Wind power climbing event identification and classification method and system, terminal and medium

Abstract

The invention discloses a wind power climbing event identification and classification method and system, a terminal and a medium. The method comprises the following steps: extracting a power time sequence within a set duration from historical power data of a target wind power plant; adaptively adjusting a window length and a power threshold value through a forward incremental sliding window mechanism so as to identify each climbing time period in the power time sequence; performing regression feature extraction on the power sequence of each climbing period, and generating a multi-dimensional feature vector; and carrying out clustering analysis on the multi-dimensional feature vector of each climbing time period, and outputting event classification labels to which each climbing time period belongs according to a clustering result, wherein the event classification labels comprise upward climbing, downward climbing and continuous climbing. According to the method, multi-scale event capture is realized by using a forward incremental sliding window, continuous fluctuation characteristics are quantified through regression residual errors, the classification precision and the anti-noise capability are remarkably improved, and data support is provided for power grid climbing risk early warning and adjustment strategy optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy data processing, and particularly to a method, system, terminal and medium for identifying and classifying wind power ramp events. Background Art

[0002] With the rapid increase in the penetration rate of wind power in the power system, the volatility and randomness of wind power generation pose significant challenges to the safe operation of the power grid. Ramp events (phenomena of sharp increases or decreases in power within a short period) are the core risk sources leading to frequency violations and insufficient reserve capacity, and their accurate detection and classification have become key requirements for power grid stability control. However, existing methods have theoretical limitations in feature extraction dimensions, event definition frameworks, and anti-noise mechanism designs.

[0003] Traditional methods for identifying wind power ramp events usually use the power difference between the start and end points or simple slopes to divide up / down ramp events. Essentially, they belong to a one-dimensional feature decision-making mechanism. Such methods cannot effectively identify continuous ramp events (such as complex fluctuations where power first rises, then falls, and then rises again), resulting in significant biases in the assessment of the power grid's frequency regulation capacity requirements. The fundamental reason is that relying solely on the start and end point information cannot characterize the complex fluctuation characteristics within the time series. In addition, traditional methods lack clear definitions and detection criteria for continuous ramp events and usually rely on manual experience to set thresholds for the number of fluctuations or amplitudes. This subjective judgment method is difficult to adapt to the geographical and meteorological characteristics of different wind farms and cannot ensure the robustness of the classification results from a statistical perspective, leading to a high missed detection rate for complex fluctuation events. Local power fluctuations (such as short-term disturbances caused by sudden changes in wind speed) can easily cause the difference between the start and end points to deviate from the true trend, and traditional methods lack a mechanism to suppress the fluctuation amplitude, seriously affecting the accuracy of identifying and classifying wind power ramp events. Summary of the Invention

[0004] To address the deficiencies in the prior art, the present invention provides a method, system, terminal and medium for identifying and classifying wind power ramp events, which uses a forward-increasing sliding window to achieve multi-scale event capture and quantifies continuous fluctuation characteristics through regression residuals, significantly improving the classification accuracy and anti-noise ability, and providing data support for power grid ramp risk early warning and adjustment strategy optimization.

[0005] The present invention adopts the following technical solutions.

[0006] In a first aspect, the present invention provides a method for identifying and classifying wind power ramp events, the method comprising:

[0007] extracting a power time series within a set time period from the historical power data of the target wind farm;

[0008] identifying each ramp period in the power time series by adaptively adjusting the window length and power threshold through a forward-increasing sliding window mechanism;

[0009] Regression feature extraction is performed on the power sequences of each ramping period respectively, and multi-dimensional feature vectors are generated;

[0010] Cluster analysis is performed on the multi-dimensional feature vectors of each ramping period, and according to the clustering results, the event classification labels to which each ramping period belongs are output: up-ramping, down-ramping, and continuous ramping.

[0011] Optionally, the adaptive adjustment of the window length and power threshold through the forward incremental sliding window mechanism to identify each ramping period in the power time series includes the following steps:

[0012] Step a: Set the initial window length L0 of the sliding window, the initial power threshold P0, the sliding window step ΔL, the window length increment Δt, and the power threshold increment ΔP, and locate the starting sliding window position of the power time series;

[0013] Step b: Determine whether the maximum power change rate within the current sliding window is greater than the current power threshold;

[0014] Step c: If the judgment result in step 2.2 is no, then increase the window length of the current sliding window by Δt and increase the power threshold by ΔP, and determine whether the increased power threshold is greater than the rated power of the fan. If it is not greater, then return to execute step b; if it is greater, then execute step d;

[0015] Step d: Move the starting position of the current sliding window forward by ΔL, reset the length of the current window to L0 and the power threshold to P0, and then return to execute step b;

[0016] Step e: If the judgment result in step b is yes, then record the current sliding window period as a ramping period, and return to execute step d until the sliding window traverses all the data of the power time series, and output all the power sequences marked as ramping periods.

[0017] Optionally, in step b, the calculation formula of the maximum power change rate within the current sliding window is as follows:

[0018]

[0019] In the formula, ΔP max is the maximum power change rate of the power sequence within the time window [t i , t j of the current sliding window, and P(t i ) and P(t j ) are the power values corresponding to the starting time t i and the ending time t j of the window respectively.

[0020] Optionally, in step c, the expressions for increasing the window length of the current sliding window by Δt and the power threshold by ΔP are as follows:

[0021] L k+1 = L k + Δt

[0022] P k+1 = P k + ΔP

[0023] Wherein, L k and L k+1 are the original window length and the window length after increment of the sliding window respectively, and P k and P t+1 are the original power threshold and the power threshold after increment respectively.

[0024] Optionally, the regression feature extraction for the power sequences of each ramp period respectively includes the following steps:

[0025] Use a linear regression model to fit a trend line to the power sequence data of each ramp period: P(t) = a·t + b, where P(t) represents the power at time t, a is the regression coefficient, and b is the intercept;

[0026] Calculate the coefficient of determination R 2 and the residual index between each trend line and the power sequence data of each corresponding ramp period respectively;

[0027] Combine the regression coefficient a, the coefficient of determination R 2 and the residual index corresponding to each ramp period to form a multi-dimensional feature vector for each ramp period.

[0028] Optionally, the residual index includes: root mean square error RMSE, mean absolute error MAE, and / or mean absolute percentage error MAPE.

[0029] Optionally, the algorithm used for clustering analysis of the multi-dimensional feature vectors of each ramp period is: Kmeans++ clustering algorithm, and the number of clusters is 3.

[0030] In a second aspect, the present invention provides a wind power ramp event identification and classification system, which operates according to the steps of any one of the first aspects of the present invention. The system includes:

[0031] A power sequence extraction module, configured to extract a power time sequence within a set time length from the historical power data of a target wind farm;

[0032] A ramp period identification module, configured to adaptively adjust the window length and the power threshold through a forward incremental sliding window mechanism to identify each ramp period in the power time sequence;

[0033] A regression feature extraction module, which is used to respectively extract regression features from the power sequences of each climbing period and generate multi-dimensional feature vectors;

[0034] A climbing event classification module, which is used to perform clustering analysis on the multi-dimensional feature vectors of each climbing period and output the event classification labels to which each climbing period belongs according to the clustering results: up climbing, down climbing, and continuous climbing.

[0035] Thirdly, the present invention provides a terminal, including a processor and a storage medium;

[0036] The storage medium is used to store instructions;

[0037] The processor is used to operate according to the instructions to execute the steps of the method described in any one of the first aspects of the present invention.

[0038] Fourthly, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the method described in any one of the first aspects of the present invention are implemented.

[0039] The beneficial effects of the present invention are as follows. Compared with the prior art:

[0040] The present invention uses a forward-increasing sliding window to achieve multi-scale event capture, and quantifies the continuous fluctuation characteristics through regression residuals, significantly improving the classification accuracy and anti-noise ability, and providing data support for grid climbing risk early warning and regulation strategy optimization. Specifically as follows:

[0041] 1. It has stronger dynamic self-adaptability. Compared with the problem that traditional methods rely on fixed window parameters or static thresholds, resulting in insufficient adaptability to complex fluctuation events. The present invention uses a forward-increasing sliding window mechanism, and dynamically adjusts the window length and power threshold according to the data fluctuation situation in the window to accurately identify the climbing period. This dynamic adjustment can more flexibly adapt to climbing events with different time scales and fluctuation intensities.

[0042] 2. It has higher multi-dimensional feature extraction and classification accuracy. Traditional methods usually only rely on the difference in window mean and one-dimensional threshold for judgment and classification, and cannot capture the non-linear fluctuations inside the sequence; or through trend segment integration and fuzzy classification, but its feature extraction dimension is relatively low (such as only relying on power difference and duration), and it is easy to miss complex fluctuation patterns. The present invention constructs a linear regression model for each climbing period, extracts slope regression coefficients, determination coefficients and residual indexes to construct multi-dimensional feature vectors, and classifies them through a clustering algorithm. These features quantify the trend intensity, goodness of fit and residual fluctuation characteristics, and can effectively distinguish up climbing, down climbing and continuous climbing, improving the recognition and classification accuracy of wind power climbing events.

[0043] 3. Significantly improved anti-noise ability. Traditional methods are sensitive to mutations and vulnerable to disturbances caused by transient power (such as sudden changes in wind speed), and are prone to misjudgment due to local fluctuations, resulting in a relatively high false alarm rate. The present invention uses regression residual indicators (RMSE / MAE / MAPE) to quantify fluctuating noise and filter out local disturbances, greatly enhancing the anti-noise ability.

[0044] 4. More scientific definition and identification of continuous ramping events. Traditional methods do not clearly distinguish continuous ramping events and are prone to mis-segmenting complex fluctuations into multiple independent events; different from the traditional binary classification framework, the present invention clearly defines continuous ramping events through data-driven (clustering algorithm) without relying on manual experience rules (such as "the number of fluctuations exceeds the threshold"), ensuring the robustness of the classification results.

[0045] 5. Higher computational efficiency and generalization. Traditional methods usually need to combine prediction models such as neural networks for classification, resulting in a high computational cost; or need to iterate and optimize the gate width multiple times, with a high algorithm complexity. The present invention only needs to use an adaptive sliding window mechanism for rapid identification, increasing the window only when an anomaly is detected, greatly reducing the redundant computational amount, and the feature extraction and clustering processes do not rely on complex optimization processes, improving the computational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of the method for identifying and classifying wind power ramping events in an embodiment of the present invention;

[0047] Figure 2 It is a schematic flow chart of the forward incremental sliding window mechanism in an embodiment of the present invention;

[0048] Figure 3 It is a schematic diagram of the clustering results of wind power ramping events in an embodiment of the present invention;

[0049] Figure 4 It is a schematic block diagram of the structure principle of the system for identifying and classifying wind power ramping events in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. The embodiments described herein are only a part of the embodiments of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0051] Embodiment 1:

[0052] Referring to Figure 1 , an embodiment of the present invention provides a method for identifying and classifying wind power ramping events, which specifically includes the following steps:

[0053] S1. Extract the power time series within a set duration from the historical power data of the target wind farm;

[0054] Specifically, extract the power data according to the set time interval scale, and obtain the power time series after normalization. According to the actual calculation requirements, data extraction can be performed according to time scales such as 15 min, 30 min, or 45 min, etc.; preferably, in this embodiment, the power data is extracted according to the hourly scale.

[0055] S2. Adaptive adjust the window length and power threshold through a forward-incremental sliding window mechanism to identify each ramp period in the power time series;

[0056] Refer to Figure 2 , and the process of the forward-incremental sliding window mechanism is as follows:

[0057] Step a: Set the initial window length L0, initial power threshold P0, sliding window step ΔL, window length increment Δt, and power threshold increment ΔP of the sliding window, and locate the starting sliding window position of the power time series;

[0058] Step b: Determine whether the maximum power change rate within the current sliding window is greater than the current power threshold;

[0059] Step c: If the judgment result of step 2.2 is no, increase the window length of the current sliding window by Δt and the power threshold by ΔP, and determine whether the increased power threshold is greater than the rated power of the fan. If it is not greater, return to execute step b; if it is greater, execute step d;

[0060] Step d: Move the starting position of the current sliding window forward by ΔL, reset the length of the current window to L0 and the power threshold to P0, and then return to execute step b;

[0061] Step e: If the judgment result of step b is yes, record the current sliding window period as a ramp period, and return to execute step d until the sliding window traverses all the data of the power time series, and output all the power sequences marked as ramp periods.

[0062] A preferred but non-limiting embodiment, when performing a forward-increasing sliding window: Set the initial sliding window length to 2h, and set the initial threshold to 20% of the rated power; Increase the forward sliding window length in a fixed step. For every 1h increase, the threshold is correspondingly increased by 10% of the rated power. At each time point of the power sequence, dynamically adjust the duration of the sliding window and the threshold according to the fluctuation of the data within the current sliding window; When the maximum change rate of the data within the sliding window exceeds the current threshold, the sliding window is incremented forward by 1h, and the threshold is correspondingly increased, and then the next time point is detected. When it is found that the data change rate exceeds the threshold within a certain sliding window, that sliding window is determined as a potential ramping period and recorded. Subsequently, move the starting point of the sliding window forward to the end of that period, continue to increment the sliding window forward and perform a new round of detection. If no abnormality is detected, the sliding window continues to move forward, repeating the detection process until the incremental sliding window analysis of the entire data set is completed; The corresponding process includes:

[0063] (1) Define variables:

[0064] The window length L of the sliding window k : The window length after the k-th increment; The initial value L0 = 2h.

[0065] The power threshold P k : The threshold after the k-th increment; The initial value P0 = 20%·P c , P c is the rated power.

[0066] Step size: ΔL = 1h; Window length increase: Δt Threshold increment ΔP = 10%·P c .

[0067] (2) Dynamic adjustment rules:

[0068] When the power change rate within the sliding window does not exceed the current threshold, trigger an increment:

[0069] L k+1 = L k + Δt

[0070] P k+1 = P k + ΔP

[0071] In the formula, L k and L k+1 are the original window length and the incremented window length of the sliding window respectively, and P k and P t+1 are the original power threshold and the incremented power threshold respectively.

[0072] Otherwise, keep L k+1 = L k , P k+1 = P k。

[0073] (3) Power change calculation

[0074] For the power sequence {P(t i , t j )} within the time window [t i , t i+1 ,..., P(t j )} within the time window [t

[0075]

[0076] In the formula, ΔP max is the maximum power change rate of the power sequence within the time window [t i , t j of the current sliding window, and P(t i ) and P(t j ) are the power values corresponding to the start time t i and the end time t j of the window respectively.

[0077] If ΔP max ≥ P k , it is marked as a climbing event.

[0078] (4) Sliding window movement rule: When the threshold of the current round of sliding window is not less than the rated power, or after the current round of window is marked as a climbing event, reset the sliding window length L k to the initial value, reset the power threshold P k to the initial value, move the sliding window forward by 1 h, and continue the next round of detection until all data is covered.

[0079] S3. Respectively extract the regression features of the power sequences of each climbing period and generate multi-dimensional feature vectors;

[0080] A: Use a linear regression model to fit the trend line to the power sequence data of each climbing period: P(t) = a·t + b, where P(t) represents the power at time t, a is the regression coefficient, and b is the intercept;

[0081] B: Calculate the coefficient of determination R 2 and the residual index of each trend line and the power sequence data of each corresponding climbing period respectively;

[0082] C: Combine the regression coefficient a, the coefficient of determination R 2 and the residual index corresponding to each climbing period to form the multi-dimensional feature vector of each climbing period.

[0083] Specifically, the residual indicators include one or more of root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). In a preferred but non-limiting embodiment, the present invention constructs a five-dimensional feature vector: F = [a, R 2 , RMSE, MAE, MAPE].

[0084] Furthermore, the coefficient of determination R 2 measures the goodness of fit of the regression model to the data and reflects the proportion of the model in explaining the fluctuations of the dependent variable; the closer R 2 is to 1, the better the model fitting effect; if it is negative, the model is invalid; the calculation formula of R 2 is as follows:

[0085]

[0086] The root mean square error (RMSE) represents the square root of the mean square error between the predicted value and the actual value, reflecting the prediction accuracy. The RMSE is sensitive to outliers, and the smaller the value, the more accurate the prediction; the calculation formula of RMSE is as follows:

[0087]

[0088] The mean absolute error (MAE) represents the average of the absolute errors between the predicted value and the actual value, reflecting the average error size; the MAE is not sensitive to outliers and more robustly reflects the average error; the calculation formula of MAE is as follows:

[0089]

[0090] The mean absolute percentage error (MAPE) converts the absolute error into a percentage form, facilitating cross-data magnitude comparison; the smaller the MAPE, the lower the relative prediction error; the calculation formula of MAPE is as follows:

[0091]

[0092] Where y i is the actual power value, is the predicted value of the regression model, is the mean of the actual values, and n represents the duration of the calculated ramping period.

[0093] S4. Perform cluster analysis on the multi-dimensional feature vectors of each ramping period, and output the event classification labels to which each ramping period belongs according to the clustering results: up-ramping, down-ramping, and continuous ramping.

[0094] Further, those skilled in the art can adopt algorithms such as K-Means clustering, density clustering, and hierarchical clustering when performing clustering analysis. However, in order to obtain better technical effects, the Kmeans++ clustering algorithm in the embodiments of the present invention is used to cluster and divide each climbing period, and the number of clusters is 3, corresponding to uphill climbing, downhill climbing, and continuous climbing respectively.

[0095] As an optimized method of the Kmeans algorithm, the main optimization point of the Kmeans++ algorithm lies in the optimization of the method of randomly initializing the clustering centers in the Kmeans algorithm; the traditional Kmeans method determines the clustering centers in a completely random manner, while Kmeans++ introduces a more refined initialization mechanism, which is based on a key principle: ensuring that the initial clustering centers are as far apart from each other as possible; in Kmeans++, by calculating the distances between the sample points and the already selected clustering centers (randomly select one as the starting point first), a probability-weighted method is adopted, so that the sample points farther from the current clustering center have a higher probability of being selected as the next clustering center, and this process is iterated until all clustering centers are selected. Kmeans++ not only improves the clustering effect, but also eliminates the uncertainty brought by random initialization, thereby enhancing the computational accuracy and stability of the algorithm.

[0096] When performing clustering analysis, the initial number of clusters k and its corresponding center C are first determined by randomly selecting data objects in the dataset, and then the Euclidean distances between these centers and other objects in the dataset are calculated. This distance is used as a key metric in the clustering process, and the calculation formula is as follows:

[0097]

[0098] where x is the data object, μ is the clustering center, x i and μ i are the attribute values of the i-th dimensional feature of the data x and the clustering center μ respectively.

[0099] In this embodiment, when performing the simulation experiment, according to the hourly wind power sequence of the target wind farm in 2021, the initial sliding window length is set to 2 hours, and the rated power p c is 100MW, and the method provided by the present invention is used for identification and classification; since the determination coefficient R 2 , root mean square error RMSE, mean absolute error MAE, and mean absolute percentage error MAPE in the five-dimensional index can all reflect the fitting effect of the trend line on the climbing period, for the convenience of drawing, only the slope a and the root mean square error RMSE are retained as the horizontal and vertical coordinates in the clustering result graph. The clustering results are as Figure 3 and shown in Table 1, Figure 3The horizontal coordinate is the slope, and the vertical coordinate is the RMSE. Each point represents a ramping event, and different colors represent the clustering results. The values of the characteristic indicators for each event type in Table 1 are the averages of the characteristic indicators of all ramping events within their respective clusters.

[0100] Table 1 Five-dimensional feature vector indicators

[0101]

[0102] As can be seen from Table 1, the up-ramping events (purple) are concentrated in the positive slope region (a > 200), and the RMSE is relatively low (about 600 - 700); the down-ramping events (yellow) are distributed in the negative slope region (a ≈ -200), and the RMSE is slightly higher (about 600); the continuous ramping events (green) have a slope close to zero (a ≈ 2.3), but the RMSE is significantly higher (about 1800). Figure 3 As can be seen, the up, down, and continuous ramping events are concentrated at the corresponding clustering centers and are divided into different categories. The chart results verify the discrimination ability of the characteristic indicators provided by the present invention and verify the effectiveness of the method for identifying and classifying wind power ramping events proposed by the present invention.

[0103] The beneficial effects of the present invention are as follows compared with the prior art:

[0104] The present invention uses a forward-increasing sliding window to achieve multi-scale event capture and quantifies the continuous fluctuation characteristics through regression residuals, significantly improving the classification accuracy and anti-noise ability, and providing data support for grid ramping risk early warning and regulation strategy optimization. Specifically as follows:

[0105] 1. Stronger dynamic self-adaptability. Compared with the traditional method that relies on fixed window parameters or static thresholds, resulting in insufficient adaptability to complex fluctuation events. The present invention uses a forward-increasing sliding window mechanism and dynamically adjusts the window length and power threshold according to the data fluctuation within the window to accurately identify the ramping period. This dynamic adjustment can more flexibly adapt to ramping events of different time scales and fluctuation intensities.

[0106] 2. Higher multi-dimensional feature extraction and classification accuracy. Traditional methods usually rely only on the difference in window mean and one-dimensional threshold for judgment and classification, unable to capture the non-linear fluctuations within the sequence; or through trend segment integration and fuzzy classification, but their feature extraction dimension is relatively low (such as only relying on power difference and duration), and it is easy to miss complex fluctuation patterns. The present invention constructs a linear regression model for each ramping period, extracts slope regression coefficients, determination coefficients, and residual indicators to construct a multi-dimensional feature vector, and classifies through a clustering algorithm. These features quantify the trend intensity, goodness of fit, and residual fluctuation characteristics, and can effectively distinguish up-ramping, down-ramping, and continuous ramping, improving the identification and classification accuracy of wind power ramping events.

[0107] 3. The anti-noise ability is significantly improved. Traditional methods are sensitive to mutations and vulnerable to disturbances caused by transient power (such as sudden changes in wind speed), and are prone to misjudgment due to local fluctuations, resulting in a high false alarm rate. The present invention uses regression residual indicators (RMSE / MAE / MAPE) to quantify the fluctuating noise and filter out local disturbances, greatly enhancing the anti-noise ability.

[0108] 4. The definition and recognition of continuous ramping events are more scientific. Traditional methods do not clearly distinguish continuous ramping events and are prone to mis-segmenting complex fluctuations into multiple independent events; different from the traditional binary classification framework, the present invention clearly defines continuous ramping events through data-driven (clustering algorithm), without relying on manual experience rules (such as "the number of fluctuations exceeds the threshold"), ensuring the robustness of the classification results.

[0109] 5. The computational efficiency and generalization are better. Traditional methods usually need to combine prediction models such as neural networks for classification, with a high computational cost; or need to iteratively optimize the gate width multiple times, with a high algorithm complexity. The present invention only needs to use an adaptive sliding window mechanism for rapid identification, and only increases the window when an anomaly is detected, greatly reducing the redundant computational amount, and the feature extraction and clustering processes do not rely on complex optimization processes, improving the computational efficiency.

[0110] Embodiment 2:

[0111] As Figure 4 shown, the present invention provides a wind power ramping event identification and classification system, which is used to implement the steps of the method in Embodiment 1 above. The system specifically includes:

[0112] A power sequence extraction module, which is used to extract the power time series within a set time period from the historical power data of the target wind farm;

[0113] A ramping period identification module, which is used to adaptively adjust the window length and power threshold through a forward-incremental sliding window mechanism to identify each ramping period in the power time series;

[0114] A regression feature extraction module, which is used to extract regression features from the power sequences of each ramping period and generate multi-dimensional feature vectors;

[0115] A ramping event classification module, which is used to perform clustering analysis on the multi-dimensional feature vectors of each ramping period and output the event classification labels to which each ramping period belongs according to the clustering results: up-ramping, down-ramping, and continuous ramping.

[0116] The wind power ramping event identification and classification system provided by the embodiment of the present invention and the wind power ramping event identification and classification method provided by Embodiment 1 are based on the same technical concept, and can produce the beneficial effects as described in Embodiment 1. The content not described in detail in this embodiment can be referred to Embodiment 1.

[0117] Embodiment 3:

[0118] A terminal provided by an embodiment of the present invention includes a processor and a storage medium;

[0119] The storage medium is used to store instructions;

[0120] The processor is configured to operate according to the instructions to execute the steps of the method according to any one of Embodiment 1.

[0121] Embodiment 4:

[0122] A computer-readable storage medium provided by an embodiment of the present invention stores a computer program, and when the program is executed by a processor, the steps of the method according to any one of Embodiment 1 are implemented.

[0123] The present disclosure may be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0124] The computer-readable storage medium may be a tangible device that can retain and store instructions used by an instruction execution device. The computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanically encoded device such as a punched card or raised structures in grooves having instructions stored thereon, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not construed as an instantaneous signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0125] The computer-readable program instructions described herein can be downloaded to various computing / processing devices from a computer-readable storage medium or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0126] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer-readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field-programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer-readable program instructions to implement various aspects of the present disclosure.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. A method for identifying and classifying wind power ramp events, characterized in that, The method includes: extracting a power time series within a set time period from the historical power data of the target wind farm; adapting to adjust the window length and power threshold through a forward-incremental sliding window mechanism to identify each ramp period in the power time series; respectively performing regression feature extraction on the power sequences of each ramp period and generating multi-dimensional feature vectors; performing clustering analysis on the multi-dimensional feature vectors of each ramp period and outputting the event classification labels to which each ramp period belongs according to the clustering results: up-ramp, down-ramp, and continuous ramp.

2. The method for identifying and classifying wind power ramp events according to claim 1, wherein, The step of adapting to adjust the window length and power threshold through a forward-incremental sliding window mechanism to identify each ramp period in the power time series includes the following steps: Step a: Set the initial window length L0, initial power threshold P0, sliding window step ΔL, window length increment Δt, and power threshold increment ΔP of the sliding window, and locate the starting sliding window position of the power time series; Step b: Determine whether the maximum power change rate within the current sliding window is greater than the current power threshold; Step c: If the judgment result in step 2.2 is no, increase the window length of the current sliding window by Δt and increase the power threshold by ΔP, and determine whether the increased power threshold is greater than the rated power of the fan. If it is not greater, return to execute step b; if it is greater, execute step d; Step d: Move the starting position of the current sliding window forward by ΔL, reset the length of the current window to L0 and the power threshold to P0, and then return to execute step b; Step e: If the judgment result in step b is yes, record the current sliding window period as a ramp period, and return to execute step d until the sliding window traverses all the data of the power time series, and output all the power sequences marked as ramp periods.

3. The method for identifying and classifying wind power ramp events according to claim 2, characterized in that In step b, the calculation formula for the maximum power change rate within the current sliding window is as follows: where ΔP max is the maximum rate of change of the power sequence within the time window [t i , t j of the current sliding window, and P(t i ) and P(t j ) are the power values corresponding to the start time t i and the end time t j of the window, respectively.

4. The method for identifying and classifying wind power ramp events according to claim 2, characterized in that In step c, the expressions for increasing the window length of the current sliding window by Δt and increasing the power threshold by ΔP are as follows: L k+1 = L k + Δt P k+1 = P k + ΔP Wherein, L k and L k+1 are respectively the original window length and the window length after increment of the sliding window, and P k and P t+1 are respectively the original power threshold and the power threshold after increment.

5. The method for identifying and classifying wind power ramp events according to claim 1, wherein The step of respectively performing regression feature extraction on the power sequences of each ramp period and generating multi-dimensional feature vectors includes: using a linear regression model to fit a trend line to the power sequence data of each ramp period: P(t) = at + b, where P(t) represents the power at time t, a is the regression coefficient, and b is the intercept; Calculate the coefficient of determination R of each trend line and the power sequence data of each corresponding ramp-up period respectively 2 and the residual index; Combine the regression coefficient a and the coefficient of determination R corresponding to each ramping period 2 with the residual index to form a multi-dimensional feature vector for each ramping period.

6. The method for identifying and classifying wind power ramp events according to claim 5, wherein The residual indicators include: root mean square error RMSE, mean absolute error MAE, and / or mean absolute percentage error MAPE.

7. The method for identifying and classifying wind power ramp events according to claim 1, characterized in that The algorithm used for performing clustering analysis on the multi-dimensional feature vectors of each ramp period is: Kmeans++ clustering algorithm, and the number of clusters is 3.

8. A wind power ramp event identification and classification system, which operates the wind power ramp event identification and classification method according to any one of claims 1-7, wherein the system including: a power sequence extraction module for extracting a power time series within a set time period from the historical power data of the target wind farm; a ramp period identification module for adapting to adjust the window length and power threshold through a forward-incremental sliding window mechanism to identify each ramp period in the power time series; a regression feature extraction module for respectively performing regression feature extraction on the power sequences of each ramp period and generating multi-dimensional feature vectors; A ramp event classification module, which is used to perform clustering analysis on the multi-dimensional feature vectors of each ramp period, and output the event classification labels to which each ramp period belongs according to the clustering results: uphill ramp, downhill ramp, and continuous ramp.

9. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, the steps of the method according to any one of claims 1-7 are implemented.

Images