Fan current detection method and system, computer equipment and storage medium

Abstract

The invention relates to the technical field of draught fan detection, in particular to a draught fan current detection method and system, computer equipment and a storage medium, and the draught fan current detection method comprises the following steps: collecting draught fan current data, analyzing the time sequence characteristics of the draught fan current data, monitoring and comparing the change trend of the distance between adjacent points, and determining the distance between adjacent points; and determining an optimal embedding dimension and time delay through a phase space reconstruction method, and generating a fan current phase space trajectory. According to the method, the time sequence characteristics of the fan current data are analyzed, the embedded dimensions and the time delay parameters are optimized by adopting the phase-space reconstruction method, and the track point distribution mode is deeply analyzed, so that the capability of capturing abnormal behaviors is effectively improved. The transient characteristic signals are classified and indexed by combining the signal decomposition and frequency band screening technology, the short-time abnormal fluctuation in the current signals is accurately identified, and the response capability to the short-time high-frequency signal change is improved.

Description

Technical Field

[0001] The present invention relates to the field of wind turbine detection technology, and in particular to a wind turbine current detection method, system, computer equipment and storage medium. Background Art

[0002] The field of fan detection technology belongs to the branch of mechanical equipment monitoring and diagnosis, and is mainly used to monitor and evaluate the operating status of fan devices such as wind turbines, industrial ventilation equipment, and aerodynamic equipment. This field covers the detection and analysis of fan operating parameters, including vibration, temperature, current, speed, power and other key parameters, in order to achieve fault warning, energy efficiency optimization and analysis of operating life.

[0003] Among them, the fan current detection method is a monitoring and analysis technology for the changes in current parameters during the operation of the fan. By collecting and processing the fan operating current data in real time, this method can effectively identify abnormal conditions in the fan operation, such as motor overload, coil aging, or reduced operating efficiency. Its main uses include fan fault diagnosis, energy consumption analysis, and the formulation of optimized control strategies.

[0004] When monitoring changes in current parameters, existing technologies have difficulty effectively capturing the complex dynamic characteristics of current signals, such as high-frequency short-term fluctuations or transient anomalies, resulting in the inability to timely identify and locate critical faults. For example, when a wind turbine is overloaded for a short period of time, existing technologies usually only focus on the overall current trend and may ignore potential fault signs in subtle fluctuations. In addition, for the assessment of the propagation range of abnormalities, existing technologies usually rely on single-node analysis, which may lead to inaccurate estimates of the impact range of chain abnormalities and increase maintenance difficulties. In terms of fault trend prediction, existing technologies find it difficult to accurately simulate future dynamic changes through historical data. When the operating status of the wind turbine is complex and changeable, this will cause the early warning results to lag or have large errors, affecting the operating safety and maintenance efficiency of the wind turbine. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a wind turbine current detection method, system, computer equipment and storage medium.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a fan current detection method, comprising the following steps: S1: Collect wind turbine current data, analyze the time series characteristics of wind turbine current data, monitor and compare the distance change trend between neighboring points, determine the optimal embedding dimension and time delay through the phase space reconstruction method, and generate the wind turbine current phase space trajectory; S2: based on the wind turbine current phase space trajectory, decomposing the wind turbine current data into multiple characteristic frequency bands, screening the decomposed characteristic frequency bands of the wind turbine current data, identifying transient characteristic signals in the wind turbine current data, and establishing a transient signal feature set; S3: Based on the transient signal feature set, construct current feature nodes of wind turbines in the wind farm, determine the dependency strength between current feature nodes by comparing frequency band features between current feature nodes, identify abnormal current feature nodes and their influence range, and generate an abnormal propagation path set; S4: Based on the transient signal feature set, a sliding window is used to extract local features of the transient signal feature set, abnormal fluctuation intervals and fault sources of wind turbine current data are marked, and a local fluctuation abnormal feature set is generated; S5: Based on the abnormal propagation path set and the local fluctuation abnormal feature set, analyze the change of wind turbine current data in the wind turbine current phase space trajectory, predict the dynamic change trend of wind turbine current data at future moments, and generate wind turbine current abnormality warning results.

[0007] As a further solution of the present invention, the wind turbine current phase space trajectory includes an optimized time series embedding dimension, a determined time delay range and a corresponding distribution pattern of trajectory points in the phase space; the transient signal feature set includes a screened characteristic frequency band set, a transient signal after amplitude normalization and baseline drift correction, and a classification mark and feature description of the transient signal; the abnormal propagation path set includes the characteristic frequency band distribution of abnormal nodes, a propagation path network diagram between abnormal nodes, and a quantitative result of the abnormal impact range; the local fluctuation anomaly feature set includes a marked significant fluctuation interval, the fluctuation amplitude statistical results of the first-order difference and the second-order difference, and a set of significant change points identified based on local variance analysis; the wind turbine current abnormal warning result includes a dynamic change trend prediction diagram of future wind turbine current data, an abnormal pattern distribution of prediction deviation, and multi-dimensional abnormal warning information combining propagation path and fluctuation characteristics.

[0008] As a further solution of the present invention, the wind turbine current data is collected, the time series characteristics of the wind turbine current data are analyzed, the distance change trend between neighboring points is monitored and compared, the optimal embedding dimension and time delay are determined by the phase space reconstruction method, and the specific steps of generating the wind turbine current phase space trajectory are as follows: S101: Based on the collected wind turbine current data, analyze its time series characteristics, extract the time point distribution of the current data, the data amplitude variation range, and the time dependency characteristics in sequence, and generate a time series characteristic set; S102: Based on the time series characteristic set, reconstruct the time point distribution of the current data in phase space, gradually adjust the embedding dimension and time delay parameter, analyze the trend of trajectory point neighbor distance change, and obtain the optimal embedding parameter set; S103: Based on the optimal embedding parameter set, the wind turbine current data is embedded in dimension adjustment and time delay processing is performed, a phase space trajectory diagram is constructed, and a trajectory point distribution pattern is analyzed to generate a wind turbine current phase space trajectory.

[0009] As a further solution of the present invention, based on the wind turbine current phase space trajectory, the wind turbine current data is decomposed into multiple characteristic frequency bands, the decomposed characteristic frequency bands of the wind turbine current data are screened, and the transient characteristic signals in the wind turbine current data are identified. The specific steps of establishing the transient signal feature set are as follows: S201: Based on the wind turbine current phase space trajectory, the wind turbine current data is subjected to amplitude normalization processing, and the overall signal range is rescaled using the upper and lower limits of the amplitude data to eliminate the amplitude difference and obtain a normalized current signal; S202: performing a baseline drift correction operation based on the normalized current signal, eliminating the influence of low-frequency errors through a signal mean trend, decomposing the signal into multiple characteristic frequency bands, and obtaining a frequency band decomposition signal; S203: Based on the frequency band decomposition signal, characteristic frequency bands are screened and transient signals are classified, and a transient signal feature set is established by extracting the frequency, amplitude and duration of each transient signal and classifying them in sections.

[0010] As a further solution of the present invention, based on the transient signal feature set, the current feature nodes of the wind turbines in the wind farm are constructed, the frequency band features between the current feature nodes are compared, the dependency strength between the current feature nodes is determined, the abnormal current feature nodes and their influence range are identified, and the specific steps of generating the abnormal propagation path set are as follows: S301: Based on the transient signal feature set, analyze the frequency band characteristics of the transient signal, including frequency range, signal strength, duration and change mode, extract characteristic indicators of the transient signal, and generate a transient signal frequency band indicator set; S302: Based on the transient signal frequency band indicator set and in combination with the distribution of transient signal characteristics among nodes, the current characteristic nodes of wind turbines in the wind farm are identified and classified, the frequency band range and signal characteristics of the nodes are defined, and a current characteristic node distribution map is generated; S303: Based on the current characteristic node distribution diagram, calculate the dependency strength of characteristic frequency bands between nodes, analyze the influence range and propagation path between nodes, identify the propagation relationship and range of abnormal nodes, and generate an abnormal propagation path set; The dependence strength of the characteristic frequency bands between nodes is calculated using the formula: ; Get Node and nodes The strength of the dependence ; in, Is a node At the point in time The signal strength of the frequency band at Is a node At the point in time The signal strength of the frequency band at is the total number of time points of the signal.

[0011] As a further solution of the present invention, based on the transient signal feature set, a sliding window is used to extract local features of the transient signal feature set, abnormal fluctuation intervals and fault sources of wind turbine current data are marked, and a local fluctuation abnormal feature set is generated; the specific steps are as follows: S401: Based on the transient signal feature set, the signal data is segmented by a sliding window, the change rate and trend of the first-order difference and the second-order difference are obtained section by section, the differential dynamic features of each section are extracted, and a sliding window differential dynamic set is generated; S402: Based on the sliding window difference dynamic set, analyze the local characteristics in the segment, obtain the local variance and identify the segment with significant fluctuation, locate the key characteristics with abnormal behavior through the analysis of the change pattern, and generate a local fluctuation dynamic characteristic set; S403: Based on the local fluctuation dynamic characteristic set and in combination with the marked significant fluctuation area, the abnormal fluctuation intervals are classified and the fault source is located, key abnormal points and fault characteristics in the wind turbine current data are marked, and a local fluctuation abnormal feature set is generated.

[0012] As a further solution of the present invention, based on the abnormal propagation path set and the local fluctuation abnormal feature set, the changes of the wind turbine current data in the wind turbine current phase space trajectory are analyzed, the dynamic change trend of the wind turbine current data at future moments is predicted, and the specific steps of generating the wind turbine current abnormal warning result are as follows: S501: Based on the abnormal propagation path set and the local fluctuation abnormal feature set, the abnormal nodes and fluctuation intervals are mapped to the wind turbine current phase space trajectory, the distribution characteristics and dynamic change trends of the abnormal points in the trajectory are analyzed, and a trajectory abnormal dynamic characteristic set is generated; S502: Based on the trajectory abnormality dynamic characteristic set and in combination with the time series characteristics of the wind turbine current data, a dynamic change trend prediction of the current data value at a future time point is performed, the difference between the predicted value and the actual observed value is determined, and a current data prediction deviation set is generated; The dynamic change trend of the current data value at a future time point is predicted using the formula: ; Calculate future time points Predicted current data value ; in, Indicates the current time point The actual current value, Indicates current data value The weight of the predicted value, Indicates the prediction start time, It is the current moment The time step index after , is the time step between the future moment and the present moment; S503: Based on the current data prediction deviation set, classify the abnormal patterns appearing in the deviation and analyze the impact range, mark the specific characteristics and occurrence conditions of the abnormal patterns, and generate a wind turbine current abnormality warning result.

[0013] The wind turbine current detection system is used to perform the wind turbine current detection method, and the system comprises: The data processing module analyzes the time series characteristics of the collected wind turbine current data, extracts the time point distribution of the current data, the range of data amplitude variation, and the time dependence characteristics, reconstructs the characteristic set in phase space, gradually adjusts the embedding dimension and time delay parameters, and generates the wind turbine current phase space trajectory; The transient feature analysis module is based on the wind turbine current phase space trajectory, uses the signal mean trend to eliminate the influence of low-frequency errors, decomposes the signal into multiple characteristic frequency bands, extracts the frequency, amplitude and duration of each transient signal, performs classification marking and segmentation classification, and establishes a transient signal feature set; The abnormal propagation path analysis module analyzes the frequency band characteristics of the transient signal one by one based on the transient signal feature set, extracts the frequency range, signal strength and change mode, identifies the current feature nodes with abnormal frequency bands and their influence range through similarity comparison of feature frequency bands, and generates an abnormal propagation path set; The abnormal warning module performs sliding window segmentation on the transient signal features based on the abnormal propagation path set and the transient signal feature set, analyzes the local fluctuation characteristics of each section, predicts the dynamic change trend of the wind turbine current data, and generates abnormal wind turbine current warning results.

[0014] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the wind turbine current detection system as described above when executing the computer program.

[0015] A computer-readable storage medium stores a computer program, which implements the steps of the wind turbine current detection method as described above when executed by a processor.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by analyzing the time series characteristics of wind turbine current data, the phase space reconstruction method is used to optimize the embedding dimension and time delay parameters, and the trajectory point distribution pattern is deeply analyzed, which effectively improves the ability to capture abnormal behaviors. Combined with signal decomposition and frequency band screening technology, transient characteristic signals are classified and indexed, and short-term abnormal fluctuations in current signals are accurately identified, thereby improving the response capability to short-term high-frequency signal changes. By constructing feature nodes and analyzing their dependency strength and propagation paths, the quantification and impact assessment of the abnormal propagation range are achieved, and the diagnostic capability of complex chain abnormalities is enhanced. In local dynamic analysis, the first-order and second-order differential characteristics are extracted using a sliding window, and significant fluctuation intervals are marked and key fault sources are located, thereby optimizing the accuracy of abnormal characteristic extraction and future trend prediction, improving the monitoring accuracy of the wind turbine operating status and fault warning capability, and providing comprehensive data support for abnormal diagnosis and preventive maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 This is a detailed flow chart of S1 of the present invention; Figure 3 This is a detailed flow chart of S2 of the present invention; Figure 4 This is a detailed flow chart of S3 of the present invention; Figure 5 This is a detailed flow chart of S4 of the present invention; Figure 6 This is a detailed flow chart of S5 of the present invention; Figure 7 is a system flow chart of the present invention; DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0020] See also Figure 1 The present invention provides a technical solution: a method for detecting a fan current, comprising the following steps: S1: Collect wind turbine current data, analyze the time series characteristics of wind turbine current data, monitor and compare the distance change trend between neighboring points, determine the optimal embedding dimension and time delay through the phase space reconstruction method, and generate the wind turbine current phase space trajectory; S2: Based on the wind turbine current phase space trajectory, the wind turbine current data is decomposed into multiple characteristic frequency bands, the decomposed characteristic frequency bands of the wind turbine current data are screened, the transient characteristic signals in the wind turbine current data are identified, and a transient signal feature set is established; S3: Based on the transient signal feature set, construct the current feature nodes of the wind turbines in the wind farm, determine the dependency strength between the current feature nodes by comparing the frequency band features between the current feature nodes, identify the abnormal current feature nodes and their impact range, and generate the abnormal propagation path set; S4: Based on the transient signal feature set, a sliding window is used to extract local features of the transient signal feature set, mark the abnormal fluctuation interval and the fault source of the wind turbine current data, and generate a local fluctuation abnormal feature set; S5: Based on the abnormal propagation path set and the local fluctuation abnormal feature set, analyze the changes of wind turbine current data in the wind turbine current phase space trajectory, predict the dynamic change trend of wind turbine current data at future moments, and generate wind turbine current abnormality warning results.

[0021] The wind turbine current phase space trajectory includes the optimized time series embedding dimension, the determined time delay range and the corresponding distribution pattern of trajectory points in the phase space. The transient signal feature set includes the screened characteristic frequency band set, the transient signal after amplitude normalization and baseline drift correction, and the classification mark and feature description of the transient signal. The abnormal propagation path set includes the characteristic frequency band distribution of abnormal nodes, the propagation path network diagram between abnormal nodes, and the quantitative results of the abnormal impact range. The local fluctuation anomaly feature set includes the marked significant fluctuation interval, the fluctuation amplitude statistical results of the first-order difference and the second-order difference, and the significant change point set identified based on local variance analysis. The wind turbine current abnormal warning results include the dynamic change trend prediction diagram of future wind turbine current data, the abnormal pattern distribution of prediction deviation, and multi-dimensional abnormal warning information combining the propagation path and fluctuation characteristics.

[0022] See also Figure 2 , collect wind turbine current data, analyze the time series characteristics of wind turbine current data, monitor and compare the distance change trend between neighboring points, determine the optimal embedding dimension and time delay through the phase space reconstruction method, and generate the wind turbine current phase space trajectory. The specific steps are as follows: S101: Based on the collected wind turbine current data, analyze its time series characteristics, extract the time point distribution of the current data, the data amplitude variation range, and the time dependency characteristics in sequence, and generate a time series characteristic set; Collect wind turbine current data, including three-phase current values, corresponding timestamps, and wind turbine operating status marks, and import the data into MATLAB for processing. Extract the time point distribution of current data. Arrange all timestamps in order, analyze whether the intervals between two adjacent timestamps are consistent, record the time points when the intervals are interrupted or fluctuate significantly, and determine whether these time points correspond to load changes or shutdown events in combination with the wind turbine operation log, so as to establish the distribution characteristics of the time points. Extract the amplitude variation range of current data. First, group the current values ​​of each time period. According to the current range during normal operation of the wind turbine (such as the rated current fluctuates by 10%), filter out the peak and valley values ​​that exceed the range, and mark the period with abnormal amplitude in combination with the operating load information in the time period. Extract the time dependency characteristics of current data. Perform trend analysis on the change trend of each phase current, observe the increase and decrease relationship of current in adjacent time periods, match the change trend of current value with the change of wind turbine speed or operating conditions, extract the characteristic time periods with significant dependence, and combine these steps to form the wind turbine current time series feature set.

[0023] S102: Based on the time series characteristic set, the time point distribution of the current data is reconstructed in phase space, the embedding dimension and time delay parameter are gradually adjusted, the trend of the change of the trajectory point neighbor distance is analyzed, and the optimal embedding parameter set is obtained; The sorted wind turbine current time series data is imported into the Python environment, and the phase space reconstruction process is operated using NumPy and Matplotlib tools. Taking wind turbine current detection as the scenario, the time series data of three-phase current during operation is selected. First, the initial phase space trajectory diagram is generated by setting the embedding dimension to 3 and the time delay to one-tenth of the wind turbine rotation period. The diagram shows the distribution characteristics of the current value in the three-dimensional space during wind turbine operation over time. It is observed that when the wind turbine is in a stable operating state, the phase space trajectory diagram appears as a periodic closed curve, indicating that the current changes smoothly; when the wind turbine encounters a sudden load change (such as a sharp increase in wind speed), the trajectory diagram is significantly offset and forms a bifurcated line, reflecting a large fluctuation in the current value. In the cyclic comparison process, by increasing the embedding dimension to 4 and redrawing the trajectory diagram, the details of the current fluctuation of the wind turbine are more obvious; at the same time, the time delay is adjusted to one-fifth of the wind turbine rotation period, and it is found that the distribution density of the trajectory is more uniform, which can clearly identify the current fluctuation characteristics under abnormal load. Finally, according to the periodic closure characteristics and distribution density in the trajectory diagram, the parameter combination suitable for the wind turbine current data is selected, and the corresponding embedding dimension and time delay results are saved as the optimal embedding parameter set.

[0024] S103: Based on the optimal embedding parameter set, the embedding dimension is adjusted and the time delay is processed for the wind turbine current data, a phase space trajectory diagram is constructed and the trajectory point distribution pattern is analyzed to generate a wind turbine current phase space trajectory; The three-phase current time series data collected during the operation of the wind turbine are mapped to the phase space. After importing the determined embedding dimension and time delay parameters, the three-phase current values ​​are grouped by time axis using Pandas in Python to generate a multidimensional representation of each sampling point in the time series. The spatial relationship of adjacent sampling points in the time series is analyzed and mapped in combination with the SciPy library. The three-dimensional coordinates of each trajectory point in the phase space are obtained, and the continuous sampling points in the time series are connected to form a three-dimensional trajectory line. The Plotly library is used to generate a visual trajectory diagram to specifically show the dynamic change behavior of the wind turbine current over time. For the normal operation of the wind turbine, the trajectory line usually presents a periodic closed curve, showing the regularity of the current change; when the wind turbine load fluctuates or the operation is abnormal, the trajectory diagram will show obvious offset, divergence or mutation characteristics, indicating the irregularity of the current change. Combined with the data characteristics of the trajectory map, the trajectory line is divided into several areas, the starting and ending point coordinates are marked, the spatial coverage of the area and the dense distribution of the trajectory line are analyzed, and specific abnormal areas are marked, such as line bifurcations caused by abnormal current fluctuations or trajectory deviations caused by abnormal current amplitude. Through these data analysis and area marking, the wind turbine current phase space trajectory dataset is finally generated.

[0025] See also Figure 3 Based on the wind turbine current phase space trajectory, the wind turbine current data is decomposed into multiple characteristic frequency bands, the decomposed characteristic frequency bands of the wind turbine current data are screened, and the transient characteristic signals in the wind turbine current data are identified. The specific steps of establishing the transient signal feature set are as follows: S201: Based on the wind turbine current phase space trajectory, the wind turbine current data is subjected to amplitude normalization processing, and the upper and lower limits of the amplitude data are used to rescale the overall signal range to eliminate the amplitude difference and obtain a normalized current signal; The phase space trajectory data is loaded through the MATLAB analysis tool, the current value amplitude corresponding to each data point in the trajectory is extracted, and the global maximum amplitude and minimum amplitude of the current trajectory are recorded. The normalization process first calculates the amplitude proportional coefficient based on the extracted upper and lower limits, and scales the amplitude of each current data point to the standard range. For example, the normalization range is 0 to 1, and the current value of each point in the trajectory is mapped to this range through proportional transformation, while ensuring that the trajectory shape remains unchanged. During the normalization process, the extreme points of the signal are individually checked to prevent the error diffusion caused by extreme values. The normalized trajectory retains the relative amplitude change characteristics of the original data, provides a unified amplitude scale for subsequent baseline drift correction and signal decomposition, and finally generates a normalized current signal.

[0026] S202: performing a baseline drift correction operation based on the normalized current signal, eliminating the influence of low-frequency errors through the signal mean trend, decomposing the signal into multiple characteristic frequency bands, and obtaining a frequency band decomposition signal; First, the normalized three-phase current data is analyzed point by point, and the low-frequency drift is identified by extracting the amplitude center in each time period. The process of extracting the amplitude center includes the following steps: segmenting the normalized current signal according to the time period, sorting all the data points of each current signal, extracting the median value of the signal as the amplitude center, and recording the starting amplitude and ending amplitude of each signal at the same time, calculating the difference between the two and judging whether there is a drift trend in the amplitude center. If the difference exceeds the set threshold, it is marked as low-frequency drift. For the marked drift signal, the trend line fitting correction method is used to adjust the drift part to the horizontal line with the amplitude center as zero, and the current signal after drift correction is redrawn. Finally, each corrected signal is decomposed into frequency bands, and the high-frequency, low-frequency and intermediate-frequency signals are separated by a bandpass filter. The signals of different frequency bands are marked separately to generate frequency band decomposition signals.

[0027] S203: based on the frequency band decomposition signal, the characteristic frequency band is screened and the transient signal is classified, and the frequency, amplitude and duration of each transient signal are extracted and classified into segments to establish a transient signal feature set; First, the time axis of the signal is divided according to the frequency band decomposition, and the start time, end time, main frequency and maximum amplitude of each signal are recorded to construct the initial feature list of the transient signal. Then, for each transient signal, by comparing the frequency fluctuation amplitude and amplitude change rate of the signal, the transient signal is classified into three categories: fast mutation type, slow change type and stable type. For the classification of each signal, the peak time point of the mutation signal, the time period boundary of the slow change signal and the stable interval range of the stable signal are marked in combination with the duration and frequency fluctuation range, and these annotation information are associated with the segmented signal characteristics and stored. Through this process, a transient signal feature set is established, which includes the time range, frequency distribution and classification label of the signal.

[0028] See also Figure 4 ,Based on the transient signal feature set, the current feature nodes of the wind turbines in the wind farm are constructed.,By comparing the frequency band features between the current feature nodes,,the dependency strength between the current feature nodes is determined, and the abnormal current feature nodes and their,influence range are identified.,The specific steps for generating the abnormal propagation path set are as follows: S301: Based on the transient signal feature set, analyze the frequency band characteristics of the transient signal, including frequency range, signal strength, duration and change mode, extract characteristic indicators of the transient signal, and generate a transient signal frequency band indicator set; First, the signal after the decomposition of the three-phase current of the wind turbine is divided into multiple segments according to the time period. The main frequency and signal strength peak value are extracted for each signal segment, and the timestamp and amplitude corresponding to the peak value are recorded to mark the characteristic points in the signal segment; secondly, by observing the waveform of each signal segment, the continuous signal with an amplitude exceeding the set threshold is analyzed as a segment, and by comparing the amplitude changes of adjacent data points, the start time of the segment is determined as the time point when the amplitude first exceeds the threshold, and the end time is the time point when the amplitude last exceeds the threshold, and the start and end range of each segment is recorded; finally, the change pattern of the signal is analyzed, and by comparing the frequency fluctuation amplitude and amplitude change rate within the segment, the change trend is marked as three types: stable, rising or falling, and the specific change characteristics of the frequency band are recorded. By sorting and summarizing the above content, a transient signal frequency band indicator set is generated to provide basic data for subsequent node feature analysis.

[0029] S302: Based on the transient signal frequency band indicator set and in combination with the distribution of transient signal characteristics among nodes, the current characteristic nodes of wind turbines in the wind farm are identified and classified, the frequency band range and signal characteristics of the nodes are defined, and a current characteristic node distribution map is generated; Combined with the signal data collected by each node in the wind farm, the frequency band characteristics are associated with the node distribution, the frequency band range and signal strength characteristics of the node are extracted, and the nodes are classified into three categories: stable, fluctuating and sudden changes according to their distribution rules. The specific judgment process is as follows: Judging stable nodes: Draw a trend chart of the main frequency band range and signal strength of the node in time series. If the fluctuation amplitude of the main frequency range of the node is less than the set threshold (for example, 5%) during the entire observation time, and the difference between the peak and valley values ​​of the signal strength does not exceed 10% of its mean, then mark the node as stable, and record its main frequency range and strength range. Judging fluctuating nodes: For nodes whose main frequency range fluctuation amplitude exceeds the stable threshold but is less than the larger sudden change threshold (for example, between 5% and 20%), by comparing the signal strength change trend, if the peak-to-valley difference of the intensity fluctuates greatly compared with the mean, but there is no drastic jump or sudden drop, then mark it as a fluctuating node, and record the frequency band range and change trend during the fluctuation period. Determine mutation nodes: If the fluctuation range of the node's main frequency range is greater than the set mutation threshold (for example, more than 20%), and the signal strength changes dramatically in a short period of time (for example, the peak value increases by more than 50% of the mean or the valley value decreases by more than 50% within a few seconds), it is marked as a mutation node, and the mutation period and the main frequency band range and intensity characteristics of the mutation are marked. Based on the above classification standards, draw a node characteristic distribution map, mark each node's classification label, frequency band range, main change characteristics and distribution position in the wind farm operation, generate a current characteristic node distribution map, and provide a clear data basis for subsequent node dependency strength analysis and propagation path identification.

[0030] S303: Based on the current characteristic node distribution diagram, calculate the dependency strength of the characteristic frequency bands between the nodes, analyze the influence range and propagation path between the nodes, identify the propagation relationship and range of abnormal nodes, and generate an abnormal propagation path set; The dependence strength of the characteristic frequency bands between nodes is calculated using the formula: ; Get Node and nodes The strength of the dependence ; in, The smaller the range, the closer the signal characteristics of the two nodes are. Is a node At the point in time The frequency band signal strength at the time is obtained by extracting the signal amplitude in the characteristic frequency band at each time point after frequency band decomposition. The amplitude information is directly obtained from the data record. Is a node At the point in time The frequency band signal strength at that time can be directly obtained by extracting the frequency band characteristics. Is a node and nodes At the point in time The absolute value of the signal strength difference is used to eliminate the influence of positive and negative signs and ensure that the dependent strength is non-negative. It is the total number of time points of the signal, which means the number of sampling points in the time series, which is determined by the time resolution and the total time length, for example, by experimentally recording the total running time and sampling frequency.

[0031] For example, the node and The signal frequency band distribution is recorded as the sampling data of the three-phase current, the sampling time interval is 1 millisecond, and the total sampling time is 10 seconds. time point. Signal amplitude and Extracted from frequency band characteristics, for example, in the high frequency range of 100Hz-200Hz, the peak amplitude value at each time point is taken as the signal strength of the node.

[0032] Assume Node The signal strength sequence is ,node The signal strength sequence is , total number of time points . Substituting these data into the formula: ; The results show that the dependence of the calculation , based on the calculation results of the dependency strength between nodes, analyze the influence range and propagation path between nodes. First, calculate the dependency strength of all nodes According to the node Construct a dependency matrix, where each element represents the strength value between corresponding nodes. By analyzing the dependency matrix, identify node pairs with low strength values, which indicate high correlation between nodes, and mark these node pairs as key connection points of the preliminary propagation path. Further analyze the propagation path, combined with the dependency strength threshold, for example, Node pairs with a value less than a set value (such as 2.0) are considered as connection points with direct propagation relationships, and a propagation network diagram is drawn using graph theory. For multiple paths formed in the network, the importance and priority of the propagation path are determined by recording the number of connected nodes and the total dependency strength of each path. Finally, combined with the marking of abnormal propagation nodes, the propagation path is associated with the time axis, the timing of signal propagation is analyzed, and the position of abnormal nodes on the propagation path and the scope of influence on surrounding nodes are identified. For example, when a node When an abnormal signal occurs, the adjacent nodes on its propagation path When the dependence intensity of the signal drops significantly, these nodes can be marked as affected areas, and the impact of abnormal signals on the entire propagation path can be further analyzed. Through the above process, an abnormal propagation path set is finally generated, which contains abnormal nodes, propagation paths, affected node ranges and propagation timing information, providing data support for the rapid location of abnormal sources in wind farms and the assessment of the impact range.

[0033] See also Figure 5 , based on the transient signal feature set, a sliding window is used to extract local features of the transient signal feature set, mark the abnormal fluctuation interval and the fault source of the wind turbine current data, and generate a local fluctuation abnormal feature set; the specific steps are as follows: S401: Based on the transient signal feature set, the signal data is segmented by a sliding window, the change rate and trend of the first-order difference and the second-order difference are obtained region by region, the differential dynamic features of each section are extracted, and a sliding window differential dynamic set is generated; The signal data is segmented by sliding windows, and the normalized signal of the fan current is processed by MATLAB. The fixed window size is set to segment the complete time series signal into multiple continuous segments. Extracting the key change characteristics of the signal includes the following steps: Extracting the first-order difference: In each sliding window, the difference in signal amplitude between adjacent time points is calculated point by point, and the difference of each pair of data points is recorded to generate the first-order difference sequence of the segment, which is used to describe the instantaneous change rate of the signal. For example, if a window contains the signal sequence {2,4,6,5}, its first-order difference is {2,2,-1}, indicating the increase or decrease trend of the signal. Extracting the second-order difference: Based on the first-order difference sequence, the change difference of the first-order difference value is continued to be calculated point by point in the same window, and the change rate of the second-order difference is recorded to describe the acceleration characteristics of the signal change. For example, the second-order difference of the above first-order difference sequence {2,2,-1} is {0,-3}, indicating the process of the signal turning from stable change to significant fluctuation. Record dynamic characteristics: By summarizing the range of changes in the first-order and second-order differences in each window, extract the dynamic change characteristics of the section, including the minimum, maximum and average change trend of the difference value, and generate a sliding window difference dynamic set for subsequent local fluctuation analysis. Through the above processing steps, the change rate and trend characteristics of the signal are extracted area by area, and the dynamic characteristics of each sliding window are annotated in combination with the characteristic change pattern.

[0034] S402: Based on the sliding window difference dynamic set, analyze the local characteristics in the segment, obtain the local variance and identify the segment with significant fluctuation, locate the key characteristics with abnormal behavior through the analysis of the change pattern, and generate the local fluctuation dynamic characteristic set; Analyze the local characteristics in each segment and locate the key characteristics with abnormal behavior. The process is as follows: Extract local variance: Calculate the average of the square of the difference of each difference sequence (including first-order difference and second-order difference) in each sliding window point by point as the local variance, which is used to evaluate the fluctuation amplitude of the signal. For example, if the first-order difference of a sliding window is {2,4,1,3}, then the local variance is . Identify significant fluctuation sections: Compare the local variance with the benchmark fluctuation range during normal operation of the wind turbine. When the local variance of a window exceeds the set threshold (for example, 1.5 times the normal benchmark fluctuation value), mark the window as a significant fluctuation section. Combined with the differential trend, observe whether the signal in this section shows drastic fluctuations, such as a continuous upward or downward trend, and record the time range of the fluctuation interval. Locate key characteristics: In the significant fluctuation section, determine the key abnormal characteristics by analyzing the change pattern of the differential trend. For example, if the second-order difference of the fluctuation section shows continuous negative values, it may point to the behavior of rapid signal attenuation; if the maximum value of the first-order difference far exceeds the mean, and the corresponding peak duration is short, it may be marked as a sudden abnormal point. Combined with the fan operation log and sampling period information, associate the key characteristics with the fan operation status (such as sudden wind speed changes or load fluctuations), and record the time period, change amplitude and main characteristic range of the abnormal behavior.

[0035] S403: Based on the local fluctuation dynamic characteristic set and in combination with the marked significant fluctuation area, the abnormal fluctuation interval is classified and the fault source is located, key abnormal points and fault characteristics in the wind turbine current data are marked, and a local fluctuation abnormal feature set is generated; The abnormal fluctuation intervals are classified and the fault sources are located, and the key abnormal points and fault characteristics in the wind turbine current data are marked. The specific process is as follows: Classify the abnormal fluctuation intervals: For the marked significant fluctuation areas, the abnormal intervals are classified into sudden abnormal intervals and continuous abnormal intervals according to the change range of the local variance and the differential dynamic characteristics. The sudden abnormal interval is usually manifested as a sharp change in the signal in a short period of time, such as a sharp increase in the local variance and a second-order difference value peak value far exceeding the baseline level; the continuous abnormal interval is manifested as a signal fluctuation exceeding the normal range for a long time, such as the first-order difference value remaining deviated from zero for a long time. Associate the wind turbine operation status: The classified abnormal fluctuation intervals are associated with the operation log of the wind turbine, and the operation status of the wind turbine during the operation period (such as speed, load change, wind speed fluctuation, etc.) is compared in combination with the time range of the abnormal interval. For example, if the abnormal interval coincides with the period of sudden increase in wind speed, it may point to the current abnormality caused by load overload; if it matches the period of rapid drop in speed, it may point to the signal abnormality caused by mechanical failure. Mark the key abnormal points: In each classified abnormal interval, the key characteristic points of the signal are further extracted. For example, record the peak amplitude and corresponding time point of the sudden abnormal interval, and mark it as a key abnormal point; in the continuous abnormal interval, mark the starting point and end point of the fluctuation, and record the signal fluctuation amplitude range and change trend in the interval. Through the above marking, a characteristic description of each abnormal point is formed, including the type of abnormal point, time range, amplitude change and possible fault source. Generate a fault feature set: Integrate the marking information of all abnormal intervals and key abnormal points to form a fault feature set of wind turbine current data. The feature set contains the classification results of the abnormal interval, the detailed characteristics of the key abnormal points, and the corresponding fault type and time range.

[0036] See also Figure 6 ,Based on the abnormal propagation path set and the local fluctuation abnormal feature set, the changes of wind turbine current data in the wind turbine current phase space trajectory are analyzed, the dynamic change trend of wind turbine current data at future moments is predicted, and the specific steps of generating wind turbine current abnormal warning results are as follows: S501: Based on the abnormal propagation path set and the local fluctuation abnormal feature set, the abnormal nodes and fluctuation intervals are mapped to the wind turbine current phase space trajectory, the distribution characteristics and dynamic change trends of the abnormal points in the trajectory are analyzed, and the trajectory abnormal dynamic characteristic set is generated; The abnormal nodes and fluctuation intervals are mapped to the phase space trajectory of the wind turbine current. The specific process is as follows: Determine the mapping basis: Extract the time range and propagation path of each abnormal node from the abnormal propagation path set, and combine the significant fluctuation intervals marked in the local fluctuation abnormal feature set to form a node-time interval association table. This table is used to identify the abnormal behavior of a specific node within a specific time range and provide input for subsequent trajectory mapping. Map to the phase space trajectory: According to the above time interval, the abnormal behavior of the corresponding node is projected to the phase space trajectory. The specific operation is to regenerate the phase space trajectory by analyzing the three-phase current data of the node within the abnormal time range and combining the phase space reconstruction method (such as time delay embedding). The abnormal points in the trajectory are determined by the characteristic changes of the signal within the time interval, such as the sudden offset of the trajectory point or the abnormal stretching of the trajectory shape. Analyze the distribution characteristics of the abnormal points: In the generated phase space trajectory, analyze the spatial distribution characteristics of the abnormal points in the trajectory point by point. Count the density of the abnormal points, for example, by calculating the local density of the abnormal points, marking the high-density area as an abnormal point group, and identifying the isolated abnormal points. In addition, the propagation mode of abnormal behavior can be determined by the distribution form of abnormal points in the trajectory (such as linear arrangement or scattered distribution). Analyze dynamic change trends: Analyze the dynamic change trend of the trajectory in combination with the time attributes of the abnormal points. By comparing the trajectory forms of different time periods, observe the trajectory deviation direction, speed, and overall trajectory changes of the abnormal points (such as from a smooth curve to an irregular distribution). Record the amplitude and speed of the deviation of the abnormal points in the trajectory to mark the dynamic characteristics of the abnormal behavior.

[0037] S502: Based on the trajectory abnormality dynamic characteristic set and the time series characteristics of the wind turbine current data, the current data value at a future time point is dynamically predicted, the difference between the predicted value and the actual observed value is determined, and a current data prediction deviation set is generated; The dynamic change trend of the current data value at a future time point is predicted using the formula: ; Calculate future time points Predicted current data value ; in, Indicates the current time point The actual current value is obtained from the current signal sequence collected by the wind turbine in real time. The data is obtained and recorded in the data set through the sensor. Each data point is the current amplitude information of a single time step. Indicates current data value The weight of the predicted value is determined by the distribution characteristics of the abnormal points in the trajectory abnormality dynamic characteristic set. The weight is obtained by normalizing the influence of each point trajectory shape on the future trajectory. For example, when the trajectory shape changes faster, the weight is appropriately increased, indicating that the point has a significant impact on the prediction result. Indicates the prediction start time, which is determined by the real-time requirements of the data and is the benchmark time point for current series sampling. It is the current moment The index of the next time step, from 1 to Represents the position index of each data point in the time window, It is the time step between the future moment and the current moment, which is usually determined by the needs of the forecasting scenario (such as short-term forecast or medium-term forecast).

[0038] Assume that the time series data is , the weight is , time window length , the target prediction time point is .

[0039] Substitute into the formula: ; If the actual observed value is 6.0, the prediction deviation is , predicted value It indicates that the current value in the next 2 seconds will have a slight upward trend, which is consistent with the changing trend in the current time period.

[0040] By comparing the predicted values ​​and actual values ​​at all time points, the deviation is calculated and the difference characteristics at each moment are recorded, including the size of the deviation, the positive and negative direction, and the change trend. The deviation value is correlated with the abnormal pattern for analysis. For example, a large deviation may point to a data mutation or the instability of the prediction model. The abnormal deviation points at specific moments are marked and classified into a current data prediction deviation set for subsequent abnormal pattern recognition and early warning system optimization.

[0041] S503: Based on the current data prediction deviation set, classify the abnormal modes appearing in the deviation and analyze the impact range, mark the specific characteristics and occurrence conditions of the abnormal modes, and generate the abnormal current warning result of the wind turbine; First, the deviation values ​​in the predicted deviation set are classified according to the amplitude and change trend. For example, by setting a threshold, the data points with large deviation amplitudes are marked as sudden anomalies, and the data intervals with continuous deviations exceeding the threshold are marked as cumulative anomalies. Secondly, by analyzing the time distribution of the abnormal pattern and the rate of change of the deviation, its specific characteristics are extracted, such as the peak value of the sudden anomaly, the duration of the cumulative anomaly and the total deviation. These characteristics are then correlated and analyzed in combination with the wind turbine operation data. For example, it is analyzed whether the sudden anomaly occurs in the time period when the wind speed changes sharply or the load rises suddenly, so as to determine the occurrence conditions of the abnormal pattern. Furthermore, the abnormal pattern is cross-compared with the key operating parameters of the wind turbine to evaluate its potential impact on the operation of the wind turbine, including the system components that may be affected, the propagation path of the abnormal signal, and the degree of disturbance to the current load. Finally, all the analysis results are integrated, and the specific characteristics, occurrence conditions and impact range of each abnormal pattern are marked to generate a complete wind turbine current abnormality warning result to support real-time monitoring and early warning management of the wind turbine operation status.

[0042] See also Figure 7 The fan current detection system is used to execute the fan current detection method, and the system includes: The data processing module analyzes the time series characteristics of the collected wind turbine current data, extracts the time point distribution of the current data, the range of data amplitude variation, and the time dependence characteristics, reconstructs the characteristic set in phase space, gradually adjusts the embedding dimension and time delay parameters, and generates the wind turbine current phase space trajectory; The transient feature analysis module is based on the wind turbine current phase space trajectory, uses the signal mean trend to eliminate the influence of low-frequency errors, decomposes the signal into multiple characteristic frequency bands, extracts the frequency, amplitude and duration of each transient signal, performs classification and segmentation, and establishes a transient signal feature set; The abnormal propagation path analysis module analyzes the frequency band characteristics of transient signals one by one based on the transient signal feature set, extracts the frequency range, signal strength and change mode, and identifies the current feature nodes with abnormal frequency bands and their influence range by comparing the similarity of the feature frequency bands, thus generating an abnormal propagation path set. The abnormal warning module performs sliding window segmentation on the transient signal features based on the abnormal propagation path set and the transient signal feature set, analyzes the local fluctuation characteristics of each section, predicts the dynamic change trend of the wind turbine current data, and generates abnormal wind turbine current warning results.

[0043] A computer device comprises a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the above-mentioned wind turbine current detection system is implemented.

[0044] A computer-readable storage medium stores a computer program, which implements the steps of the above-mentioned wind turbine current detection method when executed by a processor.

[0045] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting fan current, characterized in that: The following steps are involved: Collect wind turbine current data, analyze the time series characteristics of wind turbine current data, monitor and compare the distance change trend between neighboring points, determine the optimal embedding dimension and time delay through phase space reconstruction method, and generate wind turbine current phase space trajectory; Based on the wind turbine current phase space trajectory, the wind turbine current data is decomposed into a plurality of characteristic frequency bands, the decomposed characteristic frequency bands of the wind turbine current data are screened, the transient characteristic signals in the wind turbine current data are identified, and a transient signal characteristic set is established; Based on the transient signal feature set, current feature nodes of wind turbines in the wind farm are constructed, the dependency strength between the current feature nodes is determined by comparing the frequency band features between the current feature nodes, abnormal current feature nodes and their influence range are identified, and an abnormal propagation path set is generated; Based on the transient signal feature set, a sliding window is used to extract local features of the transient signal feature set, abnormal fluctuation intervals and fault sources of wind turbine current data are marked, and a local fluctuation abnormal feature set is generated; Based on the abnormal propagation path set and the local fluctuation abnormal feature set, the change of the wind turbine current data in the wind turbine current phase space trajectory is analyzed, the dynamic change trend of the wind turbine current data at future moments is predicted, and the wind turbine current abnormality warning result is generated.

2. The wind turbine current detection method according to claim 1, characterized in that: The wind turbine current phase space trajectory includes an optimized time series embedding dimension, a determined time delay range, and a distribution pattern of trajectory points in the corresponding phase space. The transient signal feature set includes a screened characteristic frequency band set, a transient signal that has been amplitude normalized and baseline drift corrected, and a classification label and feature description of the transient signal. The abnormal propagation path set includes the characteristic frequency band distribution of abnormal nodes, a propagation path network diagram between abnormal nodes, and a quantitative result of the abnormal impact range. The local fluctuation anomaly feature set includes a marked significant fluctuation interval, the fluctuation amplitude statistics of the first-order difference and the second-order difference, and a set of significant change points identified based on local variance analysis. The wind turbine current abnormal warning result includes a dynamic change trend prediction diagram of future wind turbine current data, an abnormal pattern distribution of prediction deviations, and multi-dimensional abnormal warning information combining propagation paths and fluctuation characteristics.

3. The wind turbine current detection method according to claim 1, characterized in that: Collect wind turbine current data, analyze the time series characteristics of wind turbine current data, monitor and compare the distance change trend between neighboring points, determine the optimal embedding dimension and time delay through phase space reconstruction method, and generate the wind turbine current phase space trajectory. The specific steps are as follows: Based on the collected wind turbine current data, its time series characteristics are analyzed, and the time point distribution of the current data, the range of data amplitude variation, and the time dependence characteristics are extracted in sequence to generate a time series feature set; Based on the time series characteristic set, the time point distribution of the current data is reconstructed in phase space, the embedding dimension and time delay parameter are gradually adjusted, the trend of the change of the trajectory point neighbor distance is analyzed, and the optimal embedding parameter set is obtained; Based on the optimal embedding parameter set, the embedding dimension of the wind turbine current data is adjusted and time-delayed, a phase space trajectory diagram is constructed, and the trajectory point distribution pattern is analyzed to generate the wind turbine current phase space trajectory.

4. The wind turbine current detection method according to claim 1, characterized in that: Based on the wind turbine current phase space trajectory, the wind turbine current data is decomposed into multiple characteristic frequency bands, the decomposed characteristic frequency bands of the wind turbine current data are screened, and the transient characteristic signals in the wind turbine current data are identified. The specific steps of establishing the transient signal feature set are as follows: Based on the wind turbine current phase space trajectory, amplitude normalization processing is performed on the wind turbine current data, and the upper and lower limits of the amplitude data are used to rescale the overall signal range to eliminate the amplitude difference and obtain a normalized current signal; Based on the normalized current signal, a baseline drift correction operation is performed, the influence of low-frequency errors is eliminated through the signal mean trend, and the signal is decomposed into multiple characteristic frequency bands to obtain a frequency band decomposition signal; Based on the frequency band decomposition signal, the characteristic frequency bands are screened and the transient signals are classified, and the transient signal feature set is established by extracting the frequency, amplitude and duration of each transient signal and classifying them in sections.

5. The wind turbine current detection method according to claim 1, characterized in that: Based on the transient signal feature set, the current feature nodes of the wind turbines in the wind farm are constructed, the frequency band features between the current feature nodes are compared, the dependency strength between the current feature nodes is determined, the abnormal current feature nodes and their influence range are identified, and the specific steps of generating the abnormal propagation path set are as follows: Based on the transient signal feature set, the frequency band characteristics of the transient signal are analyzed, including the frequency range, signal strength, duration and change mode, characteristic indicators of the transient signal are extracted, and a transient signal frequency band indicator set is generated; Based on the transient signal frequency band indicator set and in combination with the distribution of transient signal characteristics among nodes, the current characteristic nodes of wind turbines in the wind farm are identified and classified, the frequency band range and signal characteristics of the nodes are defined, and a current characteristic node distribution map is generated; Based on the current characteristic node distribution diagram, the dependency strength of the characteristic frequency bands between the nodes is calculated, the influence range and propagation path between the nodes are analyzed, the propagation relationship and range of the abnormal nodes are identified, and an abnormal propagation path set is generated; The dependence strength of the characteristic frequency bands between nodes is calculated using the formula: ； Get Node and nodes The strength of the dependence ; in, Is a node At the point in time The signal strength of the frequency band at Is a node At the point in time The signal strength of the frequency band at is the total number of time points of the signal.

6. The wind turbine current detection method according to claim 1, characterized in that: Based on the transient signal feature set, a sliding window is used to extract local features of the transient signal feature set, abnormal fluctuation intervals and fault sources of wind turbine current data are marked, and a local fluctuation abnormal feature set is generated; the specific steps are as follows: Based on the transient signal feature set, the signal data is segmented using a sliding window, the change rate and trend of the first-order difference and the second-order difference are obtained region by region, the differential dynamic features of each section are extracted, and a sliding window differential dynamic set is generated; Based on the sliding window difference dynamic set, the local characteristics within the segment are analyzed to obtain the local variance and identify the segment with significant fluctuations, and the key characteristics with abnormal behavior are located by analyzing the change pattern to generate a local fluctuation dynamic characteristic set; Based on the local fluctuation dynamic characteristic set and in combination with the marked significant fluctuation area, the abnormal fluctuation intervals are classified and the fault source is located, the key abnormal points and fault characteristics in the wind turbine current data are marked, and the local fluctuation abnormal feature set is generated.

7. The wind turbine current detection method according to claim 1, characterized in that: Based on the abnormal propagation path set and the local fluctuation abnormal feature set, the specific steps of analyzing the change of wind turbine current data in the wind turbine current phase space trajectory, predicting the dynamic change trend of wind turbine current data at future moments, and generating wind turbine current abnormal warning results are as follows: Based on the abnormal propagation path set and the local fluctuation abnormal feature set, the abnormal nodes and fluctuation intervals are mapped to the wind turbine current phase space trajectory, the distribution characteristics and dynamic change trends of the abnormal points in the trajectory are analyzed, and the trajectory abnormal dynamic characteristic set is generated; Based on the trajectory abnormality dynamic characteristic set and combined with the time series characteristics of the wind turbine current data, the current data value at a future time point is dynamically predicted, the difference between the predicted value and the actual observed value is determined, and a current data prediction deviation set is generated; The dynamic change trend of the current data value at a future time point is predicted using the formula: ； Calculate future time points The predicted current data value ; in, Indicates the current time point The actual current value, Indicates current data value The weight of the predicted value, Indicates the prediction start time, It is the current moment The time step index after , is the time step between the future moment and the present moment; Based on the current data prediction deviation set, the abnormal patterns appearing in the deviation are classified and the impact range is analyzed, the specific characteristics and occurrence conditions of the abnormal patterns are marked, and the abnormal current warning results of the wind turbine are generated.

8. The fan current detection system is characterized by: The wind turbine current detection system is used to execute the wind turbine current detection method according to any one of claims 1 to 7, and the system comprises: The data processing module analyzes the time series characteristics of the collected wind turbine current data, extracts the time point distribution of the current data, the range of data amplitude variation, and the time dependence characteristics, reconstructs the characteristic set in phase space, gradually adjusts the embedding dimension and time delay parameters, and generates the wind turbine current phase space trajectory; The transient feature analysis module is based on the wind turbine current phase space trajectory, uses the signal mean trend to eliminate the influence of low-frequency errors, decomposes the signal into multiple characteristic frequency bands, extracts the frequency, amplitude and duration of each transient signal, performs classification marking and segmentation classification, and establishes a transient signal feature set; The abnormal propagation path analysis module analyzes the frequency band characteristics of the transient signal one by one based on the transient signal feature set, extracts the frequency range, signal strength and change mode, identifies the current feature nodes with abnormal frequency bands and their influence range through similarity comparison of feature frequency bands, and generates an abnormal propagation path set; The abnormal warning module performs sliding window segmentation on the transient signal features based on the abnormal propagation path set and the transient signal feature set, analyzes the local fluctuation characteristics of each section, predicts the dynamic change trend of the wind turbine current data, and generates abnormal wind turbine current warning results.

9. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor implements the wind turbine current detection system according to claim 8 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind turbine current detection method according to any one of claims 1 to 8 are implemented.

Images