Wind power plant intelligent management system based on Internet of Things

Abstract

The invention discloses a wind power plant intelligent management system based on the Internet of Things, relates to the technical field of wind power prediction, and improves the generalization ability of wind power prediction through deep learning and multi-source data fusion. Performing feature extraction on the environmental data by adopting a long-short term memory network and convolutional neural network combined model, and optimizing the time sequence weight in combination with historical wind power data to reduce the dependence on historical data; in order to solve the problem that short-term wind speed fluctuation is difficult to capture, a high-frequency wind speed data acquisition and quick response prediction module is designed, high-frequency wind speed data is acquired by using laser radar wind speed measurement equipment and an ultrasonic anemometer, short-term change characteristics are extracted in combination with wavelet transform, and short-term and long-term information is dynamically fused by using a double-flow deep neural network. And the prediction precision of wind speed change in a short time is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power prediction, and specifically to an intelligent management system for a wind farm based on the Internet of Things. Background Art

[0002] The intelligent management system for a wind farm stems from the development of smart grid and Internet of Things technologies. In the early days, wind farms mainly relied on manual inspections and regular maintenance, making it difficult to achieve precise monitoring and prediction. With the development of sensor technology, cloud computing, big data analysis, and artificial intelligence, the management of wind farms has gradually evolved towards automation and intelligence. In recent years, combined with edge computing and 5G communication, the intelligent management system can achieve remote real-time monitoring, fault prediction, and optimal scheduling, improving the operating efficiency and economy of wind farms.

[0003] However, when the existing intelligent management systems for wind farms are applied to wind power prediction and optimal scheduling, they often have the following technical drawbacks: 1. Strong dependence on historical data: Many prediction models rely on historical power generation data. However, the wind farm environment changes rapidly, which may lead to insufficient generalization ability of the model and reduce prediction accuracy.

[0004] 2. Difficulty in capturing short-term fluctuations: Traditional methods are difficult to accurately depict the instantaneous changes in wind speed, resulting in large prediction errors in power fluctuations within a short period of time and affecting the stability of the power grid.

[0005] 3. Challenges in grid connection security: Wind power fluctuations may cause grid instability, and the existing optimal scheduling strategies are still insufficient in adapting to extreme weather or emergencies, affecting the security and economy of wind power grid connection. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides an intelligent management system for a wind farm based on the Internet of Things, which solves the technical drawbacks mentioned in the background art.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: An intelligent management system for a wind farm based on the Internet of Things, including a data perception module, a dynamic correction module, an instantaneous modeling module, a security assessment module, and an optimal scheduling module; The data perception module uses Internet of Things terminals to collect wind speed, wind direction, turbulence intensity, and unit vibration data in real time, constructs a spatio-temporal data set, and calculates and evaluates the environmental mutation index Hd to trigger a dynamic correction instruction; After receiving the dynamic correction instruction, the dynamic correction module uses the sliding time window algorithm to perform timeliness screening on historical power generation data, eliminates invalid data segments, generates a dynamic training set, and updates the prediction model parameters in combination with online learning; The instantaneous modeling module extracts turbulence characteristics based on high-frequency sampling data, uses a hybrid model to predict instantaneous power fluctuations and outputs the instantaneous power prediction results; then calculates the fluctuation entropy value Bd; finally, evaluates the fluctuation entropy value Bd to activate the safety warning. The safety assessment module is used to monitor the grid frequency deviation, voltage volatility and line load rate, calculate the grid connection risk coefficient Rx in combination with the fluctuation entropy value Bd, and evaluate the system robustness under extreme weather through Monte Carlo simulation to obtain the safety margin Sy. The optimization scheduling module dynamically adjusts the unit output according to the instantaneous power prediction results, grid connection risk coefficient Rx and safety margin Sy, and adopts a virtual inertia compensation strategy to provide inertial response during power mutation.

[0008] Preferably, the data perception module includes a collection unit, a fusion calculation unit and a trigger unit. The collection unit is used to collect multi-dimensional dynamic data of the wind farm in real time through the Internet of Things terminal, generate a spatio-temporal data set, and transmit it to the central processing platform in real time. By performing spatio-temporal alignment and standardization processing on the wind speed, wind direction, turbulence intensity and unit vibration data, an environmental feature database is constructed. Among them, the spatio-temporal data set includes the sequence variation coefficient Sx of the real-time wind speed, the turbulence intensity distribution matrix Tij, the unit vibration spectrum Vk and the environmental temperature and humidity gradient ΔH.

[0009] Preferably, the fusion calculation unit is used to perform multi-scale fusion calculation on the spatio-temporal data set by taking the dispersion of the sequence variation coefficient Sx of the real-time wind speed, the dispersion of the turbulence intensity distribution matrix Tij and the energy entropy of the unit vibration spectrum Vk to obtain the environmental mutation index Hd. The specific calculation formula is as follows: ; In the formula, represents the dispersion of the turbulence intensity distribution matrix Tij, represents the energy entropy of the unit vibration spectrum Vk; w1, w2 and w3 respectively correspond to the contribution weights of the sequence variation coefficient Sx of the real-time wind speed, the dispersion of the turbulence intensity distribution matrix Tij and the energy entropy of the unit vibration spectrum Vk, and w1 + w2 + w3 = 1.

[0010] Preferably, the trigger unit is used to preset an adaptive threshold J and compare and evaluate it with the environmental mutation index Hd to determine whether it is necessary to start the model correction mechanism. The specific determination logic is as follows: If the environmental mutation index Hd ≥ the adaptive threshold J, it means that the current environmental mutation degree exceeds the model adaptation range, generates a dynamic correction instruction and transmits it to the dynamic correction module; If the environmental mutation index Hd < the adaptive threshold J, it means that the current environmental state is within the prediction model tolerance, and the system default operation mode is maintained.

[0011] Preferably, after receiving the dynamic correction instruction, the dynamic correction module uses a sliding time window algorithm to perform timeliness screening on historical power generation data, dynamically eliminates invalid data segments beyond the time window range, and generates a dynamic training set based on the screened valid data; then, in combination with the online learning mechanism, the dynamic training set is used to update the parameters of the prediction model in real time to dynamically adjust the prediction accuracy and adaptability of the model.

[0012] Preferably, the instantaneous modeling module includes a feature prediction unit and an entropy value evaluation unit; The feature prediction unit extracts turbulence features based on high-frequency sampling data, including the instantaneous wind speed fluctuation amplitude, spectral distribution characteristics, and spatial correlation coefficient; uses a hybrid model to model the turbulence features, predicts the instantaneous power fluctuation, and outputs the instantaneous power prediction result; A hybrid modeling method combining a physical model and data-driven is adopted, and the specific prediction formula is as follows:

[0013] In the formula, Ppred represents the predicted value of instantaneous power, Pphy represents the prediction component of the physical model based on hydrodynamic calculation, v represents the wind speed, ρ represents the air density, Pdata represents the prediction component of the data-driven model based on the turbulence feature Xturb, α and β are the weight coefficients of the physical model and the data-driven model respectively, and ϵ is the correction error term.

[0014] Preferably, the entropy value evaluation unit is used to calculate the fluctuation entropy value Bd according to the predicted value Ppred of instantaneous power, and evaluate the fluctuation entropy value Bd to activate the safety warning mechanism; among them, the information entropy algorithm is used to calculate the fluctuation entropy value Bd, and the specific calculation formula is as follows: ; In the formula, represents the probability distribution of the instantaneous power fluctuation in the i-th amplitude interval; The preset first safety threshold Bdth-1 and the second safety threshold Bdth-2 are compared with the fluctuation entropy value Bd for evaluation to activate the safety warning; and the first safety threshold Bdth-1 is greater than the second safety threshold Bdth-2, and the specific evaluation content is as follows: When the fluctuation entropy value Bd ≥ the first safety threshold Bdth-1, it is determined as the first abnormal fluctuation, triggering a first-level safety warning, and the following operations are performed: Immediately start the emergency load reduction control strategy of the fan, limit the output power to the safe range; if the fluctuation continues to intensify, trigger the forced shutdown protection; synchronously send a high-level alarm signal to the monitoring system to prompt manual intervention; When the second safety threshold Bdth-2 ≤ fluctuation entropy value Bd < the first safety threshold Bdth-1, it is determined as the second abnormal fluctuation, triggering a secondary safety warning, and the following operations are performed: Automatically adjust the dynamic damping parameter or pitch rate of the fan to suppress power oscillation; Real-time optimize the prediction model parameters to enhance the adaptability to short-term fluctuations; Generate a warning log and mark the abnormal period for subsequent analysis; When the fluctuation entropy value Bd < the second safety threshold Bdth-2, it is determined as the normal fluctuation range, maintain the current operating state of the fan, and continuously monitor the change trend of the entropy value.

[0015] Preferably, the safety assessment module includes a real-time monitoring unit and a risk simulation unit; The real-time monitoring unit is used to continuously collect the grid frequency deviation Ra, voltage volatility Rb, and line load rate Rc and perform normalization processing, and combine with the fluctuation entropy value Bd output by the instantaneous modeling module to fit and calculate the grid connection risk coefficient Rx; The specific calculation formula of the grid connection risk coefficient Rx is: ; In the formula, all parameters participate in the calculation with equal weights. When any parameter increases abnormally, the grid connection risk coefficient Rx will rise, reflecting the risk accumulation effect.

[0016] Preferably, the risk simulation unit is used to simulate the dynamic response of the power grid under fault conditions by generating the wind speed-load joint probability distribution under extreme weather scenarios based on the Monte Carlo method, and output the system safety margin Sy; when performing the Monte Carlo simulation, the wind speed-load joint probability density function f(v,P) is used as the input, and the system instability probability Ps is calculated through N random samplings. The safety margin Sy is defined as: ; Where N ≥ 10000 times, and the action timing constraints of the protection device are considered in each simulation.

[0017] Preferably, the power prediction result output by the instantaneous modeling module, the grid connection risk coefficient Rx and the safety margin Sy calculated by the safety assessment module are received in real time, and the unit output command is generated through the dynamic optimization algorithm; when a power mutation is detected, the virtual inertia compensation strategy is activated, the virtual inertia coefficient Kv is dynamically adjusted according to the grid frequency change rate dt, and the inertial power compensation is realized through additional torque control. At the same time, the output adjustment amplitude is constrained by the coupling relationship between the grid connection risk coefficient Rx and the safety margin Sy to form the final control signal.

[0018] The present invention provides a wind farm intelligent management system based on the Internet of Things. It has the following beneficial effects: (1) The intelligent management system for wind farms based on the Internet of Things improves the generalization ability of wind power prediction by introducing deep learning methods and multi-source data fusion mechanisms. Specifically, a model combining a long short-term memory network and a convolutional neural network is used to perform deep feature extraction on environmental feature data including wind speed Fsd, temperature Wdj, humidity Sds, and air pressure Qyl, and a model is built in combination with historical wind power data (Fdl). At the same time, an attention mechanism is used to optimize the time series weight allocation, reduce the over-reliance on historical data, and improve the adaptability to environmental changes, thus solving the problem of strong dependence on historical data in traditional methods. (2) The intelligent management system for wind farms based on the Internet of Things designs a high-frequency wind speed data acquisition and fast response prediction module to solve the problem of difficult capture of short-term wind speed fluctuations. First, a lidar wind speed measurement device and an ultrasonic anemometer are used to collect high-frequency wind speed data, and the wavelet transform is combined to decompose the wind speed signal to extract short-term change characteristics. Second, a dual-stream deep neural network is used to process long-term trend and short-term fluctuation information respectively, and the prediction weight is adaptively adjusted through a dynamic fusion mechanism to improve the sensitivity to wind speed changes within a short time, thereby reducing the short-term power prediction error. (3) The intelligent management system for wind farms based on the Internet of Things combines wind power optimization scheduling and extreme weather response mechanisms in enhancing the safety of wind power grid connection. First, a wind farm power scheduling model is constructed based on an adaptive reinforcement learning algorithm, with parameters such as current wind power Fdl, grid load Wfh, reserve power capacity Bdr, and weather anomaly index Tyyz as inputs to optimize the active power output. Second, an extreme weather recognition network is used to classify abnormal meteorological events such as sudden storms, lightning, and low-temperature icing, and the dispatching system is linked to adjust the wind power output curve to ensure the stability and economy of wind power grid connection. Brief Description of the Drawings

[0019] Figure 1 It is a schematic diagram of the framework structure of the intelligent management system for wind farms based on the Internet of Things of the present invention. Detailed Embodiments

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] Embodiment 1

[0022] Please refer to Figure 1, the present invention provides an intelligent management system for a wind farm based on the Internet of Things, including a data perception module, a dynamic correction module, an instantaneous modeling module, a safety assessment module, and an optimal scheduling module; The data perception module collects wind speed, wind direction, turbulence intensity, and unit vibration data in real time through Internet of Things terminals, constructs a spatio-temporal data set, calculates and evaluates the environmental mutation index Hd, and triggers a dynamic correction instruction; After receiving the dynamic correction instruction, the dynamic correction module uses a sliding time window algorithm to perform timeliness screening on historical power generation data, eliminates invalid data segments, generates a dynamic training set, and updates the prediction model parameters in combination with online learning; The instantaneous modeling module extracts turbulence characteristics based on high-frequency sampling data, uses a hybrid model to predict instantaneous power fluctuations and outputs an instantaneous power prediction result; then calculates the fluctuation entropy value Bd; finally evaluates the fluctuation entropy value Bd to activate a safety warning; The safety assessment module is used to monitor the grid frequency deviation, voltage volatility, and line load rate, calculate the grid connection risk coefficient Rx in combination with the fluctuation entropy value Bd, and evaluate the system robustness under extreme weather through Monte Carlo simulation to obtain the safety margin Sy; the optimal scheduling module dynamically adjusts the unit output according to the instantaneous power prediction result, the grid connection risk coefficient Rx, and the safety margin Sy, and adopts a virtual inertia compensation strategy to provide inertial response during power mutation.

[0023] In this embodiment, the stability and safety of wind power grid connection are improved through the collaborative work of each module; the data perception module collects wind speed Wf, wind direction Wfx, turbulence intensity Wt, and unit vibration data Vd in real time through Internet of Things terminals, constructs a spatio-temporal data set and calculates and evaluates the environmental mutation index Hd to trigger a dynamic correction instruction to ensure the timeliness and reliability of the data; After receiving the dynamic correction instruction, the dynamic correction module uses a sliding time window algorithm to perform timeliness screening on historical power generation data, eliminates invalid data segments, generates a dynamic training set, and updates the prediction model parameters in combination with online learning, thereby enhancing the adaptability and generalization ability of the model; The instantaneous modeling module extracts turbulence characteristics based on high-frequency sampling data, uses a hybrid model to predict instantaneous power fluctuations and outputs an instantaneous power prediction result Pg, then calculates the fluctuation entropy value Bd, evaluates the fluctuation entropy value Bd to activate a safety warning, and realizes accurate prediction and control of short-term power fluctuations; The safety assessment module is used to monitor the grid frequency deviation Fv, voltage volatility Vw, and line load rate Ll, calculate the grid connection risk coefficient Rx in combination with the fluctuation entropy value Bd, and at the same time evaluate the system robustness under extreme weather through Monte Carlo simulation to obtain the safety margin Sy to ensure the grid connection stability of the wind farm in a complex environment; The optimization scheduling module dynamically adjusts the output of the unit according to the instantaneous power prediction result Pg, the grid connection risk coefficient Rx, and the safety margin Sy, and adopts a virtual inertia compensation strategy to provide inertial response during power mutation, improving the adaptability and regulation ability of the wind farm in the power grid.

[0024] Embodiment 2

[0025] The data perception module includes a collection unit, a fusion calculation unit, and a trigger unit; The collection unit is used to collect multi-dimensional dynamic data of the wind farm in real time through the Internet of Things terminal, generate a spatio-temporal data set, and transmit it to the central processing platform in real time. By performing spatio-temporal alignment and standardization processing on the wind speed, wind direction, turbulence intensity, and unit vibration data, an environmental feature database is constructed; Among them, the spatio-temporal data set includes the sequence variation coefficient Sx of the real-time wind speed, the turbulence intensity distribution matrix Tij, the unit vibration spectrum Vk, and the environmental temperature and humidity gradient ΔH.

[0026] The fusion calculation unit is used to perform multi-scale fusion calculation on the sequence variation coefficient Sx of the real-time wind speed, the dispersion degree of the turbulence intensity distribution matrix Tij, and the energy entropy of the unit vibration spectrum Vk based on the spatio-temporal data set to obtain the environmental mutation index Hd. The specific calculation formula is as follows: ; In the formula, represents the dispersion degree of the turbulence intensity distribution matrix Tij, represents the energy entropy of the unit vibration spectrum Vk; w1, w2, and w3 respectively correspond to the contribution weights of the sequence variation coefficient Sx of the real-time wind speed, the dispersion degree of the turbulence intensity distribution matrix Tij, and the energy entropy of the unit vibration spectrum Vk, and w1 + w2 + w3 = 1.

[0027] The trigger unit is used to preset an adaptive threshold J and compare and evaluate it with the environmental mutation index Hd to determine whether to start the model correction mechanism. The specific determination logic is as follows: If the environmental mutation index Hd ≥ the adaptive threshold J, it means that the current environmental mutation degree exceeds the model adaptation range, and a dynamic correction instruction is generated and transmitted to the dynamic correction module; If the environmental mutation index Hd < the adaptive threshold J, it means that the current environmental state is within the tolerance of the prediction model, and the system default operation mode is maintained.

[0028] In this embodiment, is the standard deviation of the wind speed sequence within the sliding time window, is the mean value of the wind speed sequence within the sliding time window; , M×N represents the spatial grid division dimension of the wind farm; The global mean value of the turbulence intensity matrix; , where pk represents the proportion of the energy of the k-th frequency band in the total energy; K represents the total number of frequency bands for spectrum analysis; The data perception module undertakes the functions of real-time perception, fusion calculation, and early warning triggering of multi-dimensional dynamic data of the wind farm in the system, ensuring that environmental changes can be captured in a timely manner and fed back to subsequent modules to optimize the adaptability and reliability of the prediction model; the acquisition unit obtains the wind speed Wf, wind direction Wfx, turbulence intensity Wt, and unit vibration data Vd in real time through the Internet of Things terminal, constructs a spatio-temporal data set, and performs spatio-temporal alignment and standardization processing on the data to form an environmental feature database to improve the comparability and calculation efficiency of the data; among them, the spatio-temporal data set includes the sequence variation coefficient Sx of the real-time wind speed, which is used to measure the stability of the wind speed over time and reflect the short-term fluctuation characteristics of wind power output; the turbulence intensity distribution matrix Tij, which is used to describe the spatial distribution and dispersion degree of turbulence in the wind farm and is crucial for the load fluctuation of the wind turbine; the unit vibration spectrum Vk, which is used to analyze the vibration characteristics of the wind turbine under different operating conditions to identify potential mechanical failures or abnormal operating conditions; and the environmental temperature and humidity gradient ΔH, which is used to evaluate the impact of external meteorological conditions on the operating efficiency and material fatigue loss of the fan. The fusion calculation unit performs multi-scale fusion calculation based on the spatio-temporal data set, using the sequence variation coefficient Sx of the real-time wind speed, the dispersion degree of the turbulence intensity distribution matrix Tij, and the energy entropy of the unit vibration spectrum Vk to obtain the environmental mutation index Hd to quantify the environmental dynamic changes of the wind farm. Among them, the dispersion degree of the turbulence intensity distribution matrix Tij measures the inhomogeneity of the turbulence intensity, and the energy entropy of the unit vibration spectrum Vk is used to quantify the complexity of the vibration signal. The combination of the two can accurately describe the operating stability of the wind farm; the trigger unit conducts a comparative evaluation of the environmental mutation index Hd based on the adaptive threshold J, and generates a dynamic correction instruction and transmits it to the dynamic correction module according to the evaluation content to ensure that the prediction model can be adaptively adjusted for abnormal operating conditions.

[0029] Embodiment 3

[0030] After receiving the dynamic correction instruction, the dynamic correction module uses the sliding time window algorithm to perform timeliness screening on the historical power generation data, dynamically eliminates the invalid data segments outside the time window range, and generates a dynamic training set based on the screened valid data; then, in combination with the online learning mechanism, it uses the dynamic training set to update the parameters of the prediction model in real time to dynamically adjust the prediction accuracy and adaptability of the model.

[0031] The instantaneous modeling module includes a feature prediction unit and an entropy value evaluation unit; The feature prediction unit extracts turbulence features based on high-frequency sampling data, including the instantaneous wind speed fluctuation amplitude, spectrum distribution characteristics, and spatial correlation coefficient; uses a hybrid model to model the turbulence features, predicts the instantaneous power fluctuation, and outputs the instantaneous power prediction result; Adopt a hybrid modeling method that combines a physical model and data-driven approach. The specific prediction formula is as follows:

[0032] In the formula, Ppred represents the predicted value of instantaneous power, Pphy represents the predicted component of the physical model based on hydrodynamic calculations, v represents the wind speed, ρ represents the air density, Pdata represents the predicted component of the data-driven model based on the turbulence characteristic Xturb, α and β are the weight coefficients of the physical model and the data-driven model respectively, and ϵ is the correction error term.

[0033] The entropy value evaluation unit is used to calculate the fluctuation entropy value Bd according to the predicted value of instantaneous power Ppred and evaluate the fluctuation entropy value Bd to activate the safety warning mechanism. Among them, the information entropy algorithm is used to calculate the fluctuation entropy value Bd, and the specific calculation formula is as follows: ; In the formula, represents the probability distribution of the instantaneous power fluctuation in the i-th amplitude interval; Preset the first safety threshold Bdth-1 and the second safety threshold Bdth-2, compare and evaluate them with the fluctuation entropy value Bd to activate the safety warning. And the first safety threshold Bdth-1 is greater than the second safety threshold Bdth-2. The specific evaluation content is as follows: When the fluctuation entropy value Bd ≥ the first safety threshold Bdth-1, it is determined as the first abnormal fluctuation, triggering a first-level safety warning, and the following operations are performed: Immediately start the emergency load reduction control strategy of the fan to limit the output power to the safe range; if the fluctuation continues to intensify, trigger the forced shutdown protection; simultaneously send a high-level alarm signal to the monitoring system to prompt manual intervention; When the second safety threshold Bdth-2 ≤ the fluctuation entropy value Bd < the first safety threshold Bdth-1, it is determined as the second abnormal fluctuation, triggering a second-level safety warning, and the following operations are performed: Automatically adjust the dynamic damping parameters or pitch rate of the fan to suppress power oscillation; Real-time optimize the parameters of the prediction model to enhance the adaptability to short-term fluctuations; Generate a warning log and mark the abnormal period for subsequent analysis; When the fluctuation entropy value Bd < the second safety threshold Bdth-2, it is determined as the normal fluctuation range, maintain the current operating state of the fan, and continuously monitor the change trend of the entropy value.

[0034] In this embodiment, after receiving the dynamic correction instruction, the dynamic correction module performs timeliness screening on the historical power generation data through the sliding time window algorithm, eliminates the invalid data segments beyond the time window range, so as to ensure the timeliness and accuracy of the training data, generates a dynamic training set based on the screened valid data, combines with the online learning mechanism to update the parameters of the prediction model in real time, and improves the prediction accuracy and adaptability of the model; the instantaneous modeling module includes a feature prediction unit and an entropy value evaluation unit. Among them, the feature prediction unit extracts the turbulence feature Xturb based on the high-frequency sampling data, including the instantaneous wind speed fluctuation amplitude Xamp, the spectral distribution characteristic Xfreq, and the spatial correlation coefficient Xcorr, so as to accurately describe the instantaneous change characteristics of the wind speed, and uses a hybrid model for modeling. The instantaneous power prediction value Ppred is calculated by combining the physical model prediction component Pphy and the data-driven model prediction component Pdata. Among them, Pphy provides physical constraints based on the hydrodynamic calculation of the wind speed v and the air density ρ, and Pdata uses the change trend of Xturb for data-driven prediction. α and β are the weight coefficients of the two models, and the correction error term ϵ is used to compensate for the prediction error to improve the accuracy and robustness of the instantaneous power prediction; the entropy value evaluation unit calculates the fluctuation entropy value Bd based on Ppred, which is used to quantify the fluctuation degree of the instantaneous power. Among them, Bd is calculated by the information entropy algorithm, which reflects the probability distribution of the power fluctuation in different amplitude intervals, and is compared and evaluated with the first safety threshold Bdth-1 and the second safety threshold Bdth-2 to activate the safety warning mechanism; when Bd≥Bdth-1, a first-level safety warning is triggered, the emergency load reduction control strategy of the fan is started, the output power is limited to the safe range, forced shutdown protection is performed if necessary, and a high-level alarm signal is sent to the monitoring system to prompt manual intervention to ensure the safe operation of the fan; when Bdth-2≤Bd<Bdth-1, a second-level safety warning is triggered, the power oscillation is suppressed by adjusting the dynamic damping parameter or the pitch rate of the fan, and at the same time the parameters of the prediction model are optimized to enhance the short-term fluctuation adaptability, and a warning log is generated to record the abnormal period for subsequent analysis; when Bd<Bdth-2, it is determined to be in the normal fluctuation range, the current operating state of the fan is maintained, and the change trend of the entropy value is continuously monitored to ensure the long-term stable operation of the system.

[0035] Embodiment 4

[0036] The safety assessment module includes a real-time monitoring unit and a risk simulation unit; The real-time monitoring unit is used to continuously collect the grid frequency deviation Ra, the voltage volatility Rb, and the line load rate Rc and perform normalization processing, and combine with the fluctuation entropy value Bd output by the instantaneous modeling module to fit and calculate the grid connection risk coefficient Rx; The specific calculation formula of the grid connection risk coefficient Rx is: ; In the formula, all parameters participate in the calculation with equal weights. When any parameter increases abnormally, the grid connection risk coefficient Rx rises, reflecting the risk accumulation effect.

[0037] The risk simulation unit is used to simulate the dynamic response of the power grid under fault conditions and output the system safety margin Sy by generating the joint probability distribution of wind speed - load under extreme weather scenarios based on the Monte Carlo method; when performing the Monte Carlo simulation, the joint probability density function f(v, P) of wind speed - load is used as the input, and the system instability probability Ps is calculated through N random samplings. The safety margin Sy is defined as: ; where N ≥ 10000 times, and the action timing constraints of the protection device are considered in each simulation.

[0038] It receives in real - time the power prediction results output by the instantaneous modeling module, the grid connection risk coefficient Rx and the safety margin Sy calculated by the safety assessment module, and generates the unit output command through the dynamic optimization algorithm; when a power mutation is detected, the virtual inertia compensation strategy is activated, and the virtual inertia coefficient Kv is dynamically adjusted according to the grid frequency change rate dt, and the inertial power compensation is achieved through additional torque control. At the same time, the coupling relationship between the grid connection risk coefficient Rx and the safety margin Sy is combined to constrain the output adjustment amplitude to form the final control signal.

[0039] In this embodiment, the safety assessment module includes a real - time monitoring unit and a risk simulation unit, which are used to comprehensively evaluate the safety of wind power grid connection; the real - time monitoring unit continuously collects the grid frequency deviation Ra, the voltage volatility Rb and the line load rate Rc, and normalizes them to ensure the comparability of different physical quantities, and combines the fluctuation entropy value Bd output by the instantaneous modeling module to comprehensively fit and calculate the grid connection risk coefficient Rx. The calculation formula of the grid connection risk coefficient Rx ensures that all parameters participate in the calculation with equal weights, so as to accurately reflect the stable state of the grid connection. The risk simulation unit uses the Monte Carlo method to simulate the dynamic response of the power grid under fault conditions and output the system safety margin Sy by generating the joint probability distribution f(v, P) of wind speed - load under extreme weather scenarios. The Monte Carlo simulation calculates the system instability probability Ps through N ≥ 10000 random samplings, and considers the action timing constraints of the protection device in each simulation to ensure that Sy can accurately evaluate the safety margin of the power grid in extreme cases. The system receives in real time the power prediction result Ppred of the instantaneous modeling module, the grid connection risk coefficient Rx and the safety margin Sy calculated by the safety assessment module, and generates the unit output command through the dynamic optimization algorithm. When a power mutation (|dt| ≥ threshold) is detected, the virtual inertia compensation strategy is activated, and the virtual inertia coefficient Kv is dynamically adjusted according to the grid frequency change rate dt to achieve inertial power compensation through additional torque control. At the same time, the coupling relationship between Rx and Sy is combined to constrain the output adjustment amplitude, forming the final control signal to optimize the dynamic response ability of wind power grid connection while ensuring the stability of the power grid.

[0040] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The wind farm intelligent management system based on the Internet of Things is characterized by: It includes data perception module, dynamic correction module, instantaneous modeling module, safety assessment module and optimization scheduling module; The data perception module collects wind speed, wind direction, turbulence intensity and unit vibration data in real time through the Internet of Things terminal, constructs a spatiotemporal data set, and calculates and evaluates the environmental mutation index Hd to trigger dynamic correction instructions; After receiving the dynamic correction instruction, the dynamic correction module uses the sliding time window algorithm to screen the historical power generation data for timeliness, removes the invalid data segments, generates a dynamic training set, and combines online learning to update the prediction model parameters; The instantaneous modeling module extracts turbulence characteristics based on high-frequency sampling data, uses a hybrid model to predict instantaneous power fluctuations and outputs instantaneous power prediction results; then calculates the fluctuation entropy value Bd; finally, the fluctuation entropy value Bd is evaluated and a safety warning is activated; The safety assessment module is used to monitor the grid frequency deviation, voltage fluctuation rate and line load rate, calculate the grid connection risk factor Rx based on the fluctuation entropy value Bd, and evaluate the system robustness under extreme weather conditions through Monte Carlo simulation to obtain the safety margin Sy; The optimization scheduling module dynamically adjusts the unit output according to the instantaneous power prediction results, the grid-connected risk factor Rx and the safety margin Sy, and adopts a virtual inertia compensation strategy to provide inertial response when the power changes suddenly.

2. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized by: The data perception module includes a collection unit, a fusion calculation unit, and a trigger unit; The acquisition unit is used to collect multi-dimensional dynamic data of wind farms in real time through the Internet of Things terminal, generate spatiotemporal data sets, and transmit them to the central processing platform in real time. By aligning and standardizing wind speed, wind direction, turbulence intensity and unit vibration data in time and space, an environmental feature database is constructed. Among them, the spatiotemporal data set includes the sequence variation coefficient Sx of the real-time wind speed, the turbulence intensity distribution matrix Tij, the unit vibration spectrum Vk and the ambient temperature and humidity gradient ΔH.

3. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized by: The fusion calculation unit is used to obtain the environmental mutation index Hd based on the spatiotemporal data set by performing multi-scale fusion calculations on the sequence variation coefficient Sx of the real-time wind speed, the discreteness of the turbulence intensity distribution matrix Tij, and the energy entropy of the unit vibration spectrum Vk. The specific calculation formula is as follows: ； In the formula, represents the discreteness of the turbulence intensity distribution matrix Tij, represents the energy entropy of the unit vibration spectrum Vk; w1, w2 and w3 correspond to the sequence variation coefficient Sx of the real-time wind speed, the discreteness of the turbulence intensity distribution matrix Tij and the contribution weight of the energy entropy of the unit vibration spectrum Vk, respectively, and w1+w2+w3=1.

4. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized in that: The trigger unit is used to preset the adaptive threshold J and compare and evaluate it with the environmental mutation index Hd to determine whether the model correction mechanism needs to be activated. The specific judgment logic is as follows: If the environmental mutation index Hd ≥ adaptive threshold J, it means that the current environmental mutation degree exceeds the model adaptation range, and a dynamic correction instruction is generated and transmitted to the dynamic correction module; If the environmental mutation index Hd is less than the adaptive threshold J, it means that the current environmental state is within the tolerance of the prediction model and the system maintains the default operation mode.

5. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized in that: After receiving the dynamic correction instruction, the dynamic correction module uses a sliding time window algorithm to screen the historical power generation data for timeliness, dynamically eliminates invalid data segments that exceed the time window range, and generates a dynamic training set based on the screened valid data; then, combined with the online learning mechanism, the dynamic training set is used to update the parameters of the prediction model in real time to dynamically adjust the prediction accuracy and adaptability of the model.

6. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized by: The instantaneous modeling module includes a feature prediction unit and an entropy value evaluation unit; The feature prediction unit extracts turbulence features based on high-frequency sampling data, including instantaneous wind speed fluctuation amplitude, spectrum distribution characteristics and spatial correlation coefficient; uses a hybrid model to model turbulence features, predict instantaneous power fluctuations, and output instantaneous power prediction results; A hybrid modeling method based on physical model and data-driven is adopted. The specific prediction formula is as follows: ； Where Ppred represents the instantaneous power prediction value, Pphy represents the prediction component of the physical model based on fluid dynamics calculation, v represents the wind speed, ρ represents the air density, Pdata represents the prediction component of the data-driven model based on the turbulence feature Xturb, α and β are the weight coefficients of the physical model and the data-driven model respectively, and ϵ is the correction error term.

7. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized by: The entropy value evaluation unit is used to calculate the fluctuation entropy value Bd according to the instantaneous power prediction value Ppred, and evaluate the fluctuation entropy value Bd to activate the safety warning mechanism; the information entropy algorithm is used to calculate the fluctuation entropy value Bd, and the specific calculation formula is as follows: ; In the formula, represents the probability distribution of instantaneous power fluctuation in the i-th amplitude interval; The first safety threshold Bdth-1 and the second safety threshold Bdth-2 are preset, and compared with the fluctuation entropy value Bd to activate the safety warning; and the first safety threshold Bdth-1 is greater than the second safety threshold Bdth-2. The specific evaluation content is as follows: When the fluctuation entropy value Bd ≥ the first safety threshold Bdth-1, it is determined to be the first abnormal fluctuation, triggering the first-level safety warning, and the following operations are performed: Immediately initiate the wind turbine emergency load reduction control strategy to limit the output power to a safe range; if the fluctuation continues to intensify, trigger the forced shutdown protection; simultaneously send a high-level alarm signal to the monitoring system to prompt manual intervention; When the second safety threshold Bdth-2≤the fluctuation entropy value Bd<the first safety threshold Bdth-1, it is determined to be the second abnormal fluctuation, triggering the second-level safety warning, and performing the following operations: Automatically adjust wind turbine dynamic damping parameters or pitch rate to suppress power oscillations; Real-time optimization of forecast model parameters to enhance adaptability to short-term fluctuations; Generate early warning logs and mark abnormal periods for subsequent analysis; When the fluctuation entropy value Bd is less than the second safety threshold value Bdth-2, it is determined to be within the normal fluctuation range, the current operating state of the wind turbine is maintained, and the entropy value change trend is continuously monitored.

8. The wind farm intelligent management system based on the Internet of Things according to claim 1 is characterized by: The safety assessment module includes a real-time monitoring unit and a risk simulation unit; The real-time monitoring unit is used to continuously collect and normalize the grid frequency deviation Ra, voltage fluctuation rate Rb and line load rate Rc, and fit and calculate the grid connection risk coefficient Rx in combination with the fluctuation entropy value Bd output by the instantaneous modeling module; The specific calculation formula of the grid connection risk coefficient Rx is: ; In the formula, all parameters are calculated with equal weights. When any parameter increases abnormally, the grid connection risk coefficient Rx increases, reflecting the cumulative effect of risk.

9. The wind farm intelligent management system based on the Internet of Things according to claim 1, characterized in that: The risk simulation unit is used to simulate the dynamic response of the power grid under fault conditions by generating the wind speed-load joint probability distribution under extreme weather scenarios based on the Monte Carlo method, and output the system safety margin Sy. When performing the Monte Carlo simulation, the wind speed-load joint probability density function f(v,P) is used as input, and the system instability probability Ps is calculated through N random samplings. The safety margin Sy is defined as: ; N is ≥ 10,000 times, and the timing constraints of the protection device action are considered in each simulation.

10. The wind farm intelligent management system based on the Internet of Things according to claim 1, characterized in that: It receives the power prediction results output by the instantaneous modeling module, the grid-connected risk coefficient Rx and the safety margin Sy calculated by the safety assessment module in real time, and generates the unit output command through the dynamic optimization algorithm; when a power mutation is detected, it activates the virtual inertia compensation strategy, dynamically adjusts the virtual inertia coefficient Kv according to the grid frequency change rate dt, realizes inertia power compensation through additional torque control, and at the same time constrains the output adjustment range by combining the coupling relationship between the grid-connected risk coefficient Rx and the safety margin Sy to form the final control signal.

Images