PID (Proportion Integration Differentiation) controller parameter adaptive adjustment method and device, electronic equipment and medium

Abstract

The invention relates to the technical field of PID parameter setting, and provides a PID controller parameter adaptive adjustment method and device, electronic equipment and a medium. Multi-source heterogeneous data collected by a distributed sensor array are acquired in real time, feature quantization processing is performed on the multi-source heterogeneous data to obtain a quantization feature matrix, and working condition pattern recognition based on depth fuzzy clustering is performed on the quantization feature matrix to obtain working condition pattern vectors. Performing parameter space adaptive mapping on the working condition mode vector to obtain a PID parameter initial set and an optimized space, and performing parallel parameter optimization processing on the PID parameter initial set and the optimized space according to a hybrid optimization algorithm to obtain an optimized parameter set, and performing parameter dynamic fusion and robustness enhancement processing on the optimized parameter set to obtain a target PID parameter set. Through combination of data processing, deep learning, algorithm optimization and robustness enhancement, the precision of PID controller parameter adaptive adjustment is improved, and efficient and stable operation of a target system under different working conditions is ensured.

Description

Technical Field

[0001] This application relates to the technical field of PID parameter setting, and in particular, to a method, device, electronic device, and medium for adaptively adjusting PID controller parameters. Background Art

[0002] The Proportion Integer Differential (PID) controller plays an important role in industrial automation and control systems. By adjusting the three parameters of proportion, integral, and differential, it responds to, accumulates, and predicts the system error, thereby achieving precise control of the system output. Due to its simplicity, robustness, and ease of implementation, the PID controller is widely used in various industrial control scenarios such as temperature, pressure, and flow.

[0003] Existing adaptive adjustment of PID controller parameters can adjust PID parameters in real time according to the dynamic characteristics of the system and environmental changes to adapt to system changes and external disturbances. For example, in motor control, the adaptive PID algorithm can automatically adjust parameters according to the load change of the motor to ensure the stable operation of the motor. Although existing adaptive PID control algorithms can adjust parameters in real time, they are difficult to provide the best performance for nonlinear or time-varying systems, require more complex control strategies, and cannot maintain a stable control effect. Summary of the Invention

[0004] In view of this, this application provides a method, device, electronic device, and medium for adaptively adjusting PID controller parameters to solve the problem of inaccurate adaptive adjustment of PID controller parameters.

[0005] The first aspect of this application provides a method for adaptively adjusting PID controller parameters, and the method includes: Real-time acquisition of multi-source heterogeneous data of the controlled system collected by a distributed sensor array, and performing feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix representing the dynamic characteristics of the controlled system; Performing working condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to classify the operating conditions of the controlled system and generating a working condition mode vector according to the working condition category; Performing non-linear mapping processing on the working condition mode vector through a preset parameter space adaptive mapping method to obtain an initial set of PID parameters and an optimization space matching the current working condition category; According to a preset hybrid optimization strategy, performing parallel parameter optimization processing on the initial set of PID parameters and the optimization space to obtain an optimized parameter set that meets the dynamic response index; The optimization parameter set is subjected to evidence theory fusion and anti-disturbance correction processing to obtain a target PID parameter set, and the proportional coefficient, integral coefficient and differential coefficient of the PID controller are updated in real time according to the target PID parameter set.

[0006] In an optional implementation, the performing feature quantization processing on the multi-source heterogeneous data to construct a quantized feature matrix characterizing the dynamic characteristics of the controlled system includes: Performing timestamp alignment processing on the multi-source heterogeneous data to obtain a time-aligned data set; Performing wavelet basis function denoising processing on the time-aligned data set to obtain a standardized feature data set; Performing time domain feature extraction processing on the feature data set to obtain a time domain feature set; Performing frequency domain feature extraction processing on the feature data set to obtain a frequency domain feature set; Performing spatiotemporal correlation feature extraction processing on the feature data set to obtain a spatiotemporal correlation feature set; A matrix is ​​constructed according to the time domain feature set, the frequency domain feature set and the spatiotemporal correlation feature set to obtain the quantization feature matrix.

[0007] In an optional embodiment, the operating condition category includes an operating condition classification data set, a coupling strength feature set, and a dynamic level, and the operating condition pattern recognition based on deep fuzzy clustering of the quantized feature matrix is ​​performed to classify the operating conditions of the controlled system, and the operating condition pattern vector is generated according to the operating condition category, including: Performing deep fuzzy mean clustering processing on the quantized feature matrix to obtain a membership matrix; Performing a classification process of the operating condition continuity according to the membership matrix to obtain the operating condition classification data set; Performing dynamic characteristic analysis on the working conditions in the working condition classification data set according to the frequency domain feature set to obtain the dynamic level; Perform coupling strength analysis according to the spatiotemporal correlation feature set, the operating condition classification data set, and the dynamic level to obtain the coupling strength feature set; Vector construction is performed on the operating condition classification data set, the coupling strength feature set, and the dynamic level to obtain the operating condition mode vector.

[0008] In an optional embodiment, the nonlinear mapping process is performed on the operating mode vector by a preset parameter space adaptive mapping method to obtain an initial set of PID parameters matching the current operating condition category and an optimization space, including: Perform coefficient mapping processing on the working condition classification data set in the working condition mode vector through a preset coefficient model to obtain the mapping coefficients of each working condition category; Construct a parameter space mapping function for each working condition level according to the mapping coefficients and the dynamic level; Perform non-linear mapping processing on the working condition mode vector according to the parameter space mapping function to obtain the initial set of PID parameters, and select the corresponding preset constraint range according to the initial set of PID parameters; Adjust the constraint range according to the coupling strength feature set to obtain an optimization domain; Adjust each initial parameter range in the optimization domain according to the working condition classification data set, the dynamic level, and the coupling strength feature set to obtain the optimization space of each parameter in the initial set of PID parameters.

[0009] In an alternative embodiment, the parallel parameter optimization processing of the initial set of PID parameters and the optimization space according to a preset hybrid optimization strategy to obtain an optimized parameter set that meets the dynamic response index includes: Perform particle swarm algorithm processing on the initial set of PID parameters to obtain a global optimal solution; Perform genetic algorithm processing on the optimization space to obtain multiple candidate optimization solutions; Perform weighted average calculation on the global optimal solution and the candidate optimization solutions to obtain a candidate optimized parameter set; Perform fitness evaluation on the candidate optimized parameter set to screen out a sub-optimal parameter set; Perform constraint verification on the sub-optimal parameter set to obtain the optimized parameter set.

[0010] In an alternative embodiment, the evidence theory fusion and anti-interference correction processing on the optimized parameter set to obtain a target PID parameter set, and the proportional coefficient, integral coefficient, and differential coefficient of the PID controller are updated in real time according to the target PID parameter set, including: Perform evidence theory fusion processing on the optimized parameter set to obtain the basic probability assignment function values of each candidate solution in the optimized parameter set; Perform weighted combination on the candidate solutions according to the basic probability assignment function values through a preset Dempster combination rule to obtain an optimized solution set; Perform dynamic adjustment on the optimized solution set based on the anti-interference correction term to obtain an anti-interference corrected optimized solution set; Perform performance verification on the anti-interference correction and optimization solution set to screen out a target PID parameter set that meets the preset stability requirements, and update the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

[0011] In an optional implementation manner, the performing performance verification on the anti-interference correction and optimization solution set to screen out a target PID parameter set that meets the preset stability requirements includes: Update the PID control calculation formula according to the anti-interference correction and optimization solution set, and calculate the control input data at the current moment according to the output data at the current moment and the PID control calculation formula; Perform a rolling optimization window calculation on the control input data through a preset prediction model to obtain optimized control input data and parameter response change data; Perform calculation processing on the parameters to be verified according to the parameter response change data to obtain the Lyapunov exponent and dynamic indicators; Compare the Lyapunov exponent, the dynamic indicators, and the adjustment time in the parameter response change data with a preset verification threshold set one by one to screen out a target PID parameter set that meets the preset stability requirements.

[0012] The second aspect of the present application provides a PID controller parameter adaptive adjustment device, and the device includes: A feature matrix module, configured to obtain multi-source heterogeneous data of a controlled system collected by a distributed sensor array in real time, and perform feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix representing the dynamic characteristics of the controlled system; A mode vector module, configured to perform working condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to classify the running working conditions of the controlled system, and generate a working condition mode vector according to the working condition category; An initial parameter module, configured to perform non-linear mapping processing on the working condition mode vector through a preset parameter space adaptive mapping method to obtain a PID parameter initial set and an optimization space that match the current working condition category; An optimized parameter module, configured to perform parallel parameter optimization processing on the PID parameter initial set and the optimization space according to a preset hybrid optimization strategy to obtain an optimized parameter set that meets the dynamic response index; A dynamic fusion module, configured to perform evidence theory fusion and anti-interference correction processing on the optimized parameter set to obtain a target PID parameter set, and update the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

[0013] In a third aspect of the present application, an electronic device is provided. The electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the PID controller parameter adaptive adjustment method described above are implemented.

[0014] In a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the PID controller parameter adaptive adjustment method described above are implemented.

[0015] In summary, the present application includes at least the following beneficial technical effects: 1. By introducing technologies such as multi-source heterogeneous data processing, deep fuzzy clustering, and dynamic parameter mapping, the PID controller parameters are dynamically adjusted according to real-time working conditions, so that the PID controller can better adapt to complex working conditions with large changes in environment and load, and improve the control accuracy.

[0016] 2. Combining two excellent global search algorithms, the particle swarm algorithm and the genetic algorithm, can effectively avoid the trouble of local optimal solutions and improve the effect and stability of parameter adjustment.

[0017] 3. By adding a disturbance rejection correction term and dynamically adjusting it during the optimization process, the robustness and anti-interference ability of the PID controller can be improved, ensuring that the PID controller can still operate stably in a complex and changeable working environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0019] Figure 1 is a flowchart of a method for adaptively adjusting PID controller parameters provided by an embodiment of the present application; Figure 2 is a functional module diagram of a device for adaptively adjusting PID controller parameters provided by an embodiment of the present application; Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

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

[0021] As Figure 1 shown, it is a flowchart of the PID controller parameter adaptive adjustment method provided by the embodiment of the present application. The PID controller parameter adaptive adjustment method provided by the embodiment of the present application includes the following steps.

[0022] Step S1: Real-time obtain the multi-source heterogeneous data of the controlled system collected by the distributed sensor array, and perform feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix representing the dynamic characteristics of the controlled system.

[0023] Obtain multi-source heterogeneous data from the distributed sensor array. These multi-source heterogeneous data come from different types of sensors, such as temperature sensors, pressure sensors, flow sensors, electromagnetic interference sensors, network load sensors, etc. The data collected by each sensor will have different sampling frequencies, different timestamps, and different units. Therefore, during the process of collecting multi-source heterogeneous data, it is first necessary to synchronize these data. The present application uses timestamp alignment processing to align the collected data of all sensors according to the same time reference, so as to ensure the correct time correspondence relationship of the data of different sensors. By introducing a global time reference system or by using an interpolation method to align the multi-source heterogeneous data, the time deviation caused by different sampling frequencies of different sensors is eliminated. The processed multi-source heterogeneous data is summarized and stored in the time-aligned dataset.

[0024] During the process of multi-source heterogeneous data, the sensors affected by noise will affect the quality of the data, thereby affecting the accuracy of analysis and decision-making. To remove this noise, the present application uses wavelet basis functions to remove the noise in the time-aligned dataset. Among them, the operation of performing data transformation through wavelet basis functions is a time-frequency analysis method. By convolving with a set of basic wavelet functions, information of different scales can be extracted from the time-aligned dataset. During the denoising process, the multi-source heterogeneous data in the time-aligned dataset is decomposed into different frequency bands (i.e., the low-frequency part and the high-frequency part) by using wavelet transform, and then the noise is removed by removing the high-frequency part. The expression formula of the basic wavelet function is as follows: Among them, is the original collected signal. is the layer wavelet basis function. Represents the inner product of the original signal and the wavelet basis function. is the noise standard deviation of the wavelet at the th layer.

[0025] is the regularization coefficient used to avoid excessive noise removal. Wavelet denoising is used to eliminate high-frequency noise in the signal, making the multi-source heterogeneous data in the processed time-aligned dataset cleaner.

[0026] On the denoised feature dataset, feature extraction is performed on the multi-source heterogeneous data in the feature dataset to facilitate the extraction of useful patterns or rules therefrom. Among them, the data features extracted in this application include: time-domain features, frequency-domain features, and spatio-temporal correlation features. Among them, is the time-domain data of the signal, is the mean value of the signal, is the variance of the signal.

[0027] Frequency-domain features can reveal the frequency components of the signal. Common frequency-domain features include but are not limited to the main frequency component, spectral entropy, etc. Through frequency-domain features, the frequency distribution of the signal can be understood. In this application, the Fourier transform is used to perform spectral analysis on the multi-source heterogeneous data to obtain the main frequency-domain components of the multi-source heterogeneous data, and the main frequency component is used to represent the frequency-domain features of the multi-source heterogeneous data.

[0028] Spatio-temporal correlation features are used to describe the spatio-temporal relationship between signals, especially the correlation between multiple sensors. In this application, the Granger causality test method is used to analyze the degree of association between different sensor data (i.e., spatio-temporal correlation features). Through spatio-temporal correlation features, the correlation and causal relationship between different sensor data can be revealed, facilitating the understanding of the interaction between multiple sensors at different time points.

[0029] After extracting the time-domain features, frequency-domain features, and spatio-temporal correlation features, these features are combined into a quantization feature matrix. The quantization feature matrix contains multiple feature dimensions of all sensors and is used for data analysis, pattern recognition, or optimization. The quantization feature matrix can be represented by the following formula: Among them, is the quantization feature matrix, are respectively the mean value in the time-domain data and the variance of the signal, For representing frequency-domain features (i.e., main frequency components), For representing spatio-temporal correlation features (i.e., Granger causality). The quantization feature matrix is used to represent and store the feature information of all sensors, providing input for machine learning models or optimization algorithms.

[0030] Step S2: Perform condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to classify the operating conditions of the controlled system, and generate a condition mode vector according to the condition categories. Among them, each condition category includes a condition classification data set, a coupling strength feature set, and a dynamic level.

[0031] Perform deep fuzzy c-means clustering processing on the quantization feature matrix, so as to classify the conditions according to the data features collected by different sensors. Among them, deep fuzzy c-means is a technique that extends the traditional c-means method, which not only allows data points to belong to multiple categories, but also controls the fuzziness of classification by adjusting the fuzzy exponent. Specifically, input the quantization feature matrix into the following improved deep fuzzy c-means clustering algorithm, and obtain the membership degree of each data point (i.e., sensor) to each category through clustering calculation.

[0032] Among them, is the fuzzy clustering objective function, representing the total cost of clustering is the th data point to the th cluster, and its value range is [0,1], indicating the degree to which the data point belongs to a certain cluster. is the fuzzy exponent, usually taking the value 2 (default value). This parameter controls the fuzziness of clustering. The larger the value, the fuzzier the clustering; the smaller the value, the stricter the clustering. is the th data point and the th cluster center Euclidean distance. is the regularization coefficient, used to balance the entropy value of the membership degree. The clustering algorithm can classify the sensor data according to the distance and membership degree, and obtain the membership degree matrix of each sensor to each condition category. The membership degree matrix is used to describe the behavior characteristics of each sensor under different conditions.

[0033] After deep fuzzy c-means clustering, the obtained membership matrix can be used for continuous classification of working conditions. The working condition classification is determined based on the membership degrees of different sensors to identify the working condition types of the system at different time periods or in different states. It should be understood that the membership matrix also defines the duration of each working condition category in the time series. For example, if the membership degree of a working condition remains high for a period of time, it can be considered that the working condition lasts for this period of time. According to the change of the membership degree, the multi-source heterogeneous data is divided into different working condition categories, thereby obtaining a working condition classification data set. In this application, a preset threshold is used to determine which working condition the multi-source heterogeneous data at a certain moment belongs to.

[0034] After obtaining the working condition classification data set, dynamic characteristic analysis is performed on each working condition to quantitatively describe the dynamic behavior under different working conditions and obtain the dynamic level of each working condition. Common dynamic characteristics include but are not limited to response speed, overshoot, stability, etc. Frequency domain features are used to judge the dynamic characteristics of the working conditions. For example, some working conditions may have rapid frequency changes, while other working conditions may be relatively stable. Thus, the dynamic level of each working condition is judged through frequency domain features, adjustment time, overshoot, etc. Among them, the dynamic level is a preset feature standard. Exemplarily, the following shows the preset recommended feature standards and their corresponding dynamic levels: The data features corresponding to each data in the working condition classification data set are compared one by one with the dynamic characteristics in the preset standard to assign a dynamic level label to each working condition data in the working condition classification data set.

[0035] Coupling strength analysis is used to evaluate the relationships and interactions between multiple sensors. For complex systems, there may be strong or weak coupling relationships between various sensors. In this application, the Granger causality test method is used to calculate the coupling strength between different sensors. Sensors with high coupling strength have a greater impact on the system. Therefore, spatio-temporal correlation features can be used to represent the coupling strength of each working condition data in the working condition classification data set. Specifically, each working condition data is sorted according to the dynamic level of the working condition data, and the spatio-temporal correlation features of the working conditions with higher dynamic levels are placed in the priority position, and so on to form a complete coupling strength feature set with the spatio-temporal correlation features. The coupling strength feature set reflects the influence of each sensor under different working conditions. The construction of the coupling strength feature set provides an important basis for the subsequent formation of the working condition mode vector and the adjustment of the PID controller parameters.

[0036] Finally, the working condition classification data set, the coupling strength feature set, and the dynamic level are integrated into a working condition mode vector. The working condition mode vector contains various state features of the current system for subsequent analysis and can provide a basis for subsequent control decisions. The working condition mode vector can be expressed by the following formula: Among them, is the operating condition mode vector, is the classification data of the th operating condition, is the dynamic level of the th operating condition, is the coupling strength between sensor and sensor

[0037] Step S3: Perform a non - linear mapping process on the operating condition mode vector through a preset parameter space adaptive mapping method to obtain an initial set of PID parameters and an optimization space that match the current operating condition category.

[0038] The operating condition mode vector contains multi - dimensional features under different operating conditions, including time - domain features, frequency - domain features, and coupling strength and other information. Train the parameter mapping coefficient according to the operating condition classification data set in the operating condition mode vector to obtain the mapping coefficient of each operating condition category, so as to train the model through the known operating condition data and provide a basis for subsequent parameter mapping. Use the classification data in the operating condition mode vector as input to represent different operating condition categories. And through regression analysis, the least - squares method or other parameter fitting methods, train the mapping coefficient between the operating condition category and the PID control parameters. Among them, the mapping coefficient is a coefficient related to the operating condition type, used to reflect the change law of the PID controller parameters under different operating conditions.

[0039] After obtaining the mapping coefficient of each operating condition category, construct a parameter space mapping function for each operating condition level according to the dynamic level. Among them, the dynamic level represents the complexity of the operating condition and the response characteristics of the system. Constructing the mapping function according to the dynamic level can ensure that different complexity operating conditions have different control parameter mappings. The PID controller of the system includes three main parameters: proportional coefficient, integral coefficient, and differential coefficient. The PID controller parameters will be adjusted according to the dynamic characteristics of the operating condition level. The dynamic characteristics of the operating condition level are represented by where each represents the dynamic response characteristics of a certain operating condition (for example, overshoot, adjustment time, etc.). The parameter space mapping function is specifically as follows: Among them, , , are the proportional, integral, and differential parameters of the PID controller respectively, and the parameter values will be adjusted according to the dynamic level of the operating condition. represents the dynamic level of the operating condition, and each ​Reflects the performance of the operating condition on a certain characteristic (such as response speed, stability, etc.). Is the mapping coefficient, used to adjust the relationship between the operating condition characteristics and the PID parameters Is the coefficient used to adjust the integral part of the PID parameters. Is the coefficient used to adjust the derivative part of the PID parameters. The proportional coefficient is usually related to the response speed of the system. Under certain operating conditions, a higher dynamic level (such as a faster response requirement) requires a higher proportional coefficient. Therefore, the parameter space mapping function uses an exponential function to map the dynamic level To the proportional coefficient. By selecting appropriate coefficients , the growth rate of the proportional coefficient under different operating conditions can be adjusted. The integral coefficient is mainly used to eliminate the steady-state error. A higher dynamic level usually corresponds to a need for faster error correction. When the dynamic level of the operating condition is high, a larger integral coefficient may be required to accelerate the steady-state regulation. The parameter space mapping function adjusts the integral coefficient through a hyperbolic tangent mapping to ensure that the system can respond to changes more smoothly. Adding the absolute value of the operating condition level to the mapping is to increase the sensitivity under specific operating conditions and avoid too small an integral under complex or unstable operating conditions resulting in a steady-state error. The derivative coefficient mainly affects the anti-interference ability of the system, especially the suppression of rapid changes during the response process. A higher dynamic level usually requires a higher derivative coefficient to enhance the response suppression ability of the system. Therefore, the parameter space mapping function uses a square function to adjust the derivative coefficient to ensure that the system can handle faster dynamic changes. To avoid excessive suppression caused by too large a derivative coefficient, a logarithmic function is used to limit the growth to ensure that the derivative action does not get out of control at a higher dynamic level.

[0040] After converting the operating condition mode vector into the initial set of PID parameters through the parameter space mapping function, according to the non-linear mapping result of the parameter space and the actual system requirements, a suitable parameter range (i.e., the constraint range) is selected. The constraint range of this application is specifically as follows: Among them, Are the proportional, integral, and derivative coefficients of the PID controller. Is the initial set of PID parameters obtained through parameter space mapping. The constraint range is used to ensure that the PID parameters are not too large or too small, avoid system instability, and prevent over-adjustment from causing system instability or sluggishness.

[0041] Furthermore, adjust the constraint range of the PID parameter set according to the coupling strength feature set to obtain the final optimization domain. The coupling strength feature set can reveal which sensors have a greater impact on the system. Therefore, more refined adjustment is required for these parameters with greater impact. Evaluate which sensors have a greater impact on the system behavior through the coupling strength feature set. For these sensors with greater impact, adjust the constraint range of their PID parameters to ensure that their impact on the system is effectively controlled. Optimize the parameter constraint range according to the analysis results of the coupling strength. For example, for sensors with greater impact, it may be necessary to relax the constraint range to increase their control flexibility. The optimization domain can be expressed by the following formula: Wherein, is the PID parameter after adjustment by the coupling strength. is the coupling strength feature set, representing the interaction strength between sensors. is a function for adjusting the PID parameters according to the coupling strength. Through the adjustment of the coupling strength, ensure that the PID parameters of key sensors are properly adjusted, thereby optimizing the control effect of the entire system.

[0042] Finally, according to the working condition classification data set, dynamic level, and coupling strength feature set, adjust each initial parameter range in the optimization domain. This process further refines the parameter space of the PID controller to ensure optimal control under different working conditions. Based on the working condition classification, dynamic level, and coupling strength analysis, refine the optimization space of the PID parameters. For working conditions with a higher dynamic level, a larger range of control parameter changes may be required to cope with more complex dynamic responses. Construct the adjusted parameter range into the following optimization space: Wherein, is the optimization space of the PID parameters. are the minimum and maximum values of the proportional coefficient are the minimum and maximum values of the integral coefficient. are the minimum and maximum values of the derivative coefficient. The adjustment of the optimization space can ensure that under different working conditions, the change range of the PID parameters can meet the control requirements, improving the robustness and flexibility of the system.

[0043] Step S4: According to the preset hybrid optimization strategy, perform parallel parameter optimization processing on the initial PID parameter set and the optimization space to obtain an optimized parameter set that meets the dynamic response index.

[0044] Among them, the hybrid optimization algorithm includes the particle swarm optimization algorithm and the genetic algorithm. Through the hybrid optimization algorithm, global search and local optimization are carried out to find the optimal PID controller parameters to meet the requirements such as system stability, response speed, and robustness.

[0045] The particle swarm optimization algorithm is applied to the processing of the initial set of PID parameters to find the global optimal solution in the parameter space through global search. Specifically, the particle swarm optimization algorithm regards the entire parameter space as the search space and initializes a group of particles, where each particle represents a set of PID controller parameters. Each particle has a position and a velocity, representing a certain point in the parameter space, and the position of the particle corresponds to a candidate solution of the PID parameters. Each particle updates its position according to its current position, velocity, and historical optimal position. During the iterative process of the particle swarm optimization algorithm, position update and fitness evaluation are carried out alternately. In each round of iteration, first, the velocity and position of the particle are updated according to its current position, and then the updated particle is subjected to fitness evaluation. According to the results of the fitness evaluation, the historical optimal position and the global optimal position of the particle are updated. This alternating process continues, and the particle swarm gradually finds the global optimal solution.

[0046] The update formula of the particle swarm optimization algorithm is as follows: Among them, is expressed as the velocity of particle at the th iteration. is expressed as the position of particle at the th iteration (i.e., the PID parameter is expressed as the historical optimal position of particle and is the global optimal position of all particles. is expressed as the learning factor, which controls the learning ability of the particle towards the local optimal and the global optimal. is expressed as a random number, with a range of [0, 1], which introduces randomness to help the particle jump out of the local optimal. is expressed as the inertia weight, which determines the driving force of the particle's forward movement and controls the globality and locality of the search.

[0047] Fitness evaluation is required both before and after updating the particle position. Before updating the particle position, fitness evaluation is first carried out to determine whether the current position of the particle is a good solution. If the fitness of the current particle is poor, it adjusts its position according to the guidance of the local optimal position and the global optimal position to search for a better solution. Therefore, fitness evaluation provides a feedback mechanism for the update of the particle. After the particle position is updated, the fitness evaluation of the particle also changes because the updated particle position represents a new combination of PID parameters. Therefore, each position update of the particle introduces a new solution, and it is necessary to verify the quality of this new solution through fitness evaluation. This makes the search process of the particle swarm algorithm dynamic and real-time. The fitness evaluation formula of the particle swarm algorithm is as follows: where, is the fitness function value of particle , is the system error, is the control input, is the change in the control input. The lower the fitness value of the particle, the better the PID parameter combination represented by the particle. The higher the fitness, the worse the solution. In the particle swarm algorithm, the global optimal solution is the solution with the lowest fitness value, that is, this solution has the best performance in the control system. After each particle update, fitness evaluation will calculate the fitness value of the current particle. If this fitness value is lower than the fitness value of the preset global optimal particle, the current particle will be updated to the global optimal particle.

[0048] Meanwhile, the genetic algorithm is used to further process the optimization space to obtain multiple candidate optimization solutions. The genetic algorithm performs local search and improvement by simulating the mechanisms of natural selection and genetics, and optimally selects suitable control parameters. The genetic algorithm first initializes a set of solutions (i.e., PID control parameters), and these solutions form the initial population. Each solution represents a set of PID parameters. The population size is usually set to a relatively large value to increase the diversity of the exploration space. The fitness of each solution is also evaluated by the objective function, which measures how well the parameter set performs in the system. The higher the fitness value, the better the solution. According to the results of fitness evaluation, solutions with higher fitness are selected for reproduction. Commonly used selection operations include roulette wheel selection, tournament selection, etc. The selected solutions are combined through crossover operations to generate new solutions. The crossover operation simulates the process of gene recombination in nature, and exchanges part of the genes (i.e., PID parameters) from the parent solutions to generate new solutions. The crossover operation can be expressed as the following specific formula: where, represents the newly generated offspring solution represents the parent solution. It represents the crossover rate and controls the exchange ratio of parental genes.

[0049] After the crossover operation generates new solutions, the mutation operation makes minor modifications to the solutions to simulate gene mutations. The mutation operation increases the diversity of the solutions and avoids falling into local optima. The mutation operation can be expressed as the following specific formula: Among them, : The mutation amount of the solution. : The mutation rate, which controls the mutation amplitude. RandomNoise: Random noise, used to change the solution.

[0050] The fitness of the offspring solutions generated by the crossover operation and the mutation operation is evaluated, and some individuals in the population are replaced according to the fitness values to maintain the quality of the population. The genetic algorithm gradually optimizes the quality of the solutions by continuously iterating the crossover and mutation operations. After a certain number of generations, the genetic algorithm will provide a set of candidate solutions with better performance.

[0051] After obtaining the global optimal solution and candidate solutions, through weighted averaging, the advantages of the two optimization methods can be combined to obtain a solution with better comprehensive performance. The global optimal solution and multiple candidate solutions are weighted averaged as shown below to obtain a new set of optimized parameters.

[0052] Among them, The set of optimized parameters after weighted averaging represents the global optimal solution obtained by the particle swarm optimization algorithm and the candidate solutions obtained by the genetic algorithm. represents the weights of each solution, usually determined by their fitness values. Through weighted averaging, the global search ability of the particle swarm optimization algorithm and the local optimization ability of the genetic algorithm are combined to obtain a more stable and high-quality set of optimized parameters.

[0053] Furthermore, the fitness of the candidate set of optimized parameters obtained by weighted averaging is evaluated, and its performance in the control system is calculated. Thus, according to the fitness evaluation results, a sub-optimal set of parameters with better performance is selected. And the sub-optimal set of parameters is subjected to constraint verification to ensure that it meets the parameter range and the performance requirements of the system. The constraint conditions include the maximum and minimum values of the PID parameters, response speed, stability, etc. The set of parameters that passes the verification is the final set of optimized parameters.

[0054] Step S5: Perform evidence theory fusion and anti-interference correction processing on the set of optimized parameters to obtain a target PID parameter set, and update the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

[0055] Parameter dynamic fusion and robustness enhancement processing includes evidence theory fusion processing, Dempster combination rule, and anti-interference correction term. The optimization parameter set is subjected to parameter dynamic fusion and robustness enhancement processing to obtain an optimization parameter set that can not only ensure the system performance but also maintain stability and robustness under various working conditions.

[0056] The evidence theory evaluates the credibility of each solution by calculating the basic probability assignment function of each solution, and outputs the reliability of the solution in the optimization process. The basic probability assignment function can be expressed as the following specific formula: Among them, is the basic probability assignment function value of solution , representing the credibility of this solution. is the fitness value of solution . is the minimum fitness value among all solutions, representing the fitness of the optimal solution. is a regularization parameter used to control the smoothness of probability assignment. The denominator part is the normalization factor of the fitness of all solutions, which is used to ensure that the sum of all BPAs is 1. The BPA value of each optimized solution reflects the relative credibility of this solution, and solutions with higher BPA values will be given higher weights in the fusion process. Through evidence theory fusion, the credibility of different solutions can be quantified.

[0057] The Dempster combination rule generates a comprehensive solution by weighted combining the BPA values of multiple solutions. For each pair of candidate solutions, they are combined according to their BPA values, and the combined BPA value is calculated. The Dempster combination rule formula can be expressed as the following specific formula: Among them, is the combined BPA value, representing the credibility of the combined result of two solutions and . and are the BPA values of solutions and respectively. represents the possible intersection in solutions and , represents the overlap of the two solutions in this part. The denominator part is Dempster's normalization factor, which ensures that the sum of the combined BPA values is 1. By applying the Dempster combination rule, multiple candidate solutions can be weighted combined to obtain an optimized solution set. The optimized solution set represents the optimal solution comprehensively evaluated according to the credibility of different sources and solutions.

[0058] The anti-interference correction term is a dynamic adjustment method based on system errors and disturbances. The PID parameters in the optimal solution set are adjusted according to the real-time state and errors of the system, thereby improving the anti-interference ability of the adjusted system. The anti-interference correction term can be expressed by the following specific formula: Wherein, is the change amount of the corrected PID parameter. is the system error (i.e., the difference between the target signal and the actual output signal). sgn is the sign function of the error, indicating the positive or negative of the error. is the error change rate, indicating the change speed of the error. is the time constant, used to control the attenuation speed of the correction term. is the correction intensity, controlling the intensity of the correction term. According to the current error and error change rate, the PID parameters in the optimal solution set are dynamically adjusted. By introducing the anti-interference correction term, the adaptability of the system to external disturbances or system changes can be enhanced.

[0059] Finally, the performance of the optimal solution set after anti-interference correction is verified to ensure that the obtained PID parameter set meets the system stability and control performance requirements. The stability verification and dynamic response verification are carried out for each parameter set in the anti-interference correction optimal solution set. Common stability verification methods include Lyapunov exponents, dynamic indicators, etc. The dynamic response verification includes evaluating performance indicators such as overshoot, adjustment time, steady-state error, etc.

[0060] Specifically, the collected historical control data set contains the input data and output data of the system under different working conditions. The input data usually includes control signals (i.e., the control inputs after PID parameter calculation), and the output data includes the responses of the system (such as feedback signals such as motor speed, temperature, etc.). Through the historical control data set, a prediction model can be trained to predict the response of the system at future moments. The historical data set can be collected through a sensor array during actual operation, recording the input and output data of the system. Common inputs include voltage, current, adjustment signals of the PID controller, etc., and the output is usually the feedback signal of the system (such as motor speed, pressure, temperature, etc.). Using the input and output data in the historical data set, a prediction model is constructed through regression analysis, time series analysis or machine learning algorithms (such as ARX model, neural network, etc.). This model is used to predict the response of the system at future moments. This application uses the ARX model as the prediction model, and the ARX model can be expressed by the following specific formula: Wherein, represents the predicted system output, representing the system response at time ​Denoted as historical output data, representing the moment of the system output. Denoted as historical control input data, representing the moment of the input signal. Denoted as the model coefficients, obtained through training with historical data. Denoted as the order of the model, which determines the window size of the input and output historical data in the model. After constructing the prediction model, the response of the future system can be predicted based on the input data at the current moment.

[0061] Meanwhile, according to the obtained anti-interference correction and optimization solution set, update the calculation formula of the PID controller to calculate the control input at the current moment. Specifically, according to the optimized target PID parameter set , update the calculation formula of the PID controller. The calculation formula of the PID controller can be expressed as the following specific formula: where, Denoted as the control input at the current moment, the signal output by the controller, used to adjust the behavior of the system. Denoted as the error of the system, the difference between the target output and the current output, that is . are the proportional, integral, and derivative parameters of the PID controller. Denoted as the rate of change of the error, representing the rate of change of the error over time. Through the updated PID formula, combined with the error at the current moment and the target PID parameter set , calculate the control input at the current moment .

[0062] Perform a rolling optimization window calculation on the control input data through the prediction model, aiming to predict the impact of the control input on the system response and optimize the parameters at future moments. Specifically, through the prediction model, perform a rolling calculation with a preset time window (for example, the next 5 moments). For each time step, calculate the impact of the current control input on the future system output and obtain the predicted system response. The objective function of the rolling window can be expressed as the following specific formula: where, Denoted as the target reference output. Denoted as the prediction moment of the system output. Denoted as the change amount of the control input. Denoted as the size of the rolling optimization window (for example, 5 moments). Denoted as the weight coefficient, it controls the balance between the control error and the change in the control input. According to the calculation results of the rolling window, the control input is adjusted to minimize the objective function, thereby optimizing the control input. During the rolling optimization process, the parameter response change data at each moment (such as error, control input change, etc.) is recorded.

[0063] According to the changes in the parameter response change data, the Lyapunov exponent is calculated to judge the stability of the system. Among them, the Lyapunov exponent is used to describe the sensitivity and stability of the system. If the exponent is negative, it indicates that the system is stable. The calculation formula of the Lyapunov exponent can be expressed as the following specific formula: Among them, Denoted as the Lyapunov exponent. Denoted as the small perturbation of the system state. Initial perturbation. Denoted as time. If the Lyapunov exponent is greater than zero, it means that the system will diverge rapidly under the small perturbation of the initial conditions, that is, the system is unstable. If the Lyapunov exponent is equal to zero, it means that the system is marginally stable and the perturbation will not increase or decrease with time. If the Lyapunov exponent is less than zero, it means that the system is stable, the perturbation decays with time, and the system will eventually return to the steady state.

[0064] At the same time, according to the parameter response change data, the overshoot and settling time are calculated to evaluate the dynamic response of the system. Among them, the overshoot measures the overshoot of the system response, and the settling time measures the time required for the system to stabilize to the target value. The overshoot formula can be expressed as the following specific formula: Among them, Denoted as the maximum output value of the system (i.e., the peak value of the system response). Denoted as the target reference output value of the system.

[0065] The settling time refers to the time when the system output first enters and remains within ±2% of the target value. The settling time can be expressed as the following specific formula: Among them, Denoted as the system output And the target reference output Between the errors. Denoted as this is the Error range of the target reference output. Denoted as the time when it first reaches and remains within this error range, that is, the settling time It is the system output First enter the target reference value within the range of ±2%, and the minimum time to maintain within this range in subsequent time.

[0066] After obtaining the Lyapunov exponent, dynamic index, and regulation time, the Lyapunov exponent, dynamic index, and regulation time are compared one by one with a preset threshold set to determine whether the PID parameter set meets the requirements of stability and dynamic response. The verification threshold set includes the Lyapunov exponent threshold, overshoot threshold, and regulation time threshold. Among them, it is set to , that is, it is required that the Lyapunov exponent is less than zero, which means the system must be stable; it is set to , that is, it is required that the overshoot of the system must be less than 15% to ensure that the system response does not exceed the target value by too much; it is set to , that is, the regulation time needs to be less than a certain time constant, indicating that the system should be stable within a reasonable time and remain near the target value.

[0067] During the comparison process of the Lyapunov exponent, if , it indicates that the system is stable. If , it is considered that the system is unstable and does not meet the verification requirements, and parameter adjustment needs to be performed before returning to the previous step. During the comparison process of the overshoot, if , it indicates that the overshoot of the system is within the acceptable range and meets the dynamic response requirements. If , it is considered that the overshoot of the system is too high, violating the dynamic response requirements, and parameter adjustment is required. During the comparison process of the regulation time, if , it indicates that the system response speed is qualified and meets the dynamic response requirements. If , it indicates that the system response is too slow and the regulation time is too long, and further parameter adjustment is required.

[0068] If the Lyapunov exponent, overshoot, and regulation time all meet the preset threshold conditions, it is considered that the current target PID parameter set is appropriate and the system can continue to run without adjustment. At this time, it can be considered that the current target PID controller parameter set meets the target stability and dynamic response requirements (that is, the anti-disturbance correction optimization solution set that passes the verification is the target PID parameter set), and the proportional coefficient, integral coefficient, and differential coefficient of the PID controller are updated in real time according to the target PID parameter set If any one of the indicators does not meet the preset threshold, it is considered that the current PID parameter set does not meet the requirements and needs to be adjusted. If the system is unstable (Lyapunov exponent is greater than or equal to zero), the PID parameters (such as proportional, integral, and differential coefficients) need to be adjusted to enhance the stability of the system. If the overshoot is too large, it means that the overshoot of the system does not meet the dynamic response requirements. The proportional coefficient in the PID parameters can be considered to be reduced, or the integral coefficient can be adjusted to reduce the overshoot. If the adjustment time is too long, it means that the response speed of the system is slow. The response of the system can be accelerated by increasing the proportional coefficient of the PID controller or adjusting the differential coefficient.

[0069] This application is applied to the technical field of PID parameter setting. By obtaining multi-source heterogeneous data collected by a distributed sensor array in real time, and performing feature quantization processing on the multi-source heterogeneous data to obtain a quantization feature matrix, performing working condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to obtain a working condition mode vector, performing parameter space adaptive mapping on the working condition mode vector to obtain an initial set of PID parameters and an optimization space, performing parallel parameter optimization processing on the initial set of PID parameters and the optimization space according to a hybrid optimization algorithm to obtain an optimized parameter set, and performing parameter dynamic fusion and robustness enhancement processing on the optimized parameter set to obtain a target PID parameter set. This application combines data processing, deep learning, optimization algorithms, and robustness enhancement to improve the accuracy of adaptive adjustment of PID controller parameters and ensure the efficient and stable operation of the target system under different working conditions.

[0070] As Figure 2 shown, it is a functional module diagram of a device for adaptive adjustment of PID controller parameters provided by an embodiment of this application.

[0071] In some embodiments, the device 2 for adaptive adjustment of PID controller parameters may include multiple functional modules composed of computer program segments. The computer programs of each program segment in the device 2 for adaptive adjustment of PID controller parameters can be stored in the memory of the server and executed by at least one processor to execute (see details in Figure 1 description) the functions of the method for adaptive adjustment of PID controller parameters.

[0072] In this embodiment, the device 2 for adaptive adjustment of PID controller parameters can be divided into multiple functional modules according to the functions it performs. The functional modules may include: a feature matrix module 21, a mode vector module 22, an initial parameter module 23, an optimized parameter module 24, a dynamic fusion module 25, and a threshold verification module 26. The module referred to in the present invention means a series of computer program segments that can be executed by at least one processor and can complete fixed functions, and are stored in the memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0073] The feature matrix module 21 is configured to acquire in real time the multi-source heterogeneous data of the controlled system collected by the distributed sensor array, and perform feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix representing the dynamic characteristics of the controlled system.

[0074] In an optional embodiment, the feature matrix module 21 is specifically configured to: Perform timestamp alignment processing on the multi-source heterogeneous data to obtain a time-aligned data set; Perform wavelet basis function denoising processing on the time-aligned data set to obtain a standardized feature data set; Perform time-domain feature extraction processing on the feature data set to obtain a time-domain feature set; Perform frequency-domain feature extraction processing on the feature data set to obtain a frequency-domain feature set; Perform spatio-temporal correlation feature extraction processing on the feature data set to obtain a spatio-temporal correlation feature set; Perform matrix construction according to the time-domain feature set, the frequency-domain feature set, and the spatio-temporal correlation feature set to obtain the quantization feature matrix.

[0075] The mode vector module 22 is configured to perform operating condition mode recognition based on deep fuzzy clustering on the quantization feature matrix, classify the operating conditions of the controlled system, and generate an operating condition mode vector according to the operating condition category.

[0076] In an optional embodiment, the mode vector module 22 is specifically configured to: Perform deep fuzzy c-means clustering processing on the quantization feature matrix to obtain a membership matrix; Perform operating condition continuity classification processing according to the membership matrix to obtain an operating condition classification data set; Perform dynamic characteristic analysis on the operating conditions in the operating condition classification data set according to the frequency-domain feature set to obtain a dynamic level; Perform coupling strength analysis according to the spatio-temporal correlation feature set, the operating condition classification data set, and the dynamic level to obtain a coupling strength feature set; Perform vector construction on the operating condition classification data set, the coupling strength feature set, and the dynamic level to obtain the operating condition mode vector.

[0077] The initial parameter module 23 is configured to perform non-linear mapping processing on the operating condition mode vector through a preset parameter space adaptive mapping method to obtain an initial set of PID parameters and an optimization space that match the current operating condition category.

[0078] In an optional embodiment, the initial parameter module 23 is specifically configured to: Perform coefficient mapping processing on the working condition classification data set in the working condition mode vector through a preset coefficient model to obtain the mapping coefficients of each working condition category; Construct a parameter space mapping function for each working condition level according to the mapping coefficient and the dynamic level; Perform non-linear mapping processing on the working condition mode vector according to the parameter space mapping function to obtain the initial set of PID parameters, and select the corresponding preset constraint range according to the initial set of PID parameters; Adjust the constraint range according to the coupling strength feature set to obtain an optimization domain; Adjust each initial parameter range in the optimization domain according to the working condition classification data set, the dynamic level, and the coupling strength feature set respectively to obtain the optimization space of each parameter in the initial set of PID parameters.

[0079] The optimization parameter module 24 is used to perform parallel parameter optimization processing on the initial set of PID parameters and the optimization space according to a preset hybrid optimization strategy to obtain an optimized parameter set that meets the dynamic response index.

[0080] In an alternative embodiment, the optimization parameter module 24 is specifically used for: Perform particle swarm optimization on the initial set of PID parameters to obtain a global optimal solution; Perform genetic algorithm processing on the optimization space to obtain multiple candidate optimization solutions; Perform weighted average calculation on the global optimal solution and the candidate optimization solutions to obtain a candidate optimized parameter set; Perform fitness evaluation on the candidate optimized parameter set to screen out a sub-optimal parameter set; Perform constraint verification on the sub-optimal parameter set to obtain the optimized parameter set.

[0081] The dynamic fusion module 25 is used to perform evidence theory fusion and anti-interference correction processing on the optimized parameter set to obtain a target PID parameter set, and update the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

[0082] In an alternative embodiment, the dynamic fusion module 25 is specifically used for: Perform evidence theory fusion processing on the optimized parameter set to obtain the basic probability assignment function values of each candidate solution in the optimized parameter set; Perform weighted combination on the candidate solutions according to the basic probability assignment function values through a preset Dempster combination rule to obtain an optimized solution set; Perform dynamic adjustment on the optimized solution set based on the anti-interference correction term to obtain an anti-interference corrected optimized solution set; Perform performance verification on the anti-interference corrected optimized solution set to screen out a target PID parameter set that meets the preset stability requirements.

[0083] In an optional embodiment, the PID controller parameter adaptive adjustment device 2 further includes a threshold verification electrical module 26, and the threshold verification electrical module 26 is used for: Update the PID control calculation formula according to the anti-interference corrected optimized solution set, and calculate the control input data at the current moment according to the output data at the current moment and the PID control calculation formula; Perform a rolling optimization window calculation on the control input data through a preset prediction model to obtain optimized control input data and parameter response change data; Perform calculation processing on the parameters to be verified according to the parameter response change data to obtain the Lyapunov exponent and dynamic indicators; Compare the Lyapunov exponent, the dynamic indicators, and the adjustment time in the parameter response change data with the preset verification threshold set one by one to screen out a target PID parameter set that meets the preset stability requirements.

[0084] It should be understood that the various change methods and specific embodiments in the methods provided in the above embodiments are equally applicable to the PID controller parameter adaptive adjustment device in this embodiment. Through the foregoing detailed description of the PID controller parameter adaptive adjustment method, those skilled in the art can clearly know the implementation method of the PID controller parameter adaptive adjustment device in this embodiment. For the sake of simplicity of the specification, it will not be described in detail here.

[0085] As Figure 3 shown, it is a schematic structural diagram of an electronic device provided by an embodiment of the present application.

[0086] In a preferred embodiment of the present invention, the electronic device 3 may include, but is not limited to: a memory 31, at least one processor 32, and at least one communication bus 33.

[0087] Those skilled in the art should understand that Figure 3 the structure of the electronic device 3 shown does not constitute a limitation on the embodiments of the present invention. The electronic device 3 may further include more or fewer other hardware or software than shown, or different component arrangements.

[0088] In some embodiments, the electronic device 3 is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application specific integrated circuits, programmable gate arrays, digital signal processors, and embedded devices, etc.

[0089] It should be noted that the electronic device 3 is only an example, and other existing or future electronic products that can be adapted to this application should also be included within the protection scope of this application and are hereby incorporated by reference.

[0090] In some embodiments, a computer program is stored in the memory 31, and when the computer program is executed by the at least one processor 32, all or part of the steps in the PID controller parameter adaptive adjustment method as described above are implemented. The memory 31 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc memories, magnetic disk memories, tape memories, or any other computer-readable medium capable of carrying or storing data. Further, the computer-readable storage medium mainly includes a storage program area and a storage data area. Among them, the storage program area can store an operating system, application programs required for at least one function, etc.

[0091] In some embodiments, the at least one processor 32 is the control core (Control Unit) of the electronic device 3, connecting various components of the entire electronic device 3 through various interfaces and circuits. By running or executing programs or modules stored in the memory 31, and calling data stored in the memory 31, it performs various functions of the electronic device 3 and processes data. For example, when the at least one processor 32 executes the computer program stored in the memory 31, it implements all or part of the steps of the PID controller parameter adaptive adjustment method described in the embodiments of the present application; or implements all or part of the functions of the PID controller parameter adaptive adjustment device. The at least one processor 32 may be composed of integrated circuits. For example, it may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same or different functions packaged, including a combination of one or more central processing units (Central Processing Unit, CPU), microprocessors, digital processing chips, graphics processors, and various control chips, etc.

[0092] In some embodiments, the at least one communication bus 33 is configured to implement connection communication between the memory 31 and the at least one processor 32, etc. Although not shown, the electronic device 3 may further include a power supply (such as a battery) for powering each component. Preferably, the power supply can be logically connected to the at least one processor 32 through a power management device, so as to implement functions such as management of charging, discharging, and power consumption management through the power management device. The power supply may also include any components such as one or more DC or AC power supplies, a recharge device, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device 3 may also include various sensors, a Bluetooth module, a Wi-Fi module, etc., which will not be elaborated here.

[0093] The above-mentioned integrated unit implemented in the form of software function modules can be stored in a computer-readable storage medium. The above-mentioned software function modules are stored in a storage medium, including several instructions for causing an electronic device (which may be a personal computer, an electronic device, or a network device, etc.) or a processor to execute part of the methods described in the various embodiments of the present application.

[0094] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0095] The module described as a separation component may or may not be physically separated. The component shown as a module may or may not be a physical unit, and it may be located in one place or distributed across multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0096] The above are all preferred embodiments of this application. The protection scope of this application is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of this application shall be covered within the protection scope of this application.

Claims

1. A method for adaptively adjusting the parameters of a PID controller, characterized in that, The method includes: Obtaining in real time the multi-source heterogeneous data of the controlled system collected by the distributed sensor array, and performing feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix characterizing the dynamic characteristics of the controlled system; Performing condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to classify the operating conditions of the controlled system, and generating a condition mode vector according to the condition category; Performing non-linear mapping processing on the condition mode vector through a preset parameter space adaptive mapping method to obtain an initial set of PID parameters and an optimization space matching the current condition category; Performing parallel parameter optimization processing on the initial set of PID parameters and the optimization space according to a preset hybrid optimization strategy to obtain an optimized parameter set that meets the dynamic response index; Performing evidence theory fusion and disturbance rejection correction processing on the optimized parameter set to obtain a target PID parameter set, and updating the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

2. The PID controller parameter self - adaptive adjustment method according to claim 1, wherein, The performing feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix characterizing the dynamic characteristics of the controlled system includes: Performing timestamp alignment processing on the multi-source heterogeneous data to obtain a time-aligned data set; Performing wavelet basis function denoising processing on the time-aligned data set to obtain a standardized feature data set; Performing time-domain feature extraction processing on the feature data set to obtain a time-domain feature set; Performing frequency-domain feature extraction processing on the feature data set to obtain a frequency-domain feature set; Performing spatio-temporal correlation feature extraction processing on the feature data set to obtain a spatio-temporal correlation feature set; Performing matrix construction according to the time-domain feature set, the frequency-domain feature set, and the spatio-temporal correlation feature set to obtain the quantization feature matrix.

3. The PID controller parameter adaptive adjustment method according to claim 2, wherein The condition category includes a condition classification data set, a coupling strength feature set, and a dynamic level. The performing condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to classify the operating conditions of the controlled system and generating a condition mode vector according to the condition category includes: Performing deep fuzzy c-means clustering processing on the quantization feature matrix to obtain a membership matrix; Performing condition continuity classification processing according to the membership matrix to obtain the condition classification data set; Performing dynamic characteristic analysis on the conditions in the condition classification data set according to the frequency-domain feature set to obtain the dynamic level; Performing coupling strength analysis according to the spatio-temporal correlation feature set, the condition classification data set, and the dynamic level to obtain the coupling strength feature set; Performing vector construction on the condition classification data set, the coupling strength feature set, and the dynamic level to obtain the condition mode vector.

4. The PID controller parameter self - adaptive adjustment method according to claim 3, wherein, The performing non-linear mapping processing on the condition mode vector through a preset parameter space adaptive mapping method to obtain an initial set of PID parameters and an optimization space matching the current condition category includes: Performing coefficient mapping processing on the condition classification data set in the condition mode vector through a preset coefficient model to obtain mapping coefficients for each condition category; Construct a parameter space mapping function for each working condition level according to the mapping coefficient and the dynamic level; Perform a non-linear mapping process on the working condition mode vector according to the parameter space mapping function to obtain the initial set of PID parameters, and select the corresponding preset constraint range according to the initial set of PID parameters; Adjust the constraint range according to the coupling strength feature set to obtain an optimization domain; Adjust each initial parameter range in the optimization domain according to the working condition classification data set, the dynamic level, and the coupling strength feature set to obtain the optimization space of each parameter in the initial set of PID parameters.

5. The PID controller parameter adaptive adjustment method according to claim 1, characterized in that The parallel parameter optimization process for the initial set of PID parameters and the optimization space according to a preset hybrid optimization strategy to obtain an optimized parameter set that meets the dynamic response index includes: Perform a particle swarm algorithm process on the initial set of PID parameters to obtain a global optimal solution; Perform a genetic algorithm process on the optimization space to obtain multiple candidate optimization solutions; Perform a weighted average calculation on the global optimal solution and the candidate optimization solutions to obtain a candidate optimized parameter set; Perform a fitness evaluation on the candidate optimized parameter set to screen out a sub-optimal parameter set; Perform a constraint verification on the sub-optimal parameter set to obtain the optimized parameter set.

6. The PID controller parameter self - adaptive adjustment method according to claim 1, characterized in that The evidence theory fusion and anti-interference correction process on the optimized parameter set to obtain a target PID parameter set, and the proportional coefficient, integral coefficient, and differential coefficient of the PID controller are updated in real time according to the target PID parameter set, including: Perform an evidence theory fusion process on the optimized parameter set to obtain the basic probability assignment function values of each candidate solution in the optimized parameter set; Perform a weighted combination of the candidate solutions according to the basic probability assignment function values through a preset Dempster combination rule to obtain an optimized solution set; Perform a dynamic adjustment on the optimized solution set based on an anti-interference correction term to obtain an anti-interference corrected optimized solution set; Perform a performance verification on the anti-interference corrected optimized solution set to screen out a target PID parameter set that meets the preset stability requirements, and update the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

7. The PID controller parameter adaptive adjustment method according to claim 6, wherein The performance verification on the anti-interference corrected optimized solution set to screen out a target PID parameter set that meets the preset stability requirements includes: Update the PID control calculation formula according to the anti-interference corrected optimized solution set, and calculate the control input data at the current moment according to the output data at the current moment and the PID control calculation formula; Perform a rolling optimization window calculation on the control input data through a preset prediction model to obtain optimized control input data and parameter response change data; Perform a calculation process on the parameters to be verified according to the parameter response change data to obtain the Lyapunov exponent and dynamic index; Compare the Lyapunov exponent, the dynamic index, and the adjustment time in the parameter response change data with a preset verification threshold set one by one to screen out a target PID parameter set that meets the preset stability requirements.

8. A PID controller parameter adaptive adjustment device, characterized in that, The device includes: A feature matrix module, configured to obtain in real time multi-source heterogeneous data of a controlled system collected by a distributed sensor array, and perform feature quantization processing on the multi-source heterogeneous data to construct a quantization feature matrix characterizing the dynamic characteristics of the controlled system; A mode vector module, configured to perform operating condition mode recognition based on deep fuzzy clustering on the quantization feature matrix to classify the operating conditions of the controlled system, and generate an operating condition mode vector according to the operating condition category; An initial parameter module, configured to perform a non-linear mapping process on the operating condition mode vector through a preset parameter space adaptive mapping method to obtain an initial set of PID parameters and an optimization space matching the current operating condition category; An optimized parameter module, configured to perform parallel parameter optimization on the initial set of PID parameters and the optimization space according to a preset hybrid optimization strategy to obtain an optimized parameter set that meets the dynamic response index; A dynamic fusion module, configured to perform evidence theory fusion and anti-interference correction processing on the optimized parameter set to obtain a target PID parameter set, and update the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in real time according to the target PID parameter set.

9. An electronic device, characterized in that, The electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the PID controller parameter adaptive adjustment 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 the processor, the steps of the PID controller parameter adaptive adjustment method according to any one of claims 1 to 7 are implemented.

Images