coding; S3. Extract equipment operation characteristics and construct a standardized characteristic library; S4. Perform data self-learning operation and determine the upper and lower limits of the operating range of the equipment operating characteristics; S5. Determine faults and issue alarms based on real-time values ​​of equipment operating characteristic data; S6. After maintenance, the operation and maintenance personnel confirm the alarm results. The self-learning model regularly combines the alarm results and equipment operation characteristic data to perform self-learning.

[0005] Furthermore, the multi-source data includes: analog quantity data, digital quantity data, and operating condition data; Operating conditions include the following types: shutdown, idling, no-load, power generation, and pumping; Standardized processing includes: error value removal, missing value filling, and jitter repair operations.

[0006] Furthermore, the detailed process of step S2 is as follows: S201. Using the national standard KKS coding rules as the standard, assign a unique KKS code to each independent device as the primary key of the database, and build a four-level tree-structured hierarchical structure of plant-level, system-level, unit-level, and equipment-level and a unified retrieval dictionary for the whole plant. S202. Configure complete attributes for a single device model and simultaneously establish a device topology association table; S203. Establish a standardized equipment template library; The template includes: a complete digital model of the equipment, a script for calculating the equipment's operating characteristics, and a standardized feature library; S204. Set judgment rules for each type of operating condition of the unit, and automatically identify the real-time operating condition of the equipment through the combination logic of measurement points.

[0007] Furthermore, the equipment operating characteristics include: steady-state distribution characteristics, time-series trend characteristics, and abnormal and sensor degradation characteristics.

[0008] Furthermore, the process of constructing the standardized feature library is as follows: S301. Normalize the extracted equipment operating characteristics; S302. Bind the normalized equipment operating characteristics to the corresponding equipment KKS code and the current operating condition; S303. Store the equipment operation characteristics in the real-time characteristic library and the historical characteristic library respectively in a standard storage format; The real-time feature library is used to store equipment operation features within the historical feature time threshold, while the historical feature library is used to store equipment operation features outside the historical feature time threshold. The standard storage format is: one operating condition of one device corresponds to a set of equipment operation features. S304. Set a storage time threshold and delete historical device operation feature data in the historical feature library that has been stored for a longer period than the storage time threshold.

[0009] Furthermore, the detailed process of step S4 is as follows: S401. Extract equipment operation feature data from the standard feature library and classify the equipment operation feature data according to the operation records; The types of equipment operation characteristic data are: normal operation samples, fault and abnormal samples, and invalid samples from maintenance and isolation. S402. Select an appropriate self-learning model to analyze the feature data and determine the upper and lower limits of the operating range of the equipment operating characteristics; self-learning models include: normal distribution model, box plot model and ARIMA time series prediction model; S403. Determine the upper and lower limits of the operating range based on the undetermined upper and lower limits of the operating range for different sample types; the upper and lower limits of the operating range include: the upper and lower limits of the operating range for normal operation samples and the upper and lower limits of the operating range for faulty samples, specifically: Calculate the undetermined upper and lower limits of the operating range for normal operation samples and fault / abnormal samples respectively; the undetermined upper and lower limits of the operating range for normal operation samples are the upper and lower limits of the operating range for normal operation samples, and the undetermined upper and lower limits of the operating range for fault / abnormal samples are the upper and lower limits of the operating range for fault samples.

[0010] Furthermore, the process for determining the upper and lower limits of the operating range of the feature data is as follows: For steady-state distribution characteristic data, we have: The mean and standard deviation of the data are calculated using a normal distribution model, as shown in the following expressions: (1); in, and These are the mean and standard deviation, respectively. For the first Feature values ​​of each sample Total sample size; The operating range of steady-state distribution characteristic data is to be determined, with upper and lower limits based on... The principle is established, that is, it satisfies: (2); in, , These represent the undetermined lower and upper limits of the operating range of the feature data; here, the feature data refers to steady-state distributed feature data. For abnormal and sensor degradation characteristic data, we have: The interquartile range is calculated using a box plot model, as shown in the following expression: (3); in, Interquartile range, , These are the lower quartile and upper quartile of the sample, respectively; The upper and lower limits of the operating range for anomaly and sensor degradation characteristic data are as follows: (4); Here , This corresponds to abnormal and sensor degradation characteristic data; Regarding time-series trend characteristics, we have: The ARIMA time series forecasting model is used to establish a fitting equation for the time series trend eigenvalues, as shown in the following expression: (5); in, For the first The difference between the time-series trend feature values ​​of each sample This is the offset of the model constants. For the first The autoregressive coefficient of order, For the first The moving average coefficient of the order, for The predicted value, This is the differential measurement error; The expression for the fitted objective function is determined as follows: (6); in, The objective function value; The fitting target is determined as follows: Take the minimum value and set the initial value of the coefficients in the coefficient column. The difference measurement error of the sample in the equation can be calculated according to the fitting equation with the initial value of the coefficients. Then, the fitting algorithm is used to fit the coefficients in the coefficient column to calculate the optimal value of the coefficients in the coefficient column. At this time, equation (5) can be rewritten as: (7); in, For the first Predicted values ​​of time-series trend characteristics of each sample; The expressions for the above coefficient column are as follows: (8); in, For coefficient column; The subsequent differential measurement error can be calculated based on equations (5) and (7); the normal random fluctuation amplitude of the model can be calculated based on the differential measurement error, as shown in the following expression: (9); in, This represents the normal random fluctuation range of the model; The upper and lower limits of the operating range for time series trend feature data are as follows: (10); It should be noted that here , This corresponds to time-series trend feature data.

[0011] Furthermore, the detailed process of step S5 is as follows: S501. Determine the fault judgment threshold; the fault judgment threshold includes: normal threshold and fault threshold; The expression for the normal threshold is as follows: (11); in, , These represent the upper and lower limits of the normal operating threshold for any device. , These represent the upper and lower limits of the normal operating range for this feature; The expression for the fault threshold is as follows: (12); in, , These represent the upper and lower limits of the fault threshold for any device's operational characteristics. , These represent the upper and lower limits of the operating range for faulty and abnormal samples of this feature. This is the adjustment coefficient; the value of the adjustment coefficient ranges from 0 to 1. S502. Perform real-time status determination of equipment operating characteristics. For any equipment operating characteristic, the following applies: when the real-time value of the equipment operating characteristic does not exceed the upper or lower limit of the normal threshold, it is in a normal state; when the real-time value of the equipment operating characteristic exceeds the upper or lower limit of the normal threshold but does not exceed the upper or lower limit of the fault threshold, it is in an offset state; when the real-time value of the equipment operating characteristic exceeds the upper or lower limit of the fault threshold, it is in a fault state. S503. Take corresponding alarm measures based on the fault determination results; The appropriate action under normal circumstances is to not trigger an alarm. The measures corresponding to the offset transition are: continuously record the data of the equipment operation characteristics of the offset, and submit alarm reports to the operation and maintenance personnel according to a fixed period; The corresponding measures for the fault status are: immediately send the faulty equipment's operating characteristic data to the maintenance personnel in the form of an alarm report; S504. Store the alarm report in the alarm storage repository.

[0012] Furthermore, the detailed process of step S6 is as follows: S601. Maintenance personnel confirm the alarm results; alarm results are divided into two types: correct and false alarms; a false alarm refers to a discrepancy between the actual state of the equipment and the real-time state of the equipment's operating characteristics reported in the alarm; the actual state of the equipment includes normal operating state and non-operating state; the correspondence between the actual state of the equipment and the real-time state of the equipment's operating characteristics under normal circumstances is as follows: The normal working state corresponds to the normal state and the offset state; The inoperable state corresponds to the fault state; False alarms include: An alarm report indicates an offset state, but the device is not in a working state. This situation is recorded as a false alarm scenario. An alarm report indicates a fault state, but the equipment is in normal working condition. This situation is recorded as a normal false alarm scenario. S602. Add alarm result tags to alarm reports in the alarm repository based on alarm results, and add type tags to device operation characteristics in the standardized feature library; The type labels here are: normal operation sample, faulty / abnormal sample, and maintenance / isolation invalid sample; S603. The self-learning model periodically performs self-learning operations based on updated equipment operation characteristic data, and iteratively adjusts the upper and lower limits of the equipment operation characteristic range. S604. Adjust the adjustment coefficient according to the false alarm rate; the false alarm rate is divided into: normal false alarm rate and fault false alarm rate.

[0013] Furthermore, the expression for the adjustment coefficient is as follows: (13); in, This is the adjusted adjustment coefficient. To adjust the step size, , These are the normal false alarm rate and the fault false alarm rate, respectively.

[0014] Beneficial effects: (1) Bind the KKS code to the equipment and associate it with the equipment operation characteristics used for fault diagnosis, so that faulty equipment can be quickly located; (2) Using equipment operating characteristics as the basis for fault judgment can overcome the limitations of single-point threshold judgment; (3) Using normal and abnormal samples to determine the upper and lower limits of the operating range can improve the accuracy of fault judgment threshold selection; (4) The self-learning model learns from real-time and historical data, which can ensure the real-time applicability of the fault judgment threshold; (5) Setting the adjustment coefficient to adaptive adjustment can reduce the false alarm rate. Attached Figure Description

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the steps of the present invention. Detailed Implementation

[0016] Example 1 like Figure 1 As shown, the intelligent alarm method for unmanned hydropower stations based on equipment models and historical self-learning includes the following steps: S1. Extract multi-source data and perform standardization processing; Multi-source data includes: analog data, digital data, and operating condition data; Analog data includes the following types of data: temperature data, vibration data, swing data, pressure data, flow rate data, and water level data; Switching data includes, but is not limited to: equipment start / stop status, switch position, and protection action data; protection action data includes: alarm, trip, shutdown, and other action signals. Operating data includes: start-up and shutdown data of operating process, pumping and power generation conversion signal data; the unit's operating conditions include the following types: shutdown, idling, no-load, power generation and pumping; Standardized processing includes: error value removal, missing value imputation, and jitter repair operations; For missing values, linear interpolation is used for imputation. The jitter repair method uses a moving average filtering method to address the problem of instantaneous pulse jumps in the sensor. The filtered data expression is as follows: (1); in, The length of the sliding window. For the first One original data value, For the filtered first There are several data values. It should be noted that when filtering the first few data points, since there are not enough original data points to calculate the mean, the first data point in the sliding window is set as the first data point. For example, in the filtering process of the second original data point, the sliding window length is 4, and only the first and second data points are used to calculate the mean.

[0017] S2. Based on KKS encoding, a hierarchical topology information model for all devices in the site is built. The detailed process is as follows: S201. Using the national standard KKS coding rules as the standard, assign a unique KKS code to each independent device as the primary key of the database, and build a four-level tree-structured hierarchical structure of plant-level, system-level, unit-level, and equipment-level and a unified retrieval dictionary for the whole plant. Each individual device is also bound to a standard Chinese name and device type keywords, which can be searched through a unified search dictionary built for the entire plant. KKS encoding is adapted to a four-level tree-like hierarchical structure, including four parts: unit code, system code, unit code, and equipment code. S202. Configure complete attributes for a single device model and simultaneously establish a device topology association table; Equipment relationships include: master-slave relationship, hydraulic upstream and downstream relationship, and electrical linkage relationship. Equipment relationships can characterize the subordinate, pipeline, and control linkage logical relationships between equipment. Complete attributes include: equipment setpoint parameters, basic attributes, control attributes, relevant quantities, real-time characteristic attributes, historical characteristic attributes, alarm attributes, and emergency response attributes; Equipment setpoint parameters refer to static rated parameters that are fixed and unchanging during the equipment's manufacturing and design phases. These are inherent standard values ​​of the hardware and are not modified over a long period of time. Basic attributes refer to the static basic information of the device management class; Control attributes refer to the control logic configuration related to equipment start-up, shutdown, interlocking, and operation; Relevant quantities refer to analog data, digital data, and operating condition data; Real-time characteristic attributes refer to dynamic indicators calculated in real time during the current operation of the device; Historical feature attributes refer to time-series feature datasets that are archived and stored for a long time, and belong to historical statistical results; Alarm attributes refer to the device-specific alarm rules and self-learning output threshold configurations; Emergency response attributes refer to a standardized knowledge base for handling equipment malfunctions. S203. Establish a standardized equipment template library; The template includes: a complete digital model of the equipment, a script for calculating the equipment's operating characteristics, and a standardized feature library; For the same type of unit, oil pump, water pump and transformer, the template can be directly imported to generate equipment models in batches, which can greatly reduce the workload of manual modeling. The template also has a template version management mechanism, specifically: Each time the template is modified, a new version number is automatically generated, recording the account of the person making the modification, the modification time, and details of the changes. If the new version of the template has poor compatibility, you can switch back to a previous stable version. It supports fine-tuning of template-differentiated parameters. For the same type of equipment with different capacities and different years of commissioning, the set parameters can be individually modified based on the standard template, taking into account both reusability and equipment differentiation. S204. Set judgment rules for each type of operating condition of the unit, and automatically identify the real-time operating condition of the equipment through measurement point combination logic; operating conditions include: shutdown, no-load, power generation, and pumping. The shutdown determination rule is that the following conditions must be met simultaneously: The actual speed of the generator unit does not exceed 5% of the rated speed; the rotor is stationary; the generator outlet circuit breaker is open; The main inlet valve is fully closed; there is no start-up process execution signal and no excitation activation signal. The rule for determining idling is that the following conditions must be met simultaneously: The unit speed is maintained at 99%~101% of the rated range; the generator outlet circuit breaker is open; the main inlet valve is fully open, and the speed governor maintains the closed-loop speed control; the excitation system is engaged, and the generator terminal voltage reaches the rated voltage range; there are no grid connection or load commands. The rule for determining no-load is that the following conditions must be met simultaneously: The unit speed is maintained at 99%~101% of the rated range; the generator outlet circuit breaker is turned on, and the unit is connected to the grid; the active power does not exceed 5% of the unit's rated active power; the speed governor and excitation system are automatically connected to the grid; the guide vanes maintain the no-load opening, and there is no load increase or decrease action; The rule for determining power generation is that the following conditions must be met simultaneously: The unit speed is maintained at 99%~101% of the rated range; the generator outlet circuit breaker is turned on and the unit is connected to the grid; the active power is greater than 5% of the unit's rated active power; the main inlet valve is fully open; the phase-reversing isolating switch is placed in the generating position; The criteria for determining whether to pump water are to do so are as follows: The unit speed is maintained at 99%~101% of the rated range; the generator outlet circuit breaker is turned on and the unit is connected to the grid; the active power is negative and less than 5% of the rated active power of the unit; the main inlet valve is fully open; the phase-reversing isolating switch is set in the pumping direction; the runner chamber pressure water and air replenishment valve is engaged.

[0018] S3. Extract equipment operation characteristics and construct a standardized characteristic library; Equipment operating characteristics include: steady-state distribution characteristics, time-series trend characteristics, and abnormal and sensor degradation characteristics; Steady-state distribution characteristics include: mean, variance, standard deviation, maximum value, minimum value, and median of the analog quantity within a fixed time window, which are used to determine the long-term steady-state operation deviation of the equipment; The time-series trend features include: analog parameter values ​​within a fixed window, hourly changes, and daily cumulative changes, which are used to capture early signs of slow, gradual deterioration faults. Anomalies and sensor degradation characteristics include: abrupt changes in analog signal amplitude, specifically identifying sensor disconnection, drift, and aging failure issues; The standardized feature library includes: a real-time feature library and a historical feature library; The process of constructing the standardized feature library is as follows: S301. Normalize the extracted equipment operation characteristics and adjust the values ​​of the equipment operation characteristics to the range of 0~1. S302. Bind the normalized equipment operating characteristics to the corresponding equipment KKS code and the current operating condition; S303. Store the equipment operation characteristics in a standard storage format in the real-time characteristic library and the historical characteristic library respectively; the real-time characteristic library is used to store the equipment operation characteristics within the historical characteristic time threshold, and the historical characteristic library is used to store the equipment operation characteristics outside the historical characteristic time threshold. The standard storage format is: one operating condition of a device corresponds to a set of device operating characteristics; S304. Set a storage time threshold and delete historical device operation feature data in the historical feature library that has been stored for a longer period than the storage time threshold.

[0019] S4. Perform data self-learning operations and determine the upper and lower limits of the operating range of the equipment's operating characteristics. The detailed process is as follows: S401. Extract equipment operation feature data from the standard feature library and classify the equipment operation feature data according to the operation records; The types of equipment operation characteristic data are: normal operation samples, fault and abnormal samples, and invalid samples from maintenance and isolation. Normal operating sample consists of characteristic data under steady-state conditions with no alarms, no protection actions, no maintenance records; The fault and anomaly samples are: alarm time period, protection action, fault shutdown, and characteristic data corresponding to manually marked faults; Invalid samples for maintenance isolation are: characteristic data of the equipment maintenance isolation phase; it should be noted that invalid samples for maintenance isolation are directly removed and not used in training. The device operation characteristic data here includes data from the real-time characteristic library and the historical characteristic library; S402. Select an appropriate self-learning model to analyze the feature data and determine the undetermined upper and lower limits of the operating range of the equipment's operating characteristics; self-learning models include: normal distribution model, box plot model, and ARIMA time series prediction model. The correspondence between self-learning models and feature data types is as follows: The normal distribution model is used to analyze steady-state distribution characteristic data; Box plot models are used to analyze anomaly and sensor degradation characteristics data; ARIMA time series forecasting models are used to analyze time series trend characteristics; The process for determining the upper and lower limits of the operating range of feature data is as follows: For steady-state distribution characteristic data, we have: The mean and standard deviation of the data are calculated using a normal distribution model, as shown in the following expressions: (1); in, and These are the mean and standard deviation, respectively. For the first Feature values ​​of each sample Total sample size; The operating range of steady-state distribution characteristic data is to be determined, with upper and lower limits based on... The principle is established, that is, it satisfies: (2); in, , These represent the undetermined lower and upper limits of the operating range of the feature data; here, the feature data refers to steady-state distributed feature data. For abnormal and sensor degradation characteristic data, we have: The interquartile range is calculated using a box plot model, as shown in the following expression: (3); in, Interquartile range, , These are the lower quartile and upper quartile of the sample, respectively; The upper and lower limits of the operating range for anomaly and sensor degradation characteristic data are as follows: (4); It should be noted that here , This corresponds to abnormal and sensor degradation characteristic data; Regarding time-series trend characteristics, we have: The ARIMA time series forecasting model is used to establish a fitting equation for the time series trend eigenvalues, as shown in the following expression: (5); in, For the first The difference between the time-series trend feature values ​​of each sample; This is the model constant offset. This value represents the average normal increment of parameters within a single sampling period under steady-state operation of the equipment. When the equipment slowly deteriorates, this value will continue to shift in one direction. For the first The autoregressive coefficient of the order, the term in which this value is located uses the historical changes of the previous 3 periods to fit the normal incremental trend of the current period; For the first The moving average coefficient of the order is the output of the model that uses the historical prediction error correction model from the previous two periods to reduce the prediction bias caused by random disturbances. for The predicted value, i.e., the first The predicted value of the time series trend feature of the first sample and the first The difference between the predicted values ​​of the time-series trend features of each sample; This is the differential measurement error; The expression for the fitted objective function is determined as follows: (6); in, The objective function value; The fitting target is determined as follows: Take the minimum value and set the initial value of the coefficients in the coefficient column. The difference measurement error of the sample in the equation can be calculated according to the fitting equation with the initial value of the coefficients. Then, the fitting algorithm such as the least squares method is used to fit the coefficients in the coefficient column to calculate the optimal value of the coefficients in the coefficient column. At this time, equation (5) can be rewritten as: (7); in, For the first Predicted values ​​of time-series trend characteristics of each sample; The expressions for the above coefficient column are as follows: (8); in, For coefficient column; The subsequent differential measurement error can be calculated based on equations (5) and (7); the normal random fluctuation amplitude of the model can be calculated based on the differential measurement error, as shown in the following expression: (9); in, This represents the normal random fluctuation range of the model; The upper and lower limits of the operating range for time series trend feature data are as follows: (10); It should be noted that here , This corresponds to time-series trend feature data; S403. Determine the upper and lower limits of the operating range based on the undetermined upper and lower limits of the operating range for different sample types; the upper and lower limits of the operating range include: the upper and lower limits of the operating range for normal operation samples and the upper and lower limits of the operating range for faulty samples. This step is specifically as follows: Calculate the undetermined upper and lower limits of the operating range for normal operation samples and fault / abnormal samples respectively; the undetermined upper and lower limits of the operating range for normal operation samples are the upper and lower limits of the operating range for normal operation samples, and the undetermined upper and lower limits of the operating range for fault / abnormal samples are the upper and lower limits of the operating range for fault samples. The undetermined upper and lower limits of the operating range here include the undetermined upper and lower limits of the operating range for three types of equipment operating characteristic data.

[0020] S5. Based on the real-time values ​​of the equipment's operating characteristic data, fault diagnosis and alarm are triggered. The detailed process is as follows: S501. Determine the fault judgment threshold; the fault judgment threshold includes: normal threshold and fault threshold; The expression for the normal threshold is as follows: (11); in, , These represent the upper and lower limits of the normal operating threshold for any device. , These represent the upper and lower limits of the normal operating range for this feature; The expression for the fault threshold is as follows: (12); in, , These represent the upper and lower limits of the fault threshold for any device's operational characteristics. , These represent the upper and lower limits of the operating range for faulty and abnormal samples of this feature. This is the adjustment coefficient; the value of the adjustment coefficient ranges from 0 to 1; it should be noted that the adjustment coefficient may not be the same for different equipment operating characteristics. S502. Perform real-time status determination of equipment operating characteristics. For any equipment operating characteristic, the following applies: when the real-time value of the equipment operating characteristic does not exceed the upper or lower limit of the normal threshold, it is in a normal state; when the real-time value of the equipment operating characteristic exceeds the upper or lower limit of the normal threshold but does not exceed the upper or lower limit of the fault threshold, it is in an offset state; when the real-time value of the equipment operating characteristic exceeds the upper or lower limit of the fault threshold, it is in a fault state. S503. Take corresponding alarm measures based on the fault determination results; The appropriate action under normal circumstances is to not trigger an alarm. The measures corresponding to the offset transition are: continuously record the data of the equipment operation characteristics of the offset, and submit alarm reports to the operation and maintenance personnel according to a fixed period; The corresponding measures for the fault status are: immediately send the faulty equipment operation characteristic data to the operation and maintenance personnel in the form of an alarm report; it should be noted that the operation and maintenance personnel are only responsible for the operation and maintenance of the unmanned hydropower station and will not participate in the operation of the hydropower station; The alarm report includes: the KKS device code corresponding to the device's operating characteristics, the device name, the fault status, the precise timestamp of occurrence, the real-time value of the device's operating characteristics, the self-learning normal baseline range, the list of associated devices, and the current device operating status; S504. Store the alarm report in the alarm storage repository.

[0021] S6. After maintenance, the operation and maintenance personnel confirm the alarm results. The self-learning model periodically combines the alarm results and equipment operation characteristic data to perform self-learning. The detailed process is as follows: S601. Maintenance personnel confirm the alarm results; alarm results are divided into two types: correct and false alarms; a false alarm refers to a discrepancy between the actual state of the equipment and the real-time state of the equipment's operating characteristics reported in the alarm; the actual state of the equipment includes normal operating state and non-operating state; the correspondence between the actual state of the equipment and the real-time state of the equipment's operating characteristics under normal circumstances is as follows: The normal working state corresponds to the normal state and the offset state; The inoperable state corresponds to the fault state; False alarms include: An alarm report indicates an offset state, but the device is not in a working state. This situation is recorded as a false alarm scenario. An alarm report indicates a fault state, but the equipment is in normal working condition. This situation is recorded as a normal false alarm scenario. S602. Add alarm result tags to the alarm reports in the alarm repository based on the alarm results, and add type tags to the device operation characteristics in the standardized feature library; The type labels here are: normal operation sample, fault and abnormal sample, and maintenance and isolation ineffective sample, which are the three types of equipment operation characteristics. S603. The self-learning model periodically performs self-learning operations based on updated equipment operation characteristic data, and iteratively adjusts the upper and lower limits of the equipment operation characteristic range. The self-learning operation here refers to the self-learning model combining the updated data to perform the operation in step S4, calculating the upper and lower limits of the new device operating characteristic operating range, and then updating it. S604. Adjust the adjustment coefficient according to the false alarm rate; the false alarm rate is divided into: normal false alarm rate and fault false alarm rate; the normal false alarm rate refers to the proportion of the number of times a normal false alarm scenario occurs in the total number of false alarms; the fault false alarm rate refers to the proportion of the number of times a fault false alarm scenario occurs in the total number of false alarms; Analysis shows that theoretically, there should exist an optimal adjustment coefficient value that makes both the normal false alarm rate and the fault false alarm rate zero, representing an ideal and perfect situation where there are no false alarms. If the adjustment coefficient value is decreased, the upper limit of the operating range decreases, leading to normal false alarm scenarios. Conversely, increasing the adjustment coefficient value leads to fault false alarm scenarios. In other words, when there are many normal false alarm scenarios, the adjustment coefficient value should be increased; when there are many fault false alarm scenarios, the adjustment coefficient value should be decreased. Although such an ideal and perfect scenario will not occur in practical applications, the difference between the normal false alarm rate and the fault false alarm rate can characterize this phenomenon to some extent. Therefore, this embodiment uses the difference between the normal false alarm rate and the fault false alarm rate to adjust the adjustment coefficient, as expressed below: (13); in, This is the adjusted adjustment coefficient. To adjust the step size, , These are the normal false alarm rate and the fault false alarm rate, respectively.

[0022] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A smart alarm method for unmanned hydropower stations based on equipment models and historical self-learning, characterized in that, Includes the following steps: S1. Extract multi-source data and perform standardization processing; S2. Construct a hierarchical topology information model for all equipment in the station based on KKS coding; S3. Extract equipment operation characteristics and construct a standardized characteristic library; S4. Perform data self-learning operation and determine the upper and lower limits of the operating range of the equipment operating characteristics; S5. Determine faults and issue alarms based on real-time values ​​of equipment operating characteristic data; S6. After maintenance, the operation and maintenance personnel confirm the alarm results. The self-learning model regularly combines the alarm results and equipment operation characteristic data to perform self-learning.

2. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 1, characterized in that, Multi-source data includes: analog quantity data, digital quantity data, and operating condition data; Operating conditions include the following types: shutdown, idling, no-load, power generation, and pumping; Standardized processing includes: error value removal, missing value filling, and jitter repair operations.

3. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 2, characterized in that, The detailed process of step S2 is as follows: S201. Using the national standard KKS coding rules as the standard, assign a unique KKS code to each independent device as the primary key of the database, and build a four-level tree-structured hierarchical structure of plant-level, system-level, unit-level, and equipment-level and a unified retrieval dictionary for the whole plant. S202. Configure complete attributes for a single device model and simultaneously establish a device topology association table; S203. Establish a standardized equipment template library; The template includes: a complete digital model of the equipment, a script for calculating the equipment's operating characteristics, and a standardized feature library; S204. Set judgment rules for each type of operating condition of the unit, and automatically identify the real-time operating condition of the equipment through the combination logic of measurement points.

4. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 3, characterized in that, Equipment operating characteristics include: steady-state distribution characteristics, time-series trend characteristics, and abnormal and sensor degradation characteristics.

5. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 4, characterized in that, The process of constructing the standardized feature library is as follows: S301. Normalize the extracted equipment operating characteristics; S302. Bind the normalized equipment operating characteristics to the corresponding equipment KKS code and the current operating condition; S303. Store the equipment operation characteristics in the real-time characteristic library and the historical characteristic library respectively in a standard storage format; The real-time feature library is used to store device operation features within the historical feature time threshold, while the historical feature library is used to store device operation features outside the historical feature time threshold. The standard storage format is: one operating condition of a device corresponds to a set of device operating characteristics; S304. Set a storage time threshold and delete historical device operation feature data in the historical feature library that has been stored for a longer period than the storage time threshold.

6. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 5, characterized in that, The detailed process of step S4 is as follows: S401. Extract equipment operation feature data from the standard feature library and classify the equipment operation feature data according to the operation records; The types of equipment operation characteristic data are: normal operation samples, fault and abnormal samples, and invalid samples from maintenance and isolation. S402. Select an appropriate self-learning model to analyze the feature data and determine the upper and lower limits of the operating range of the equipment operating characteristics; self-learning models include: normal distribution model, box plot model and ARIMA time series prediction model; S403. Determine the upper and lower limits of the operating range based on the undetermined upper and lower limits of the operating range for different sample types; the upper and lower limits of the operating range include: the upper and lower limits of the operating range for normal operation samples and the upper and lower limits of the operating range for faulty samples, specifically: Calculate the undetermined upper and lower limits of the operating range for normal operation samples and fault / abnormal samples respectively; the undetermined upper and lower limits of the operating range for normal operation samples are the upper and lower limits of the operating range for normal operation samples, and the undetermined upper and lower limits of the operating range for fault / abnormal samples are the upper and lower limits of the operating range for fault samples.

7. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 6, characterized in that, The process for determining the upper and lower limits of the operating range of feature data is as follows: For steady-state distribution characteristic data, we have: The mean and standard deviation of the data are calculated using a normal distribution model, as shown in the following expressions: （1）； in, and These are the mean and standard deviation, respectively. For the first Feature values ​​of each sample Total sample size; The operating range of steady-state distribution characteristic data is to be determined, with upper and lower limits based on... The principle is established, that is, it satisfies: （2）； in, , These represent the undetermined lower and upper limits of the operating range of the feature data; here, the feature data refers to steady-state distributed feature data. For abnormal and sensor degradation characteristic data, we have: The interquartile range is calculated using a box plot model, as shown in the following expression: （3）； in, Interquartile range, , These are the lower quartile and upper quartile of the sample, respectively; The upper and lower limits of the operating range for anomaly and sensor degradation characteristic data are as follows: （4）； Here , This corresponds to abnormal and sensor degradation characteristic data; Regarding time-series trend characteristics, we have: The ARIMA time series forecasting model is used to establish a fitting equation for the time series trend eigenvalues, as shown in the following expression: （5）； in, For the first The difference between the time-series trend feature values ​​of each sample This is the offset of the model constants. For the first The autoregressive coefficient of order, For the first The moving average coefficient of the order, for The predicted value, This is the differential measurement error; The expression for the fitted objective function is determined as follows: （6）； in, The objective function value; The fitting target is determined as follows: Take the minimum value and set the initial value of the coefficients in the coefficient column. The difference measurement error of the sample in the equation can be calculated according to the fitting equation with the initial value of the coefficients. Then, the fitting algorithm is used to fit the coefficients in the coefficient column to calculate the optimal value of the coefficients in the coefficient column. At this time, equation (5) can be rewritten as: （7）； in, For the first Predicted values ​​of time-series trend characteristics of each sample; The expressions for the above coefficient column are as follows: （8）； in, For coefficient column; The subsequent differential measurement error can be calculated based on equations (5) and (7); the normal random fluctuation amplitude of the model can be calculated based on the differential measurement error, as shown in the following expression: （9）； in, This represents the normal random fluctuation range of the model; The upper and lower limits of the operating range for time series trend feature data are as follows: （10）； It should be noted that here , This corresponds to time-series trend feature data.

8. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 7, characterized in that, The detailed process of step S5 is as follows: S501. Determine the fault judgment threshold; the fault judgment threshold includes: normal threshold and fault threshold; The expression for the normal threshold is as follows: （11）； in, , These represent the upper and lower limits of the normal operating threshold for any device. , These represent the upper and lower limits of the normal operating range for this feature; The expression for the fault threshold is as follows: （12）； in, , These represent the upper and lower limits of the fault threshold for any device's operational characteristics. , These represent the upper and lower limits of the operating range for faulty and abnormal samples of this feature. This is the adjustment coefficient; the value of the adjustment coefficient ranges from 0 to 1. S502. Perform real-time status determination of equipment operating characteristics. For any equipment operating characteristic, the following applies: when the real-time value of the equipment operating characteristic does not exceed the upper or lower limit of the normal threshold, it is in a normal state; when the real-time value of the equipment operating characteristic exceeds the upper or lower limit of the normal threshold but does not exceed the upper or lower limit of the fault threshold, it is in an offset state; when the real-time value of the equipment operating characteristic exceeds the upper or lower limit of the fault threshold, it is in a fault state. S503. Take corresponding alarm measures based on the fault determination results; The appropriate action under normal circumstances is to not trigger an alarm. The measures corresponding to the offset transition are: continuously record the data of the equipment operation characteristics of the offset, and submit alarm reports to the operation and maintenance personnel according to a fixed period; The corresponding measures for the fault status are: immediately send the faulty equipment's operating characteristic data to the maintenance personnel in the form of an alarm report; S504. Store the alarm report in the alarm storage repository.

9. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 8, characterized in that, The detailed process of step S6 is as follows: S601. Maintenance personnel confirm the alarm results; alarm results are divided into two types: correct and false alarms; a false alarm refers to a discrepancy between the actual state of the equipment and the real-time state of the equipment's operating characteristics reported in the alarm; the actual state of the equipment includes normal operating state and non-operating state; the correspondence between the actual state of the equipment and the real-time state of the equipment's operating characteristics under normal circumstances is as follows: The normal working state corresponds to the normal state and the offset state; The inoperable state corresponds to the fault state; False alarms include: An alarm report indicates an offset state, but the device is not in a working state. This situation is recorded as a false alarm scenario. An alarm report indicates a fault state, but the equipment is in normal working condition. This situation is recorded as a normal false alarm scenario. S602. Add alarm result tags to alarm reports in the alarm repository based on alarm results, and add type tags to device operation characteristics in the standardized feature library; The type labels here are: normal operation sample, faulty / abnormal sample, and maintenance / isolation invalid sample; S603. The self-learning model periodically performs self-learning operations based on updated equipment operation characteristic data, and iteratively adjusts the upper and lower limits of the equipment operation characteristic range. S604. Adjust the adjustment coefficient according to the false alarm rate; the false alarm rate is divided into: normal false alarm rate and fault false alarm rate.

10. The intelligent alarm method for unmanned hydropower stations based on equipment model and historical self-learning as described in claim 9, characterized in that, The expression for the adjustment coefficient is as follows: （13）； in, This is the adjusted adjustment coefficient. To adjust the step size, , These are the normal false alarm rate and the fault false alarm rate, respectively.

Images