Intelligent water body oxygenation pump

Abstract

The application discloses an intelligent water body oxygenation pump and belongs to the technical field of aquaculture. The application solves the problem that the existing method relies on a high-cost real dissolved oxygen sensor, oxygenation control is extensive, the ability of graded adjustment of oxygen release according to the degree of oxygen deficiency is lacked, and fine oxygenation management is difficult to realize. Through the combination of multi-source water body state parameter acquisition and joint filtering, pre-feature coupling extraction and evidence fusion, three-dimensional sparse tensor membership mapping, gated attention dynamic confidence correction, bivariate risk association rule reasoning, change rate prediction and inertia compensation PID parameter online optimization, the application realizes the precise evaluation of the oxygen deficiency risk in the home fish tank scene and the fine graded oxygen release control of the air pump and air valve linkage. Meanwhile, through the lightweight collaborative architecture of the fuzzy logic risk assessment model and the improved particle swarm PID parameter optimization, the overall technical effect of low detection cost, fine control granularity, simple system structure and strong adaptability to home scenes is achieved.

Description

Technical Field

[0001] This invention belongs to the field of aquaculture technology, specifically relating to intelligent water aeration pumps. Background Technology

[0002] Dissolved oxygen in water is essential for the survival of aquatic organisms such as fish. In home aquariums and small ornamental water bodies, the small water volume and relatively high organism density lead to rapid consumption of dissolved oxygen, easily resulting in oxygen deficiency, which can negatively impact the health of aquatic organisms and even cause death. Therefore, timely and effective replenishment of oxygen in the water is a crucial aspect of home fishkeeping.

[0003] Traditional aeration methods primarily rely on air pumps to force air into the water, where oxygen dissolves through the contact between the air bubbles and the water. However, traditional air pumps are typically manually controlled, requiring users to rely on experience or visual observation to determine if the water is oxygen-deficient and manually start or stop the equipment. This manual control method has significant drawbacks: firstly, users cannot monitor the dissolved oxygen levels in real time, often only taking action after fish exhibit obvious signs of oxygen deficiency such as surfacing, resulting in a severely delayed response; secondly, manual operation is time-consuming and labor-intensive, and the aeration intensity cannot be adjusted according to the degree of oxygen deficiency, frequently leading to over-aeration or under-aeration.

[0004] To address these issues, automatic oxygenation technology has emerged. In existing technologies, one approach uses a dissolved oxygen sensor to monitor the dissolved oxygen concentration in real time. When the concentration falls below a set threshold, a microcontroller controls a relay to activate an oxygenation pump, achieving automatic oxygenation. The core of this approach lies in its closed-loop control of detection, judgment, and execution, which can replace manual judgment to some extent. However, this approach has two shortcomings: First, its core detection method relies on a physical dissolved oxygen sensor. High-precision dissolved oxygen sensors are expensive and require regular calibration and maintenance. The sensor probes have limited lifespan, making them unsuitable as a basic component of ordinary home aquariums, thus hindering cost control and widespread adoption. Second, the control method of this approach primarily involves starting and stopping the air pump, simply switching between on and off states. It cannot adjust the oxygen release based on the severity of oxygen deficiency, resulting in the same oxygenation intensity being applied to both mild and severe oxygen deficiency, wasting energy and making precise oxygenation management difficult.

[0005] With the development of IoT technology, another type of solution combines oxygenation with remote control. For example, some solutions propose water aerators controlled remotely via a mobile app. Through an IoT communication module, cloud server, and mobile app, users can remotely view the water status and control the aerator's operating time or mode. While this type of solution improves user convenience, it essentially still adjusts the overall operating parameters of the aerator, resulting in coarse-grained control and a lack of fine-grained linkage control of actuators such as air pumps and valves. It also fails to achieve graded adjustment of oxygen release based on the real-time oxygen deficiency level of the water.

[0006] In addition, some solutions integrate oxygenation functions into comprehensive intelligent aquarium control systems. These systems connect to various detection devices such as temperature sensors and dissolved oxygen sensors via an STM32 microcontroller, and simultaneously control multiple actuators including an air pump, water circulation filter, feeder, heater, and lighting. Remote monitoring and operation are achieved through a wireless communication module and a mobile app. For example, Chinese patent CN213939357U discloses an intelligent aquarium control system. This system includes an aquarium body connected to the STM32 microcontroller's AD port via temperature and dissolved oxygen sensors for feeding fish and regulating the aquarium's water environment. Actuators such as the water circulation filter, feeder, heater, lighting cover, and air pump are connected to the STM32 microcontroller's GPIO for feeding fish and regulating the aquarium's water environment. A touchscreen is connected to the STM32 microcontroller's USART interface via a serial bus for locally displaying the aquarium's operating status and setting operating parameters. The aquarium body can exchange data with a server via a wireless module, and aquarium water environment data, feeding parameters, and remote feeding can be obtained via a mobile app or PC. However, the system has a complex structure, high hardware costs, and cumbersome installation and debugging. For users who only need intelligent oxygenation, it suffers from functional redundancy and low cost-effectiveness. Furthermore, its control logic focuses on the collaborative management and status display of multiple devices, rather than the fine-grained and graded adjustment of the oxygenation process itself.

[0007] Chinese patent CN106647835A discloses a smart aerator and a three-dimensional dissolved oxygen monitoring method based on narrowband Internet of Things (IoT). Even during power outages, it can still monitor water quality and report relevant information to the backend. It also features multi-channel water quality monitoring and automatic aerator start / stop functions, providing aquaculture operators with an efficient and convenient experience. Through stratified water quality monitoring and different colored LED warnings for abnormalities, the monitoring is reliable and the results are intuitive. By establishing a linkage mechanism for the aerators arranged in the same aquaculture pond, data sharing and coordinated control are achieved, which is more in line with the actual manual control methods on site, making it scientific and convenient. By executing pre-set program rules, it supplements oxygen at specific time periods, transmitting first-hand practical results through… The scientific approach demonstrates the technology, making it easier for frontline technicians to accept. Furthermore, the priority of control commands via a remote backend management platform or accompanying mobile app allows for rapid control of the aerator in emergencies and data traceability of monitoring data, indicating broad application prospects in aquaculture. However, the monitoring architecture of this intelligent aerator relies on a multi-node, three-dimensional layout and narrowband IoT wide-area communication, resulting in complex hardware configurations and high deployment costs, making it unsuitable for the miniaturization and plug-and-play requirements of home aquariums. Moreover, the intelligent aerator's control logic primarily relies on timed aeration based on preset program rules and remote manual intervention, lacking a risk assessment model that integrates multiple parameters, and thus cannot automatically adjust the aeration intensity according to the real-time water conditions, resulting in coarse control granularity.

[0008] Chinese patent CN113126490A discloses an intelligent variable frequency oxygenation control method and device. The method includes: determining multiple key environmental data variables that affect the ideal dissolved oxygen content in the aquaculture environment based on the ideal dissolved oxygen content; inputting the multiple key environmental data variables and the ideal dissolved oxygen content into a trained prediction model to obtain the predicted dissolved oxygen content of the aquaculture environment at the current moment; inputting the error and error change rate between the predicted dissolved oxygen content and the actual dissolved oxygen content collected at the current moment into a trained fuzzy PID control algorithm model to obtain the oxygenation regulation amount, and regulating the dissolved oxygen content in the aquaculture environment according to the oxygenation regulation amount; however, the oxygenation regulation amount of the existing device is a single output quantity, which is only used for variable frequency speed regulation and does not involve dual-output linkage control of air pump start / stop and air valve opening, so it cannot achieve graded and fine adjustment of oxygen release.

[0009] In summary, existing technologies for intelligent aeration in home aquariums and small water bodies suffer from the following key shortcomings: First, the detection method is singular, relying primarily on high-cost real dissolved oxygen sensors, which hinders the widespread adoption of low-cost products. Second, aeration control is rudimentary, mostly involving pump start / stop control or operating time adjustments, lacking the ability to adjust oxygen release in stages based on the degree of oxygen deficiency, making it difficult to achieve refined aeration management. Third, the system structure is complex or functionally redundant, either integrating too many unnecessary functions leading to increased costs, or adopting complex technologies designed for large-scale aquariums, unsuitable for the miniaturization, low cost, and ease of installation requirements of home aquariums. To address these issues, we propose an intelligent water aerator pump. Summary of the Invention

[0010] The purpose of this invention is to address the shortcomings of existing technologies by providing an intelligent water oxygenation pump. This solves the problems of existing methods relying on high-cost real dissolved oxygen sensors, having a crude oxygenation control, lacking the ability to adjust oxygen release in stages according to the degree of oxygen deficiency, and being unable to achieve refined oxygenation management.

[0011] This invention is implemented as follows: an intelligent water aeration pump, the intelligent water aeration pump comprising:

[0012] The detection module is used to collect multi-source water state parameters in the aquarium, preprocess the multi-source water state parameters, and output the preprocessed multi-source parameter set. The multi-source water state parameters include water temperature data, cumulative inactivity time of the air pump, water circulation status data, and simulated oxygen content data.

[0013] The main control module communicates with the detection module and is used to receive the pre-processed multi-source parameter set, identify and analyze the multi-source parameter set based on the pre-built fuzzy logic risk assessment model, output the state assessment result, use the state assessment result as an index to retrieve the preset multi-source inference rule table, output the water body hypoxia risk level, generate an adaptive dual-output control strategy based on the water body hypoxia risk level, and generate an adaptive dual-output control command based on the adaptive dual-output control strategy.

[0014] The air pump drive module responds to the adaptive dual-output control command, synchronously outputs the air pump control signal, and executes the air pump start-up and stop actions;

[0015] The air volume regulation module responds to the adaptive dual-output control command, synchronously outputs the air valve opening control signal, and executes the air valve opening regulation action;

[0016] The wireless communication module is connected to the main control module via a serial port and is used to upload the status data of the main control module to the user terminal and receive control commands from the user terminal.

[0017] The user terminal is used to display the water condition assessment results, the water hypoxia risk level, the air pump status and air valve opening, and to receive control commands and user-set thresholds.

[0018] Preferably, the detection module includes:

[0019] The water temperature detection unit includes a thermistor temperature sensor, which is placed in the aquarium water and collects aquarium water temperature data based on a preset sampling period. The water temperature data is used to characterize the changing trend of dissolved oxygen saturation concentration in the water. When the water temperature rises, the risk of water hypoxia increases.

[0020] The timing unit, integrated inside the main control module, is used to accumulate the timing duration when the air pump is off, and to reset it when the air pump is on. It outputs the cumulative non-running time of the air pump, which is used to characterize the degree of continuous consumption of dissolved oxygen in the water body since the last oxygenation.

[0021] A water circulation status detection unit includes at least one of a water flow velocity sensor or a water level fluctuation detection sensor, used to collect water flow and exchange capacity data and output water circulation status data. When the water circulation status is good, the oxygen exchange efficiency at the water-air interface is high, and when the water circulation status is weak, the risk of hypoxia increases.

[0022] The simulated oxygen content trend input unit includes a general analog input interface, which is configured as an ADC input channel. A potentiometer is connected to generate an analog voltage signal corresponding to the preset virtual dissolved oxygen state. This signal is used to verify the hypoxia risk assessment algorithm and the control logic of the air pump and air valve, and to output simulated oxygen content data.

[0023] The preprocessing unit is electrically connected to the water temperature detection unit, timing unit, water circulation status detection unit, and simulated oxygen trend input unit. It is used to acquire water temperature data, cumulative inactivity time of the air pump, water circulation status data, and simulated oxygen status data in real time and integrate them into multi-source water status parameters. It preprocesses the multi-source water status parameters and outputs the preprocessed multi-source parameter set.

[0024] Preferably, the main control module includes:

[0025] The model building unit is used to pre-build a fuzzy logic risk assessment model. It uses a reference-level dissolved oxygen detection device to collect the actual dissolved oxygen concentration corresponding to the time of the multi-source water body state parameters, and generates hypoxia risk labels based on the actual dissolved oxygen concentration. The user-defined threshold is used as the control preference parameter, and together with the multi-source water body state parameters, it is used as the model input. The hypoxia risk label is used as the supervision target to optimize the fuzzy membership parameters and rule weights, and outputs a converged fuzzy logic risk assessment model.

[0026] The fuzzy logic risk assessment unit receives the preprocessed multi-source parameter set and user-defined thresholds. Based on the pre-built fuzzy logic risk assessment model, it performs membership degree division on the parameters in the multi-source parameter set and outputs the state assessment results. When dividing the membership degree, the water temperature data is mapped to three membership degree levels: normal, high, and high temperature; the cumulative non-running time of the air pump is mapped to three membership degree levels: short, medium, and long; the water circulation state data is mapped to three membership degree levels: good, average, and weak; and the simulated oxygen content state data is mapped to three membership degree levels: normal, critical, and low.

[0027] The dynamic confidence correction unit is used to traverse the multi-source parameter set, dynamically adjust the membership confidence of the state evaluation result based on the parameter coupling coefficient of the multi-source parameter set, and determine the dynamic weight of the parameter by combining the membership confidence.

[0028] The rule reasoning unit communicates with the fuzzy logic risk assessment unit, and combines the dynamic weights of parameters, the state assessment results and the multi-source reasoning rule table matching analysis to output the fuzzy risk level.

[0029] The fuzzy risk level is defuzzified and the area center method is used to perform integral calculation on the fuzzy risk level to output the water body hypoxia risk level.

[0030] A dual-output control strategy generation unit is communicatively connected to the defuzzification unit. Using the water body hypoxia risk level as an index, it generates an adaptive dual-output control strategy based on rate of change prediction and inertial compensation mechanism. The adaptive dual-output control strategy includes a pump control strategy and a valve opening control strategy corresponding to the water body hypoxia risk level. Specifically, the low-risk level corresponds to a pump shutdown strategy and a valve small opening strategy, the medium-risk level corresponds to a pump start strategy and a valve medium opening strategy, and the high-risk level corresponds to a pump continuous operation strategy and a valve large opening strategy.

[0031] The control command output unit is communicatively connected to the dual-output control strategy generation unit and is used to convert the adaptive dual-output control strategy into adaptive dual-output control commands, which are then output to the air pump drive module and the air volume regulation module, respectively.

[0032] Preferably, the user terminal includes:

[0033] The status display unit is used to display the current water body oxygen deficiency risk level, air pump operating status, air valve opening degree, real-time water temperature, cumulative air pump inactivity time, and abnormal reminder information;

[0034] The mode switching unit is used to switch between automatic mode and manual mode. In automatic mode, it operates autonomously based on an adaptive dual-output control strategy. In manual mode, user-input control commands are executed first.

[0035] The manual control unit is used to receive commands from the user to manually turn the air pump on or off and adjust the opening of the air valve.

[0036] The threshold parameter setting unit is used to receive user input of oxygenation preference settings and user-set thresholds. The oxygenation preference settings include energy-saving mode, normal mode or strong oxygenation mode, and the user-set thresholds include water temperature alarm threshold, air pump maximum continuous running time threshold, and automatic recovery time threshold.

[0037] The safety priority determination unit is used to monitor abnormal operating conditions in manual mode. When an abnormal operating condition is detected, a safety protection command is generated, and the safety protection command has a higher priority than the manual control command.

[0038] Preferably, the method for preprocessing the state parameters of multi-source water bodies includes:

[0039] Acquire multi-source water body state parameters, and mark the collection timestamps for water temperature data, cumulative inactivity time of air pump, water circulation state data and simulated oxygen state data respectively. Use a preset reference period as the time window to align the parameters on the time axis, and group the parameter data collected within the same reference period into the same parameter frame. Parameter data that were not collected within the reference period are filled into the corresponding parameter frame by nearest neighbor interpolation. Output the time-domain aligned multi-source water body state parameters.

[0040] Load the time-domain aligned multi-source water body state parameters, perform sliding window anomaly detection on the time-domain synchronized multi-source water body state parameters, calculate the mean and standard deviation of the parameter data in the current window, determine the data points that deviate from the mean by more than three times the standard deviation as outliers, replace the outliers with the median of the window where the outlier is located, and output the multi-source water body state parameters after outlier removal.

[0041] Obtain the state parameters of the multi-source water bodies after outlier removal, perform joint filtering on the state parameters of the multi-source water bodies, and output the filtered state parameters of the multi-source water bodies.

[0042] The filtered multi-source water state parameters are mapped to a unified numerical range, which is a closed interval [0,1]. The mapped parameter data constitute the preprocessed multi-source parameter set.

[0043] Preferably, the method for joint filtering of multi-source water state parameters includes:

[0044] After removing outliers, obtain the multi-source water body state parameters, determine the parameter coupling coefficient based on the weights and mutual information of the multi-source water body state parameters, and establish a parameter correlation matrix in combination with the parameter coupling coefficient. The parameter correlation matrix represents the first coupling coefficient between water temperature data and simulated oxygen-containing state data, the second coupling coefficient between the cumulative non-running time of the air pump and simulated oxygen-containing state data, and the third coupling coefficient between water body circulation state data and simulated oxygen-containing state data.

[0045] Among them, when determining the parameter coupling coefficient based on the weights and mutual information of multi-source water state parameters, the formula for calculating the parameter coupling coefficient is expressed as follows:

[0046]

[0047]

[0048] in, This represents the mutual information between the state parameter i of the multi-source water body and the associated state parameter j. This represents the joint probability density distribution of the multi-source water body state parameter i and the associated state parameter j. These are the marginal probability density distributions of multi-source water body state parameter i and associated state parameter j, respectively. These are the parameter coupling coefficient and the weight of the multi-source water state parameter i, respectively.

[0049] Using the parameter correlation matrix as a constraint, the state parameters of multiple water bodies are collaboratively corrected, and the coupling deviation between the state parameters of multiple water bodies and the associated state parameters is calculated.

[0050] Determine whether the coupling deviation between the state parameters of multiple water sources and the associated state parameters exceeds a preset deviation threshold;

[0051] If the coupling deviation between the multi-source water body state parameters and the associated state parameters exceeds the preset deviation threshold, the multi-source water body state parameters are constrained and corrected by the weighted combination of the associated state parameters, and the multi-source water body state parameters after parameter correlation constraint filtering are output.

[0052] The state parameters of multi-source water bodies after parameter correlation constraint filtering are obtained. The dynamic window length of the sliding window is determined based on the parameter dynamic response. Multi-scale wavelet packet decomposition is performed on the state parameters of multi-source water bodies in combination with the dynamic window length, separating the state parameters of multi-source water bodies into low-frequency approximate components and high-frequency detail components. The trend term of the low-frequency components is extracted by long-window moving average filtering, and the high-frequency components are retained after removing impulse noise by improved median filtering. The number of decomposition layers of multi-scale wavelet packet decomposition is adaptively determined according to the dynamic window length, with a long window corresponding to 3 layers of decomposition and a short window corresponding to 1 layer of decomposition. The filtered low-frequency and high-frequency components are reconstructed into time-domain signals, and the multi-scale decomposed and filtered state parameters of multi-source water bodies are output.

[0053] Based on the multi-source water body state parameters after multi-scale decomposition filtering, a three-dimensional state vector containing dissolved oxygen concentration, water temperature, and cumulative inactivity time of air pumps is constructed. With water temperature as the known input, dissolved oxygen concentration as the state to be estimated, and cumulative inactivity time of air pumps as the control variable, a nonlinear state space model is established. Combining the nonlinear state space model, a volumetric Kalman filter is performed to output the multi-source water body state parameters after time-consistent filtering.

[0054] Preferably, the fuzzy logic risk assessment model includes an input layer, a risk assessment architecture, and an output layer. The risk assessment architecture includes a pre-feature extraction layer and a risk association layer. The input layer is used to obtain a multi-source parameter set, encode and reduce the dimensionality of the multi-source parameter set, and output the encoded four-dimensional parameter vector. The pre-feature extraction layer is used to perform feature coupling extraction and evidence correction operations on the four-dimensional parameter vector and output the corrected comprehensive feature vector.

[0055] Preferably, the method for classifying the membership of parameters in a multi-source parameter set based on a pre-constructed fuzzy logic risk assessment model includes:

[0056] Obtain a multi-source parameter set, encode and reduce the dimensionality of the multi-source parameter set, and output the encoded four-dimensional parameter vector;

[0057] The water temperature data and the cumulative inactivity time of the air pump are nonlinearly coupled and calculated. Based on the temperature correction coefficient of dissolved oxygen saturation concentration and the cumulative effect coefficient of the cumulative inactivity time of the air pump, the environmental stress index is calculated.

[0058] The formula for calculating the environmental stress index is as follows:

[0059]

[0060] In the formula, , representing the baseline oxygen consumption rate coefficient, water temperature data, and standard water temperature data, respectively. b and d represent the temperature sensitivity coefficient and the cumulative effect decay index (used to characterize the natural decay effect of oxygen consumption rate after a decrease in dissolved oxygen concentration), respectively. The temperature sensitivity coefficient is the relative change rate of the environmental stress index caused by a unit change in water temperature, and its value range is calibrated based on the exponential relationship between fish metabolic rate and water temperature. The cumulative effect decay index d is calibrated through oxygen consumption experiments, is dimensionless, and ranges from 0.5 to 1.5. Based on oxygen consumption experiments in closed water bodies, during the process of dissolved oxygen concentration decreasing from saturation to criticality, the oxygen consumption rate decays exponentially with time. By fitting the correspondence between the cumulative inactivity time of the air pump and the environmental stress index in the historical multi-source parameter set using the least squares method, the calibration value of d is obtained as 0.85. This value characterizes that the effect of pump shutdown time on the environmental stress index exhibits sublinear growth (0 < d < 0). 1) In the initial stage of pump shutdown, dissolved oxygen is consumed rapidly and the environmental stress index increases rapidly. As the pump shutdown time is extended and the dissolved oxygen concentration decreases, the rate of biological oxygen consumption naturally slows down due to hypoxia stress, and the growth of the environmental stress index slows down, which is consistent with the physical self-limiting mechanism of oxygen consumption in water bodies.

[0061] The water circulation state data and the simulated oxygen state data are dynamically coupled and calculated. The air-water interface update rate is calculated based on the water circulation state data, and the water-air exchange efficiency index is calculated by combining the change slope of the simulated oxygen state data.

[0062] The formula for calculating the water-air exchange efficiency index is as follows:

[0063]

[0064] In the formula, Indicates the water-air exchange efficiency index. This represents the gas-water interface renewal rate and the oxygen content trend compensation coefficient. To simulate the slope of changes in oxygen-containing state data;

[0065] The environmental stress index and water-air exchange efficiency index are obtained and used to resolve evidence conflicts and perform weighted fusion on the environmental stress index and water-air exchange efficiency index, and output the corrected comprehensive feature vector and water-air exchange correction vector.

[0066] The integrated feature vector and the water-air exchange correction vector are expressed as follows:

[0067]

[0068]

[0069] in, These represent the integrated eigenvector and the water-air exchange correction vector, respectively. These represent the basic probability allocation values ​​of the environmental stress index and the water-air exchange efficiency index, respectively. The degree of conflict of evidence between the environmental stress index and the water-air exchange efficiency index;

[0070] Obtain the corrected integrated feature vector and water-air exchange correction vector, construct a three-dimensional sparse tensor lookup table based on the integrated feature vector and water-air exchange correction vector, traverse the preset parameter-level membership mapping relationship based on the three-dimensional sparse tensor lookup table, calculate the interpolated membership degree between the integrated feature vector and the parameter-level membership degree using the trilinear interpolation method, and output the state evaluation result composed of 12-dimensional membership degree values.

[0071] Obtain the state assessment results composed of 12-dimensional membership values ​​and parameter coupling coefficients, and construct the basic confidence of the state assessment results based on the state assessment results;

[0072] By dynamically adjusting the basic confidence of the state assessment results in conjunction with the parameter coupling coefficient, the membership confidence of the state assessment results is obtained. Using the membership confidence of the state assessment results as a priori, the membership confidence is directly linearly mapped to the dynamic weight of the parameters.

[0073] Obtain the comprehensive feature vector and the water-air exchange correction vector, and construct a joint membership distribution based on the comprehensive feature vector and the water-air exchange correction vector;

[0074] A bivariate risk association rule matrix is ​​established by combining the joint membership degree distribution and the dynamic weight of the parameters. The activation intensity of the bivariate risk association rule matrix in the multi-source inference rule table is calculated based on the fuzzy clustering algorithm. The activation intensity of the multi-source inference rule table is classified and superimposed according to the output risk level to obtain the fuzzy risk level.

[0075] The area center method is used to perform integral calculations on the fuzzy risk level and output the water body hypoxia risk level.

[0076] Preferably, the method for generating an adaptive dual-output control strategy based on rate of change prediction and inertia compensation mechanism includes:

[0077] Obtain the current period's water body hypoxia risk level and the historical period's water body hypoxia risk level sequence, and calculate the risk change rate using the backward difference method;

[0078] The predicted gain coefficient is determined based on the current cycle risk level, and the predicted risk level for the next cycle is predicted based on the rate of change of the risk level.

[0079] Obtain the mechanical response parameters of the servo motor, and construct the mechanical hysteresis compensation based on the mechanical response parameters of the servo motor; use the predicted risk level of the next cycle as the set value, the water hypoxia risk level of the current cycle as the feedback value, and the difference between the two as the control error;

[0080] A multi-objective fitness function is constructed based on the control error, mechanical hysteresis compensation, and the number of pump start-stop cycles. An improved particle swarm optimization algorithm is used to iteratively optimize the multi-objective fitness function. An adaptive inertial weight mechanism is introduced into the improved particle swarm optimization algorithm, and a differential evolution mutation operator is introduced to enhance particle diversity. When the number of iterations reaches 30 generations or the improvement of the population's optimal fitness value is less than 1% for 10 consecutive generations, the iteration is terminated and the optimal PID parameter set corresponding to the population's historical optimal position is output.

[0081] The multi-objective fitness function is expressed as:

[0082]

[0083]

[0084] in, These represent the control error item, the servo wear item, and the air pump start / stop item, respectively. These represent the initial weights of the control error term, the servo wear term, and the air pump start / stop term, respectively. Indicates the evaluation time window;

[0085] The optimal PID parameter set is used as the adaptive dual-output control strategy. The adaptive dual-output control strategy is converted into adaptive dual-output control commands, which are output to the air pump drive module and the air volume regulation module respectively.

[0086] Compared with the prior art, the embodiments of this application have the following main advantages:

[0087] This invention combines multi-source water state parameter acquisition and joint filtering, pre-feature coupling extraction and evidence fusion, three-dimensional sparse tensor membership mapping, gated attention dynamic confidence correction, bivariate risk association rule reasoning, and online optimization of PID parameters through rate of change prediction and inertia compensation. Without relying on high-cost real dissolved oxygen sensors, it achieves accurate assessment of hypoxia risk in home aquarium scenarios and refined, graded oxygen release control via air pump and valve linkage. Simultaneously, through a lightweight collaborative architecture combining a fuzzy logic risk assessment model and an improved particle swarm optimization PID parameter optimization, it overcomes the computational latency problem of existing complex algorithms on low-cost embedded devices, achieving overall technical effects of low detection cost, fine control granularity, simple system structure, and strong adaptability to home scenarios.

[0088] In this embodiment of the invention, the main control module comprises a model building unit, a fuzzy logic risk assessment unit, a dynamic confidence correction unit, a rule reasoning unit, a defuzzification unit, a dual-output control strategy generation unit, and a control command output unit, forming a progressive collaborative architecture. The model building unit provides a basic reasoning framework for fuzzy logic risk assessment. The fuzzy logic risk assessment unit collaborates with the dynamic confidence correction unit to achieve adaptive discrimination of multi-source signal quality and real-time optimization of parameter dynamic weights. The rule reasoning unit integrates parameter dynamic weights and state assessment results, taking into account parameter reliability and physical correlation to perform risk level reasoning. The defuzzification unit completes the mapping conversion from fuzzy to precise. The dual-output control strategy generation unit optimizes the control strategies of the air pump and air valve based on rate of change prediction and inertia compensation mechanisms. The control command output unit realizes the synchronous drive of the two actuators. The main control module's sub-units form a closed-loop adaptive collaborative control system by using dynamic weights of parameters to inversely constrain the accuracy of risk assessment and by using predictive risk levels to feed forward and compensate for execution delays. This enables the system to achieve accurate fusion of multi-source parameters, dynamic assessment of hypoxia risk, and millisecond-level linkage response of air pumps and valves with an entry-level microcontroller, without relying on high-cost dissolved oxygen sensors. This results in low detection costs, fine control granularity, and a simple system structure.

[0089] In this embodiment of the invention, when preprocessing multi-source water state parameters, temporal consistency of heterogeneous multi-source data is ensured through temporal domain alignment, instantaneous noise pollution is eliminated through sliding window anomaly detection and median replacement, cross-validation of physical correlations between parameters and multi-scale time-frequency denoising are achieved through joint filtering, and dimensional differences are eliminated and physical distribution characteristics are preserved through normalization mapping. This forms a progressive collaborative system of alignment-cleaning-fusion-standardization: temporal domain alignment provides a synchronous benchmark for anomaly detection, anomaly detection reduces noise interference for joint filtering, joint filtering uses parameter coupling relationships to reverse the interpolation errors that may be introduced by temporal domain alignment, and normalization mapping uniformly inputs the fused high-quality parameters into the fuzzy logic risk assessment model.

[0090] In this embodiment of the invention, when jointly filtering multi-source water state parameters, the physical correlation representation of multi-source parameters is constructed through parameter coupling coefficients and parameter correlation matrices. Cross-validation and physical consistency restoration of abnormal parameters are achieved through coupling deviation determination and collaborative constraint correction. Refined noise separation and trend extraction in the time-frequency domain are achieved through multi-scale wavelet packet decomposition and adaptive median filtering. Optimal state estimation at the dynamic level of the nonlinear system is achieved through capacitive Kalman filtering. This joint filtering mechanism enables the system to perform cross-validation and complementary enhancement using the physical coupling information between multi-source parameters without increasing sensor hardware costs, thereby suppressing local noise and occasional faults of a single sensor. Finally, a high-quality multi-source parameter set that is physically self-consistent, temporally stable, and complete with multi-scale features is output.

[0091] In this embodiment of the invention, the fuzzy logic risk assessment model achieves accurate risk level assessment of multi-source heterogeneous parameters from raw data collection through a progressive collaborative architecture consisting of an input layer, a pre-feature extraction layer, a risk association layer, and an output layer. The pre-feature extraction layer reduces the four-dimensional parameter vector into a comprehensive feature vector with clear physical meaning through dual-branch coupling extraction and evidence fusion of the environmental stress index and the water-air exchange efficiency index, thereby reducing the complexity of subsequent calculations while retaining key risk information. The risk association layer achieves dynamic reshaping of user preferences for membership boundaries through a three-dimensional sparse tensor lookup table, achieves dynamic weight allocation of parameters with adaptive signal quality through a gated attention weighting layer, and achieves collaborative reasoning of multi-source evidence through a bivariate risk association rule matrix. This enables the system to accurately assess hypoxia risk by comprehensively considering multi-source information such as water temperature, pump shutdown duration, circulation status, and oxygen content trends without relying on real dissolved oxygen sensors. The output layer completes the fuzzy-to-precise conversion using the area center method, providing a reliable decision-making basis for the linkage control of air pumps and valves.

[0092] In this embodiment of the invention, when generating an adaptive dual-output control strategy based on rate of change prediction and inertial compensation mechanisms, the backward difference method is used to extract the changing trend from the historical risk sequence. The interval-differential configuration of the predicted gain coefficient achieves a sensitive response and smooth transition to the risk evolution trend. The pre-estimation of mechanical hysteresis compensation offsets the servo motor execution delay. A multi-objective fitness function coordinates the diverse demands of control accuracy, smoothness, and equipment lifespan. An improved particle swarm optimization algorithm with adaptive inertial weights and differential evolution mutation achieves global optimal optimization of PID parameters. Finally, the optimal PID parameter set is synchronously configured in both the air pump and valve loops. This achieves precise linkage between air pump start / stop and valve opening, and smooth, graded adjustment of oxygen release, effectively avoiding the frequent start / stop and coarse adjustment problems of traditional threshold control. It also extends the mechanical lifespan of relays and servos, and improves the comfort and energy efficiency of oxygenation control in home aquarium scenarios. Attached Figure Description

[0093] Figure 1 This is a schematic diagram of the control architecture of the intelligent water oxygenation pump provided by the present invention.

[0094] Figure 2 This is a schematic diagram of the main control module circuit connection provided by the present invention.

[0095] Figure 3 This is a schematic diagram of the main control module provided by the present invention.

[0096] Figure 4 A schematic diagram of the process for preprocessing state parameters of multi-source water bodies is shown.

[0097] Figure 5 A schematic diagram of the fuzzy logic risk assessment model architecture is shown.

[0098] Figure 6 The diagram illustrates the implementation process of a method for classifying the membership of parameters in a multi-source parameter set based on a pre-built fuzzy logic risk assessment model.

[0099] Figure 7 A schematic diagram illustrating the implementation process of an adaptive dual-output control strategy based on rate of change prediction and inertia compensation mechanism is shown.

[0100] In the diagram: 100-Detection module, 110-Water temperature detection unit, 120-Timing unit, 130-Water circulation status detection unit, 140-Simulated oxygen content trend input unit, 150-Preprocessing unit, 200-Main control module, 210-Model building unit, 220-Fuzzy logic risk assessment unit, 230-Dynamic confidence correction unit, 240-Rule reasoning unit, 250-Defuzzification unit, 260-Dual output control strategy generation unit, 270-Control command output unit, 300-Air pump drive module, 400-Air volume adjustment module, 500-Wireless communication module, 600-User terminal, 610-Status display unit, 620-Mode switching unit, 630-Manual control unit, 640-Threshold parameter setting unit, 650-Safety priority determination unit. Detailed Implementation

[0101] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0102] This invention addresses the industry pain points of high detection costs, coarse control granularity, and complex system structures in the oxygenation control of home aquariums and small water bodies, and has conducted targeted research and development. During the research and development process, it was found that existing technologies (CN213939357U A smart control system for aquariums, CN106647835A A smart aerator and dissolved oxygen three-dimensional monitoring method based on narrowband Internet of Things, CN113126490A A smart variable frequency oxygenation control method and device) can only solve single problems and cannot simultaneously achieve low-cost multi-source detection, fine-grained graded oxygen release control, lightweight embedded computing, and adaptation to home scenarios.

[0103] Based on existing conventional designs, this invention innovatively proposes an overall technical solution for an intelligent water aerator pump based on fuzzy logic risk assessment and adaptive dual-output PID parameter optimization. By organically combining multi-source water state parameter acquisition and joint filtering, pre-feature coupling extraction and evidence fusion, three-dimensional sparse tensor membership mapping, gated attention dynamic confidence correction, bivariate risk association rule reasoning, and online optimization of PID parameters with rate of change prediction and inertia compensation, it realizes several core actions such as collaborative perception of multi-source heterogeneous parameters, intelligent assessment of hypoxia risk level, and hierarchical control of air pump and air valve linkage.

[0104] This solution not only addresses the problems of existing technologies relying on a single, high-cost dissolved oxygen sensor, oxygenation control being limited to a simple start-stop binary switching, and complex algorithms being difficult to run in real-time on low-cost microcontrollers, but also achieves the following overall technical effects: Without configuring a real dissolved oxygen sensor in the basic hardware, it achieves oxygen deficiency risk assessment accuracy comparable to a real dissolved oxygen sensor through multi-source coupled analysis of water temperature, cumulative air pump downtime, water circulation status, and simulated oxygenation trends; simultaneously, using entry-level microcontrollers such as STM32 as the computing platform, and through a collaborative architecture of fuzzy logic risk assessment models and improved particle swarm optimization (PSO) PID parameter optimization, it achieves millisecond-level linkage response of air pump start-stop and air valve opening, and three-level fine adjustment of oxygen release, reducing oxygenation energy consumption in home aquarium scenarios while also lowering the false alarm rate for oxygen deficiency. It should be noted that the aforementioned algorithms, such as joint filtering and evidence fusion, have been adapted for lightweighting through methods such as table lookup, fixed-point arithmetic, and dimensionality reduction. For example, the three-dimensional sparse tensor lookup table transforms the complex membership calculation into a trilinear interpolation lookup operation, the volumetric Kalman filter is performed only in a 16×16 dimensionality reduction state, and the wavelet packet decomposition layer is adaptively limited to 1-3 layers according to the dynamic window length, ensuring that the overall computational load meets the real-time requirements of entry-level microcontrollers such as STM32F103RET6. This effect is the result of the combined action of multiple mechanisms, including the multi-source parameter joint filtering of the detection module 100, the fuzzy logic risk assessment and dynamic confidence correction of the main control module 200, and the dual-output PID collaborative control of the air pump drive module 300 and the air volume regulation module 400, rather than a simple superposition of the effects of a single component, nor a conventional combination of common knowledge in the field. It is something that those skilled in the art could not have predicted based on existing technology.

[0105] It should be noted that this invention achieves accurate assessment of oxygen deficiency risk and refined graded oxygen release control of air pump and valve linkage in home aquarium scenarios by combining multi-source water state parameter acquisition and joint filtering, pre-feature extraction and evidence fusion, three-dimensional sparse tensor membership mapping, gated attention dynamic confidence correction, bivariate risk association rule reasoning, and online optimization of PID parameters with rate of change prediction and inertia compensation. The above-mentioned algorithmic steps are not isolated but form a progressive collaborative chain of "alignment-cleaning-fusion-standardization-dimensionality reduction-table lookup-weighting-reasoning-defuzzing-optimization": joint filtering uses parameter coupling relationships to inversely correct interpolation errors that may be introduced by time-domain alignment; the pre-feature extraction layer reduces the four-dimensional parameter vector to a comprehensive feature vector with clear physical meaning to reduce the complexity of subsequent calculations; and the three-dimensional sparse tensor lookup table... The complex membership calculation is transformed into a lookup table operation to adapt to the computing power of entry-level microcontrollers. The gated attention weighting layer realizes the dynamic weight allocation of parameters for adaptive signal quality. The bivariate risk association rule matrix realizes collaborative reasoning of multi-source evidence. Finally, the area center method completes the fuzzy to precise conversion and provides continuous risk scores for PID parameter optimization for rate of change prediction and inertia compensation. Each algorithm forms a closed-loop adaptive collaborative control system by inversely constraining the risk assessment accuracy through parameter dynamic weights and compensating for execution delay through feedforward of predicted risk level. At the same time, through the lightweight collaborative architecture of fuzzy logic risk assessment model and improved particle swarm PID parameter optimization, the computational delay problem of existing complex algorithms on low-cost embedded devices is overcome, achieving the overall technical effect of low detection cost, fine control granularity, simple system structure, and strong adaptability to home scenarios.

[0106] In summary, this invention achieves accurate assessment of oxygen deficiency risk in home aquariums and refined, graded oxygen release control through a combination of multi-source water state parameter acquisition and joint filtering, pre-feature coupling extraction and evidence fusion, three-dimensional sparse tensor membership mapping, gated attention dynamic confidence correction, bivariate risk association rule reasoning, and online optimization of PID parameters with rate of change prediction and inertia compensation, without relying on high-cost real dissolved oxygen sensors. Furthermore, through a lightweight collaborative architecture combining a fuzzy logic risk assessment model and an improved particle swarm optimization PID parameter optimization, it overcomes the computational latency problem of existing complex algorithms on low-cost embedded devices, achieving overall technical effects such as low detection cost, fine control granularity, simple system structure, and strong adaptability to home scenarios.

[0107] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0108] This invention provides an intelligent water oxygenation pump. Figure 1 The control architecture diagram of the intelligent water aeration pump is shown. The intelligent water aeration pump specifically includes:

[0109] The intelligent water oxygenation pump includes:

[0110] The detection module 100 is used to collect multi-source water state parameters in the aquarium, preprocess these parameters, and output a preprocessed set of multi-source parameters. These parameters include water temperature data, cumulative inactivity time of the air pump, water circulation status data, and simulated oxygen content data. The detection module 100 can acquire these parameters using a low-cost indirect detection method, primarily collecting water temperature data, cumulative inactivity time of the air pump, water circulation status data, and simulated oxygen content data. It can also connect to a real dissolved oxygen sensor via a universal analog input interface. The detection module 100 of this invention employs low-cost multi-source detection technology and does not rely solely on a real dissolved oxygen sensor. This is because high-precision dissolved oxygen sensors are expensive and require calibration and maintenance, making them unsuitable for basic home aquarium products. Therefore, this invention uses a comprehensive assessment of water temperature data, cumulative inactivity time of the air pump, water circulation status data, and simulated oxygen content data.

[0111] The detection module 100 includes:

[0112] The water temperature detection unit 110 includes a thermistor temperature sensor, which is installed in the aquarium water. It collects aquarium water temperature data based on a preset sampling period. This water temperature data is used to characterize the changing trend of dissolved oxygen saturation concentration in the water. Increased water temperature increases the risk of oxygen deficiency in the water. The system classifies water temperature into three membership levels: normal, slightly high, and high temperature.

[0113] The timing unit 120, integrated within the main control module 200, is used to accumulate the timing duration when the air pump is off, and to reset it when the air pump is on, outputting the cumulative inactivity time of the air pump. This cumulative inactivity time characterizes the continuous consumption of dissolved oxygen in the water body since the last aeration. The system classifies the cumulative inactivity time into three membership levels: short, medium, and long.

[0114] The water circulation status detection unit 130 includes at least one of a water flow velocity sensor or a water level fluctuation detection sensor, used to collect water flow and exchange capacity data and output water circulation status data. When the water circulation status is good, the oxygen exchange efficiency at the water-air interface is high, and when the water circulation status is weak, the risk of hypoxia increases.

[0115] The simulated oxygen trend input unit 140 includes a general-purpose analog input interface configured as an ADC input channel. A potentiometer is connected to generate an analog voltage signal corresponding to a preset virtual dissolved oxygen state. This signal is used to verify the hypoxia risk assessment algorithm and the control logic of the air pump and valve, and to output simulated oxygen state data. In the basic version of the product, the oxygen is simulated by a potentiometer; in the upgraded version, this interface can be replaced by a real analog dissolved oxygen sensor.

[0116] The preprocessing unit 150 is electrically connected to the water temperature detection unit 110, the timing unit 120, the water circulation status detection unit 130, and the simulated oxygen content trend input unit 140. It is used to acquire water temperature data, cumulative inactivity time of the air pump, water circulation status data, and simulated oxygen content status data in real time and integrate them into multi-source water state parameters. It preprocesses the multi-source water state parameters and outputs the preprocessed multi-source parameter set.

[0117] The main control module 200, which is communicatively connected to the detection module 100, is used to receive the preprocessed multi-source parameter set, identify and analyze the multi-source parameter set based on the pre-built fuzzy logic risk assessment model, output the state assessment result, retrieve the preset multi-source inference rule table using the state assessment result as an index, output the water body hypoxia risk level, generate an adaptive dual-output control strategy based on the water body hypoxia risk level, and generate adaptive dual-output control instructions based on the adaptive dual-output control strategy. The main control module 200 uses an STM32F103RET6 as the main control chip, which is responsible for receiving the preprocessed multi-source parameter set, identifying and analyzing the multi-source parameter set based on the pre-built fuzzy logic risk assessment model, outputting the state assessment result, retrieving the preset multi-source inference rule table using the state assessment result as an index, outputting the water body hypoxia risk level, generating an adaptive dual-output control strategy based on the water body hypoxia risk level, and generating adaptive dual-output control instructions based on the adaptive dual-output control strategy.

[0118] In this embodiment of the invention, Figure 2A circuit connection diagram of the main control module 200 is shown. The simulated oxygen trend input unit 140 in the detection module 100 is connected to the main control module 200 through a general analog input interface. Its VCC terminal is connected to the 3.3V power output of the main control module 200, its OUT terminal is connected to the PA0 pin of the main control chip STM32F103RET6, and its GND terminal shares a common ground with the main control module 200. The PA0 pin is configured as an ADC input channel to read simulated oxygen state data in the range of 0-3.3V; the gas pump drive module 300... A relay module is used, with its IN terminal connected to the PB0 pin of the main control chip, VCC and GND terminals connected to an external 5V DC power supply, COM terminal connected to the positive terminal of the external 5V power supply, and NO terminal connected to the positive terminal of the air pump drive module 300. The PB0 pin is configured as a GPIO output to control the relay to engage or disengage, thereby executing the air pump start and stop actions. In the default state of the relay module, the normally closed contacts of COM and NC are connected to de-energize and shut down the air pump. When the PB0 output is high, the relay coil is energized and engages, switching to the NO contact to energize and start the air pump. The air volume regulation module 400 uses a servo motor / valve module. Its SIG signal line is connected to the PA6 pin of the main control chip, and its VCC and GND pins are connected to an external 5V DC power supply. The PA6 pin is configured as a PWM output to output a valve opening control signal, driving the servo motor to rotate and change the opening of the air volume regulation valve, thereby executing the valve opening regulation action. The wireless communication module 500 uses an ESP32 / ESP8266 communication module. Its RX pin is connected to the TX pin of the main control chip, its TX pin is connected to the RX pin of the main control chip, and its VCC pin is connected to the main control chip. The 3.3V power output and GND terminal of module 200 are grounded with the main control module 200, enabling serial communication between the main control module 200 and the user terminal 600 for uploading status data and receiving control commands. The GND terminals of the main control module 200, relay module, servo / valve module, ESP32 / ESP8266 communication module, and analog oxygen trend input unit 140 are all grounded to the GND terminal of the external 5V DC power supply, forming a unified reference ground potential to ensure the electrical integrity of signal transmission between modules.

[0119] Meanwhile, the main control chip is not limited to STM32F103RET6, and can be replaced with other STM32 series microcontrollers, ESP32, Arduino control boards, or other microcontrollers with ADC, GPIO, PWM, and UART serial communication functions; the relay in the air pump drive module 300 can be replaced with a MOSFET drive module, a transistor drive module, or a dedicated motor drive chip; the fuzzy logic risk assessment model can be replaced with a rule control model, a PID control model, a fuzzy PID control model, or an adaptive threshold algorithm model based on historical data, but retains the basic architecture of multi-source parameter input and adaptive dual-output hierarchical control; the user terminal 600 can be replaced with a WeChat mini-program, a web-based control interface, or a local touchscreen interface.

[0120] Among them, such as Figure 3 As shown, the main control module 200 includes:

[0121] Model building unit 210 is used to pre-build a fuzzy logic risk assessment model. It uses a reference-level dissolved oxygen detection device to collect the actual dissolved oxygen concentration corresponding to the time of the multi-source water body state parameters, and generates hypoxia risk labels based on the actual dissolved oxygen concentration. The user-defined threshold is used as a control preference parameter, which, together with the multi-source water body state parameters, is used as the model input. The hypoxia risk label is used as the supervision target to optimize the fuzzy membership parameters and rule weights, and outputs a converged fuzzy logic risk assessment model.

[0122] The fuzzy logic risk assessment unit 220 receives the preprocessed multi-source parameter set and user-defined thresholds. Based on a pre-built fuzzy logic risk assessment model, it performs membership degree classification on the parameters in the multi-source parameter set and outputs the state assessment results. During membership degree classification, water temperature data is mapped to three membership levels: normal, slightly high, and high temperature; the cumulative inactivity time of the air pump is mapped to three membership levels: short, medium, and long; water circulation state data is mapped to three membership levels: good, average, and weak; and simulated oxygen content state data is mapped to three membership levels: normal, critical, and slightly low. Among these, the water temperature data is considered normal. The membership function center value for the grade is 22℃, and the boundary range is 18℃ to 26℃. The membership function center value for the higher grade is 28℃, and the boundary range is 24℃ to 32℃. The membership function center value for the high temperature grade is 34℃, and the boundary range is 30℃ to 40℃. The membership function center value for the cumulative inactivity time of the air pump is 15 minutes, and the boundary range is 0 minutes to 30 minutes. The membership function center value for the medium grade is 45 minutes, and the boundary range is 30 minutes to 60 minutes. The membership function center value for the long grade is 90 minutes, and the boundary range is 60 minutes to 120 minutes.

[0123] The dynamic confidence correction unit 230 is used to traverse the multi-source parameter set, dynamically adjust the membership confidence of the state evaluation result based on the parameter coupling coefficient of the multi-source parameter set, and determine the dynamic weight of the parameter by combining the membership confidence.

[0124] The rule reasoning unit 240 is communicatively connected to the fuzzy logic risk assessment unit 220. It combines the dynamic weight of parameters, the state assessment results and the multi-source reasoning rule table matching analysis to output the fuzzy risk level.

[0125] The multi-source inference rule table is stored in the STM32 control program. After each sampling, the system matches the corresponding rule according to the current input state. Table 1 shows an example of the rule table of the multi-source inference rule table in this embodiment of the invention. Through the above rules, the present invention realizes multi-factor comprehensive judgment, rather than relying on only a dissolved oxygen sensor value or a fixed time threshold.

[0126] Table 1. Examples of rules in the multi-source inference rule table

[0127]

[0128] The defuzzification unit 250 uses the area center method to perform integral calculations on the fuzzy risk levels, outputting the water body hypoxia risk level. The water body hypoxia risk level is divided into three categories: low risk, medium risk, and high risk. Low risk indicates that the current water body condition is basically normal and no forced oxygenation is needed. Medium risk indicates that the water body may begin to show signs of hypoxia and moderate oxygenation is needed. High risk indicates that the water body is highly likely to be hypoxic and immediate enhanced oxygenation is required.

[0129] In this embodiment of the invention, the implementation process of the area center method is as follows: First, obtain the fuzzy risk level output by the rule inference block. This fuzzy risk level includes the membership values ​​of three output levels: low risk, medium risk, and high risk. Second, for each output level, construct the output membership function of that level based on its membership value and a preset level center value. The output membership function is a triangular or trapezoidal membership function, with its height corresponding to the membership value of that level, its base covering the preset numerical range of that level, and the bases of the membership functions of adjacent levels overlapping to ensure the continuity of the defuzzification results. Then, perform a point-by-point enlargement operation on the output membership functions of the three output levels in the same coordinate system, summing the results. The overall output membership function is generated, and its upper envelope is the comprehensive contribution distribution of all activation rules. Next, the total area of ​​the figure enclosed by the overall output membership function and the horizontal axis is calculated. Specifically, the overall output membership function is numerically integrated over the interval from the lower limit of the low-risk level to the upper limit of the high-risk level, with the integration step size preset according to the defuzzification accuracy requirements. At the same time, the first moment of the figure about the vertical axis is calculated. Specifically, the product of the overall output membership function and the horizontal axis coordinate value is numerically integrated over the same interval. Finally, the first moment is divided by the total area to obtain the horizontal coordinate of the center point of the area. This horizontal coordinate is the continuous risk level value after defuzzification, denoted as R, and the value range of R is the closed interval [0, 1]. Then, the continuous risk score R is compared with the preset segmented thresholds. When R is less than the first threshold, it is determined to be a low risk level. When R is between the first threshold and the second threshold, it is determined to be a medium risk level. When R is greater than or equal to the second threshold, it is determined to be a high risk level. Discrete water body hypoxia risk levels are output, where the first threshold and the second threshold are 0.4 and 0.75, respectively.

[0130] The dual-output control strategy generation unit 260 is communicatively connected to the defuzzification unit 250. Using the water body hypoxia risk level as an index, it generates an adaptive dual-output control strategy based on the rate of change prediction and inertial compensation mechanism. The adaptive dual-output control strategy includes a pump control strategy and a valve opening control strategy corresponding to the water body hypoxia risk level. Specifically, the low-risk level corresponds to a pump shutdown strategy and a valve small opening strategy, the medium-risk level corresponds to a pump start strategy and a valve medium opening strategy, and the high-risk level corresponds to a pump continuous operation strategy and a valve large opening strategy.

[0131] In this embodiment of the invention, the dual-output control strategy generation unit 260 outputs not a single pump-on or pump-off command, but two control results simultaneously: a pump control strategy and a valve opening control strategy.

[0132] When the risk of hypoxia is low, the main control module 200 controls the air pump to shut down or maintain low-frequency short-term operation; at the same time, the air valve is in a closed or slightly open state.

[0133] When the risk of hypoxia is medium, the main control module 200 controls the relay to activate and start the air pump; at the same time, it controls the servo motor through the PWM signal to keep the air valve at a medium opening to achieve normal oxygenation.

[0134] When the risk of oxygen deficiency is high, the main control module 200 controls the air pump to run continuously and controls the servo motor to adjust the air valve to a large opening, so that more air enters the water through the air stone to achieve strong oxygenation.

[0135] In other words, this invention maps the level of oxygen deficiency risk to two actions: "air pump start / stop + air valve opening," enabling ordinary small air pumps to achieve graded oxygen release without the need for costly variable frequency oxygenation equipment.

[0136] This invention employs a staged oxygen release technology that links an air pump and an air valve. Traditional air pumps can only be turned on or off, while this invention controls the air valve opening to allow the oxygen release to vary according to the degree of oxygen deficiency. In this embodiment, the air pump drive module 300 responds to an adaptive dual-output control command, synchronously outputting air pump control signals to execute air pump start-up and shutdown actions. When the system determines a high risk of oxygen deficiency, the STM32 controller PB0 outputs a high level, energizing the relay and starting the air pump drive module 300. When the system determines a low risk of oxygen deficiency, the STM32 controller PB0 outputs a low level, deactivating the relay and shutting down the air pump.

[0137] The air volume regulation module 400 responds to adaptive dual-output control commands, synchronously outputting air valve opening control signals to execute air valve opening adjustment actions. When the system determines a high risk of hypoxia, the STM32 outputs a PWM signal via PA6, causing the servo motor to drive the air valve to a large opening position, increasing the air output. When the system determines a medium risk of hypoxia, the air pump drive module 300 starts, and the air valve maintains a medium opening, achieving normal oxygenation. When the system determines a low risk of hypoxia, the air pump can be shut off or run at low intensity for a short period, and the air valve is at a small opening. When the system determines that the water condition is normal, the air pump shuts off, and the air valve returns to the closed or minimum opening position.

[0138] The control command output unit 270 is communicatively connected to the dual-output control strategy generation unit 260, and is used to convert the adaptive dual-output control strategy into adaptive dual-output control commands, which are output to the air pump drive module 300 and the air volume adjustment module 400 respectively.

[0139] The air pump drive module 300 responds to adaptive dual-output control commands, synchronously outputs air pump control signals, and executes air pump start-stop actions. The air pump drive module 300 uses a 5V relay module. The main control module 200 controls the relay to engage or disengage by outputting high and low levels through the PB0 pin, thereby controlling whether the air pump drive module 300 is working.

[0140] The gas volume regulation module 400, responding to adaptive dual-output control commands, synchronously outputs a gas valve opening control signal to execute the gas valve opening regulation action. The gas volume regulation module 400 uses a servo motor and a gas volume regulation valve. The main control module 200 outputs a PWM signal through the PA6 pin to control the servo motor angle. The servo motor drives the gas volume regulation valve to change its opening, thereby changing the oxygen release.

[0141] The wireless communication module 500 is connected to the main control module 200 via a serial port. It uploads status data from the main control module 200 to the user terminal 600 and receives control commands from the user terminal 600. The wireless communication module 500 uses an ESP32 or ESP8266. The main control module 200 is connected to the communication module via a serial port, and the communication module then communicates with the user terminal 600 to achieve status display, manual control, and threshold parameter setting, etc.

[0142] The user terminal 600 is used to display the water condition assessment results, the water hypoxia risk level, the air pump status and air valve opening, and to receive user-input control commands and user-set thresholds.

[0143] In this embodiment of the invention, the user terminal 600 includes:

[0144] The status display unit 610 is used to display the current water body oxygen deficiency risk level, air pump operating status, air valve opening degree, real-time water temperature, cumulative air pump inactivity time and abnormal reminder information;

[0145] The mode switching unit 620 is used to switch between automatic mode and manual mode. In automatic mode, it operates autonomously based on an adaptive dual-output control strategy. In manual mode, the control commands input by the user are executed first.

[0146] The manual control unit 630 is used to receive user commands to manually turn the air pump on or off and adjust the air valve opening.

[0147] The threshold parameter setting unit 640 is used to receive the user's input oxygenation preference settings and user-set thresholds. The oxygenation preference settings include energy-saving mode, normal mode or strong oxygenation mode, and the user-set thresholds include water temperature alarm threshold, air pump maximum continuous running time threshold, and automatic recovery time threshold.

[0148] The safety priority determination unit 650 is used to monitor abnormal operating conditions in manual mode. When an abnormal operating condition is detected, a safety protection command is generated. The safety protection command has a higher priority than the manual control command. The abnormal operating conditions include: the air pump operating current is lower than 60% of the normal operating current for the first 5 seconds (achieved by connecting a sampling resistor in series with the power supply circuit and acquiring data via an ADC), the deviation between the air valve opening feedback and the set value exceeds the allowable range, the wireless communication module is disconnected for more than a set time, and the sensor output signal remains unchanged for 10 consecutive sampling cycles or exceeds the physically reasonable range. When any of the above abnormal operating conditions is detected, a safety protection command is generated. The safety protection command includes: immediately shutting down the air pump, adjusting the air valve opening to the minimum or closing it, and pushing alarm information through the user terminal. The execution of the safety protection command is still prioritized in manual mode.

[0149] In this embodiment of the invention, the control process of the intelligent water aeration pump includes initialization, data acquisition, data processing, risk assessment, hierarchical control, user terminal 600 interaction, and anomaly protection. After the system is powered on, the STM32 first initializes the ADC, GPIO, PWM, and serial communication. After initialization, the system enters a cyclic detection state. The STM32 periodically reads water temperature data, simulated oxygen content data, air pump running time, and water circulation status data. The acquired data is first filtered by the preprocessing unit 150 to reduce the impact of instantaneous fluctuations on the judgment results. Then, the system inputs the data into the fuzzy logic risk assessment model to obtain the current water hypoxia risk level. According to the risk level, the STM32 outputs air pump control signals and air valve opening control signals respectively. Finally, the system uploads the current status to the user terminal 600 and waits for the next detection cycle.

[0150] This invention operates using a periodic control method. Each control cycle includes a sampling phase, a judgment phase, an execution phase, and a communication phase.

[0151] During the sampling phase, the STM32 reads the data from the detection end.

[0152] During the judgment phase, the STM32 runs a fuzzy logic risk assessment model and outputs the risk level of water body hypoxia.

[0153] During the execution phase, PB0 outputs a relay control signal, and PA6 outputs a PWM control signal.

[0154] During the communication phase, the STM32 sends the status to the ESP32 / ESP8266 via serial port, and then the communication module uploads it to the user terminal 600.

[0155] This timing design ensures that the system operates stably according to a fixed cycle, rather than being controlled randomly. Simultaneously, the system allows setting minimum start-up and minimum shutdown times to prevent the air pump from frequently starting and stopping under critical conditions.

[0156] In addition, the present invention also includes an anomaly protection mechanism to improve equipment reliability.

[0157] When the detected signal remains unchanged for an extended period or exceeds a reasonable range, the system determines that the detection module 100 is malfunctioning and alerts the user to check the sensor via the user terminal 600. The preprocessing unit 150 continuously monitors the output values ​​of each sensor. If a signal remains unchanged for 10 consecutive sampling cycles or exceeds a physically reasonable range (where the physically reasonable range refers to a pre-calibrated effective interval based on the physical characteristics of each sensor, i.e., the reasonable range for water temperature data is 0℃-50℃, the reasonable range for cumulative inactivity of the air pump is 0min-1440min, the reasonable range for water circulation status data is 0%-100%, and the reasonable range for simulated oxygen content data is 0-3.3V), the sensor is determined to be faulty. An alarm message requesting sensor inspection is pushed to the APP, and the weight of this parameter in the risk assessment is automatically reduced to prevent misjudgment due to erroneous data.

[0158] When the air pump runs for an extended period without improving the system's condition, the system determines that there may be a blockage in the air tube, a blockage in the air stone, or a malfunction in the air pump, and reminds the user to check the air path. Specifically, within the first 5 seconds after each air pump startup, the main control module 200 monitors the air pump's operating current (achieved by connecting a sampling resistor in series in the power supply circuit and acquiring the data via an ADC). If the current value is lower than 60% of the normal operating current (indicating that the air tube has come loose or the air stone is blocked, and the air pump is in an unloaded state), it is determined that the air path is abnormal, and the air pump is immediately shut down. A maintenance reminder for air path blockage is pushed through the APP, requesting the user to check the air tube or air stone.

[0159] When the communication module is disconnected, the system does not stop local control, but continues to operate in automatic mode to ensure that the aquarium can still automatically oxygenate.

[0160] When a user keeps the system in manual mode for an extended period of time, the system can automatically switch back to automatic mode after a set time to prevent the user from forgetting to turn off the air pump.

[0161] In this embodiment of the invention, the main control module 200 comprises a model building unit 210, a fuzzy logic risk assessment unit 220, a dynamic confidence correction unit 230, a rule reasoning unit 240, a defuzzification unit 250, a dual-output control strategy generation unit 260, and a control command output unit 270, forming a progressive collaborative architecture. The model building unit 210 provides a basic reasoning framework for fuzzy logic risk assessment. The fuzzy logic risk assessment unit 220 collaborates with the dynamic confidence correction unit 230 to achieve adaptive discrimination of multi-source signal quality and real-time optimization of parameter dynamic weights. The rule reasoning unit 240 integrates parameter dynamic weights and state assessment results, taking into account parameter reliability and physical correlation to perform risk level reasoning. The defuzzification unit 250 completes the mapping conversion from fuzzy to precise. The dual-output control strategy generation unit 260 optimizes the control strategies of the air pump and air valve based on rate of change prediction and inertia compensation mechanisms. The control command output unit 270 realizes the synchronous drive of the two actuators. The various sub-units in the main control module 200 form a closed-loop adaptive collaborative control system by using dynamic weights of parameters to inversely constrain the accuracy of risk assessment and by using predictive risk levels to feedforward and compensate for execution delay. This enables the system to achieve accurate fusion of multi-source parameters, dynamic assessment of hypoxia risk, and millisecond-level linkage response of air pumps and valves with an entry-level microcontroller, without relying on high-cost dissolved oxygen sensors. This results in low detection cost, fine control granularity, and a simple system structure.

[0162] In a further preferred embodiment of the present invention, such as Figure 4 As shown, the method for preprocessing state parameters of multi-source water bodies includes:

[0163] S101: Acquire multi-source water body state parameters, mark the collection timestamps for water temperature data, cumulative inactivity time of air pump, water circulation state data and simulated oxygen content state data, use a preset reference period as the time window, align the parameters on the time axis, group the parameter data collected within the same reference period into the same parameter frame, and fill the parameter data that was not collected within the reference period into the corresponding parameter frame using nearest neighbor interpolation, and output the time-domain aligned multi-source water body state parameters;

[0164] S102. The aquarium water environment is affected by external disturbances (such as water changes, fish movement, bubble disturbances, etc.), and the sensor data is prone to instantaneous pulse noise or occasional jumps. If data containing outliers is directly input into the fuzzy logic risk assessment model, it will lead to distortion of membership degree division, and thus misjudgment of risk level. Therefore, in this step, the time-domain aligned multi-source water state parameters are loaded, and a sliding window anomaly detection is performed on the time-domain aligned multi-source water state parameters. The mean and standard deviation of the parameter data in the current window are calculated, and data points that deviate from the mean by more than three times the standard deviation are judged as outliers. The outlier is replaced by the median of the window where the outlier is located, and the multi-source water state parameters after outlier removal are output.

[0165] S103: Obtain the state parameters of the multi-source water bodies after outlier removal, perform joint filtering on the state parameters of the multi-source water bodies, and output the filtered state parameters of the multi-source water bodies.

[0166] S104, the filtered multi-source water state parameters are mapped to a unified numerical range, which is a closed interval [0,1]. The mapped parameter data constitute the preprocessed multi-source parameter set.

[0167] In this embodiment of the invention, when preprocessing multi-source water state parameters, temporal consistency of heterogeneous multi-source data is ensured through temporal domain alignment, instantaneous noise pollution is eliminated through sliding window anomaly detection and median replacement, cross-validation of physical correlations between parameters and multi-scale time-frequency denoising are achieved through joint filtering, and dimensional differences are eliminated and physical distribution characteristics are preserved through normalization mapping. This forms a progressive collaborative system of alignment-cleaning-fusion-standardization: temporal domain alignment provides a synchronous benchmark for anomaly detection, anomaly detection reduces noise interference for joint filtering, joint filtering uses parameter coupling relationships to reverse the interpolation errors that may be introduced by temporal domain alignment, and normalization mapping uniformly inputs the fused high-quality parameters into the fuzzy logic risk assessment model.

[0168] In a further preferred embodiment of the present invention, the method for joint filtering of multi-source water state parameters includes:

[0169] S1031, Obtain the multi-source water state parameters after outlier removal. Determine the parameter coupling coefficients based on the weights and mutual information of the multi-source water state parameters. Establish a parameter correlation matrix based on the parameter coupling coefficients. The parameter correlation matrix represents the first coupling coefficient between water temperature data and simulated oxygen-containing state data, the second coupling coefficient between the cumulative inactivity of the air pump and simulated oxygen-containing state data, and the third coupling coefficient between water circulation state data and simulated oxygen-containing state data. Mutual information is a quantitative index in information theory used to measure the degree of interdependence between two random variables, representing the reduction in uncertainty of one variable after knowing the other. A larger mutual information value indicates a stronger nonlinear correlation between the two parameters. The parameter coupling coefficient is a normalized correlation strength index calculated based on the mutual information between the multi-source water state parameters, with a value range of [0,1]. A value closer to 1 indicates a tighter physical coupling relationship between the parameters. The parameter correlation matrix is ​​a square matrix constructed using the parameter coupling coefficients as elements, used to systematically describe the mutual constraints between multi-source parameters. The diagonal elements of the matrix are 1, and the off-diagonal elements are the coupling coefficients of the corresponding parameter pairs.

[0170] Among them, when determining the parameter coupling coefficient based on the weights and mutual information of multi-source water state parameters, the formula for calculating the parameter coupling coefficient is expressed as follows:

[0171]

[0172]

[0173] in, This represents the mutual information between the state parameter i of the multi-source water body and the associated state parameter j. This represents the joint probability density distribution of the multi-source water body state parameter i and the associated state parameter j. These are the marginal probability density distributions of multi-source water body state parameter i and associated state parameter j, respectively. These are the parameter coupling coefficient and the weight of the multi-source water state parameter i, respectively.

[0174] S1032, using the parameter correlation matrix as a constraint, performs collaborative correction on the state parameters of multiple water bodies and calculates the coupling deviation between the state parameters of multiple water bodies and the associated state parameters.

[0175] S1033, determine whether the coupling deviation between the state parameters of the multi-source water body and the associated state parameters exceeds the preset deviation threshold;

[0176] S1034, if the coupling deviation between the multi-source water state parameters and the associated state parameters exceeds a preset deviation threshold, the multi-source water state parameters are constrained and corrected using a weighted combination of the associated state parameters. The multi-source water state parameters are then output after parameter correlation constraint filtering. The collaborative constraint correction utilizes the physical coupling relationship between parameters and performs weighted estimation based on the real-time values ​​of the associated parameters, ensuring that the corrected parameters conform to the physical consistency constraints of the multi-source parameters while retaining the current state characteristics. This mechanism implements a cross-validation strategy of correcting suspicious parameters with reliable parameters, improving the system's robustness under conditions of sensor performance degradation or disturbance.

[0177] If the coupling deviation between the multi-source water state parameters and the associated state parameters does not exceed the preset deviation threshold, proceed to step S1035.

[0178] S1035: Obtain the multi-source water body state parameters after parameter correlation constraint filtering. Determine the dynamic window length of the sliding window based on the parameter dynamic response. Combine the dynamic window length with the multi-scale wavelet packet decomposition of the multi-source water body state parameters to separate the multi-source water body state parameters into low-frequency approximate components and high-frequency detail components. The low-frequency components are filtered by long-window moving average filtering to extract the trend term, and the high-frequency components are retained after removing impulse noise by improved median filtering. The number of decomposition layers of multi-scale wavelet packet decomposition is adaptively determined according to the dynamic window length, with a long window corresponding to 3 layers of decomposition and a short window corresponding to 1 layer of decomposition. Reconstruct the filtered low-frequency and high-frequency components into time-domain signals and output the multi-scale decomposed and filtered multi-source water body state parameters.

[0179] S1036 constructs a three-dimensional state vector containing dissolved oxygen concentration, water temperature, and cumulative inactivity time of the air pump based on the multi-source water body state parameters after multi-scale decomposition filtering. With water temperature as the known input, dissolved oxygen concentration as the state to be estimated, and cumulative inactivity time of the air pump as the control variable, a nonlinear state-space model is established. The nonlinear state-space model refers to a mathematical model that describes the dynamic behavior of the system using state equations and observation equations. The state equations characterize the evolution of the system state over time, and the observation equations characterize the relationship between the state variables and the measurable outputs. Nonlinear functional relationships are allowed in the model. Combining the nonlinear state-space model with volumetric Kalman filtering, the multi-source water body state parameters after time-consistent filtering are output.

[0180] In this embodiment of the invention, when jointly filtering multi-source water state parameters, the physical correlation representation of multi-source parameters is constructed through parameter coupling coefficients and parameter correlation matrices. Cross-validation and physical consistency restoration of abnormal parameters are achieved through coupling deviation determination and collaborative constraint correction. Refined noise separation and trend extraction in the time-frequency domain are achieved through multi-scale wavelet packet decomposition and adaptive median filtering. Optimal state estimation at the dynamic level of the nonlinear system is achieved through capacitive Kalman filtering. This joint filtering mechanism enables the system to perform cross-validation and complementary enhancement using the physical coupling information between multi-source parameters without increasing sensor hardware costs, thereby suppressing local noise and occasional faults of a single sensor. Finally, a high-quality multi-source parameter set that is physically self-consistent, temporally stable, and complete with multi-scale features is output.

[0181] In a further preferred embodiment of the present invention, Figure 5 A schematic diagram of a fuzzy logic risk assessment model architecture is shown. The fuzzy logic risk assessment model includes an input layer, a risk assessment architecture, and an output layer. The risk assessment architecture includes a pre-feature extraction layer and a risk association layer. The input layer is used to obtain a multi-source parameter set, encode and reduce the dimensionality of the multi-source parameter set, and output the encoded four-dimensional parameter vector. The purpose of the encoding and dimensionality reduction process is to uniformly map the multi-source parameters after joint filtering into a standardized input format that the model can recognize, thereby eliminating the influence of dimensional differences on subsequent feature extraction. The pre-feature extraction layer is used to perform feature coupling extraction and evidence correction operations on the four-dimensional parameter vector and output the corrected comprehensive feature vector.

[0182] The pre-feature extraction layer includes an environmental stress index extraction branch, an exchange feature extraction branch, and an index fusion branch. The environmental stress index extraction branch performs nonlinear coupling calculations between water temperature data and the cumulative downtime of the air pump. Based on the temperature correction coefficient for dissolved oxygen saturation concentration from water temperature data and the cumulative effect coefficient of the cumulative downtime of the air pump, the environmental stress index is calculated. The environmental stress index characterizes the degree of dissolved oxygen deficit accumulated in the water body due to pump shutdown under the current temperature conditions. The water-air exchange efficiency index characterizes the comprehensive level of oxygen replenishment capacity achieved by the water body through circulation and the current oxygen content trend. When the water body circulation is good, the oxygen exchange efficiency at the water-air interface is high, and the risk of hypoxia is low. When the water body circulation is weak, the system increases the hypoxia risk assessment level, and the system classifies the water body circulation state into three levels: good, average, and weak. Simulated oxygen content data is used to represent the current trend change in water body oxygen content. In the basic version of the product, it is simulated by a potentiometer, and in the upgraded version, it can be replaced by a real analog dissolved oxygen sensor. The system classifies the simulated oxygen content data into three levels: normal, critical, and low. The exchange feature extraction branch dynamically couples the water circulation state with the oxygen content trend, enabling the system to assess the oxygen replenishment efficiency of the water body through natural circulation and artificial aeration.

[0183] The feature extraction branch is used to dynamically couple water circulation state data with simulated oxygenation state data. Based on the water circulation state data, the air-water interface update rate is calculated. Combined with the slope of change in simulated oxygenation state data, the water-air exchange efficiency index is calculated. The water-air exchange efficiency index represents the comprehensive level of oxygen replenishment capacity achieved by the water body through circulation flow and the current oxygenation trend. The index fusion branch is used to obtain the environmental stress index and the water-air exchange efficiency index. It is used to resolve evidence conflicts and perform weighted fusion on the environmental stress index and the water-air exchange efficiency index, and output the corrected comprehensive feature vector and the water-air exchange correction vector.

[0184] The risk association layer includes a membership mapping block, a dynamic confidence correction block, a rule reasoning block, and a defuzzification layer;

[0185] The membership mapping block is used to obtain the corrected comprehensive feature vector and water-air exchange correction vector. A three-dimensional sparse tensor lookup table is constructed based on the comprehensive feature vector and water-air exchange correction vector. The coordinates of the center value grid points of the three-dimensional sparse tensor lookup table are shifted as a whole based on a preset user-defined threshold. In the energy-saving mode, the coordinates of each center value are shifted by 0.15 to the high confidence interval, and in the enhanced mode, they are shifted by 0.15 to the low confidence interval, realizing the dynamic reshaping of the membership division boundary by user preferences. The tensor size is 16×16×16. Based on the three-dimensional sparse tensor lookup table, the preset parameter-level membership mapping relationship is traversed. The interpolated membership degree between the comprehensive feature vector and the parameter-level membership degree is calculated using the trilinear interpolation method, and the state evaluation result composed of 12-dimensional membership degree values ​​is output.

[0186] The dynamic confidence correction block is a gated attention weighting layer. The dynamic confidence correction block is used to obtain the state evaluation result composed of 12-dimensional membership values ​​and parameter coupling coefficients. Based on the state evaluation result, the basic confidence of the state evaluation result is constructed. The basic confidence of the state evaluation result is dynamically adjusted in combination with the parameter coupling coefficients to obtain the membership confidence of the state evaluation result. Using the membership confidence of the state evaluation result as a priori, the membership confidence is directly linearly mapped to the dynamic weight of the parameters.

[0187] The rule inference block is used to obtain the comprehensive feature vector and the water-air exchange correction vector. Based on the comprehensive feature vector and the water-air exchange correction vector, a joint membership distribution is constructed. The comprehensive feature vector is mapped to three membership levels: low stress, medium stress, and high stress. The water-air exchange correction vector is mapped to three membership levels: high efficiency, medium efficiency, and low efficiency. A bivariate risk association rule matrix is ​​established by combining the joint membership distribution and the dynamic weight of the parameters. The activation intensity of the bivariate risk association rule matrix in the multi-source inference rule table is calculated based on the fuzzy clustering algorithm. The activation intensity of the multi-source inference rule table is classified and superimposed according to the output risk level to obtain the fuzzy risk level.

[0188] The defuzzification layer uses the area center method to perform integral calculations on the fuzzy risk level and outputs the water body hypoxia risk level.

[0189] In this embodiment of the invention, the fuzzy logic risk assessment model achieves accurate risk level assessment of multi-source heterogeneous parameters from raw data collection through a progressive collaborative architecture consisting of an input layer, a pre-feature extraction layer, a risk association layer, and an output layer. The pre-feature extraction layer reduces the four-dimensional parameter vector into a comprehensive feature vector with clear physical meaning through dual-branch coupling extraction and evidence fusion of the environmental stress index and the water-air exchange efficiency index, thereby reducing the complexity of subsequent calculations while retaining key risk information. The risk association layer achieves dynamic reshaping of user preferences for membership boundaries through a three-dimensional sparse tensor lookup table, achieves dynamic weight allocation of parameters with adaptive signal quality through a gated attention weighting layer, and achieves collaborative reasoning of multi-source evidence through a bivariate risk association rule matrix. This enables the system to accurately assess hypoxia risk by comprehensively considering multi-source information such as water temperature, pump shutdown duration, circulation status, and oxygen content trends without relying on real dissolved oxygen sensors. The output layer completes the fuzzy-to-precise conversion using the area center method, providing a reliable decision-making basis for the linkage control of air pumps and valves.

[0190] In a further preferred embodiment of the present invention, such as Figure 6 As shown, the method for classifying the membership of parameters in a multi-source parameter set based on a pre-built fuzzy logic risk assessment model includes:

[0191] S201: Obtain the multi-source parameter set, encode and reduce the dimensionality of the multi-source parameter set, and output the encoded four-dimensional parameter vector.

[0192] S202, nonlinearly coupled calculations are performed between water temperature data and cumulative non-running time of air pump. Based on the temperature correction coefficient of dissolved oxygen saturation concentration and the cumulative effect coefficient of cumulative non-running time of air pump, the environmental stress index is calculated.

[0193] The formula for calculating the environmental stress index is as follows:

[0194]

[0195] In the formula, , representing the baseline oxygen consumption rate coefficient, water temperature data, and standard water temperature data, respectively. b and d represent the temperature sensitivity coefficient and the cumulative effect decay index (used to characterize the natural decay effect of oxygen consumption rate after a decrease in dissolved oxygen concentration), respectively. The temperature sensitivity coefficient is the relative change rate of the environmental stress index caused by a unit change in water temperature, and its value range is calibrated based on the exponential relationship between fish metabolic rate and water temperature. The cumulative effect decay index d is calibrated through oxygen consumption experiments, is dimensionless, and ranges from 0.5 to 1.5. Based on oxygen consumption experiments in closed water bodies, during the process of dissolved oxygen concentration decreasing from saturation to criticality, the oxygen consumption rate decays exponentially with time. By fitting the correspondence between the cumulative inactivity time of the air pump and the environmental stress index in the historical multi-source parameter set using the least squares method, the calibration value of d is obtained as 0.85. This value characterizes that the effect of pump shutdown time on the environmental stress index exhibits sublinear growth (0 < d < 0). 1) In the initial stage of pump shutdown, dissolved oxygen is consumed rapidly and the environmental stress index increases rapidly. As the pump shutdown time is extended and the dissolved oxygen concentration decreases, the rate of biological oxygen consumption naturally slows down due to hypoxia stress, and the growth of the environmental stress index slows down, which is consistent with the physical self-limiting mechanism of oxygen consumption in water bodies.

[0196] S203, dynamically couples water circulation state data with simulated oxygenation state data, calculates the air-water interface update rate based on water circulation state data, and calculates the water-air exchange efficiency index by combining the change slope of simulated oxygenation state data. The water-air exchange efficiency index is an indicator that comprehensively evaluates the water body's ability to replenish oxygen through circulation flow and the current oxygenation trend by integrating water circulation state data and simulated oxygenation state data.

[0197] The formula for calculating the water-air exchange efficiency index is as follows:

[0198]

[0199] In the formula, Indicates the water-air exchange efficiency index. This represents the gas-water interface renewal rate and the oxygen content trend compensation coefficient. To simulate the slope of changes in oxygen-containing state data;

[0200] S204, obtains the environmental stress index and water-air exchange efficiency index, and uses them to resolve evidence conflicts and perform weighted fusion, outputting the corrected comprehensive feature vector and water-air exchange correction vector.

[0201] The integrated feature vector and the water-air exchange correction vector are expressed as follows:

[0202]

[0203]

[0204] in, These represent the integrated eigenvector and the water-air exchange correction vector, respectively. These represent the basic probability allocation values ​​of the environmental stress index and the water-air exchange efficiency index, respectively. The degree of conflict of evidence between the environmental stress index and the water-air exchange efficiency index is used to characterize the degree of contradiction between the risk directions indicated by the two indices. The environmental stress index assesses the risk of oxygen deficiency in water bodies from the "oxygen-consuming side," with a higher value indicating a more severe cumulative dissolved oxygen deficit due to pump shutdowns at the current temperature. The water-air exchange efficiency index assesses the oxygen replenishment capacity of water bodies from the "oxygen-supply side," with a higher value indicating more sufficient oxygen replenishment through circulation and a better current oxygenation trend. When the environmental stress index indicates a high risk of oxygen deficiency while the water-air exchange efficiency index simultaneously indicates high-efficiency replenishment, the risk conclusions supported by the two evidence sources directly contradict each other. In this case, the degree of conflict of evidence is high, and directly merging them would produce absurd conclusions that do not conform to physical reality.

[0205] The formula for calculating the degree of conflict of evidence is as follows:

[0206]

[0207] In the formula, These represent the basic probability allocation values ​​of the environmental stress index and the water-air exchange efficiency index, respectively. These represent the normalized values ​​of the basic probability distribution of the environmental stress index and the basic probability distribution of the water-air exchange efficiency index, respectively. It is a very small constant.

[0208] S205: Obtain the corrected comprehensive feature vector and water-air exchange correction vector. Construct a three-dimensional sparse tensor lookup table based on the comprehensive feature vector and water-air exchange correction vector. Traverse the preset parameter-level membership mapping relationship based on the three-dimensional sparse tensor lookup table. Calculate the interpolated membership degree between the comprehensive feature vector and the parameter-level membership degree using trilinear interpolation. Output the state evaluation result composed of 12-dimensional membership degree values. The three-dimensional sparse tensor lookup table refers to a sparse data lookup structure constructed with the comprehensive feature vector, water-air exchange correction vector, and user-defined threshold as three dimensions. It has a size of 16×16×16 and stores the membership degree mapping relationship only at key grid points. Non-key areas are calculated through interpolation, reducing storage overhead. Trilinear interpolation refers to an interpolation method that uses the known values ​​of the eight vertices surrounding the target point in the three-dimensional grid space to calculate the target point value through cubic linear weighting. It is suitable for the rapid calculation of the membership degree of non-grid points in the sparse tensor lookup table.

[0209] S206, obtain the state assessment result composed of 12-dimensional membership degree values ​​and parameter coupling coefficients, and construct the basic confidence of the state assessment result based on the state assessment result;

[0210] S207, combined with the parameter coupling coefficient, dynamically adjusts the basic confidence of the state assessment result to obtain the membership confidence of the state assessment result. Using the membership confidence of the state assessment result as a priori, the membership confidence is directly linearly mapped to the parameter dynamic weight.

[0211] S208, obtain the comprehensive feature vector and the water-air exchange correction vector, and construct the joint membership distribution based on the comprehensive feature vector and the water-air exchange correction vector;

[0212] S209. A bivariate risk association rule matrix is ​​established by combining the joint membership degree distribution and the dynamic weight of the parameters. The activation intensity of the bivariate risk association rule matrix in the multi-source inference rule table is calculated based on the fuzzy clustering algorithm. The activation intensity of the multi-source inference rule table is classified and superimposed according to the output risk level to obtain the fuzzy risk level.

[0213] S2010 uses the area center method to perform integral calculations on fuzzy risk levels and outputs the water body hypoxia risk level.

[0214] In this embodiment, when performing integral calculations on fuzzy risk levels using the area center method, firstly, the fuzzy risk levels output by the rule inference block are obtained. These fuzzy risk levels include the membership values ​​of three output levels: low risk, medium risk, and high risk. Secondly, for each output level, an output membership function is constructed based on its membership value and a preset level center value. This output membership function is a triangular or trapezoidal membership function, with its height corresponding to the membership value of the level, its base covering a preset numerical range for the level, and overlapping areas between the membership functions of adjacent levels to ensure the continuity of the defuzzification results. Then, the output membership functions of the three output levels are subjected to a point-by-point maximization operation in the same coordinate system to synthesize the overall output membership function. The upper envelope of the overall output membership function represents the comprehensive contribution distribution of all activated rules. The maximization operation ensures that the maximum membership contribution in any risk level direction is not weakened by other levels. Next, the total area of ​​the graph enclosed by the overall output membership function and the horizontal axis is calculated. Specifically, the overall output membership function is numerically integrated within the interval from the lower limit of the low-risk level to the upper limit of the high-risk level, with the integration step size preset according to the defuzzification accuracy requirements. Simultaneously, the first moment of the graph about the vertical axis is calculated. Specifically, the product of the overall output membership function and the horizontal axis coordinate value is numerically integrated within the same interval. Finally, the first moment is divided by the total area to obtain the horizontal coordinate of the center point of the area. This horizontal coordinate is the determined value of the water body hypoxia risk level after defuzzification. The determined value is in a continuous numerical interval between the center value of the low-risk level and the center value of the high-risk level. When the center point of the area is biased towards the center value of the low-risk level, a low-risk judgment is output; when it is biased towards the center value of the high-risk level, a high-risk judgment is output; and when it is in the middle region, a medium-risk judgment is output. The determined value can be directly used as an index for subsequent strategy matching and actuator driving of the dual-output control strategy generation unit 260.

[0215] In a further preferred embodiment of the present invention, such as Figure 7 As shown, the method for generating an adaptive dual-output control strategy based on rate of change prediction and inertia compensation mechanism includes:

[0216] S301: Obtain the current period's water body hypoxia risk level and the historical period's water body hypoxia risk level sequence, and use the backward difference method to calculate the risk change rate.

[0217] The backward difference method is implemented as follows: the difference between the current period's water hypoxia risk level and the previous period's water hypoxia risk level is divided by the control period duration to obtain an estimate of the risk change rate. The backward difference method relies only on known sampled values ​​from the current and historical periods, without requiring predicted values ​​for future times, thus conforming to the practical constraints of real-time control systems where only current and historical information is available.

[0218] S302, determine the predictive gain coefficient based on the current cycle risk level, and predict the predictive risk level of the next cycle based on the risk level change rate;

[0219] S303: Obtain the mechanical response parameters of the servo motor, and construct the mechanical hysteresis compensation based on the mechanical response parameters of the servo motor; use the predicted risk level of the next cycle as the set value, the water hypoxia risk level of the current cycle as the feedback value, and the difference between the two as the control error;

[0220] S304 constructs a multi-objective fitness function based on control error, mechanical hysteresis compensation, and the number of times the air pump starts and stops. Iterative optimization of the multi-objective fitness function is achieved using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm introduces an adaptive inertial weight mechanism and a differential evolution mutation operator to enhance particle diversity. When the number of iterations reaches 30 generations or the improvement of the population's optimal fitness value is less than 1% for 10 consecutive generations, the iteration is terminated and the optimal PID parameter set corresponding to the population's historical optimal position is output.

[0221] The multi-objective fitness function is expressed as:

[0222]

[0223]

[0224] in, These represent the control error item, the servo wear item, and the air pump start / stop item, respectively. These represent the initial weights of the control error term, the servo wear term, and the air pump start / stop term, respectively. Indicates the evaluation time window;

[0225] S305 uses the optimal PID parameter set as the adaptive dual-output control strategy, converts the adaptive dual-output control strategy into adaptive dual-output control commands, and outputs them to the air pump drive module 300 and the air volume regulation module 400 respectively.

[0226] In this embodiment of the invention, when generating an adaptive dual-output control strategy based on rate of change prediction and inertial compensation mechanisms, the backward difference method is used to extract the changing trend from the historical risk sequence. The interval-differential configuration of the predicted gain coefficient achieves a sensitive response and smooth transition to the risk evolution trend. The pre-estimation of mechanical hysteresis compensation offsets the servo motor execution delay. A multi-objective fitness function coordinates the diverse demands of control accuracy, smoothness, and equipment lifespan. An improved particle swarm optimization algorithm with adaptive inertial weights and differential evolution mutation achieves global optimal optimization of PID parameters. Finally, the optimal PID parameter set is synchronously configured in both the air pump and valve loops. This achieves precise linkage between air pump start / stop and valve opening, and smooth, graded adjustment of oxygen release, effectively avoiding the frequent start / stop and coarse adjustment problems of traditional threshold control. It also extends the mechanical lifespan of relays and servos, and improves the comfort and energy efficiency of oxygenation control in home aquarium scenarios.

[0227] In summary, this invention provides an intelligent water aerator. By combining multi-source water state parameter acquisition and joint filtering, pre-feature coupling extraction and evidence fusion, three-dimensional sparse tensor membership mapping, gated attention dynamic confidence correction, bivariate risk association rule reasoning, and online optimization of PID parameters through rate of change prediction and inertia compensation, this invention achieves accurate assessment of oxygen deficiency risk in home aquarium scenarios and refined, graded oxygen release control of the air pump and valve linkage without relying on high-cost real dissolved oxygen sensors. Furthermore, through a lightweight collaborative architecture combining a fuzzy logic risk assessment model and an improved particle swarm optimization PID parameter optimization, it overcomes the computational latency problem of existing complex algorithms on low-cost embedded devices, achieving overall technical effects such as low detection cost, fine control granularity, simple system structure, and strong adaptability to home scenarios.

[0228] It should be noted that, for the sake of simplicity, the foregoing embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0229] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. An intelligent water aeration pump, characterized in that, The intelligent water oxygenation pump includes: The detection module is used to collect multi-source water state parameters in the aquarium, preprocess the multi-source water state parameters, and output the preprocessed multi-source parameter set. The main control module communicates with the detection module and is used to receive the pre-processed multi-source parameter set, identify and analyze the multi-source parameter set based on the pre-built fuzzy logic risk assessment model, output the state assessment result, use the state assessment result as an index to retrieve the preset multi-source inference rule table, output the water body hypoxia risk level, generate an adaptive dual-output control strategy based on the water body hypoxia risk level, and generate an adaptive dual-output control command based on the adaptive dual-output control strategy. The air pump drive module responds to the adaptive dual-output control command, synchronously outputs the air pump control signal, and executes the air pump start-up and stop actions; The air volume regulation module responds to the adaptive dual-output control command, synchronously outputs the air valve opening control signal, and executes the air valve opening regulation action; The wireless communication module is connected to the main control module via a serial port and is used to upload the status data of the main control module to the user terminal and receive control commands from the user terminal. The user terminal is used to display the water condition assessment results, the water hypoxia risk level, the air pump status and air valve opening, and to receive control commands and user-set thresholds.

2. The intelligent water aeration pump as described in claim 1, characterized in that: The detection module includes: The water temperature detection unit includes a thermistor temperature sensor, which is placed in the aquarium water and collects aquarium water temperature data based on a preset sampling period; The timing unit, integrated inside the main control module, is used to accumulate the timing duration when the air pump is off, reset it to zero when the air pump is on, and output the accumulated non-running time of the air pump. The water circulation status detection unit is used to collect data on water flow and exchange capacity, and output water circulation status data. The simulated oxygen content trend input unit includes a general analog input interface, which is configured as an ADC input channel. A potentiometer is connected to generate an analog voltage signal corresponding to the preset virtual dissolved oxygen state. This signal is used to verify the hypoxia risk assessment algorithm and the control logic of the air pump and air valve, and to output simulated oxygen content data. The preprocessing unit is used to acquire water temperature data, cumulative inactivity time of air pump, water circulation status data and simulated oxygen content data in real time and integrate them into multi-source water state parameters. It preprocesses the multi-source water state parameters and outputs the preprocessed multi-source parameter set.

3. The intelligent water aeration pump as described in claim 2, characterized in that: The main control module includes: The model building unit is used to pre-build a fuzzy logic risk assessment model. It uses a reference-level dissolved oxygen detection device to collect the actual dissolved oxygen concentration corresponding to the time of the multi-source water body state parameters, and generates hypoxia risk labels based on the actual dissolved oxygen concentration. The user-defined threshold is used as the control preference parameter, and together with the multi-source water body state parameters, it is used as the model input. The hypoxia risk label is used as the supervision target to optimize the fuzzy membership parameters and rule weights, and outputs a converged fuzzy logic risk assessment model. The fuzzy logic risk assessment unit is used to receive the preprocessed multi-source parameter set and user-defined thresholds, perform membership division on the parameters in the multi-source parameter set based on the pre-built fuzzy logic risk assessment model, and output the state assessment results. The dynamic confidence correction unit is used to traverse the multi-source parameter set, dynamically adjust the membership confidence of the state evaluation result based on the parameter coupling coefficient of the multi-source parameter set, and determine the dynamic weight of the parameter by combining the membership confidence. The rule reasoning unit communicates with the fuzzy logic risk assessment unit, and combines the dynamic weights of parameters, the state assessment results and the multi-source reasoning rule table matching analysis to output the fuzzy risk level. The fuzzy risk level is defuzzified and the area center method is used to perform integral calculation on the fuzzy risk level to output the water body hypoxia risk level. A dual-output control strategy generation unit is communicatively connected to the defuzzification unit. Using the water body hypoxia risk level as an index, it generates an adaptive dual-output control strategy based on rate of change prediction and inertial compensation mechanism. The adaptive dual-output control strategy includes a pump control strategy and a valve opening control strategy corresponding to the water body hypoxia risk level. The control command output unit is communicatively connected to the dual-output control strategy generation unit and is used to convert the adaptive dual-output control strategy into adaptive dual-output control commands, which are then output to the air pump drive module and the air volume regulation module, respectively.

4. The intelligent water aeration pump as described in claim 1, characterized in that: The user terminal includes: The status display unit is used to display the current water body oxygen deficiency risk level, air pump operating status, air valve opening degree, real-time water temperature, cumulative air pump inactivity time, and abnormal reminder information; The mode switching unit is used to switch between automatic mode and manual mode. In automatic mode, it operates autonomously based on an adaptive dual-output control strategy. In manual mode, user-input control commands are executed first. The manual control unit is used to receive commands from the user to manually turn the air pump on or off and adjust the opening of the air valve. The threshold parameter setting unit is used to receive the user's input oxygenation preference settings and the user-set threshold. The oxygenation preference settings include energy-saving mode, normal mode or strong oxygenation mode. The safety priority determination unit is used to monitor abnormal operating conditions in manual mode and generate safety protection commands when abnormal operating conditions are detected.

5. The intelligent water aeration pump as described in claim 2, characterized in that: Methods for preprocessing state parameters of multi-source water bodies include: Acquire multi-source water body state parameters, and mark the collection timestamps for water temperature data, cumulative inactivity time of air pump, water circulation state data and simulated oxygen state data respectively. Use a preset reference period as the time window to align the parameters on the time axis, and group the parameter data collected within the same reference period into the same parameter frame. Parameter data that were not collected within the reference period are filled into the corresponding parameter frame by nearest neighbor interpolation. Output the time-domain aligned multi-source water body state parameters. Load the time-domain aligned multi-source water body state parameters, perform sliding window anomaly detection on the time-domain synchronized multi-source water body state parameters, identify data points that deviate from the mean by more than three times the standard deviation as outliers, replace the outliers with the median of the window where the outlier is located, and output the multi-source water body state parameters after outlier removal. Obtain the state parameters of the multi-source water bodies after outlier removal, perform joint filtering on the state parameters of the multi-source water bodies, and output the filtered state parameters of the multi-source water bodies. The filtered multi-source water state parameters are mapped to a unified numerical range, and the mapped parameter data constitute the preprocessed multi-source parameter set.

6. The intelligent water aeration pump as described in claim 5, characterized in that: Methods for joint filtering of multi-source water state parameters include: Obtain the state parameters of multi-source water bodies after outlier removal, determine the parameter coupling coefficients based on the weights and mutual information of the multi-source water body state parameters, and establish a parameter correlation matrix based on the parameter coupling coefficients. Using the parameter correlation matrix as a constraint, the state parameters of multiple water bodies are collaboratively corrected, and the coupling deviation between the state parameters of multiple water bodies and the associated state parameters is calculated. Determine whether the coupling deviation between the state parameters of multiple water sources and the associated state parameters exceeds a preset deviation threshold; If the coupling deviation between the multi-source water body state parameters and the associated state parameters exceeds the preset deviation threshold, the multi-source water body state parameters are constrained and corrected by the weighted combination of the associated state parameters, and the multi-source water body state parameters after parameter correlation constraint filtering are output. The state parameters of the multi-source water bodies after parameter correlation constraint filtering are obtained. The dynamic window length of the sliding window is determined based on the parameter dynamic response. The state parameters of the multi-source water bodies are decomposed into multi-scale wavelet packets in combination with the dynamic window length, and the state parameters of the multi-source water bodies after multi-scale decomposition and filtering are output. Based on the multi-source water body state parameters after multi-scale decomposition filtering, a three-dimensional state vector containing dissolved oxygen concentration, water temperature, and cumulative inactivity time of air pumps is constructed. With water temperature as the known input, dissolved oxygen concentration as the state to be estimated, and cumulative inactivity time of air pumps as the control variable, a nonlinear state space model is established. Combining the nonlinear state space model, a volumetric Kalman filter is performed to output the multi-source water body state parameters after time-consistent filtering.

7. The intelligent water aeration pump as described in claim 3, characterized in that: The fuzzy logic risk assessment model includes an input layer, a risk assessment architecture, and an output layer. The risk assessment architecture includes a pre-feature extraction layer and a risk association layer. The risk association layer includes a membership mapping block, a dynamic confidence correction block, a rule reasoning block, and a defuzzification layer. The defuzzification layer uses the area center method to perform integral calculations on the fuzzy risk level and outputs the water body hypoxia risk level.

8. The intelligent water aeration pump as described in claim 7, characterized in that: Methods for classifying the membership of parameters in a multi-source parameter set based on a pre-built fuzzy logic risk assessment model include: Obtain a multi-source parameter set, encode and reduce the dimensionality of the multi-source parameter set, and output the encoded four-dimensional parameter vector; The water temperature data and the cumulative inactivity time of the air pump are nonlinearly coupled and calculated. Based on the temperature correction coefficient of dissolved oxygen saturation concentration and the cumulative effect coefficient of the cumulative inactivity time of the air pump, the environmental stress index is calculated. The water circulation state data and the simulated oxygen state data are dynamically coupled and calculated. The air-water interface update rate is calculated based on the water circulation state data, and the water-air exchange efficiency index is calculated by combining the change slope of the simulated oxygen state data. The environmental stress index and water-air exchange efficiency index are obtained and used to resolve evidence conflicts and perform weighted fusion on the environmental stress index and water-air exchange efficiency index, and output the corrected comprehensive feature vector and water-air exchange correction vector. Obtain the corrected integrated feature vector and water-air exchange correction vector, construct a three-dimensional sparse tensor lookup table based on the integrated feature vector and water-air exchange correction vector, traverse the preset parameter-level membership mapping relationship based on the three-dimensional sparse tensor lookup table, calculate the interpolated membership degree between the integrated feature vector and the parameter-level membership degree using the trilinear interpolation method, and output the state evaluation result composed of 12-dimensional membership degree values. Obtain the state assessment results composed of 12-dimensional membership values ​​and parameter coupling coefficients, and construct the basic confidence of the state assessment results based on the state assessment results; By dynamically adjusting the basic confidence level of the state assessment results using the parameter coupling coefficient, the membership confidence level of the state assessment results is obtained. Using the membership confidence level of the state assessment results as a priori, the membership confidence level is directly linearly mapped to the dynamic weights of the parameters.

9. The intelligent water aeration pump as described in claim 8, characterized in that: The method for classifying the membership of parameters in a multi-source parameter set based on a pre-built fuzzy logic risk assessment model also includes: Obtain the comprehensive feature vector and the water-air exchange correction vector, and construct a joint membership distribution based on the comprehensive feature vector and the water-air exchange correction vector; A bivariate risk association rule matrix is ​​established by combining the joint membership degree distribution and the dynamic weight of the parameters. The activation intensity of the bivariate risk association rule matrix in the multi-source inference rule table is calculated based on the fuzzy clustering algorithm. The activation intensity of the multi-source inference rule table is classified and superimposed according to the output risk level to obtain the fuzzy risk level. The area center method is used to perform integral calculations on the fuzzy risk level and output the water body hypoxia risk level.

10. The intelligent water aeration pump as described in claim 9, characterized in that: Methods for generating adaptive dual-output control strategies based on rate-of-change prediction and inertia compensation mechanisms include: Obtain the current period's water body hypoxia risk level and the historical period's water body hypoxia risk level sequence, and calculate the risk change rate using the backward difference method; The predicted gain coefficient is determined based on the current cycle risk level, and the predicted risk level for the next cycle is predicted based on the rate of change of the risk level. Obtain the mechanical response parameters of the servo motor, and construct the mechanical hysteresis compensation based on the mechanical response parameters of the servo motor; use the predicted risk level of the next cycle as the set value, the water hypoxia risk level of the current cycle as the feedback value, and the difference between the two as the control error; A multi-objective fitness function is constructed based on the control error, mechanical lag compensation, and the number of times the air pump starts and stops. An improved particle swarm algorithm is used to iteratively optimize the multi-objective fitness function. When the number of iterations reaches 30 generations or the improvement of the population's optimal fitness value is less than 1% for 10 consecutive generations, the iteration is terminated and the optimal PID parameter set corresponding to the population's historical optimal position is output. The optimal PID parameter set is used as the adaptive dual-output control strategy. The adaptive dual-output control strategy is converted into adaptive dual-output control commands, which are output to the air pump drive module and the air volume regulation module respectively.

Images