Intelligent frequency conversion control method and system

Abstract

The invention provides an intelligent frequency conversion control method and system. The method comprises the following steps: acquiring an environment parameter set and equipment operation data of intelligent equipment according to a preset multi-dimensional sensor; the method comprises the following steps: processing an environment parameter set and multi-dimensional parameters of equipment operation data to generate a multi-dimensional feature vector; constructing a neural network wind speed prediction model according to the multi-dimensional feature vector; dynamically adjusting a hidden layer node connection path of the wind speed prediction model to generate a dynamic wind speed set value; the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the discourse domain range of the deviation change rate are dynamically adjusted, and a PID parameter adjusting instruction is generated in combination with the fuzzy rule base; and generating a real-time adjustment parameter of the fan frequency of the intelligent equipment according to the PID parameter adjustment instruction. PID parameters can be dynamically adjusted under the condition of different wind speed deviations, it is ensured that the frequency of a draught fan is accurately adjusted, and therefore the dynamically-changing ventilation requirement is met.

Description

Technical Field

[0001] The present application relates to the field of intelligent device control, and in particular to an intelligent frequency conversion control method and system. Background Art

[0002] With the increasing popularity of smart home systems, many smart devices (such as air purifiers, smart fans, and smart ventilation systems) are being widely deployed in residential, office, and public spaces. These devices need to intelligently adjust wind speed or ventilation intensity based on multiple environmental factors, such as indoor air quality, temperature and humidity, and occupancy density, to maintain comfort while saving energy. However, existing intelligent control methods generally rely on fixed-parameter PID controllers or simple rule-based triggering logic. These methods lack a deep understanding of multidimensional environmental data, making it difficult to effectively predict and dynamically respond to future ventilation needs. Summary of the Invention

[0003] The present application provides an intelligent variable frequency control method and system for solving the problem in related technologies that fixed frequency or simple variable frequency control strategies cannot meet the dynamic changes in future ventilation needs.

[0004] A first aspect of the present application provides an intelligent frequency conversion control method, the intelligent frequency conversion control method comprising: Collect environmental parameter sets and device operation data of smart devices based on preset multi-dimensional sensors; Generate a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the equipment operation data; Constructing a neural network wind speed prediction model based on the multidimensional feature vector; Dynamically adjusting the connection paths of hidden layer nodes of the neural network wind speed prediction model through an ant colony algorithm to generate a dynamic wind speed setting value within a prediction time period; Dynamically adjust the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate through the variable domain expansion factor of fuzzy control, and generate PID parameter adjustment instructions in combination with the fuzzy rule base; A real-time adjustment parameter of the fan frequency of the smart device is generated according to the PID parameter adjustment instruction.

[0005] Optionally, in a first implementation of the first aspect of the present application, the step of generating a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the device operation data includes: Analyzing the correlation between the multi-dimensional parameters and system energy consumption and purification efficiency, and determining a core parameter set whose correlation is higher than a preset threshold; De-noising the core parameter set based on sliding window filtering to generate a de-noised normalized data matrix; Performing dimensionality reduction processing on the redundant dimensions of the noise-normalized data matrix by covariance matrix eigenvalue decomposition; The low-dimensional feature vector after dimensionality reduction processing is cross-validated by a redundant sensor group, and abnormal data is replaced or interpolated and corrected according to the validation result to generate the multi-dimensional feature vector.

[0006] Optionally, in a second implementation of the first aspect of the present application, the step of dynamically adjusting the hidden layer node connection path of the neural network wind speed prediction model by using an ant colony algorithm to generate a dynamic wind speed set value within the prediction time period includes: Performing a global search on the initial weights of the neural network wind speed prediction model, and generating an optimized initial weight matrix through selection, crossover, and mutation operations; Initializing the pheromone concentration distribution of the ant colony algorithm based on the optimized initial weight matrix, and iteratively screening the connection paths between the hidden layer nodes through the pheromone volatility factor and the path selection probability function; Dynamically update the weight distribution of the connection path according to the pheromone concentration between the hidden layer nodes, and generate the hidden layer node connection path adjustment instruction; The environmental parameters collected in real time are input into the adjusted neural network wind speed prediction model, and the dynamic wind speed set value within the prediction time period is generated through forward propagation calculation.

[0007] Optionally, in a third implementation of the first aspect of the present application, the step of dynamically adjusting the domain range of the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the deviation change rate by using the variable universe scaling factor of fuzzy control, and generating a PID parameter adjustment instruction in combination with a fuzzy rule base includes: Calculating a wind speed deviation based on a difference between the dynamic wind speed setting value and the real-time wind speed; Performing differential calculation on the wind speed deviation through a sliding time window to generate a deviation change rate; Dynamically adjusting the domain range of the variable domain scaling factor based on the relationship between the absolute value of the wind speed deviation and a preset deviation threshold; Input the wind speed deviation, the deviation change rate sequence and the adjusted domain range into a fuzzy rule base, and generate a fuzzy output set through membership function mapping; The fuzzy output set is defuzzified and a PID parameter adjustment instruction is generated in combination with the current parameter value of the PID controller.

[0008] Optionally, in a fourth implementation of the first aspect of the present application, the step of generating a real-time adjustment parameter of the fan frequency of the smart device according to the PID parameter adjustment instruction includes: Performing weight decoupling on the proportional, integral and differential components in the PID parameter adjustment instruction to generate a parameter adjustment matrix corresponding to each control component; A transfer function model is constructed based on the parameter adjustment matrix, and overshoot suppression is performed on the step input of the transfer function model through dynamic response characteristics to generate a segmented frequency adjustment amount sequence; According to the relationship between the cumulative interval of the integral component in the segmented frequency adjustment sequence and the absolute value of the real-time wind speed deviation, the target integral component exceeding the preset deviation threshold is frozen through the integral separation strategy to generate an anti-saturation corrected frequency adjustment; The anti-saturation correction frequency adjustment amount is superimposed on the current fan frequency, and the superimposed result is dynamically compensated and corrected through a feedforward compensation model to generate a real-time adjustment parameter of the fan frequency.

[0009] Optionally, in a fifth implementation of the first aspect of the present application, the method further includes: Based on the real-time adjustment parameters and historical equipment operation data, a digital twin model synchronized with the physical equipment state is constructed through a time domain synchronization algorithm; wherein the input parameters of the digital twin model include real-time wind speed and wind turbine current harmonic distortion rate; Performing frequency domain energy feature extraction on the device vibration signal output by the digital twin model to generate a corresponding feature vector; Determining a fault type that matches the feature vector using a preset fault prediction model; Matching the predicted deviation of the real-time wind speed with the historical wind speed deviation sequence corresponding to the fault type, and performing weighted correction on the matching result in combination with the wind turbine current harmonic distortion rate to generate a comprehensive fault risk index; When the comprehensive failure risk index exceeds a preset risk threshold, the early warning mechanism of the smart device is triggered.

[0010] Optionally, in a sixth implementation of the first aspect of the present application, the method further includes: Extracting time-frequency domain degradation features of the fan bearing vibration signal based on the multidimensional feature vector and historical equipment degradation data to generate a degradation feature vector; Inputting the degradation feature vector into a pre-trained degradation prediction model to generate a confidence interval prediction value of the remaining life of the wind turbine bearing; According to the weight distribution relationship between the confidence interval prediction value and the preset maintenance cost model, the maintenance cycle and frequency adjustment parameters are jointly optimized to generate a dynamic maintenance decision matrix; A gradient maintenance instruction is triggered according to the risk priority parameter in the dynamic maintenance decision matrix, and the wind turbine bearing is maintained through the gradient maintenance instruction.

[0011] A second aspect of the present application provides an intelligent frequency conversion control device, the intelligent frequency conversion control device comprising: The acquisition module is used to collect the environmental parameter set and device operation data of the smart device based on the preset multi-dimensional sensors; a processing module, configured to generate a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the device operation data; A construction module, configured to construct a neural network wind speed prediction model based on the multidimensional feature vector; A first adjustment module is used to dynamically adjust the connection path of the hidden layer nodes of the neural network wind speed prediction model through an ant colony algorithm to generate a dynamic wind speed setting value within a prediction time period; A second adjustment module is used to dynamically adjust the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate through a variable domain expansion factor of fuzzy control, and generate a PID parameter adjustment instruction in combination with a fuzzy rule base; A generating module is used to generate a real-time adjustment parameter of the fan frequency of the smart device according to the PID parameter adjustment instruction.

[0012] A third aspect of an embodiment of the present application provides an electronic device, comprising a memory and a processor, wherein the processor is used to execute a computer program stored on the memory. When the processor executes the computer program, it implements the steps of the intelligent frequency conversion control method provided in the first aspect of the embodiment of the present application.

[0013] The fourth aspect of the embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the intelligent frequency conversion control method provided in the first aspect of the embodiment of the present application are implemented.

[0014] In summary, according to the intelligent variable frequency control method and system provided by the present application, the environmental parameter set and device operation data of the intelligent device are collected according to the preset multi-dimensional sensor; the multi-dimensional parameters of the environmental parameter set and the device operation data are standardized to generate a multi-dimensional feature vector; a neural network wind speed prediction model is constructed according to the multi-dimensional feature vector; the hidden layer node connection path of the neural network wind speed prediction model is dynamically adjusted by the ant colony algorithm to generate a dynamic wind speed set value within the prediction time period; the wind speed deviation and the domain range of the deviation change rate between the dynamic wind speed set value and the real-time wind speed are dynamically adjusted by the variable domain scaling factor of fuzzy control, and a PID parameter adjustment instruction is generated in combination with the fuzzy rule base; the real-time adjustment parameter of the fan frequency of the intelligent device is generated according to the PID parameter adjustment instruction. Through the implementation of the present application, the hidden layer structure of the neural network is optimized by the ant colony algorithm to improve the wind speed prediction accuracy, and the variable domain fuzzy PID control is introduced to dynamically adjust the PID parameters under different wind speed deviations to ensure accurate adjustment of the fan frequency, thereby meeting the dynamically changing ventilation needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of the intelligent frequency conversion control method provided in an embodiment of the present application; Figure 2 A schematic diagram of a program module of an intelligent variable frequency control device provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of this application.

[0017] In order to solve the problem that the fixed frequency or simple variable frequency control strategy in the related art cannot meet the dynamic changes in future ventilation needs, the embodiment of the present application provides an intelligent variable frequency control method, such as Figure 1 This is a flow chart of the intelligent frequency conversion control method provided in this embodiment, which includes the following steps: Step 110: Collect the environmental parameter set and device operation data of the smart device according to the preset multi-dimensional sensor.

[0018] Specifically, pre-set multi-dimensional sensors are used to collect environmental parameter sets and equipment operation data. This involves deploying multiple sensors at the exhaust inlet, purification unit outlet, and fan shaft end, including but not limited to wind speed sensors, temperature and humidity sensors, load current sensors, and oil fume concentration sensors. Real-time sensor data collection is performed through a data acquisition module, using high-frequency sampling technology to improve data accuracy. A time synchronization mechanism is also used to align the data from each sensor to ensure temporal consistency, thereby obtaining a complete set of environmental parameter and equipment operation data.

[0019] Step 120 : Generate a multi-dimensional feature vector by standardizing the environmental parameter set and the multi-dimensional parameters of the equipment operation data.

[0020] Specifically, the collected environmental parameter sets and equipment operation data are standardized to generate a multidimensional feature vector. To ensure the validity of the data, the correlation between each parameter and the system energy consumption and purification efficiency is first calculated. Grey correlation analysis or mutual information calculation methods are used to screen out the core parameter set with a correlation higher than the set threshold. Based on sliding window filtering, the core parameter set is de-noised and a denoised standardized data matrix is ​​constructed. Furthermore, the data matrix is ​​subjected to dimensionality reduction processing using the covariance matrix eigenvalue decomposition method to reduce redundant features and improve computational efficiency. Finally, the low-dimensional feature vector after dimensionality reduction is cross-validated using a redundant sensor group, and abnormal data is replaced or interpolated to correct it, ultimately generating a multidimensional feature vector with a high signal-to-noise ratio.

[0021] In an optional implementation of this embodiment, the step of generating a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the equipment operation data includes: analyzing the correlation between the multidimensional parameters and the system energy consumption and purification efficiency, and determining a core parameter set whose correlation is higher than a preset threshold; performing noise elimination on the core parameter set based on sliding window filtering to generate a denoised standardized data matrix; performing dimensionality reduction processing on the redundant dimensions of the noisy standardized data matrix through eigenvalue decomposition of the covariance matrix; cross-validating the low-dimensional feature vector after dimensionality reduction processing through a redundant sensor group, and replacing or interpolating and correcting abnormal data according to the verification results to generate the multidimensional feature vector.

[0022] Specifically, the correlation between multi-dimensional parameters and system energy consumption and purification efficiency is analyzed. The grey correlation analysis method is first used for calculation. The influence of each parameter on the overall performance of the system is quantified by comparing the similarities between different parameter sequences. Parameters with high correlation are considered to have a significant impact on the system energy consumption and purification efficiency, thus forming a core parameter set. For example, in actual applications, by collecting data such as wind speed, temperature and humidity, load current and oil fume concentration, it can be found that changes in oil fume concentration and wind speed are closely related to fluctuations in purification efficiency. Their correlation values ​​exceed the preset threshold, and therefore they are selected into the core parameter set. This analysis can not only filter out variables with weak correlation with system performance, but also simplify the complexity of subsequent data processing. Next, by defining a fixed-length window on the data sequence and performing mean or weighted average calculations on the data within the window, the random noise in the signal is smoothed, and a denoised standardized data matrix is ​​generated. The significance of standardization is to eliminate the influence between different dimensions, so that the data distribution of each parameter is on a similar scale, which is convenient for subsequent algorithm processing. For example, filtering the wind speed data within a certain window can reduce the error caused by instantaneous wind speed fluctuations, making the data more stable. The denoised standardized data matrix is ​​then reduced in dimension by eigenvalue decomposition of the covariance matrix. By calculating the covariance matrix between the variables and performing eigenvalue decomposition on the matrix, the main eigencomponents are selected to retain most of the data information, thereby reducing the data dimension, eliminating data redundancy, and reducing the computational burden. For example, when there is a high correlation between some variables in the multidimensional features, the eigenvalue decomposition can compress this redundant information into one or a small number of principal components, which not only retains the main features of the original data but also greatly improves data processing efficiency. After dimensionality reduction, in order to ensure that no errors are introduced or key information is omitted during the dimensionality reduction process, a redundant sensor group is used for cross-validation. By acquiring data from multiple sensors at different locations or using the same measurement principle, the low-dimensional feature vector is verified by using the similarity of the data between sensors, and replacement or interpolation correction is performed when abnormal data is found. For example, if the wind speed data after dimensionality reduction shows abnormal values ​​that deviate significantly from the normal range, by comparing the data with other sensors in the redundant sensor group, it can be determined whether the data is abnormal, and reasonable interpolation correction can be performed using adjacent data or historical trends to ensure that the final generated multi-dimensional feature vector has a high signal-to-noise ratio and high reliability.

[0023] Through systematic data preprocessing, the original collected multi-dimensional data undergoes correlation analysis, noise filtering, dimensionality reduction and data verification, and finally generates a multi-dimensional feature vector that meets the actual operating conditions and accurately reflects the system status, thereby providing a solid data foundation for subsequent neural network prediction, control strategy formulation and fault prediction, ensuring that the entire system can achieve precise real-time control and efficient operation even under complex working conditions.

[0024] Step 130: Construct a neural network wind speed prediction model based on the multi-dimensional feature vector.

[0025] Step 140: Dynamically adjust the connection paths of the hidden layer nodes of the neural network wind speed prediction model through the ant colony algorithm to generate a dynamic wind speed setting value within the prediction time period.

[0026] Specifically, based on multi-dimensional feature vectors, a neural network wind speed prediction model is constructed to achieve accurate prediction of wind speed demand in future time periods. First, the basic architecture of the neural network is determined using the training data set, including the number of nodes and connection weights in the input layer, hidden layer, and output layer. In order to prevent the model from falling into local optimality, a global search is performed on the initial weights of the neural network, and the initial weight matrix is ​​optimized through selection, crossover, and mutation operations in the genetic algorithm. Based on the optimized initial weight matrix, the ant colony algorithm is further introduced to iteratively optimize the connection paths of the hidden layer nodes under the dynamic update mechanism of the pheromone concentration distribution. The pheromone volatilization factor and the path selection probability function are used to screen the optimal connection path, and finally generate the hidden layer node connection path adjustment instruction, and the real-time collected environmental parameters are input into the adjusted neural network wind speed prediction model. After forward propagation calculation, the dynamic wind speed set value within the prediction time period is generated.

[0027] In an optional implementation of this embodiment, the step of dynamically adjusting the hidden layer node connection paths of the neural network wind speed prediction model through an ant colony algorithm to generate a dynamic wind speed set value within a prediction time period includes: performing a global search on the initial weights of the neural network wind speed prediction model, and generating an optimized initial weight matrix through selection, crossover and mutation operations; initializing the pheromone concentration distribution of the ant colony algorithm based on the optimized initial weight matrix, and iteratively screening the connection paths between the hidden layer nodes through the pheromone volatility factor and the path selection probability function; dynamically updating the weight distribution of the connection paths according to the pheromone concentration between the hidden layer nodes, and generating a hidden layer node connection path adjustment instruction; inputting the real-time collected environmental parameters into the adjusted neural network wind speed prediction model, and generating the dynamic wind speed set value within the prediction time period through forward propagation calculation.

[0028] Specifically, when conducting a global search for the initial weights of the neural network wind speed prediction model, a global optimization strategy based on a genetic algorithm is first adopted. A randomly generated initial weight matrix is ​​used as the search space. Through selection, crossover, and mutation operations, the weight matrix is ​​searched for initial parameters that are more suitable for the current prediction task. This allows the neural network to converge faster and avoid falling into local optimality, thereby providing an optimized initial weight matrix for subsequent iterations. For example, in a training set of simulated wind speed data, the optimized initial weights can better reflect the inherent distribution characteristics of the data and shorten the model training time. Next, the ant colony algorithm uses this optimized initial weight matrix to initialize the pheromone concentration distribution. The connections between each hidden layer node are considered as information transmission paths. Multiple rounds of iterative screening are performed under the influence of the pheromone volatility factor and the path selection probability function. This process can screen out more optimal node connection combinations from a large number of possible connection methods, thereby establishing a more reasonable structural mapping relationship within the neural network, ensuring that the wind speed prediction model responds more quickly to input data. Subsequently, by dynamically updating the pheromone concentration between hidden layer nodes, the connection weights are redistributed based on the performance of each path in historical predictions, and corresponding connection path adjustment instructions are generated, so that the connection structure within the neural network is continuously optimized to adapt to real-time changes in environmental parameters. For example, in a certain round of iterations, if the pheromone concentration of a path decreases significantly, the information transmission function of this path will be weakened, and the path with a higher pheromone concentration will receive a greater weight, thus occupying a dominant position in the overall network. The process of dynamically updating pheromone concentration can be expressed by the following formula: , in, represents the pheromone concentration between hidden layer node i and node j at time t, which reflects the quality of the connection structure between the two nodes, that is, the effectiveness of information transmission in the network; The pheromone volatilization factor is used to describe the rate at which pheromones decay naturally over time. Its existence ensures that old and possibly outdated connection information is gradually forgotten, allowing the system to respond to environmental changes. is the performance deviation of the connection path at the current moment, is the weight factor of the connection path, which reflects the importance or basic reliability of the connection path in the overall prediction performance of the smart device. A high weight value means that the path shows good stability in historical data. represents the incremental replenishment of pheromones based on current prediction performance, used to inject real-time improved information into node connections, thereby strengthening high-performing paths in subsequent iterations. The significance of this formula lies in its comprehensive consideration of the natural attenuation of pheromones and dynamic replenishment based on prediction errors, ensuring that high-quality connection paths are strengthened in each iteration, while poorer paths are gradually weakened, thereby achieving adaptive adjustment of the global optimal connection structure. Finally, the real-time collected environmental parameters are used as input, and forward propagation calculations are performed through the adjusted neural network model. Utilizing nonlinear mapping between network layers, the dynamic wind speed setpoint within the prediction time period is output, allowing intelligent devices to update wind speed targets in real-time based on real-time environmental changes, ensuring efficient and stable operation of the system under complex and changing working conditions. The entire process organically combines the global search of initial weights, the dynamic connection path adjustment of the ant colony algorithm, and the forward propagation prediction, achieving accurate modeling and real-time prediction of complex wind speed variation environments.

[0029] It should be noted that pheromone concentration is a metric used in the ant colony algorithm to measure the quality of the connection between any two nodes in the hidden layer of a neural network. A higher concentration indicates a greater contribution of the connection to the wind speed prediction model's performance during previous iterations, reflecting the superiority of the information transmission path. A lower concentration, on the other hand, indicates poor performance or instability of the connection. A path represents the information transmission route between two nodes in the hidden layer of a neural network. Specifically, it represents the connection relationships and weight combinations between each node. These paths together constitute the information flow structure within the neural network. By dynamically screening and optimizing these paths, a network structure more suitable for real-time wind speed prediction can be constructed, thereby achieving more accurate variable frequency control for intelligent devices. A node is an independent computational unit in a neural network, often also called a neuron. Each node receives input signals, performs a weighted sum on the input data, and performs a nonlinear transformation using an activation function to generate an output signal. In intelligent variable frequency control, these nodes constitute the information processing layer within the neural network. The connections between them determine the data propagation path within the network, ultimately affecting the accuracy of wind speed prediction. By dynamically optimizing the pheromone concentration of the connection paths between nodes, the paths that contribute more to the prediction performance can be screened out, thereby optimizing the structure and performance of the entire neural network.

[0030] Step 150 : Dynamically adjust the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate through the variable universe expansion factor of fuzzy control, and generate a PID parameter adjustment instruction in combination with the fuzzy rule base.

[0031] Specifically, the deviation and deviation change rate between the dynamic wind speed set value and the real-time wind speed are dynamically adjusted through the variable domain expansion factor of fuzzy control, and the PID parameter adjustment instructions are generated in combination with the fuzzy rule base. First, the difference between the dynamic wind speed set value and the real-time wind speed is calculated to obtain the wind speed deviation. Subsequently, the wind speed deviation is differentially calculated through a sliding time window to obtain the deviation change rate. Based on the relationship between the absolute value of the wind speed deviation and the preset deviation threshold, a piecewise linear function is used to dynamically adjust the domain range of the variable domain expansion factor. The adjusted domain range is input into the fuzzy rule base together with the wind speed deviation and the deviation change rate sequence, and a fuzzy output set is generated through membership function mapping. Finally, the fuzzy output set is defuzzified and combined with the current PID control parameters to calculate the PID parameter adjustment instructions.

[0032] In an optional implementation of this embodiment, the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate are dynamically adjusted through the variable domain scaling factor of fuzzy control, and the step of generating a PID parameter adjustment instruction in combination with the fuzzy rule base includes: calculating the wind speed deviation according to the difference between the dynamic wind speed set value and the real-time wind speed; performing a differential operation on the wind speed deviation through a sliding time window to generate a deviation change rate; dynamically adjusting the domain range of the variable domain scaling factor based on the relationship between the absolute value of the wind speed deviation and a preset deviation threshold; inputting the wind speed deviation, the deviation change rate sequence and the adjusted domain range into the fuzzy rule base, and generating a fuzzy output set through membership function mapping; defuzzifying the fuzzy output set, and generating a PID parameter adjustment instruction in combination with the current parameter value of the PID controller.

[0033] Specifically, the calculation is performed based on the difference between the dynamic wind speed setting value and the real-time wind speed. The difference between the two is first defined as the wind speed deviation. This value reflects the error between the current system setting and the actual measurement. For example, in an actual application scenario, if the wind speed setting value is 3.5m / s and the real-time measurement value is 3.0m / s, then the deviation is 0.5m / s. This value can intuitively represent the difference in the operation of the system at a specific moment. Then, by introducing the sliding time window technology, the continuously collected wind speed deviation is differentially calculated to generate the deviation change rate. This process uses the change in wind speed deviation at previous and subsequent time points to quantitatively describe the upward or downward trend of the wind speed deviation. It is of great significance for judging the response speed of the system. For example, if the continuously collected data shows that the deviation increases sharply in a short period of time, it indicates that there may be a problem of unbalanced regulation. Subsequently, the absolute value of the wind speed deviation is compared with the preset deviation threshold, and the domain range of the variable domain scaling factor is dynamically adjusted through a piecewise linear function. The variable domain scaling factor is a control parameter that dynamically adjusts the domain range of the input variable. In the fuzzy control system, it is used to correct the mapping relationship between the dependent variable and the independent variable. Its function is to allow the input variable to vary within a wider range to obtain a smooth response when the error between the actual measured value and the set value (such as the wind speed deviation) is small; and when the error is large, by narrowing the input domain, the output of the fuzzy control is made more refined, avoiding excessive control amplitude, so that the entire control system can maintain stability and response accuracy under nonlinear and uncertain conditions. The domain refers to the range of values ​​that a variable or parameter can take, and the variable domain means that the range of the domain can be dynamically adjusted rather than fixed. The function for the dynamic adjustment of the variable domain scaling factor can be expressed as: , in, Indicates the absolute value of the wind speed deviation at the current moment, T is the preset deviation threshold, To adjust the shape parameters, this formula dynamically generates a scaling factor based on the deviation size, which is used to adjust the range of the fuzzy control input domain. Next, the calculated wind speed deviation, the deviation change rate sequence, and the adjusted domain range are used as input and imported into a pre-built fuzzy rule base. The fuzzy rule base is a set of logical rules constructed based on experience and experimental data. It maps precise numerical values ​​into fuzzy sets through membership functions. This process can handle the uncertainty and nonlinear relationships in the system, so that each input variable is mapped through different membership functions to form a fuzzy output set. For example, triangular or trapezoidal membership functions can be used to divide wind speed deviations, so that fuzzy quantities such as small, medium, and large are clearly defined. After obtaining the fuzzy output set, it is converted into specific PID parameter adjustment instructions through defuzzification operation. The defuzzification process adopts the center of gravity method or the maximum membership method to comprehensively consider the weight information in the fuzzy set and combine it with the current parameter value of the PID controller to generate a PID adjustment instruction with real-time adjustment function. The function of the instruction is to change the proportional, integral and differential components so that the system can achieve faster and more stable response when facing wind speed changes; for example, in the case of large wind speed fluctuations, the defuzzified adjustment instruction may increase the proportional coefficient in the PID controller to enhance the system's immediate response to the error, thereby achieving the purpose of smoothly controlling the fan frequency.

[0034] Step 160: Generate a real-time adjustment parameter for the fan frequency of the smart device according to the PID parameter adjustment instruction.

[0035] Specifically, after obtaining the PID parameter adjustment instruction, the real-time adjustment parameters of the fan frequency are further generated to drive the fan to achieve precise variable frequency control. First, the proportional, integral and differential components in the PID parameter adjustment instruction are weighted decoupled to construct a parameter adjustment matrix. Based on the parameter adjustment matrix, a transfer function model is constructed, and through dynamic response characteristic analysis, the step input of the transfer function model is overshoot suppressed to generate a segmented frequency adjustment amount sequence. According to the relationship between the cumulative interval of the integral component in the frequency adjustment amount sequence and the absolute value of the real-time wind speed deviation, an integral separation strategy is adopted to freeze the target integral component that exceeds the preset deviation threshold to generate an anti-saturation correction frequency adjustment amount. Finally, the anti-saturation correction frequency adjustment amount is superimposed on the current fan frequency, and the superimposed result is dynamically compensated and corrected through the feedforward compensation model to finally generate the real-time adjustment parameters of the fan frequency.

[0036] In an optional implementation of this embodiment, the step of generating a real-time adjustment parameter of the fan frequency of the smart device according to the PID parameter adjustment instruction includes: performing weighted decoupling on the proportional, integral and differential components in the PID parameter adjustment instruction to generate a parameter adjustment matrix corresponding to each control component; constructing a transfer function model based on the parameter adjustment matrix, and suppressing the overshoot of the step input of the transfer function model through dynamic response characteristics to generate a segmented frequency adjustment amount sequence; freezing the target integral component that exceeds the preset deviation threshold through an integral separation strategy according to the relationship between the cumulative interval of the integral component in the segmented frequency adjustment amount sequence and the absolute value of the real-time wind speed deviation to generate an anti-saturation correction frequency adjustment amount; superimposing the anti-saturation correction frequency adjustment amount with the current fan frequency, and dynamically compensating and correcting the superimposed result through a feedforward compensation model to generate the real-time adjustment parameter of the fan frequency.

[0037] Specifically, by weighting the proportional, integral, and differential components in the PID parameter adjustment command, each control component is first treated as an independent control variable, and a corresponding parameter adjustment matrix is ​​constructed. This matrix separates the influence of the proportional, integral, and differential control parts, and then independently calculates the weight of each component. The significance of this is to clarify the independent contribution of each control component to the overall control process. For example, when the wind speed deviation is large, the proportional component may require a higher weight to quickly respond to the error, while the integral component needs to control its cumulative effect to prevent integral windup. Next, based on the constructed parameter adjustment matrix, a transfer function model is established. This transfer function describes the dynamic relationship between the system's input command and output response. The transfer function model is usually expressed using the Laplace transform, such as the process of converting the input signal into the output response. By analyzing the dynamic response characteristics, a step input is applied to the transfer function model. With the help of advanced dynamic overshoot suppression technology, a segmented frequency adjustment amount sequence is generated. The purpose is to reduce the overshoot caused by excessively fast response and ensure a smooth system response without violent oscillation. For example, in a step response test, if the output response exceeds the target value, the segmented adjustment sequence will reduce the adjustment force within the overshoot interval to smoothly transition to a steady-state response. Subsequently, based on the relationship between the cumulative interval of the integral component in the segmented frequency adjustment sequence and the absolute value of the real-time wind speed deviation, an integral separation strategy is introduced to freeze the integral part when it exceeds the preset deviation threshold. That is, when there is a risk of potential overcompensation due to excessive integral accumulation, this part of the regulation is frozen, and an anti-saturation corrected frequency adjustment is generated. This strategy can prevent the integral link from having an excessive cumulative effect on the system, thereby maintaining the stability of the system. For example, when it is detected that the real-time wind speed deviation is greater than a certain value, the corresponding integral accumulation value will no longer continue to increase, but will remain within a safe range. Finally, the obtained anti-saturation correction frequency adjustment is superimposed on the current fan frequency, and the feedforward compensation model is used to perform dynamic compensation correction on the superimposed result. The role of the feedforward compensation model here is to consider in advance the impact of external disturbances and dynamic load changes on the system, and to generate real-time adjustment parameters for the fan frequency by compensating for expected changes in advance. This process can not only make up for the shortcomings caused by closed-loop control delays, but also achieve more accurate frequency regulation. Overall, from the independent decoupling of PID control parameters, to the use of transfer function models to achieve regulation of step input response, to the generation of anti-saturation correction based on the integral separation strategy, and combined with feedforward compensation for the final dynamic correction, each link is designed to ensure that the intelligent device can achieve high-precision, stable and fast real-time frequency adjustment when facing wind speed fluctuations and load changes, thereby ensuring the efficient operation of the system under complex working conditions.

[0038] In an optional implementation of this embodiment, based on the real-time adjustment parameters and the historical operation data of the equipment, a digital twin model synchronized with the physical equipment state is constructed through a time domain synchronization algorithm; the frequency domain energy characteristics of the equipment vibration signal output by the digital twin model are extracted to generate a corresponding feature vector; the fault type matching the feature vector is determined through a preset fault prediction model; the predicted deviation of the real-time wind speed is matched with the historical wind speed deviation sequence corresponding to the fault type, and the matching result is weighted and corrected in combination with the wind turbine current harmonic distortion rate to generate a comprehensive fault risk index; when the comprehensive fault risk index exceeds the preset risk threshold, the early warning mechanism of the smart device is triggered.

[0039] Specifically, a time-domain synchronization algorithm is used to build a digital twin model based on real-time adjustment parameters and historical equipment operating data, ensuring synchronization with the physical equipment status. The core of this algorithm is to align operating data at different time points to ensure that the physical equipment status can be mapped to the digital twin in real time. Key parameters such as real-time wind speed and wind turbine current harmonic distortion (HTD) are used as input variables to construct a virtual model reflecting the wind turbine's operating status. HTD is an important indicator of motor stability and can reveal potential problems such as internal electromagnetic imbalance, winding aging, or mechanical failure. For example, when a wind turbine's load changes due to blade oil accumulation or bearing wear, the motor's HTD fluctuates significantly. This information, captured by the digital twin model, can be used to predict wind turbine operating trends. The constructed digital twin model outputs a vibration signal from the equipment, which is then subjected to frequency-domain energy feature extraction to obtain a feature vector. This frequency-domain energy feature extraction process uses a Fourier transform to convert the vibration signal from the time domain to the frequency domain, enabling a clearer representation of the vibration characteristics under different fault modes. For example, a bearing fault may produce a significant energy peak within a specific frequency range, while an unbalanced wind turbine blade may result in an increase in the energy of low-frequency components. The extracted feature vector is then input into a fault prediction model to determine the fault type that matches it. The fault prediction model uses a pretrained neural network or pattern matching algorithm to compare the current feature vector with historical fault data to accurately identify the fault type that may correspond to the current equipment status. For example, in the case of a bearing fault, if the spectral components of the feature vector closely match the typical vibration patterns of bearing wear, the model determines that the current equipment has a bearing anomaly. Furthermore, the predicted deviation of real-time wind speed is matched with the historical wind speed deviation sequence corresponding to the fault type. The matching results are weighted and corrected using the wind turbine current harmonic distortion rate to generate a comprehensive fault risk index. Wind speed prediction deviation reflects the difference between the system's ideal control conditions and actual operating conditions. Different types of faults have different impacts on wind speed stability, so comparison with historical data is necessary to determine whether the current wind speed deviation shows an abnormal trend. For example, blade contamination may cause a decrease in wind speed, while motor aging may lead to increased wind speed fluctuations. Therefore, a comprehensive analysis combining historical data and harmonic distortion rate is necessary. Ultimately, when the comprehensive failure risk index exceeds a preset risk threshold, the system triggers an early warning mechanism, enabling timely maintenance measures. This mechanism can take various forms, such as an alarm, automatic adjustment of turbine operating parameters, or maintenance notifications. This ensures that the smart device can take effective preventative measures before a failure occurs, thereby extending the equipment's lifespan and reducing operating costs.

[0040] In an optional implementation of this embodiment, based on the multidimensional feature vector and historical equipment degradation data, the time-frequency domain degradation characteristics of the fan bearing vibration signal are extracted to generate a degradation feature vector; the degradation feature vector is input into a pre-trained degradation prediction model to generate a confidence interval prediction value of the remaining life of the fan bearing; according to the weight distribution relationship between the confidence interval prediction value and the preset maintenance cost model, the maintenance cycle and frequency adjustment parameters are jointly optimized to generate a dynamic maintenance decision matrix; according to the risk priority parameter in the dynamic maintenance decision matrix, a gradient maintenance instruction is triggered, and the fan bearing is maintained through the gradient maintenance instruction.

[0041] Specifically, based on multidimensional feature vectors and historical equipment degradation data, the wind turbine bearing vibration signal is analyzed in the time-frequency domain to extract degradation features. Using methods such as short-time Fourier transforms or wavelet transforms, the time-frequency domain degradation features convert the vibration signal from the time domain to the joint time-frequency domain, enabling identification of characteristic changes under different degradation conditions, such as bearing wear, poor lubrication, or fatigue damage. For example, in the early stages of bearing wear, the vibration signal may exhibit low-amplitude, high-frequency components within a specific frequency range. In the advanced degradation stage, the energy of the impact signal is concentrated in the mid- and high-frequency bands. Therefore, analysis of the time-frequency distribution helps accurately determine the degree of bearing degradation. By extracting multiple features, such as energy entropy, kurtosis factor, and envelope spectrum, a degradation feature vector describing the bearing's health status is generated. This vector is then fed into a pre-trained degradation prediction model. The degradation prediction model learns the evolution of bearing wear from long-term operating data and, combined with the current degradation feature vector, calculates a confidence interval prediction for the wind turbine bearing's remaining life. This prediction typically consists of an upper confidence limit, a lower confidence limit, and an expected lifespan value, which measure the uncertainty of bearing degradation. For example, under conditions of good lubrication and stable loads, the predicted remaining life of a bearing may be concentrated within a narrow range. However, under high loads or harsh environments, uncertainty increases, and the prediction interval widens. Based on the confidence interval predictions and a preset maintenance cost model, the maintenance cycle and wind turbine frequency adjustment parameters are jointly optimized through weight distribution to find the optimal balance between equipment life and maintenance costs. For example, if the predicted remaining life approaches a certain maintenance critical point, the wind turbine speed is appropriately adjusted to slow the bearing degradation rate, thereby extending the maintenance interval and avoiding the additional costs caused by excessive maintenance. This optimization process generates a dynamic maintenance decision matrix that includes maintenance requirements, risk levels, and adjusted maintenance cycles for different time periods. Ultimately, based on the risk priority parameters in this matrix, a gradient maintenance instruction is triggered, and graded maintenance is performed on the wind turbine bearings accordingly. The core idea of ​​gradient maintenance instructions is to adopt different levels of maintenance measures based on the degree of bearing degradation. For example, in the early stages of bearing degradation, only lubrication optimization is performed, while when degradation worsens, preventive replacement or deep maintenance is arranged. This minimizes maintenance costs and improves system operating efficiency while ensuring safe equipment operation.

[0042] According to the intelligent variable frequency control method provided by the present application, the environmental parameter set and device operation data of the intelligent device are collected based on a preset multi-dimensional sensor; a multi-dimensional feature vector is generated by processing the multi-dimensional parameters of the environmental parameter set and the device operation data; a neural network wind speed prediction model is constructed based on the multi-dimensional feature vector; the connection path of the hidden layer nodes of the wind speed prediction model is dynamically adjusted to generate a dynamic wind speed set value; the domain range of the wind speed deviation and the deviation change rate between the dynamic wind speed set value and the real-time wind speed is dynamically adjusted, and a PID parameter adjustment instruction is generated in combination with a fuzzy rule base; and the real-time adjustment parameters of the fan frequency of the intelligent device are generated based on the PID parameter adjustment instruction. The PID parameters can be dynamically adjusted under different wind speed deviations to ensure accurate adjustment of the fan frequency, thereby meeting dynamically changing ventilation needs.

[0043] Figure 2 The embodiment of the present application provides an intelligent frequency conversion control device, which can be used to implement the intelligent frequency conversion control method in the above embodiment. Figure 2 As shown, the intelligent frequency conversion control device mainly includes: The acquisition module 10 is used to collect the environmental parameter set and device operation data of the smart device based on the preset multi-dimensional sensors; A processing module 20 is configured to generate a multi-dimensional feature vector by normalizing the environmental parameter set and the multi-dimensional parameters of the device operation data; A construction module 30 is used to construct a neural network wind speed prediction model based on the multidimensional feature vector; A first adjustment module 40 is configured to dynamically adjust the connection paths of the hidden layer nodes of the neural network wind speed prediction model by using an ant colony algorithm to generate a dynamic wind speed setting value within a prediction time period; A second adjustment module 50 is configured to dynamically adjust the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate by using a variable domain expansion factor of fuzzy control, and generate a PID parameter adjustment instruction in combination with a fuzzy rule base; The generating module 60 is configured to generate a real-time adjustment parameter of the fan frequency of the smart device according to the PID parameter adjustment instruction.

[0044] In an optional implementation of this embodiment, the processing module is specifically used to: analyze the correlation between the multidimensional parameters and the system energy consumption and purification efficiency, and determine a core parameter set whose correlation is higher than a preset threshold; perform noise elimination on the core parameter set based on sliding window filtering to generate a denoised standardized data matrix; perform dimensionality reduction processing on the redundant dimensions of the noisy standardized data matrix through eigenvalue decomposition of the covariance matrix; perform cross-validation on the low-dimensional feature vector after dimensionality reduction processing through a redundant sensor group, and replace or interpolate and correct abnormal data according to the verification results to generate the multidimensional feature vector.

[0045] In an optional implementation of this embodiment, the first adjustment module is specifically used to: perform a global search on the initial weights of the neural network wind speed prediction model, and generate an optimized initial weight matrix through selection, crossover and mutation operations; initialize the pheromone concentration distribution of the ant colony algorithm based on the optimized initial weight matrix, and iteratively screen the connection paths between the hidden layer nodes through the pheromone volatility factor and the path selection probability function; dynamically update the weight distribution of the connection paths according to the pheromone concentration between the hidden layer nodes, and generate a hidden layer node connection path adjustment instruction; input the real-time collected environmental parameters into the adjusted neural network wind speed prediction model, and generate a dynamic wind speed set value within the prediction time period through forward propagation calculation.

[0046] In an optional implementation of this embodiment, the second adjustment module is specifically used to: calculate the wind speed deviation based on the difference between the dynamic wind speed set value and the real-time wind speed; perform differential operation on the wind speed deviation through a sliding time window to generate a deviation change rate; dynamically adjust the domain range of the variable domain scaling factor based on the relationship between the absolute value of the wind speed deviation and a preset deviation threshold; input the wind speed deviation, the deviation change rate sequence and the adjusted domain range into a fuzzy rule base, and generate a fuzzy output set through membership function mapping; defuzzify the fuzzy output set, and generate a PID parameter adjustment instruction in combination with the current parameter value of the PID controller.

[0047] In an optional implementation of this embodiment, the generation module is specifically used to: perform weighted decoupling on the proportional, integral and differential components in the PID parameter adjustment instruction to generate a parameter adjustment matrix corresponding to each control component; construct a transfer function model based on the parameter adjustment matrix, and suppress overshoot of the step input of the transfer function model through dynamic response characteristics to generate a segmented frequency adjustment amount sequence; according to the relationship between the cumulative interval of the integral component in the segmented frequency adjustment amount sequence and the absolute value of the real-time wind speed deviation, freeze the target integral component that exceeds the preset deviation threshold through the integral separation strategy to generate an anti-saturation correction frequency adjustment amount; superimpose the anti-saturation correction frequency adjustment amount on the current fan frequency, and dynamically compensate and correct the superposition result through the feedforward compensation model to generate the real-time adjustment parameter of the fan frequency.

[0048] In an optional implementation of this embodiment, the intelligent variable frequency control device also includes an early warning module. The early warning module is used to: construct a digital twin model synchronized with the physical device state based on the real-time adjustment parameters and the historical operation data of the device through a time domain synchronization algorithm; wherein the input parameters of the digital twin model include real-time wind speed and wind turbine current harmonic distortion rate; extract frequency domain energy features of the device vibration signal output by the digital twin model to generate a corresponding feature vector; determine the fault type matching the feature vector through a preset fault prediction model; match the predicted deviation of the real-time wind speed with the historical wind speed deviation sequence corresponding to the fault type, and perform weighted correction on the matching result in combination with the wind turbine current harmonic distortion rate to generate a comprehensive fault risk index; when the comprehensive fault risk index exceeds the preset risk threshold, trigger the early warning mechanism of the intelligent device.

[0049] In an optional implementation of this embodiment, the intelligent variable frequency control device further includes a maintenance module. The maintenance module is configured to: extract the time-frequency domain degradation characteristics of the fan bearing vibration signal based on the multidimensional feature vector and historical equipment degradation data to generate a degradation feature vector; input the degradation feature vector into a pre-trained degradation prediction model to generate a confidence interval prediction value for the remaining life of the fan bearing; jointly optimize the maintenance cycle and frequency adjustment parameters based on the weight distribution relationship between the confidence interval prediction value and a preset maintenance cost model to generate a dynamic maintenance decision matrix; trigger a gradient maintenance instruction based on the risk priority parameter in the dynamic maintenance decision matrix, and perform maintenance on the fan bearing through the gradient maintenance instruction.

[0050] According to the intelligent variable frequency control device provided by the present application, the environmental parameter set and device operation data of the intelligent device are collected based on a preset multi-dimensional sensor; a multi-dimensional feature vector is generated by processing the multi-dimensional parameters of the environmental parameter set and the device operation data; a neural network wind speed prediction model is constructed based on the multi-dimensional feature vector; the connection path of the hidden layer nodes of the wind speed prediction model is dynamically adjusted to generate a dynamic wind speed set value; the domain range of the wind speed deviation and the deviation change rate between the dynamic wind speed set value and the real-time wind speed is dynamically adjusted, and a PID parameter adjustment instruction is generated in combination with a fuzzy rule base; and the real-time adjustment parameters of the fan frequency of the intelligent device are generated based on the PID parameter adjustment instruction. The PID parameters can be dynamically adjusted under different wind speed deviations to ensure accurate adjustment of the fan frequency, thereby meeting dynamically changing ventilation needs.

[0051] According to the application plan Figure 3 An electronic device provided in an embodiment of the present application can be used to implement the intelligent frequency conversion control method in the aforementioned embodiment, mainly comprising: Memory 301, processor 302, and computer program 303 stored in memory 301 and executable on processor 302. Memory 301 and processor 302 are connected via communication. When processor 302 executes computer program 303, the intelligent variable frequency control method described in the aforementioned embodiment is implemented. The number of processors may be one or more.

[0052] The memory 301 can be a high-speed random access memory (RAM) memory or a non-volatile memory such as a disk drive. The memory 301 is used to store executable program code. The processor 302 is coupled to the memory 301 .

[0053] Furthermore, the embodiment of the present application also provides a computer-readable storage medium, which can be provided in the electronic device in the above embodiments. The computer-readable storage medium can be the above Figure 3 Memory in the illustrated embodiment.

[0054] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the intelligent variable frequency control method described in the aforementioned embodiment. Furthermore, the computer-readable storage medium may be any medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), RAM, a magnetic disk, or an optical disk.

[0055] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0056] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0057] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent frequency conversion control method, characterized in that: include: Collect environmental parameter sets and device operation data of smart devices based on preset multi-dimensional sensors; Generate a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the device operation data; Constructing a neural network wind speed prediction model based on the multidimensional feature vector; Dynamically adjusting the connection paths of the hidden layer nodes of the neural network wind speed prediction model by using an ant colony algorithm to generate a dynamic wind speed setting value within a prediction time period; Dynamically adjust the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate through the variable domain expansion factor of fuzzy control, and generate PID parameter adjustment instructions in combination with the fuzzy rule base; A real-time adjustment parameter of the fan frequency of the smart device is generated according to the PID parameter adjustment instruction.

2. The intelligent frequency conversion control method according to claim 1, characterized in that: The step of generating a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the device operation data includes: Analyzing the correlation between the multi-dimensional parameters and system energy consumption and purification efficiency, and determining a core parameter set whose correlation is higher than a preset threshold; De-noising the core parameter set based on sliding window filtering to generate a de-noised normalized data matrix; Performing dimensionality reduction processing on the redundant dimensions of the noise-normalized data matrix by covariance matrix eigenvalue decomposition; The low-dimensional feature vector after dimensionality reduction processing is cross-validated by a redundant sensor group, and abnormal data is replaced or interpolated and corrected according to the validation result to generate the multi-dimensional feature vector.

3. The intelligent frequency conversion control method according to claim 1, characterized in that: The step of dynamically adjusting the connection paths of the hidden layer nodes of the neural network wind speed prediction model by using an ant colony algorithm to generate a dynamic wind speed setting value within a prediction time period includes: Performing a global search on the initial weights of the neural network wind speed prediction model, and generating an optimized initial weight matrix through selection, crossover, and mutation operations; Initializing the pheromone concentration distribution of the ant colony algorithm based on the optimized initial weight matrix, and iteratively screening the connection paths between the hidden layer nodes through the pheromone volatility factor and the path selection probability function; Dynamically update the weight distribution of the connection path according to the pheromone concentration between the hidden layer nodes, and generate the hidden layer node connection path adjustment instruction; The environmental parameters collected in real time are input into the adjusted neural network wind speed prediction model, and the dynamic wind speed set value within the prediction time period is generated through forward propagation calculation.

4. The intelligent frequency conversion control method according to claim 3, characterized in that: The step of dynamically adjusting the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate by using the variable domain expansion factor of fuzzy control, and generating a PID parameter adjustment instruction in combination with a fuzzy rule base includes: Calculating a wind speed deviation based on a difference between the dynamic wind speed setting value and the real-time wind speed; Performing differential calculation on the wind speed deviation through a sliding time window to generate a deviation change rate; Dynamically adjusting the domain range of the variable domain scaling factor based on the relationship between the absolute value of the wind speed deviation and a preset deviation threshold; Input the wind speed deviation, the deviation change rate sequence and the adjusted domain range into a fuzzy rule base, and generate a fuzzy output set through membership function mapping; The fuzzy output set is defuzzified and a PID parameter adjustment instruction is generated in combination with the current parameter value of the PID controller.

5. The intelligent frequency conversion control method according to claim 4, characterized in that: The step of generating a real-time adjustment parameter of the fan frequency of the smart device according to the PID parameter adjustment instruction includes: Performing weight decoupling on the proportional, integral and differential components in the PID parameter adjustment instruction to generate a parameter adjustment matrix corresponding to each control component; A transfer function model is constructed based on the parameter adjustment matrix, and overshoot suppression is performed on the step input of the transfer function model through dynamic response characteristics to generate a segmented frequency adjustment amount sequence; According to the relationship between the cumulative interval of the integral component in the segmented frequency adjustment sequence and the absolute value of the real-time wind speed deviation, the target integral component exceeding the preset deviation threshold is frozen through the integral separation strategy to generate an anti-saturation corrected frequency adjustment; The anti-saturation correction frequency adjustment amount is superimposed on the current fan frequency, and the superimposed result is dynamically compensated and corrected through a feedforward compensation model to generate a real-time adjustment parameter of the fan frequency.

6. The intelligent frequency conversion control method according to claim 5, characterized in that: The method further comprises: Based on the real-time adjustment parameters and historical equipment operation data, a digital twin model synchronized with the physical equipment state is constructed through a time domain synchronization algorithm; wherein the input parameters of the digital twin model include real-time wind speed and wind turbine current harmonic distortion rate; Performing frequency domain energy feature extraction on the device vibration signal output by the digital twin model to generate a corresponding feature vector; Determining a fault type that matches the feature vector using a preset fault prediction model; Matching the predicted deviation of the real-time wind speed with the historical wind speed deviation sequence corresponding to the fault type, and performing weighted correction on the matching result in combination with the wind turbine current harmonic distortion rate to generate a comprehensive fault risk index; When the comprehensive failure risk index exceeds a preset risk threshold, the early warning mechanism of the smart device is triggered.

7. The intelligent frequency conversion control method according to claim 6, characterized in that: The method further comprises: Extracting time-frequency domain degradation features of the fan bearing vibration signal based on the multidimensional feature vector and historical equipment degradation data to generate a degradation feature vector; Inputting the degradation feature vector into a pre-trained degradation prediction model to generate a confidence interval prediction value of the remaining life of the wind turbine bearing; According to the weight distribution relationship between the confidence interval prediction value and the preset maintenance cost model, the maintenance cycle and frequency adjustment parameters are jointly optimized to generate a dynamic maintenance decision matrix; A gradient maintenance instruction is triggered according to the risk priority parameter in the dynamic maintenance decision matrix, and the wind turbine bearing is maintained through the gradient maintenance instruction.

8. An intelligent frequency conversion control device, characterized in that: The intelligent frequency conversion control device comprises: The acquisition module is used to collect the environmental parameter set and device operation data of the smart device based on the preset multi-dimensional sensors; a processing module, configured to generate a multidimensional feature vector by standardizing the environmental parameter set and the multidimensional parameters of the device operation data; A construction module, configured to construct a neural network wind speed prediction model based on the multidimensional feature vector; A first adjustment module is used to dynamically adjust the connection path of the hidden layer nodes of the neural network wind speed prediction model through an ant colony algorithm to generate a dynamic wind speed setting value within a prediction time period; A second adjustment module is used to dynamically adjust the wind speed deviation between the dynamic wind speed set value and the real-time wind speed and the domain range of the deviation change rate through a variable domain expansion factor of fuzzy control, and generate a PID parameter adjustment instruction in combination with a fuzzy rule base; A generating module is used to generate a real-time adjustment parameter of the fan frequency of the smart device according to the PID parameter adjustment instruction.

9. An electronic device, characterized in that: Comprising a memory and a processor, wherein: The processor is configured to execute a computer program stored in the memory; When the processor executes the computer program, the steps of the intelligent variable frequency control method according to any one of claims 1 to 7 are implemented.

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

Images