Data security transmission method and system for wind power plant

Abstract

The invention belongs to the field of data transmission of wind power plants, and particularly discloses a secure data transmission method and system for a wind power plant, and the method comprises the steps: firstly obtaining wind speed, wind direction and other wind power plant parameters, grouping the parameters in sequence, mining a data sequence dependency relationship by means of an optimized long-short-term memory network, and extracting a feature vector; associating and integrating the equipment space and the operation relationship into a feature matrix; constructing a multi-dimensional data encryption model based on multiple dimensions of parameters, and encrypting the feature matrix by adopting a dynamic key generation mechanism; and establishing a secure channel to transmit encrypted data through a self-adaptive routing selection strategy based on a link state, and restoring the data by a receiving end by using a corresponding key and a decryption algorithm. The system is composed of a parameter acquisition unit, a data grouping unit and other eight units, all the units work cooperatively, the safety guarantee of the whole process of wind power plant data from acquisition and encryption processing to safe transmission and decryption reduction is achieved, and the data transmission safety and reliability are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data transmission of wind power plants, in particular to a data secure transmission method and system for wind power plants. BACKGROUND

[0002] With the increasing proportion of wind power in the global energy structure, the scale of wind power plants continues to expand and the intelligent degree of equipment is significantly improved. The massive data generated during the operation of wind power plants, such as wind speed, wind direction, equipment operating parameters, etc., need to be transmitted efficiently and securely. These data not only concern the real-time operation state monitoring and power generation efficiency optimization of wind power plants, but also play a decisive role in equipment maintenance, fault warning and other key links. Data secure transmission is the key foundation to ensure the stable and reliable operation of wind power plants and to realize intelligent management.

[0003] However, the existing wind power plant data transmission technology has many shortcomings in data security protection. On the one hand, traditional data transmission methods mostly use fixed encryption modes, which are difficult to adapt to the complex and variable operating environment of wind power plants. The change frequency and importance of various parameters (such as wind speed, generator power, etc.) in wind power plants are different, and the fixed encryption mode cannot flexibly adjust the encryption strategy according to the parameter characteristics, resulting in the risk of important data leakage due to insufficient encryption strength, and the influence of transmission efficiency of secondary data due to excessive encryption.

[0004] On the other hand, in the management of data transmission channels, the existing technology lacks dynamic perception and adaptive adjustment capability of communication link state. Wind power plant equipment is widely distributed, and communication links are easily affected by natural environment (such as weather changes) and network congestion, etc. The stability and bandwidth utilization rate fluctuate greatly. The traditional routing selection strategy cannot be optimized in real time according to the link state, which easily causes data transmission delay, packet loss, and even data transmission interruption, seriously affecting the timeliness and integrity of wind power plant data transmission. SUMMARY

[0005] In order to overcome the shortcomings and deficiencies of the prior art, the present application provides a data secure transmission method and system for wind power plants.

[0006] The technical solution adopted by the present application is a data secure transmission method for wind power plants, comprising the following steps: Step S1: obtaining various parameters of the wind power plant, the various parameters including wind speed, wind direction, impeller speed, generator power, nacelle temperature and yaw angle; Step S2: grouping the obtained various parameters of the wind power plant according to time sequence to form a plurality of data sequence groups; Step S3: Input each data sequence group into a pre-trained optimized Long Short-Term Memory network, and use the gating mechanism of the Long Short-Term Memory network to mine and process the long-term and short-term dependencies in the data sequence group to obtain the feature vector corresponding to each data sequence group. Step S4: Based on the spatial location and operational relationships of the wind power plant equipment, the feature vectors are integrated in terms of spatial and logical dimensions to generate an integrated feature matrix. Step S5: Construct a multidimensional data encryption model. The multidimensional data encryption model is based on the attribute dimension, time dimension and spatial dimension of various parameters of the wind power plant. It adopts a dynamic key generation mechanism. The dynamic key generation mechanism allocates different key update strategies according to the change frequency and importance of different parameters. Step S6: Use the multidimensional data encryption model to encrypt the integrated feature matrix to obtain encrypted feature data; Step S7: Establish a secure transmission channel. The secure transmission channel adopts an adaptive routing strategy based on the stability of the communication link of the wind power plant equipment and the bandwidth utilization rate to transmit the encrypted feature data to the target receiving end through the secure transmission channel. Step S8: The target receiver receives the encrypted feature data, decrypts the encrypted feature data using the corresponding decryption key and multidimensional data decryption algorithm, recovers the original feature matrix, and restores various parameters of the wind power plant based on the original feature matrix.

[0007] Further, step S3, the training process of the pre-trained optimized Long Short-Term Memory network, includes: constructing a loss function. ,in, The number of training samples. For the first The actual output of each sample For the first The network prediction output for each sample The weighting coefficients are set according to the importance of various parameters of the wind power plant. The weighting coefficients corresponding to key parameters such as wind speed and generator power are greater than the weighting coefficients corresponding to secondary parameters such as nacelle temperature. The parameters of the optimized long short-term memory network are iteratively updated using the stochastic gradient descent algorithm until the loss function converges.

[0008] Furthermore, in step S4, the feature vectors are correlated and integrated along both spatial and logical dimensions, specifically through the following formula: ,in, The integrated feature matrix, The feature vector obtained in step S3, A spatial correlation function for correlating the feature vectors based on spatial position relationships of the wind power plant equipment, A logical correlation function for correlating the feature vectors based on operational correlation relationships of the equipment, and The weighting coefficient set according to the importance of the spatial position relationship and the operational correlation relationship is greater than The value of .

[0009] Further, the step S5, the dynamic key generation mechanism of the multi-dimensional data encryption model adopts the following formula: Wherein, is the dynamic key generated at time , is the change period of the parameter, is the importance index of the parameter, which is set according to the influence degree of the parameter on the operation of the wind power plant, is the real-time change rate of the parameter; the function According to different combinations of and and , the corresponding dynamic key is generated according to the preset rule.

[0010] Further, the step S6, the integrated feature matrix is encrypted by using the multi-dimensional data encryption model, and the specific process is as follows: the integrated feature matrix is divided into a plurality of sub-matrices according to the attribute dimension, for each sub-matrix, according to the corresponding parameter attribute, different encryption transformation matrix is used for encryption through matrix operation , wherein is the divided sub-matrix, and the encryption transformation matrix is generated according to the attribute characteristics of the corresponding parameter and the dynamic key .

[0011] Further, the step S7, the adaptive routing selection strategy of the secure transmission channel is based on the following formula: Wherein, is the probability of selecting a communication link for data transmission, is the stability index of the communication link, which is calculated by the error code rate and the packet loss rate parameters of the link, is the bandwidth utilization rate of the communication link; the function According to the values of and , the transmission probability of each communication link is determined according to the preset probability distribution rule.

[0012] Further, the step S8, the multi-dimensional data decryption algorithm of the target receiving end, according to the encryption transformation matrix used in the encryption process and the dynamic key , decrypts the encrypted feature data through inverse matrix operation and key reverse processing, and the specific operation process is as follows: for each encrypted sub-matrix, the inverse matrix of the encryption transformation matrix corresponding to the encrypted sub-matrix is used to decrypt through matrix operation , wherein is the encrypted sub-matrix, and the data is restored in combination with the dynamic key .

[0013] Further, when grouping according to the time sequence in the step S2, the grouping time interval is dynamically adjusted according to the change frequency of each parameter of the wind power plant, the parameter with a high change frequency adopts a shorter grouping time interval, and the parameter with a low change frequency adopts a longer grouping time interval, and the specific adjustment formula is as follows: , wherein, is the reference time interval, is the change frequency of the parameter.

[0014] Further, in the step S3, the optimized long short-term memory network further introduces an attention mechanism, different attention weights are allocated to different parameters according to the influence degree of each parameter of the wind power plant on the power generation efficiency and equipment safety, and the attention weight is calculated through the following formula: , wherein, is the attention weight vector, is the activation function, is the weight matrix, is the hidden state of the long short-term memory network at time , and is the bias vector.

[0015] A data security transmission system for a wind farm, the system comprising: A multi-dimensional parameter acquisition unit is configured to acquire wind speed, wind direction, impeller speed, generator power, nacelle temperature, and yaw angle parameters of a wind power plant. A time sequence dynamic grouping unit is connected to the multi-dimensional parameter acquisition unit and is configured to group the acquired parameters according to a time sequence to form a plurality of data sequence groups. An LSTM feature extraction unit is connected to the time sequence dynamic grouping unit and is configured to input each data sequence group into a pre-trained optimized long short-term memory network to obtain a feature vector corresponding to each data sequence group. The space-time correlation integration unit is connected with the LSTM feature extraction unit, and is used for correlating and integrating the feature vectors in the space dimension and the logic dimension according to the spatial position relationship and the operation correlation relationship of the wind power plant equipment, to generate an integrated feature matrix; The three-dimensional encryption model construction unit is connected with the space-time correlation integration unit, and is used for constructing a multi-dimensional data encryption model based on the attribute dimension, the time dimension and the space dimension of each parameter of the wind power plant, and adopting a dynamic key generation mechanism; The matrix block encryption processing unit is connected with the three-dimensional encryption model construction unit, and is used for performing encryption processing on the integrated feature matrix by using the multi-dimensional data encryption model, to obtain encrypted feature data; The adaptive routing transmission unit is connected with the matrix block encryption processing unit, and is used for establishing a secure transmission channel adopting an adaptive routing selection strategy, and transmitting the encrypted feature data to a target receiving end through the secure transmission channel; The key-driven decryption restoration unit is connected with the adaptive routing transmission unit, and is used for receiving the encrypted feature data, decrypting the encrypted feature data by using a corresponding decryption key and a multi-dimensional data decryption algorithm, restoring the original feature matrix, and restoring each parameter of the wind power plant according to the original feature matrix.

[0016] Advantages: the method and system for data secure transmission of a wind farm are provided, and the multi-dimensional data encryption model is constructed according to the attributes, time and space dimensions of the parameters of the wind power plant, and the dynamic key generation mechanism updates the key according to the parameter change frequency and importance, so that the limitation of the traditional fixed encryption mode is changed. For the key parameters such as wind speed and generator power that change frequently, the high-frequency key update and high-intensity encryption strategy are adopted to ensure the data security; for the secondary parameters such as engine room temperature, the transmission efficiency is improved by using the adaptive encryption mode, so that the problems of insufficient encryption of important data and excessive encryption of secondary data are solved. In the transmission channel management aspect, the secure transmission channel adopts the adaptive routing selection strategy based on the link stability and bandwidth utilization, and the communication link state is sensed in real time. When the link is unstable due to the change of the natural environment or network congestion, the system automatically adjusts the routing, selects a better link to transmit data, reduces the delay and packet loss rate, avoids transmission interruption, and compared with the traditional fixed routing strategy, the timeliness and integrity of data transmission are greatly improved. At the same time, the optimized long short-term memory network combined with the attention mechanism accurately extracts the data features, the space-time correlation integration further excavates the data value, and each unit cooperates to construct a complete, efficient and safe data transmission system, which effectively ensures the safe and reliable transmission of the wind farm data. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is a method step flowchart of the application. Figure 2 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, a method for secure data transmission in wind farms includes the following steps: Step S1: Obtain various parameters of the wind power plant, including but not limited to wind speed, wind direction, rotor speed, generator power, nacelle temperature and yaw angle; Specifically, this step is fundamental to secure data transmission, involving real-time acquisition of operating parameters through sensors distributed across key locations within the wind power plant. Wind speed sensors are typically installed at the hub height, employing a three-cup or ultrasonic principle for measurement, with a typical range of 0-75 m / s and an accuracy of ±0.3 m / s. Wind direction sensors are mostly photoelectric or magnetoelectric, with a measurement range of 0-360° and an accuracy of ±3°. Rotor speed sensors monitor speed via gears or encoders, with a measurement range of 0-25 rpm and a resolution of 0.01 rpm. Generator power is acquired through current and voltage transformers, covering 0-120% of the generator's rated power with an accuracy of ±0.5%. Nacelle temperature sensors use PT100 RTDs or thermocouples, with a measurement range of -40℃ to 120℃ and an accuracy of ±0.5℃. Yaw angle sensors are mostly absolute encoders, with a measurement range of 0-360° and an accuracy of ±0.1°. These sensors continuously collect data at a set sampling frequency (usually 10-100Hz) to provide raw material for subsequent analysis.

[0020] The completeness and accuracy of parameter acquisition directly affect the system's judgment of the wind farm's operating status. Wind speed reflects the wind energy resource status and is a key input for power prediction; wind direction guides the yaw system to adjust the nacelle orientation to ensure the rotor is aligned with the incoming flow direction; rotor speed is related to generator power and reflects wind energy conversion efficiency; nacelle temperature reflects the equipment's operating temperature rise, and an abnormal increase may indicate a fault; yaw angle is related to the accuracy of the unit's alignment with the wind and affects power generation efficiency. The sampling frequency of each parameter needs to be set reasonably according to its variation characteristics. Parameters that change rapidly (such as wind speed) should use a high sampling frequency, while parameters that change slowly (such as nacelle temperature) can have their sampling frequency appropriately reduced to balance the amount of data and the system's processing capacity.

[0021] Step S2: Group the acquired parameters of the wind power plant according to the time series to form multiple data sequence groups; Specifically, this step is based on the time dimension to discretize the continuous data stream, and adopts a sliding window mechanism to realize data grouping. The window length is set according to the parameter variation period and analysis requirements, and is usually 10-60 seconds, and the sliding step is 1-10 seconds. For example, for the fast-changing wind speed parameter, the window length can be set to 10 seconds, and the sliding step is 1 second; for the slow-changing cabin temperature parameter, the window length can be extended to 60 seconds, and the sliding step is 5 seconds. Each data sequence group contains multi-parameter data in the same time period, forming a multi-dimensional time sequence, such as [wind speed t1, wind direction t1, speed t1, power t1, temperature t1, angle t1; wind speed t2, wind direction t2, speed t2, power t2, temperature t2, angle t2; …; wind speed tn, wind direction tn, speed tn, power tn, temperature tn, angle tn], where n is the number of sampling points in the window.

[0022] The granularity of data grouping directly affects the accuracy of subsequent feature extraction and system real-time performance. Too short window length will result in insufficient information in the data sequence group, affecting the model's capture of the dependence between parameters; too long window length will reduce the system's real-time response ability and increase the processing delay. The grouping process needs to ensure time continuity and parameter synchronization, that is, the sampling time of each parameter in the same sequence group is strictly aligned. For asynchronous sampling parameters, interpolation algorithms (such as linear interpolation, spline interpolation) are used for time alignment to ensure the spatiotemporal consistency of the data sequence group, laying a foundation for subsequent time series analysis.

[0023] Step S3: input each data sequence group into a pre-trained optimized long short-term memory network, and use the gating mechanism of the long short-term memory network to mine and process the long-term and short-term dependence relationships in the data sequence group to obtain the corresponding feature vector of each data sequence group; Specifically, this step uses an optimized LSTM network for time series feature extraction, and the network structure includes an input layer, a hidden layer and an output layer. The input layer has a dimension of window length x parameter number (such as 10-second window x 6 parameters), the hidden layer includes multiple LSTM units, and each unit includes an input gate, a forget gate and an output gate. The input gate controls the entry of the current input information, the forget gate determines the retention degree of the previous time state, and the output gate adjusts the output of the current state. Through the gating mechanism, the network can effectively handle the long-term and short-term dependence relationships in the data, such as the immediate influence of wind speed change on generator power (short-term dependence) and the influence of long-term wind speed pattern on equipment wear (long-term dependence). The network training adopts the stochastic gradient descent algorithm, optimizes the weight parameters by minimizing the error between the predicted value and the true value, and the training data covers historical sequence groups in normal operation and abnormal conditions.

[0024] The optimized LSTM network introduces an attention mechanism and a batch normalization layer on the basis of a traditional structure. The attention mechanism dynamically allocates weights according to the importance of parameters, such as increasing the weights of wind speed and rotor speed in the power prediction scenario; the batch normalization layer accelerates network convergence and enhances model stability. After training, the data sequence group is input into the network, and the fixed-dimension feature vector (such as 128 dimensions) is obtained from the output layer through forward propagation calculation. The vector condenses the key information in the original data, retains the time sequence correlation between parameters, and reduces the data dimension, providing compact and representative feature expression for subsequent integration processing.

[0025] Step S4: according to the spatial position relationship and operation correlation of the wind power plant equipment, the spatial dimension and logical dimension of the feature vector are associated and integrated to generate an integrated feature matrix; Specifically, this step associates and integrates the feature vector from the spatial and logical dimensions. In the spatial dimension, an association model is established based on the physical layout of the wind farm. For example, there is a wake effect between adjacent wind turbines, and the wake of the upstream wind turbine will affect the wind speed and direction of the downstream wind turbine. The spatial correlation is quantified by a distance weight matrix, and the closer the distance, the greater the weight. In the logical dimension, an association relationship is constructed according to the operation principle of the equipment, such as the nonlinear mapping relationship between wind speed and rotor speed, generator power, and the synchronous adjustment relationship between yaw angle and wind direction. The logical dependence strength between parameters is determined by a causal analysis algorithm (such as Granger causality test) to form a logical association matrix.

[0026] The integration process adopts a weighted summation strategy, multiplies the spatial association matrix and the logical association matrix with the feature vector respectively to obtain a spatial feature matrix and a logical feature matrix, and then combines them according to a certain weight coefficient to generate the final integrated feature matrix. The weight coefficient is dynamically adjusted according to the layout characteristics and operation mode of the wind farm. For densely laid wind farms, the spatial association weight is increased; for wind farms with high requirements for cooperative control, the logical association weight is increased. The integrated feature matrix not only retains the time sequence characteristics of individual equipment, but also integrates the spatial coupling and logical interaction information between equipment, providing a data basis for comprehensively sensing the overall operation state of the wind farm.

[0027] Step S5: constructing a multi-dimensional data encryption model based on the attribute dimension, time dimension and spatial dimension of each parameter of the wind power plant, using a dynamic key generation mechanism that allocates different key update strategies according to the change frequency and importance of different parameters; Specifically, the multi-dimensional data encryption model constructed in this step designs encryption strategies from three dimensions. In the attribute dimension, security levels are divided according to the sensitivity of parameters, such as high security level for core parameters such as generator power and control instructions, and low security level for environmental temperature and non-critical device status. In the time dimension, the changing characteristics of parameters are considered, and fast key update mechanism is adopted for parameters with high real-time requirements (such as wind speed), and the key update frequency is reduced for parameters with slow changes (such as device maintenance records). In the spatial dimension, different spatial keys are generated for devices in different regions based on their physical locations and data flow directions, such as using different keys for devices at the bottom of the tower and devices at the top of the cabin.

[0028] The dynamic key generation mechanism generates keys based on the change period, importance index and real-time change rate of parameters. For parameters with high security level and frequent changes, the key update period is set to minutes; for parameters with low security level and slow changes, the key update period can be extended to hours. The key generation process introduces random number factors and time stamps to enhance the randomness and timeliness of the keys. At the same time, a key management center is established to be responsible for key generation, distribution, storage and update, to ensure the security and traceability of the keys. Through multi-dimensional dynamic encryption strategy, the security of critical data is guaranteed, and the system resource consumption is optimized.

[0029] Step S6: encrypting the integrated feature matrix using the multi-dimensional data encryption model to obtain encrypted feature data; Specifically, this step uses a block encryption strategy to process the integrated feature matrix. First, the feature matrix is divided into multiple sub-matrices according to the attribute dimension, each sub-matrix corresponding to a type of parameter (such as an electrical parameter sub-matrix and a mechanical parameter sub-matrix). For each sub-matrix, according to its security level and associated dynamic key, select the corresponding encryption algorithm (such as AES, RSA) and encryption mode (such as CBC, CTR). For example, for the electrical parameter sub-matrix with high security level, use AES-256 algorithm and CTR mode; for the environmental parameter sub-matrix with low security level, use AES-128 algorithm and CBC mode.

[0030] In the encryption process, first, perform row / column permutation operation on each sub-matrix to disrupt the original data order and enhance the confusion; then further diffuse the data through linear transformation and nonlinear transformation (such as S-box transformation); finally, perform XOR operation with the dynamically generated key to generate ciphertext sub-matrix. The encryption processes of different sub-matrices are independent of each other and are processed in parallel to improve the encryption efficiency. The encrypted sub-matrices are reorganized according to the original matrix structure to form complete encrypted feature data. This data effectively hides sensitive information while maintaining the original matrix dimension and structure, ensuring the confidentiality and integrity of the data during transmission.

[0031] Step S7: Establish a secure transmission channel, which uses an adaptive routing strategy based on the stability and bandwidth utilization of the wind power plant equipment communication link to transmit the encrypted feature data to the target receiving end through the secure transmission channel. Specifically, the secure transmission channel established in this step is based on a software-defined network (SDN) architecture, and the core switch uses the OpenFlow protocol to achieve programmable control of traffic. The system monitors the stability indicators (such as bit error rate and packet loss rate) and bandwidth utilization of each communication link in real time, and collects link state data every 500 milliseconds through probe technology. The stability indicators are obtained by calculating the sliding average of the bit error rate and packet loss rate over 10 consecutive sampling periods, and the bandwidth utilization is determined by monitoring the ratio of instantaneous traffic to maximum bandwidth.

[0032] The adaptive routing strategy uses a multi-index comprehensive evaluation algorithm to map the stability indicators and bandwidth utilization to the 0-1 interval, and obtains the link quality score by weighted summation. When the link quality score of a certain link is detected to be below a threshold value (such as 0.6), the routing reselection mechanism is triggered. In the new routing selection process, links with high stability and sufficient bandwidth are preferentially selected, while path hop count and load balancing are also considered. The data transmission uses segmented encryption tunnel technology, with independent keys used for each network segment to ensure end-to-end security. When the primary link fails, the backup link automatically switches within 100 milliseconds, ensuring the continuity of data transmission.

[0033] Step S8: The target receiving end receives the encrypted feature data, decrypts it using the corresponding decryption key and multi-dimensional data decryption algorithm, restores the original feature matrix, and restores the parameters of the wind power plant based on the original feature matrix.

[0034] Specifically, the receiving end uses a decryption architecture that is symmetrical to the sending end, and first obtains the current valid decryption key from the key management center. This key strictly corresponds to the dynamic key used in the encryption process and contains a timestamp, parameter attribute identifier, and spatial region information. The decryption process is performed in reverse order of the encryption, first dividing the encrypted feature data into sub-matrices, and then decrypting each sub-matrix using the corresponding decryption algorithm and key. For example, for a sub-matrix encrypted using the AES-256 / CTR mode, the same key and algorithm are used for inverse transformation to restore the original sub-matrix data.

[0035] The decrypted sub-matrix is reconstructed into a complete feature matrix through inverse permutation and inverse linear transformation operations. The receiving end verifies the integrity and authenticity of the feature matrix, compares the calculated hash value with the hash check value attached by the sending end, and ensures that the data has not been tampered with during transmission. Finally, according to the mapping relationship in the feature extraction stage, the feature matrix is restored to the original parameter value. For the parameters that have been time-aligned and interpolated, the original sampling time data is restored through the reverse interpolation algorithm, realizing the complete and accurate restoration of various parameters of the wind power plant, and providing a reliable basis for subsequent data analysis and decision-making.

[0036] Preferably, the training process of the pre-trained optimized long short-term memory network in step S3 includes: constructing a loss function wherein, is the number of training samples, is the true output of the th sample, is the network prediction output of the th sample, is a weight coefficient set according to the importance of various parameters of the wind power plant, and the weight coefficient corresponding to key parameters such as wind speed and generator power is greater than the weight coefficient corresponding to secondary parameters such as cabin temperature; the random gradient descent algorithm is used to iteratively update the parameters of the optimized long short-term memory network until the loss function converges.

[0037] Specifically, by constructing a weighted loss function, different weight coefficients are set for parameters of different importance, and the weight of key parameters is higher, so that the network training process pays more attention to data that has a greater impact on the operation of the wind farm, and the prediction accuracy of the model for key parameters is improved. The random gradient descent algorithm is used to iteratively update the network parameters to ensure the convergence of the loss function, so that the network can accurately capture the dependencies in the data sequence, optimize the feature extraction effect, and enhance the adaptability and prediction ability of the model for wind farm data.

[0038] Preferably, the association and integration of the feature vectors in step S4 in the spatial dimension and the logical dimension are performed through the following formula: wherein, is the integrated feature matrix, is the feature vector obtained in step S3, is a spatial correlation function for associating the feature vector based on the spatial position relationship of the wind power plant equipment, is a logical correlation function for associating the feature vector based on the operational association relationship of the equipment, and are weighting coefficients set according to the importance of the spatial position relationship and the operational association relationship, and the value of is greater than the value of the parameter.

[0039] Specifically, the spatial correlation function and the logical correlation function are used to process the spatial positions and the operating relationships of the devices respectively to generate an integrated feature matrix. The weighting coefficients are dynamically adjusted according to the layout of the wind farm and the cooperation requirements of the devices. In a wind farm with a compact layout and strong correlation between the devices, the weight of the logical correlation function is increased, so that the integrated feature matrix can more comprehensively reflect the interaction between the devices and enhance the spatial and logical correlation of the data, thereby providing a richer and more accurate information basis for subsequent encryption and analysis.

[0040] Preferably, the dynamic key generation mechanism of the multi-dimensional data encryption model in the step S5 adopts the following formula: wherein, is the dynamic key generated at time is the change period of the parameter, is an importance index of the parameter, which is set according to the influence degree of the parameter on the operation of the wind power plant, is the real-time change rate of the parameter; the function generates the corresponding dynamic key according to the different combinations of and and the preset rules.

[0041] Specifically, the dynamic key is generated based on the change period, the importance index and the real-time change rate of the parameter. For parameters that change rapidly and are important, a high-frequency key update strategy is set to ensure the security of the key data. For secondary parameters that change slowly, the key update frequency is reduced to balance the security and system resource consumption. Through this mechanism, the key is dynamically adjusted according to the characteristics of the parameter, enhancing the pertinence and timeliness of the encryption, effectively resisting attacks on static keys and improving the security of data encryption.

[0042] Preferably, in the step S6, the integrated feature matrix is encrypted by using the multi-dimensional data encryption model. The specific process is as follows: the integrated feature matrix is divided into a plurality of sub-matrices according to the attribute dimensions. For each sub-matrix, a different encryption transformation matrix is used according to the corresponding parameter attribute, and the matrix operation is performed for encryption, wherein is the divided sub-matrix, and the encryption transformation matrix is generated according to the attribute characteristics of the corresponding parameter and the dynamic key .

[0043] ​Specifically, the integrated feature matrix is divided into blocks according to attributes, and different encryption transformation matrices are used for different sub-matrices. The encryption transformation matrix is combined with the parameter attribute feature and the dynamic key to generate an encryption strategy unique to each sub-matrix. Encryption is achieved through matrix operations, which not only maintains the integrity of the data structure but also enhances the encryption complexity, effectively protecting the confidentiality of the data during transmission and preventing unauthorized access and tampering.

[0044] Preferably, the adaptive routing strategy of the secure transmission channel in step S7 is based on the following formula: wherein, is the probability of selecting a communication link for data transmission, is a stability index of the communication link, calculated by the bit error rate, packet loss rate and other parameters of the link, is the bandwidth utilization of the communication link; the function According to the values of and , the transmission probability of each communication link is determined according to the preset probability distribution rule.

[0045] Specifically, by evaluating the stability index and bandwidth utilization of the communication link, the transmission probability of each link is calculated. When the link stability decreases or the bandwidth is insufficient, the system preferentially selects high-quality links to transmit data, avoiding data delay or loss due to link problems. This strategy real-time perceives the network state, dynamically adjusts the transmission path, ensures the reliability and efficiency of data transmission, and is especially suitable for complex and variable communication environments in wind farms.

[0046] Preferably, the multi-dimensional data decryption algorithm of the target receiving end in step S8 is based on the encryption transformation matrix and the dynamic key used in the encryption process. The encrypted feature data is decrypted through inverse matrix operation and key reverse processing. The specific operation process is as follows: for each encrypted sub-matrix, the inverse matrix of its corresponding encryption transformation matrix is used to decrypt through matrix operation , wherein is the encrypted sub-matrix, and the data is restored in combination with the dynamic key .

[0047] Specifically, the inverse matrix of the encryption transformation matrix and the dynamic key are used to decrypt the encrypted sub-matrix. The original sub-matrix data is restored through inverse matrix operation, and the decryption accuracy is ensured by combining the key reverse processing. The decryption process strictly corresponds to the encryption process, ensuring that the data can be accurately restored after transmission, verifying the integrity and authenticity of the data, and ensuring that the parameters obtained by the receiving end are consistent with those of the sending end, providing a reliable basis for subsequent data analysis.

[0048] Preferably, the grouping time interval in step S2 is According to the change frequency of each parameter of the wind power plant, the parameter with high change frequency is grouped with short time interval, and the parameter with low change frequency is grouped with long time interval. The specific adjustment formula is: , wherein, is the reference time interval, is the change frequency of the parameter.

[0049] Specifically, the grouping interval is set inversely according to the parameter change frequency. The parameter with fast change is grouped with short interval to capture the rapid change characteristics of the data; the parameter with slow change is grouped with long interval to reduce data redundancy and improve processing efficiency. This mechanism makes the data grouping more in line with the parameter characteristics, optimizes the information content of the data sequence group, provides more reasonable data units for subsequent feature extraction, and improves the analysis ability of the model for different parameters.

[0050] Preferably, the long short-term memory network optimized in step S3 further introduces an attention mechanism, assigns different attention weights to different parameters according to the influence of each parameter of the wind power plant on the power generation efficiency and equipment safety, and calculates the attention weight through the following formula: , wherein, is the attention weight vector, is the activation function, is the weight matrix, is the hidden state of the long short-term memory network at time , and is the bias vector.

[0051] Specifically, the attention weight is assigned according to the influence of the parameter on the power generation efficiency and equipment safety. Higher weight is given to core parameters such as wind speed and power to make the network focus on feature extraction of these parameters; lower weight is given to secondary parameters to reduce irrelevant information interference. Through this mechanism, the network can automatically focus on key data and enhance the ability to capture important features, thereby improving the perception accuracy and prediction reliability of the model for the operating state of the wind farm.

[0052] As shown in Figure 2 , a data security transmission system for a wind farm includes: A multi-dimensional parameter acquisition unit is configured to acquire wind speed, wind direction, impeller speed, generator power, nacelle temperature, yaw angle, and other parameters of a wind power plant. A time sequence dynamic grouping unit is connected to the multi-dimensional parameter acquisition unit and is configured to group the acquired parameters in time sequence to form a plurality of data sequence groups. An LSTM feature extraction unit is connected with the time sequence dynamic grouping unit, configured to input each data sequence group into a pre-trained optimized long short-term memory network to obtain a corresponding feature vector of each data sequence group. A space-time correlation integration unit is connected with the LSTM feature extraction unit, configured to perform correlation integration of the feature vector in a space dimension and a logic dimension according to a spatial position relationship and an operation correlation relationship of the wind power plant equipment to generate an integrated feature matrix. A three-dimensional encryption model construction unit is connected with the space-time correlation integration unit, configured to construct a multi-dimensional data encryption model based on attribute dimensions, time dimensions and space dimensions of various parameters of the wind power plant, and adopt a dynamic key generation mechanism. A matrix block encryption processing unit is connected with the three-dimensional encryption model construction unit, configured to perform encryption processing on the integrated feature matrix by using the multi-dimensional data encryption model to obtain encrypted feature data. An adaptive routing transmission unit is connected with the matrix block encryption processing unit, configured to establish a secure transmission channel adopting an adaptive routing selection strategy, and transmit the encrypted feature data to a target receiving end through the secure transmission channel. A key-driven decryption restoration unit is connected with the adaptive routing transmission unit, configured to receive the encrypted feature data, decrypt the encrypted feature data by using a corresponding decryption key and a multi-dimensional data decryption algorithm, restore an original feature matrix, and restore various parameters of the wind power plant according to the original feature matrix.

[0053] The wind farm data secure transmission method and system provided by the application can solve the problems of insufficient encryption of important data and excessive encryption of secondary data in the traditional fixed encryption mode, and can improve the security of core data and the overall data transmission efficiency.

[0054] In the transmission channel management, the traditional method is difficult to deal with the unstable communication link of wind farm, low bandwidth utilization. The safe transmission channel of the system selects the optimal transmission path intelligently by monitoring the stability and bandwidth utilization of each communication link in real time. When the signal of a certain link is weakened due to bad weather, or the bandwidth is insufficient due to network congestion, the system can quickly perceive and automatically switch to a more stable and bandwidth-rich link for data transmission, effectively reducing data transmission delay and packet loss, greatly improving the timeliness and integrity of data transmission, and being more environmentally adaptable and reliable than the traditional transmission method.

[0055] In the data feature processing layer, the system uses an optimized long short-term memory network combined with an attention mechanism to accurately identify the influence of different parameters on the power generation efficiency and equipment safety of the wind farm. For key parameters with significant impact, the system gives them higher attention and digs deeper into their data features. Secondary parameters are reasonably processed, so that data value is extracted comprehensively and efficiently. At the same time, based on the spatial position and running correlation of the equipment, the feature vectors are integrated, further enhancing the correlation and logic between data, providing a solid foundation for the deep analysis and application of wind farm data.

[0056] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "arrangement", "installation", "connection", "connection", "fixing" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be connected inside two elements. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0057] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, replacements and variations of these embodiments can be made without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalent scope.

Claims

1. A method for secure data transmission in wind farms, characterized in that, Includes the following steps: Step S1: Obtain various parameters of the wind power plant, including wind speed, wind direction, rotor speed, generator power, nacelle temperature, and yaw angle; Step S2: Group the acquired parameters of the wind power plant according to the time series to form multiple data sequence groups; Step S3: Input each data sequence group into a pre-trained optimized Long Short-Term Memory network, and use the gating mechanism of the Long Short-Term Memory network to mine and process the long-term and short-term dependencies in the data sequence group to obtain the feature vector corresponding to each data sequence group. Step S4: Based on the spatial location and operational relationships of the wind power plant equipment, the feature vectors are integrated in terms of spatial and logical dimensions to generate an integrated feature matrix. Step S5: Construct a multidimensional data encryption model. The multidimensional data encryption model is based on the attribute dimension, time dimension and spatial dimension of various parameters of the wind power plant. It adopts a dynamic key generation mechanism. The dynamic key generation mechanism allocates different key update strategies according to the change frequency and importance of different parameters. Step S6: Use the multidimensional data encryption model to encrypt the integrated feature matrix to obtain encrypted feature data; Step S7: Establish a secure transmission channel. The secure transmission channel adopts an adaptive routing strategy based on the stability of the communication link of the wind power plant equipment and the bandwidth utilization rate to transmit the encrypted feature data to the target receiving end through the secure transmission channel. Step S8: The target receiver receives the encrypted feature data, decrypts the encrypted feature data using the corresponding decryption key and multidimensional data decryption algorithm, recovers the original feature matrix, and restores various parameters of the wind power plant based on the original feature matrix.

2. The data security transmission method for wind farms according to claim 1, characterized in that, Step S3, the training process of the pre-trained optimized Long Short-Term Memory network, includes: constructing a loss function. ,in, The number of training samples. For the first The actual output of each sample For the first The network prediction output for each sample The weighting coefficients are set according to the importance of various parameters of the wind power plant. The weighting coefficients corresponding to key parameters such as wind speed and generator power are greater than the weighting coefficients corresponding to secondary parameters such as nacelle temperature. The parameters of the optimized long short-term memory network are iteratively updated using the stochastic gradient descent algorithm until the loss function converges.

3. The data security transmission method for wind farms according to claim 1, characterized in that, Step S4 involves integrating the feature vectors along both spatial and logical dimensions, specifically using the following formula: ,in, The integrated feature matrix, The feature vector obtained in step S3, This is a spatial association function that performs association processing on feature vectors based on the spatial location relationships of equipment in a wind power plant. This is a logical association function that performs association processing on feature vectors based on device operation relationships. and These are weighting coefficients set according to the importance of spatial location and operational correlation, in situations where wind farms are densely distributed and the requirements for coordinated equipment operation are high. The value is greater than The value of .

4. The data security transmission method for wind farms according to claim 1, characterized in that, In step S5, the dynamic key generation mechanism of the multidimensional data encryption model adopts the following formula: ,in, In time The generated dynamic key, The period of parameter variation These parameters are set as importance indicators based on their impact on the operation of wind power plants. The real-time rate of change of the parameter; function according to and Different combinations of [keys] generate corresponding dynamic keys according to preset rules.

5. A data security transmission method for wind farms according to claim 1, characterized in that, Step S6 involves encrypting the integrated feature matrix using a multidimensional data encryption model. Specifically, the integrated feature matrix is ​​divided into multiple sub-matrices according to attribute dimensions. For each sub-matrix, a different encryption transformation matrix is ​​applied based on its corresponding parameter attributes. Through matrix operations Encryption is performed, among which The encryption transformation matrix is ​​used to divide the submatrix. Based on the attribute characteristics of the corresponding parameters and the dynamic key generate.

6. The data security transmission method for wind farms according to claim 1, characterized in that, In step S7, the adaptive routing selection strategy for the secure transmission channel is based on the following formula: ,in, The probability of selecting a particular communication link for data transmission. It is a stability indicator of the communication link, calculated using parameters such as the link's bit error rate and packet loss rate. For the bandwidth utilization of the communication link; function according to and The value is used to determine the transmission probability of each communication link according to a preset probability allocation rule.

7. The data security transmission method for wind farms according to claim 1, characterized in that, In step S8, the multi-dimensional data decryption algorithm at the target receiving end is based on the encryption transformation matrix used in the encryption process. and dynamic key The encrypted feature data is decrypted through inverse matrix operations and key reversal. Specifically, for each encrypted submatrix, the inverse of its corresponding encryption transformation matrix is ​​used... Through matrix operations Decryption is performed, among which This is the encrypted submatrix, combined with a dynamic key. Perform data restoration.

8. A data security transmission method for wind farms according to claim 1, characterized in that, In step S2, when grouping by time series, the grouping time interval is... The parameters of the wind power plant are dynamically adjusted based on their frequency of change. Parameters with high frequency of change use shorter time intervals, while parameters with low frequency of change use longer time intervals. The specific adjustment formula is as follows: ,in, As the reference time interval, The frequency of parameter change.

9. A data security transmission method for wind farms according to claim 1, characterized in that, In step S3, the optimized Long Short-Term Memory network also introduces an attention mechanism. Based on the impact of various parameters of the wind power plant on power generation efficiency and equipment safety, different attention weights are assigned to different parameters. Specifically, the attention weights are calculated using the following formula: ,in, For attention weight vectors, For activation function, This is the weight matrix. For Long Short-Term Memory Networks in time The hidden state, This is the bias vector.

10. A data security transmission system for wind farms, characterized in that, The system includes: The multi-dimensional parameter acquisition unit is used to acquire various parameters of the wind power plant, such as wind speed, wind direction, rotor speed, generator power, nacelle temperature, and yaw angle. The time-series dynamic grouping unit is connected to the multi-dimensional parameter acquisition unit and is used to group the acquired parameters according to the time series to form multiple data sequence groups; The LSTM feature extraction unit, connected to the temporal dynamic grouping unit, is used to input each data sequence group into a pre-trained optimized long short-term memory network to obtain the feature vector corresponding to each data sequence group. The spatiotemporal correlation integration unit, connected to the LSTM feature extraction unit, is used to perform spatial and logical dimension correlation integration on the feature vectors based on the spatial location relationship and operational correlation relationship of the wind power plant equipment, and generate an integrated feature matrix. The three-dimensional encryption model construction unit is connected to the spatiotemporal correlation integration unit and is used to construct a multi-dimensional data encryption model based on the attribute dimension, time dimension and spatial dimension of various parameters of the wind power plant, and adopts a dynamic key generation mechanism. The matrix block encryption processing unit is connected to the three-dimensional encryption model construction unit and is used to encrypt the integrated feature matrix using a multi-dimensional data encryption model to obtain encrypted feature data. An adaptive routing transmission unit, connected to the matrix block encryption processing unit, is used to establish a secure transmission channel using an adaptive routing selection strategy and transmit the encrypted feature data to the target receiving end through the secure transmission channel. The key-driven decryption and restoration unit is connected to the adaptive routing transmission unit. It is used to receive the encrypted feature data, decrypt the encrypted feature data using the corresponding decryption key and multidimensional data decryption algorithm, recover the original feature matrix, and restore various parameters of the wind power plant based on the original feature matrix.

Images