Intelligent control system of microcirculation purifier based on negative ion concentration feedback

Abstract

The application discloses a microcirculation purifier intelligent control system based on negative ion concentration feedback and belongs to the technical field of intelligent equipment control. Specifically, the system comprises an environment sensing module, a dynamic working condition prediction module, an intelligent regulation and control module and a negative ion concentration feedback module. The environment sensing module collects multi-dimensional environment and equipment state data in real time. The dynamic working condition prediction module constructs a fusion prediction model and performs dynamic working condition prediction and parameter setting processing. The intelligent regulation and control module adjusts the frequency, pulse width and operation mode of a high-voltage pack switch, realizes accurate regulation and control of the tungsten gold needle corona discharge state, and performs target site effective ion concentration distribution optimization and arrival time estimation processing. The negative ion concentration feedback module outputs a personalized physiotherapy scheme. The application constructs a closed-loop control model of "environment-feedback-optimization", dynamically adjusts the key parameters of a negative ion generator, and realizes stable generation of high-concentration, small-particle-size and high-quality negative ions.

Description

Technical Field

[0001] This invention relates to the field of intelligent device control technology, and in particular to an intelligent control system for a micro-circulation purifier based on negative ion concentration feedback. Background Technology

[0002] The integration of intelligent control technology and negative ion purification technology has become an important trend in the industry. Early micro-circulation purifiers mostly used a constant power control mode, which could not adjust its operating status according to environmental changes. With the development of sensor technology, some devices have introduced simple concentration detection modules, but lack precise feedback control mechanisms, making it difficult to achieve a dynamic balance of negative ion concentration. In recent years, advancements in technologies such as the Internet of Things and intelligent sensing have driven the development of purification equipment towards intelligence and precision. Intelligent control systems based on feedback mechanisms have become the key to resolving the contradiction between purification efficiency and user experience.

[0003] Although significant progress has been made in negative ion microcirculation purification technology, existing technologies still have obvious shortcomings: 1. Integer-order PID control has poor adaptability to the nonlinear and time-varying characteristics of negative ion generators; 2. Existing equipment has slow output response and lacks execution feedback closed loop; 3. Traditional equipment has a one-size-fits-all output, which cannot meet the personalized needs of users. Summary of the Invention

[0004] This application presents an intelligent control system for a micro-circulation purifier based on negative ion concentration feedback. The system aims to dynamically adjust the key parameters of the negative ion generator by constructing a closed-loop control model of "environment-feedback-optimization". This system can quickly and stably generate high-concentration, small-particle-size high-quality negative ions under various environmental conditions, thereby solving the above-mentioned defects in the prior art (points 1 to 3).

[0005] This application provides the following technical solution: This application discloses an intelligent control system for a microcirculation purifier based on negative ion concentration feedback. The system includes an environmental sensing module, a dynamic operating condition prediction module, an intelligent control module, and a negative ion concentration feedback module. The environmental sensing module collects multi-dimensional environmental and equipment status data in real time through built-in temperature, humidity, dust particle, and ion concentration sensors. After preprocessing, it generates a structured raw feature dataset. The dynamic operating condition prediction module, based on the raw feature dataset, constructs a FOPID-RBFNN fusion prediction model using fractional-order PID and radial basis function neural networks to perform dynamic operating condition prediction and parameter tuning, obtaining a set of optimized negative ion control parameters. The intelligent control module obtains control commands from the optimized negative ion control parameter set and adjusts the high-voltage transformer switching frequency, pulse width, and operating mode using high-voltage drive and mode self-adaptation methods to achieve precise control of the corona discharge state of the tungsten gold needle, ensuring stable output from the negative ion generator. The negative ion concentration feedback module obtains user-personalized parameters and real-time feedback data. Using the dynamic operating condition prediction module combined with response curve simulation methods, it optimizes the effective ion concentration distribution and arrival time prediction of the target area, outputting a personalized physiotherapy plan.

[0006] Furthermore, the environmental perception module includes a heterogeneous data acquisition submodule, a signal preprocessing submodule, and a data fusion submodule; The heterogeneous data acquisition submodule acquires multi-source heterogeneous data by deploying multi-source heterogeneous data sensors to form a raw data stream. The signal preprocessing submodule acquires the raw data stream, uses physical range verification to remove outliers and performs linear interpolation to fill them, and combines sliding window mean and median filtering to suppress noise and obtain the cleaned effective data stream. The data fusion submodule acquires the cleaned effective data stream, performs linear interpolation resampling based on a unified time base, and aligns multi-source data of different frequencies into synchronized multi-dimensional data frames.

[0007] Furthermore, the environment perception module also includes a feature engineering submodule; The feature engineering submodule acquires multidimensional data frames, calculates derived features, and standardizes the features using the Min-Max normalization method to obtain a structured original feature dataset.

[0008] Furthermore, the dynamic operating condition prediction module includes a feature parsing submodule, a target concentration deviation calculation submodule, and an RBFNN feedforward parameter prediction submodule; The feature parsing submodule uses data caching technology to cache, extract features, and verify the integrity of the structured original feature dataset to obtain the feature vector at the current time. The target concentration deviation calculation submodule uses an environmental dynamic correction algorithm to dynamically set the target concentration based on the environmental parameters in the feature vector and the user-defined physiotherapy mode data, and calculates the control deviation signal by combining the measured ion concentration. The RBFNN feedforward parameter prediction submodule uses a radial basis function neural network (RBFNN) to perform forward calculations on the current environmental feature vector and obtains a set of PID control coefficients through nonlinear mapping.

[0009] Furthermore, the dynamic operating condition prediction module also includes a controller calculation submodule and a parameter encapsulation submodule; The controller calculation submodule uses the Oustaloup filtering algorithm to perform fractional-order calculus operations and synthesis on the PID control coefficients and control deviation signals to obtain the target duty cycle control quantity. The parameter encapsulation submodule uses a parameter group encapsulation method to structure and package the target duty cycle control quantity, operating mode flag, and PID control coefficient, and transmits them through internal inter-process communication to generate a negative ion optimized control parameter group.

[0010] Furthermore, the intelligent control module includes a parameter parsing submodule, a mode self-adaptation submodule, and a discharge control submodule; The parameter parsing submodule uses parameter parsing and integrity verification technology to perform data verification, field extraction and PWM instruction conversion processing on the negative ion optimized control parameter group to obtain the PWM control instruction package; The mode self-adaptive submodule performs dynamic amplitude limiting and first-order low-pass filtering smoothing on the target duty cycle control quantity in the PWM control instruction packet according to the running mode flag bit, so as to obtain the smoothed execution duty cycle. The discharge control submodule uses PWM pulse width modulation technology to generate signals and perform high voltage transformation on the smoothed execution duty cycle, driving the corona discharge of the tungsten gold needle tip to output a negative ion flow.

[0011] Furthermore, the intelligent control module also includes an execution feedback submodule and an exception handling submodule; The execution feedback submodule uses execution feedback acquisition technology to collect concentration and duty cycle data of the actual output negative ion flow, and sends the feedback data packet back to the dynamic operating condition prediction module for online fine-tuning of the RBFNN model and calibration of the FOPID integral term. The anomaly handling submodule utilizes a multi-threshold hierarchical protection mechanism to perform real-time safety monitoring and hierarchical response to the high-voltage transformer current, input voltage, and equipment temperature, ensuring safe shutdown and limited operation of the equipment under abnormal conditions.

[0012] Furthermore, the negative ion concentration feedback module includes a personalized modeling submodule, a response curve simulation submodule, and a physiotherapy interaction submodule; The personalized modeling submodule acquires the user's height, weight, treatment area, preference pattern, and sensitive body constitution data. It then uses the personalized physiotherapy constraint model to perform constraint modeling and correction factor calculation to obtain personalized correction parameters and physiotherapy area codes. The response curve simulation submodule uses the system transfer function model in the FOPID controller to perform numerical simulation and prediction of the dynamic characteristic parameters after the physiotherapy site coding is corrected and the current measured concentration, so as to obtain the remaining time to reach the optimal concentration and the maximum overshoot. The physiotherapy interaction submodule utilizes dynamic simulation display technology and real-time feedback control technology to perform visualization rendering and threshold determination processing on the remaining time to reach the optimal concentration and the maximum overshoot data, thereby obtaining a real-time interactive interface and dynamic mode adjustment instructions.

[0013] Furthermore, the negative ion concentration feedback module also includes a collaborative optimization submodule and a model evolution submodule; The collaborative optimization submodule uses a collaborative control parameter generation method based on preference pattern mapping to process the dynamic pattern adjustment instruction data, transforming control target correction parameters and embedding constraints to obtain a collaborative control parameter set. The model evolution submodule uses the logistic regression algorithm to incrementally learn from user behavior time series data, and updates the model weights through gradient descent to obtain the updated user preference model parameters and the personalized physiotherapy plan for the next time step.

[0014] Compared with the prior art, this application achieves at least one of the following beneficial effects: 1. By utilizing multi-source heterogeneous sensor fusion technology, heterogeneous data is collected and preprocessed in real time to obtain a normalized structured feature dataset. Compared with existing technologies that rely on single concentration feedback or simple threshold judgment, this solves the problems of existing technologies lacking comprehensive perception of environmental parameters and equipment status and the inability to coordinate between data sources.

[0015] 2. By using the FOPID-RBFNN fusion prediction model, feedforward parameter prediction and fractional-order PID control quantity synthesis are performed on the feature vector to obtain the optimal five-dimensional control parameter set that adapts to different working conditions, which solves the problem of poor adaptability of the integer-order PID control of the existing technology to the nonlinearity and time-varying characteristics of the negative ion generator.

[0016] 3. By utilizing PWM pulse width modulation and multi-threshold hierarchical protection technology, the control parameters are adapted to different modes, dynamically limited, driven by high voltage, and monitored for safety. This results in accurate, stable, and safe negative ion output and complete execution feedback data, solving the problems of slow output response and lack of execution feedback closed loop in existing technology equipment.

[0017] 4. By utilizing personalized constraint modeling and logistic regression incremental learning techniques, the physiological parameters, preference patterns, and historical behavioral data input by users are constrained and transformed, and then subjected to response simulation, interface interaction, and model evolution. This results in personalized physiotherapy plans, effect predictions, and continuously optimized user preference models, solving the problem that traditional equipment in existing technologies provides a one-size-fits-all output that cannot meet the personalized needs of users. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments or examples of the present invention, the drawings used in the embodiments or examples will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained according to these drawings without creative effort.

[0019] Figure 1 This is an architecture diagram of an intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to the present invention. Figure 2 This is an architecture diagram of the environmental perception module of the present invention; Figure 3 This is an architecture diagram of the dynamic operating condition prediction module of the present invention; Figure 4 This is an architecture diagram of the intelligent control module of the present invention; Figure 5 This is an architecture diagram of the negative ion concentration feedback module of the present invention. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0022] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0024] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0026] Specific embodiments of the present invention are described below: See appendix Figure 1A microcirculation purifier intelligent control system based on negative ion concentration feedback is disclosed. The system includes an environmental sensing module, a dynamic operating condition prediction module, an intelligent control module, and a negative ion concentration feedback module. The environmental sensing module collects multi-dimensional environmental and equipment status data in real time through built-in temperature, humidity, microparticle, and ion concentration sensors, and generates a structured raw feature dataset after preprocessing. The dynamic operating condition prediction module, based on the raw feature dataset, constructs a FOPID-RBFNN fusion prediction model using fractional-order PID and radial basis function neural networks to perform dynamic operating condition prediction and parameter tuning, obtaining a set of optimized negative ion control parameters. The intelligent control module obtains control commands from the optimized negative ion control parameter set and adjusts the high-voltage transformer switching frequency, pulse width, and operating mode using high-voltage drive and mode self-adaptation methods to achieve precise control of the tungsten gold needle corona discharge state, ensuring stable output from the negative ion generator. The negative ion concentration feedback module obtains user-personalized parameters and real-time feedback data, and uses the dynamic operating condition prediction module combined with response curve simulation methods to optimize the effective ion concentration distribution and arrival time prediction of the target area, outputting a personalized physiotherapy plan.

[0027] The aforementioned intelligent control system for a micro-circulation air purifier based on negative ion concentration feedback is centered on constructing a closed-loop control model of "environment-feedback-optimization". It dynamically adjusts the key parameters of the negative ion generator, enabling it to quickly and stably generate high-concentration, small-particle-size high-quality negative ions under various environmental conditions, and optimizes energy utilization efficiency. Around the "optimized control model", it constructs a closed-loop intelligent system integrating sensing, prediction, regulation, and interaction through the FOPID-RBFNN fusion algorithm.

[0028] For specific implementation details, please refer to the appendix. Figure 2 The environmental perception module includes a heterogeneous data acquisition submodule, a signal preprocessing submodule, and a data fusion submodule; The heterogeneous data acquisition submodule acquires multi-source heterogeneous data by deploying multi-source heterogeneous data sensors to form a raw data stream. The signal preprocessing submodule acquires the raw data stream, uses physical range verification to remove outliers and performs linear interpolation to fill them, and combines sliding window mean and median filtering to suppress noise and obtain the cleaned effective data stream. The data fusion submodule acquires the cleaned effective data stream, performs linear interpolation resampling based on a unified time base, and aligns multi-source data of different frequencies into synchronized multi-dimensional data frames.

[0029] In practice, the environment perception module also includes a feature engineering submodule; The feature engineering submodule acquires multidimensional data frames, calculates derived features, and standardizes the features using the Min-Max normalization method to obtain a structured original feature dataset.

[0030] Its multi-source heterogeneous data includes: 1-temperature, 2-relative humidity, 3-PM2.5 concentration, 4-negative ion concentration, 5-input voltage, and 6-high voltage transformer current; for analog signals, sampling is performed through a built-in 12-bit ADC converter with a resolution of 0.1%; for digital signals, data values ​​are directly read through the UART bus, and all data is packaged into data frames as the raw data stream.

[0031] The physical range verification eliminates anomalies. The physical range inspection rules include: 1- Temperature range set to 0-50℃; 2- Relative humidity range set to 0-100%; 3- Negative ion concentration range set to 0-1000W / cm³. 3 The electrical range check rules include: 1- Input voltage range is set to 100-240V; high voltage transformer current is set to 0-2A; data points exceeding the above ranges are removed, and the data is filled by linear interpolation using the value of the previous valid moment to maintain the continuity of the data stream.

[0032] It combines sliding window mean and median filtering for noise suppression, including: 1- For slowly changing physical quantities (i.e., temperature, relative humidity), sliding window mean filtering is used, and its calculation formula is: , where x t The current filtered output value; N is the window size (set to 3); x t-i 1. The original values ​​of the i sampling points in the window; 2. For physical quantities that change rapidly and are susceptible to electromagnetic interference (i.e., negative ion concentration, input voltage and high voltage transformer current), use the median filtering of the sliding window (i.e., take the median).

[0033] The above feature standardization using the Min-Max normalization method includes the following steps: Step 1) Based on the original features, calculate derived features to enhance the model's ability to perceive dynamic processes. These derived features include: 1- Temperature change rate (the temperature difference between the current moment and 3 seconds ago, divided by 3 seconds); 2- Humidity change rate (similar to temperature, calculate the rate of change of relative humidity); 3- Negative ion concentration stability index (the standard deviation of ion concentration within the last 5 seconds); 4- Voltage fluctuation amplitude (the difference between the maximum and minimum values ​​of the input voltage within the last second). Combine these four derived features into an extended feature set. Step 2) Based on Step 1), for the extended feature set, the system stores the historical maximum and minimum values, and uses the Min-Max normalization method to standardize the features. The normalization formula is as follows: , where x归 x is the normalized eigenvalue; x is the eigenvalue at the current time; x min x is the minimum value of the feature in historical data; max The maximum value of the feature in the historical data; for the validity flag, non-numerical features are not normalized and the original flag is retained to obtain the normalized feature vector; Step 3) Based on Step 2), perform an integrity check on the feature vector of the current frame. The check must meet the following conditions: 1- All necessary features are valid values; 2- If one of the features is marked as invalid for 3 consecutive seconds, the entire frame is marked as "unavailable" and a device alarm is triggered. When both conditions are met, the integrity check passes and the frame data is encapsulated into a structured raw feature dataset.

[0034] For specific implementation details, please refer to the appendix. Figure 3 The dynamic operating condition prediction module includes a feature parsing submodule, a target concentration deviation calculation submodule, and an RBFNN feedforward parameter prediction submodule. The feature parsing submodule uses data caching technology to cache, extract features, and verify the integrity of the structured original feature dataset to obtain the feature vector at the current time. The target concentration deviation calculation submodule uses an environmental dynamic correction algorithm to dynamically set the target concentration based on the environmental parameters in the feature vector and the user-defined physiotherapy mode data, and calculates the control deviation signal by combining the measured ion concentration. The RBFNN feedforward parameter prediction submodule uses a radial basis function neural network (RBFNN) to perform forward calculations on the current environmental feature vector and obtains a set of PID control coefficients through nonlinear mapping.

[0035] The above feature extraction and integrity verification process includes the following steps: Step 1) Allocate a circular buffer in memory, which is a fixed-size array with a capacity of 10 frames. Set a read pointer and a write pointer. Each time a new frame of data is received, it is written to the position pointed to by the write pointer. Then the write pointer moves forward one position. When the buffer is full of 10 frames, when new data is written later, it will overwrite the oldest frame. This will obtain the structured feature data of the 10 most recent frames stored in the circular buffer. Step 2) Based on Step 1), locate the most recently written frame from the buffer, extract the complete feature vector from the frame and denote it as X1. At the same time, extract the measured ion concentration sequence of the last 5 frames from the buffer and denote it as C1. Step 3) Based on step 2), perform a range check on each feature value in X1. The value should be within the range of [0, 1]. If it exceeds this range, it indicates that there is an abnormality in the data processing or sensor system alarm. Check the validity flag of each feature value. If "invalid" appears, the feature value is unreliable. Output the verification result. Step 4) Based on step 3), pack X1, C and the timestamp into a standard output data packet (which includes the feature vector of the current time).

[0036] Its target concentration deviation calculation submodule includes the following steps in its implementation process: Step 1) The system presets a baseline target concentration value C. 基准 =500w pieces / cm 3 The baseline value is stored in the device firmware; Step 2) Perform dynamic environmental corrections, including humidity correction coefficients, temperature correction coefficients, and PM2.5 correction coefficients. Regarding the humidity correction coefficient, negative ions readily combine with moisture to form heavy ions in high humidity environments, reducing the proportion of small particles; in low humidity environments, diffusion attenuation is rapid. Therefore, the humidity correction formula set is as follows: , where k RH Here are the temperature correction coefficients (used to adjust the target concentration based on relative humidity); RH represents relative humidity; 30% is the low humidity threshold; 60% is the high humidity threshold; 1.2 is the initial correction value for low humidity (the coefficient was verified through environmental chamber experiments; the attenuation rate of negative ions during transmission reaches 20%-30%, therefore the target concentration is increased by 20%, i.e., the upper limit correction coefficient of 1.2); 0.005 is the correction slope in the low humidity region (obtained through linear interpolation fitting, with optimal compensation coefficients of 10%, 20%, and 30%); 1.0 is the baseline correction value (i.e., the upper limit of compensation under extreme conditions); 0.015 is the correction slope in the high humidity region (obtained through linear interpolation fitting, with optimal compensation coefficients of 60%, 70%, and 80%); for the temperature correction coefficient, temperature affects molecular activity and ionization efficiency, therefore the temperature correction formula set is as follows: , where k T Here, T is the temperature correction coefficient (used to compensate for the effect of temperature on the ionization efficiency of air molecules); 10℃ is the low-temperature threshold; 35℃ is the high-temperature threshold; 1.1 is the initial correction value for the low-temperature zone (i.e., the set upper limit value to compensate for low-temperature loss); 0.01 is the low-temperature zone correction slope (the slope is fitted by testing three points at 0℃, 5℃, and 10℃); 1.0 is the baseline correction value; 0.015 is the high-temperature correction slope (obtained by fitting the slope by testing three points at 35℃, 40℃, and 45℃). For the PM2.5 correction coefficient, since particulate matter in the air adsorbs and consumes negative ions, high concentrations of PM2.5 require increasing the target concentration to maintain an effective therapeutic concentration. Based on this principle, the PM2.5 correction coefficient calculation formula is as follows: , where k PM This is the PM2.5 correction factor (used to compensate for the adsorption and consumption of negative ions by particulate matter); PM is the PM2.5 concentration; 35 μg / m³ 3 The upper limit for good air quality; 150 μg / m³ 3 1.0 is the severe pollution threshold; 0.002 is the PM2.5 correction slope (obtained by fitting a linear relationship with a slope of 0.002 by testing the optimal compensation coefficients at four PM2.5 concentrations of 35, 70, 100, and 150 μg / m³); 1.23 is the maximum correction upper limit (set according to the linear relationship to prevent equipment overload caused by excessively high target concentrations in extreme pollution environments); multiplying the three correction coefficients and limiting the range to [0.7, 1.3] yields the total environmental correction factor k. 环境 ; Step 3) The system is set to three physiotherapy modes: high efficiency, comfort, and energy saving. The influence factors on the target concentration are set to high efficiency = 1.0, comfort = 0.9, and energy saving = 0.8 respectively. At the same time, a body shape correction factor is embedded, based on the baseline value range [0,1], where the larger the body surface area, the larger the value. The influence factor on the target concentration is used as the user mode correction factor and denoted as k. 模式 ; Step 4) Calculate the comprehensive target concentration value based on steps 1 to 3), using the formula: C 目标 =C 基准 +k 环境 +k 模式 C 目标 The overall target concentration value (ensuring it is within the safe output range of the equipment); C 基准 The baseline target concentration value; k 环境 For total environmental correction factor; k 模式 Adjustment factor for user mode; Step 5) Based on Step 4), calculate the control deviation, including the current deviation, the cumulative deviation, and the rate of change of deviation; for the current deviation, the calculation formula is: e1(t) = C 目标 -C 实际 Where e1(t) is the current deviation value; C 目标 The overall target concentration value; C 实际 This represents the actual deviation; the formula for calculating the cumulative deviation is: , where I 累积 (t) represents the cumulative deviation at the current moment (as input to the integral term of the fractional PID); ∑ represents the summation sign (summing the deviation values ​​over 5 time points from 0 to 4); e(tk·Δt) represents the historical deviation value (cumulative compensation for persistent deviations); the formula for the deviation change rate is: Where e2(t) is the deviation change rate at the current moment; e1(t) is the current deviation value; e(t-Δt) is the deviation of the previous cycle; Δt is the control cycle time interval; the purpose of the division is to approximate the derivative of the deviation with respect to time by dividing the difference between the deviations of two adjacent frames by the time interval, and finally package the current deviation, the cumulative deviation, and the deviation change rate into a deviation data packet.

[0037] The above-mentioned use of radial basis function neural network (RBFNN) to perform forward computation on the current environment feature vector includes the following process: Step 1) The hidden layer of the RBFNN consists of M radial basis function neurons. In this system, M=20, covering the typical working condition space. Each hidden layer neuron j corresponds to a Gaussian radial basis function, which calculates the similarity (i.e., Euclidean distance) between the input vector and the predefined "typical working condition center point". The closer the distance, the larger the output value. The formula for calculating the Gaussian radial basis function is as follows: ,in Let be the output value of the j-th neuron; exp be the exponential function; X1 be the feature vector of the current time step in the standard output data packet; c j Let be the center point vector of the j-th radial basis neuron (the parameter is determined from historical operating data during the training phase); σ is the square of the Euclidean distance (representing the degree of difference between the current operating condition and the j-th typical operating condition); j Let be the width parameter of the j-th radial basis neuron (this parameter is determined during model training using the average distance method); Step 2) Based on Step 1), the output layer contains a total of 3 neurons, which correspond to the three coefficients K of the PID controller. p K i and K d The outputs of the hidden layer are linearly weighted and summed using each output neuron, plus a bias term, to obtain the final PID coefficients. The formulas for calculating the three coefficients are as follows: K p K i and K d These are the optimal proportional coefficient, integral coefficient, and differential coefficient, respectively; M is the number of neurons in the hidden layer; j is the neuron filament; w pj w ij and w dj These are the connection weights from the j-th hidden layer neuron to the proportional coefficient output neuron, the connection weight from the j-th hidden layer neuron to the integral coefficient output neuron, and the connection weight from the j-th hidden layer neuron to the differential coefficient output neuron, respectively. (The activation value of the j-th hidden layer neuron is calculated via step 1); b p b i and bd These are the bias terms for the proportional coefficient output neuron, the integral coefficient output neuron, and the derivative coefficient output neuron, respectively; the three coefficients obtained through the formula serve as a set of PID control coefficients.

[0038] In practice, the dynamic operating condition prediction module also includes a controller calculation submodule and a parameter encapsulation submodule; The controller calculation submodule uses the Oustaloup filtering algorithm to perform fractional-order calculus operations and synthesis on the PID control coefficients and control deviation signals to obtain the target duty cycle control quantity. The parameter encapsulation submodule uses a parameter group encapsulation method to structure and package the target duty cycle control quantity, operating mode flag, and PID control coefficient, and transmits them through internal inter-process communication to generate a negative ion optimized control parameter group.

[0039] It performs fractional calculus operations and synthesis, including: Step 1) Extract the current deviation e1(t) from the deviation data packet, and introduce the two core parameters λ and μ of the fractional PID controller, which represent the integral order and the derivative order respectively. Combine them with the PID control coefficients to form the FOPID input parameter set. Step 2) Based on the parameters of the Oustaloup filter preset in Step 1), including the approximate order N=5 (a balance between accuracy and real-time performance), the operating frequency band is between [0.01Hz, 100Hz] (to cover the dynamic response range of the negative ion generator), and the integral order operator λ=0.8 and the differential order operator μ=0.6 (values ​​set by the fractional-order characteristics experiment of the corona discharge process), two independent Oustaloup filters are constructed respectively, and their filter transfer function formulas are as follows: ,in ω is the fractional-order Laplace operator (used to describe fractional-order differentiation and integration of a signal); K is the gain adjustment coefficient; π is the multiplication symbol (multiplying all terms from -N to N using k); s is the Laplace operator; ω k and These are the k-th pole and k-th zero, respectively (determining the gain drop / rise characteristics of the filter at the corresponding frequency); N is the approximation order (used to control the approximation accuracy, the larger N is, the more accurate the approximation); by implementation, the coefficient tables of two independent Oustaloup filters are obtained (i.e., the fractional-order integrator filter and the fractional-order derivative filter). Step 3) Based on Step 2), calculate the integral term of the fractional-order integrator filter and the differential term of the fractional-order differential filter respectively. The formula for the integral term is: , where I f(n) represents the fractional integral output at the nth sampling time (used for the synthesis of subsequent FOPID control quantities); n is the discrete-time index; M is the filter order; a k b is the forward coefficient; k Here, e(nk) represents the feedback coefficient; e(nk) represents the deviation input at the nk-th sampling time; I f (nk) represents the historical output at the nk-th sampling time; the integral of the fractional integral term is calculated using the formula; the formula for its differential term is: D f (n) represents the fractional-order differential output at the nth sampling time (used for synthesizing the FOPID control quantity, providing a damping effect); D f (nk) represents the historical output value of the differentiating filter; Step 4) Based on Step 3), calculate the output control quantity of the FOPID controller. The formula for calculating the output FOPID control quantity is: U(t) = K p ·e1(t)+K i ·I f (n)+K d ·D f (n), where U(t) is the output control quantity of the FOPID controller (i.e., the target duty cycle control quantity); K p K i and K d These are the optimal proportional coefficient, integral coefficient, and differential coefficient, respectively; I f (n) represents the fractional integral output at the nth sampling time; D f (n) represents the fractional-order differential output at the nth sampling time; e1(t) represents the current deviation; after the calculation is completed, U(t) is limited, and the limited U(t) is used as the target duty cycle control quantity.

[0040] Its motion pattern is determined as follows: 1. Fast response mode: U(t) > 0.9 and |e1(t)| > 0.2 × C 目标 (If the deviation exceeds 20% of the target concentration), the mode flag m=01; 2. Steady-state maintenance mode: |e1(t)| < 0.05 × C 目标 (The deviation is less than 5% of the target concentration), and the mode flag m=10; 3. Transition mode: Except for the above two modes, the mode flag m=00.

[0041] For specific implementation details, please refer to the appendix. Figure 4 The intelligent control module includes a parameter parsing submodule, a mode self-adaptation submodule, and a discharge control submodule. The parameter parsing submodule uses parameter parsing and integrity verification technology to perform data verification, field extraction and PWM instruction conversion processing on the negative ion optimized control parameter group to obtain the PWM control instruction package; The mode self-adaptive submodule performs dynamic amplitude limiting and first-order low-pass filtering smoothing on the target duty cycle control quantity in the PWM control instruction packet according to the running mode flag bit, so as to obtain the smoothed execution duty cycle. The discharge control submodule uses PWM pulse width modulation technology to generate signals and perform high voltage transformation on the smoothed execution duty cycle, driving the corona discharge of the tungsten gold needle tip to output a negative ion flow.

[0042] Its data verification, field extraction, and PWM command conversion processing include the following steps: Step 1) Verify that U(t)∈[0,1], mode flag m∈[00,01,10] and PID control coefficient are within the preset range; verify whether the timestamps are continuous to prevent frame loss; if any verification fails, discard the current frame, use the parameter group that was successfully executed in the previous cycle, and record the exception log. Step 2) Based on Step 1), extract U(t) and m, and convert the floating-point duty cycle U(t) into the value of the PWM timer's compare value register. Step 3) Based on step 2), encapsulate the PWM comparison value, the running mode flag bit m, and the timestamp into a PWM control instruction package.

[0043] Its pattern self-adaptive sub-module, the implementation steps include: Step 1) Set the maximum single adjustment step size u according to the mode flag m. max Including: 1- Fast Response Mode (u max =0.1); 2-transition mode (u max =0.05); 3-steady-state maintenance mode (u max =0.02), calculate the difference between the current duty cycle and the target duty cycle, when it exceeds u max Then it is restricted to be less than u max To prevent sudden changes in duty cycle from causing electrical stress impact on the high-voltage module; Step 2) Based on Step 1), filter the duty cycle after amplitude limiting to suppress high-frequency fluctuations. The filtering formula is as follows: u smooth (n)=a 滤波 ·u limit (n)+(1-a 滤波 )·u smooth (n-1), where u smooth (n) represents the smoothed duty cycle of the current period; a 滤波The filter coefficients (serving as smoothing factors, their values ​​range from 0 to 1 depending on the motion pattern); (1-a 滤波 ) represents the smoothed duty cycle of the previous period; n is the discrete-time index; where the smoothing factor a 滤波 The value is dynamically adjusted according to the motion mode as follows: 1 - Fast response mode a 滤波 =60%; 2-Steady-state maintenance mode a 滤波 =30%; 3-Transition mode a 滤波 =10%, the smoothed duty cycle of the current cycle after filtering, is used as the smoothed execution duty cycle.

[0044] In practice, the intelligent control module also includes an execution feedback submodule and an exception handling submodule; The execution feedback submodule uses execution feedback acquisition technology to collect concentration and duty cycle data of the actual output negative ion flow, and sends the feedback data packet back to the dynamic operating condition prediction module for online fine-tuning of the RBFNN model and calibration of the FOPID integral term. The anomaly handling submodule utilizes a multi-threshold hierarchical protection mechanism to perform real-time safety monitoring and hierarchical response to the high-voltage transformer current, input voltage, and equipment temperature, ensuring safe shutdown and limited operation of the equipment under abnormal conditions.

[0045] The online fine-tuning of its RBFNN model includes the following steps: Step 1) Set the control cycle to 100ms. When it ends, trigger the execution result acquisition process, read the actual ion concentration value at the current moment from the shared data area of ​​the environmental sensing module, record the duty cycle actually written to the PWM register in this cycle, read the high voltage pack current, read the flag bit of whether the protection has been triggered in this cycle, and construct the original execution dataset as a feedback data packet. Step 2) Based on Step 1), the feedback data packet is sent back to the dynamic operating condition prediction module through the internal inter-process communication (IPC) mechanism. The system uses a lock-free shared memory ring buffer to achieve low-latency data transmission. Step 3) Based on Step 2), the dynamic operating condition prediction module compares the feedback measured concentration with the expected concentration predicted by the FOPID model in each control cycle, and calculates the prediction error. The formula is as follows: ,in For prediction error; C 实际 For actual deviation; C 预期 The expected concentration predicted by the FOPID model; C 目标 The overall target concentration value is determined; when the prediction error reaches 10 consecutive control cycles (1 second), the prediction accuracy of the current RBFNN model is determined to have decreased, triggering fine-tuning; Step 4) Based on step 3), fine-tune the weights of the RBFNN output layer using gradient descent. The gradient descent formula is as follows: , where w new For the updated RBFNN weights; w old The weights are the unupdated RBFNN weights; η is the learning rate (set to 0.01). is the gradient (used to indicate the direction and magnitude of weight adjustment); E is the cost function (used to measure the difference between the model's predicted value and the actual optimal value).

[0046] Its FOPID integral term calibration includes: 1. Monitor the deviation between the real-time concentration and the target concentration. The steady-state condition is: |e²(t)| = |C 目标 -C 实际 |<0.05×C 目标 If the conditions are met for more than 3 seconds (i.e. 30 control cycles), the system is determined to have entered a steady state. 2. Check the fractional integral term I of the FOPID controller. f (n) Whether it exceeds the preset threshold (i.e., duty cycle 80%). When the integral term is too large but the system has entered a steady state, it indicates that there is a risk of integral saturation. 3. When the system has an excessively large integral term, the integral is reset. The formula is: I f (n)=min(I f (n),I max ·0.5), where I max The upper limit of the integration threshold is set to 0.8, i.e., 80%, and finally the calibrated integration term is obtained.

[0047] For specific implementation details, please refer to the appendix. Figure 5 The negative ion concentration feedback module includes a personalized modeling submodule, a response curve simulation submodule, and a physiotherapy interaction submodule. The personalized modeling submodule acquires the user's height, weight, treatment area, preference pattern, and sensitive body constitution data. It then uses the personalized physiotherapy constraint model to perform constraint modeling and correction factor calculation to obtain personalized correction parameters and physiotherapy area codes. The response curve simulation submodule uses the system transfer function model in the FOPID controller to perform numerical simulation and prediction of the dynamic characteristic parameters after the physiotherapy site coding is corrected and the current measured concentration, so as to obtain the remaining time to reach the optimal concentration and the maximum overshoot. The physiotherapy interaction submodule utilizes dynamic simulation display technology and real-time feedback control technology to perform visualization rendering and threshold determination processing on the remaining time to reach the optimal concentration and the maximum overshoot data, thereby obtaining a real-time interactive interface and dynamic mode adjustment instructions.

[0048] In practice, the negative ion concentration feedback module also includes a collaborative optimization submodule and a model evolution submodule; The collaborative optimization submodule uses a collaborative control parameter generation method based on preference pattern mapping to process the dynamic pattern adjustment instruction data, transforming control target correction parameters and embedding constraints to obtain a collaborative control parameter set. The model evolution submodule uses the logistic regression algorithm to incrementally learn from user behavior time series data, and updates the model weights through gradient descent to obtain the updated user preference model parameters and the personalized physiotherapy plan for the next time step.

[0049] Its correction factor calculation and physiotherapy site coding include: Step 1) Use the internationally accepted DuBois formula for calculating body surface area, which is: S 体表 =0.007184×H 0.725 ×W 0.425 S 体表 The user's body surface area is 0.007184; the DuBois constant is 0.007184; the user's height is 0.007184; the user's weight is 0.007184; the height power exponent is 0.007184; the weight power exponent is 0.007184; the user's body surface area is calculated using the formula. Step 2) Based on Step 1), set the reference body surface area S. 基准 =1.7m 2 (For a standard body type of 170cm height and 70kg weight), calculate the body type correction factor using the following formula: , where k 修正 0.8 is the body size correction factor (adjusting the target concentration by a factor; the larger the body size, the larger the factor); 0.8 is the lower limit baseline value; 0.4 is the amplification factor (used to control the sensitivity of changes in body surface area to the correction factor); S 体表 S represents the user's body surface area. 基准 The reference surface area is set.

[0050] Step 3) Convert the user-selected area into a unique thermal encoding vector, and set the transmission loss coefficient k according to the treatment area. loss Includes: 1- Facial therapy (release window 10-15cm from the face, minimizing damage, k loss =1.0); 2-Top scalp therapy (release window 15-20cm from the top of the head, less damage, k loss =0.95); 3-Back physiotherapy (release window 30-50cm from the back, moderate wear and tear, k loss =0.7); 4-Leg physiotherapy (release window 50-80cm from the leg, maximum wear and tear, k loss =0.5); Step 4) Based on Step 3), when the user selects multiple parts, the system takes the weighted average as the comprehensive loss coefficient, and the weight is determined by the priority of the parts; Step 5) Based on Step 4), the user-selected mode is converted into two control factors, including: 1- Response speed preference factor (affects the response speed of the FOPID controller, set as high efficiency mode = 1.2, comfort mode = 0.8 and energy saving mode = 0.6); 2- Target concentration correction factor (affects the setting of the target concentration, set as high efficiency mode = 1.0, comfort mode = 0.9 and energy saving mode = 0.8). Step 6), Based on steps 1) to 5), the comprehensive target concentration correction factor is: k 综合 =k 修正 ×Target concentration correction factor, which integrates the target concentration correction factor as the final personalized correction parameter, and the process obtains the treatment site code.

[0051] The dynamic characteristic parameters after the above-mentioned correction of the physiotherapy site coding are compared with the current measured concentration using numerical simulation prediction, including: Step 1) Request the current system transfer function model from the dynamic operating condition prediction module. The returned transfer function is in the form of a standard second-order system with pure delay: Where G(s) is the transfer function (used to predict the response of negative ion concentration after a change in duty cycle); s is the complex frequency variable; K is the steady-state gain; T is the time constant; ξ is the damping ratio; and τ is the pure delay time. This is a pure delay factor; the current target concentration C is obtained from the dynamic operating condition prediction module. 目标 and C 实际 As initial conditions for simulation; Step 2) Based on Step 1), the steady-state gain K is corrected, and the correction formula is: K 修 =K×k loss K 修 K is the corrected steady-state gain; K is the original steady-state gain; k loss Here is the transmission loss coefficient; the corrected transfer function model is: Where G1(s) is the corrected transfer function; K 修 This is the corrected steady-state gain; Step 3) Based on step 2), G1(s) is converted into a state-space equation. The fourth-order Runge-Kutta method is used for numerical integration to solve the time-domain response of the system. The simulation time window is set to 120 seconds and the simulation step size is 0.1 seconds (consistent with the control cycle to ensure simulation accuracy). At each simulation step, the concentration value at the current moment is calculated and stored in an array to form a complete theoretical response curve. Step 4) Based on the time point in Step 3) when the concentration first enters and remains within ±5% of the target concentration, the minimum time to meet the condition is: |current concentration value - C 目标 |≤0.005×C 目标 Furthermore, the concentration remains within this range at all subsequent time points from this moment onwards. Find the first time t in the theoretical response curve array that satisfies the above conditions. set ; Step 5) Calculate the remaining time based on Step 4), using the formula: t 剩余 =max(0,t set -t 计时 ), where t 剩余 The remaining time to reach the expected optimal concentration; t 计时 Time the system modules; Step 6) Based on the maximum deviation of the theoretical response curve from the target concentration defined in Step 5), find the maximum concentration value from the theoretical response curve array and denote it as C. max Its absolute overshoot is: OS abs =max(0,C max -C 目标 ), where OS abs This is the absolute overshoot; the relative overshoot is calculated using the following formula: OS rel This is the relative overshoot; it is taken as the maximum overshoot (percentage).

[0052] Its incremental learning, which updates the model weights through gradient descent, includes: Step 1) Using user behavior time series data (20-50 times) as training samples for logistic regression, calculate the satisfaction probability for each sample using the following formula: ,in Let be the predicted satisfaction probability; e is the natural constant (≈2.718). w is the weight coefficient for each feature; x is the feature vector; w T x is a linear combination of weights and features; b is a bias term; Step 2), based on Step 1), the formula for calculating the cross-entropy loss is: Where L is the loss value (measuring the difference between the model's prediction and the true label); K is the number of samples (20-50); i is the sample index; y i The labels are real (1 for satisfied, 0 for dissatisfied); is the predicted satisfactory probability (obtained in step 1); log is the natural logarithm (used to amplify the loss penalty when the prediction is wrong); For unsatisfactory samples, this is the loss term. The loss term for the satisfactory sample; Step 3), based on step 2), calculate the gradient formula set as follows: ,in The gradient of the loss with respect to the weights (used to indicate the direction and magnitude by which each weight should be adjusted); The gradient of the loss with respect to the bias (used to indicate the direction and magnitude of the bias adjustment). For prediction error; x i Let be the feature vector of the i-th sample; This means averaging the gradients over K samples to stabilize the update direction; Step 4) Based on Step 3), update the parameter formula group as follows: , where w is the weight before the update; w← is the assignment operation (meaning the right-hand weight calculation result is assigned to the left-hand variable to complete the parameter update); η is the learning rate (taken as 0.01); b is the original bias term; b← is the assignment operation (meaning the right-hand bias term calculation result is assigned to the left-hand variable to complete the parameter update).

[0053] Working Principle: The system uses an environmental sensing module to deploy sensors to collect multi-source heterogeneous data in real time. After cleaning and alignment, the data is input into a dynamic operating condition prediction module. This module utilizes a FOPID-RBFNN fusion model to feed forward and output optimal control parameters based on the current operating conditions and user-specific parameters. The intelligent control module then precisely adjusts the duty cycle of the high-voltage generator to drive the tungsten gold needle corona discharge, stably outputting high-concentration, small-particle-size negative ions. Execution feedback data is sent back to the negative ion concentration feedback module for online model fine-tuning and integral term calibration. The system displays the estimated arrival time and concentration trend in real time, forming a closed-loop intelligent control of "sensing-decision-execution-optimization" to achieve adaptive and personalized microcirculation purification therapy.

[0054] Those skilled in the art will recognize that the algorithmic steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

Claims

1. A smart control system for a micro-circulation air purifier based on negative ion concentration feedback, characterized in that the system... include: The system includes an environmental sensing module, a dynamic operating condition prediction module, an intelligent control module, and a negative ion concentration feedback module. The environmental sensing module, through its built-in temperature, humidity, dust particle, and ion concentration sensors, collects multi-dimensional environmental and equipment status data in real time, and generates a structured raw feature dataset after preprocessing. The dynamic operating condition prediction module, based on the original feature dataset, uses fractional PID and radial basis function neural network to construct FOPID-RBFNN fusion prediction model, performs dynamic operating condition prediction and parameter tuning, and obtains negative ion optimized control parameter set; The intelligent control module obtains control commands from the negative ion optimization control parameter group, and uses a high-voltage drive and mode self-adaptation method to adjust the high-voltage transformer switching frequency, pulse width and operating mode to achieve precise control of the tungsten gold needle corona discharge state, so that the negative ion generator can output stably. The negative ion concentration feedback module acquires user-specific parameters and real-time feedback data. Using a dynamic working condition prediction module combined with response curve simulation, it optimizes the effective ion concentration distribution and estimates the arrival time at the target site, and outputs a personalized physiotherapy plan.

2. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 1, characterized in that, The environmental perception module includes a heterogeneous data acquisition submodule, a signal preprocessing submodule, and a data fusion submodule; The heterogeneous data acquisition submodule acquires multi-source heterogeneous data by deploying multi-source heterogeneous data sensors to form a raw data stream. The signal preprocessing submodule acquires the raw data stream, uses physical range verification to remove outliers and performs linear interpolation to fill them, and combines sliding window mean and median filtering to suppress noise and obtain the cleaned effective data stream. The data fusion submodule acquires the cleaned effective data stream, performs linear interpolation resampling based on a unified time base, and aligns multi-source data of different frequencies into synchronized multi-dimensional data frames.

3. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 2, characterized in that, The environmental perception module also includes a feature engineering submodule; The feature engineering submodule acquires multidimensional data frames, calculates derived features, and standardizes the features using the Min-Max normalization method to obtain a structured original feature dataset.

4. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 3, characterized in that, The dynamic operating condition prediction module includes a feature parsing submodule, a target concentration deviation calculation submodule, and an RBFNN feedforward parameter prediction submodule. The feature parsing submodule uses data caching technology to cache, extract features, and verify the integrity of the structured original feature dataset to obtain the feature vector at the current time. The target concentration deviation calculation submodule uses an environmental dynamic correction algorithm to dynamically set the target concentration based on the environmental parameters in the feature vector and the user-defined physiotherapy mode data, and calculates the control deviation signal by combining the measured ion concentration. The RBFNN feedforward parameter prediction submodule uses a radial basis function neural network (RBFNN) to perform forward calculations on the current environmental feature vector and obtains a set of PID control coefficients through nonlinear mapping.

5. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 4, characterized in that, The dynamic operating condition prediction module also includes a controller calculation submodule and a parameter encapsulation submodule; The controller calculation submodule uses the Oustaloup filtering algorithm to perform fractional-order calculus operations and synthesis on the PID control coefficients and control deviation signals to obtain the target duty cycle control quantity. The parameter encapsulation submodule uses a parameter group encapsulation method to structure and package the target duty cycle control quantity, operating mode flag, and PID control coefficient, and transmits them through internal inter-process communication to generate a negative ion optimized control parameter group.

6. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 5, characterized in that, The intelligent control module includes a parameter parsing submodule, a mode self-adaptation submodule, and a discharge control submodule. The parameter parsing submodule uses parameter parsing and integrity verification technology to perform data verification, field extraction and PWM instruction conversion processing on the negative ion optimized control parameter group to obtain the PWM control instruction package; The mode self-adaptive submodule performs dynamic amplitude limiting and first-order low-pass filtering smoothing on the target duty cycle control quantity in the PWM control instruction packet according to the running mode flag bit, so as to obtain the smoothed execution duty cycle. The discharge control submodule uses PWM pulse width modulation technology to generate signals and perform high voltage transformation on the smoothed execution duty cycle, driving the corona discharge of the tungsten gold needle tip to output a negative ion flow.

7. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 6, characterized in that, The intelligent control module also includes an execution feedback submodule and an exception handling submodule; The execution feedback submodule uses execution feedback acquisition technology to collect concentration and duty cycle data of the actual output negative ion flow, and sends the feedback data packet back to the dynamic operating condition prediction module for online fine-tuning of the RBFNN model and calibration of the FOPID integral term. The anomaly handling submodule utilizes a multi-threshold hierarchical protection mechanism to perform real-time safety monitoring and hierarchical response to the high-voltage transformer current, input voltage, and equipment temperature, ensuring safe shutdown and limited operation of the equipment under abnormal conditions.

8. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 7, characterized in that, The negative ion concentration feedback module includes a personalized modeling submodule, a response curve simulation submodule, and a physiotherapy interaction submodule. The personalized modeling submodule acquires the user's height, weight, treatment area, preference pattern, and sensitive body constitution data. It then uses the personalized physiotherapy constraint model to perform constraint modeling and correction factor calculation to obtain personalized correction parameters and physiotherapy area codes. The response curve simulation submodule uses the system transfer function model in the FOPID controller to perform numerical simulation and prediction of the dynamic characteristic parameters after the physiotherapy site coding is corrected and the current measured concentration, so as to obtain the remaining time to reach the optimal concentration and the maximum overshoot. The physiotherapy interaction submodule utilizes dynamic simulation display technology and real-time feedback control technology to perform visualization rendering and threshold determination processing on the remaining time to reach the optimal concentration and the maximum overshoot data, thereby obtaining a real-time interactive interface and dynamic mode adjustment instructions.

9. The intelligent control system for a micro-circulation purifier based on negative ion concentration feedback according to claim 8, characterized in that, The negative ion concentration feedback module also includes a collaborative optimization submodule and a model evolution submodule; The collaborative optimization submodule uses a collaborative control parameter generation method based on preference pattern mapping to process the dynamic pattern adjustment instruction data, transforming the control target correction parameters and embedding constraints to obtain a collaborative control parameter set. The model evolution submodule uses the logistic regression algorithm to incrementally learn from user behavior time series data, and updates the model weights through gradient descent to obtain the updated user preference model parameters and the personalized physiotherapy plan for the next time step.

Images