Event detection method and system for mechanical overspeed test of steam turbine generator unit

Abstract

The invention provides an event detection method and system for a mechanical overspeed test of a steam turbine generator unit, relates to the technical field of state monitoring and fault diagnosis of electric power equipment, and aims to solve the problems that existing mechanical overspeed test recognition is prone to false alarm and missing alarm, and positioning at a critical moment is not accurate. The method comprises the following steps: acquiring the active power of a generator and the rotating speed of a steam turbine, carrying out preprocessing and rolling smoothing processing to obtain a smoothing power and smoothing rotating speed sequence, determining a first key time point based on the smoothing power sequence, carrying out segmented backtracking and descending trend detection on the smoothing power sequence, and determining a second key time point. And carrying out descending track detection on the smooth rotating speed sequence, determining a third key time point, backtracking the smooth rotating speed sequence, determining a fourth key time point, determining a fifth key time point based on the original rotating speed, and judging whether a mechanical overspeed test is carried out or not. According to the invention, problems in the prior art are solved, and reliable and automatic identification and accurate event definition of the mechanical overspeed test are realized.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and fault diagnosis technology, and particularly relates to an event detection method and system for mechanical overspeed testing of steam turbine generator sets. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In the safety testing and condition monitoring of steam turbine generator sets, mechanical overspeed testing is a core verification method for examining the mechanical integrity and safety performance of rotating components under overspeed conditions. Reliable automatic identification and accurate event definition in mechanical overspeed testing are key technologies for improving the intelligent operation and maintenance level of the unit, ensuring the integrity of operational audit tracks, and meeting compliance verification requirements.

[0004] Currently, existing automatic identification methods for mechanical overspeed tests mainly rely on simple manual rules or single threshold judgment logic. For example, when "generator active power P drops to zero" or "turbine speed exceeds a certain set limit," a mechanical overspeed test is simply identified. However, this method has certain shortcomings: it lacks joint modeling and analysis of the trend, amplitude, and continuity of the entire shutdown process; it is insensitive to noise, short-term fluctuations, and phased plateaus, easily leading to false alarms or missed alarms; and it is difficult to accurately define and extract multiple key moments in the mechanical overspeed test process, such as the starting point of the decline, the power zeroing anchor point, the stable reference point, and the peak confirmation point. Especially under typical complex operating conditions, such as when the generator active power first decreases and then oscillates, the turbine speed exhibits a slight rebound, or the on-site measured noise is too high, the continuous decline process is easily misjudged as an interruption, resulting in deviations in the definition of the event window and the location of the time point.

[0005] Furthermore, although existing technologies employ data smoothing techniques to suppress noise, they only search for speed peaks based on the smoothed data sequence, ignoring the transient peak features in the original data, which leads to deviations in the identification of peak confirmation points. At the same time, the identification of stable reference points lacks neighborhood continuity correction, making it impossible to accurately trace back and define the true end boundary of the speed range maintained at the rated high level (≥ 1500 rpm), resulting in inaccurate positioning of the starting point of the event window.

[0006] It is evident that existing mechanical overspeed test event detection technologies for steam turbine generator sets rely on manual rules or single thresholds, lack joint modeling and analysis of the dynamic characteristics of the entire shutdown process, and have poor robustness to noise and short-term fluctuations, making them prone to false alarms and missed alarms. Furthermore, their positioning at critical moments is inaccurate, making it difficult to meet the requirements for reliable automated identification and accurate event definition in mechanical overspeed tests during shutdown. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, this invention provides an event detection method and system for mechanical overspeed testing of steam turbine generator sets. Under noise, fluctuation and complex operating conditions, it achieves reliable identification of mechanical overspeed testing and accurate positioning of the time of national construction by combining segmented trend, tolerance two-phase decrease and threshold backtracking.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention provides an event detection method for a mechanical overspeed test of a steam turbine generator set, comprising: Obtain the generator's active power and turbine speed, and perform preprocessing; The preprocessed data is subjected to rolling smoothing to obtain smoothed power sequences and smoothed speed sequences, while the original speed sequences are retained. The first critical time point was determined based on the smoothed power sequence; Using the first critical time point as the endpoint, the smoothed power sequence is backtracked in segments to determine the second critical time point; Based on the second key time point, the descent trajectory of the smooth rotation speed sequence is detected, and the third key time point is determined. Using the first critical time point as the endpoint, the smoothed rotational speed sequence is backtracked, and based on the continuity of the original rotational speed sequence, the fourth critical time point is determined. Within the time interval from the fourth critical time point to the third critical time point, the fifth critical time point is determined based on the original rotational speed sequence; Based on the fifth critical time point, determine whether it is a mechanical overspeed test.

[0009] As one implementation method, the preprocessing process includes: aligning and resampling the generator active power and turbine speed according to the time axis; Missing values ​​are filled with forward padding or amplitude-limited linear interpolation, while preserving the original missing value markers; Outliers are removed and their width is limited.

[0010] As one implementation method, the first key time point is determined based on the smoothed power sequence. Specifically, a power threshold is set, and the time point corresponding to the first smoothed power that is less than or equal to the power threshold in the smoothed power sequence is determined and used as the first key time point.

[0011] As one implementation method, the smoothed power sequence is backtracked in segments and a downward trend is detected, with the first key time point as the endpoint. The specific process is as follows: Set the backtracking segment width and continuous descent tolerance; Using the first critical time point as the endpoint, the smoothed power sequence is iteratively segmented forward along the time axis according to the backtracking segment width; Based on the continuous decline tolerance, a downward trend is detected for each segment. If the current smoothing power value is less than or equal to the sum of the previous sequence value and the continuous decline tolerance, it is determined that the downward trend is maintained; otherwise, it is determined that the downward trend is interrupted. If the segment maintains a downward trend and the number of valid samples within the segment is not less than the set minimum segment length, the current segment is confirmed as a valid downward segment.

[0012] As one implementation method, the second critical time point is determined, and the specific process is as follows: Set the minimum drop threshold; If the current segment shows a downward trend and the magnitude is greater than the minimum downward magnitude threshold, then continue to segment forward; otherwise, take the time point corresponding to the maximum smooth power of the current segment as the second critical time point. When tracing back to the boundary of the smoothed power data, the time point corresponding to the maximum smoothed power value of the last segment is taken as the second key time point.

[0013] As one implementation method, based on the second key time point, a descent trajectory detection is performed on the smoothed rotational speed sequence. The specific process is as follows: Set the target threshold, minimum duration, and reference value, and iterate through the smoothed rotation speed sequence; The tolerance-based continuous decrease algorithm is used to determine the state of each smooth speed. If the current smooth speed is less than or equal to the sum of the reference value and the tolerance, it is determined to maintain the decreasing state; otherwise, it is determined to interrupt the decreasing state and reset the continuous count. When the smoothed speed sequence first drops to the target threshold, and the duration of the drop is greater than or equal to the shortest duration, the index of the current smoothed speed is returned.

[0014] As one implementation method, and to determine the third key time point, the specific process is as follows: Set a low threshold termination condition and use the time point corresponding to the first smooth speed that is greater than the low threshold termination condition as the new starting index. Based on the new starting index, the descent trajectory of the smoothed rotational speed sequence is detected by the tolerance model, and the index that is first less than or equal to the low threshold termination condition and meets the continuous descent condition is determined as the third key time point.

[0015] As one implementation method, the smoothed rotational speed sequence is backtracked with the first critical time point as the endpoint, and the fourth critical time point is determined based on the continuity of the original rotational speed sequence. The specific process is as follows: In the smoothed speed sequence, the time corresponding to the first smoothed speed greater than or equal to the speed threshold is determined and used as the initial fourth key time point; Within a predetermined neighborhood after the initial fourth key time point, the original rotational speed sequence is retrieved to determine the first continuous interval where the original rotational speed is greater than or equal to a set threshold. Update the initial fourth key time point to the last index of the current first continuous interval to obtain the fourth key time point; otherwise, the initial fourth key time point remains unchanged.

[0016] As one implementation method, within the time interval from the fourth key time point to the third key time point, the fifth key time point is determined based on the original rotational speed sequence. The specific process is as follows: Within the time interval defined by the fourth and third key time points, determine the peak value of the original rotational speed sequence and its corresponding time index. Set a peak speed threshold. If the peak value of the original speed sequence is greater than or equal to the peak speed threshold, then the time index corresponding to the current maximum value is used as the fifth key time point.

[0017] A second aspect of the present invention provides an event detection system for mechanical overspeed testing of a steam turbine generator set, comprising: The data acquisition and preprocessing module is used to acquire the generator's active power and the turbine's speed, and to perform preprocessing. The signal smoothing module is used to perform rolling smoothing on the preprocessed data to obtain smoothed power sequences and smoothed speed sequences, while retaining the original speed sequences. The critical time point detection module is used to determine the first critical time point based on the smoothed power sequence; Using the first critical time point as the endpoint, the smoothed power sequence is backtracked in segments to determine the second critical time point; Based on the second key time point, the descent trajectory of the smooth rotation speed sequence is detected, and the third key time point is determined. Using the first critical time point as the endpoint, the smoothed rotational speed sequence is backtracked, and based on the continuity of the original rotational speed sequence, the fourth critical time point is determined. Within the time interval from the fourth critical time point to the third critical time point, the fifth critical time point is determined based on the original rotational speed sequence; The judgment module is used to determine whether a mechanical overspeed test is being conducted based on the fifth key time point.

[0018] The above one or more technical solutions have the following beneficial effects: This embodiment achieves precise location and clear event definition at key time points by constructing an integrated detection scheme that includes power anchor triggering, segmented backtracking positioning, continuous tolerance decrease judgment, and original peak value verification. It is both robust and efficient. Joint modeling and analysis of the coordinated changes in power and speed during shutdown demonstrates high robustness to noise, short-term fluctuations, and phased plateaus. This fundamentally overcomes the shortcomings of existing single threshold or manual rule methods, which are prone to misjudgment under complex operating conditions, thus significantly improving the accuracy and reliability of event identification. Compared with existing technologies, the method in this embodiment has high identification accuracy and significantly reduced false alarm and missed alarm rates.

[0019] This embodiment employs layered positioning at five key time points, resulting in clear boundaries and significantly reducing false positives and false negatives. By introducing segmented backtracking and minimum amplitude constraints, the starting point of the speed decrease is objectively and accurately located, avoiding interference from plateau periods or micro-oscillations. A combined strategy of initial threshold locking and continuous neighborhood updates precisely defines the end boundary of the high-speed range, ensuring the authenticity of the event window's starting point. Peak value verification using the original speed sequence accurately captures transient speed peaks, effectively preventing peak clipping and peak positioning deviations caused by data smoothing. Ultimately, automated and high-precision positioning of five full-series key moments is achieved, providing a reliable data foundation for subsequent analysis and auditing.

[0020] This embodiment employs a smoothing and original dual-track parallel approach and a tolerance-continuous model, which can effectively suppress deviations caused by noise, short-term bounce, and peak clipping, ensuring stable performance even in real industrial environments.

[0021] In terms of engineering implementation, the algorithm relies solely on rolling average and segmented detection, making it suitable not only for offline data analysis but also for lightweight deployment on edge devices or real-time monitoring systems with limited computing resources. Key parameters (threshold, window, tolerance) are configurable, facilitating rapid migration across different machine types and operating conditions. Its continuous update strategy requires the existence of the original rotational speed after the stable point. The stable end of the segment can further filter out occasional spikes and improve the reliability of the judgment. At the same time, the method emphasizes interpretability, and each step produces a visual anchor point and evidence interval. The output also provides five key time points, start and end times and operating condition markers, which are convenient for auditing, review and direct calling of upper-level strategies.

[0022] Compared with existing technologies, this embodiment achieves a balanced advantage in terms of accurate positioning, noise resistance and fault tolerance, computational cost, portability and interpretability. It can significantly improve the event detection rate and boundary accuracy, and reduce maintenance and false alarm costs. It is very suitable for scenarios with strict requirements for process traceability, such as operation auditing, compliance verification, and fault diagnosis review.

[0023] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0024] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0025] Figure 1 This is a flowchart of an event detection method for a mechanical overspeed test of a steam turbine generator set in this embodiment 1; Figure 2 This is a schematic diagram of the overspeed event timing curve and threshold labeling in Implementation 1. Detailed Implementation

[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0028] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0029] Terminology Explanation: Power threshold ( ): Used to determine whether the power is "zero / close to zero", positioning And suppress noise-induced false triggering; Speed ​​threshold ( ): Used to determine whether the rotational speed has reached the stable / baseline condition, assisting in backtracking. The descent phase is defined and set according to the relevant threshold values ​​of the unit's rated capacity; Smooth Window The window length for smoothing and denoising balances robustness and response latency; Continuous decrease in tolerance ( A slight rebound is still considered a decline, to prevent the fluctuations from interrupting the trend; Backtracking segment width ( ): Upper limit of search range, limiting backtracking / lookahead distance to avoid out-of-bounds errors and mismatches; Minimum drop threshold ( ): Effective descent threshold, the minimum velocity descent required to confirm an effective descent, used for positioning. ; Peak speed threshold ( ): Used for The threshold is confirmed to prevent peak clipping or missed detection; Low threshold termination condition ( ): Used for determination The termination threshold near the bottom; the shortest segment length ( ): Used to confirm the minimum effective sample length of the falling / stable segment and filter out short pseudo-segments; Shortest duration ( ): Minimum duration / sample size required to maintain stability during continuous decline; Target threshold ( ): A numerical threshold used to determine the end of descent; the end determination is triggered when the sequence value is first less than or equal to this threshold. Reference value (last_value): The reference value used for comparison with the current value in continuous descent detection, which is updated to the current value as the descent continues; Tolerance model: Let tolerance be a factor. The value is determined to continue decreasing if and only if current_value ≤ last_value + tolerance. Otherwise, the decrease is considered to be interrupted.

[0030] Example 1 This embodiment discloses an event detection method for mechanical overspeed testing of steam turbine generator sets.

[0031] To more clearly illustrate this embodiment, the event detection process for a mechanical overspeed test of a steam turbine generator set can be specifically described as follows: A method for detecting events during mechanical overspeed testing of a steam turbine generator set, comprising: S1. Obtain the generator active power and turbine speed, and perform preprocessing; S2. Perform rolling smoothing on the preprocessed data to obtain smoothed power sequence and smoothed speed sequence, while retaining the original speed sequence; S3. Based on the smoothed power sequence, determine the first critical time point; Using the first critical time point as the endpoint, the smoothed power sequence is backtracked in segments to determine the second critical time point; Based on the second key time point, the descent trajectory of the smooth rotation speed sequence is detected, and the third key time point is determined. Using the first critical time point as the endpoint, the smoothed rotational speed sequence is backtracked, and based on the continuity of the original rotational speed sequence, the fourth critical time point is determined. Within the time interval from the fourth critical time point to the third critical time point, the fifth critical time point is determined based on the original rotational speed sequence; S4. Based on the fifth key time point, determine whether it is a mechanical overspeed test.

[0032] like Figure 1 As shown, in step S1, the generator active power and turbine speed are obtained and preprocessed.

[0033] S101. Obtain the generator active power and turbine speed.

[0034] The operation data of the steam turbine generator unit within a certain time period is collected and sampled at intervals to obtain the generator power and steam turbine speed data.

[0035] In this embodiment, unit operation data within the time range of 2024-03-30 00:00:00 to 2024-05-02 00:00:00 is acquired, with a sampling interval of 1 minute, to obtain time series data containing generator power and turbine speed columns, and with an alignable time index. That is, the acquired data includes power columns, speed columns, and time index columns.

[0036] S102. Preprocess the acquired generator active power and turbine speed data.

[0037] The acquired data is preprocessed, specifically as follows: (1) The generator active power and turbine speed are aligned and resampled according to the time axis.

[0038] This ensures that the generator active power (P) and turbine speed (N), two dissimilar signals, are synchronized in time, providing an effective data foundation for subsequent collaborative analysis of the generator active power (P) and turbine speed (N), and avoiding the failure of the associated logic and the positioning deviation at critical moments caused by time misalignment.

[0039] (2) For missing values, forward filling or amplitude-limited linear interpolation is used, and the original missing value is preserved.

[0040] If the power is missing one non-empty point, it is filled in according to the time neighborhood rule or the amplitude is limited by interpolation, which ensures the continuity of the time series and prevents misjudgment due to interruption in the descent process caused by missing data.

[0041] (3) Remove outliers and perform amplitude limiting and clipping.

[0042] For example, removing non-physical values ​​such as negative speed and limiting and pruning abnormal jumps proactively eliminates abnormal data that may lead to subsequent misjudgment of trends and false detection of peaks.

[0043] The above steps provide high-quality, consistent data input, thereby fundamentally improving the robustness and accuracy of the entire method.

[0044] like Figure 1 As shown, in step S2, the preprocessed data is subjected to rolling smoothing to obtain a smoothed power sequence and a smoothed speed sequence, while retaining the original speed sequence.

[0045] Rolling averages were applied to the preprocessed generator active power (P) and turbine speed (N), respectively, where the window length for the rolling average was [value missing]. Using a centered window calculation method And the minimum number of data points in the window is 1, that is After processing, a smoothed power sequence is obtained ( ) and smooth speed sequence ( ).

[0046] And retain the original rotational speed sequence, denoted as This is used to determine the authenticity of subsequent transient peak values ​​and peak value exceeding the threshold, preventing missed detections caused by smoothing and peak clipping.

[0047] like Figure 1 As shown, in step S3, the first critical time point is determined based on the smoothed power sequence; Using the first critical time point as the endpoint, the smoothed power sequence is backtracked in segments to determine the second critical time point; Based on the second key time point, the descent trajectory of the smooth rotation speed sequence is detected, and the third key time point is determined. Using the first critical time point as the endpoint, the smoothed rotational speed sequence is backtracked, and based on the continuity of the original rotational speed sequence, the fourth critical time point is determined. Within the time interval from the fourth critical time point to the third critical time point, the fifth critical time point is determined based on the original rotational speed sequence.

[0048] S301. Based on the smoothed power sequence, determine the first critical time point.

[0049] The specific process for determining the first critical time point is as follows: (1) Set the power threshold to It is generally set to 0. If there is a slightly positive residual, a tolerance window close to 0 is used.

[0050] (2) Based on the power threshold, in the smooth power sequence, determine the time point corresponding to the first smooth power less than or equal to the power threshold, and use it as the first key time point, i.e. the power zero anchor point.

[0051] In this embodiment, from the smoothed power sequence The first search result in China The time point, and as the first key time point. That is, the power zero anchor point, when hour, This moment is used as the benchmark anchor point for event window and trend backtracking.

[0052] Through the above steps, the zero-power event is robustly detected and false triggering caused by small positive residual values ​​is suppressed, providing a stable time reference for subsequent retrospective analysis of the speed decline trend and precise definition of the event window.

[0053] S302. Using the first critical time point as the endpoint, the smoothed power sequence is backtracked in segments and the downward trend is detected, and the second critical time point is determined.

[0054] (1) Taking the first key time point as the endpoint, the smoothed power sequence is backtracked in segments and a downward trend is detected. The specific process is as follows: 1) Set the backtracking segment width and continuous descent tolerance; In this embodiment, the backtracking segment width is set to... The range is 10~20, preferably. =20, continuously decreasing tolerance tolerance = max(2 rpm, 0.003×rated speed).

[0055] 2) Using the first critical time point as the endpoint, iterate forward along the time axis by taking segments according to the backtracking segment width.

[0056] by As the endpoint, according to Iterate forward to extract segments based on the segment width.

[0057] Obtain the values ​​of the segment boundary points, and denote the smoothed power value of the first sample in the segment as V_start and the smoothed power value of the last sample in the segment as V_end. The segments are arranged in chronological order.

[0058] 3) Based on the continuous decline tolerance, a downward trend is detected for each segment. If the current smoothing power value is less than or equal to the sum of the previous sequence value and the continuous decline tolerance, it is determined that the downward trend is maintained; otherwise, it is determined that the downward trend is interrupted.

[0059] Perform a downward trend detection within each segment. Introduce a continuous decline tolerance to determine if the segment maintains a downward trend. Small fluctuations in the sequence value are allowed, not exceeding the continuous decline tolerance. If the current smoothing power value is not greater than the sum of the previous sequence value and the continuous decline tolerance, then... current ≤ last If the current smoothing power value is greater than the sum of the previous sequence value and the continuous decline tolerance, then the downward trend is considered to continue; if the current smoothing power value is greater than the sum of the previous sequence value and the continuous decline tolerance, then the downward trend is considered to continue. current last If so, it is determined that the segmented trend is interrupted.

[0060] 4) The current segment is confirmed as a valid descending segment only if the trend remains valid and the number of valid samples within the segment is not less than the set minimum segment length.

[0061] The effective sample size within a segment is no less than the shortest segment length. That is, the number of valid samples within a segment needs to be 100. The shortest segment length is set to The commonly used values ​​are 5 or 6.

[0062] 5) Calculate the amplitude for each segment.

[0063] For confirmed valid descent segments, the segment amplitude will be... Defined as the difference between the start and end of a segment, i.e., amplitude = V_start - V_end; for segments that are not confirmed, the amplitude is set to 0.

[0064] Segmentation Amplitude As an indicator of the descent intensity of this segment, it is used in conjunction with the minimum descent threshold. The comparison is used to drive subsequent forward segmentation or to select the point of maximum smooth power within a segment as the second key time point to provide prerequisite support.

[0065] (2) Based on the segmented downward trend and magnitude, the second critical time point is determined. The specific process is as follows: 1) Set the minimum descent threshold =10.

[0066] 2) If the current segment shows a downward trend and the magnitude is greater than the minimum downward magnitude threshold, then continue to segment forward; otherwise, take the time point corresponding to the maximum smooth power of the current segment as the second critical time point.

[0067] 3) When backtracking to the boundary of the smoothed power data, the time point corresponding to the maximum smoothed power value of the last segment is taken as the second key time point.

[0068] If the current segment shows a decrease and the magnitude If the current segment continues to move forward, then... At that time, the maximum value point of that segment is taken as the second critical time point. If the sequence reaches its boundary, the maximum value of the last segment is taken as the second critical time point. This ensures that the starting point of the decline can be robustly determined under plateau / oscillation conditions.

[0069] The backtracking result of this embodiment is: , serving as the starting point and anchor for a significant decline.

[0070] S303. Based on the second key time point, perform descent trajectory detection on the smooth rotation speed sequence and determine the third key time point.

[0071] S3031. Detect the descent trajectory of the smoothed rotational speed sequence. The specific process is as follows: (1) Set the target threshold, the shortest duration and the reference value, and iterate through the smooth speed sequence.

[0072] 1) Set target threshold Shortest duration And the reference value last_value.

[0073] In this embodiment, the target threshold Shortest duration It is an integer, commonly 4 or 5. The reference value `last_value` is initialized to `speed_smooth` at the start of the iteration. Specifically, based on the data fragment or upstream event position, the corresponding start_idx is calculated. A low-threshold cross-search is performed on the smooth speed sequence speed_smooth to find the first value that satisfies speed_smooth[i] > end_threshold_low, typically set to 5, i.e., index. =5, the reference value last_value is initially set to speed_smooth[ The reference value `last_value` is then dynamically updated according to the logic of continuously decreasing tolerance.

[0074] 2) Traverse the smoothed rotation speed sequence starting from the given starting index and maintain the reference value last_value.

[0075] enter: ; Starting index Target threshold Continuous decrease in tolerance Shortest duration .

[0076] (2) The tolerance continuous decrease algorithm is adopted to determine the state of each smooth speed. If the current smooth speed is less than or equal to the sum of the reference value and the tolerance, it is determined to maintain the decreasing state; otherwise, it is determined to interrupt the decreasing state and reset the continuous count.

[0077] If the current smooth speed If, then it is recorded as a decrease and hold. If it occurs, it is recorded as a drop interrupt and reset.

[0078] (3) When the smoothed speed sequence drops to the target threshold for the first time, and the duration of the drop is greater than or equal to the shortest duration, the index of the current smoothed speed is returned.

[0079] When the smoothed speed sequence first drops to the target threshold And satisfy the duration of the descent state When that happens, return the index.

[0080] The arrival time that satisfies the threshold and persistence check is used to robustly identify the end of the shutdown segment to the threshold point.

[0081] S3032. Determine the third critical time point.

[0082] 1) Set a low threshold termination condition and use the time point corresponding to the first smooth speed greater than the low threshold termination condition as the new starting index.

[0083] The low threshold termination condition is In a smooth speed sequence Select the first one The time point as the new starting index .

[0084] In this embodiment, .

[0085] 2) Based on the new starting index, the smooth speed sequence is subjected to a descent trajectory detection using a tolerance model. The index that is first less than or equal to the low threshold termination condition and meets the continuous descent condition is determined as the third key time point.

[0086] Based on the new starting index Execute step S3031, that is = , return to the first drop ( An index that meets the persistence condition is used as the third key time point. This is the point at which the speed decreases to the lower threshold.

[0087] The detection result of this embodiment is: .

[0088] By continuously reducing tolerance With the shortest duration To prevent a short-term rebound from being mistaken for the end.

[0089] S304. Using the first critical time point as the endpoint, backtrack the smoothed speed sequence and determine the fourth critical time point based on the continuity of the original speed sequence.

[0090] Starting from the first critical time point t1, the smoothed rotational speed sequence is backtracked to determine the time corresponding to the first smoothed rotational speed not falling below the speed threshold, which is then used as the initial fourth critical time point. Based on the continuity of the original rotational speed sequence, the fourth critical time point t2 is updated. The specific process is as follows: (1) Using the first critical time point as the endpoint, backtrack the smooth speed sequence to determine the time corresponding to the first smooth speed that is not lower than the speed threshold and use it as the initial fourth critical time point.

[0091] Determining the third key time point Previously, smooth speed sequences Perform a backtracking process from the nearest stable point, searching backwards for points that satisfy the condition. The most recent index, that is, the time corresponding to the smooth rotational speed that is first greater than or equal to the speed threshold, is defined as the initial fourth key time point. .

[0092] To enhance stability, it is required The nearby consecutive points (e.g.) )all .

[0093] Initial results of this embodiment: .

[0094] This strategy balances trend and threshold stability, reducing early triggering caused by noise.

[0095] (2) Based on the continuity of the original rotational speed sequence, the fourth key time point is updated.

[0096] 1) Within a predetermined neighborhood after the initial fourth key time point, the original rotational speed sequence is searched to determine the first continuous interval where the original rotational speed is greater than or equal to a set threshold.

[0097] 2) Update the initial fourth key time point to the last index of the current first continuous interval, which is the fourth key time point. Otherwise, the initial fourth key time point remains unchanged, which is the fourth key time point.

[0098] At the initial fourth critical time point Within the subsequent predetermined neighborhood, based on the original rotational speed sequence A search is performed to identify the first continuous interval with a speed value not lower than 1500 rpm, wherein short-term gaps of no more than 2 data points are allowed within the continuous interval, and the speed value within the gap is not lower than (1500 - tolerance) rpm.

[0099] If the continuous interval is successfully identified, the time index corresponding to the last data point of the continuous interval is updated to the final fourth key time point t2; if no continuous interval that meets the conditions is identified, the initial fourth key time point t2_init is maintained as the final t2. This yields the threshold stabilization point.

[0100] Update results for this embodiment: Ensure that the event window is based on the actual end of the high-speed segment.

[0101] This update mechanism ensures that the defined start point of the event window is consistent with the end boundary of the actual high-speed section.

[0102] S305. Within the time interval from the fourth critical time point to the third critical time point, determine the fifth critical time point based on the original rotational speed sequence.

[0103] Within the time interval from the fourth critical time point to the third critical time point, the original speed peak value is calculated based on the original speed sequence. If the original speed peak value is not lower than the peak speed threshold, the time corresponding to the original speed peak value is taken as the fifth critical time point t4. The specific process is as follows: 1) Within the time interval defined by the fourth critical time point and the third critical time point, determine the maximum value of the original rotational speed sequence and its corresponding time index.

[0104] Within the time interval defined by the fourth key time point t2 and the third key time point t3, that is... Within this timeframe, in this embodiment, it is from 00:54:00 on April 10, 2024 to 04:03:00 on April 10, 2024, based on the original rotational speed sequence. Find its global maximum value and its corresponding time index.

[0105] 2) Set a peak speed threshold. If the peak value of the original speed sequence is greater than or equal to the peak speed threshold, then the time index corresponding to the current maximum value is taken as the fifth key time point.

[0106] Set the peak speed threshold to If the peak value is the maximum value (e.g., 1550) Then the time index corresponding to this maximum value is defined as the fifth key time point. And it was determined that the conditions for the mechanical overspeed test were met, that is Otherwise, Mark it as a null value and determine that the mechanical overspeed test conditions are not met, i.e. .

[0107] The detection results of this embodiment are as follows: the original peak value is 1633 rpm (index 2024-04-10 01:28:00 ≥ 1550 rpm, indicating the presence of...). Based on this, we can obtain , .

[0108] like Figure 1 As shown, in step S4, based on the fifth key time point, it is determined whether it is a mechanical overspeed test.

[0109] Based on steps S1-S3, five key time points are obtained in a structured manner: , , , , .

[0110] The event window is as follows: ;

[0111] Based on the fifth critical time point t4, if the fifth critical time point exists, that is, the mechanical overspeed test conditions are met, it is judged to be a mechanical overspeed test; otherwise, it is an atypical working condition.

[0112] If the fifth critical time point exists, the mechanical overspeed test conditions... If the condition is as described in Table 1, then it corresponds to a typical mechanical overspeed test curve 1 (shutdown); otherwise, it corresponds to an atypical operating condition. Figure 2 As shown.

[0113] Table 1. Key Moments and Judgment Results of Mechanical Overspeed Tests

[0114] The test results are consistent with the factory list, indicating that the data was accurately captured at a 1-minute resolution. – At critical moments, both the event window and peak value determination meet the rule requirements.

[0115] This embodiment, based on a combination of segmented trend analysis and threshold correction, successfully identified the critical moments of mechanical overspeed test events: For 2024-04-09 19:26, It is 01:23. 00:54 =04:03, peak The time is 01:28; the maximum speed of 1633 rpm exceeds the 1550 threshold, which is judged as a typical shutdown curve, such as... Figure 2 As shown in the figure. The method in this embodiment is robust to short-term fluctuations with a 1-minute sampling time, and the event window is accurate and the peak detection is reliable, meeting the needs of engineering applications and audit traceability.

[0116] Example 2 The purpose of this embodiment is to provide an event detection system for mechanical overspeed testing of steam turbine generator sets, including: The data acquisition and preprocessing module is used to acquire the generator's active power and the turbine's speed, and to perform preprocessing. The signal smoothing module is used to perform rolling smoothing on the preprocessed data to obtain smoothed power sequences and smoothed speed sequences, while retaining the original speed sequences. The critical time point detection module is used to determine the first critical time point based on the smoothed power sequence; Using the first critical time point as the endpoint, the smoothed power sequence is backtracked in segments to determine the second critical time point; Based on the second key time point, the descent trajectory of the smooth rotation speed sequence is detected, and the third key time point is determined. Using the first critical time point as the endpoint, the smoothed rotational speed sequence is backtracked, and based on the continuity of the original rotational speed sequence, the fourth critical time point is determined. Within the time interval from the fourth critical time point to the third critical time point, the fifth critical time point is determined based on the original rotational speed sequence; The judgment module is used to determine whether a mechanical overspeed test is being conducted based on the fifth key time point.

[0117] Based on an event detection system for mechanical overspeed testing of a steam turbine generator set, the steps of the event detection method for mechanical overspeed testing of a steam turbine generator set in Embodiment 1 are implemented.

[0118] Example 3 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0119] Example 4 The purpose of this embodiment is to provide a computer-readable storage medium.

[0120] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0121] Example 5 The purpose of this embodiment is to provide a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods and functions involved in any of the above embodiments. The steps and methods involved in the apparatus of the above embodiments correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0122] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0123] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for event detection in a mechanical overspeed test of a turbogenerator unit, characterized in that, The method comprises the following steps: Obtain the active power of the generator and the rotating speed of the steam turbine, and perform preprocessing; Perform rolling smoothing processing on the preprocessed data to obtain a smoothed power sequence and a smoothed rotating speed sequence respectively, and retain the original rotating speed sequence; Determine a first key time point based on the smoothed power sequence; Segment and backtrack the smoothed power sequence forward from the first key time point as the end point, and determine a second key time point; Perform descending trajectory detection on the smoothed rotating speed sequence based on the second key time point, and determine a third key time point; Backtrack the smoothed rotating speed sequence from the first key time point as the end point, and determine a fourth key time point based on the continuity of the original rotating speed sequence; Determine a fifth key time point based on the original rotating speed sequence within the time interval from the fourth key time point to the third key time point; Determine whether it is a mechanical overspeed test based on the fifth key time point.

2. A method of event detection for mechanical overspeed testing of a turbogenerator unit as claimed in claim 1, wherein, The preprocessing process comprises the following steps: Align and resample the active power of the generator and the rotating speed of the steam turbine according to the time axis; Perform forward filling or amplitude limiting linear interpolation processing on the missing values, and retain the original missing labels; 3. A method of event detection for mechanical overspeed testing of a turbogenerator unit as claimed in claim 1, wherein, Remove and clip the abnormal values.

4. A method of event detection for mechanical overspeed testing of a turbogenerator unit as set forth in claim 1, wherein, Determine the first key time point based on the smoothed power sequence, specifically: Set a power threshold, and determine the time point corresponding to the first smoothed power less than or equal to the power threshold in the smoothed power sequence as the first key time point. Segment and backtrack the smoothed rotating speed sequence forward from the first key time point as the end point, and perform descending trend detection, specifically: Set the backtracking segment width and the continuous descending tolerance; Iteratively segment the smoothed power sequence forward along the time axis by the backtracking segment width from the first key time point as the end point; 5. A method of event detection for mechanical overspeed testing of a turbogenerator unit as defined in claim 1, wherein Perform descending trend detection on each segment based on the continuous descending tolerance, if the current smoothed power value is less than or equal to the sum of the previous sequence value and the continuous descending tolerance, it is determined that the descending trend is maintained, otherwise, it is determined that the descending trend is interrupted; When the segment maintains the descending trend and the number of valid samples in the segment is not less than the set minimum segment length, the current segment is confirmed as an effective descending segment. Determine the second key time point, specifically: Set a minimum descending amplitude threshold; 6. A method of event detection for mechanical overspeed testing of a turbogenerator unit as defined in claim 1, wherein, If the current segment has a descending trend and the amplitude is greater than the minimum descending amplitude threshold, continue to segment forward, otherwise, take the time point corresponding to the maximum smoothed power value in the current segment as the second key time point; When the backtracking reaches the boundary of the smoothed power data, take the time point corresponding to the maximum smoothed power value in the last segment as the second key time point. Perform descending trajectory detection on the smoothed rotating speed sequence based on the second key time point, specifically: Set a target threshold, a minimum duration, and a reference value, and traverse the smoothed rotating speed sequence; 7. A method of event detection for mechanical overspeed testing of a turbogenerator unit as defined in claim 1, wherein Use the tolerance continuous descending algorithm to judge each smoothed rotating speed state, if the current smoothed rotating speed is less than or equal to the sum of the reference value and the tolerance, it is determined that the descending state is maintained, otherwise, it is determined that the descending is interrupted and the duration count is reset; When the smoothed rotating speed sequence first falls below the target threshold, and the descending state duration is greater than or equal to the minimum duration, return the index of the current smoothed rotating speed. Determine the third key time point, specifically: A low threshold end condition is set, and a time point corresponding to a first smooth rotation speed greater than the low threshold end condition is taken as a new starting index; Based on the new starting index, the smooth rotation speed sequence is subjected to a descending trajectory detection through a tolerance model, and an index first less than or equal to the low threshold end condition and satisfying continuous descending is taken as a third key time point.

8. A method of event detection for mechanical overspeed testing of a turbogenerator unit as defined in claim 1, wherein, The first key time point is taken as an end point, the smooth rotation speed sequence is traced back, and based on the continuity of the original rotation speed sequence, a fourth key time point is determined, and the specific process is as follows: In the smooth rotation speed sequence, a time corresponding to a first smooth rotation speed greater than or equal to a speed threshold is determined, and taken as an initial fourth key time point; In a predetermined neighborhood after the initial fourth key time point, the original rotation speed sequence is searched, and a first continuous interval in which the original rotation speed is greater than or equal to a set threshold is determined; The initial fourth key time point is updated to the last index of the current first continuous interval, that is, the fourth key time point is obtained, otherwise, the initial fourth key time point remains unchanged.

9. A method of event detection for mechanical overspeed testing of a turbogenerator unit as defined in claim 1, wherein, In the time interval from the fourth key time point to the third key time point, based on the original rotation speed sequence, a fifth key time point is determined, and the specific process is as follows: In the time interval defined by the fourth key time point and the third key time point, the peak maximum value of the original rotation speed sequence and the time index corresponding thereto are determined; A peak speed threshold is set, and if the peak maximum value of the original rotation speed sequence is greater than or equal to the peak speed threshold, the time index corresponding to the current maximum value is taken as the fifth key time point.

10. An event detection system for mechanical overspeed testing of a turbogenerator unit, characterized by It comprises: A data acquisition and preprocessing module for acquiring generator active power and turbine rotation speed and performing preprocessing; A signal smoothing processing module for performing rolling smoothing processing on the preprocessed data to obtain a smooth power sequence and a smooth rotation speed sequence, and retaining an original rotation speed sequence; A key time point detection module for determining a first key time point based on the smooth power sequence; The first key time point is taken as an end point, the smooth power sequence is segmented and traced back forward, and a second key time point is determined; Based on the second key time point, the smooth rotation speed sequence is subjected to a descending trajectory detection, and a third key time point is determined; The first key time point is taken as an end point, the smooth rotation speed sequence is traced back, and based on the continuity of the original rotation speed sequence, a fourth key time point is determined; In the time interval from the fourth key time point to the third key time point, based on the original rotation speed sequence, a fifth key time point is determined; A judgment module for judging whether it is a mechanical overspeed test based on the fifth key time point.

Images