Monitoring management method and system based on wind power plant

Abstract

The invention discloses a monitoring management method and system based on a wind power plant, and relates to the technical field of computer platform load balancing, and the method comprises the steps: installing a sensor, and collecting the operation data of the wind power plant; cleaning and filtering the data, and extracting state features by using a time-frequency domain analysis technology; establishing a baseline state model of the equipment in a normal operation period, and monitoring the operation condition of the wind power plant equipment; and in combination with a machine learning algorithm, predicting the remaining service life of the key part. According to the method, full space-time monitoring and intelligent analysis of the wind power plant operation data are realized, and the equipment operation efficiency and the overall reliability of the wind power plant are improved. The invention belongs to the field of monitoring and management of a wind power generation system, and provides an efficient and stable technical means for realizing intelligent operation of a wind power plant.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power management, and particularly to a monitoring and management method and system based on a wind farm. Background Art

[0002] With the continuous deepening of the global utilization of renewable energy, wind power generation, as a clean and sustainable energy form, has developed rapidly worldwide. The scale of modern wind farms has gradually expanded from a single wind turbine generator to large-scale centralized wind farms, and even formed a multi-region linked wind power generation network. In order to achieve efficient and safe operation of wind farms, wind farm monitoring and management technologies have been widely studied and applied. Traditional wind farm monitoring mainly relies on the SCADA (Supervisory Control and Data Acquisition) system to monitor the status and give basic fault alarms by collecting the operation data of wind turbines. However, with the increase in the scale and complexity of wind farms, the traditional SCADA system has gradually exposed problems such as insufficient capabilities in data integration, fault prediction, and intelligent scheduling support. In recent years, a new generation of monitoring systems combining Internet of Things, artificial intelligence, and big data technologies has gradually emerged, aiming to achieve intelligent, real-time, and refined management of wind farm operations.

[0003] Although there have been many studies and technological applications in the field of wind farm monitoring and management, the current technologies still have many deficiencies. First, traditional monitoring methods mostly rely on threshold alarm methods based on fixed rules, which cannot predict potential faults of key components in advance, easily causing sudden shutdowns and high maintenance costs. Second, most existing monitoring systems rely on a centralized data processing architecture, and when facing large-scale wind farms, the processing efficiency is limited and it is difficult to meet real-time requirements. In addition, the scheduling and management problems in the coordinated operation of cross-regional wind farms have not been fully solved, especially in the coordinated control between grid fluctuations and unstable wind power generation, there is still a lack of effective means. Existing technologies also lack a comprehensive perception and optimization of the operating states of various devices in the wind farm, and cannot achieve full-life cycle intelligent management from single devices to the entire wind farm. Summary of the Invention

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

[0005] Therefore, the technical problems solved by the present invention are: the problems of insufficient fault prediction of wind power equipment, limited real-time monitoring ability, and insufficient cross-regional collaborative management ability existing in the prior art.

[0006] To solve the above technical problems, the present invention provides the following technical solution: A monitoring and management method based on a wind farm, including

[0007] Install sensors to collect the operation data of the wind farm;

[0008] Clean and filter the data, and extract the state features by using time-frequency domain analysis technology;

[0009] Establish a baseline state model of the equipment during normal operation to monitor the operation of the wind farm equipment;

[0010] Combine machine learning algorithms to predict the remaining service life of key components.

[0011] As a preferred solution of the monitoring and management method based on the wind farm according to the present invention, wherein: the installation of sensors includes determining the equipment components and operating parameters to be monitored, and installing vibration, temperature, current and voltage sensors on key components such as gearbox bearings, generator stators, and wind turbine blades; ensuring the accurate position of the sensors, reducing external noise interference, using the SCADA system to collect sensor data in real time, and setting the sampling frequency to more than 1 kHz to capture high-frequency vibration signals;

[0012] Data acquisition formula, collecting wind turbine and environmental data through sensors, wind speed v, power generation P

[0013]

[0014] Among them, η represents the conversion efficiency, ρ represents the air density, and A represents the blade swept area; the data sampling period formula is expressed as:

[0015]

[0016] Among them, f s represents the sensor sampling frequency.

[0017] As a preferred solution of the monitoring and management method based on the wind farm according to the present invention, wherein: the cleaning and filtering of the data includes removing redundant data with the same timestamp or extremely small data value changes for frequently collected repeated data packets; performing signal filtering, setting the cut-off frequency, selecting a frequency higher than the signal bandwidth according to the operating frequency range of the equipment, applying a filter to retain low-frequency information while suppressing high-frequency noise, and using a low-pass filter for filtering:

[0018]

[0019] Among them, H(f0 represents the filter transfer function, f c represents the cut-off frequency;

[0020] Use discrete wavelet transform to decompose the signal, select a suitable mother wavelet to perform multi-layer decomposition on the signal, threshold the high-frequency components, and reconstruct the decomposed signal to generate a denoised signal.

[0021] As a preferred solution of the monitoring and management method based on a wind farm according to the present invention, wherein: the extraction of state features by using time-frequency domain analysis technology includes converting a time-domain signal into a frequency domain, analyzing the intensity of different frequency components in the signal to identify possible faults, and performing a fast Fourier transform (FFT);

[0022] The formula for the mean value of feature extraction features is expressed as:

[0023]

[0024] The variance formula is expressed as:

[0025]

[0026] Power spectral density (PSD):

[0027] P(f) = |X(f)| 2

[0028] wherein, X(f) represents the spectrum of the signal, and the short-time Fourier transform:

[0029]

[0030] wherein, w(t) represents the window function, and x(t) represents the time signal.

[0031] As a preferred solution of the monitoring and management method based on a wind farm according to the present invention, wherein: the establishment of the baseline state model of the equipment includes constructing a baseline model, normal distribution hypothesis or adaptive baseline;

[0032] Perform anomaly detection, assuming that the data distribution in the normal state can be represented by the weighted sum of K Gaussian distributions:

[0033]

[0034] wherein, π k represents the weight of the k-th Gaussian component, satisfying μ k represents the mean vector, and Σ k represents the covariance matrix;

[0035] Construct an autoencoder, and the encoder maps the input features to the latent variable space:

[0036] h = f(x) = σ(Wx + b)

[0037] wherein, f(x) represents the encoder function, W represents the encoder weight, and b represents the encoder bias; the decoder maps the latent variable back to the original space:

[0038]

[0039] Among them, g represents the decoder function, W' represents the decoder weight, and b' represents the decoder bias; the loss function is the reconstruction error:

[0040]

[0041] Among them, L represents the model reconstruction loss, and x i represents the original input data, represents the reconstructed output data; the model is trained using normal operating data, and the goal is to minimize the reconstruction error loss function;

[0042] The real-time data X real ={x 1,real , x 2,real ,...} is input into the baseline model for comparison detection, and the probability model detection is expressed as:

[0043]

[0044] Anomaly determination, ∈ represents the reconstruction error detection threshold; when p(x real ) < ∈, the data is considered abnormal; the reconstruction error detection formula is expressed as:

[0045]

[0046] Among them, represents the reconstructed output data of the real-time data; when the reconstruction error exceeds the set threshold for anomaly determination, the data is considered abnormal

[0047] As a preferred solution of the monitoring and management method based on a wind farm according to the present invention, wherein: the monitoring of the operation of wind farm equipment includes optimizing the maintenance plan according to the monitoring situation, and the dynamic programming optimization total cost formula is expressed as:

[0048] C total = C m + C d

[0049] Among them, C m represents the maintenance cost, and C d represents the outage loss;

[0050] Maintenance priority, using the risk assessment method, the risk score:

[0051] R = P(failure)·C impact

[0052] Among them, P(failure) represents the failure probability, and C impact represents the failure impact cost.

[0053] Using swarm intelligence algorithms and genetic algorithms to optimize the maintenance order: the objective function is to minimize C total , and the constraint condition is to meet the maintainability of the fan.

[0054] As a preferred solution of the monitoring and management method based on a wind farm according to the present invention, wherein: predicting the remaining service life of the key components includes predicting the remaining useful life RUL, and calculating the remaining life based on a linear degradation model:

[0055]

[0056] wherein, x t represents the current health state, x th represents the set fault threshold, and Δx represents the degradation rate, which can be fitted through historical data;

[0057] Construct a machine learning model, use the operation historical data of the wind farm equipment, which includes health state characteristics, operating conditions, environmental factors, and actual fault times, to train a long short-term memory network LSTM to predict the health state trend, and the state update formula is expressed as:

[0058]

[0059] wherein, c t represents the current memory state, f t , i t represent the forget and input gates, and output the predicted value:

[0060] h t = o t ·tanh(c t )

[0061] wherein, h t represents the predicted output at the current time step; when the predicted remaining useful life RUL is lower than the preset threshold, an alarm signal is triggered to prompt the operation and maintenance team to perform early maintenance; different levels of remaining useful life correspond to different maintenance strategies.

[0062] As a preferred solution of the monitoring and management system based on a wind farm according to the present invention, wherein:

[0063] A data acquisition unit that integrates the data input of multiple sensors;

[0064] A communication module that supports wireless and wired transmissions to ensure real-time transmission to the central control system or the cloud platform;

[0065] A data processing and analysis module that processes the collected data, extracts features, and performs status monitoring and anomaly detection;

[0066] The fault prediction module predicts faults based on machine learning algorithms, formulates optimized maintenance plans, calculates the optimal maintenance strategy through dynamic programming, and obtains the immediate costs of different optimal maintenance strategies.

[0067] A computer device includes a memory and a processor. The memory stores a computer program, and the execution of the computer program by the processor implements the steps of a monitoring and management method based on a wind farm.

[0068] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, it implements the steps of a monitoring and management method based on a wind farm.

[0069] Advantages of the present invention: The monitoring and management method based on a wind farm provided by the present invention presents an innovative solution that integrates Internet of Things sensing technology, a distributed data processing architecture, and an intelligent prediction algorithm. The present invention realizes the full-time and full-space monitoring and intelligent analysis of wind farm operation data, improves the equipment operation efficiency and the overall reliability of the wind farm, and provides an efficient and stable technical means for realizing the intelligent operation of the wind farm. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0071] Figure 1 It is the overall flowchart of a monitoring and management method based on a wind farm provided for the first embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0072] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will give a detailed description of the specific embodiments of the present invention with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0073] Embodiment 1, referring to Figure 1 , which is an embodiment of the present invention, provides a monitoring and management method based on a wind farm, including:

[0074] S1: Install sensors to collect wind farm operation data.

[0075] Further, the installed sensors include determining the equipment components and operating parameters to be monitored, and installing vibration, temperature, current, and voltage sensors on key components such as gearbox bearings, generator stators, and wind turbine blades; ensuring the accurate position of the sensors to reduce external noise interference, and using the SCADA system to collect sensor data in real time. The sampling frequency is set above 1 kHz to capture high-frequency vibration signals;

[0076] Data acquisition formula: Collect wind turbine and environmental data through sensors, including wind speed v and power generation P

[0077]

[0078] where η represents the conversion efficiency, ρ represents the air density, and A represents the blade swept area; the data sampling period formula is expressed as:

[0079]

[0080] where f s represents the sensor sampling frequency.

[0081] The cleaning and filtering of the data include removing redundant data with the same timestamp or extremely small data value changes for frequently collected repeated data packets; performing signal filtering, setting the cut-off frequency, selecting a frequency higher than the signal bandwidth according to the operating frequency range of the equipment, applying a filter to retain low-frequency information while suppressing high-frequency noise, and using a low-pass filter for filtering:

[0082]

[0083] where H(f0 represents the filter transfer function and f c represents the cut-off frequency;

[0084] It should be noted that the discrete wavelet transform is used to decompose the signal, a suitable mother wavelet is selected to perform multi-layer decomposition on the signal, the high-frequency components are thresholded, and the decomposed signal is reconstructed to generate a denoised signal.

[0085] S2: Clean and filter the data, and use time-frequency domain analysis technology to extract state features.

[0086] Using time-frequency domain analysis technology to extract state features includes converting the time-domain signal to the frequency domain, analyzing the intensity of different frequency components in the signal to identify possible faults, and performing a fast Fourier transform FFT;

[0087] The formula for the feature mean of feature extraction is expressed as:

[0088]

[0089] The variance formula is expressed as:

[0090]

[0091] Power Spectral Density PSD:

[0092] P(f) = |X(f)| 2

[0093] where X(f) represents the spectrum of the signal, and the short-time Fourier transform:

[0094]

[0095] where w(t) represents the window function and x(t) represents the time signal.

[0096] S3: Establish a baseline state model of the equipment during normal operation and monitor the operation of the wind farm equipment.

[0097] Collect the operation data of the equipment during normal operation as the input of the baseline state. Long-term data collection: Select the time period when the equipment operates under stable working load and environmental conditions, and collect the time series data of key parameters (such as vibration, temperature, current). Preprocessing: Clean and filter the collected data to remove noise, missing values and outliers to ensure data quality. Feature extraction: Extract the time domain, frequency domain and time-frequency domain features, such as mean value, peak value, spectral amplitude, etc.

[0098] Reconstruction error, calculate the probability p(X) of the real-time data X in the Gaussian mixture model.

[0099] If the probability value is lower than a certain threshold (such as p(X) < 0.05), it is determined as abnormal.

[0100] Reconstruct the real-time data through an autoencoder and calculate the reconstruction error If the error exceeds the threshold (such as the 95% confidence interval), it is determined as abnormal.

[0101] Combine multiple features, calculate a health score (such as a normalized value between 0 and 1), and set a score threshold to construct a baseline model, normal distribution hypothesis or adaptive baseline. Perform anomaly detection, assuming that the data distribution in the normal state can be represented by the weighted sum of K Gaussian distributions:

[0102]

[0103] where, π k represents the weight of the k-th Gaussian component, satisfying μ k represents the mean vector, and Σ k represents the covariance matrix. Use the Expectation-Maximization algorithm EM to estimate the model parameters:

[0104]

[0105] Construct an autoencoder. The encoder maps the input features to the latent variable space:

[0106] h = f(x) = σ(Wx + b)

[0107] where f(x) represents the encoder function, W represents the encoder weights, and b represents the encoder bias; the decoder maps the latent variable back to the original space:

[0108]

[0109] where g represents the decoder function, W' represents the decoder weights, and b' represents the decoder bias; the loss function is the reconstruction error:

[0110]

[0111] where L represents the model reconstruction loss, x i represents the original input data, represents the reconstructed output data. The model is trained using normal operating data, and the goal is to minimize the reconstruction error loss function.

[0112] Real-time data X real ={x 1,real , x 2,real ,...} is input into the baseline model for comparison detection. The probability model detection is expressed as:

[0113]

[0114] Anomaly determination, ∈ represents the reconstruction error detection threshold.

[0115] Model training. The model is trained using normal operating data X, and the goal is to minimize the following loss function: The GMM minimizes the negative log-likelihood function:

[0116]

[0117] The autoencoder minimizes the reconstruction error:

[0118]

[0119] When p(x real ) < ∈, the data is considered abnormal. The reconstruction error detection formula is expressed as:

[0120]

[0121] where, Represents the reconstructed output data of real-time data; performs anomaly determination. When the reconstruction error exceeds the set threshold, the data is considered abnormal.

[0122] The monitoring of the operation status of wind farm equipment includes optimizing the maintenance plan according to the monitoring situation. The dynamic programming optimization total cost formula is expressed as:

[0123] C total = C m + C d

[0124] Among them, C m represents the maintenance cost, and C d represents the outage loss.

[0125] Maintenance priority, using a risk assessment method, risk score:

[0126] R = P(failure)·C impact

[0127] Among them, P(failure) represents the failure probability, and C impact represents the failure impact cost.

[0128] Using a swarm intelligence algorithm, a genetic algorithm is used to optimize the maintenance order: the objective function is to minimize C total , and the constraint condition is to meet the maintainability of the fan.

[0129] It should be noted that for rolling window update, the latest normal operation data is used to retrain the model in a sliding window manner. Perform incremental training on the existing model and gradually update the model parameters without completely reconstructing. Regularly verify the detection performance of the model to ensure that the updated model can accurately reflect the latest normal state.

[0130] S4: Combining machine learning algorithms, predict the remaining useful life of key components.

[0131] The prediction of the remaining useful life of key components includes the remaining useful life prediction RUL, and calculates the remaining life based on a linear degradation model:

[0132]

[0133] Among them, x t represents the current health status, x th represents the set failure threshold, and Δx represents the degradation rate, which can be fitted through historical data.

[0134] Build a machine learning model. Using the operation history data of wind farm equipment, which includes health status features, operating conditions, environmental factors, and actual failure times, train a long short-term memory network (LSTM) to predict the health status trend. The state update formula is expressed as:

[0135]

[0136] where c t represents the current memory state, f t , i t represent the forget and input gates, and output the predicted value:

[0137] h t = o t ·tanh(c t )

[0138] where h t represents the predicted output at the current time step; when the predicted remaining useful life (RUL) is lower than a preset threshold, an alarm signal is triggered to prompt the operation and maintenance team to perform early maintenance; different levels of remaining useful life correspond to different maintenance strategies.

[0139] It should be noted that real-time operation data input: Input the real-time operation data of the equipment into the trained model to calculate the current health factor R and the predicted remaining useful life RUL. Regularly update the model, and use the newly collected data to perform incremental training on the model to ensure the accuracy of the prediction. When the equipment operation environment changes (such as load fluctuations, extreme weather), introduce environmental factors to correct the prediction results.

[0140] Example 2, an embodiment of the present invention, provides a monitoring and management method based on a wind farm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0141] In order to verify the effectiveness of a monitoring and management method and system based on a wind farm, 10 wind turbines in a certain wind farm are selected as test objects, and vibration and temperature sensors are installed on key components such as gearbox bearings and generator stators. Data acquisition uses a SCADA system, and the sampling frequency is set to 1 kHz to capture high-frequency signal characteristics. The data is transmitted to the central monitoring system in real time through a fiber optic network.

[0142] First, record data such as wind speed, power generation, and component vibration frequency through sensors.

[0143] In the data cleaning stage, a low-pass filter is used to remove high-frequency noise.

[0144] The cut-off frequency is set to 500 Hz. Subsequently, a fast Fourier transform (FFT) is performed on the vibration signal to extract the main frequency domain features. To capture the time-frequency characteristics of non-stationary signals, a short-time Fourier transform (STFT) is also used.

[0145] The Hanning window function is used to optimize the time-frequency resolution. Then, a device baseline state model is constructed based on normal operation data, adopting a normal distribution hypothesis and an anomaly scoring function. Anomaly detection combines a support vector machine (SVM) to identify the abnormal states of key components.

[0146] Finally, the remaining useful life (RUL) of key components is predicted through a long short-term memory network (LSTM), using a linear degradation model.

[0147] Table 1 Test data table

[0148]

[0149] It can be seen from the tabular data that the fan data in this embodiment has significant differences. In terms of anomaly scoring, the anomaly scores of Fan 1 and Fan 8 are relatively low (1.2 and 0.9), indicating that their operating states are close to the baseline state and the anomaly risk is low; while the scores of Fan 5 and Fan 7 are relatively high (2.5 and 2.8), suggesting that key attention is required. This reflects that the anomaly scoring function can effectively quantify the degree to which the fan state deviates from the baseline.

[0150] Embodiment 3 is an embodiment of the present invention, which provides a monitoring and management system based on a wind farm, where a data acquisition unit integrates the data input of multiple sensors.

[0151] A communication module supports wireless such as LoRa, 5G or wired transmission such as optical fiber to ensure real-time transmission to a central control system or a cloud platform.

[0152] A data processing and analysis module processes the collected data, extracts features, and performs status monitoring and anomaly detection.

[0153] A fault prediction module predicts faults based on machine learning algorithms, formulates an optimized maintenance plan, calculates the optimal maintenance strategy through dynamic programming, and obtains the immediate costs of different optimal maintenance strategies.

[0154] If a 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, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0155] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a predefined sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or used in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0156] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which a program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0157] It should be understood that each part of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc. 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 preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

[0158] 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 preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A monitoring and management method based on a wind farm, characterized in that: include: Install sensors to collect data on wind farm operations; Clean and filter the data, and extract state features using time-frequency domain analysis techniques; Establish a baseline status model for equipment during normal operation to monitor the operation of wind farm equipment; Combined with machine learning algorithms, it predicts the remaining useful life of key components.

2. The wind farm-based monitoring and management method according to claim 1, characterized in that: The installation of sensors includes determining the equipment components and operating parameters that need to be monitored, installing vibration, temperature and current and voltage sensors on key components such as gearbox bearings, generator stators and fan blades; ensuring that the sensor positions are accurate, reducing external noise interference, using the SCADA system to collect sensor data in real time, and setting the sampling frequency to above 1kHz to capture high-frequency vibration signals; Data collection formula, collect wind turbine and environmental data through sensors, wind speed v, power generation P Among them, η represents the conversion efficiency, ρ represents the air density, and A represents the blade sweep area; the data sampling period formula is expressed as: Among them, f s Indicates the sensor sampling frequency.

3. The wind farm-based monitoring and management method according to claim 2, characterized in that: The cleaning and filtering of the data includes, for frequently collected repeated data packets, removing redundant data with the same timestamp or with very small data value changes; performing signal filtering, setting a cutoff frequency, selecting a frequency higher than the signal bandwidth according to the operating frequency range of the device, applying a filter to retain low-frequency information while suppressing high-frequency noise, and filtering using a low-pass filter: Where H(f) represents the filter transfer function, f c represents the cut-off frequency; The signal is decomposed using discrete wavelet transform, and a suitable mother wavelet is selected to perform multi-layer decomposition of the signal. The high-frequency component is thresholded and the decomposed signal is reconstructed to generate a denoised signal.

4. The wind farm-based monitoring and management method according to claim 3, characterized in that: The state feature extraction using time-frequency domain analysis technology includes converting the time domain signal into the frequency domain, analyzing the strength of different frequency components in the signal to identify possible faults, and performing fast Fourier transform FFT; The feature extraction feature mean formula is expressed as: The variance formula is expressed as: Power spectral density PSD: P(f)=|X(f)| 2 Among them, X(f) represents the spectrum of the signal, short-time Fourier transform: Among them, w(t) represents the window function and x(t) represents the time signal.

5. The wind farm-based monitoring and management method according to claim 4, characterized in that: The establishing of the baseline state model of the device includes constructing a baseline model, a normal distribution assumption or an adaptive baseline; For anomaly detection, it is assumed that the data distribution in the normal state can be represented by the weighted sum of K Gaussian distributions: Among them, π k represents the weight of the kth Gaussian component, satisfying μ k represents the mean vector, ∑ k represents the covariance matrix; Construct an autoencoder that maps input features to a latent variable space: h=f(x)=σ(Wx+b) Among them, f(x) represents the encoder function, W represents the encoder weight, and b represents the encoder bias; the decoder maps the latent variables back to the original space: Among them, g represents the decoder function, W′ represents the decoder weight, b′ represents the decoder bias; the loss function is the reconstruction error: Among them, L represents the model reconstruction loss, x i represents the original input data, Represents the reconstructed output data; the model is trained using normal operating data, with the goal of minimizing the reconstruction error loss function; Real-time dataX real ={x 1,real ,x 2,real ,…} Input the baseline model for comparative detection. The probability model detection is expressed as: Abnormal judgment, ∈ represents the reconstruction error detection threshold; when p(x real )<∈, the data is considered abnormal; the reconstruction error detection formula is expressed as: in, Represents the reconstructed output data of real-time data; when the reconstruction error exceeds the set threshold, the data is considered abnormal.

6. The wind farm-based monitoring and management method according to claim 5, characterized in that: The monitoring of the operation of the wind farm equipment includes optimizing the maintenance plan according to the monitoring situation. The dynamic programming optimization total cost formula is expressed as: C total =C m +C d Among them, C m represents the maintenance cost, C d Indicates downtime loss; Maintain priorities, using risk assessment methods, risk scoring: R=P(failure)·C impact Among them, P (failure) represents the failure probability, C impact Indicates the failure impact cost. Use swarm intelligence algorithm and genetic algorithm to optimize the maintenance order: the objective function is to minimize C total , the constraint condition is to meet the maintainability of the fan.

7. The wind farm-based monitoring and management method according to claim 6, characterized in that: The prediction of the remaining useful life of key components includes a remaining useful life prediction RUL, which is calculated based on a linear degradation model: Among them, x t Indicates the current health status, x th represents the set fault threshold, Δx represents the degradation rate, which can be fitted by historical data; A machine learning model is constructed using the historical operation data of wind farm equipment, which includes health status characteristics, operating conditions, environmental factors, and actual failure time. The long short-term memory network LSTM is trained to predict the health status trend. The status update formula is expressed as: Among them, c t Indicates the current memory state, f t ,i t Represents forget and input gates, outputting predicted values: h t =o t ·tanh(c t ) Among them, h t Represents the predicted output of the current time step; when the predicted remaining life RUL is lower than the preset threshold, an alarm signal is triggered to prompt the operation and maintenance team to perform maintenance in advance; different levels of remaining life correspond to different maintenance strategies.

8. A system using the wind farm monitoring and management method according to any one of claims 1 to 7, characterized in that: Data acquisition unit, integrating data input from multiple sensors; Communication module, supporting wireless and wired transmission, ensuring real-time transmission to the central control system or cloud platform; The data processing and analysis module processes the collected data, extracts features, and performs status monitoring and anomaly detection; The fault prediction module predicts faults based on machine learning algorithms, formulates optimized maintenance plans, calculates the optimal maintenance strategy through dynamic planning, and obtains the optimal instant cost of different maintenance strategies.

9. 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-based monitoring and management method according to any one of claims 1 to 7 are implemented.

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

Images