A high-speed-rail-line pseudo-periodic-maximum wind speed prediction method and system

Abstract

The application discloses a high-speed rail line pseudo-periodic maximum wind speed prediction method and system, which comprises the following steps: obtaining a high-speed rail historical second-level wind speed sequence, selecting a first local extreme point, and setting an initial period window width; extracting the extreme point in the high-speed rail historical second-level wind speed sequence to obtain a maximum point sequence; adjusting the initial period window width in real time according to the historical second-level wind speed to obtain an online maximum point wind speed sequence; and using a bidirectional long short memory neural network to perform rolling prediction on the future maximum point position and value. The application provides support for a high-speed rail line wind speed early warning system and enhances the predictability of wind speed.

Description

Technical Field

[0001] This invention relates to the field of wind speed prediction along high-speed railways, specifically to a method and system for predicting pseudo-periodic maximum wind speeds along high-speed railways. Background Technology

[0002] High-speed trains are extremely sensitive to sudden strong winds. If the instantaneous wind speed reaches the critical overturning speed, it can cause the train to derail in just 1 to 2 seconds. At present, the high-speed train wind warning system still uses monitoring and alarm technology, which does not realize prediction and early warning. This poses a serious hidden danger of insufficient deceleration window, affecting safe operation. Therefore, it is urgent to develop wind speed prediction methods to realize high-speed train wind prediction and early warning.

[0003] Existing wind speed forecasting methods primarily originate from the meteorology and wind power fields. In meteorology, the International Meteorological Organization recommends using 10-minute average wind speeds for forecasting, while in wind power, average wind speeds over 15 minutes or longer are typically predicted. However, for high-speed rail, a 1-2 second gale is enough to overturn a train. The second-level wind speed fluctuations along high-speed rail lines are extremely volatile and random. Therefore, monitoring extreme second-level wind speeds along high-speed rail lines is crucial, and existing wind speed forecasting methods from meteorology and wind power are ineffective for this purpose. Furthermore, existing wind speed forecasting methods pre-fix the time interval before forecasting, using this fixed interval for single or multi-step predictions. This not only fails to capture the periodicity of wind speed fluctuations but also smooths out extreme values. For high-speed rail, focusing on instantaneous gales, requiring methods that accurately predict winds exceeding the alarm threshold for 10 consecutive seconds is too difficult and overly demanding.

[0004] Therefore, how to utilize the pseudo-periodicity of the second-level wind speed of high-speed rail to model and predict the extreme wind speed of high-speed rail is an important problem that needs to be solved. Summary of the Invention

[0005] The technical problem this invention aims to solve is to provide a method and system for predicting pseudo-periodic maximum wind speeds along high-speed railway lines. This method approaches the problem from the perspective of local maximum wind speeds, determines whether a value is a major extreme based on distance, and uses extreme points to form an envelope, combined with alarm rules, to achieve prediction and early warning. This not only improves the accuracy of early warnings but also simplifies model construction.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] This invention proposes a method for predicting pseudo-periodic maximum wind speeds along high-speed railway lines, comprising:

[0008] S1. Obtain a historical second-level wind speed sequence Ts for high-speed rail, where Ts: x = x1, x2, ..., x n x nLet e1 represent the wind speed at the nth second. Select the first point that is greater than both the previous and the next wind speed as the first local extreme point. Set the initial period window width T0 based on experience.

[0009] S2. In order to obtain the non-fixed period extreme wind speed that conforms to the second-level wind speed fluctuation characteristics, local extreme points are extracted from the historical second-level wind speed sequence of high-speed rail. By judging the distance at the local extreme points, the maximum value sequence is obtained. This sequence includes the maximum value position sequence and the maximum value value sequence.

[0010] S3. Due to the pseudo-periodicity of wind speed time series, wind speed fluctuations occur within a certain period, and fixed-period searches may result in the absence of extreme points. Therefore, based on the maximum point sequence, the initial period window width T0 is adjusted in real time according to historical second-level wind speeds to obtain an online maximum point wind speed sequence, providing training and testing data for maximum point prediction.

[0011] S4. Divide the online maximum wind speed sequence into training and test sets according to a set ratio, and use a bidirectional long short-term memory neural network (BiLSTM) to make rolling predictions of the location and value of future maximum points.

[0012] Furthermore, in step S2, obtaining the sequence of maximum points includes the following sub-steps:

[0013] S201. Take the first local extremum point e1 as the first extremum point, and search for all local extremum points within the T0 period window width starting from e1, for a total of m points.

[0014] S202. Connect e1 with the p-th local extremum point using a dashed line. Calculate the sum of distances from all second-level wind speeds between these two local extremum points to this dashed line. The formula for calculating the sum of distances at the local extremum point p is as follows:

[0015]

[0016] Where, x j This represents the wind speed in the j-th second interval between the dashed lines, where J represents the total number of wind speeds in the second interval between the dashed lines, and Line represents the dashed line. p Ev represents the sum of distances at the local extremum point p. p Let ev represent the value of the p-th local extremum. p-1 This represents the value of the (p-1)th local extremum point; Used to avoid local spurious extrema.

[0017] Therefore, the distance and vector obtained at all local extrema within the T0 period window width is d, where d = {d1, ... d2}. p ,...d m}

[0018] S203. Sort each element in the distance and vector, locate the position with the minimum distance and set it as the second extreme point e2.

[0019] S204. Starting from the second extreme point e2, search for all local extreme points within a T0-period window. If there are no extreme points within the T0-period window, extend the search to a T0 / 2-period window and repeat steps S202-S203 to obtain the third extreme point e3. Starting from e3, repeat the search process until the extreme points at the end of the sequence Ts are searched. False extreme points are filtered out to obtain the extreme point sequence e, e = e1, e2, ..., e N e N This represents the Nth extreme point.

[0020] S205. The extreme point sequence contains maxima and minima. Extract the maxima sequence E from the extreme point sequence. The maxima sequence includes the maxima position sequence EI and the maxima value sequence EV, where EI = EI1, EI2, ..., EI l EV = EV1, EV2, ..., EV l EI l Indicates the position of the l-th maximum point, EV l This represents the value of the l-th maximum point; the sequence of maximum points is represented as E = (EI, EV) = (EI1, EV1), (EI2, EV2), ..., (EI... l EV l ).

[0021] Furthermore, in step S3, obtaining the online maximum wind speed sequence includes the following sub-steps:

[0022] S301. Using the l2-constrained least squares method of the Gaussian kernel model, the sequence of maximum point positions in step S2 is fitted and predicted to obtain the next maximum point E′. l+1 Position EI′ l+1 .

[0023] S302. A pseudo-period is defined by setting the time interval between two maxima. For pseudo-periods with wind speeds on the order of seconds, their variability needs to be quantified. The slope variability is obtained using historical statistics of the maxima slope, specifically using the following formula:

[0024]

[0025] Where l represents the length of the maximum sequence, C K The slope variation refers to the rate of change of the slope of the maximum point sequence E; k i This represents the slope at the i-th maximum point. EV iEI represents the value of the i-th maximum point. i EV represents the position of the i-th maximum point. i-1 EI represents the value of the (i-1)th maximum point. i-1 This indicates the position of the (i-1)th maximum point; This represents the average slope of the sequence of maxima.

[0026] The variation in period is influenced by both slope variation and second-level wind speed fluctuations. The size of a pseudo-period and wind speed fluctuations are mutually influential and constraining; their relationship can be approximated as an inverse proportional function. Furthermore, second-level wind speed fluctuations are calculated using the wind speed standard deviation and mean. Therefore, the period variation C... T The formula is:

[0027]

[0028] Among them, C T This indicates periodic variation, and std represents standard deviation. denoted as mean, representing the wind speed in seconds between the (l+1)th maximum point and the lth maximum point, and mean representing the average value.

[0029] The updated periodic window width T′0 formula is as follows:

[0030] T′0=2·C T (EI′ l+1 -EI l )

[0031] Where T′0 represents the updated periodic window width.

[0032] S303. Based on the updated periodic window width, repeat steps S201-S302 to determine the next maximum point in the maximum point sequence, until the maximum point at the current moment is obtained, thus obtaining the online maximum point wind speed sequence. This sequence includes the online maximum point location sequence Ei = EI1, EI2, ... EI l ,...Ei t-1 Ei t And the numerical sequence of maximum points Ev = EV1, EV2, ..., EV l ...Ev t-1 Ev t t represents the current time.

[0033] This invention also proposes a system for predicting pseudo-periodic maximum wind speeds along high-speed railway lines, comprising:

[0034] The data acquisition module is used to obtain a historical second-level wind speed sequence Ts for high-speed rail, where Ts: x = x1, x2, ..., x n x nLet e1 represent the wind speed at the nth second. Select the first point that is greater than both the previous and the next wind speed as the first local extreme point. Set the initial period window width T0 based on experience.

[0035] The maximum value sequence acquisition module is used to extract local extreme values ​​from the historical second-level wind speed sequence of high-speed rail in order to obtain non-fixed period extreme wind speeds that conform to the second-level wind speed fluctuation characteristics. By judging the distance at the local extreme values, the maximum value sequence is obtained. This sequence includes the maximum value position sequence and the maximum value value sequence.

[0036] The online maximum wind speed sequence acquisition module addresses the issue that the pseudo-periodic nature of wind speed time series causes wind speed fluctuations to oscillate within a certain period, while fixed-period searches may fail to find extreme points. Therefore, based on the maximum point sequence, the initial period window width T0 is adjusted in real time according to historical second-level wind speeds to obtain the online maximum wind speed sequence, providing training and testing data for maximum point prediction.

[0037] The maximum point prediction module is used to divide the online maximum wind speed sequence into training and test sets according to a set ratio, and use a bidirectional long short-term memory neural network (BiLSTM) to make rolling predictions of the future maximum point location and value.

[0038] The present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the pseudo-periodic maximum wind speed prediction method along the high-speed railway described above.

[0039] The present invention also proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the aforementioned method for predicting the pseudo-periodic maximum wind speed along high-speed railway lines.

[0040] The present invention adopts the above technical solution, and compared with the prior art, its significant technical effects are as follows:

[0041] This invention predicts the time, location, and magnitude of future wind speed maxima based on extreme wind speeds along high-speed railways. It also adaptively and dynamically adjusts the extreme value search period window width to remove pseudo-periods based on the slope and periodic variations of wind speeds, and uses BiLSTM to perform rolling prediction of the location and magnitude of the maximum value point, thus providing support for a high-speed railway wind speed early warning system.

[0042] The method proposed in this invention breaks with the existing prediction framework. First, it changes the prediction target from mean to maximum value prediction, which is more reasonable for high-speed trains where a wind speed of 1-2 seconds can overturn them. Second, the proposed pseudo-periodic extreme value extraction method can extract the main period of second-level wind speed fluctuations and remove non-interfering wind speeds, thereby enhancing the predictability of wind speed. Finally, this prediction method can transform traditional fixed-time interval wind speed prediction into non-fixed-time interval prediction that focuses on the main period, which can improve the accuracy of prediction and obtain the time when future extreme values ​​will occur. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the overall implementation of the present invention.

[0044] Figure 2 This is a schematic diagram illustrating the dynamic changes in the search period in an embodiment of the present invention.

[0045] Figure 3 This is a schematic diagram of the fitting and prediction results of the maximum value location sequence in an embodiment of the present invention.

[0046] Figure 4 This is a schematic diagram of the wind speed pseudo-period division result based on the extraction of the maximum point in an embodiment of the present invention. Detailed Implementation

[0047] The technical solution of the present invention will now be clearly and completely described in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0048] This invention proposes a method for predicting pseudo-periodic maximum wind speeds along high-speed railway lines, such as... Figure 1 As shown, it includes the following steps:

[0049] S1. Obtain the second-level wind speed sequence Ts of a high-speed railway station from 9:00 AM to 12:08 PM on January 3, 2021, where Ts: x = x1, x2, ..., x 11280 x 11280 Let e ​​represent the wind speed at second 11280. Data from the first 3 hours is used as the offline part of the algorithm to form a sufficient offline extreme value sequence. Data from the last 8 minutes is used as the online part of the algorithm for testing. The first point that is greater than both the previous and subsequent wind speeds is selected as the first local extreme point e1. Based on the statistical analysis of the fluctuation cycle of second-level wind speeds throughout 2020, the initial period T0 = 20 is set.

[0050] S2. To obtain non-fixed-period extreme wind speeds that conform to the characteristics of second-level wind speed fluctuations, local extreme points are extracted from the historical second-level wind speed sequence of high-speed rail. By determining the distances to these local extreme points, an extreme point sequence e is obtained. This sequence includes the extreme point location sequence ei and the extreme point value sequence ev. Furthermore, a maximum value sequence is obtained, which includes the maximum value location sequence and the maximum value sequence, such as... Figure 2 As shown, the specific content is as follows:

[0051] S201. Take the first local extremum point e1 as the first extremum point, and search for all local extremum points within the T0 period window width starting from e1, for a total of m points.

[0052] In this embodiment, e1 = (ei1, ev1) = (14, 12.0), m = 1.

[0053] S202. Connect e1 with the first local extremum point using a dashed line. Calculate the sum of distances from all second-level wind speeds between these two local extremum points to this dashed line. The formula for calculating the sum of distances obtained at the first local extremum point is:

[0054]

[0055] Where, x j This represents the wind speed in the j-th second interval between the dashed lines, where J represents the total number of wind speeds in the second interval between the dashed lines, and Line represents the dashed line. p Ev represents the sum of distances at the local extremum point p. p Let ev represent the value of the p-th local extremum. p-1 This represents the value of the (p-1)th local extremum point; Used to avoid local spurious extrema.

[0056] Therefore, the distance and vector obtained within the T0 period window width is d.

[0057] In this embodiment, d = {d1} = {-0.4075}.

[0058] S203. Sort each element in the distance and vector, locate the position with the minimum distance and sum at 40, and set it as the second extreme point e2, e2 = (40, 11.7).

[0059] S204. Starting from the second extreme point e2, search for all local extreme points within a T0-period window. If there are no extreme points within the T0-period window, extend the search to a T0 / 2-period window and repeat steps S202-S203 to obtain the third extreme point e3. Starting from e3, repeat the search process until the extreme points at the end of the sequence Ts are searched. False extreme points are filtered out to obtain the extreme point sequence e, e = e1, e2, ..., e Ne N This represents the Nth extreme point.

[0060] S205. The extreme point sequence contains maxima and minima. Extract the maxima sequence E from the extreme point sequence. The maxima sequence includes the maxima position sequence EI and the maxima value sequence EV, where EI = EI1, EI2, ..., EI l EV = EV1, EV2, ..., EV l EI l Indicates the position of the l-th maximum point, EV l This represents the value of the l-th maximum point; the sequence of maximum points is represented as E = (EI, EV) = (EI1, EV1), (EI2, EV2), ..., (EI... l EV l ).

[0061] S3. Due to the pseudo-periodicity of wind speed time series, wind speed fluctuations occur within a certain period, while fixed-period searches may result in the absence of extreme points. Therefore, based on the maximum point sequence, the initial period window width T0 is adjusted in real time according to the second-level wind speed of the previous 3 hours to obtain an online maximum point wind speed sequence, providing training and testing data for maximum point prediction. The specific details are as follows:

[0062] S301, such as Figure 3 As shown, the l2-constrained least squares method of the Gaussian kernel model is used to fit and predict the sequence of maximum point positions in step S2, thus obtaining the next maximum point E′. l+1 Position EI′ l+1 . Figure 3 These are the predicted locations of the 14 maximum points.

[0063] S302. A pseudo-period is defined by setting the time interval between two maxima. For pseudo-periods with wind speeds on the order of seconds, their variability needs to be quantified. The slope variability is obtained using historical statistics of the maxima slope, specifically using the following formula:

[0064]

[0065] Where l represents the length of the maximum sequence, C K The slope variation refers to the rate of change of the slope of the maximum point sequence E; k i This represents the slope at the i-th maximum point. EV i EI represents the value of the i-th maximum point. i EV represents the position of the i-th maximum point. i-1 EI represents the value of the (i-1)th maximum point. i-1 This indicates the position of the (i-1)th maximum point; This represents the average slope of the sequence of maxima.

[0066] The variation in period is influenced by both slope variation and second-level wind speed fluctuations. The size of a pseudo-period and wind speed fluctuations are mutually influential and constraining; their relationship can be approximated as an inverse proportional function. Furthermore, second-level wind speed fluctuations are calculated using the wind speed standard deviation and mean. Therefore, the period variation C... T The formula is:

[0067]

[0068] Among them, C T This indicates periodic variation, and std represents standard deviation. denoted as mean, representing the wind speed in seconds between the (l+1)th maximum point and the lth maximum point, and mean representing the average value.

[0069] The updated periodic window width T′0 formula is as follows:

[0070] T′0=2·C T (EI′ l+1 -EI l )

[0071] Where T′0 represents the updated periodic window width.

[0072] S303. Based on the updated periodic window width, repeat steps S201-S302 to determine the next maximum point in the maximum point sequence, until the maximum point at the current moment is obtained, thus obtaining the online maximum point wind speed sequence. This sequence includes the online maximum point location sequence Ei = EI1, EI2, ... EI l ,...Ei t-1 Ei t And the numerical sequence of maximum points Ev = EV1, EV2, ..., EV l ...Ev t-1 Ev t t represents the current time. For example... Figure 4 As shown, this represents the location and magnitude of the maximum points in the online maximum wind speed sequence.

[0073] S4. Divide the online maximum wind speed sequence into training and testing sets according to a set ratio, and use a bidirectional long short-term memory neural network to make rolling predictions of the location and value of future maximum points.

[0074] This invention also proposes a system for predicting pseudo-periodic maximum wind speeds along high-speed railway lines, including a data acquisition module, a maximum point sequence acquisition module, an online maximum point wind speed sequence acquisition module, a maximum point prediction module, and a computer program that can run on a processor. It should be noted that each module in the above system corresponds to a specific step of the method provided in this invention embodiment, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention embodiment.

[0075] This invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. It should be noted that when the processor executes the computer program, it corresponds to the specific steps of the method provided in this invention, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention.

[0076] This invention also proposes a computer-readable storage medium storing a computer program. It should be noted that when the computer program is executed by a processor, it corresponds to the specific steps of the method provided in this invention, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention.

[0077] The above description is merely an exemplary embodiment of the present invention and does not limit the scope of patent protection of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for predicting pseudo-periodic maximum wind speed along high-speed railway lines, characterized in that, include: S1. Obtain a historical high-speed rail wind speed sequence down to the second level. , , Indicates the first For wind speeds measured in seconds, select the first point that is greater than both the previous and the next wind speed as the first local extremum. Set the initial period window width ; S2. Extract local extreme points from the historical second-level wind speed sequence of high-speed rail. By judging the distances at the local extreme points, obtain the maximum value sequence, which includes the maximum value position sequence and the maximum value value sequence. S3. Based on the maximum point sequence, adjust the initial period window width in real time according to the historical second-level wind speed. Adjustments were made to obtain the online maximum wind speed sequence; specifically: S301. Using the l2-constrained least squares method of the Gaussian kernel model, the sequence of maximum point positions in step S2 is fitted and predicted to obtain the next maximum point. Location ; S302. Set the time interval between two maximum points to form a pseudo-period. For the pseudo-period of wind speed at the second level, quantify the variability. The slope variation is obtained by using historical statistics of the slope at the maximum value, and the specific formula is as follows: ； in, Indicates the length of the maximum value sequence. Indicates slope variation; Indicates the first The slope at each maximum point , Indicates the first The value of each maximum point Indicates the first The location of the maximum point Indicates the first -1 is the value of the maximum point. Indicates the first -1 location of the maximum point; This represents the average slope of the sequence of maxima. ; Based on slope variation and second-level wind speed fluctuations, the period variation is calculated using the following formula: ； in, Indicates periodic variation, Indicates standard deviation, Indicates the first The maximum point and the th maximum point Wind speeds measured in seconds between each maximum point This represents the average value; Updated periodic window width The formula is as follows: ； in, Indicates the window width of the update cycle; S303. Based on the updated periodic window width, repeat steps S201-S302 to determine the next maximum point in the maximum point sequence, until the maximum point at the current moment is obtained, thus obtaining the online maximum point wind speed sequence, which includes the online maximum point location sequence. and the numerical sequence of maximum points , Indicates the current moment; S4. Divide the online maximum wind speed sequence into training and testing sets according to a set ratio, and use a bidirectional long short-term memory neural network to make rolling predictions of the location and value of future maximum points.

2. The method for predicting pseudo-periodic maximum wind speed along high-speed railways according to claim 1, characterized in that, Step S2, obtaining the sequence of maximum points, includes the following sub-steps: S201, the first local extreme point As the first extreme point, Starting from, moving backward Search for all local extrema within the periodic window width, total indivual; S202, will Draw a dashed line connecting the p-th local extremum point and the p-th local extremum point. Calculate the sum of the distances from the dashed line to all second-level wind speeds between these two local extremum points. The distance obtained at the location and the calculation formula are as follows: ； in, Indicates the number of dashed lines. The wind speed is measured in seconds, and J represents the total number of wind speeds measured in seconds between the dashed lines. Indicates a dashed line. Represents local extreme points The distance at the location and, Indicates the first The values ​​of local extrema. This represents the value of the (p-1)th local extremum point; exist The distance and vector obtained at all local extrema within the periodic window width is d. ; S203. Sort each element in the distance and vector, locate the position with the minimum distance and vector, and set it as the second extreme point. ; S204, at the second extreme point Starting from, moving backward Search for all local extrema within the periodic window width, if If there are no extreme values ​​within the periodic window width, then the window is extended further. Searching using a periodic window width of / 2, and repeating steps S202-S203, we obtain the third extreme point. ,by Starting from the sequence, repeat the search process until the sequence is found. The search for extreme points at the end is completed, and false extreme points are filtered out to obtain the sequence of extreme points. , , This represents the Nth extreme point; S205. The extreme point sequence contains both maximum and minimum points. Extract the maximum point sequence from the extreme point sequence. The sequence of maximum points includes the sequence of maximum point locations. and the numerical sequence of maximum points ,in , , This indicates the location of the l-th maximum point. This represents the value of the l-th maximum point; the sequence of maximum points is represented as... .

3. A system applied to the method for predicting pseudo-periodic maximum wind speed along high-speed railways as described in claim 1, characterized in that, include: The data acquisition module is used to acquire a historical second-level wind speed sequence for high-speed rail. , , Indicates the first For wind speeds measured in seconds, select the first point that is greater than both the previous and the next wind speed as the first local extremum. Set the initial period window width ; The maximum value sequence acquisition module is used to extract local extreme points in the historical second-level wind speed sequence of high-speed rail. By judging the distance at the local extreme points, the maximum value sequence is obtained. This sequence includes the maximum value position sequence and the maximum value value sequence. The online maximum wind speed sequence acquisition module is used to adjust the initial period window width in real time based on the maximum wind speed sequence and historical second-level wind speed. Adjustments were made to obtain the online maximum wind speed sequence; The maximum point prediction module is used to divide the online maximum wind speed sequence into training and test sets according to a set ratio, and use a bidirectional long short-term memory neural network to make rolling predictions of the future maximum point location and value.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for predicting the pseudo-periodic maximum wind speed along a high-speed railway as described in any one of claims 1 to 2.

5. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is run by the processor, it executes any one of the pseudo-periodic maximum wind speed prediction methods along high-speed railways as described in claims 1 to 2.

Images