Intelligent water affair system driven sewage treatment sludge backflow accurate regulation and control method

Abstract

The invention discloses an intelligent water affair system driven sewage treatment sludge backflow accurate regulation and control method, and relates to the technical field of sewage treatment.The method comprises the steps that water quality sensing signals, water inlet spectral signals and sludge floc images of water inlet, an aeration tank, a secondary sedimentation tank and a backflow pipeline in a sewage treatment system are collected; time sequence water quality characteristics are extracted from the water quality sensing signals at edge calculation nodes, pollutant concentration characteristics are inverted from the inflow spectral signals, floc morphological characteristics are extracted from the sludge floc images, and the time sequence water quality characteristics, the pollutant concentration characteristics and the floc morphological characteristics are fused to generate sludge state characteristic vectors; calculating a sludge health index representing the sludge settling performance and the active state based on the sludge state feature vector; and inputting the sludge state feature vector and the sludge health index into a backflow strategy model which is trained based on a digital twin environment and deployed at an edge node after model compression, and outputting a recommended backflow ratio.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, and in particular to a method for precise control of sludge recirculation in wastewater treatment driven by a smart water system. Background Technology

[0002] Wastewater treatment is a technological system that purifies water bodies using physical, chemical, and biological methods to meet discharge or reuse standards. It is mainly divided into two categories: industrial wastewater treatment and domestic wastewater treatment. The treatment process includes three stages: primary treatment removes suspended solids using physical methods such as screens and sedimentation; secondary treatment uses biological technologies such as activated sludge and biofilm processes to degrade organic matter; and tertiary treatment uses technologies such as membrane separation and ion exchange for deep purification. The industry is gradually promoting high-efficiency processes such as membrane bioreactors and electron beam treatment.

[0003] However, existing technologies rely solely on MLSS or flow signals for feedback, lacking synchronous characterization of influent load dynamics, floc microstructure, and biological activity. This results in regulation lagging behind actual operating conditions. Mainstream reflux control still employs fixed ratios or simple threshold switching logic, failing to balance multiple conflicting objectives such as effluent quality compliance, energy consumption optimization, and system stability. In particular, it lacks proactive intervention capabilities during influent shock or sludge bulking precursor stages. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a precise control method for sludge recirculation in wastewater treatment driven by a smart water system. This method addresses the problem of relying solely on MLSS or flow signals for feedback, which lacks synchronous characterization of influent load dynamics, floc microstructure, and biological activity. As a result, the control lags behind actual operating conditions, and the mainstream recirculation control still adopts a fixed ratio or simple threshold switching logic. This fails to balance the conflicting objectives of achieving effluent quality standards, optimizing energy consumption, and maintaining system stability. In particular, it lacks the ability to proactively intervene in the early stages of influent impact or sludge bulking.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for precise control of sludge recirculation in wastewater treatment driven by a smart water management system, comprising,

[0008] The system collects water quality sensor signals, influent spectral signals, and sludge floc images from the influent, aeration tank, secondary sedimentation tank, and return pipeline of the wastewater treatment system.

[0009] At the edge computing node, time-series water quality features are extracted from water quality sensing signals, pollutant concentration features are inverted from influent spectral signals, floc morphology features are extracted from sludge floc images, and time-series water quality features, pollutant concentration features and floc morphology features are fused to generate sludge state feature vector.

[0010] The sludge health index, which characterizes the settling performance and activity state of sludge, is calculated based on the sludge state feature vector.

[0011] The sludge state feature vector and sludge health index are input into the recirculation strategy model trained in a digital twin environment and deployed on edge nodes after model compression, and the recommended recirculation ratio is output.

[0012] The recommended reflow ratio is adjusted based on the sludge health index and the preset process safety boundary to obtain the commanded reflow ratio.

[0013] The command reflux ratio is converted into a device control signal to drive the reflux pump and adjust the sludge reflux flow rate;

[0014] The actual return flow rate and the suspended solids concentration of the mixed liquor in the aeration tank after the return flow is obtained are used as feedback data to update the sludge state feature vector.

[0015] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the intelligent water system described in this invention, the steps of extracting time-series water quality features from water quality sensor signals at edge computing nodes, inverting pollutant concentration features from influent spectral signals, extracting floc morphology features from sludge floc images, and fusing the time-series water quality features, pollutant concentration features, and floc morphology features to generate a sludge state feature vector are as follows:

[0016] Raw water quality signals were collected from dissolved oxygen, oxidation-reduction potential, pH, and temperature sensors. For each type of signal, a length of [length missing] was used. The sliding window is normalized, and the first-order difference mean of the signal within the window is calculated to obtain the time-series water quality characteristics.

[0017] The incoming water spectral signal was acquired from the ultraviolet-visible spectrometer. The incoming water spectral signal is of wavelength. Absorbance sequence in the 200nm to 800nm ​​range The absorbance sequence is input into a partial least squares regression model pre-trained in the cloud. The partial least squares regression model uses historically synchronized laboratory measurements of chemical oxygen demand (COD) and total nitrogen concentration as labels to output an estimated value of the current influent COD. Total nitrogen estimate This constitutes the characteristics of pollutant concentration;

[0018] Images of sludge flocs from an industrial camera are acquired and input into a lightweight visual transform network deployed at edge nodes. The lightweight visual transform network performs pixel-level segmentation of the floc region using an encoder-decoder structure and calculates the floc projection area-weighted average particle size based on the segmentation results. Normalized value of the standard deviation of floc grayness per unit area Together, they constitute the morphological characteristics of flocs;

[0019] The time-series water quality characteristics, pollutant concentration characteristics, and floc morphology characteristics are concatenated according to a fixed field order to generate a sludge state feature vector with fixed dimensions. .

[0020] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the intelligent water system described in this invention, the specific steps for calculating the sludge health index characterizing sludge settling performance and activity state based on sludge state feature vectors are as follows:

[0021] From sludge state feature vector Extracting the average particle size of flocs With floc density index Substituting the values ​​into the estimated sludge volume index, the expression is:

[0022] ;

[0023] in, , , It is a positive real constant, determined by fitting the measured sludge volume index with the corresponding floc image features from historical operating data. These are unitless estimated values; the higher the value, the worse the sludge settling performance. The average particle size of the flocs is... The density index of flocs;

[0024] From sludge state feature vector Extract the current dissolved oxygen concentration and compared with the preset optimal dissolved oxygen reference value Construct normalized deviation terms;

[0025] Estimate sludge volume index Flocculation density index Current dissolved oxygen concentration Reference density and optimal dissolved oxygen reference value Substitute all the components into the nonlinear health scoring function to calculate the sludge health index. The expression is:

[0026] ;

[0027] in, For sludge health index, , , These are positive weighting coefficients, determined by multi-objective optimization calibration. The value range of is [0,1]. When A value close to 1 indicates that the sludge settles quickly, has a dense structure, and is metabolically active. A value below 0.4 indicates a risk of expansion or aging, requiring the triggering of an emergency control mechanism.

[0028] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the smart water system described in this invention, the method involves inputting the sludge state feature vector and sludge health index into a recirculation strategy model trained in a digital twin environment and deployed on edge nodes after model compression, and outputting a recommended recirculation ratio. The specific steps are as follows:

[0029] A high-fidelity digital twin simulation environment is built in the cloud. The high-fidelity digital twin simulation environment takes the ASM2d activated sludge mechanism model as the core and integrates the influent disturbance module, the biochemical reaction kinetics module, the sedimentation tank sludge layer dynamics module and the return pump energy consumption module. The model parameters are calibrated online using historical operating data.

[0030] In a high-fidelity digital twin simulation environment, the state space of reinforcement learning is defined as containing sludge state feature vectors. With sludge health index The joint vector, the action space is the continuous reflux ratio The expression for the reward function is:

[0031] ;

[0032] in, The concentration of ammonia nitrogen in the effluent. Its emission standard limits, Instantaneous power consumption of the reflux pump This represents the absolute deviation of the concentration of suspended solids in the mixture from the set value. , , For adjustable weights, This is the reward value;

[0033] A policy network is trained in a high-fidelity digital twin simulation environment using a deep deterministic policy gradient algorithm until the policy converges.

[0034] Knowledge distillation is performed on the trained policy network. Using the original network as the teacher model, a three-layer fully connected student network is trained. Channel pruning and 8-bit quantization are then applied to generate a lightweight backflow policy model with an inference latency of less than 20 milliseconds.

[0035] Deploy the lightweight backflow strategy model on edge computing nodes and set the current cycle's... and Input model, output recommended reflux ratio .

[0036] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the intelligent water system described in this invention, the step of limiting and correcting the recommended recirculation ratio based on the sludge health index and preset process safety boundaries to obtain the commanded recirculation ratio includes the following steps:

[0037] Set a basic safety boundary for the reflux ratio, including a lower limit. and upper limit Set sludge health emergency thresholds. When the sludge health index When the system is determined to be in a high-risk state, the instruction return ratio is forcibly set to a preset high return emergency value. ;

[0038] Instruction return ratio Determined according to the rules:

[0039] When recommending reflux ratio Less than the lower limit of the reflux ratio season ;

[0040] when Between and In between, let ;

[0041] when Greater than the upper limit of the reflux ratio season Obtain the final instruction return ratio This ensures that control commands are always within the process feasible domain.

[0042] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the intelligent water system described in this invention, the specific steps of converting the command recirculation ratio into a device control signal to drive the recirculation pump and adjust the sludge recirculation flow rate are as follows:

[0043] Real-time acquisition of the outflow rate for the current cycle The outflow rate is provided by an electromagnetic flow meter installed in the main outflow pipe;

[0044] Based on command return ratio With water flow rate Calculate the target return flow rate The expression is:

[0045] ;

[0046] in, This refers to the instruction return ratio. For water flow rate, For target return traffic;

[0047] Under the control of the frequency converter, multiple frequency points are set sequentially. After the system stabilizes, the corresponding actual return flow rate and frequency are recorded synchronously. The flow rate-frequency characteristic curve is then fitted. The flow rate-frequency characteristic curve of the return pump is obtained through on-site calibration, characterizing the output frequency of the frequency converter. With actual return flow The monotonic mapping relationship between them;

[0048] Based on target return flow The required inverter frequency can be deduced from the characteristic curve by looking up a table. ;

[0049] Frequency commands are transmitted via Modbus TCP protocol. The signal is sent to the return pump frequency converter driver, which drives the motor to adjust its speed, thus achieving closed-loop regulation.

[0050] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the intelligent water system described in this invention, the specific steps for obtaining the actual recirculation flow rate and the suspended solids concentration of the mixed liquor in the aeration tank after recirculation execution, and using this as feedback data to update the sludge state feature vector, are as follows:

[0051] Set a time delay after the command is issued. Then, the actual return flow rate output by the electromagnetic flowmeter on the return pipeline is collected. The concentration of suspended solids in the mixed liquor output by the online MLSS sensor in the aeration tank ;

[0052] Using actual return flow Compared with the same period of outflow The actual reflux ratio is calculated using the following expression:

[0053] ;

[0054] in, This represents the actual return flow. This refers to the outflow rate during the same period. This is the actual reflux ratio;

[0055] The current concentration of suspended solids in the mixture Compared with the historical average under normal operating conditions The comparison is performed, and if the deviation exceeds the preset tolerance zone, the reference density is dynamically updated. Compared with the optimal dissolved oxygen reference value The update rule is exponential smoothing, and the expression is:

[0056] ;

[0057] in, This is the smoothing coefficient, and its value range is... , For the updated reference density, This represents the floc density index for the current cycle. The reference density used for the previous cycle or historical cycles. This represents the dissolved oxygen concentration for the current period.

[0058] Based on the updated and The feature extraction process is repeated to generate the sludge state feature vector for the next control cycle. .

[0059] As a preferred embodiment of the precise control method for sludge recirculation in wastewater treatment driven by the intelligent water system described in this invention, the specific steps for recommending the recirculation ratio are as follows:

[0060] Based on the actual return flow rate and mixed liquor suspended solids concentration collected over multiple consecutive historical control cycles, the sludge return stability index was calculated. The expression is:

[0061] ;

[0062] in, For the first The actual return flow rate of the cycle, This refers to the outflow rate during the same period. This refers to the concurrent instruction return rate. The length of the sliding window. As an indicator of sludge return stability;

[0063] When the sludge return stability index When the preset tolerance threshold is exceeded, it is determined that there is an abnormal fluctuation in the current sludge settling behavior;

[0064] In response to the assessment, the emergency threshold for the sludge health index will be dynamically lowered. New threshold The expression is:

[0065] ;

[0066] in, As the baseline emergency threshold, This is the stability tolerance threshold. The maximum allowable fluctuation value, The sensitivity coefficient and ;

[0067] The updated The logic for limiting and correcting the recommended return ratio is applied in the next regulatory cycle.

[0068] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the precise control method for sludge return in wastewater treatment driven by the smart water system as described in the first aspect of the present invention.

[0069] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the precise control method for sludge return in wastewater treatment driven by the smart water system as described in the first aspect of the present invention.

[0070] The beneficial effects of this invention are as follows: multi-source heterogeneous sensing data, such as water quality sensor signals, ultraviolet-visible influent spectral signals, and sludge floc images, are simultaneously collected from the influent, aeration tank, secondary sedimentation tank, and return pipeline of the sewage treatment system. At the edge computing node, the three types of signals are processed in a targeted manner to extract the time-series dynamic features of water quality. The COD / TN concentration is inverted using a cloud-based pre-trained partial least squares regression model. The image is segmented using a lightweight visual network, and the weighted average particle size and grayscale standard deviation normalized value of the floc projection area are quantified. Then, the three are fused into a sludge state feature vector with fixed dimensions and clear physical meaning. This not only ensures real-time performance but also provides a structured, low-redundancy, high-quality input for subsequent intelligent decision-making. Attached Figure Description

[0071] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0072] Figure 1 A flowchart of a precise control method for sludge recirculation in wastewater treatment driven by a smart water management system. Detailed Implementation

[0073] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0074] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0075] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0076] Reference Figure 1 This embodiment of the invention provides a method for precise control of sludge recirculation in wastewater treatment driven by a smart water system, comprising the following steps:

[0077] S1 collects water quality sensor signals, influent spectral signals, and sludge floc images from the influent, aeration tank, secondary sedimentation tank, and return pipeline of the sewage treatment system.

[0078] Furthermore, by deploying distributed sensors and imaging equipment at the inlet, the end of the aeration tank, the sludge layer interface of the secondary sedimentation tank, and key nodes of the return pipeline, the entire chain of water quality dynamics, pollutant components, and sludge floc morphology can be synchronously perceived.

[0079] It should be noted that by simultaneously collecting multi-source heterogeneous sensing data from the influent, aeration tank, secondary sedimentation tank, and return pipeline, comprehensive coverage of the key states of the entire wastewater treatment process was achieved, providing a high-dimensional, real-time, and multimodal data foundation for subsequent precise modeling and control.

[0080] S2. At the edge computing node, extract time-series water quality features from water quality sensing signals, invert pollutant concentration features from influent spectral signals, extract floc morphology features from sludge floc images, and fuse time-series water quality features, pollutant concentration features, and floc morphology features to generate a sludge state feature vector.

[0081] Furthermore, raw water quality signals were collected from dissolved oxygen, oxidation-reduction potential, pH, and temperature sensors. For each type of signal, a length of [length missing] was used. The sliding window is normalized, and the first-order difference mean of the signal within the window is calculated to obtain the time-series water quality characteristics.

[0082] The incoming water spectral signal was acquired from the ultraviolet-visible spectrometer. The incoming water spectral signal is of wavelength. Absorbance sequence in the 200nm to 800nm ​​range The absorbance sequence is input into a partial least squares regression model pre-trained in the cloud. The partial least squares regression model uses historically synchronized laboratory measurements of chemical oxygen demand (COD) and total nitrogen concentration as labels to output an estimated value of the current influent COD. Total nitrogen estimate This constitutes the characteristics of pollutant concentration;

[0083] Images of sludge flocs from an industrial camera are acquired and input into a lightweight visual transform network deployed at edge nodes. The lightweight visual transform network performs pixel-level segmentation of the floc region using an encoder-decoder structure and calculates the floc projection area-weighted average particle size based on the segmentation results. Normalized value of the standard deviation of floc grayness per unit area Together, they constitute the morphological characteristics of flocs;

[0084] The time-series water quality characteristics, pollutant concentration characteristics, and floc morphology characteristics are concatenated according to a fixed field order to generate a sludge state feature vector with fixed dimensions. .

[0085] It should be noted that by fusing time-series water quality features, spectral inversion pollutant concentration features, and image-driven floc morphology features at edge nodes, a sludge state feature vector with fixed dimensions and clear physical meaning was constructed, which improved the completeness of feature expression and the consistency of model input.

[0086] S3. Calculate the sludge health index, which characterizes the sludge settling performance and activity state, based on the sludge state feature vector.

[0087] Furthermore, from the sludge state feature vector Extracting the average particle size of flocs With floc density index Substituting the values ​​into the estimated sludge volume index, the expression is:

[0088] ;

[0089] in, , , It is a positive real constant, determined by fitting the measured sludge volume index with the corresponding floc image features from historical operating data. These are unitless estimated values; the higher the value, the worse the sludge settling performance. The average particle size of the flocs is... The density index of flocs;

[0090] From sludge state feature vector Extract the current dissolved oxygen concentration and compared with the preset optimal dissolved oxygen reference value Construct normalized deviation terms;

[0091] Estimate sludge volume index Flocculation density index Current dissolved oxygen concentration Reference density and optimal dissolved oxygen reference value Substitute all the components into the nonlinear health scoring function to calculate the sludge health index. The expression is:

[0092] ;

[0093] in, For sludge health index, , , These are positive weighting coefficients, determined by multi-objective optimization calibration. The value range of is [0,1]. When A value close to 1 indicates that the sludge settles quickly, has a dense structure, and is metabolically active. A value below 0.4 indicates a risk of expansion or aging, requiring the triggering of an emergency control mechanism.

[0094] It should be noted that the constructed sludge health index comprehensively reflects the floc structure, settling performance and metabolic activity, and can identify the risk of sludge bulking or aging at an early stage, providing a quantitative decision-making basis for the dynamic control of the reflux ratio.

[0095] S4. Input the sludge state feature vector and sludge health index into the recirculation strategy model, which is trained based on the digital twin environment and deployed on the edge node after model compression, and output the recommended recirculation ratio.

[0096] Furthermore, a high-fidelity digital twin simulation environment is built in the cloud. The high-fidelity digital twin simulation environment takes the ASM2d activated sludge mechanism model as the core and integrates the influent disturbance module, the biochemical reaction kinetics module, the sedimentation tank sludge layer dynamics module and the return pump energy consumption module. The model parameters are calibrated online using historical operating data.

[0097] In a high-fidelity digital twin simulation environment, the state space of reinforcement learning is defined as containing sludge state feature vectors. With sludge health index The joint vector, the action space is the continuous reflux ratio The expression for the reward function is:

[0098] ;

[0099] in, The concentration of ammonia nitrogen in the effluent. Its emission standard limits, Instantaneous power consumption of the reflux pump This represents the absolute deviation of the concentration of suspended solids in the mixture from the set value. , , For adjustable weights, This is the reward value;

[0100] A policy network is trained in a high-fidelity digital twin simulation environment using a deep deterministic policy gradient algorithm until the policy converges.

[0101] Knowledge distillation is performed on the trained policy network. Using the original network as the teacher model, a three-layer fully connected student network is trained. Channel pruning and 8-bit quantization are then applied to generate a lightweight backflow policy model with an inference latency of less than 20 milliseconds.

[0102] Deploy the lightweight backflow strategy model on edge computing nodes and set the current cycle's... and Input model, output recommended reflux ratio ;

[0103] Based on the actual return flow rate and mixed liquor suspended solids concentration collected over multiple consecutive historical control cycles, the sludge return stability index was calculated. The expression is:

[0104] ;

[0105] in, For the first The actual return flow rate of the cycle, This refers to the outflow rate during the same period. This refers to the concurrent instruction return rate. The length of the sliding window. As an indicator of sludge return stability;

[0106] When the sludge return stability index When the preset tolerance threshold is exceeded, it is determined that there is an abnormal fluctuation in the current sludge settling behavior;

[0107] In response to the assessment, the emergency threshold for the sludge health index will be dynamically lowered. New threshold The expression is:

[0108] ;

[0109] in, As the baseline emergency threshold, This is the stability tolerance threshold. The maximum allowable fluctuation value, The sensitivity coefficient and ;

[0110] The updated The logic for limiting and correcting the recommended return ratio is applied in the next regulatory cycle;

[0111] The process of training a policy network in a high-fidelity digital twin simulation environment using a deep deterministic policy gradient algorithm is as follows:

[0112] The sludge state feature vector and sludge health index are used together as the input state of the agent, and the continuously adjustable reflux ratio is used as the output action. A multi-objective reward mechanism is designed that comprehensively considers the compliance of effluent ammonia nitrogen, reflux pump energy consumption and mixed liquor suspended solids concentration stability.

[0113] Simulate various typical and extreme working conditions in a digital twin environment, allowing the agent to continuously try different return ratio strategies and record experience data such as state, action, reward and next state generated by its interaction with the environment;

[0114] The algorithm uses two cooperating neural networks, a policy network, and a value network to generate the optimal action and evaluate the long-term benefits of the action, respectively, and introduces a target network and an experience replay mechanism to improve learning stability.

[0115] Add time-dependent noise to the motion output;

[0116] After multiple rounds of iterative training, when the policy network can stably output a reflow ratio that meets the process safety boundary and has excellent overall performance under various disturbance conditions, the training is considered to have converged, and a reflow control policy model that can be deployed is obtained.

[0117] During the model compression phase, channel pruning is first performed on the trained reflow policy network:

[0118] By evaluating the importance of neurons in each convolutional or fully connected layer, redundant or low-contribution channels are removed, thereby simplifying the network structure and reducing the number of parameters and computational cost.

[0119] Then, 8-bit quantization is performed: the weights and activation values ​​in the network, which were originally represented as 32-bit floating-point numbers, are converted into 8-bit integers, which significantly reduces memory usage and computational complexity while keeping the model's inference accuracy basically unchanged.

[0120] The two steps described above work together to significantly reduce the model size and improve inference speed, meeting the requirements of edge computing nodes for low latency and low power consumption deployment.

[0121] It should be noted that the lightweight reflux strategy model, trained and deployed in a compressed digital twin environment, combines mechanistic interpretability with edge real-time reasoning capabilities. Furthermore, by dynamically adjusting the emergency threshold through the sludge reflux stability index, the system's adaptive robustness to operational disturbances is enhanced.

[0122] S5. Based on the sludge health index and the preset process safety boundary, the recommended reflux ratio is adjusted to obtain the commanded reflux ratio.

[0123] Furthermore, set a basic safety boundary for the reflux ratio, including a lower limit. and upper limit Set sludge health emergency thresholds. When the sludge health index When the system is determined to be in a high-risk state, the instruction return ratio is forcibly set to a preset high return emergency value. ;

[0124] Instruction return ratio Determined according to the rules:

[0125] When recommending reflux ratio Less than the lower limit of the reflux ratio season ;

[0126] when Between and In between, let ;

[0127] when Greater than the upper limit of the reflux ratio season Obtain the final instruction return ratio This ensures that control commands are always within the process feasible domain.

[0128] It should be noted that by combining the sludge health index with rigid process boundaries to impose dual limits on the recommended reflux ratio, the safety of the effluent quality is ensured while avoiding equipment over-limit operation, effectively balancing control performance and engineering feasibility.

[0129] S6. Convert the commanded reflux ratio into a device control signal to drive the reflux pump and adjust the sludge reflux flow rate.

[0130] Furthermore, it can obtain the outflow rate for the current cycle in real time. The outflow rate is provided by an electromagnetic flow meter installed in the main outflow pipe;

[0131] Based on command return ratio With water flow rate Calculate the target return flow rate The expression is:

[0132] ;

[0133] in, This refers to the instruction return ratio. For water flow rate, For target return traffic;

[0134] Under the control of the frequency converter, multiple frequency points are set sequentially. After the system stabilizes, the corresponding actual return flow rate and frequency are recorded synchronously. The flow rate-frequency characteristic curve is then fitted. The flow rate-frequency characteristic curve of the return pump is obtained through on-site calibration, characterizing the output frequency of the frequency converter. With actual return flow The monotonic mapping relationship between them;

[0135] Based on target return flow The required inverter frequency can be deduced from the characteristic curve by looking up a table. ;

[0136] Frequency commands are transmitted via Modbus TCP protocol. The signal is sent to the return pump frequency converter driver, which drives the motor to adjust its speed, thus achieving closed-loop regulation.

[0137] It should be noted that the closed-loop frequency control method based on the measured outflow rate and the characteristic curve of the return pump achieves high-precision tracking of the return flow rate, overcoming the execution deviation problem caused by changes in pipeline resistance in traditional open-loop control.

[0138] S7. Obtain the actual return flow rate and the suspended solids concentration of the mixed liquor in the aeration tank after the return flow is executed, and use them as feedback data to update the sludge state feature vector.

[0139] Furthermore, a time delay can be set after the command is issued. Then, the actual return flow rate output by the electromagnetic flowmeter on the return pipeline is collected. The concentration of suspended solids in the mixed liquor output by the online MLSS sensor in the aeration tank ;

[0140] Using actual return flow Compared with the same period of outflow The actual reflux ratio is calculated using the following expression:

[0141] ;

[0142] in, This represents the actual return flow. This refers to the outflow rate during the same period. This is the actual reflux ratio;

[0143] The current concentration of suspended solids in the mixture Compared with the historical average under normal operating conditions The comparison is performed, and if the deviation exceeds the preset tolerance zone, the reference density is dynamically updated. Compared with the optimal dissolved oxygen reference value The update rule is exponential smoothing, and the expression is:

[0144] ;

[0145] in, This is the smoothing coefficient, and its value range is... , For the updated reference density, This represents the floc density index for the current cycle. The reference density used for the previous cycle or historical cycles. This represents the dissolved oxygen concentration for the current period.

[0146] Based on the updated and The feature extraction process is repeated to generate the sludge state feature vector for the next control cycle. .

[0147] It should be noted that by dynamically updating the reference compaction and optimal dissolved oxygen reference values ​​through actual reflux execution feedback, the feature extraction benchmark is continuously adapted to the current operating conditions, effectively suppressing model drift caused by seasonal changes in water quality or equipment aging.

[0148] This embodiment also provides a computer device applicable to the precise control method for sludge return in wastewater treatment driven by a smart water system, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the precise control method for sludge return in wastewater treatment driven by a smart water system as proposed in the above embodiment.

[0149] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0150] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the precise control method for sludge recirculation in wastewater treatment driven by a smart water system, as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0151] In summary, this invention achieves this by simultaneously acquiring multi-source heterogeneous sensing data, including water quality sensor signals, UV-Vis influent spectral signals, and sludge floc images, from the influent, aeration tank, secondary sedimentation tank, and return pipeline of the wastewater treatment system. At edge computing nodes, these three types of signals are processed to extract dynamic time-series features of water quality. A cloud-based pre-trained partial least squares regression model is used to invert COD / TN concentrations. A lightweight visual network is used to segment images and quantify the weighted average particle size and grayscale standard deviation normalized value of the floc projection area. Finally, these three data points are fused into a sludge state feature vector with fixed dimensions and clear physical meaning. This approach ensures real-time performance and provides structured, low-redundancy, high-quality input for subsequent intelligent decision-making.

[0152] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for precise control of sludge recirculation in wastewater treatment driven by a smart water management system, characterized by: Includes the following steps: The system collects water quality sensor signals, influent spectral signals, and sludge floc images from the influent, aeration tank, secondary sedimentation tank, and return pipeline of the wastewater treatment system. At the edge computing node, time-series water quality features are extracted from water quality sensing signals, pollutant concentration features are inverted from influent spectral signals, floc morphology features are extracted from sludge floc images, and time-series water quality features, pollutant concentration features and floc morphology features are fused to generate sludge state feature vector. The sludge health index, which characterizes the settling performance and activity state of sludge, is calculated based on the sludge state feature vector. The sludge state feature vector and sludge health index are input into the recirculation strategy model trained in a digital twin environment and deployed on edge nodes after model compression, and the recommended recirculation ratio is output. The recommended reflow ratio is adjusted based on the sludge health index and the preset process safety boundary to obtain the commanded reflow ratio. The command reflux ratio is converted into a device control signal to drive the reflux pump and adjust the sludge reflux flow rate; The actual return flow rate and the suspended solids concentration of the mixed liquor in the aeration tank after the return flow is obtained are used as feedback data to update the sludge state feature vector.

2. The method for precise control of sludge return from wastewater treatment driven by a smart water system as described in claim 1, characterized in that: The specific steps are as follows: extracting time-series water quality features from water quality sensing signals at edge computing nodes, inverting pollutant concentration features from influent spectral signals, extracting floc morphology features from sludge floc images, and fusing the time-series water quality features, pollutant concentration features, and floc morphology features to generate a sludge state feature vector. Raw water quality signals were collected from dissolved oxygen, oxidation-reduction potential, pH, and temperature sensors. For each type of signal, a length of [length missing] was used. The sliding window is normalized, and the first-order difference mean of the signal within the window is calculated to obtain the time-series water quality characteristics. The incoming water spectral signal was acquired from the ultraviolet-visible spectrometer. The incoming water spectral signal is of wavelength. Absorbance sequence in the 200nm to 800nm ​​range The absorbance sequence is input into a partial least squares regression model pre-trained in the cloud. The partial least squares regression model uses historically synchronized laboratory measurements of chemical oxygen demand (COD) and total nitrogen concentration as labels to output an estimated value of the current influent COD. Total nitrogen estimate This constitutes the characteristics of pollutant concentration; Images of sludge flocs from an industrial camera are acquired and input into a lightweight visual transform network deployed at edge nodes. The lightweight visual transform network performs pixel-level segmentation of the floc region using an encoder-decoder structure and calculates the floc projection area-weighted average particle size based on the segmentation results. Normalized value of the standard deviation of floc grayness per unit area Together, they constitute the morphological characteristics of flocs; The time-series water quality characteristics, pollutant concentration characteristics, and floc morphology characteristics are concatenated according to a fixed field order to generate a sludge state feature vector with fixed dimensions. .

3. The method for precise control of sludge recirculation in wastewater treatment driven by a smart water system as described in claim 2, characterized in that: The specific steps for calculating the sludge health index, which characterizes sludge settling performance and activity state, based on sludge state feature vectors are as follows: From sludge state feature vector Extracting the average particle size of flocs With floc density index Substituting the values ​​into the estimated sludge volume index, the expression is: ; in, , , It is a positive real constant, determined by fitting the measured sludge volume index with the corresponding floc image features from historical operating data. These are unitless estimated values; the higher the value, the worse the sludge settling performance. The average particle size of the flocs is... The density index of flocs; From sludge state feature vector Extract the current dissolved oxygen concentration and compared with the preset optimal dissolved oxygen reference value Construct normalized deviation terms; Estimate sludge volume index Flocculation density index Current dissolved oxygen concentration Reference density and optimal dissolved oxygen reference value Substitute all the components into the nonlinear health scoring function to calculate the sludge health index. The expression is: ; in, For sludge health index, , , These are positive weighting coefficients, determined by multi-objective optimization calibration.

4. The method for precise control of sludge recirculation in wastewater treatment driven by a smart water system as described in claim 3, characterized in that: The specific steps for inputting the sludge state feature vector and sludge health index into the recirculation strategy model trained in a digital twin environment and deployed on edge nodes after model compression, and outputting a recommended recirculation ratio are as follows: A high-fidelity digital twin simulation environment is built in the cloud. The high-fidelity digital twin simulation environment takes the ASM2d activated sludge mechanism model as the core and integrates the influent disturbance module, the biochemical reaction kinetics module, the sedimentation tank sludge layer dynamics module and the return pump energy consumption module. The model parameters are calibrated online using historical operating data. In a high-fidelity digital twin simulation environment, the state space of reinforcement learning is defined as containing sludge state feature vectors. With sludge health index The joint vector, the action space is the continuous reflux ratio The expression for the reward function is: ; in, The concentration of ammonia nitrogen in the effluent. Its emission standard limits, Instantaneous power consumption of the reflux pump This represents the absolute deviation of the concentration of suspended solids in the mixture from the set value. , , For adjustable weights, This is the reward value; A policy network is trained in a high-fidelity digital twin simulation environment using a deep deterministic policy gradient algorithm until the policy converges. Knowledge distillation is performed on the trained policy network. Using the original network as the teacher model, a three-layer fully connected student network is trained. Channel pruning and 8-bit quantization are then applied to generate a lightweight backflow policy model with an inference latency of less than 20 milliseconds. Deploy the lightweight backflow strategy model on edge computing nodes and set the current cycle's... and Input model, output recommended reflux ratio .

5. The method for precise control of sludge recirculation in wastewater treatment driven by a smart water system as described in claim 4, characterized in that: The recommended reflux ratio is adjusted based on the sludge health index and the preset process safety boundary to obtain the commanded reflux ratio. The specific steps are as follows: Set a basic safety boundary for the reflux ratio, including a lower limit. and upper limit ; Set sludge health emergency threshold When the sludge health index When the system is determined to be in a high-risk state, the instruction return ratio is forcibly set to a preset high return emergency value. ; Instruction return ratio Determined according to the rules: When recommending reflux ratio Less than the lower limit of the reflux ratio season ; when Between and In between, let ; when Greater than the upper limit of the reflux ratio season Obtain the final instruction return ratio This ensures that control commands are always within the process feasible domain.

6. The method for precise control of sludge recirculation in wastewater treatment driven by a smart water system as described in claim 5, characterized in that: The specific steps for converting the commanded reflux ratio into a device control signal to drive the reflux pump and adjust the sludge reflux flow rate are as follows: Real-time acquisition of the outflow rate for the current cycle The outflow rate is provided by an electromagnetic flow meter installed in the main outflow pipe; Based on command return ratio With water flow rate Calculate the target return flow rate The expression is: ; in, This refers to the instruction return ratio. For water flow rate, For target return traffic; Under the control of the frequency converter, multiple frequency points are set sequentially. After the system stabilizes, the corresponding actual return flow rate and frequency are recorded synchronously. The flow rate-frequency characteristic curve is then fitted. The flow rate-frequency characteristic curve of the return pump is obtained through on-site calibration, characterizing the output frequency of the frequency converter. With actual return flow The monotonic mapping relationship between them; Based on target return flow The required inverter frequency can be deduced from the characteristic curve by looking up a table. ; Frequency commands are transmitted via Modbus TCP protocol. The signal is sent to the return pump frequency converter driver, which drives the motor to adjust its speed, thus achieving closed-loop regulation.

7. The method for precise control of sludge recirculation in wastewater treatment driven by a smart water system as described in claim 6, characterized in that: The steps for obtaining the actual return flow rate and the suspended solids concentration of the mixed liquor in the aeration tank after the return flow is executed, and using this as feedback data to update the sludge state feature vector, are as follows: Set a time delay after the command is issued. Then, the actual return flow rate output by the electromagnetic flowmeter on the return pipeline is collected. The concentration of suspended solids in the mixed liquor output by the online MLSS sensor in the aeration tank ; Using actual return flow Compared with the same period of outflow The actual reflux ratio is calculated using the following expression: ; in, This represents the actual return flow. This refers to the outflow rate during the same period. This is the actual reflux ratio; The current concentration of suspended solids in the mixture Compared with the historical average under normal operating conditions The comparison is performed, and if the deviation exceeds the preset tolerance zone, the reference density is dynamically updated. Compared with the optimal dissolved oxygen reference value The update rule is exponential smoothing, and the expression is: ; in, This is the smoothing coefficient, and its value range is... , For the updated reference density, This represents the floc density index for the current cycle. The reference density used for the previous cycle or historical cycles. This represents the dissolved oxygen concentration for the current period. Based on the updated and The feature extraction process is repeated to generate the sludge state feature vector for the next control cycle. .

8. The method for precise control of sludge recirculation in wastewater treatment driven by a smart water system as described in claim 7, characterized in that: The specific steps for recommending the reflux ratio are as follows: Based on the actual return flow rate and mixed liquor suspended solids concentration collected over multiple consecutive historical control cycles, the sludge return stability index was calculated. The expression is: ; in, For the first The actual return flow rate of the cycle, This refers to the outflow rate during the same period. This refers to the concurrent instruction return rate. The length of the sliding window. As an indicator of sludge return stability; When the sludge return stability index When the preset tolerance threshold is exceeded, it is determined that there is an abnormal fluctuation in the current sludge settling behavior; In response to the assessment, the emergency threshold for the sludge health index will be dynamically lowered. New threshold The expression is: ; in, As the baseline emergency threshold, This is the stability tolerance threshold. The maximum allowable fluctuation value, The sensitivity coefficient and ; The updated The logic for limiting and correcting the recommended return ratio is applied in the next regulatory cycle.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the precise control method for sludge return in wastewater treatment driven by any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the precise control method for sludge return in wastewater treatment driven by the smart water system according to any one of claims 1 to 8.

Images