Wind power plant intelligent simulation acquisition test method and system

Abstract

The invention discloses a wind power plant intelligent simulation acquisition test method and system, and relates to the technical field of wind power plant intelligent acquisition and state prediction control, and the method comprises the steps: obtaining the operation state parameters of a wind turbine generator, and constructing a twinborn prediction model based on a time sequence. And comparing a model prediction value with a measured value to calculate an operation state error, and estimating prediction uncertainty. And dynamically adjusting the sampling frequency of the sampling channel according to the prediction uncertainty and controlling test execution. According to the method disclosed by the invention, a smart wind power plant simulation acquisition test system from data driving to state modeling to risk perception to dynamic regulation and control to multi-node collaborative integration is realized. According to the technical scheme, independent technical contributions are provided, link-by-link logic closed loops are constructed, a breakthrough is made from a traditional static acquisition mode, a digital wind field acquisition test system with predictability, self-adaptability and linkage response capability is formed, and the monitoring efficiency, the risk control granularity and the system resource utilization rate are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data collection and state prediction control of wind farms, and in particular to a wind farm intelligent simulation data collection and testing method and system. Background Art

[0002] As the wind power industry develops toward large-scale, high-density deployment, wind farms are increasingly demanding operational safety, energy-efficient scheduling, and predictive maintenance capabilities. In recent years, thanks to advances in sensor technology, edge computing, and artificial intelligence algorithms, wind farm operational monitoring has gradually evolved from single-point acquisition to multi-dimensional dynamic perception, with state prediction and data-driven fault identification becoming research hotspots. Against this backdrop, digital twin technology has begun to be applied to wind turbine modeling and fault trend analysis. By constructing time series models, it enables feedforward prediction of wind turbine operating status, providing data support for subsequent scheduling optimization and equipment maintenance. However, existing methods still primarily rely on fixed-period acquisition and lack dynamic control mechanisms, limiting the responsiveness of prediction models and the resource efficiency of acquisition systems.

[0003] Current mainstream wind farm operational data collection methods often employ statically configured, periodic sampling mechanisms, which are unable to adapt to the dynamic fluctuations in wind turbines' demands under varying meteorological conditions and operating states. This results in data redundancy when equipment is operating stably and data lag during sudden state changes, severely impacting the real-time and accuracy of fault prediction and intelligent control. Existing technologies attempt to apply deep learning models (such as LSTM) to wind power prediction or wind speed fitting, but these methods often overlook the uncontrollable fluctuations in model prediction accuracy over time and lack a closed-loop mechanism that dynamically feeds back sampling strategies based on model credibility, resulting in a decoupling between model training and collection strategies. Existing technologies typically perform anomaly detection based on preset thresholds, failing to implement stratified sampling adjustments based on state risk levels or establish a linkage logic between sampling frequency and equipment state change trends, limiting the adaptability of wind farm digital testing strategies. Furthermore, traditional data collection systems lack the ability to respond to the state linkage between adjacent wind turbines, making it impossible to conduct coordinated testing and intensive data capture before regional wind disturbances or group failure risks arise. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the existing wind farm operation data collection and testing methods have problems such as static configuration of sampling strategies, unstable risk prediction accuracy, and inability to link response tests, as well as how to construct an adaptive sampling frequency adjustment and test control strategy based on prediction uncertainty.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a wind farm intelligent simulation acquisition and testing method, comprising obtaining wind turbine operating status parameters and constructing a twin prediction model based on time series.

[0007] Compare the model predictions with the measured values ​​to calculate the operating state error and estimate the prediction uncertainty.

[0008] Dynamically adjust the sampling frequency of the sampling channel according to the prediction uncertainty and control the test execution.

[0009] Estimating forecast uncertainty involves selecting a sliding time window based on the error sequence of each wind turbine, extracting the standard deviation of the error within the window, evaluating the rate of change of the error over time, and forming a forecast uncertainty evaluation value through a weighted combination of the standard deviation and the rate of change of the error. The weighted combination is determined using empirical parameters, and upper and lower thresholds are set to determine the current state level. Different state levels trigger a sampling frequency strategy switching mechanism.

[0010] As a preferred solution of the wind farm intelligent simulation acquisition and testing method described in the present invention, the acquisition of wind turbine operating status parameters includes real-time acquisition of operating parameters from the target wind turbine, generating a time series sample set according to a fixed sampling period, and establishing a feature input data set corresponding to each wind turbine.

[0011] As a preferred solution of the wind farm intelligent simulation acquisition test method described in the present invention, the construction of a twin prediction model based on time series includes constructing a multi-time step input sequence through a sliding time window method, inputting the operating parameters into the time series prediction network, and the prediction network adopts a recurrent neural network based on a long short-term memory structure. The model structure supports historical window updates and prediction residual corrections, and uses a rolling window training mode to adjust the model parameters in each round of updates to adapt to the operating characteristics of different wind turbines and output predictions of the operating status of future target time steps. Model copies are deployed for each wind turbine unit, and the copy model automatically triggers the parameter self-update mechanism on an hourly basis. Each update uses measured data as a training set, and uses a sliding average and minimum mean square error strategy to calibrate the model error. The predicted output includes the future value of each monitored variable and the corresponding confidence interval, and an additional state confidence indicator is added to the model output.

[0012] As a preferred embodiment of the wind farm intelligent simulation acquisition and testing method described in the present invention, the method of comparing model predictions with measured values ​​to calculate operating state errors includes calculating a sequence of differences between the predicted and measured values ​​for each type of wind turbine operating state parameter at each time step, performing Euclidean norm normalization on the difference sequence, and generating an error sequence vector for error assessment. Furthermore, a lag matching operation within a preset time range is introduced into the error calculation, and sliding statistics are performed on the errors within consecutive time periods to construct a fluctuation function representing the error volatility. The function value serves as the basic indicator for uncertainty assessment.

[0013] As a preferred solution of the wind farm intelligent simulation acquisition test method described in the present invention, the estimated prediction uncertainty includes, based on the error sequence vector generated by each wind turbine, defining a sliding time window and extracting the prediction error standard deviation of all time steps in the window, calculating the error mean change trend in consecutive time steps, and obtaining the error change rate value. The prediction error standard deviation and the error change rate value are weighted to form a prediction uncertainty evaluation value of the credibility of the current prediction state. According to the prediction uncertainty evaluation value and the preset upper and lower limit threshold ranges, the current state of the wind turbine is divided into three state levels: low risk, medium risk and high risk. When the state level changes, the sampling frequency adjustment action is automatically triggered.

[0014] As a preferred embodiment of the wind farm intelligent simulation acquisition and testing method described in the present invention, the dynamic adjustment of the sampling frequency of the sampling channels based on the prediction uncertainty includes assigning a sampling frequency category to each channel according to the current status level of the wind turbine, with the frequency category being divided into three levels: basic frequency, enhanced frequency, and high-frequency acquisition. In the control logic, if the current status level corresponding to the current channel is at the lowest level, the basic sampling frequency is implemented. If the indicator is at the medium level, the sampling frequency is increased to the enhanced frequency. If the indicator is at the highest level, the sampling frequency is adjusted to high-frequency mode, and an auxiliary redundant acquisition channel is added.

[0015] As a preferred embodiment of the wind farm intelligent simulation data acquisition and testing method described in the present invention, the control test execution includes triggering an automatic test command upon identifying that the prediction uncertainty evaluation value of a particular wind turbine unit continues to exceed the normal range, placing the current sampling channel in a high-sensitivity monitoring state, and synchronously compressing and storing the sampled data within a specified time range. Simultaneously, a coordinated check is performed on the operating status of adjacent turbines within the current wind farm area. If there is a trend of synchronization deviation between adjacent turbines, a regional test linkage mechanism is automatically activated to execute distributed high-frequency data acquisition tasks within the area, and generate the next round of data acquisition priority sequence based on the current prediction model output.

[0016] Another object of the present invention is to provide a wind farm intelligent simulation acquisition and testing system, which can compare the model prediction value with the measured value through one of the schemes to calculate the operating status error and estimate the prediction uncertainty, thereby solving the problems of static configuration of sampling strategy, unstable risk prediction accuracy, and inability to link response testing in existing wind farm operation data acquisition and testing methods, as well as the problem of how to construct an adaptive sampling frequency adjustment and test control strategy based on prediction uncertainty.

[0017] As a preferred solution of the wind farm intelligent simulation acquisition and testing system described in the present invention, it includes: a model building module, a calculation and prediction uncertainty module, and a dynamic adjustment module.

[0018] The model building module is used to obtain the operating status parameters of the wind turbine generator set and build a twin prediction model based on time series.

[0019] The prediction uncertainty calculation module is used to compare the model prediction value with the measured value to calculate the operating state error and estimate the prediction uncertainty.

[0020] The dynamic adjustment module is used to dynamically adjust the sampling frequency of the sampling channel according to the prediction uncertainty and control the test execution.

[0021] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a step of a wind farm intelligent simulation acquisition test method.

[0022] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a wind farm intelligent simulation acquisition test method.

[0023] Beneficial effects of the present invention: The wind farm intelligent simulation acquisition test method provided by the present invention obtains the operating status parameters of the wind turbines, and constructs a twin prediction model based on time series to achieve the technical effect of deeply integrating the prediction system into the acquisition control mechanism and realizing feedforward intelligent regulation. Comparing the model prediction value with the measured value to calculate the operating status error, estimating the prediction uncertainty realizes the mechanism of converting the historical error trend into real-time control input, giving the system the ability to perceive the stability of the prediction, and achieving the technical effect of providing a dynamic adaptive basis for the sampling frequency adjustment strategy and realizing the closed-loop feedback of the acquisition control. Dynamically adjusting the sampling frequency of the sampling channel according to the prediction uncertainty and controlling the test execution achieves the technical effect of building an intelligent acquisition test process with full-domain linkage, high-frequency precision, and prediction in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 This is an overall flow chart of a wind farm intelligent simulation acquisition and testing method provided by the first embodiment of the present invention.

[0026] Figure 2 This is a system flow chart of a wind farm intelligent simulation acquisition and testing method provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0027] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0028] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a wind farm intelligent simulation acquisition test method, comprising:

[0029] S1: Obtain the operating status parameters of the wind turbine and build a twin prediction model based on time series.

[0030] The operating parameters of the target wind turbines are collected in real time, and a time series sample set is generated according to a fixed sampling period. A feature input data set is established for each wind turbine.

[0031] Furthermore, the operating parameters collected by the present invention include wind speed, rotor speed, output active power, yaw angle, generated current, generated voltage, main shaft vibration, and generator temperature. These parameters are collected with a fixed sampling period of 10 seconds, forming a continuous time-series data stream.

[0032] A multi-time-step input sequence is constructed by a sliding time window method. In the present invention, the sliding window length is set to 12 time steps (i.e., covering 2 minutes of data), and the sliding step length is set to 1 time step to improve the time sensitivity and short-term prediction ability of the model. The operating parameters are input into the time series prediction network. The prediction network adopts a recurrent neural network based on a long short-term memory structure. The model structure supports historical window updates and prediction residual corrections. In each round of updates, a rolling window training mode is used to adjust the model parameters to adapt to the operating characteristics of different wind turbines and output predictions of the operating status of the future target time step. Model copies are deployed for each wind turbine unit, and the copy model automatically triggers the parameter self-update mechanism on an hourly basis. Each update uses the measured data as the training set, and the sliding average and mean square error minimum strategy are used to calibrate the model error. The predicted output includes the future value of each monitored variable and the corresponding confidence interval, and a state confidence index is added to the model output. In the present invention, the state confidence index ranges from 0 to 1. The closer the value is to 1, the more stable and reliable the model prediction is.

[0033] A preferred solution for a recurrent neural network with a long-short-term memory (LSTM) structure specifically employs a time series prediction network structure based on a LSTM network. The input is a constructed multi-time-step feature sequence, and the output is a predicted operating state value for the next three time steps, covering a short-term prediction window of 30 seconds. The model structure consists of an input layer, two layers of hidden-state LSTM units, and a fully connected output layer. Each LSTM unit layer contains 64 neurons, with a tanh activation function.

[0034] It should be noted that S1 constructs time series samples by collecting multi-dimensional operating parameters, and establishes a twin prediction model for wind turbines by combining sliding windows and LSTM structures, thereby improving the accuracy and real-time performance of state predictions. It has parameter self-update and confidence output functions, realizes feedforward perception of operating risks, and provides an intelligent decision-making basis for subsequent sampling and adjustment.

[0035] S2: Compare the model predictions with the measured values ​​to calculate the operating state error and estimate the prediction uncertainty.

[0036] For each type of wind turbine operating state parameter, the difference sequence between the predicted value and the measured value is calculated at each time step, and the difference sequence is normalized using the Euclidean norm to generate an error sequence vector for error evaluation.

[0037] Furthermore, the running state error vector is constructed as:

[0038]

[0039] Among them, ε m (t s ) indicates that the wind turbine numbered m is at time step t sThe error evaluation value of . Indicates the nth state parameter of wind turbine m at time step t s The predicted value of is calculated by the LSTM network in S1. m,n (t s ) represents the nth state parameter of wind turbine m at time step t s D represents the total number of operating status parameter dimensions.

[0040] At the same time, a lag matching operation within a preset time range is introduced into the error calculation, and sliding statistics of the errors in continuous time periods are performed to construct a fluctuation function representing the error volatility. The function value is used as the basic indicator for uncertainty assessment.

[0041] Furthermore, the wave function is expressed as:

[0042]

[0043] Among them, θ m (t s ) represents wind turbine m at time step t s The error fluctuation function value. H represents the number of time steps of the sliding window (set to 12 in this invention). m (t s ) represents the mean error of wind turbine m within the sliding window, which can be expressed as:

[0044]

[0045] Where r represents the relative time step offset within the sliding time window, and its value range is 0 to H. m (t s -r) indicates that the wind turbine number is m at time step t s -r Error evaluation value.

[0046] Based on the error sequence vector generated by each wind turbine, a sliding time window is defined and the standard deviation of the forecast error for all time steps within the window is extracted. The trend of the error mean change in consecutive time steps is calculated to obtain the error change rate value. The forecast error standard deviation and the error change rate value are weighted to form a forecast uncertainty evaluation value that reflects the credibility of the current forecast state.

[0047] Furthermore, the prediction uncertainty evaluation value is calculated as:

[0048]

[0049] Among them, U m (t s ) Wind turbine m at time step t sThe prediction uncertainty evaluation value of α1 and α2 represent the weighted coefficients of volatility and trend change rate, which are set to 0.6 and 0.4 respectively in the present invention. It represents the rate of change of the mean error, which can be calculated approximately as:

[0050]

[0051] Where Δt represents the time step interval.

[0052] Based on the prediction uncertainty evaluation value and the preset upper and lower thresholds, the current status of the wind turbine is divided into three status levels: low risk, medium risk, and high risk. When the status level changes, the sampling frequency adjustment action is automatically triggered.

[0053] If U m (t s ) is less than 0.2, which is judged as a low risk level.

[0054] If U m (t s ) is greater than or equal to 0.2 and less than 0.5, which is judged as a medium risk level.

[0055] If U m (t s ) is greater than or equal to 0.5, which is judged as a high risk level.

[0056] When the status level changes, the sampling frequency adjustment action is automatically triggered, which serves as the trigger condition for subsequent sampling control.

[0057] It should be noted that the design concept of S2 is to introduce a mechanism that integrates prediction error and uncertainty assessment to dynamically quantify changes in wind turbine operating status. By constructing an error sequence using the LSTM model's predicted and measured values, and combining the sliding standard deviation with the error change rate to construct a prediction uncertainty evaluation value, this approach not only reflects state fluctuations but also detects sudden changes in trends. This step overcomes the existing problems of fixed sampling frequency and unresponsiveness to risk changes, enabling refined, state-driven dynamic control. Compared to traditional threshold or empirical strategies, it offers high responsiveness, adaptability, and scenario generalization capabilities.

[0058] S3: Dynamically adjust the sampling frequency of the sampling channel according to the prediction uncertainty and control the test execution.

[0059] Each channel type is assigned a sampling frequency category based on the wind turbine's current status level. These frequencies are divided into three levels: basic frequency, enhanced frequency, and high-frequency acquisition, with corresponding sampling periods of 60 seconds, 10 seconds, and 1 second, respectively. In the control logic, if the current channel's corresponding status level is at the lowest level, the basic sampling frequency is used. If the indicator is at the medium level, the sampling frequency is increased to enhanced frequency. If the indicator is at the highest level, the sampling frequency is adjusted to high-frequency mode, and auxiliary redundant acquisition channels are added to achieve multi-channel parallel sampling, which is used for subsequent data cross-validation and anomaly detection.

[0060] Furthermore, if the predicted uncertainty assessment value continuously exceeds the set high-risk threshold (0.5) for more than three time steps (i.e., 30 seconds), the unit is determined to be in an abnormal trend state, automatically triggering a test control instruction and placing all sampling channels of the unit into a high-sensitivity monitoring state. In this high-sensitivity state, the collected data is no longer delayed in writing. Instead, it is immediately buffered and a compression encoding program is initiated, using a specified compression format to store the complete raw sampling data for the last five minutes, with a compression ratio of no less than 3:1, to ensure the integrity of critical data.

[0061] At the same time, based on the current turbine's geographic location within the wind farm, other turbine nodes within its proximity radius are selected from the logical topology structure, and their operating status levels are read and analyzed in parallel. In this invention, the proximity radius is set to 500 meters. If two or more adjacent turbines are found to exhibit a synchronously increasing trend in forecast uncertainty, the system initiates a regional linkage sampling mechanism. This mechanism automatically adjusts the sampling frequency of all affected turbines within the target area to above the enhanced frequency and enters the area cascade test state.

[0062] During the test control process, the system synchronously calls the output of the current twin prediction model to calculate the probability ranking of anomalies in each acquisition channel over the next period of time. Based on the ranking results, the system generates a priority queue for the next round of acquisition, prioritizing resource allocation for predicted high-risk channels. The system also uses the task scheduler to update parameters, switch sampling frequencies, and refresh compression strategy configurations.

[0063] The entire test execution control process is completed using a non-blocking instruction chain to avoid data flow interruption due to sampling frequency switching or compression calls, ensuring the continuity and integrity of various high-frequency data under dynamic conditions, and ultimately forming a closed-loop control process.

[0064] It should be noted that S3 achieves dynamic adjustment of sampling frequency and high-sensitivity test triggering by predicting state level judgment driven by uncertainty. It combines abnormal trend identification, redundant channel parallel acquisition and regional linkage mechanism to build a closed-loop test control process with global response, improve fault perception sensitivity and data reliability, and is superior to traditional fixed-frequency acquisition methods.

[0065] Example 2, reference Figure 2 , which is an embodiment of the present invention, provides a wind farm intelligent simulation acquisition and testing system, including a model building module 100, a calculation and prediction uncertainty module 200, and a dynamic adjustment module 300.

[0066] S4: Constructing a model module 100 for obtaining wind turbine operating status parameters and constructing a twin prediction model based on time series.

[0067] The model building module 100 includes a data collection submodule 101 and a model building submodule 102 .

[0068] Furthermore, the data acquisition submodule 101 is used to collect operating status parameters from the target wind turbine in the wind farm in real time. The parameters include wind speed, rotor speed, active power, voltage, current, vibration value and temperature and other information. The sampling period is 10 seconds and is organized into a time series sample set through a sliding time window.

[0069] Model construction submodule 102 is used to input time series samples into a time series prediction network based on a long short-term memory architecture to construct a twin prediction model for each wind turbine. The model structure includes an input layer, two layers of LSTM units, and a fully connected output layer. It outputs predicted operating state parameters and their confidence intervals for multiple future time steps. The model supports hourly rolling window retraining and outputs a state confidence index after each training round to evaluate the reliability of the model's predictions.

[0070] It should be noted that the data acquisition submodule 101 is the starting point for constructing the model module 100, and the model construction submodule 102 is the core unit for realizing the construction of the prediction model based on the collected data. The model construction module 100 provides the prediction capability foundation for the entire system.

[0071] S5: The prediction uncertainty calculation module 200 is used to compare the model prediction value with the measured value to calculate the operating state error and estimate the prediction uncertainty.

[0072] The prediction uncertainty calculation module 200 includes an error analysis submodule 201 and an uncertainty calculation submodule 202 .

[0073] Furthermore, the error analysis submodule 201 is used to compare the predicted value output by the model building submodule 102 with the measured value collected in real time, calculate the error vector normalized by the Euclidean norm, and construct the error time series of each dimensional operating state of the wind turbine.

[0074] Uncertainty calculation submodule 202 calculates the error standard deviation and rate of change based on a sliding time window and uses a weighted combination to form a prediction uncertainty evaluation value. The system uses this evaluation value to classify the current state of the wind turbine into low-risk, medium-risk, and high-risk levels, which serve as the logical input for subsequent sampling control.

[0075] It should be noted that the error analysis submodule 201 is a pre-step of the calculation prediction uncertainty module 200, and the uncertainty calculation submodule 202 is the key function of the module, responsible for establishing a quantitative mapping mechanism from error to sampling behavior.

[0076] S6: The dynamic adjustment module 300 is used to dynamically adjust the sampling frequency of the sampling channel according to the prediction uncertainty and control the test execution.

[0077] The dynamic adjustment module 300 includes a frequency control submodule 301 and a test control submodule 302 .

[0078] Furthermore, the frequency control submodule 301 is used to assign three sampling frequencies to the sampling channels based on the predicted uncertainty level: base frequency (60 seconds), enhanced frequency (10 seconds), and high-frequency acquisition (1 second). At high risk levels, the system automatically activates redundant channels to form multi-channel parallel sampling.

[0079] The test control submodule 302 is used to identify situations where the prediction uncertainty evaluation value continuously exceeds the threshold, trigger the high-sensitivity test process, perform instant compression storage on the sampled data, and start the regional test linkage mechanism when an abnormal trend of synchronization of adjacent units is detected, execute regional-level distributed high-frequency acquisition tasks, and update the priority of the next round of sampling.

[0080] It should be noted that the frequency regulation submodule 301 implements a prediction-driven frequency adaptation mechanism, and the test control submodule 302 builds a wind farm cascade response capability, which is a reflection of the intelligent regulation capability of the system.

[0081] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0082] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0083] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0084] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logical functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc. It should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced with equivalents without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications should be encompassed by the claims of the present invention.

Claims

1. A wind farm intelligent simulation acquisition test method, characterized in that: include: Obtain wind turbine operating status parameters and build a twin prediction model based on time series; Compare model predictions with measured values ​​to calculate operational state errors and estimate prediction uncertainty; Dynamically adjust the sampling frequency of the sampling channel according to the prediction uncertainty and control the test execution; Comparing the model predicted value with the measured value to calculate the operating state error includes calculating a difference sequence between the predicted value and the measured value for each type of wind turbine operating state parameter at each time step, and performing Euclidean norm normalization on the difference sequence to generate an error sequence vector for error evaluation; The running state error vector is constructed as: Among them, ε m (t s ) indicates that the wind turbine numbered m is at time step t s The error evaluation value of Indicates the nth state parameter of wind turbine m at time step t s The predicted value of is calculated by the LSTM network in S1; m,n (ts) represents the nth state parameter of wind turbine m at time step t s The measured value of; D represents the total number of operating status parameter dimensions; At the same time, a lag matching operation within a preset time range is introduced into the error calculation, and sliding statistics of errors in continuous time periods are performed to construct a fluctuation function that represents the error volatility. The function value is used as the basic indicator for uncertainty assessment. The wave function is expressed as: Among them, θ m (t s ) represents wind turbine m at time step t s The error fluctuation function value; H represents the number of time steps of the sliding window; μ m (t s ) represents the mean error of wind turbine m within the sliding window, which can be expressed as: Among them, r represents the relative time step offset within the sliding time window, and its value range is 0~H; ε m (t s -r) indicates that the wind turbine number is m at time step t s -r error evaluation value; the estimated prediction uncertainty includes, based on the error sequence of each wind turbine generator set, selecting a sliding time window and extracting the error standard deviation within the window, evaluating the time change rate of the error, and forming a prediction uncertainty evaluation value by weighted combination of the standard deviation value and the error change rate; using empirical parameters to determine the weight value in the weight combination, and setting upper and lower limit thresholds to determine the current state level, and different state levels trigger the sampling frequency strategy switching mechanism.

2. The wind farm intelligent simulation acquisition and testing method according to claim 1, characterized in that: The obtaining of wind turbine operating status parameters includes: The operating parameters of the target wind turbines are collected in real time, and a time series sample set is generated according to a fixed sampling period. A feature input data set is established for each wind turbine.

3. The wind farm intelligent simulation acquisition and testing method according to claim 2, characterized in that: The construction of the twin prediction model based on time series includes: A multi-time-step input sequence is constructed using a sliding time window approach, and the operating parameters are input into the time series prediction network. The prediction network uses a recurrent neural network based on a long-short-term memory structure. The model structure supports historical window updates and prediction residual correction. In each update round, a rolling window training mode is used to adjust the model parameters to adapt to the operating characteristics of different wind turbines and output predictions of the operating status of the future target time step. A model copy is deployed for each wind turbine unit, and the copy model automatically triggers the parameter self-update mechanism on an hourly basis. Each update uses the measured data as the training set, and the model error is calibrated using the sliding average and minimum mean square error strategies. The predicted output includes the future value of each monitored variable and the corresponding confidence interval, and a state confidence indicator is added to the model output.

4. The wind farm intelligent simulation acquisition and testing method according to any one of claim 3, characterized in that: The estimated forecast uncertainty includes, Based on the error sequence vector generated by each wind turbine, a sliding time window is defined and the standard deviation of the prediction error of all time steps within the window is extracted. The trend of the error mean change in consecutive time steps is calculated to obtain the error change rate value. The prediction error standard deviation and the error change rate value are weighted to form a prediction uncertainty evaluation value of the credibility of the current prediction state; Based on the prediction uncertainty evaluation value and the preset upper and lower thresholds, the current status of the wind turbine is divided into three status levels: low risk, medium risk and high risk; When the status level changes, the sampling frequency adjustment action is automatically triggered.

5. The wind farm intelligent simulation acquisition and testing method according to claim 4, characterized in that: The dynamic adjustment of the sampling frequency of the sampling channel according to the prediction uncertainty includes: Assign a sampling frequency category to each channel according to the current status level of the wind turbine, which is divided into three levels: basic frequency, enhanced frequency and high frequency acquisition; In the control logic, if the current status level corresponding to the current channel is at the lowest level, the basic sampling frequency is executed; if the indicator is at the medium level, the sampling frequency is increased to the enhanced frequency; if the indicator is at the highest level, the sampling frequency is adjusted to the high-frequency mode, and an auxiliary redundant acquisition channel is added.

6. The wind farm intelligent simulation acquisition and testing method according to any one of claim 5, characterized in that: The control test execution includes, When the prediction uncertainty evaluation value of a certain wind turbine is identified to continuously exceed the normal range, an automatic test command is triggered, the current sampling channel is placed in a high-sensitivity monitoring state, and the sampling data within the specified time range is synchronously compressed and stored; At the same time, the operating status of adjacent units in the current wind farm area is coordinated and checked. If there is a synchronization deviation trend between adjacent units, the regional test linkage mechanism is automatically started to execute distributed high-frequency collection tasks within the area, and the next round of collection priority sequence is generated in combination with the current prediction model output.

7. A wind farm intelligent simulation acquisition test system, characterized by: It includes a model building module (100), a prediction uncertainty calculation module (200), and a dynamic adjustment module (300); The model building module (100) is used to obtain wind turbine operating state parameters and build a twin prediction model based on time series; The prediction uncertainty calculation module (200) is used to compare the model prediction value with the measured value to calculate the operating state error and estimate the prediction uncertainty; The dynamic adjustment module (300) is used to dynamically adjust the sampling frequency of the sampling channel according to the prediction uncertainty and control the test execution.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the wind farm intelligent simulation acquisition and testing method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind farm intelligent simulation acquisition and testing method according to any one of claims 1 to 6 are implemented.

Images