Water turbine governor system state evaluation method

Abstract

The invention discloses a hydraulic turbine governor system state evaluation method, and aims to solve the problems of working condition splitting, one-sided evaluation, threshold stiffness and system deficiency of an existing method. According to the method, multi-source parameters are collected and preprocessed at the resolution of 0.5-1 millisecond, operation conditions are subdivided into seven typical conditions for intelligent discrimination, exclusive indexes are extracted for each condition, abnormity is compared by adopting a single parameter, multi-parameter integration and dynamic threshold multi-mode, and health indexes are calculated based on degradation indexes and score deduction rules. And outputting a hierarchical evaluation conclusion through a four-layer evaluation system. All-working-condition coverage and multi-dimensional accurate evaluation are achieved, the early warning accuracy and operation and maintenance efficiency are improved, the service life of equipment is prolonged through preventive maintenance, parameters are optimized, energy consumption is reduced, and reliable guarantee is provided for safe and stable operation of the hydropower station.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water power equipment state monitoring and intelligent diagnosis, and relates to a water turbine governor system state evaluation method. BACKGROUND

[0002] The water turbine governor system is the core control equipment of the hydropower station, and its running state directly affects the stability of the power grid and the economic benefit of the hydropower station, so accurate state evaluation of the water turbine governor system is a core requirement for operation and maintenance.

[0003] The existing governor state evaluation scheme mainly includes two types: one is a single parameter threshold comparison method, which sets a fixed threshold for key parameters to issue a warning, but only focuses on the isolated change of a single parameter, ignores the coupling relationship between parameters, and cannot adapt to different operating conditions, resulting in high false alarm and missed alarm rates; the other is a fixed weight scoring model method, which calculates a health index by summing up multiple parameters with weights, but the weights are statically set based on experience or historical data, which cannot adapt to equipment degradation, load fluctuations and environmental changes, and the evaluation results deviate greatly from the actual state. The common defects of the two types are that the operating conditions are fragmented, the evaluation is one-sided, only the mechanical or electrical single-dimensional parameters are monitored, and the threshold is rigid, which cannot adapt to the equipment life cycle degradation trend. The evaluation system is incomplete, there is no hierarchical evaluation logic from "component to device to system", and only qualitative conclusions can be output, which cannot locate the faulty component.

[0004] In summary, the existing technology cannot meet the evaluation needs of the hydropower station for the governor system, such as "accurate identification, operating condition adaptation, and hierarchical positioning", and a new state evaluation method is urgently needed. SUMMARY

[0005] The purpose of the present application is to provide a water turbine governor system state evaluation method to solve the problems of insufficient evaluation accuracy, poor operating condition adaptability, low threshold flexibility and incomplete evaluation system in the existing water turbine governor system state evaluation method.

[0006] To solve the above problems, the technical scheme of the present application is as follows: A water turbine governor system state evaluation method, comprising the following steps: S1: Real-time data acquisition and preprocessing: collecting the guide vane opening, unit frequency, circuit breaker signal, locking state, active power and fault alarm parameters of the governor, and preprocessing the collected raw data by cleaning, denoising, feature extraction and time scale alignment; the high time resolution value range is 0.5ms-1ms; S2: Intelligent condition discrimination: According to the preset condition discrimination rule, the operating conditions of the speed regulator are subdivided into seven typical conditions, including the starting process, the stopping process, the no-load operating condition, the load rejection process, the primary frequency modulation process, the power response process and the load operating process, and a unique identifier is set for each condition; the condition discrimination rule includes: the starting process is determined as the starting command being activated, the locking pin being pulled out, and the guide vane opening being greater than the preset opening threshold (value range 2%-5%); the stopping process is determined as the guide vane opening being less than the preset opening threshold, the locking pin being put in, and the fault alarm not being activated; the no-load operating condition is determined as the starting being successful, the unit frequency being stable near the rated frequency (frequency error being less than the preset frequency deviation threshold, value range 0.1Hz-0.2Hz), and the circuit breaker closing signal not being activated; S3: State index analysis: For each typical condition, the typical index parameters reflecting the control accuracy, dynamic response and stability of the speed regulator under the condition are sorted out, and the characteristic values of the typical index parameters are extracted based on the preprocessed sub-condition data; S4: Threshold comparison: A multi-threshold mode composed of single parameter threshold, multi-parameter comprehensive threshold and dynamic threshold is used to compare the extracted characteristic values of the typical index parameters with the corresponding thresholds to identify parameter abnormalities; wherein the single parameter threshold is determined based on the statistical characteristics of historical operation data, the multi-parameter comprehensive threshold is realized by constructing a typical condition parameter correlation matrix and a joint probability density function, and the dynamic threshold is dynamically adjusted in combination with historical operation data and real-time operation state; S5: Quantitative evaluation: Through data statistics and comparison, abnormality qualitative analysis, deviation calculation and score deduction, combined with the deterioration index calculation and the score deduction rule matrix, the health index of the speed regulator system is synthesized; the deterioration index is calculated based on the deviation of the measured value from the reference value and the preset weight coefficient, and the weight coefficient is determined by the analytic hierarchy process; S6: Hierarchical evaluation: Based on the four-layer health evaluation system of "equipment-equipment component-component state quantity-judgment basis", the overall evaluation conclusion of the equipment unit and system is automatically generated according to the evaluation results of the component state quantity; the four-layer health evaluation system is executed according to the process of "equipment component state quantity characteristic data-component state quantity value calculation-component health state evaluation-equipment health state evaluation".

[0007] Further, the parameters collected in step S1 also include the rotating speed, the load set value, the load increase command, and the load decrease command, and the time resolution of data collection is not less than 1 millisecond. High time resolution can accurately capture the rapid changes of parameters under dynamic conditions such as starting and load rejection, solve the problem of "low data time resolution and insufficient timeliness", and provide high-quality data support for subsequent condition discrimination and index analysis.

[0008] Further, the multiple typical working conditions in step S2 include a start-up process, a shut-down process, an idle running working condition, a load shedding process, a primary frequency modulation process, a power response process, and a load running process. The discrimination rule of the load shedding process is that a circuit breaker opening signal is in an active state at a first time and is in an inactive state within a preset time interval (a value range of 50 milliseconds-150 milliseconds) before the first time, a signal mutation of a sudden load disconnection is captured, and a load is not less than a preset load proportion (a value range of 40%-60% of a rated load), so as to exclude a low-load opening misjudgment.

[0009] Further, the discrimination rules of the primary frequency modulation process and the load running process in step S2 both include that a circuit breaker is in a closing state, a unit frequency is stable near a rated frequency (a frequency error is less than the preset frequency deviation threshold), and an active power is stable in a range of 2%-3% of a rated power; the discrimination rule of the power response process includes that the circuit breaker is in the closing state and at least one of the following conditions is met: a load increasing order is activated, a load decreasing order is activated, or a load setting value change amount is greater than a preset load deviation threshold (a value range of 3%-7% of the rated power); the process refers to that the unit actively adjusts power according to a dispatching instruction, which is distinguished from a passive response to frequency change in the primary frequency modulation.

[0010] Further, the typical index parameters in step S3 include a speed rise time (tSR) corresponding to the start-up process, a rated speed proportion rise time (80% rated speed rise time t0.8), and a rise uniformity index (tSR / t0.8); a lifting amount (a maximum value of a unit axial displacement in the shut-down process) and a guide vane closing rate index corresponding to the shut-down process; a speed fluctuation relative value (xn), a frequency change attenuation degree, a disturbance amount, a frequency overshoot amount, a regulation time, and a fluctuation frequency index corresponding to the idle running working condition; a fluctuation frequency, a minimum speed, a regulation time and peak time ratio, and a transition process time index corresponding to the load shedding process; an opening degree response lag time, a target value proportion rise time, and a stable time index corresponding to the primary frequency modulation process; a disturbance amount, an active power overshoot amount, an average power regulation amount, a power stability index, a regulation time, and a fluctuation frequency index corresponding to the power response process; and an opening degree overshoot amount, an average opening degree regulation amount, and an opening degree deviation index corresponding to the load running process.

[0011] Based on the pretreated sub-working condition data, index values are calculated by using time domain analysis, statistical analysis, and the like, for example: a speed rise time tSR: a time from a speed reaching 10% of a rated value to 90% of the rated value; a speed fluctuation relative value xn: (a maximum value of the speed-a minimum value of the speed) / rated speed×100%; ​​​Power stability index: standard deviation of active power fluctuation / rated active power × 100%.

[0012] Furthermore, the formula for calculating the single parameter threshold in step S4 is as follows: ,in: Let be the threshold of the i-th parameter. The average value of parameter i over the past n days. Let be the standard deviation of parameter i over historical n days (n≥30), and k be the confidence coefficient (k=2 for no-load conditions and k=1.5 for loaded conditions, determined based on statistical significance). The confidence coefficient k is set according to the operating condition type, where the k value ranges from 1.8 to 2.2 for no-load conditions and from 1.3 to 1.7 for loaded conditions. When the value of a single indicator parameter exceeds... (or below the lower threshold) If this occurs, the parameter is initially determined to be abnormal, such as during the power-on process. The system triggers warnings in real time. The threshold is dynamically calculated based on historical data statistical characteristics and can be adjusted synchronously with equipment deterioration and changes in operating status, solving the problem that "fixed thresholds cannot adapt to the deterioration trend of the entire equipment life cycle". At the same time, by using differentiated k values, the thresholds for no-load conditions (sensitive to grid frequency) are more stringent, while the thresholds for load conditions (allowing small fluctuations) are more lenient, reducing the false alarm rate of single-parameter warnings and improving the accuracy of warnings.

[0013] Furthermore, the joint probability density function of the multi-parameter integrated threshold pattern in step S4 is: ,in: The static baseline threshold is determined based on the 95th percentile of historical normal data. This represents the change in ambient temperature (current temperature - historical average temperature, unit: °C). This is the equipment service life coefficient (current service life / design life, dimensionless). The adjustment coefficient; the adjustment coefficient The value range is 0.01 / ℃-0.03 / ℃, and the adjustment coefficient is... The value ranges from 0.08 to 0.12, and all values ​​are obtained by fitting historical operating data. It can adapt to different climate conditions (such as taking the upper limit in high-temperature areas). It can be adapted to equipment of different service ages (such as taking the upper limit for older units), enabling the threshold to dynamically respond to environmental changes and equipment aging, further improving the flexibility of the threshold and the accuracy of the assessment.

[0014] Furthermore, in step S4, the dynamic threshold is determined using a machine learning algorithm, which includes at least one of Long Short-Term Memory (LSTM) networks, Random Forest, or Gradient Boosting Tree (GPRS) algorithms. The adjustment of the dynamic threshold is based on at least one of the following: historical operating data of the speed controller, real-time operating conditions, equipment age, and environmental parameters. Compared to statistical thresholds (based on historical data and subject to lag), machine learning algorithms can adapt to dynamic factors such as equipment degradation, operating condition fluctuations, and environmental changes in real time, adjusting the threshold range in advance to avoid "over-warning" or "missed warnings." Simultaneously, the multi-dimensional adjustment criteria combined with advanced algorithms enable the dynamic threshold to accurately match the actual operating state of the equipment, raising the warning accuracy to a higher level.

[0015] Furthermore, the formula for calculating the degradation index in step S5 is as follows: ,in: These are the measured values ​​of typical indicator parameters. The baseline values ​​for typical indicator parameters (such as the 90%-98th percentile of monthly normal operation data) are used, and A is the weighting coefficient determined using the Analytic Hierarchy Process (AHP). For example, A=0.8 for critical safety indicators and A=0.3 for general indicators. The baseline values ​​are taken from the quantile range (to accommodate the differences in normal data from different power plants), and the weighting coefficients are determined using the AHP method (to avoid subjective bias from expert experience), thereby improving the objectivity and fairness of the quantitative evaluation and providing a basis for subsequent deductions and the synthesis of the health index.

[0016] Furthermore, in step S5, the deduction rule matrix classifies the degradation level according to the range of the degradation index D, including: D < 0.05 is judged as normal (deduct 0 points), 0.05 ≤ D < 0.1 is judged as attention level (deduct 1-2 points), 0.1 ≤ D < 0.2 is judged as abnormal level (deduct 5-10 points), and D ≥ 0.2 is judged as severe level (deduct 20-100 points); the formula for calculating the health index is... The health index H is the number of typical indicator parameters involved in the evaluation, ranging from 0 to 100, with higher values ​​indicating better system health. The health index directly reflects the overall health of the system. Combined with deduction rules, it can pinpoint key deterioration indicators, providing maintenance personnel with clear decision-making directions (such as prioritizing critical indicators) and improving maintenance efficiency. Simultaneously, the quantified results allow for comparison of evaluation results from different periods and different equipment, facilitating long-term tracking of equipment degradation trends.

[0017] Furthermore, in step S6, the health levels of the four-layer health evaluation system are divided according to the range of the health index H, including: H≥90 is judged as excellent (normal inspection is recommended), 80≤H<90 is judged as good (shortening the inspection cycle is recommended), 60≤H<80 is judged as sub-healthy (planned maintenance is recommended), and H<60 is judged as faulty (emergency shutdown maintenance is recommended). The operation and maintenance recommendations corresponding to the health levels can be adaptively adjusted according to the actual operation and maintenance needs of the hydropower station.

[0018] The beneficial effects of this invention are as follows: This invention achieves multi-dimensional technical effects through high-temporal-resolution multi-source data acquisition and preprocessing, intelligent discrimination of seven typical operating conditions, construction of a multi-mode threshold system, and a four-layer hierarchical evaluation system of "equipment-equipment components-component status quantities-judgment basis": First, it significantly improves the accuracy and reliability of the assessment. Multi-source data ensures data integrity, intelligent operating condition discrimination avoids misjudgments across operating conditions, and multi-threshold collaboration (single parameter threshold, multi-parameter comprehensive threshold, dynamic threshold) effectively reduces the false alarm rate and missed alarm rate of early warnings, accurately identifying parameter anomalies and multi-parameter coupled faults. Second, it greatly improves the efficiency of assessment and the convenience of operation and maintenance. Full-process automation reduces manual intervention, significantly shortens assessment time, and the hierarchical evaluation system can accurately locate faulty components. Quantifying health indices and health levels intuitively reflects equipment status, providing a clear basis for operation and maintenance decisions. Third, it has both environmental and energy-saving benefits. By identifying potential faults in advance, preventative maintenance is achieved, extending equipment lifespan and reducing downtime losses. Simultaneously, based on the assessment results, operating parameters are optimized, improving unit operating efficiency, reducing energy consumption per unit of power generation, and reducing pollutant emissions, ultimately providing a reliable guarantee for the safe and stable operation of hydropower stations. Attached Figure Description

[0019] The invention will be further described below with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the status assessment process of the turbine governor system of this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] A method for assessing the condition of a hydro turbine governor system includes the following steps: S1: Real-time data acquisition and preprocessing; S11. Data Acquisition: With a time resolution of 0.5 milliseconds to 1 millisecond (the document mentions a high resolution range of 0.5-2 milliseconds, but in practice, ≥1ms is used to balance accuracy and efficiency), multi-dimensional parameters of the speed controller system are acquired through a high-speed data acquisition module (such as NICDAQ-9178 or Siemens S7-1500 PLC). These parameters include: Key status parameters: guide vane opening (0-100%), unit frequency (49.5-50.5Hz), circuit breaker signal (0=open / 1=closed), lock status (0=engaged / 1=disengaged), active power (0-rated value), fault alarm (0=normal / 1=alarm). Supplementary dynamic parameters: Unit speed (0-rated speed), load setpoint (0-rated power), load increase order (0=not activated / 1=activated), load decrease order (0=not activated / 1=activated).

[0022] During the data acquisition process, GPS is used to synchronize the clock to ensure the consistency of timestamps for all parameters, thus avoiding subsequent analysis errors caused by time deviations.

[0023] S12. Data preprocessing; Perform four key operations on the collected raw data to provide a high-quality data foundation for subsequent analysis: Cleaning: Remove outliers (such as unreasonable data such as "guide vane opening = 120%" or "active power is negative" caused by sensor failure), and use linear interpolation to fill in missing values ​​(such as brief signal interruption); Denoising: Wavelet transform algorithms (such as db4 wavelet basis, decomposed into 3 layers) are used to remove electromagnetic interference (such as 50Hz power grid noise) and mechanical vibration (such as guide vane oscillation noise) while preserving the true trend of parameter changes; Feature extraction: Extract features for different parameter types, such as time domain features (average, maximum, and minimum rotational speed, standard deviation of active power fluctuation), frequency domain features (frequency attenuation coefficient), and process features (acceleration time during startup). Time stamp alignment: unify the timestamps of all parameters to GPS standard time (error <1ms). For example, align the "guide vane opening 001ms data" and "unit frequency 002ms data" to 001ms to ensure the time synchronization of multi-source data.

[0024] S2: Intelligent working condition judgment; Based on the preprocessed synchronization data, the governor's operating conditions are subdivided into seven typical operating conditions according to preset discrimination rules, and a unique identifier is assigned to each operating condition (such as "Start-01" and "No-load-03"). The specific discrimination process is as follows:

[0025] After the judgment is completed, the "operating condition identifier" is bound and stored with the preprocessed data of the corresponding time period (such as binding the "load shedding-04" identifier with all data within 10 seconds after the circuit breaker state change), laying the foundation for the subsequent "operating condition index analysis".

[0026] S3: Status Indicator Analysis: For the seven typical operating conditions identified in the second step, specific index parameters reflecting the governor's control accuracy, dynamic response, and stability were identified, and the characteristic values ​​of these indicators were calculated based on the preprocessed data for each operating condition. The specific analysis process is as follows: S31. Operating Conditions - Index Matching and Eigenvalue Calculation;

[0027] S32. Indicator data storage; The calculated "operating condition-indicator-feature value" associated data is stored in the characteristic parameter library, for example, "2024-05-15 startup process: tSR=18s, t0.8=12s, tSR / t0.8=1.5", which provides an indicator basis for subsequent threshold comparison.

[0028] S4: Threshold comparison; A multi-threshold mode of "single parameter threshold + multi-parameter comprehensive threshold + dynamic threshold" is adopted. The indicator feature value calculated in the third step is compared with the corresponding threshold to accurately identify parameter anomalies. The specific process is as follows: S41. Single-parameter threshold comparison Threshold calculation: Based on more than 30 days of historical normal operation data, according to the formula The formula for calculating the upper and lower thresholds for each indicator is as follows: ,in: : The historical average value of indicator i; : Historical standard deviation of indicator i; k: Confidence coefficient (k=1.8-2.2 for no-load conditions, k=1.3-1.7 for loaded conditions). Anomaly detection: If a certain indicator feature value > Th i or <Th i-min If this is the case, the indicator is initially determined to be "abnormal with a single parameter". For example, during the boot process, tSR=21s>Th tSR =20s ( =16s, =2s, k=2), determine tSR abnormality.

[0029] S42. Multi-parameter comprehensive threshold comparison; Step 1: Construct the parameter correlation matrix: For a specific operating condition (e.g., no-load), calculate the Pearson correlation coefficients of all indicators under that condition, forming an m×m correlation matrix M (where m is the number of indicators for that operating condition). For example, under the no-load condition, " xn and The correlation coefficient is 0.82, indicating a strong positive correlation between the two. Step 2: Calculate the dynamic comprehensive threshold: according to the formula Calculate, where: : The static baseline threshold of the indicator (95th percentile of historical normal data); Current temperature - historical average temperature (°C); Current service life / Design life (dimensionless); =0.01-0.03 / ℃, =0.08-0.12 (fitted from historical fault data); Anomaly detection: If multiple parameters satisfy the "correlation matrix M" pq ≥0.7 (strong correlation) and joint parameter value If so, it is determined to be a "multi-parameter comprehensive anomaly", such as an unloaded operating condition. xn=0.12%, =0.75, joint parameter value =0.35>Th dyn =0.11, indicating an overall anomaly.

[0030] S43. Dynamic threshold comparison Threshold generation: Machine learning algorithms such as LSTM (Long Short-Term Memory Network), Random Forest or Gradient Boosting Tree are used. The training set is "historical operating condition data, indicator data, fault records, environmental parameters (temperature) and equipment service life". The current operating condition and indicator data are input in real time, and the threshold adapted to the current state is dynamically output (e.g., the tSR threshold for aging units is adjusted from 20s to 25s). Anomaly identification: If the indicator characteristic value exceeds the dynamic threshold range, it is judged as "dynamic anomaly". For example, in winter low temperature, the guide vane closing rate threshold is dynamically adjusted from 0.5% / s to 0.6% / s to avoid false anomalies caused by temperature.

[0031] S44. Summary of Abnormal Results; The abnormal results identified by the three threshold patterns (such as "tSR single anomaly") xn- The "Comprehensive Anomaly" is marked as the "Judgment Basis" and passed to the next step of quantitative evaluation.

[0032] S5: Quantitative Evaluation; The abnormal results from the fourth step are transformed into a quantitative health index through a process of "data statistics - anomaly identification - deviation calculation - health index synthesis". The specific process is as follows: S51. Data Statistics and Comparison; The statistical values ​​(average, maximum, and standard deviation) of each indicator under each working condition are calculated monthly and compared with the "monthly benchmark value" (90%-98% quantile of the previous month's normal operating data). For example, "the monthly average tSR during the startup process in May is 17s, and the monthly benchmark value is 18s."

[0033] S52. Qualitative Analysis of Abnormalities; If the statistical value exceeds the monthly baseline range (e.g., tSR monthly average = 19s > 18s), the indicator is characterized as "monthly abnormal"; if it is within the range, it is judged as "normal".

[0034] S53. Deviation Calculation and Deduction; Deterioration index calculation: For abnormal indicators, use the formula... Calculate the degradation index, where: X act : Actual measured statistical value of the indicator; X ref Monthly baseline value; A: Weighting coefficient (determined by the AHP analytic hierarchy process, with safety-critical indicators A=0.6-0.8 and general indicators A=0.3-0.5), for example, guide vane control accuracy weighting A=0.8 and power response general indicator A=0.3; Application of deduction rules: Degradation levels are determined and points are deducted based on the degradation index D, as detailed below:

[0035] For example: if the opening deviation D = 0.4 (severe level) and A = 0.6, the deduction value = 100 × 0.4 = 40 points. S54. Health Index Synthesis According to the formula Calculate the governor system health index (n is the total number of indicators involved in the evaluation). For example, if n=28 and the total deduction value=40, then H=100-40 / 28≈98.57. The higher the value, the healthier the system.

[0036] S6: Hierarchical evaluation; Based on a four-layer health assessment system of "equipment - equipment components - component status variables - judgment criteria", quantitative evaluation conclusions are output from "local components" to "system as a whole". The specific process is as follows: S61. First layer: Judgment criteria; The abnormal results from step four (such as “tSR abnormality” or “opening deviation abnormality”) are used as the original basis for evaluation.

[0037] S62. Second layer: Calculation of component state variables; Map the "judgment criteria" to the state variables of the corresponding device components, according to the formula. Calculate the state variable values, where: Q j : The value of the j-th component's state variable; K j State conversion coefficient (determined based on component factory parameters, such as K=0.9 for guide vane control components); X j : The measured value of the corresponding indicator; For example, "abnormal opening deviation" is mapped to "guide vane control component", X j =5%, K j =0.9, then Q j =0.9×5=4.

[0038] S63. Third Layer: Component Health Evaluation; Calculate the health index of each component: (Score) part The deduction value is the total deduction value for all state variables of the component (where m is the number of state variables of the component). For example, if the deduction value for the guide vane control component is 40 and m = 2, then H part =100-40 / 2=80.

[0039] S64. Fourth Layer: Equipment / System Health Assessment; System health index calculation: weighted summation based on component weights, formula. For example, the guide vane control component has a weight of W=0.3, and the frequency adjustment component has a weight of W=0.25. ), for example H sys =80×0.3+100×0.7=88; Health level and maintenance recommendations: Based on H sys The system categorizes levels and provides recommendations, as follows:

[0040] The operation and maintenance recommendations can be adjusted according to the actual needs of the hydropower station (e.g., for remote power stations, the "good" inspection cycle can be extended to 20 days). The value of H ranges from 0 to 100, with 100 indicating complete health.

[0041] The working process of this invention is a fully automated process from "millisecond-level multi-source data acquisition" to "quantitative health index and operation and maintenance suggestions output". Through the design of "different operating conditions, multiple thresholds, and hierarchical structure", it solves the problems of "fragmented operating conditions, one-sided evaluation, rigid thresholds, and lack of system" in the existing technology, and finally provides accurate and practical governor system status assessment results for hydropower stations.

[0042] The embodiments described in this specification are merely examples of implementations of the inventive concept. The scope of protection of this invention should not be considered as limited to the specific forms stated in the embodiments. The scope of protection of this invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for evaluating the state of a hydro turbine governor system, characterized in that, Includes the following steps: S1: Real-time data acquisition and preprocessing: Acquire the guide vane opening of the speed governor, unit frequency, circuit breaker signal, lock status, active power, and fault alarm parameters, and perform cleaning, noise reduction, feature extraction, and time-scale alignment preprocessing on the acquired raw data. S2: Intelligent Operating Condition Judgment: Based on preset judgment rules, the operating conditions are subdivided into multiple typical operating conditions and assigned unique identifiers; S3: State index analysis: For each typical working condition, the typical index parameters reflecting the speed governor's control accuracy, dynamic response and stability under that working condition are sorted out, and the feature values ​​of the typical index parameters are extracted based on the preprocessed sub-working condition data. S4: Threshold comparison: A multi-threshold mode consisting of single-parameter threshold, multi-parameter comprehensive threshold and dynamic threshold is used to identify anomalies by comparing the index parameter values ​​with the thresholds. S5: Quantitative evaluation: Through data statistics and comparison, qualitative analysis of anomalies, deviation calculation and deduction methods, combined with deterioration index calculation and deduction rule matrix, a health index is synthesized; S6: Hierarchical Assessment: Based on a four-layer health assessment system of "equipment-equipment components-component status quantities-judgment criteria", it automatically generates assessment conclusions for equipment units and the overall system.

2. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, The parameters collected in step S1 also include speed, load setpoint, load increase command, and load decrease command. The time resolution of the data acquisition is 0.5 milliseconds to 1 millisecond.

3. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, The various typical operating conditions in step S2 include the start-up process, shutdown process, no-load operation, load shedding process, primary frequency regulation process, power response process, and load operation process. The judgment rule for the load shedding process is: the circuit breaker trip signal is in the active state at the first moment, in the closed state within the preset time interval before the first moment, and the load is not lower than the preset load ratio.

4. The method for evaluating the state of a turbine governor system according to claim 3, characterized in that, The discrimination rules for the primary frequency regulation process and the load operation process in step S2 include: the circuit breaker is in the closed state, the unit frequency is stable near the rated frequency, and the active power is stable within the preset power fluctuation range; the discrimination rules for the power response process include: the circuit breaker is in the closed state, and at least one of the following is met: the load increase order is activated, the load decrease order is activated, or the change in the load set value is greater than the preset load deviation threshold.

5. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, Typical index parameters in step S3 include: speed rise time, rated speed proportional rise time, and rise uniformity index corresponding to the start-up process; machine lifting amount and guide vane closing rate index corresponding to the shutdown process; relative value of speed fluctuation, frequency change attenuation, disturbance amount, frequency overshoot, adjustment time, and number of fluctuations index corresponding to no-load operation; number of fluctuations, minimum speed, ratio of adjustment time to peak time, and transition process time index corresponding to the load shedding process; opening response lag time, target value proportional rise time, and stabilization time index corresponding to the primary frequency regulation process; disturbance amount, active power overshoot, average power regulation amount, power stability index, adjustment time, and number of fluctuations index corresponding to the power response process; and opening overshoot, average opening regulation amount, and opening deviation index corresponding to the load operation process.

6. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, The formula for calculating the single parameter threshold in step S4 is as follows: ,in: Let be the threshold of the i-th parameter. The average value of parameter i over the past n days. Let i be the standard deviation of parameter i over the historical n days, and k be the confidence coefficient. The confidence coefficient k is set according to the operating condition type, wherein the k value range for no-load operating conditions is 1.8-2.2, and the k value range for loaded operating conditions is 1.3-1.

7.

7. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, The joint probability density function of the multi-parameter integrated threshold pattern in step S4 is: ,in: As a static base threshold, This refers to the change in ambient temperature. This refers to the equipment service life coefficient. The adjustment coefficient; the adjustment coefficient The value range is 0.01 / ℃-0.03 / ℃, and the adjustment coefficient is... The value ranges from 0.08 to 0.12, and all values ​​are obtained by fitting historical operating data.

8. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, In step S4, the dynamic threshold is determined by a machine learning algorithm, which includes at least one of long short-term memory network, random forest, or gradient boosting tree algorithm. The adjustment of the dynamic threshold is based on at least one of the following: historical operating data of the speed governor, real-time operating conditions, equipment age, and environmental parameters.

9. The method for evaluating the state of a turbine governor system according to claim 1, characterized in that, The formula for calculating the degradation index in step S5 is as follows: ,in: These are the measured values ​​of typical indicator parameters. Here, A represents the baseline value of a typical indicator parameter, and A is the weighting coefficient.

10. A method for evaluating the state of a turbine governor system according to any one of claims 1 to 9, characterized in that, In step S5, the deduction rule matrix classifies the degradation level according to the range of the degradation index D, including: D < 0.05 is judged as normal, 0.05 ≤ D < 0.1 is judged as attention level, 0.1 ≤ D < 0.2 is judged as abnormal level, and D ≥ 0.2 is judged as severe level; the formula for calculating the health index is as follows: n is the number of typical indicator parameters involved in the evaluation.

Images