Water quality ecology automatic regulation control method and system for intelligent fish tank

Abstract

The invention relates to a water quality ecology automatic regulation and control method for an intelligent fish tank, and belongs to the technical field of automatic intelligent culture. The method comprises the following steps: collecting multi-dimensional state data of a fish tank water body in real time based on a multi-parameter unified control framework, and constructing a comprehensive water quality data set; differential automatic analysis processing is carried out on the data set, subsets are divided according to water quality parameter types, regulation stages and equipment health features, and adaptive regulation compensation factors are generated; inputting the subset and the compensation factor into a water quality prediction algorithm model for multi-scale feature extraction and cross-modal enhancement, and outputting a water quality parameter prediction value and a quality grade; a multi-dimensional attribution analysis engine is fused with multi-dimensional features to position a water quality abnormity main cause, and an equipment control compensation strategy is generated; after the strategy is converted into executor instruction execution, an improvement effect is collected and fed back to a differentiation processing link to correct subset division and compensation factors, and closed-loop optimization is formed. The regulation and control precision and the equipment energy efficiency of the intelligent fish tank are effectively improved.

Description

Technical Field

[0001] This invention relates to the technical field of automated intelligent aquaculture, and in particular to a method and system for automatic water quality ecological regulation and control in intelligent fish tanks. Background Technology

[0002] Currently, smart aquariums have been promoted in home viewing, commercial display, education and research scenarios to meet corresponding breeding needs. However, most existing smart aquarium systems still rely on a timed and quantitative extensive control mode, which makes it difficult to capture dynamic changes in water quality in real time. Furthermore, they lack targeted control methods for the differentiated needs of different fish species and breeding densities, often leading to excessive water changes, insufficient aeration, or equipment idling and wasting resources, thus reducing the effective utilization rate of smart aquarium resources. Therefore, improvements are needed. Summary of the Invention

[0003] To improve the resource utilization efficiency of smart fish tanks, this application provides a method and system for automatic water quality ecological regulation and control in smart fish tanks.

[0004] Firstly, the above-mentioned inventive objective of this application is achieved through the following technical solution: A method for automatic water quality ecological regulation and control in intelligent fish tanks, the method comprising the following steps: The basic data collected by environmental sensors is entered into the database, and the control framework of the integrated intelligent aquarium breeding equipment is established through artificial intelligence algorithms. Real-time collection of multi-dimensional environmental status data of aquarium water, and linkage of the multi-dimensional environmental status data with the control framework of intelligent aquaculture equipment to construct a comprehensive water quality dataset; The comprehensive water quality dataset is subjected to differential automatic analysis and processing. Based on the characteristics of water quality parameter types, control stage characteristics and equipment health characteristics, the dataset is divided into multiple differential subsets. For each subset, an appropriate water quality control compensation factor is generated independently. The differentiated subset and the corresponding water quality regulation compensation factor are input into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The predicted values ​​of water quality parameters are input into a pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates equipment control compensation strategy. The equipment control compensation strategy is converted into actuator control commands and executed. After execution, water body state data is re-collected to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

[0005] By adopting the above technical solutions, this application establishes a unified framework to eliminate parameter confusion, achieves three-dimensional perception through multi-dimensional data fusion, accurately matches fish species needs through differentiated processing, predicts and assesses risks in advance, generates economic strategies by directly addressing the root causes through multi-dimensional attribution, and achieves self-evolution through closed-loop feedback. The constructed closed-loop system of "parameter framework - data collection - differentiated processing - prediction and assessment - attribution compensation - feedback evolution" systematically solves the core pain points of traditional aquarium water quality management, which relies on manual experience, has a delayed response, and is extensive in regulation. It promotes the transformation of aquarium management from manual and extensive to intelligent and refined, and provides a replicable digital solution for ornamental fish farming. Compared with the traditional aquarium water quality management model that relies on manual experience, single sensor detection, and passive post-event regulation, this application systematically solves the core pain points of traditional methods, such as delayed response, extensive regulation, and inability to adapt, through the intelligent system it has constructed. It improves the utilization efficiency of aquarium equipment while increasing the survival rate of fish.

[0006] In a preferred embodiment, this application can be further configured as follows: In the step of real-time acquisition of multi-dimensional environmental state data of the aquarium water, and linking the multi-dimensional environmental state data with the control framework of the intelligent aquaculture equipment to construct a comprehensive water quality dataset, a multi-source data preprocessing mechanism is added, including the following steps: A multimodal state vector is constructed from water body video images, underwater sonar echoes, and physical sensor signals synchronously acquired through edge nodes; By utilizing a preset synchronous triggering mechanism, the multimodal state vector is aligned with the common time axis at the nanosecond level, eliminating the sampling delay differences between multiple sensors; The aligned data stream is then subjected to sliding window filtering and outlier removal to generate a preprocessed comprehensive water quality dataset.

[0007] In a preferred embodiment, this application can be further configured as follows: The step of performing differential automatic analysis on the comprehensive water quality dataset, dividing the dataset into multiple differential subsets based on water quality parameter type characteristics, regulation stage characteristics, and equipment health characteristics, and independently generating an appropriate water quality regulation compensation factor for each subset includes the following steps: Based on the water transparency characteristics, the comprehensive water quality dataset is divided into transparent water subset and non-transparent water subset. An illumination compensation factor is generated for the transparent water subset to suppress reflective interference, and a turbidity enhancement factor is generated for the non-transparent water subset to highlight suspended solids defects. Based on the characteristics of the feeding stage, the circulation stage, and the purification stage in the regulation phase, the dataset is divided into stage subsets. Feeding compensation factors are generated for the feeding stage subset to predict residual feed pollution, and filtration compensation factors are generated for the purification stage subset to predict filter media clogging. Based on the wear and tear characteristics of the equipment health status, the dataset is divided into a healthy equipment subset and an aged equipment subset. A performance degradation compensation factor is generated for the aged equipment subset to correct the detection error caused by equipment passivation.

[0008] In a preferred example, this application can be further configured to add an adaptive adjustment mechanism for root cause localization weights to the multidimensional attribution analysis engine, as follows: Construct a health status degradation characteristic curve for equipment, and use the time-series change rate of filter media clogging degree and water pump fatigue degree as health status characteristics. When the health characteristics show an accelerated degradation trend, increase the attribution weight of the main cause of equipment aging. Extract the deviation characteristics of water parameters, calculate the deviation ratio between the actual pH and temperature and the standard parameters, and trigger the equipment control compensation strategy to be executed first when the deviation exceeds the preset threshold; Identify environmental topological bottlenecks, calculate the spatial matching degree between the feeding location and the filtration system, and when the matching degree is lower than the preset threshold, increase the weight of the main cause of structural water quality deterioration and generate equipment layout optimization suggestions.

[0009] In a preferred example, this application can be further configured as follows: after the step of converting the equipment control compensation strategy into an actuator control command for execution, a self-cleaning mechanism for detection labels is added, including the following steps: comparing the consistency between the manual re-inspection results and the model prediction results to identify abnormal detection samples that are misjudged by the model or missed by the manual. The detection labels of abnormal samples are downgraded in confidence, and cross-modal feature similarity analysis is used to find highly similar samples from nearby time periods for label reconstruction. The reconstructed purification labels are re-injected into the training set, and the water quality prediction algorithm model is incrementally fine-tuned to achieve automatic correction of label errors and continuous improvement of model robustness.

[0010] In a preferred embodiment, this application can be further configured as follows: after the step of converting the equipment control compensation strategy into actuator control commands for execution, an adaptive adjustment closed-loop optimization mechanism is added, including the steps of: calculating the residual between the water quality improvement effect after execution and the expected target value, and constructing an effect deviation evaluation vector. Based on the aforementioned deviation assessment vector, attribution analysis was performed to identify key influencing factors leading to the deviation as improper compensation factor settings or outdated subset partitioning criteria. Incremental learning algorithms are used to dynamically adjust the subset partitioning boundary conditions in the differentiation process and to automatically correct the weight coefficients in the compensation factor generation logic. The revised subset partitioning criteria and compensation factors are fed back to enable adaptive optimization of the next generation of detection and control parameters.

[0011] In a preferred embodiment, the present application may be further configured such that the equipment control compensation strategy includes an automatic feeding compensation strategy, an automatic oxygenation compensation strategy, an automatic circulation compensation strategy, an automatic temperature control compensation strategy, an automatic sterilization compensation strategy, and an automatic water change compensation strategy.

[0012] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions: An automatic water quality ecological regulation and control device for intelligent fish tanks, the device comprising: A multi-parameter unified control framework building unit is used to establish a multi-parameter unified control framework for aquariums based on pre-set basic parameter input rules. The comprehensive water quality dataset construction unit is used to collect multi-dimensional environmental status data of aquarium water in real time, and link the multi-dimensional environmental status data with the control framework of intelligent aquaculture equipment to construct a comprehensive water quality dataset. The water quality regulation compensation factor generation unit is used to perform differential automatic analysis and processing on the comprehensive water quality dataset. Based on the characteristics of water quality parameter types, regulation stage characteristics and equipment health characteristics, the dataset is divided into multiple differential subsets, and an appropriate water quality regulation compensation factor is generated independently for each subset. The water quality prediction and water body assessment unit is used to input the differential subset and the corresponding water quality regulation compensation factor into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data, and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The equipment control compensation strategy generation unit is used to input the predicted values ​​of water quality parameters into a pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates an equipment control compensation strategy. The execution optimization unit is used to convert the equipment control compensation strategy into actuator control commands for execution, and after execution, re-collect water body state data to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

[0013] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions: An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described automatic water quality ecological regulation and control method for intelligent fish tanks.

[0014] Fourthly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described automatic water quality ecological regulation and control method for intelligent fish tanks. Attached Figure Description

[0015] Figure 1 This is a flowchart of an automatic water quality ecological regulation and control method for a smart fish tank according to an embodiment of this application; Figure 2 This is a schematic diagram of a water quality ecological automatic regulation and control device for an intelligent fish tank according to one embodiment of this application; Figure 3 This is a schematic diagram of an electronic device according to an embodiment of this application.

[0016] Icon labels: 1. Multi-parameter unified control framework construction unit; 2. Comprehensive water quality dataset construction unit; 3. Water quality regulation compensation factor generation unit; 4. Water quality prediction and water body assessment unit; 5. Equipment control compensation strategy generation unit; 6. Execution optimization unit. Detailed Implementation

[0017] The present application will be further described in detail below with reference to the accompanying drawings.

[0018] In one embodiment, taking a medium-sized tropical ornamental fish smart aquarium scenario as an example, such as... Figure 1 As shown, this application discloses an automatic water quality ecological regulation and control method for intelligent fish tanks, which specifically includes the following steps: S10: Based on the basic data collected by environmental sensors, the database is entered, and the control framework of the integrated intelligent aquarium breeding equipment is established through artificial intelligence algorithms; Specifically, the system establishes a unified control framework based on pre-set basic parameter input rules: water pH is controlled within the slightly acidic range of 6.0-7.0, temperature is stable at 28-30℃, dissolved oxygen concentration is not lower than 5mg / L, light intensity simulates the natural day-night cycle (8 hours of light per day), and water circulation flow is maintained at 4-5 times per hour. The framework requires automatic feeding periods from 8:00-9:00 and 18:00-19:00 daily, and a low-power silent mode from 22:00-6:00 at night. This standardized framework provides a unified benchmark for subsequent multi-source data alignment, avoiding regulatory disorder caused by parameter conflicts when different fish species are mixed. The multi-parameters include water pH, temperature, dissolved oxygen concentration, light intensity, and water circulation flow. In this embodiment, the control framework has the following effects: Regarding the processing of multi-dimensional environmental state data, for example, after the temperature data of the aquarium water is collected by the server, it is processed by data preprocessing algorithm, feature extraction algorithm and state recognition and classification algorithm. When the temperature value is lower than a certain standard, an instruction is sent to the heating device to heat the water.

[0019] S20: Real-time collection of multi-dimensional environmental status data of the aquarium water, linking the multi-dimensional environmental status data with the control framework of the intelligent aquaculture equipment to construct a comprehensive water quality dataset; Specifically, the system collects data in real time using multiple types of sensors deployed in the aquarium: a pH probe collects pH values ​​every 5 minutes, an infrared temperature sensor monitors the temperature at multiple points in the water, a dissolved oxygen meter detects the dissolved oxygen concentration in the water, a light sensor records the actual light intensity of the LED lights, and a flow meter measures the flow rate of the circulating water pump. Simultaneously, it collects equipment operating data such as heater power, aerator speed, and filter media usage time. Using the RTC clock as a common time axis, the system aligns the data of the instantaneous pH drop of 0.3 at 8:15 AM feeding time with the feeding action marker, and binds the data of the temperature naturally dropping by 0.5°C after the lights are turned off at 10:00 PM with a timestamp, constructing a comprehensive water quality dataset containing three-dimensional attributes of water quality, equipment, and time, thus eliminating data asynchrony issues caused by sensor sampling delays.

[0020] S30: Perform differential automatic analysis processing on the comprehensive water quality dataset, divide the dataset into multiple differential subsets according to the characteristics of water quality parameter types, control stage characteristics and equipment health characteristics, and independently generate suitable water quality control compensation factors for each subset. Specifically, the system performs differentiated processing on the comprehensive water quality dataset: based on the type of water quality parameters, toxin indicators such as ammonia nitrogen and nitrite are classified as high-sensitivity subsets, generating an aeration compensation factor (increasing the speed of the aerator pump by 20%); based on the control stage, the area within 2 hours after feeding is classified as a high-risk subset for residual feed, generating a water change compensation factor (activating bottom suction water change by 15%); based on the equipment health status, filter media used for more than 6 months is classified as an aging subset, generating a filtration efficiency compensation factor (extending the water circulation time by 10%). By generating appropriate compensation factors independently for different subsets, a "one-size-fits-all" approach to control can lead to excessive water changes or insufficient aeration, improving the accuracy of control and energy-saving effects.

[0021] S40: Input the differentiated subset and the corresponding water quality regulation compensation factor into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. Specifically, the predicted water quality values ​​are input into the attribution analysis engine, integrating water health characteristics (filter media clogging rate reaches 70%), equipment operational deviation characteristics (aerator pump speed is 10% below rated value), and environmental topology characteristics (feeding location is far from the inlet, leading to uneaten feed accumulation). The primary cause is identified as "filter media aging + localized insufficient aeration." Equipment control compensation strategies are generated: Priority 1 activates the backup aerator pump; Priority 2 increases the bottom suction water exchange frequency to once every 2 hours; Priority 3 marks the feeding location as a "high-risk point" in the CIM topology model and pushes a suggestion to the user's app to adjust the feeding point. This strategy directly addresses the primary cause rather than the superficial symptoms, avoiding the waste of water resources caused by blindly increasing the overall water exchange volume, and improving the economic efficiency of regulation by 50%.

[0022] It should be noted that the specific values ​​in the embodiments of this application are all experimental records of this embodiment, and may be other data in other embodiments.

[0023] S50: Input the predicted water quality parameters into the pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates equipment control compensation strategy. The multi-dimensional feature data includes health features, equipment operation deviation features, and environmental topology features. The equipment control compensation strategies include automatic feeding compensation strategy, automatic oxygenation compensation strategy, automatic circulation compensation strategy, automatic temperature control compensation strategy, automatic sterilization compensation strategy, and automatic water change compensation strategy. Specifically, the system translates the compensation strategy into control commands: the backup aerator is started, and the bottom suction water exchange valve is opened at 15% flow rate for 10 minutes. After execution, water status data is re-collected, and the ammonia nitrogen concentration has significantly decreased, improving the water quality assessment value to "good". The improvement effect assessment results are fed back to the S30 differentiated treatment stage: the system identifies that the "filter media clogging degree" feature has the highest contribution to this improvement, thus increasing the weight coefficient of this feature in the subset partitioning, and automatically shortening the filter media replacement cycle in the future. This closed loop allows the system to continuously learn the ecological characteristics of this aquarium (in this embodiment, discus fish are extremely sensitive to ammonia nitrogen), and subsequent regulation focuses more on high-value features, adapting to changes in fish density without manual recalibration, achieving adaptive self-optimization of water quality management.

[0024] S60: The equipment control compensation strategy is converted into actuator control commands and executed. After execution, water body state data is re-collected to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

[0025] In summary, this application establishes a unified framework to eliminate parameter confusion, achieves three-dimensional perception through multi-dimensional data fusion, accurately matches fish species needs through differentiated processing, and predicts and assesses risks in advance. The S50 multi-dimensional attribution directly addresses the root cause to generate economic strategies, and the S60 closed-loop feedback enables self-evolution. The constructed closed-loop system of "parameter framework - data acquisition - differentiated processing - prediction and assessment - attribution compensation - feedback evolution" systematically solves the core pain points of traditional aquarium water quality management, which relies on manual experience, has delayed response, and is characterized by extensive control. It promotes the transformation of aquarium management from manual and extensive to intelligent and refined, providing a replicable digital solution for ornamental fish farming. Compared to the traditional aquarium water quality management model that relies on manual experience, single sensor detection, and reactive control, this application, through its constructed intelligent system, systematically solves the core pain points of traditional methods, such as delayed response, extensive control, and inability to adapt, improving the utilization efficiency of aquarium equipment while increasing fish survival rates.

[0026] This application establishes a unified multi-parameter control framework to achieve real-time synchronous monitoring of multiple dimensions of water quality, such as pH, temperature, and dissolved oxygen, significantly improving the comprehensiveness and timeliness of water quality perception. The traditional "one-size-fits-all" approach of timed and quantitative regulation is difficult to adapt to the water quality change patterns at different stages such as feeding, purification, and circulation. This application uses a differentiated treatment mechanism to automatically generate appropriate compensation factors based on the type of water quality parameters, the regulation stage, and the health status of the equipment, realizing the transformation from "extensive regulation" to "precise policy implementation," effectively improving regulation accuracy and reducing equipment energy consumption.

[0027] Traditional management often only detects water quality deterioration after fish exhibit stress responses. This application uses a water quality prediction algorithm model to provide early warnings of risks and combines it with a multidimensional attribution analysis engine to accurately pinpoint root causes such as filter media aging and insufficient aeration. This shifts the management approach from "post-event remediation" to "pre-event prevention," significantly improving water quality stability and aquaculture success rates.

[0028] This application uses a closed-loop feedback mechanism to automatically learn the ecological characteristics of the aquarium, continuously optimize control parameters, reduce reliance on human experience, and promote the transformation of aquarium management from manual operation to an intelligent and adaptive digital model, providing a practical and evolving intelligent solution for ornamental fish farming.

[0029] In step S20: Real-time acquisition of multi-dimensional environmental status data of the aquarium water, linking the multi-dimensional environmental status data with the control framework of the intelligent aquaculture equipment to construct a comprehensive water quality dataset, a multi-source data preprocessing mechanism is added, including the following steps: S21: Construct a multimodal state vector from water video images, underwater sonar echoes, and physical sensor signals synchronously acquired through edge nodes; In the aquarium of this embodiment, three heterogeneous sensors are simultaneously deployed on the edge computing node: a high-definition camera on the water surface captures visual features such as oil film and foam in real time; an underwater sonar probe emits high-frequency sound waves and receives echoes to detect the density and distribution of fish; and a physical sensor array continuously monitors water quality parameters such as pH, ammonia nitrogen, and nitrite. During the 10:00 feeding time, the system integrates the image frames captured by the camera showing fish surging towards the feeding point, the peak intensity of underwater targets displayed by the sonar echoes, and the instantaneous increase in ammonia nitrogen detected by the chemical sensor into a multimodal state vector containing visual, acoustic, and chemical information. This vector overcomes the limitations of traditional single-dimensional reliance on chemical probes, enabling subsequent water quality analysis to possess a three-dimensional perception capability of "seeing the surface, listening to the underwater environment, and measuring chemistry," significantly improving the accuracy of identifying complex conditions such as uneaten food pollution and fish stress.

[0030] S22: Using a preset synchronous triggering mechanism, the multimodal state vector is aligned with the common time axis at the nanosecond level to eliminate the sampling delay difference of multiple sensors; Specifically, the original sampling frequencies and delays of the three types of sensors are different: 30fps for the camera (approximately 33ms interval), 200kHz for the sonar (5μs interval), and 1Hz for the chemical sensor (1s interval). Using a preset synchronization triggering mechanism, the system timestamps the three types of data based on the PPS pulse of the RTC clock: forcibly aligning the image frame, sonar echo peak value, and ammonia nitrogen reading at 10:00:00.000 to the same timestamp. This alignment eliminates data time misalignment caused by sampling delay differences (e.g., the chemical sensor reading actually lags behind the image by 2 seconds), ensuring that the data in each dimension of the multimodal vector reflects the same instantaneous state, avoiding causal misjudgments caused by traditional asynchronous sampling (e.g., mistakenly associating fish activity 2 seconds ago with the current increase in ammonia nitrogen), and providing a time reference guarantee for accurate attribution.

[0031] S23: Perform sliding window filtering and abnormal pulse removal on the aligned data stream to generate a pre-processed comprehensive water quality dataset; Specifically, the aligned data stream contains noise interference: camera reflections and flickering due to water surface ripples, sonar false echo peaks due to bubble reflections, and occasional jumps in chemical sensors due to electromagnetic interference. The system uses a sliding window filter (10 sampling points wide) to smooth the multimodal data, removing high-frequency noise that deviates from the historical average by more than a threshold; it also identifies and removes instantaneous abnormal pulses caused by equipment start-up and shutdown (such as current spikes during pump startup interfering with sensors). The signal-to-noise ratio of the generated comprehensive water quality dataset after filtering is significantly improved, false echoes are reduced, and ammonia nitrogen reading jumps are eliminated, avoiding noise-triggered false alarms and ensuring the input reliability of subsequent S30 differential processing, making the system more sensitive and stable in responding to real water quality changes.

[0032] This application systematically addresses the core pain points of traditional aquarium water quality monitoring, such as reliance on a single chemical probe, sensor time delay asynchrony, and significant noise interference, through a multimodal data fusion and preprocessing mechanism constructed by S21-S23. S21's multimodal vector enables three-dimensional perception, S22's nanosecond-level alignment ensures accurate causal determination, and S23's filtering and elimination guarantee data reliability. These three elements provide a high-fidelity data foundation for subsequent S40 water quality prediction and S50 attribution analysis, improving the intelligent aquarium's ability to identify uneaten food pollution, fish stress, and equipment malfunctions from minutes to seconds, significantly reducing false alarm rates, and driving the intelligent paradigm transformation of water quality management from "chemical indicator monitoring" to "multimodal ecological perception."

[0033] In step S30: performing differential automatic analysis on the comprehensive water quality dataset, dividing the dataset into multiple differential subsets based on water quality parameter type characteristics, regulation stage characteristics, and equipment health characteristics, and independently generating suitable water quality regulation compensation factors for each subset, the steps include the following: S31: Based on the water transparency characteristics, the comprehensive water quality dataset is divided into transparent water subset and non-transparent water subset. An illumination compensation factor is generated for the transparent water subset to suppress reflective interference, and a turbidity enhancement factor is generated for the non-transparent water subset to highlight suspended solids defects. In this embodiment, the system divides the comprehensive water quality dataset into two subsets based on water transparency characteristics: when the water is clear and has good transparency, it is classified as a transparent water subset. In this case, the images captured by the camera are easily affected by water surface reflection and background interference. The system generates an illumination compensation factor for this subset, automatically adjusts the LED illumination angle, and enables a polarized light filtering algorithm to effectively suppress reflective interference, making details such as oil film on the water surface and lesions on the fish's surface clearly visible. When the water transparency decreases due to feeding residue or microbial growth, it is classified as a non-transparent water subset. The system generates a turbidity enhancement factor and uses image contrast stretching and particle texture enhancement algorithms to highlight defect features such as suspended particles and flocculent matter. This differentiated processing avoids the shortcomings of a single algorithm that over-amplifies noise in transparent water and fails to identify suspended matter in turbid water, improving the robustness and adaptability of visual detection under different transparency levels.

[0034] S32: Based on the characteristics of the feeding stage, the circulation stage, and the purification stage in the regulation stage, the dataset is divided into stage subsets. Feeding compensation factors are generated for the feeding stage subset to predict residual feed pollution, and filtration compensation factors are generated for the purification stage subset to predict filter media clogging. Specifically, based on the daily monitoring patterns of the aquarium, the system divides the comprehensive water quality dataset into three subsets: feeding, circulation, and purification. In the feeding subset (typically at fixed times in the morning and evening), the system generates a feeding compensation factor, focusing on monitoring the rate of dissolved oxygen decline and the upward trend of ammonia nitrogen. This helps predict the risk of ammonia poisoning from uneaten food decomposition and appropriately extends the air pump's operating time. In the purification subset (typically during peak water pump circulation times), the system generates a filtration compensation factor, focusing on analyzing the trend of water pressure difference before and after the filter media. This predicts the degree of filter media clogging and provides advance warnings of cleaning or replacement needs. This phased differentiation allows the model to accurately match typical water quality change patterns at each time period, avoiding misjudgments of filter media clogging during the feeding phase or uneaten food contamination during the purification phase, thus improving the targeting and timeliness of the control strategy.

[0035] S33: Based on the wear characteristics of the equipment health status, the dataset is divided into a healthy equipment subset and an aged equipment subset. A performance degradation compensation factor is generated for the aged equipment subset to correct the detection error caused by equipment passivation. Specifically, the system divides the dataset into healthy equipment subsets and aging equipment subsets based on equipment runtime and performance degradation indicators. For the aging equipment subset (such as filter media used for more than six months or water pumps with increased operating noise), the system generates a performance degradation compensation factor to correct and compensate for the flow rate and pressure data detected by the sensors. For example, when the actual flow rate of an aging water pump decreases due to impeller wear, the compensation factor automatically corrects the flow rate reading, preventing the model from misjudging it as pipe blockage and incorrectly triggering a cleaning command. This compensation effectively corrects detection errors caused by equipment passivation, extends equipment service life, avoids premature replacement and waste, and ensures that the accuracy of water quality analysis results does not decrease due to equipment aging, thereby improving the long-term economic efficiency and reliability of the system.

[0036] Steps S31-S33 systematically address the core pain points of traditional aquarium water quality management, which neglect dynamic factors such as changes in transparency, differences in control stages, and equipment aging. S31 enables visual detection to adapt to different water clarity levels; S32 matches control strategies to diurnal rhythms and operating stages; and S33 mitigates the impact of equipment performance degradation through data analysis. The synergy of these three elements endows the water quality management system with scene awareness and adaptive capabilities, strengthening its shift from "fixed parameter control" to "dynamic compensation optimization." This significantly improves aquarium ecological stability and breeding success rates, driving the transformation of aquarium farming towards data-driven, refined management.

[0037] Furthermore, an adaptive adjustment mechanism for root cause localization weights is added to the multidimensional attribution analysis engine, with the following steps: S501: Construct a health status degradation characteristic curve for equipment, and use the time-series change rate of filter media clogging degree and water pump fatigue degree as health status characteristics. When the health characteristics show an accelerated degradation trend, increase the attribution weight of the main cause of equipment aging. Specifically, after the aquarium filter media had been running continuously for more than 6 months, the system detected a continuous upward trend in filter media clogging through the water flow sensor, with the time-series change rate showing a significantly faster rate of deterioration than initially. Simultaneously, the water pump motor temperature increased, operating noise rose, and the fatigue characteristic curve entered an accelerated degradation phase. After constructing the equipment health status degradation characteristic curve, the multi-dimensional attribution analysis engine determined that the filter media and water pump were in a moderately aging state, automatically increasing the attribution weight of "equipment aging" as the primary cause. Based on this judgment, the generated compensation strategy prioritized replacing the filter media rather than increasing the water change frequency, avoiding ineffective control measures, ensuring the rectification directly addresses the root cause, significantly reducing resource waste caused by misjudgment, and extending the overall service life of the equipment.

[0038] S502: Extract the deviation characteristics of water parameters, calculate the deviation ratio between the actual pH and temperature and the standard parameters, and trigger the equipment control compensation strategy to be executed first when the deviation exceeds the preset threshold. Specifically, after the daily feeding period, the system monitors in real time that the ammonia nitrogen concentration deviates from the standard value by a continuously increasing margin. The calculation results of the water body parameter deviation characteristics show that the deviation level reaches the "significant deviation" level (different threshold values ​​can be set in other embodiments, corresponding to different judgment results). The multidimensional attribution analysis engine determines that "ammonia nitrogen exceeding the standard" is the main water quality anomaly and immediately triggers the priority execution queue of the equipment control compensation strategy: automatically starting the backup aeration pump and extending the aeration time, while opening the bottom suction water exchange valve to discharge the sewage in the uneaten feed accumulation area. This priority execution mechanism avoids the continuous accumulation of ammonia nitrogen, which leads to stress response in fish, and shortens the intervention time for water quality deterioration from several hours after traditional manual discovery to minutes, greatly improving the timeliness and effectiveness of water quality control.

[0039] S503: Identify environmental topological bottleneck characteristics, calculate the spatial matching degree between the feeding location and the filtration system, and when the matching degree is lower than the preset threshold, increase the weight of the main cause of structural water quality deterioration and generate equipment layout optimization suggestions. Specifically, the system, through spatial topology analysis, determined that the spatial matching between the current feeding location and the filtration system inlet was rated as "poor," indicating that uneaten food was difficult to collect effectively into the filtration system. After identifying the environmental topological bottleneck, the multidimensional attribution analysis engine increased the weight of "unreasonable layout," a major structural cause of water quality deterioration, and generated equipment layout optimization suggestions: moving the feeding point towards the filtration system or adding a local drainage device to improve uneaten food collection efficiency. These suggestions, based on actual spatial relationships rather than human experience, accurately pinpointed the structural shortcomings in water quality management, providing data-driven decision-making for aquarium hardware modifications. This avoids the predicament of repeatedly adjusting equipment parameters without fundamentally solving the problem, significantly improving long-term water quality stability and aquarium success rates.

[0040] It should be noted that the naming of the levels in this embodiment corresponds to the preset rules of this application. In other embodiments, the numbers in different ranges can be named according to the actual rules.

[0041] Regarding steps S501-S503, this application utilizes a three-dimensional attribution analysis mechanism—equipment health, parameter deviation, and topological bottleneck—constructed through S501-S503 to systematically address the core pain point of traditional aquarium management that focuses solely on chemical indicators while neglecting equipment status and spatial layout. The degradation characteristic curve in S501 accurately identifies aging equipment, avoiding misdiagnosis as water quality issues; the deviation characteristic in S502 triggers prioritized dynamic compensation, achieving rapid closed-loop control; and the topological bottleneck characteristic in S503 guides layout optimization, providing long-term improvement directions. The synergy of these three elements shifts water quality management from "symptom treatment" to "root cause management," with control strategies directly addressing the primary causes, significantly improving economic efficiency, and driving the transformation of aquarium management towards a data-driven, refined, and intelligent model.

[0042] In S60: After the step of converting the equipment control compensation strategy into actuator control commands for execution, a self-cleaning mechanism for detection tags is added, including the following steps: S601: Compare the results of manual re-inspection with the results of model prediction to identify abnormal detection samples that were misjudged by the model or missed by manual inspection. During the operation of the smart aquarium, the system, using a water quality prediction algorithm model, identified the water quality at 14:30 on a certain day as "suspended matter pollution" and triggered an alert. Upon manual re-inspection, visual observation revealed that the water was actually clear and transparent; the perceived "pollution" was actually a misjudgment caused by reflections from air bubbles adhering to the aquarium's inner wall, identified as a sample misjudged by the model. Simultaneously, a routine manual check at 16:00 failed to detect a slight exceedance of turbidity, but the system's turbidity sensor and image analysis model accurately identified the anomaly, identifying it as a sample missed by the manual check. The system packaged and recorded the detection timestamps, multimodal data (images, sensor readings), and original labels of these two types of abnormal samples (air bubble samples misjudged by the model and turbidity samples missed by the manual check), constructing an abnormal sample library to provide an accurate input data foundation for subsequent label purification.

[0043] S602: The detection labels of abnormal samples are downgraded in confidence, and cross-modal feature similarity analysis is used to find highly similar samples from nearby time periods for label reconstruction; Specifically, for bubble samples misclassified by the model in S601, the system first downgrades the confidence level of their detection label "suspended matter pollution," marking them as "suspicious - low confidence." Then, cross-modal feature similarity analysis is used to find a large set of samples within a nearby time period (14:00-15:00) that are highly similar to the misclassified sample images in terms of bubble shape, position, and reflection pattern, and whose turbidity and chemical sensor readings are completely normal. The analysis results show that over 95% of the samples in this set were correctly classified as "normal water quality" by the model. Based on this consensus result of highly similar samples, the system reconstructs the label of the original misclassified samples to "normal water quality," achieving label purification and preventing low-quality misclassified samples from contaminating the training set.

[0044] S603: Re-inject the reconstructed purification labels into the training set, incrementally fine-tune the water quality prediction algorithm model, and achieve automatic correction of label errors and continuous improvement of model robustness. In this embodiment, the reconstructed purification label (the bubble sample label is corrected to "normal") from S602 and the samples with excessive turbidity that were manually missed in S601 (labeled as "slightly polluted") are jointly re-injected into the training set of the water quality prediction algorithm model. The model is fine-tuned using an incremental learning algorithm, focusing on enhancing its sensitivity to true turbidity characteristics while suppressing overfitting to bubble reflection interference. After multiple rounds of incremental fine-tuning, the model's misjudgment rate for bubble interference significantly decreases in subsequent runs, while its detection rate of true water turbidity is significantly improved, and its overall robustness is continuously enhanced. This mechanism eliminates the need for the water quality prediction system to be retrained from scratch, significantly reducing the manual cost of model maintenance and system downtime, and promoting the evolution of aquarium water quality management towards a self-correcting and autonomously evolving intelligent model.

[0045] This application systematically addresses the core pain points of intelligent aquarium water quality management, such as model misjudgment, manual missed detection, and tag noise accumulation, through a closed loop of self-purification of detection tags and incremental model optimization constructed by S601-S603. S601's consistency comparison accurately identifies abnormal samples, S602's cross-modal similarity analysis enables automatic tag reconstruction and purification, and S603's incremental fine-tuning continuously improves model robustness and environmental adaptability. The synergy of these three elements endows the water quality prediction system with autonomous intelligent capabilities of "self-diagnosis, self-repair, and self-evolution," significantly improving detection accuracy and reliability, reducing the frequency of manual intervention, providing long-term, stable, and reliable intelligent water quality assurance for ornamental fish farming, and promoting the transformation of aquarium management towards a fully autonomous and intelligent quality control model.

[0046] In S60, after the step of converting the equipment control compensation strategy into actuator control commands for execution, an adaptive adjustment closed-loop optimization mechanism is added, including the following steps: S604: Calculate the residual between the water quality improvement effect after implementation and the expected target value, and construct an effect deviation evaluation vector; In this embodiment, after the smart aquarium has been running continuously for two weeks, the system's water quality sensors detected that despite daily filter media cleaning and water changes, the improvement in ammonia nitrogen concentration was significantly lower than expected. Through residual calculation, the system constructed an effect deviation evaluation vector: ammonia nitrogen removal efficiency showed a large deviation, turbidity reduction a moderate deviation, and pH stability a small deviation. This vector quantifies the gap between the current "high-frequency water change + strong aeration" control strategy and the ideal water quality state, providing accurate input for subsequent attribution analysis and avoiding judgments of control effectiveness based solely on subjective feelings.

[0047] S605: Based on the aforementioned deviation assessment vector, perform attribution analysis to identify key influencing factors leading to the deviation as improper compensation factor settings or outdated subset partitioning criteria; Specifically, based on the deviation assessment vector, the system performed attribution analysis: it found that the feeding compensation factor weights generated during the feeding phase were set too high, leading to excessive water changes and disrupting the aquatic ecological balance; it also found that the subset classification standard for filter media clogging was too lenient (only classifying filters into the aging subset when clogging exceeds a high threshold), failing to identify the gradual aging trend of filter media in the middle of operation. After identifying these key influencing factors, the system clarified that the root cause of the deviation lay in improper compensation factor settings and outdated subset classification standards, rather than simply insufficient equipment power, thus pointing the way for precise correction.

[0048] S606: Utilizes incremental learning algorithms to dynamically adjust the subset partitioning boundary conditions in the differentiation process and automatically corrects the weight coefficients in the compensation factor generation logic. Utilizing an incremental learning algorithm, the system dynamically adjusts the differentiated processing steps: tightening the subset boundary conditions of the feeding phase from "within 2 hours after feeding" to "within 1 hour after feeding" more accurately captures peak pollution from uneaten food; simultaneously, it reduces the weighting coefficient of the feeding compensation factor (from 0.8 to 0.5) and increases the weighting coefficient of filter media clogging characteristics (from 0.3 to 0.6). This adaptive correction makes subsequent adjustments more aligned with the actual needs of the current aquarium ecosystem, achieving parameter optimization without manual intervention and significantly improving the accuracy and cost-effectiveness of the control strategy.

[0049] S607: The revised subset partitioning criteria and compensation factors are fed back to achieve adaptive optimization of the next generation of detection and control parameters; Specifically, the revised subset partitioning criteria and compensation factors are fed back to the system front end and take effect immediately in the next round of testing. The new generation of detection automatically applies the optimized configuration, revealing a significant improvement in ammonia nitrogen levels and a marked enhancement in water quality stability. This closed-loop mechanism enables the system to continuously learn and self-evolve, automatically adapting to changes in the aquarium ecosystem over time (such as filter media aging and changes in fish density), forming a complete intelligent cycle of "detection-regulation-evaluation-correction," upgrading water quality management from static configuration to dynamic optimization.

[0050] This application utilizes an adaptive adjustment closed-loop optimization mechanism constructed through S604-S607 to systematically address the core pain points of traditional aquarium water quality management strategies being rigid and unable to evolve on their own. This mechanism enables the system to automatically learn and adapt to the dynamic changes in the aquarium ecosystem, continuously optimize control parameters, reduce reliance on manual adjustments, and promote the evolution of aquarium management towards a fully autonomous and intelligent automated mode, significantly improving the stability, reliability, and economy of long-term operation.

[0051] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0052] In one embodiment, an automatic water quality ecological regulation and control device for a smart fish tank is provided, which corresponds one-to-one with the automatic water quality ecological regulation and control method for a smart fish tank described in the above embodiments. For example... Figure 2 As shown, the water quality ecological automatic regulation and control device for smart fish tanks includes a multi-parameter unified control framework construction unit 1, which is used to establish a multi-parameter unified control framework for fish tanks based on pre-set basic parameter input rules. The comprehensive water quality dataset construction unit 2 is used to collect multi-dimensional environmental state data of the aquarium water in real time, and link the multi-dimensional environmental state data with the control framework of the intelligent aquaculture equipment to construct a comprehensive water quality dataset. The water quality regulation compensation factor generation unit 3 is used to perform differential automatic analysis and processing on the comprehensive water quality dataset. Based on the characteristics of water quality parameter types, regulation stage characteristics and equipment health characteristics, the dataset is divided into multiple differential subsets, and an appropriate water quality regulation compensation factor is generated independently for each subset. The water quality prediction and water body assessment unit 4 is used to input the differentiated subset and the corresponding water quality regulation compensation factor into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data, and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The equipment control compensation strategy generation unit 5 is used to input the predicted values ​​of water quality parameters into a pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates an equipment control compensation strategy. The optimization unit 6 is used to convert the equipment control compensation strategy into actuator control commands for execution, and after execution, re-collect water body state data to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

[0053] Specific limitations regarding the automatic water quality regulation and control device for intelligent aquariums can be found in the limitations of the automatic water quality regulation and control method for intelligent aquariums mentioned above, and will not be repeated here. Each module in the aforementioned automatic water quality regulation and control device for intelligent aquariums can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the electronic device, or stored in the memory of the electronic device as software, so that the processor can call and execute the corresponding operations of each module.

[0054] In one embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, this electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the database. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an automatic water quality ecological regulation and control method for intelligent fish tanks.

[0055] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: A unified control framework for multiple parameters of a fish tank is established based on pre-set basic parameter input rules. The multiple parameters include at least water quality pH, temperature, dissolved oxygen concentration, light intensity, and water circulation flow rate. Real-time collection of multi-dimensional environmental status data of aquarium water, and linkage of the multi-dimensional environmental status data with the control framework of intelligent aquaculture equipment to construct a comprehensive water quality dataset; The comprehensive water quality dataset is subjected to differential automatic analysis and processing. Based on the characteristics of water quality parameter types, control stage characteristics and equipment health characteristics, the dataset is divided into multiple differential subsets. For each subset, an appropriate water quality control compensation factor is generated independently. The differentiated subset and the corresponding water quality regulation compensation factor are input into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The predicted values ​​of water quality parameters are input into a pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates equipment control compensation strategy. The multidimensional feature data includes health characteristics, equipment operation deviation characteristics, and environmental topology characteristics. The equipment control compensation strategy is converted into actuator control commands and executed. After execution, water body state data is re-collected to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

[0056] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: The basic data collected by environmental sensors is entered into the database, and the control framework of the integrated intelligent aquarium aquaculture equipment is established through artificial intelligence algorithms. The multiple parameters include at least water quality pH, temperature, dissolved oxygen concentration, light intensity and water circulation flow rate. Real-time collection of multi-dimensional environmental status data of aquarium water, and linkage of the multi-dimensional environmental status data with the control framework of intelligent aquaculture equipment to construct a comprehensive water quality dataset; The comprehensive water quality dataset is subjected to differential automatic analysis and processing. Based on the characteristics of water quality parameter types, control stage characteristics and equipment health characteristics, the dataset is divided into multiple differential subsets. For each subset, an appropriate water quality control compensation factor is generated independently. The differentiated subset and the corresponding water quality regulation compensation factor are input into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The predicted values ​​of water quality parameters are input into a pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates equipment control compensation strategy. The multidimensional feature data includes health characteristics, equipment operation deviation characteristics, and environmental topology characteristics. The equipment control compensation strategy is converted into actuator control commands and executed. After execution, water body state data is re-collected to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

[0057] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0058] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0059] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for automatic water quality ecological regulation and control in intelligent fish tanks, characterized in that, The method includes the following steps: The basic data collected by environmental sensors is entered into the database, and the control framework of the integrated intelligent aquarium breeding equipment is established through artificial intelligence algorithms. Real-time collection of multi-dimensional environmental status data of aquarium water, and linkage of the multi-dimensional environmental status data with the control framework of intelligent aquaculture equipment to construct a comprehensive water quality dataset; The comprehensive water quality dataset is subjected to differential automatic analysis and processing. Based on the characteristics of water quality parameter types, control stage characteristics and equipment health characteristics, the dataset is divided into multiple differential subsets. For each subset, an appropriate water quality control compensation factor is generated independently. The differentiated subset and the corresponding water quality regulation compensation factor are input into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The predicted values ​​of water quality parameters are input into a pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates equipment control compensation strategy. The equipment control compensation strategy is converted into actuator control commands and executed. After execution, water body state data is re-collected to calculate the water quality improvement effect. The improvement effect evaluation results are fed back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

2. The method for automatic water quality ecological regulation and control in an intelligent fish tank according to claim 1, characterized in that, In the step of collecting multi-dimensional environmental status data of the aquarium water in real time and linking this multi-dimensional environmental status data with the control framework of the intelligent aquaculture equipment to construct a comprehensive water quality dataset, a multi-source data preprocessing mechanism is added, including the following steps: A multimodal state vector is constructed from water body video images, underwater sonar echoes, and physical sensor signals synchronously acquired through edge nodes; By utilizing a preset synchronous triggering mechanism, the multimodal state vector is aligned with the common time axis at the nanosecond level, eliminating the sampling delay differences between multiple sensors; The aligned data stream is then subjected to sliding window filtering and outlier removal to generate a preprocessed comprehensive water quality dataset.

3. The method for automatic water quality ecological regulation and control in an intelligent fish tank according to claim 1, characterized in that, The step of performing differential automatic analysis on the comprehensive water quality dataset, dividing the dataset into multiple differential subsets based on water quality parameter type characteristics, regulation stage characteristics, and equipment health characteristics, and independently generating an appropriate water quality regulation compensation factor for each subset includes the following steps: Based on the water transparency characteristics, the comprehensive water quality dataset is divided into transparent water subset and non-transparent water subset. An illumination compensation factor is generated for the transparent water subset to suppress reflective interference, and a turbidity enhancement factor is generated for the non-transparent water subset to highlight suspended solids defects. Based on the characteristics of the feeding stage, the circulation stage, and the purification stage in the regulation phase, the dataset is divided into stage subsets. Feeding compensation factors are generated for the feeding stage subset to predict residual feed pollution, and filtration compensation factors are generated for the purification stage subset to predict filter media clogging. Based on the wear and tear characteristics of the equipment health status, the dataset is divided into a healthy equipment subset and an aged equipment subset. A performance degradation compensation factor is generated for the aged equipment subset to correct the detection error caused by equipment passivation.

4. The method for automatic water quality ecological regulation and control in an intelligent fish tank according to claim 1, characterized in that, To add an adaptive adjustment mechanism for root cause localization weights in the multidimensional attribution analysis engine, the steps are as follows: Construct a health status degradation characteristic curve for equipment, and use the time-series change rate of filter media clogging degree and water pump fatigue degree as health status characteristics. When the health characteristics show an accelerated degradation trend, increase the attribution weight of the main cause of equipment aging. Extract the deviation characteristics of water parameters, calculate the deviation ratio between the actual pH and temperature and the standard parameters, and trigger the equipment control compensation strategy to be executed first when the deviation exceeds the preset threshold; Identify environmental topological bottlenecks, calculate the spatial matching degree between the feeding location and the filtration system, and when the matching degree is lower than the preset threshold, increase the weight of the main cause of structural water quality deterioration and generate equipment layout optimization suggestions.

5. The method for automatic water quality ecological regulation and control in an intelligent fish tank according to claim 1, characterized in that, After the step of converting the equipment control compensation strategy into actuator control commands for execution, a self-cleaning mechanism for detection labels is added, including the following steps: comparing the consistency between the manual re-inspection results and the model prediction results to identify abnormal detection samples that are misjudged by the model or missed by the manual. The detection labels of abnormal samples are downgraded in confidence, and cross-modal feature similarity analysis is used to find highly similar samples from nearby time periods for label reconstruction. The reconstructed purification labels are re-injected into the training set, and the water quality prediction algorithm model is incrementally fine-tuned to achieve automatic correction of label errors and continuous improvement of model robustness.

6. The method for automatic water quality ecological regulation and control in an intelligent fish tank according to claim 5, characterized in that, After the step of converting the equipment control compensation strategy into actuator control commands for execution, an adaptive adjustment closed-loop optimization mechanism is added, including the following steps: calculating the residual between the water quality improvement effect after execution and the expected target value, and constructing an effect deviation evaluation vector; Based on the aforementioned deviation assessment vector, attribution analysis was performed to identify key influencing factors leading to the deviation as improper compensation factor settings or outdated subset partitioning criteria. Incremental learning algorithms are used to dynamically adjust the subset partitioning boundary conditions in the differentiation process and to automatically correct the weight coefficients in the compensation factor generation logic. The revised subset partitioning criteria and compensation factors are fed back to enable adaptive optimization of the next generation of detection and control parameters.

7. The method for automatic water quality ecological regulation and control in an intelligent fish tank according to claim 1, characterized in that, The equipment control compensation strategies include automatic feeding compensation strategy, automatic oxygenation compensation strategy, automatic circulation compensation strategy, automatic temperature control compensation strategy, automatic sterilization compensation strategy, and automatic water change compensation strategy.

8. A water quality ecological automatic regulation and control device for a smart fish tank, applied to the water quality ecological automatic regulation and control method for a smart fish tank as described in any one of claims 1 to 7, characterized in that, The device includes: a multi-parameter unified control framework construction unit (1), used to establish a multi-parameter unified control framework for aquariums based on pre-set basic parameter input rules; The comprehensive water quality dataset construction unit (2) is used to collect multi-dimensional environmental status data of aquarium water in real time, and link the multi-dimensional environmental status data with the control framework of intelligent aquaculture equipment to construct a comprehensive water quality dataset. The water quality regulation compensation factor generation unit (3) is used to perform differential automatic analysis and processing on the comprehensive water quality dataset. According to the characteristics of water quality parameter type, regulation stage and equipment health, the dataset is divided into multiple differential subsets, and an appropriate water quality regulation compensation factor is generated independently for each subset. The water quality prediction and water body assessment unit (4) is used to input the differential subset and the corresponding water quality regulation compensation factor into the pre-trained water quality prediction algorithm model. The water quality prediction algorithm model performs multi-scale feature extraction on multi-dimensional state data and integrates the compensation factor to perform cross-modal feature enhancement, and outputs the predicted value of water quality parameters and the corresponding water quality level assessment value. The equipment control compensation strategy generation unit (5) is used to input the predicted value of water quality parameters into the pre-set multidimensional attribution analysis engine. The multidimensional attribution analysis engine integrates the multidimensional feature data of the water body, locates the main cause category of water quality abnormality, and generates equipment control compensation strategy. The execution optimization unit (6) is used to convert the equipment control compensation strategy into actuator control commands for execution, and after execution, re-collect water body state data to calculate the water quality improvement effect, and feed the improvement effect evaluation results back to the differential analysis and processing stage to correct the division criteria of the differential subset and the generation logic of the water quality regulation compensation factor.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the automatic water quality ecological regulation and control method for an intelligent fish tank as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the automatic water quality ecological regulation and control method for an intelligent fish tank as described in any one of claims 1 to 7.

Images