Generator load fluctuation identification method and system based on multi-threshold fusion

Abstract

The invention discloses a generator load fluctuation identification method and system based on multi-threshold fusion, and belongs to the technical field of generator load fluctuation identification, and the method comprises the steps: obtaining the power time sequence data of a generator set; calculating and determining a moving average line and a time slice baseline of the power time sequence data of the generator set; calculating an error between the rated full load value of the generator set and the time slice base line; when the error is smaller than or equal to a set error threshold value, taking a time slice baseline as a judgment reference; when the error is greater than a set error threshold value, taking a fixed lower bound as a judgment reference; and extracting a data sequence smaller than the judgment benchmark from the moving average line, and taking each continuous interval in the extracted data sequence as a load fluctuation interval to obtain a load fluctuation identification result. The problem that a current training scene cannot be dynamically adjusted according to user behaviors is solved. Accurate recognition of the load fluctuation of the generator set is realized, and the problem of inaccurate recognition of the load fluctuation of the generator set at present is solved.

Description

A generator load fluctuation identification method and system based on multi-threshold fusion Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a generator load fluctuation identification method and system based on multi-threshold fusion. 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] With the continuous growth of energy demand and the increasing complexity of power systems, load control of generator sets plays a crucial role in the operation of power systems. In order to achieve accurate control of generator set load, it is necessary to accurately identify generator load fluctuations.

[0004] Existing generator load fluctuation identification technologies mainly rely on a single fixed threshold or simple moving average for limit determination. Such methods typically consider situations where the active power curve exceeds / falls below a preset threshold for a certain period of time as fluctuation events, and roughly extract the "full load - load reduction - recovery" segment through means such as minimum duration filtering, rule engines, or experience windows in order to eliminate instantaneous noise.

[0005] However, the single threshold and fixed window setting method used in the above technology makes it insufficiently adaptable to different unit characteristics, seasonal changes and intraday operating conditions. It is very easy to misjudge normal phenomena such as measurement point jitter and planned load adjustment as abnormal events, or to miss the slow load reduction process, resulting in inaccurate identification of generator load fluctuations. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes a generator load fluctuation identification method and system based on multi-threshold fusion. The method dynamically adjusts the judgment criteria based on the collected power time series data, thereby improving the accuracy of generator load fluctuation identification.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, a generator load fluctuation identification method based on multi-threshold fusion is proposed, comprising: acquiring generator power time series data; calculating and determining the moving average line and time slice baseline of the generator power time series data; calculating the error between the generator rated full load value and the time slice baseline; when the error is less than or equal to a set error threshold, using the time slice baseline as the judgment benchmark; when the error is greater than the set error threshold, using a fixed lower bound as the judgment benchmark; extracting data sequences smaller than the judgment benchmark from the moving average line, and taking each continuous interval in the extracted data sequence as a load fluctuation interval to obtain the load fluctuation identification result.

[0008] Furthermore, load fluctuation intervals with a duration greater than or equal to the minimum duration are selected as candidate intervals. When the time interval between a candidate interval and its adjacent load fluctuation interval is less than the set interval value, and the data segment between the candidate interval and its adjacent load fluctuation interval is not continuously greater than or equal to the judgment benchmark, the candidate interval and its adjacent load fluctuation interval are merged into a single load fluctuation interval.

[0009] Furthermore, the entry and exit times of the load fluctuation range are obtained; from the time series data, the last time point where N consecutive points before the entry time of the load fluctuation range are all greater than or equal to the judgment benchmark is selected as the starting anchor point; the last time point where N consecutive points after the exit time of the load fluctuation range are all greater than or equal to the judgment benchmark is selected as the ending anchor point; from the set width region where the starting anchor point is located and the set width region where the ending anchor point is located, the points corresponding to the maximum power are selected as the boundary points of the load fluctuation range.

[0010] Furthermore, it is determined whether there are missing segments in the time series data; when there are missing segments, and the boundary point of the load fluctuation range is located in the missing segment, the point closest to the load fluctuation range in the missing segment is taken as the boundary point of the load fluctuation range.

[0011] Furthermore, the temporal and data characteristics of each load fluctuation interval are determined; the temporal and data characteristics of each load fluctuation interval are labeled to obtain the load fluctuation identification results.

[0012] Furthermore, the acquired generator power time series data is preprocessed data.

[0013] Secondly, a generator load fluctuation identification system based on multi-threshold fusion is proposed, comprising: a data acquisition unit for acquiring generator power time series data; an average line and baseline determination unit for calculating and determining the moving average line and time slice baseline of the generator power time series data; an error calculation unit for calculating the error between the rated full load value of the generator set and the time slice baseline; a judgment benchmark determination unit for using the time slice baseline as the judgment benchmark when the error is less than or equal to a set error threshold, and using a fixed lower bound as the judgment benchmark when the error is greater than the set error threshold; and a load fluctuation identification result determination unit for extracting data sequences less than the judgment benchmark from the moving average line, and taking each continuous interval in the extracted data sequence as a load fluctuation interval to obtain the load fluctuation identification result.

[0014] Thirdly, a computer device is proposed, the device comprising: a processor adapted to execute a computer program; a computer-readable storage medium storing a computer program, wherein when the computer program is executed by the processor, it implements a generator load fluctuation identification method based on multi-threshold fusion proposed in the first aspect.

[0015] Fourthly, a computer-readable storage medium is proposed, wherein the computer-readable storage medium stores a computer program adapted to be loaded and executed by a processor, the computer program being a generator load fluctuation identification method based on multi-threshold fusion proposed in the first aspect.

[0016] Fifthly, a computer program product is proposed, which includes a computer program that, when executed by a processor, implements a generator load fluctuation identification method based on multi-threshold fusion proposed in the first aspect.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes a generator load fluctuation identification method and system based on multi-threshold fusion. After acquiring the generator power time series data, the method calculates and determines the moving average line and time slice baseline of the generator power time series data. Based on the error between the rated full load value of the generator set and the time slice baseline, the dynamic determination of the judgment benchmark is realized, so that the determined judgment benchmark can adapt to different unit characteristics, seasonal changes and intraday operating condition differences. When the judgment benchmark is used for generator load fluctuation identification, the accuracy of generator load fluctuation identification is improved.

[0018] 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

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0020] Figure 1 is a flowchart of a generator load fluctuation identification method based on multi-threshold fusion disclosed in the embodiment; Figure 2 is the process curve of the first abnormal load fluctuation interval disclosed in the embodiment; Figure 3 is the process curve of the second abnormal load fluctuation interval disclosed in the embodiment; Figure 4 is the process curve of the third abnormal load fluctuation interval disclosed in the embodiment; Figure 5 is the process curve of the fourth abnormal load fluctuation interval disclosed in the embodiment; Figure 6 is the process curve of the fifth abnormal load fluctuation interval disclosed in the embodiment. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. 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 application pertains.

[0023] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0024] First, the application background of the generator load fluctuation identification method based on multi-threshold fusion proposed in the embodiments of the present invention will be described in detail.

[0025] This invention proposes a generator load fluctuation identification method based on multi-threshold fusion, which is applied to the application scenario of generator load fluctuation identification.

[0026] Existing generator load fluctuation identification technologies mainly rely on a single fixed threshold or simple moving average for limit determination. These methods typically consider situations where the active power curve exceeds / falls below a preset threshold for a certain period as fluctuation events. They then roughly extract the "full load – load reduction – recovery" segment using methods such as minimum duration filtering, rule engines, or experience windows to eliminate instantaneous noise. The advantages of these methods are their simplicity and low computational cost. They have achieved initial success in applications such as operation monitoring, alarms, and basic reporting, and possess basic event detection capabilities in typical large fluctuation scenarios.

[0027] However, the aforementioned existing technologies have significant drawbacks: First, the setting of a single threshold and a fixed window results in insufficient adaptability to different unit characteristics, seasonal changes, and intraday operating condition differences. This makes it easy to misjudge normal phenomena such as measurement point jitter and planned load adjustments as abnormal events, or to miss the slow load reduction process. Second, existing technologies generally lack an adaptive calibration mechanism based on time window quantile statistics and moving baselines, making event boundaries highly sensitive to noise and sampling resolution. This leads to deviations in the location of the origin and destination points, as well as distortions in duration and amplitude statistics, which seriously affect the accuracy of dispatch review, performance evaluation, and fault diagnosis. Third, existing solutions generally fail to systematically model the inherent relationship, continuity, and minimum duration of the specific variable load sequence of "full load - reduction to no more than a set percentage and maintenance - recovery to full load," and lack the ability to perform local extreme value alignment and boundary optimization of the origin and destination points, resulting in limited generalization ability across units / time periods. Ultimately, the results of existing technologies mostly remain at the alarm level, lacking event-level visualization, quality filtering, and structured output, which is not conducive to forming a closed-loop operation and maintenance system and traceable data assets.

[0028] The present invention aims to solve the following technical problems in existing generator load fluctuation identification methods: 1. Existing methods rely too much on a single fixed threshold and fixed window, which easily leads to false alarms, missed alarms and inaccurate event boundaries when identifying generator load fluctuations.

[0029] 2. Existing technologies are not adaptable enough to different unit characteristics, seasonal changes and intraday operating conditions, resulting in poor robustness in identification.

[0030] 3. Existing methods are susceptible to noise and sampling resolution in determining the origin and destination points, leading to inaccurate positioning and consequently affecting the accurate statistics of the duration and amplitude of load fluctuation events.

[0031] 4. Existing technologies lack the ability to systematically model the inherent relationship, continuity, and minimum duration of the specific variable load sequence of "full load - load reduction - recovery", and lack a mechanism for local extreme value alignment and boundary optimization of the start and end points, resulting in limited generalization ability across units / time periods.

[0032] To address the aforementioned technical problems, this invention proposes a generator load fluctuation identification method based on multi-threshold fusion. It provides a generator load fluctuation identification framework that integrates fixed thresholds and dynamic baselines. This framework introduces quantile statistics / moving reference, continuity, and minimum duration constraints, and combines hysteresis and quality filtering strategies to stably identify complete fluctuation processes such as "full load—load reduction—recovery." It also improves the ability to distinguish between slow changes and small disturbances, and significantly reduces misjudgments caused by planned adjustments and measurement point jitter.

[0033] This invention proposes a generator load fluctuation identification method based on multi-threshold fusion. It also proposes a mechanism for time alignment and boundary optimization of the start and end points of fluctuation events, combining local extrema / derivative features with robust smoothing to accurately characterize event duration and amplitude. The method supports event-level visualization, quality assessment, and structured result output, thereby improving generalization ability and reusability across units and time periods. While ensuring controllable computational overhead and simplified engineering implementation, it enhances the identification accuracy, stability, and interpretability in operation monitoring, scheduling analysis, and fault early warning scenarios, and promotes the formation of engineering capabilities and traceable data assets for operation and maintenance closed loops.

[0034] As shown in Figure 1, an embodiment of the present invention proposes a generator load fluctuation identification method based on multi-threshold fusion, comprising: acquiring generator power time series data; calculating and determining the moving average line and time slice baseline of the generator power time series data; calculating the error between the rated full load value of the generator set and the time slice baseline; when the error is less than or equal to a set error threshold, using the time slice baseline as the judgment benchmark; when the error is greater than the set error threshold, using a fixed lower bound as the judgment benchmark; extracting data sequences less than the judgment benchmark from the moving average line, and taking each continuous interval in the extracted data sequence as a load fluctuation interval to obtain the load fluctuation identification result.

[0035] This invention proposes a generator load fluctuation identification method based on multi-threshold fusion. After acquiring the generator power time series data, it calculates and determines the moving average line and time slice baseline of the generator power time series data. Based on the error between the rated full load value of the generator and the time slice baseline, it realizes the dynamic determination of the judgment benchmark, so that the determined judgment benchmark can adapt to different unit characteristics, seasonal changes and intraday operating condition differences. When using this judgment benchmark to identify generator load fluctuations, the accuracy of generator load fluctuation identification is improved.

[0036] In some embodiments, the acquired generator power time series data is preprocessed data.

[0037] The time resolution of the acquired unit power time series data is at the minute level, such as 1 minute, and the data format is timestamp-value sequence, with missing values ​​marked as NaN.

[0038] Preprocessing includes time axis continuity checks and smoothing / noise reduction.

[0039] Time axis continuity check refers to building a minute-level continuous time index for the entire period of power time series data; performing continuity verification on the time index; performing forward filling FFill or linear interpolation for short-term missing data (single point to several points) while retaining quality markers; and only marking without interpolation for intervals with continuous missing data exceeding a small threshold to ensure that the review is traceable.

[0040] Smoothing and noise reduction processing involves calculating a moving average MA(t) of the original power sequence with a window length on the order of minutes, within a small to medium range, to suppress high-frequency jitter and maintain the trend. Parameter settings: window W = 30 min; min_periods is set to no less than 1 for robust startup. Note: W is within the recommended range of 10–60 min.

[0041] In some embodiments, a moving average of the generator power time series data is calculated; and the generator power time series data is divided into windows with a long window of days to obtain multiple window data. The median or a specified quantile of each window data is calculated to obtain the time slice baseline A(t) of the generator power time series data; a small scaling factor (less than but close to 1) can be applied to A(t).

[0042] This embodiment of the invention employs a center-aligned multi-day rolling window, calculating the median of the time slice of P(t) as the time slice baseline A(t). Parameter settings: window length is 7 days; quantile is q=50%; A(t) can also be multiplied by a scaling factor κ, where κ is slightly less than 1. Parameter settings: κ=0.99, within the recommended range of 0.95–1.00.

[0043] In some embodiments, the judgment criterion B(t) is dynamically updated and determined based on the time slice baseline A(t)·κ and the fixed lower bound LB. B(t) = the adaptive threshold of the fused {A(t)·κ, LB}: when A(t) is close to the rated full load value of the generator set, and the error between the rated full load value of the generator set and the time slice baseline is within the set error threshold Δ, the time slice baseline A(t)·κ is used as the judgment criterion; otherwise, the fixed lower bound LB is used as the judgment criterion. B(t) is updated over time as a subsequent judgment criterion.

[0044] Set the error threshold Δ to a small bandwidth of tens of power units; LB=full_load×0.99, to avoid excessively raising the threshold.

[0045] In some embodiments, the relationship between the moving average MA(t) and the dynamic threshold B(t) is calculated. Taking the moving average MA(t) < the determination criterion B(t) as an indication of the power reduction state, a Boolean sequence flag(t) is obtained. According to the Boolean sequence flag(t), the data sequence in the moving average that is less than the determination criterion is determined, and each continuous interval in the extracted data sequence is used as a load fluctuation interval.

[0046] Among them, when MA(t) < B(t), flag(t) is true, and flag(t) = True.

[0047] The load fluctuation intervals with a duration greater than or equal to the minimum duration MinDur are used as candidate intervals; when the time interval between a candidate interval and its adjacent load fluctuation interval is less than the set interval value, and the data segment between the candidate interval and its adjacent load fluctuation interval does not continuously exceed the determination criterion, the candidate interval and its adjacent load fluctuation interval are merged into one load fluctuation interval.

[0048] The minimum duration MinDur is within the recommended range of 10 - 30 min, such as MinDur = 15 min; the data sequence interval when the corresponding flag(t) is continuously true and the duration is ≥ MinDur is used as the candidate interval.

[0049] The minimum interval threshold is set to 10 - 30 min; when the interval between adjacent intervals is less than the minimum interval threshold and the moving average of the interval segment is not stably above B(t), it indicates that the two load fluctuation intervals are adjacent, and the two adjacent intervals are merged into one load fluctuation interval as a single load fluctuation event.

[0050] In addition, for the convenience of later review, interval merging can also be not performed, and only records are made for review.

[0051] In some embodiments, the entry time and exit time of the load fluctuation interval are obtained; from the time series data, the last time point where N consecutive points before the entry time of the load fluctuation interval are greater than or equal to the determination criterion is selected as the starting anchor point; the last time point where N consecutive points after the exit time of the load fluctuation interval are greater than or equal to the determination criterion is selected as the termination anchor point; the points corresponding to the maximum power are selected from the set width regions where the starting anchor point and the termination anchor point are located as the boundary points of the load fluctuation interval.

[0052] The time axis of the load fluctuation interval is linearly scanned, the timestamps when flag(t) = True for entry / exit are recorded, a candidate (start, end) list is generated, and the boundary points of the finally determined load fluctuation interval include the starting point S′ and the ending point E′ of the load fluctuation interval.

[0053] Before the start of the load fluctuation range, search forward for the last moment point of "N consecutive points are all greater than or equal to B(t)" as the anchor point S*; search for the point corresponding to the maximum power value within the local window near S* as the aligned starting point S′, and this starting point is recorded as the starting point of the load fluctuation range. Parameter settings: N = 10 points; the search window is a small range of ±10 min.

[0054] After the end of the load fluctuation range, search backward for the last moment point of "N consecutive points are all greater than or equal to B(t)" as the anchor point E*; search for the local maximum value within the local window near E* as the aligned end point E′, and this end point is recorded as the end point of the load fluctuation range. Parameter settings: the same as the starting point (N = 10, window ±10 min).

[0055] Through the positioning of the anchor points of "stable above the threshold" and "local maximum", reduce the boundary offset caused by noise, so that the event covers the complete process of "decline - trough / stability - recovery completion".

[0056] In some embodiments, determine whether there is a missing segment in the time series data; when there is a missing segment and the boundary points of the load fluctuation range are within the missing segment, then use the point closest to the load fluctuation range in the missing segment as the boundary point of the load fluctuation range.

[0057] In the embodiments of the present invention, after determining the abnormal fluctuation range, ultra-long and abnormal filtering and boundary consistency checking are performed on the abnormal fluctuation range.

[0058] Ultra-long and abnormal filtering means: set the upper limit of the duration MaxDur, and remove or list for review the load abnormal fluctuation ranges that exceed the upper limit. Parameter settings: MaxDur = 10 days.

[0059] Boundary consistency checking means verifying that S′ < E′ of the abnormal fluctuation range, there is a segment significantly lower than B(t) inside the edge, and the boundary does not overlap with the missing segment; if the boundary falls into the missing interval, then retreat to the nearest valid point.

[0060] In some embodiments, determine the time characteristics and data characteristics of each load fluctuation range; label the time characteristics and data characteristics of each load fluctuation range to obtain the load fluctuation recognition result.

[0061] Among them, the time characteristics include the start time, end time and duration; formatted as "X days Y hours Z minutes". Among them, the start time is also the start time, the end time is also the end time, and the duration is the time period from the start time to the end time.

[0062] The data characteristics include the start power value, end power value and power change value.

[0063] The power change value can be the difference between the maximum and minimum power values, or the power difference between the start and end points.

[0064] The load fluctuation identification method based on multi-threshold fusion proposed in this invention is used to identify the load fluctuation interval in the power time series data of a generator set. The load fluctuation identification method based on multi-threshold fusion proposed in this invention is verified, and the obtained load fluctuation identification results are shown in Table 1. The corresponding process curves are shown in Figures 2-6.

[0065] Table 1 Load Fluctuation Identification Results

[0066] The power change values ​​in this table are the difference between the maximum and minimum power values ​​within the interval.

[0067] Figures 2 through 6 correspond to the process curves of the five abnormal load fluctuation intervals mentioned above: black represents P(t), gray shading represents the event interval; red dashed line is the starting point, green dashed line is the ending point; dynamic reference line and threshold band are plotted at the same time, and the horizontal axis is the time format.

[0068] The proposed method in this invention has been verified to successfully identify all load change events online at the minute level. Specifically, as shown in Table 1, it accurately captured five abnormal load fluctuation events of the "load reduction-valley / stability-recovery" type, without any omissions, and all identification results highly matched the actual operating patterns of the power plant. In terms of event characteristics, the duration, amplitude changes, and start and end boundaries all match the actual load regulation patterns of the power plant: short-term events correspond to temporary scheduling or unit power fine-tuning within the shift, while long-term events show obvious deep load reduction and gradual recovery trajectories; the valley periods of each event are all in the negative deviation range of the power sequence relative to the dynamic baseline, and the intersection positions of the entry / exit nodes, the extreme values ​​of the rate of change, and the relative deviation thresholds are basically consistent. Combined with the constraints of hysteresis and minimum steady-state duration, the identification results did not show jittery repeated triggering, and can objectively reflect the coupling relationship between the load demand fluctuations on the grid side and the execution rhythm on the unit side. It conforms to the common intraday load allocation, maintenance / switching windows, and seasonal operating condition changes of power plants, and has the authenticity and interpretability for engineering applications.

[0069] This result fully demonstrates that the generator load fluctuation identification based on multi-threshold fusion proposed in this embodiment of the invention can not only capture abnormal load fluctuation events in full coverage, but also accurately distinguish different operating scenarios (such as intraday load allocation, maintenance window switching and seasonal operating condition changes), significantly improving the engineering practicality and interpretability of the identification results.

[0070] This invention proposes a generator load fluctuation identification method based on multi-threshold fusion. Through an integrated scheme of "dynamic baseline + multi-threshold fusion + time alignment", it systematically improves the accuracy, stability and interpretability of generator load fluctuation identification. The specific advantages are as follows: 1. Compared with a single fixed threshold method, this invention can adapt to different unit characteristics, seasonal changes and intraday operating condition differences. By combining quantile statistics and steady-state reference, it significantly reduces the false alarm rate caused by planned adjustments and measurement point jitter.

[0071] 2. By introducing constraints such as relative deviation, rate of change / acceleration, minimum duration, and maximum discontinuity tolerance, this invention takes into account both abrupt and gradual changes, effectively reduces the missed detection of small gradual fluctuations, and improves the ability to identify weak signals.

[0072] 3. The boundary optimization and sequence consistency verification mechanism adopted in this invention performs precise time alignment and robust smoothing of the start and end points and stage structure of fluctuation events, making quantitative indicators such as duration, amplitude and energy more physically consistent and reproducible.

[0073] This invention innovatively integrates multiple dynamic thresholds (including fixed percentage thresholds and dynamic baselines based on time window quantile statistics) and combines them with a series of constraints such as moving average smoothing, continuity, and minimum duration. The aim is to achieve accurate and automated identification of the typical load-changing process of a generator: "full load – reduction to no more than a set percentage and maintenance – recovery to full load." Furthermore, this invention optimizes the start and end point boundaries, achieves time alignment, and supports result visualization and report output. This technology can be widely applied in scenarios such as unit operation monitoring, dispatching support decision-making, fault early warning, and performance evaluation.

[0074] This invention also proposes a generator load fluctuation identification system based on multi-threshold fusion, comprising: a data acquisition unit for acquiring generator power time series data; an average line and baseline determination unit for calculating and determining the moving average line and time slice baseline of the generator power time series data; an error calculation unit for calculating the error between the generator rated full load value and the time slice baseline; a judgment benchmark determination unit for using the time slice baseline as the judgment benchmark when the error is less than or equal to a set error threshold, and using a fixed lower bound as the judgment benchmark when the error is greater than the set error threshold; and a load fluctuation identification result determination unit for extracting data sequences less than the judgment benchmark from the moving average line, and taking each continuous interval in the extracted data sequence as a load fluctuation interval to obtain the load fluctuation identification result.

[0075] The present invention also discloses a computer device, comprising: a processor adapted to execute a computer program; and a computer-readable storage medium storing a computer program, wherein when the computer program is executed by the processor, it implements a generator load fluctuation identification method based on multi-threshold fusion proposed in the embodiments of the present invention.

[0076] The present invention also discloses a computer-readable storage medium storing a computer program adapted for loading and execution by a processor of a generator load fluctuation identification method based on multi-threshold fusion proposed in the embodiments of the present invention.

[0077] The present invention also discloses a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements a generator load fluctuation identification method based on multi-threshold fusion proposed in the embodiments of the present invention.

[0078] The method proposed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0079] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0080] 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 generator load fluctuation identification method based on multi-threshold fusion, characterized in that, include: Obtain time series data of generator set power; Calculate and determine the moving average and time slice baseline of the generator set power time series data; Calculate the error between the rated full load value of the generator set and the time slice baseline; when the error is less than or equal to the set error threshold, the time slice baseline is used as the judgment benchmark; when the error is greater than the set error threshold, the fixed lower bound is used as the judgment benchmark; extract the data sequence that is less than the judgment benchmark from the moving average, and take each continuous interval in the extracted data sequence as a load fluctuation interval to obtain the load fluctuation identification result.

2. The generator load fluctuation identification method based on multi-threshold fusion as described in claim 1, characterized in that, Load fluctuation intervals with a duration greater than or equal to the minimum duration are selected as candidate intervals. When the time interval between a candidate interval and its adjacent load fluctuation interval is less than the set interval value, and the data segment between the candidate interval and its adjacent load fluctuation interval is not continuously greater than or equal to the judgment benchmark, the candidate interval and its adjacent load fluctuation interval are merged into a single load fluctuation interval.

3. The generator load fluctuation identification method based on multi-threshold fusion as described in claim 1, characterized in that, Obtain the entry and exit times of the load fluctuation range; from the time series data, select the last time point in which N consecutive points before the entry time of the load fluctuation range are greater than or equal to the judgment benchmark, and use it as the starting anchor point; The last point in time after the exit time of the load fluctuation range where N consecutive points are greater than or equal to the judgment benchmark is selected as the termination anchor point. Points corresponding to the maximum power values ​​are selected from the set width regions where the starting anchor point and the set width region where the ending anchor point are located, respectively, as the boundary points of the load fluctuation range.

4. The generator load fluctuation identification method based on multi-threshold fusion as described in claim 3, characterized in that, Determine if there are missing segments in the time series data; if there are missing segments, and the boundary point of the load fluctuation range is located within the missing segments, then the point closest to the load fluctuation range within the missing segments is taken as the boundary point of the load fluctuation range.

5. The generator load fluctuation identification method based on multi-threshold fusion as described in claim 1, characterized in that, Determine the time and data characteristics of each load fluctuation interval; label the time and data characteristics of each load fluctuation interval to obtain the load fluctuation identification results.

6. The generator load fluctuation identification method based on multi-threshold fusion as described in claim 1, characterized in that, The acquired generator power time series data is preprocessed data.

7. A generator load fluctuation identification system based on multi-threshold fusion, characterized in that, include: The data acquisition unit is used to acquire time series data of generator set power. The average line and baseline determination unit is used to calculate and determine the moving average line and time slice baseline of the generator power time series data; The error calculation unit is used to calculate the error between the rated full load value of the generator set and the time slice baseline. The judgment benchmark determination unit is used to determine the judgment benchmark by using the time slice baseline when the error is less than or equal to the set error threshold, and by using the fixed lower boundary when the error is greater than the set error threshold. The load fluctuation identification result determination unit is used to extract data sequences that are less than the judgment benchmark from the moving average, and to take each continuous interval in the extracted data sequence as a load fluctuation interval to obtain the load fluctuation identification result.

8. An electronic device, characterized in that, The device includes: a processor adapted to execute a computer program; and a computer-readable storage medium storing a computer program, wherein when the computer program is executed by the processor, it implements the generator load fluctuation identification method based on multi-threshold fusion as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed by a processor to provide a generator load fluctuation identification method based on multi-threshold fusion as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the generator load fluctuation identification method based on multi-threshold fusion as described in any one of claims 1-6.

Images