Wind turbine generator time domain signal processing method and device and storage medium

Abstract

The invention relates to the technical field of blade state monitoring of wind power equipment, in particular to a wind turbine generator time domain signal processing method and device and a storage medium, and the method comprises the steps: obtaining originally collected time domain signals and corresponding sampling interval time; the method comprises the following steps: inputting an originally collected time domain signal into a first-order low-pass filter, setting the length of a sliding window of the first-order low-pass filter, obtaining an output corresponding to each sliding window through low-pass filtering processing, and taking the output as a mean value of the time domain signal in the corresponding sliding window. The method is low in resource occupation, high in calculation efficiency and easy to implement.

Description

Technical Field

[0001] This invention relates to the field of wind turbine control technology, specifically to a wind turbine time-domain signal processing method, device, and storage medium. Background Technology

[0002] Throughout the entire lifecycle of a wind turbine generator, condition monitoring, performance analysis, and control decisions all heavily rely on multi-timescale average data of key operating parameters. These core parameters include, but are not limited to, wind speed, output power, and main shaft speed. Generally, the required timescales cover 1 second to 10 minutes.

[0003] Currently, the mainstream solution in the industry for calculating the average value across multiple time scales is the moving average method. The technical principle of this method dictates that it requires a large data array to be allocated in advance to continuously store historical sampling data. During each mean update, a series of operations such as queue management (including enqueuing new data and removing expired old data), summing historical data, and recalculating the mean must be performed simultaneously, which is a cumbersome process with concentrated resource consumption. Summary of the Invention

[0004] The purpose of this invention is to provide a time-domain signal processing method, device, and storage medium for wind turbine generators, which has low resource consumption, high computational efficiency, and is easy to implement.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention discloses a time-domain signal processing method for wind turbine generators, comprising: Acquire the original time-domain signal and the corresponding sampling interval; The original acquired time-domain signal is input into a first-order low-pass filter. The sliding window length of the first-order low-pass filter is set, and the output corresponding to each sliding window is obtained through low-pass filtering. The output is used as the mean of the time-domain signal within the corresponding sliding window.

[0006] Furthermore, the discretization calculation formula for the low-pass filtering process is as follows: In the formula, ; T a The sampling interval time. T b The length of the sliding window. Y ( n () represents the output at the current sampling time. Y ( n－ 1) is the output of the previous sampling time. Y (0)=0, X ( n ) represents the input at the current sampling time, i.e., the original time-domain signal acquired.n This represents the number of sampling points within a single sliding window.

[0007] Furthermore, the number of first-order low-pass filters is two or more, the two or more first-order low-pass filters are arranged in parallel, and the sliding window lengths of the two or more first-order low-pass filters are different.

[0008] Furthermore, the sliding window lengths of two or more first-order low-pass filters are independently 1s to 600s.

[0009] In a second aspect, the present invention discloses a time-domain signal processing device for wind turbine generators, comprising: The acquisition module is used to acquire the raw time-domain signal and the corresponding sampling interval time; The processing module is used to input the original acquired time-domain signal into a first-order low-pass filter, set the sliding window length of the first-order low-pass filter, and obtain the output corresponding to each sliding window through low-pass filtering processing, and use the output as the mean of the time-domain signal within the corresponding sliding window.

[0010] Thirdly, the present invention discloses a storage medium storing a computer-readable program, which, when invoked, can execute the steps of the above-described wind turbine time-domain signal processing method.

[0011] The present invention has the following unexpected beneficial effects: The present invention uses a first-order low-pass filter to process time-domain signals and achieves average output through recursive calculation logic, so that the memory usage is independent of the sliding window length. Especially in multi-sliding window scenarios, there is no need to allocate an array for each sliding window, which greatly reduces memory consumption and adapts to the hardware constraints of limited memory resources of wind turbine generator PLC. Attached Figure Description

[0012] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention.

[0013] Figure 1 A flowchart illustrating the wind turbine time-domain signal processing method provided in an embodiment of this application is shown.

[0014] Figure 2 A schematic diagram showing a comparison of the mean trends obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application is presented when the sliding window length is 1s.

[0015] Figure 3 The diagram shows a partially enlarged schematic of the mean trend obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application when the sliding window length is 1s.

[0016] Figure 4 A schematic diagram showing a comparison of the mean trends obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application is presented when the sliding window length is 10s.

[0017] Figure 5 A schematic diagram showing a comparison of the mean trends obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application is presented when the sliding window length is 30s.

[0018] Figure 6 A schematic diagram showing a comparison of the mean trends obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application is presented when the sliding window length is 60s.

[0019] Figure 7 A schematic diagram showing a comparison of the mean trends obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application is presented when the sliding window length is 180s.

[0020] Figure 8 A schematic diagram showing a comparison of the mean trends obtained by the moving average method and the wind turbine time-domain signal processing method described in the embodiments of this application is presented when the sliding window length is 600s. Detailed Implementation

[0021] For the calculation of the mean of the time domain signal of wind turbine, the existing moving average method requires an independent large array to store historical data for each time scale (sliding window length). As the sliding window length increases, the array size increases linearly, resulting in excessive PLC memory usage.

[0022] Based on this, in one embodiment, see Figure 1 As shown, the present invention provides a time-domain signal processing method for wind turbine generators, comprising: Acquire the original time-domain signal and the corresponding sampling interval; The original acquired time-domain signal is input into a first-order low-pass filter. The sliding window length of the first-order low-pass filter is set, and the output corresponding to each sliding window is obtained through low-pass filtering. The output is used as the mean of the time-domain signal within the corresponding sliding window.

[0023] This invention uses a first-order low-pass filter to process time-domain signals and achieves average output through recursive calculation logic, making memory usage independent of the sliding window length. Especially in multi-sliding window scenarios, it eliminates the need to allocate an array for each sliding window, significantly reducing memory consumption and adapting to the hardware constraints of limited memory resources in wind turbine generator PLCs.

[0024] In a preferred embodiment of the present invention, the discretization calculation formula for the low-pass filtering process is as follows: In the formula, ; T a The sampling interval time. T b The length of the sliding window. Y ( n () represents the output at the current sampling time. Y ( n－ 1) is the output of the previous sampling time. Y (0)=0, X ( n ) represents the input at the current sampling time, i.e., the original time-domain signal acquired. n This represents the number of sampling points within a single sliding window.

[0025] The moving average method relies on a sequential process of removing old values, enqueuing new values, summing, and dividing. As the window length increases, the computational complexity of the summation operation increases linearly, and the computational complexity accumulates when multiple windows are parallel, easily causing delays in PLC task cycles and affecting the real-time performance of control commands. This preferred implementation achieves efficient calculation through simplified operational logic using a first-order low-pass filter. The mean output of each window only needs to be based on the output of the previous sampling time. Y ( n－ 1) Input at the current sampling time X ( n The single sliding window method performs single multiplication and addition operations on the filter parameters, which significantly reduces the computation time compared to the moving average method, thus improving computational efficiency.

[0026] Specifically, within a certain sliding window duration, the original time-domain signal acquired... I n It is a discrete sampled signal, generated by the wind turbine's sensors at sampling intervals. T a One data collection, i.e.: I n =[ X (0), X (1), X (2), ... X ( k ... X ( n ], where represents the first k The original signal value for each sampling period, k =0,1,2… …n .

[0027] At the first sampling point, , Y (0) is the initial value, set to 0.

[0028] At the second sampling point, .

[0029] And so on, in the... n One sampling point, ,Right now Y ( n This effectively includes information from all sampling points within the sliding window length, which serves as the mean of the time-domain signal within the corresponding sliding window, providing a data foundation for subsequent signal analysis.

[0030] The derivation of the calculation formula includes: the transfer function of the first-order low-pass filter. G ( s )= = , Y ( z The output is the current time. X ( z () is the current time input. T The time constant of the filter, S For the Laplace operator.

[0031] The wind turbine generator control system is a digital system, with each sampling interval... T a Only one signal acquisition is performed, using the backward differential method, let To approximate the differential operators in continuous models S , z －1 It is a delay operator in a discrete system, representing the previous value.

[0032] Will Substituting the transfer function and simplifying, we obtain the discrete transfer function. G ( z )= = .

[0033] Simplify the denominator: .

[0034] To simplify the form, filter parameters are introduced. ,Right now After substituting, the denominator becomes: .

[0035] Therefore, discrete transfer function G ( z )= = .

[0036] Cross-multiplication of the discrete transfer function yields: .in, Output of the previous sampling time in the corresponding time domain Y ( n－1), Y ( z The output corresponding to the current sampling time Y ( n ), X ( z The input corresponding to the current sampling time X ( n ).

[0037] Dividing both sides by the same factor, the final simplified result is: .

[0038] When the system reaches steady state, the output of this recursive formula converges to the average value of the input signal. Its equivalent average time window is the sliding window length and the time constant. T Closely related. In engineering, the settling time (i.e., the time considered to reach steady state) of a first-order system is approximately 4T, with an error of less than 2%. For sampled systems, the time constant... T With filter parameters and sampling interval time T a The relationship between (i.e., the data refresh time of the control system) and is as follows: T = (-1) T a .

[0039] To make the response time of the first-order low-pass filter equal to the set sliding window length T b Let 4T = T b Substituting into the equation yields... .

[0040] In a preferred embodiment of the present invention, the number of first-order low-pass filters is two or more, the two or more first-order low-pass filters are arranged in parallel, and the sliding window lengths of the two or more first-order low-pass filters are different.

[0041] When applications require simultaneous calculation of averages over multiple time scales, such as obtaining averages for 1 second, 30 seconds, and 10 minutes, the moving average method necessitates instantiating a separate functional block for each time scale. Each functional block requires independent configuration of its own historical data array and computational logic, leading to dual resource pressures on the core PLC (Programmable Logic Controller) device in the wind turbine generator control system: on the one hand, multiple large arrays continuously occupy limited memory space; on the other hand, the parallel queue management and accumulation operations of the multi-functional blocks significantly consume CPU computing resources. This resource consumption problem not only directly restricts the overall performance of the PLC system but also significantly increases code complexity due to the need to write and maintain multiple sets of functional block logic, causing inconvenience for subsequent system debugging and upgrades.

[0042] This preferred implementation uses a unified architecture based on a first-order low-pass filter to achieve multi-window processing. By simply adjusting the filter parameters corresponding to the window length, the same set of filtering logic can be reused to calculate the mean at different time scales without modifying the core code structure. This design significantly simplifies program writing complexity and offers flexible parameter adjustment. The filter parameters corresponding to different window lengths can be quickly derived using formulas, facilitating on-site debugging and subsequent system upgrades, thus lowering the technical threshold for engineering implementation. Furthermore, parallel multi-window processing does not significantly increase the computational load, ensuring synchronous output of the mean at all time scales within the millisecond-level sampling interval required for wind power control, meeting the real-time requirements of scenarios such as condition monitoring and pitch control.

[0043] In a preferred embodiment of the present invention, the sliding window lengths of two or more first-order low-pass filters are independently 1s to 600s.

[0044] The wind turbine time-domain signal processing method of the present invention will be described below with reference to specific examples.

[0045] Taking a wind turbine PLC control system as an example, the original time-domain signal acquired is power, and the sampling interval is... T a The time-domain signal was 20ms. The raw acquired time-domain signal was input into six parallel first-order low-pass filters, and the sliding window lengths of the six first-order low-pass filters were set. T b The values ​​are 1s, 10s, 30s, 60s, 180s, and 600s, respectively. According to the formula... The filtering parameters of each first-order low-pass filter are calculated.

[0046] Specifically, the length of the sliding window T b When set to 1 second, =13.5; Sliding window length T b When set to 10 seconds, =126; Sliding window length T b When set to 30 seconds, =376; Sliding window length T b When set to 60 seconds, =751; Sliding window length T b When set to 180s, =2251; Sliding window length T bWhen set to 600s, =7501.

[0047] In each PLC sampling cycle, the raw power signal is simultaneously fed into the above six first-order low-pass filters, which immediately yields the average values ​​of six different time scales for use by subsequent control logic.

[0048] The average value obtained using the moving average method is used as a comparative example; see [link / reference]. Figures 2 to 8 The horizontal axis represents the sequence number, and the vertical axis represents the power value. The average values ​​obtained by the two algorithms within 60 seconds are not significantly different, especially the 1-second average. However, the moving average requires an array of 50 data points (refreshed at a 20ms refresh rate, 50 points refreshed per second), while the embodiment of this application only requires 2 variables. For averages longer than 60 seconds, the moving average and the low-pass filter described in this application embodiment do not overlap significantly, but both can show data trends. In fact, at the corresponding time points, the low-pass filter in this application embodiment is more pronounced. For example: see... Figure 8 When the sliding window length is 600s, 4×10 4 Around a certain time, the mean data in this embodiment of the application shows a significant rise followed by a fall, which is not reflected in the moving average curve. As the time period of the data mean increases, the length of the array required for the moving average increases significantly.

[0049] In summary, compared to the moving average method, this application completely eliminates the need for large array storage and historical data management, minimizing memory usage. It only requires storing the previous output value, and the computational load is limited to two multiplication and addition operations, significantly freeing up PLC system resources. Furthermore, it enables efficient parallel computing, eliminating waiting time and simultaneously outputting the average value across all time scales in each scan cycle, resulting in excellent real-time performance. Simultaneously, it is simple to implement, highly reliable, and consists of multiple identical and simple functional blocks connected in parallel, resulting in low code redundancy, ease of maintenance and debugging, and higher system stability. The analysis window for the average value can be easily changed by adjusting the filter constant, adapting to different application scenarios.

[0050] In one embodiment, this application also discloses a wind turbine time-domain signal processing device, which includes: The acquisition module is used to acquire the raw time-domain signal and the corresponding sampling interval time; The processing module is used to input the original acquired time-domain signal into a first-order low-pass filter, set the sliding window length of the first-order low-pass filter, and obtain the output corresponding to each sliding window through low-pass filtering processing, and use the output as the mean of the time-domain signal within the corresponding sliding window.

[0051] In one embodiment, this application discloses a storage medium storing a computer-readable program, which, when invoked, can execute the steps of the wind turbine time-domain signal processing method described above.

[0052] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments, and various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention. Furthermore, the embodiments of the present invention and the features thereof can be combined with each other unless otherwise specified.

Claims

1. A time-domain signal processing method for wind turbine generators, characterized in that, include: Acquire the original time-domain signal and the corresponding sampling interval; The original acquired time-domain signal is input into a first-order low-pass filter. The sliding window length of the first-order low-pass filter is set, and the output corresponding to each sliding window is obtained through low-pass filtering. The output is used as the mean of the time-domain signal within the corresponding sliding window.

2. The wind turbine time-domain signal processing method according to claim 1, characterized in that: The discretization formula for the low-pass filtering process is as follows: In the formula, ; T a The sampling interval time. T b The length of the sliding window. Y ( n () represents the output at the current sampling time. Y (0)=0, Y ( n－ 1) is the output of the previous sampling time. X ( n ) represents the input at the current sampling time, i.e., the original time-domain signal acquired. n This represents the number of sampling points within a single sliding window.

3. The wind turbine time-domain signal processing method according to claim 1, characterized in that: The number of first-order low-pass filters is two or more, and the two or more first-order low-pass filters are arranged in parallel, and the sliding window lengths of the two or more first-order low-pass filters are different.

4. The wind turbine time-domain signal processing method according to claim 3, characterized in that: The sliding window lengths of two or more first-order low-pass filters are independently 1s to 600s.

5. A time-domain signal processing device for wind turbine generators, characterized in that, include: The acquisition module is used to acquire the raw time-domain signal and the corresponding sampling interval time; The processing module is used to input the original acquired time-domain signal into a first-order low-pass filter, set the sliding window length of the first-order low-pass filter, and obtain the output corresponding to each sliding window through low-pass filtering processing, and use the output as the mean of the time-domain signal within the corresponding sliding window.

6. A storage medium, characterized in that: It contains a computer-readable program that, when invoked, can perform the steps of the wind turbine time-domain signal processing method as described in any one of claims 1 to 4.

Images