A method for identifying unit shutdown working conditions based on key moments extracted in a shutdown process
By using multi-dimensional time-series data processing and a dual verification mechanism, the start and end times of unit shutdown conditions are accurately identified, solving the positioning deviation and misjudgment problems existing in the prior art and achieving high-precision shutdown condition identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NUCLEAR POWER OPERATIONS RES INST (NPRI)
- Filing Date
- 2026-05-12
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies face problems such as positioning deviation at the start time, unstable identification at the end time, and contradiction between noise resistance and real-time performance when identifying unit shutdown conditions, resulting in low identification accuracy and inability to accurately reflect the physical process.
By constructing and preprocessing multidimensional time-series data, combined with outlier removal and data smoothing and noise reduction, and through trend backtracking and a dual verification mechanism of 'zeroing + steady state maintenance', the system accurately locates the start, speed drop and end times of shutdown, and introduces rebound truncation detection logic to prevent misjudgment.
It significantly improves the positioning accuracy and the purity of identification results at critical moments, can completely reproduce the unit shutdown process, has good robustness and interpretability, and is adaptable to various unit characteristics.
Smart Images
Figure CN122196349A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of industrial equipment operation status monitoring and data analysis technology, specifically relating to a method for identifying unit shutdown conditions based on key moments extracted during the shutdown process. Background Technology
[0002] In intelligent operation and maintenance systems for power, nuclear power, and various large rotating machinery, the automatic identification and retrospective analysis of unit downtime events using historical operating data is a crucial foundation for fault diagnosis, performance evaluation, lifespan management, and anomaly early warning. Accurately defining the start and end boundaries of downtime conditions is essential for constructing high-quality equipment operation maps.
[0003] Existing technologies for identifying shutdown conditions often employ a single fixed threshold method or simple logical rules. For example, they might directly determine the start of shutdown as the active power falls below a certain set value, or the end of shutdown as the rotational speed falls below a certain set value. However, in real-world industrial environments, the data collected by sensors often faces complex quality issues, leading to the following significant drawbacks in existing methods: 1. Starting time positioning deviation: Relying solely on fixed thresholds cannot dynamically perceive the changing trends of unit operating status. When power fluctuates around the threshold or decreases slowly, the identified starting time often lags behind the actual load reduction operation point, resulting in incomplete coverage of the operating condition range.
[0004] 2. Unstable End-of-Term Recognition: At the end of the shutdown process, the speed sensor is often affected by electromagnetic interference or mechanical vibration, resulting in zero drift or low-amplitude jitter. The single threshold method is prone to premature truncation due to "instantaneous zeroing" or recognition delay due to noise tailing, and may even incorrectly include the restart process that immediately follows the shutdown (crossing operating conditions), seriously affecting the recognition accuracy.
[0005] 3. The contradiction between noise resistance and real-time performance: Although conventional smoothing filtering can suppress some high-frequency noise, excessive smoothing will introduce phase delay, causing a systematic shift of key feature moments. This makes the identification results unable to accurately reflect the real time points of the physical process, reducing the interpretability and engineering usability of the analysis results.
[0006] In summary, in the face of complex situations such as missing data, sampling noise, zero-point drift, and continuity across operating conditions, there is an urgent need for a shutdown condition identification method that balances robustness and accuracy. This method should be able to stably extract key feature moments from multi-dimensional time-series data, enabling precise definition of the shutdown process to meet the high standards required for industrial-grade intelligent operation and maintenance. Summary of the Invention
[0007] The purpose of this invention is to provide a method for identifying unit shutdown conditions based on key moments extracted during the shutdown process. This method is mainly used to solve the problems of automatic and accurate positioning and complete identification of key state nodes such as the start of unit shutdown, power returning to zero, speed reduction and final stabilization in complex operating data.
[0008] The technical solution of the present invention is as follows: A method for identifying shutdown conditions based on key moments extracted during unit shutdown, comprising the following steps: Step S1: Construction and preprocessing of multidimensional time series data; Step S2: Outlier removal and data smoothing / denoising; Step S3: Determine the moment when the power returns to zero. ; Step S4: Backtrack to determine the shutdown start time ; Step S5: Determine the anchor point for speed reduction ; Step S6: Fine-tuning the end time based on steady-state maintenance and rebound suppression; Step S7: Reverse positioning to determine the point where the rotational speed decreases; Step S8: Generation and visualization of operating events; Based on the aforementioned identified feature time set Generate a record of shutdown events.
[0009] Step 1 includes acquiring historical operating data of the unit to be monitored, including active power sequences. and rotational speed sequence The raw data is cleaned and time-aligned based on a preset sampling period. We use time interpolation algorithms to fill in missing data and construct a continuous time series with uniform time intervals.
[0010] Step 2 includes processing the original power sequence separately. With the original sequence of rotational speed Robust descrambling is performed, followed by centered sliding window mean filtering, to obtain a smooth sequence. and The specific process is as follows: Step S21: Process the original power sequences respectively With the original sequence of rotational speed Perform the following deburring process; denoted as the sequence to be processed. In the Centered on, half width is Calculate the local median within a local window Absolute deviation from the median and order When the following conditions are met: At that time, the sampling point was identified as a burr point, and the following method was adopted. Alternatively, interpolation of adjacent valid points can be used to replace the spurious sequence. Based on this, the power sequences after deburring were obtained respectively. With rotational speed sequence ; Step S22: Process the deburred sequence , Centralized sliding window mean filtering is applied respectively, with the half-width of the filtering window set to 1. That is, the total window length The smoothing calculation is as follows: Thus, a smoothed power sequence is obtained. With rotational speed sequence This is used for extraction at critical moments later.
[0011] Step 3 includes traversing the smoothed power sequence. Identify and satisfy The earliest moment is marked as the moment when the power reaches zero. ,in The preset power near-zero threshold indicates that the generator set has been disconnected or no longer outputs effective electrical energy.
[0012] Step 4 includes using Starting from, in Perform a reverse backtracking search on the sequence: Setting a downward trend tolerance and amplitude threshold ; If the smoothing power at the current moment Compared to the previous sampling time The increase exceeds the tolerance threshold ,Right now This indicates that the rebound exceeds the tolerance, signifying that the downward trend has been broken and the downward trend has ended; Verify whether the power change amplitude of the backtracking segment is greater than ; Mark the trend inflection point that meets the above conditions as the shutdown start time. .
[0013] Step 5 includes using Starting from the search point, in the rotational speed sequence The algorithm uses an up-down search with a tolerance-tolerant monotonically decreasing detection algorithm to find the first time the rotational speed drops to the low-speed threshold. The moment is marked as the coarse positioning anchor point of the rotational speed. .
[0014] Step 6 includes the anchor point. Then, a controlled search window is constructed with a length of [length missing]. Execute the following refined search logic: Springback cutoff detection: Within the search window, monitor in real time whether the rotational speed exhibits a significant rebound. If such a rebound occurs, [the system will detect it]. satisfy If the bounce threshold is reached, the search will be terminated immediately; Steady-state zeroing verification: Find the earliest moment that satisfies the following dual constraints without triggering the bounce cutoff. : Conditions are set to zero: ,in This represents the minimum rotational speed threshold. Condition 2: Maintain low speed: The continuous Within each sampling point, the rotational speed remained at the steady-state threshold. The following is: Output results: If a match is found that meets the above conditions Then mark it as the final shutdown end time. ; If the search window ends or a bounce truncation is triggered and no point matching the criteria is found, then backtrack using the anchor point, i.e. .
[0015] Step 7 includes finalizing... Starting from, in Backtracking in reverse order of the sequence to find the most recent satisfying condition. The moment when the rotational speed decreases is marked as the moment when the rotational speed decreases. ,in This is the rated speed threshold.
[0016] The beneficial effects of this invention are as follows: By introducing a trend backtracking and a dual verification mechanism of "zero touch + steady state maintenance," this invention effectively suppresses noise glitches and zero-point drift interference, solves the problem of misjudgment caused by sensor jitter in the low-speed range, and significantly improves the accuracy of positioning at critical moments. Simultaneously, the innovatively introduced rebound cutoff detection logic can automatically identify speed rebound, effectively preventing misjudgments under complex operating conditions such as "immediate restart after shutdown," ensuring the purity of the identification results. Furthermore, this method completely reproduces the physical process of the unit from load reduction and disconnection to stable shutdown by locating critical moments. It is logically rigorous, highly interpretable, and robust to data loss and sampling differences, flexibly adapting to the characteristics of various units, providing a reliable basis for shutdown analysis and event backtracking. Attached Figure Description
[0017] Figure 1 The flowchart of a unit shutdown condition identification method based on key moments extracted during the shutdown process provided by the present invention is shown below. Figure 2 A visualization of the shutdown condition identification results. Detailed Implementation
[0018] The present invention will be further described below with reference to specific embodiments. The descriptions of the embodiments below are merely for the purpose of aiding understanding the present invention. It should be noted that those skilled in the art can make various modifications to the present invention without departing from its principles, and these modifications and improvements also fall within the scope of protection of the claims of the present invention.
[0019] The present invention provides a unit shutdown condition identification method based on the extraction of key moments in the shutdown process. This method addresses the problems of existing shutdown condition identification methods, such as delayed start moment positioning and misjudgment of cross-condition due to tailing or rebound of the end moment, which are prone to occur when facing complex conditions such as noise interference, zero-point drift and data loss. The present invention provides a unit shutdown condition identification method that can stably extract key feature moments from multi-dimensional time series data and accurately define the shutdown boundary.
[0020] A method for identifying shutdown conditions based on key moments extracted during unit shutdown includes the following steps: Step S1: Construction and Preprocessing of Multidimensional Time Series Data Acquire historical operating data of the unit to be monitored, including active power sequences. and rotational speed sequence The raw data is cleaned and time-aligned based on a preset sampling period. We use time interpolation algorithms to fill in missing data and construct a continuous time series with uniform time intervals.
[0021] Step S2: Outlier Removal and Data Smoothing & Noise Reduction To eliminate the interference of high-frequency noise and transient abnormal protrusions (glitch) from the sensor on feature point recognition, the original power sequence was processed separately. With the original sequence of rotational speed First, robust descrambling is performed, followed by centered sliding window mean filtering to obtain a smooth sequence. and The specific process is as follows: Step S21: Process the original power sequences respectively With the original sequence of rotational speed Perform the following deburring process; for ease of consistent description, the sequence to be processed will be collectively referred to as... In the Centered on, half width is Calculate the local median within a local window Absolute deviation from the median and order When the following conditions are met: ,in, This is a preset threshold coefficient for burr detection, used to adjust the sensitivity of burr point recognition; The larger the value, the higher the threshold for puncture detection, and the more conservative the anomaly identification. The smaller the value, the lower the threshold for puncture detection, and the more sensitive the anomaly detection. In this case, the sampling point is identified as a puncture point, and the appropriate method is used. Replace with (or interpolate adjacent valid points) to obtain the despiking sequence. Based on this, the power sequences after deburring were obtained. With rotational speed sequence .
[0022] Step S22: Process the deburred sequence , Centralized sliding window mean filtering is applied to both. Let the half-width of the filtering window be... (i.e., total window length) The smoothing calculation is as follows: in, The relative index of the sampling point within the sliding window, with a value range of 1. to ; The sampling time interval between adjacent sampling points. Indicates the current time Offset to the center, forward or backward The sampling time after one sampling period.
[0023] Thus, a smoothed power sequence is obtained. With rotational speed sequence This is used for extraction at critical moments later.
[0024] Step S3: Determine the moment when the power returns to zero. Traversing the smooth power sequence ( t ), identify satisfying The earliest moment is marked as the moment when the power reaches zero. .in The preset power near-zero threshold indicates that the generator set has been disconnected or no longer outputs effective electrical energy.
[0025] Step S4: Backtrack to determine the shutdown start time by Starting from, in ( t Perform a reverse backtracking search on the sequence: Setting a downward trend tolerance and amplitude threshold ; If the smoothing power at the current moment Compared to the previous sampling time The increase exceeds the tolerance threshold ,Right now This indicates that the rebound exceeds the tolerance, signifying that the downward trend has been broken and the downward trend has ended; Verify whether the power change amplitude of the backtracking segment is greater than ; Mark the trend inflection point that meets the above conditions as the shutdown start time. .
[0026] Step S5: Determine the anchor point for speed reduction by Starting from the search point, in the rotational speed sequence The algorithm uses an up-down search with a tolerance-tolerant monotonically decreasing detection algorithm to find the first time the rotational speed drops to the low-speed threshold. The moment is marked as the coarse positioning anchor point of the rotational speed. .
[0027] Step S6: Fine-tuning the end time based on steady-state maintenance and rebound suppression To address the noise trailing and cross-condition misjudgment issues of speed sensors in the low-speed range, at the anchor point... Then, a controlled search window is constructed with a length of [length missing]. Execute the following refined search logic: Springback cutoff detection: Within the search window, monitor in real time whether the rotational speed exhibits a significant rebound. If a springback occurs... satisfy If the rebound threshold is reached, the search will be terminated immediately to prevent the climb phase of the next working condition from being mistaken for the current shutdown process.
[0028] Steady-state zeroing verification: Find the earliest moment that satisfies the following dual constraints without triggering the bounce cutoff. : Condition 1 (Touching Zero): ,in This represents the minimum rotational speed threshold. Condition 2 (low speed maintenance): self The continuous Within each sampling point, the rotational speed remained at the steady-state threshold. The following is: in, The preset number of low-speed hold confirmation points indicates the number of times since the candidate time. Start, continuous Each sampling point must satisfy the condition that the rotational speed does not exceed the steady-state threshold. This is to confirm that the unit has entered a sustained low-speed stable state.
[0029] Output results: If a match is found that meets the above conditions Then mark it as the final shutdown end time. ; If the search window ends or a bounce truncation is triggered and no point matching the criteria is found, then backtrack using the anchor point, i.e. .
[0030] Step S7: Reverse positioning of the speed drop point With the final determination Starting from, in Backtracking in reverse order of the sequence to find the most recent satisfying condition. The moment when the rotational speed decreases is marked as the moment when the rotational speed decreases. .in This is the rated speed threshold.
[0031] Step S8: Generation and Visualization of Operating Events Based on the key moments identified above This generates a stop condition event log. Specifically, it will... The shutdown start time will be determined. Determining the moment when the power returns to zero, Determining the moment when the rotational speed decreases, The shutdown end time is determined; and based on the shutdown start time... With the end of the shutdown Calculate the duration of the downtime condition: The final output includes the shutdown start time, power zeroing time, speed decrease time, shutdown end time, and shutdown duration, forming a shutdown event record. Furthermore, these key moments and shutdown intervals can be marked on the corresponding curves of the power and speed sequences, generating a visualization of the shutdown conditions.
[0032] Example: A method for identifying shutdown conditions based on key moments extracted during unit shutdown includes the following steps: Step S1: Construction and Preprocessing of Multidimensional Time Series Data First, historical operating data of the unit to be monitored is obtained. In this embodiment, the data comes from the operating records of Unit 2 of a nuclear power plant from 00:00:00 on March 30, 2024 to 00:00:00 on May 2, 2024, including active power sequences. and rotational speed sequence The original data was sampled at 1-minute intervals, totaling 47,521 records. The original data was then cleaned and time-aligned based on a preset sampling period. Minutes are used to fill in missing data using time interpolation algorithms (such as linear interpolation) to construct a continuous time series with uniform time intervals.
[0033] Step S2: Outlier Removal and Data Smoothing & Noise Reduction To eliminate the interference of high-frequency noise and transient abnormal protrusions (glitch) from the sensor on feature point recognition, the original power sequence was processed separately. With the original sequence of rotational speed First, robust descrambling is performed, followed by centered sliding window mean filtering to obtain a smooth sequence. and .
[0034] For the original power sequence respectively With the original sequence of rotational speed Perform the following deburring process; for ease of consistent description, the sequence to be processed will be collectively referred to as... In the Centered on, half width is Calculate the local median within a local window Absolute deviation from the median and order (In this embodiment, we take...) That is, the window length is (number of sampling points) when the following condition is met: At that time, the sampling point was identified as a burr point, and the following method was adopted. The replacement is performed to obtain the deburred sequence. (In this embodiment, the threshold coefficient is taken as...) Based on this, the power sequences after deburring were obtained respectively. With rotational speed sequence .
[0035] Deburred sequence , Centralized sliding window mean filtering is applied to both. Let the half-width of the filtering window be... (i.e., total window length) The smoothing calculation is as follows: Thus, a smoothed power sequence is obtained. With rotational speed sequence This is used for subsequent extraction at critical moments. (In this embodiment, it is taken as...) ,Right now The corresponding time span is approximately minute.) Step S3: Determine the moment when the power returns to zero ( ) Traversing the smooth power sequence Identify and satisfy The earliest moment is marked as the moment when the power reaches zero. .in The preset power near-zero threshold is set to 0.1MW in this embodiment.
[0036] Algorithm identification result: Locked time 2024-04-10 01:23:00, which indicates that the generator set has been disconnected or no longer outputs effective electrical energy.
[0037] Step S4: Backtrack to determine the shutdown start time ( ) by Starting from (01:23:00), in Perform a reverse backtracking search on the sequence: Setting a downward trend tolerance MW and amplitude threshold MW; If the smoothing power at the current moment Compared to the previous sampling time The increase exceeds the tolerance threshold ,Right now If the rebound exceeds the tolerance, it indicates that the downward trend has been broken, and the downward trend is judged to have ended.
[0038] Is the power change amplitude of the backtracking segment greater than The trend inflection point that meets the above conditions is marked as the shutdown start time. Algorithm identification result: The time of the backtracking lock is 2024-04-09 19:26:00.
[0039] Step S5: Determine the anchor point for the speed reduction ( ) by Starting from the search point, in the rotational speed sequence The algorithm uses an up-down search with a tolerance-tolerant monotonically decreasing detection algorithm to find the first time the rotational speed drops to the low-speed threshold. The moment is marked as the coarse positioning anchor point of the rotational speed. In this embodiment, it is set rpm. Algorithm recognition result: Locked time 2024-04-10 04:03:00.
[0040] Step S6: Fine-tuning the end time based on steady-state hold and rebound suppression ( ) To address the noise trailing and cross-condition misjudgment issues of speed sensors in the low-speed range, at the anchor point... (04:03:00) After that, a controlled search window is constructed with a length of [length missing]. For each point, execute the following refined search logic: Spin-off detection: Within the search window, monitor in real time whether the rotational speed exhibits a significant rebound. If a momentary... satisfy If the rebound threshold is reached, the search is immediately terminated to prevent the climb phase of the next operating condition from being mistakenly identified as the current shutdown process. This embodiment sets... The rpm was tested and no rebound was triggered.
[0041] State-to-zero verification: Find the earliest moment that satisfies the following dual constraints without triggering the bounce cutoff. : Condition 1 (Touching Zero): ,in The minimum speed threshold is set to 1.0 rpm in this example; Condition 2 (low speed maintenance): self The continuous Within each sampling point, the rotational speed remained at the steady-state threshold. The following is: This example is set up (i.e., 30 minutes) rpm.
[0042] Output results: Find those that meet the above conditions Then mark it as the final shutdown end time. ; If the search window ends or a bounce truncation is triggered and no point matching the criteria is found, then backtrack using the anchor point, i.e. .
[0043] Algorithm recognition result: At 04:03:00 on 2024-04-10, all the above strict conditions were met (reaching zero and remaining below 10 rpm for the next 30 minutes), therefore, the final judgment was... The time was 2024-04-10 04:03:00.
[0044] Step S7: Reverse positioning of the speed drop point ( ) With the final determination Starting from, in Backtracking in reverse order of the sequence to find the most recent satisfying condition. The moment when the rotational speed decreases is marked as the moment when the rotational speed decreases. .in The rated speed threshold is set to 1500 rpm in this example. Algorithm recognition result: Locked time 2024-04-10 00:25:00.
[0045] Step S8: Generation and Visualization of Operating Events Based on the aforementioned identified feature time set Generate a record of shutdown events.
[0046] Based on the above implementation steps, the final recognition results output by the system are shown in Table 1: Table 1: Identification Results of Characteristic Moments of Shutdown Conditions for a Nuclear Power Plant Unit As shown in Figure 2, the figure contains two curves: the solid black line represents active power, and the dashed black line represents rotational speed; the vertical dashed line in the figure indicates... The specific location is visually indicated by the gray shaded area. arrive The downtime operating range.
[0047] Based on the analysis of the results from the implementation examples, the identified characteristic time series ( The standard physical process of the unit from load reduction and disconnection to coasting and stabilization is fully reproduced. The steady-state maintenance verification and rebound suppression mechanism introduced in step S6 effectively eliminates sensor noise interference in the low-speed zone and potential cross-condition misjudgment risks, thus proving that the technical solution has extremely high logical consistency, recognition accuracy and anti-interference robustness when processing real and complex industrial data.
Claims
1. A method for identifying shutdown conditions based on key moments extracted during unit shutdown, characterized in that, Includes the following steps: Step S1: Construction and preprocessing of multidimensional time series data; Step S2: Outlier removal and data smoothing / denoising; Step S3: Determine the moment when the power returns to zero. ; Step S4: Backtrack to determine the shutdown start time ; Step S5: Determine the anchor point for speed reduction ; Step S6: Fine-tuning the end time based on steady-state maintenance and rebound suppression; Step S7: Reverse positioning to determine the point where the rotational speed decreases; Step S8: Generation and visualization of operating events; Based on the aforementioned identified feature time set Generate a record of shutdown events.
2. The shutdown condition identification method based on key moments extracted during unit shutdown as described in claim 1, characterized in that: Step 1 includes acquiring historical operating data of the unit to be monitored, including active power sequences. and rotational speed sequence The raw data is cleaned and time-aligned based on a preset sampling period. We use time interpolation algorithms to fill in missing data and construct a continuous time series with uniform time intervals.
3. The method for identifying shutdown conditions based on key moments extracted during unit shutdown as described in claim 1, characterized in that, Step 2 includes processing the original power sequence separately. With the original sequence of rotational speed Robust descrambling is performed, followed by centered sliding window mean filtering, to obtain a smooth sequence. and The specific process is as follows: Step S21: Process the original power sequences respectively With the original sequence of rotational speed Perform the following deburring process; The sequences to be processed are collectively referred to as... In the Centered on, half width is Calculate the local median within a local window Absolute deviation from the median and order When the following conditions are met: At that time, and adopted Alternatively, interpolation of adjacent valid points can be used to replace the spurious sequence. Based on this, the power sequences after deburring were obtained respectively. With rotational speed sequence ; Step S22: Process the deburred sequence , Centralized sliding window mean filtering is applied respectively, with the half-width of the filtering window set to 1. That is, the total window length The smoothing calculation is as follows: Thus, a smoothed power sequence is obtained. With rotational speed sequence This is used for extraction at critical moments later.
4. The shutdown condition identification method based on key moments extracted during unit shutdown as described in claim 1, characterized in that: Step 3 includes traversing the smoothed power sequence. Identify and satisfy The earliest moment is marked as the moment when the power reaches zero. ,in The preset power near-zero threshold indicates that the generator set has been disconnected or no longer outputs effective electrical energy.
5. The shutdown condition identification method based on key moments extracted during unit shutdown as described in claim 1, characterized in that: Step 4 includes using Starting from, in Perform a reverse backtracking search on the sequence: Setting a downward trend tolerance and amplitude threshold ; If the smoothing power at the current moment Compared to the previous sampling time The increase exceeds the tolerance threshold ,Right now This indicates that the rebound exceeds the tolerance, signifying that the downward trend has been broken and the downward trend has ended; Verify whether the power change amplitude of the backtracking segment is greater than ; Mark the trend inflection point that meets the above conditions as the shutdown start time. .
6. The shutdown condition identification method based on key moments extracted during unit shutdown as described in claim 1, characterized in that: Step 5 includes using Starting from the search point, in the rotational speed sequence The algorithm uses an up-down search with a tolerance-tolerant monotonically decreasing detection algorithm to find the first time the rotational speed drops to the low-speed threshold. The moment is marked as the coarse positioning anchor point of the rotational speed. .
7. The shutdown condition identification method based on key moments extracted during unit shutdown as described in claim 1, characterized in that: Step 6 includes the anchor point. Then, a controlled search window is constructed with a length of [length missing]. Execute the following refined search logic: Springback cutoff detection: Within the search window, monitor in real time whether the rotational speed exhibits a significant rebound. If such a rebound occurs, [the system will detect it]. satisfy If the bounce threshold is reached, the search will be terminated immediately; Steady-state zeroing verification: Find the earliest moment that satisfies the following dual constraints without triggering the bounce cutoff. : Conditions are set to zero: ,in This represents the minimum rotational speed threshold. Condition 2: Maintain low speed: The continuous Within each sampling point, the rotational speed remained at the steady-state threshold. The following is: Output results: If a match is found that meets the above conditions Then mark it as the final shutdown end time. ; If the search window ends or a bounce truncation is triggered and no point matching the criteria is found, then backtrack using the anchor point, i.e. .
8. The shutdown condition identification method based on key moments extracted during unit shutdown as described in claim 1, characterized in that: Step 7 includes finalizing... Starting from, in Backtracking in reverse order of the sequence to find the most recent satisfying condition. The moment when the rotational speed decreases is marked as the moment when the rotational speed decreases. ,in This is the rated speed threshold.