Dissolved oxygen prediction method for sewage aeration tank based on adaptive mode decomposition

Abstract

The present application relates to the technical field of intelligent water affairs and intelligent monitoring and optimal control of sewage treatment process, and discloses a sewage aeration tank dissolved oxygen prediction method based on adaptive modal decomposition, comprising the following steps: collecting operation data of the sewage aeration tank; calculating the theoretical dissolved oxygen change rate of the current period based on the dissolved oxygen mass balance relationship of the aeration tank; performing frequency domain analysis on the physical residual sequence to obtain the high-frequency residual energy proportion; adaptively updating the modal layer number or the penalty factor of the variational modal decomposition according to the high-frequency residual energy proportion; inputting the multi-scale sub-sequence and the operation data after alignment and fusion into a prediction model to output the dissolved oxygen prediction value at the target time. The present application overcomes the defects of traditional data-driven models, such as no physical constraints and prediction results easily deviating from the process boundary; and solves the problem of fixed parameters of variational modal decomposition, which cannot adapt to the fluctuation of working conditions.

Description

Technical Field

[0001] This invention relates to the fields of smart water management and intelligent monitoring and optimization control of wastewater treatment processes, specifically a method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition. Background Technology

[0002] Dissolved oxygen is a core control variable in activated sludge wastewater treatment systems, as its concentration directly determines the efficiency of biochemical reactions, the stability of effluent quality, and the level of aeration energy consumption. Accurate and proactive prediction of dissolved oxygen levels in the aeration tank is a crucial prerequisite for achieving intelligent aeration regulation and cost reduction while increasing efficiency.

[0003] Existing dissolved oxygen prediction technologies are mainly divided into two categories: The first category is mechanism-driven methods, which rely on activated sludge models and mass balance equations to construct prediction models. These models have physical interpretability, but key parameters such as oxygen transfer coefficients and oxygen consumption rates are difficult to identify accurately online. Under scenarios of fluctuating influent loads and sudden changes in operating conditions, the model deviations are significant, and the prediction accuracy is insufficient to meet control requirements. The second category is data-driven methods, which combine time-series neural networks, signal decomposition, and prediction models to achieve predictions. These methods have high accuracy under normal operating conditions, but they have obvious drawbacks: First, the models lack process physical constraints, and are prone to outputting unreasonable results that exceed physical boundaries under sudden operating conditions. Second, signal decomposition methods such as variational mode decomposition rely on offline preset parameters and cannot adaptively adjust according to the degree of non-stationarity of operating conditions. Third, the mechanism constraints, signal decomposition, and prediction models are independent of each other, failing to form a closed-loop drive from mechanism to data, resulting in insufficient overall robustness and generalization ability.

[0004] In summary, existing technologies cannot simultaneously achieve physical rationality, adaptability to operating conditions, and predictive accuracy, making it difficult to support the stable and efficient operation of wastewater treatment plants under complex and fluctuating operating conditions. Summary of the Invention

[0005] This invention provides a method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition, comprising the following steps: S1. Collect operational data of the wastewater aeration tank. The operational data includes at least dissolved oxygen concentration, saturated dissolved oxygen concentration, oxygen consumption rate, influent flow rate, effective volume of the aeration tank, influent dissolved oxygen concentration, aeration volume, and process variables characterizing influent load or sludge state. The data is filled with missing values ​​and outliers are removed to form a time-aligned multivariate time series. S2. Based on the dissolved oxygen mass balance relationship in the aeration tank, calculate the theoretical dissolved oxygen change rate for the current period; obtain the measured dissolved oxygen change rate by the difference between the measured dissolved oxygen values ​​at adjacent times; construct a physical residual sequence by subtracting the theoretical change rate from the measured change rate; this residual characterizes the degree to which the biochemical reaction process deviates from the ideal mass balance, is independent of the prediction model, and can truly reflect the fluctuation state of the operating conditions. S3. Perform frequency domain transformation on the physical residual sequence, and calculate the ratio of the spectral energy in the preset high-frequency interval to the total energy of the entire spectrum to obtain the high-frequency residual energy ratio; this index quantifies the non-stationarity of the working condition and the intensity of the disturbance, providing a basis for adjusting the decomposition parameters; S4. Adaptively update the number of mode layers or penalty factor of variational mode decomposition according to the proportion of high-frequency residual energy, and use the updated parameters to perform variational mode decomposition on the dissolved oxygen sequence to obtain multi-scale subsequences; S5. After aligning and fusing the multi-scale subsequence with the running data, input it into the prediction model and output the predicted dissolved oxygen value at the target time.

[0007] Furthermore, the dissolved oxygen mass balance relationship in S2 is as follows: ; in, The rate of change of dissolved oxygen concentration. Oxygen transfer coefficient, This represents the saturated dissolved oxygen concentration. Dissolved oxygen concentration, Oxygen consumption rate, This refers to the inlet water flow rate. The effective volume of the aeration tank, This refers to the dissolved oxygen concentration in the influent.

[0008] Furthermore, the high-frequency residual energy ratio is the ratio of the spectral energy in the preset high-frequency interval to the total energy of the entire spectrum after the physical residual sequence is transformed in the frequency domain. This indicator quantifies the non-stationarity and disturbance intensity of the working condition, providing a basis for adjusting the decomposition parameters.

[0009] Furthermore, the adaptive update of the modal layer number or penalty factor of the variational mode by the high-frequency residual energy ratio includes: setting at least two judgment thresholds; increasing the modal layer number when the high-frequency residual energy ratio exceeds the first threshold but does not exceed the second threshold; and simultaneously increasing the modal layer number and adjusting the penalty factor or shortening the decomposition window length when the high-frequency residual energy ratio exceeds the second threshold. The adaptive update also includes a parameter rollback mechanism: when the high-frequency residual energy ratio is lower than the first threshold within a consecutive preset number of sliding windows, the modal layer number and penalty factor are restored to the basic configuration.

[0010] Furthermore, the multi-scale subsequence includes at least two of the following: trend subsequence, periodic subsequence, and high-frequency perturbation subsequence.

[0011] Furthermore, the prediction model is a neural network model with temporal feature extraction capability. The prediction model integrates channel attention mechanism and temporal attention mechanism. The channel attention mechanism is used to adaptively weigh the contributions of different process variables and modal subsequences, and the temporal attention mechanism is used to locate the time step in the historical sequence that has the highest correlation with the current prediction. The forward propagation process of the prediction model includes: performing global average pooling on the multi-scale feature matrix through a channel attention mechanism, outputting feature channel weight vectors and completing feature recalibration; inputting the recalibrated features into a bidirectional long short-term memory network to complete temporal encoding, obtaining a hidden state sequence; calculating the weights of historical time steps through a temporal attention mechanism, and weighting and summing the hidden states to obtain a context vector; mapping the context vector through a fully connected layer to output the dissolved oxygen prediction value.

[0012] Furthermore, it also includes: generating abnormal alert information based on the deviation between the predicted and measured dissolved oxygen values, combined with dynamic statistical thresholds; or outputting aeration adjustment auxiliary decision-making information based on the changing trend of the predicted dissolved oxygen values.

[0013] Furthermore, a dissolved oxygen prediction system for wastewater aeration tank includes: a data acquisition module for collecting operational data from the wastewater aeration tank; a physical residual construction module for calculating the theoretical dissolved oxygen change rate based on the dissolved oxygen mass balance relationship, and subtracting the theoretical dissolved oxygen change rate from the measured dissolved oxygen change rate to obtain the physical residual sequence; a parameter adaptive update module for performing frequency domain analysis on the physical residual sequence, and updating the mode layer number or penalty factor of the variational mode decomposition according to the proportion of high-frequency residual energy; a mode decomposition module for performing variational mode decomposition on the dissolved oxygen sequence using the updated parameters to obtain multi-scale sub-sequences; and a prediction module for fusing the multi-scale sub-sequences with the operational data and inputting them into a prediction model to output the predicted dissolved oxygen value at the target time. The high-frequency residual energy ratio is the ratio of the spectral energy in the preset high-frequency range to the total energy of the entire spectrum after the physical residual sequence has undergone frequency domain transformation. The adaptive update of the modal layer number or penalty factor of variational mode decomposition based on the proportion of high-frequency residual energy includes: setting at least two judgment thresholds; increasing the modal layer number when the proportion of high-frequency residual energy exceeds the first threshold but does not exceed the second threshold; and simultaneously increasing the modal layer number and adjusting the penalty factor or shortening the decomposition window length when the proportion of high-frequency residual energy exceeds the second threshold. The adaptive update also includes a parameter rollback mechanism: when the proportion of high-frequency residual energy is lower than the first threshold within a consecutive preset number of sliding windows, the modal layer number and penalty factor are restored to the basic configuration. The multi-scale subsequence includes at least two of the following: trend subsequence, periodic subsequence, and high-frequency perturbation subsequence.

[0014] Furthermore, an electronic device includes a processor and a memory, the memory storing computer-executable instructions that can be executed by the processor, wherein the processor executes the computer-executable instructions to implement a method for predicting dissolved oxygen in a wastewater aeration tank based on adaptive mode decomposition.

[0015] Furthermore, a computer-readable storage medium stores computer-executable instructions that, when executed by a processor, implement a method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition.

[0016] The present invention has the following beneficial effects: This adaptive mode decomposition-based method for predicting dissolved oxygen in wastewater aeration tanks overcomes the shortcomings of traditional data-driven models, such as the lack of physical constraints and the tendency for prediction results to deviate from process boundaries. It also solves the problem of fixed variational mode decomposition parameters that cannot adapt to fluctuations in operating conditions. Furthermore, it breaks the limitation that the mechanistic model, signal decomposition, and prediction model are independent of each other, and constructs a closed-loop driven prediction system. This method improves the accuracy and stability of dissolved oxygen prediction under sudden operating conditions such as influent shocks and load changes. By constructing physical residuals using the dissolved oxygen mass balance equation, mechanistic constraints directly drive the adjustment of signal decomposition parameters, overcoming the limitations of traditional methods that separate mechanism from data, and ensuring that the prediction results possess rigorous physical rationality. It has strong adaptability to operating conditions. Based on the frequency domain characteristics of physical residuals, it realizes online adaptive updating of variational mode decomposition parameters. It improves decomposition accuracy under sudden shock conditions and automatically backs up under stable conditions, balancing prediction effect and computational efficiency. Improved prediction accuracy and robustness: The fusion of multi-scale decoupling and dual attention mechanism accurately extracts temporal features and key information, significantly improving prediction accuracy and model generalization ability under fluctuating conditions. The project is highly practical and does not rely on offline parameter optimization. It can operate online adaptively, adapt to the process characteristics of different wastewater treatment plants, directly support intelligent optimization control of aeration, reduce energy consumption, and improve effluent stability. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the process for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition, as described in this invention. Figure 2 This is a schematic diagram of the physical residual-driven adaptive update logic for variational mode decomposition parameters in this invention. Figure 3 This is a schematic diagram of the structure of the multi-scale feature enhancement and temporal prediction model of the present invention; Figure 4 This is a schematic diagram of the dissolved oxygen prediction system for the wastewater aeration tank of the present invention. Figure 5 This is a schematic diagram of the structure of the electronic device of the present invention.

[0018] In the diagram: 100, Data Acquisition Module; 200, Physical Residual Construction Module; 300, Parameter Adaptive Update Module; 400, Mode Decomposition Module; 500, Prediction Module. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0020] Dissolved oxygen (DO) is a key variable affecting treatment efficiency and operating energy consumption in activated sludge wastewater treatment. However, the volatility of influent water quality poses a significant challenge to DO prediction, and traditional management methods struggle to cope with complex operating conditions, resulting in high aeration energy consumption and low control precision.

[0021] Based on this, the present invention provides a method, system, electronic device and storage medium for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition. By establishing a variational mode decomposition parameter adaptive update mechanism driven by physical residuals, it achieves deep coupling between mechanism constraints and data-driven prediction.

[0022] To facilitate understanding of this embodiment, a detailed description of the method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition disclosed in this embodiment of the invention will be provided first.

[0023] This invention provides a method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition. Figure 1 This is a flowchart illustrating a method for predicting dissolved oxygen in a wastewater aeration tank based on adaptive mode decomposition, provided as an embodiment of the present invention. Figure 1 As shown, the method may include the following steps: Step S1: Collect operational data of the wastewater aeration tank.

[0024] The operational data includes at least dissolved oxygen concentration, influent flow rate, aeration rate, and one or more process variables characterizing influent load or sludge condition.

[0025] Specifically, process variables may include: influent chemical oxygen demand (COD), influent ammonia nitrogen (NH3-N), sludge concentration (MLSS), water temperature, pH value, reflux ratio, etc.

[0026] The data collection frequency can be once every 5 minutes. The original data is filled with missing values ​​by linear interpolation. Outliers that exceed the physical reasonable range (such as DO less than 0 or DO greater than 15 mg / L) are removed and replaced with adjacent valid values, and finally a time-aligned multivariate sequence is formed.

[0027] Step S2: Based on the dissolved oxygen mass balance relationship in the aeration tank, calculate the theoretical dissolved oxygen change rate in the current time period.

[0028] The dissolved oxygen mass balance relationship is as follows: ; in, This represents the rate of change in dissolved oxygen concentration. The oxygen transfer coefficient under actual working conditions can be estimated in real time based on the correlation between aeration volume and experience. This represents the saturated dissolved oxygen concentration, which is related to temperature and air pressure. This represents the current dissolved oxygen concentration within the reactor. The biological oxygen demand rate of activated sludge; This refers to the influent flow rate; The effective volume of the aeration tank; The dissolved oxygen concentration carried by the influent; Specifically, the measured dissolved oxygen change rate is approximated by the difference between the measured DO values ​​at adjacent sampling times.

[0029] It should be noted that the physical residual defined in this invention is not prediction error in the general sense. Prediction error is the difference between the output value of the prediction model and the measured value, reflecting the model's fitting ability, and it decreases as the model's training level improves. Physical residual, on the other hand, is the difference between the theoretical rate of change of dissolved oxygen derived from the mass conservation equation and the measured rate of change. It reflects the degree to which the current biochemical reaction process deviates from the ideal mass balance state. Even with excellent prediction model training, the physical residual will not be zero because the mass balance equation itself is a simplified approximation of a complex biochemical process. It is this characteristic that allows the physical residual to reflect the changing state of the operating conditions independently of the prediction model, thus providing a rationale for adjusting the driving signal decomposition strategy.

[0030] Step S3: Construct a physical residual sequence based on the difference between the measured dissolved oxygen change rate and the theoretical dissolved oxygen change rate, and perform frequency domain analysis on the physical residual sequence to obtain the proportion of high-frequency residual energy.

[0031] Specifically, the physical residual sequence is obtained by subtracting the theoretical rate of change from the measured rate of change. The physical residual is defined as follows: ; Perform Fast Fourier Transform or equivalent frequency domain analysis on the physical residual sequence, and statistically analyze the proportion of residual energy in the preset high-frequency interval to the total energy of the entire spectrum, denoted as the high-frequency residual energy proportion Rh. The calculation formula is as follows: ;in, These are the Fourier transform coefficients of the physical residual sequence. The high-frequency range can be set to the range with a normalized frequency greater than 0.3. The specific value can be adjusted according to the sampling rate and process characteristics.

[0032] Step S4: Adaptively update the mode number K or penalty factor of variational mode decomposition according to the proportion of high-frequency residual energy, and use the updated parameters to perform variational mode decomposition on the dissolved oxygen sequence in the current window to obtain multi-scale subsequences.

[0033] Specifically, at least two threshold levels, T1 and T2, are set, where T2 is greater than T1. When the proportion of high-frequency residual energy Rh is less than or equal to the first threshold T1, the current operating condition is determined to be within the normal fluctuation range, and the basic decomposition parameter configuration is maintained. When Rh is greater than T1 and less than or equal to T2, the current operating condition is determined to have a moderate degree of non-stationary disturbance, and the number of mode layers K is increased. When Rh is greater than T2, the current operating condition is determined to have a strong sudden impact or abnormal fluctuation, and the penalty factor is adjusted or the decomposition window length is shortened while increasing the number of mode layers K.

[0034] Furthermore, when the proportion of high-frequency residual energy in a consecutive preset number of M sliding windows is lower than the first threshold T1, the modal layer number K and the penalty factor are restored to their base values ​​to avoid over-decomposition under stable operating conditions, which would lead to wasted computing resources and amplified noise.

[0035] Specifically, variational mode decomposition is performed on the DO sequence within the current window using the updated parameters to obtain multi-scale subsequences, including trend subsequences, periodic subsequences, and high-frequency disturbance subsequences. The trend subsequence corresponds to the intrinsic mode function with the lowest center frequency, reflecting the slow change trend of DO; the periodic subsequence corresponds to the intrinsic mode function with a medium center frequency, reflecting daily periodic fluctuations; and the high-frequency disturbance subsequence corresponds to the intrinsic mode function with the highest center frequency, reflecting sudden disturbances and noise.

[0036] Step S5: Combine the multi-scale subsequences with the running data and input them into the prediction model to output the predicted dissolved oxygen value at the target time.

[0037] Specifically, the multi-scale subsequences are aligned and concatenated with the original process variable sequence in the time dimension to form a multi-scale enhanced feature matrix, which is then input into a pre-trained time series prediction model to output the predicted dissolved oxygen concentration at the future target time.

[0038] The prediction model is a neural network model with the ability to extract temporal features.

[0039] Optionally, the method also includes step S6: generating an abnormality alert based on the deviation between the predicted and measured dissolved oxygen values, combined with a dynamic statistical threshold, or outputting aeration adjustment auxiliary decision-making information based on the changing trend of the predicted dissolved oxygen values.

[0040] The dissolved oxygen prediction method for wastewater aeration tanks based on adaptive mode decomposition provided in this invention establishes a continuous technical action chain of "mass balance equation - physical residual - online update of decomposition parameters - multi-scale prediction". This allows the mechanistic constraints to no longer remain at the level of parallel auxiliary or loss function additions, but to directly participate in the online adjustment of the eigenvalue decomposition strategy. It can improve the decomposition precision in a timely manner when there is influent impact or sensor anomaly, and automatically restore the basic configuration under stable operating conditions, thereby improving the accuracy and robustness of dissolved oxygen prediction under different operating conditions. Example 2

[0041] This invention also provides another method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition; this method is implemented based on the method in the above embodiments; the method focuses on describing the specific implementation of the adaptive update of VMD parameters based on the frequency domain characteristics of physical residuals.

[0042] Figure 2 This is a schematic diagram of the VMD parameter adaptive update logic based on physical residual frequency domain characteristics provided in an embodiment of the present invention. Figure 2 As shown, the adaptive update process may include the following steps: Step S401: Within a sliding window of length L, calculate the theoretical DO change rate for each time step based on the aeration tank mass balance equation. The oxygen transfer coefficient and oxygen consumption rate can be estimated using empirical correlations. ; in, This represents the theoretical rate of change in dissolved oxygen. The measured rate of change of dissolved oxygen (DO) is given. Simultaneously, the measured rate of change is obtained by the difference between the measured DO values ​​at adjacent time points. Subtracting the two yields the physical residual sequence.

[0043] Step S402: Perform a fast Fourier transform on the physical residual sequence, set the high-frequency interval as the interval with a normalized frequency greater than 0.3, and calculate the proportion Rh of the residual spectral energy in this interval to the total energy of the full spectrum.

[0044] Step S403: Perform multi-level threshold judgment based on the numerical range of Rh. Specifically, set a first threshold T1 and a second threshold T2, the number of basic modal layers K0, and a basic penalty factor. When Rh is less than or equal to T1, maintain the basic parameters; when T1 is less than Rh and less than or equal to T2, increase K; when Rh is greater than T2, increase K while increasing the penalty factor and shortening the decomposition window length. When Rh is lower than T1 for M consecutive windows, restore K and the penalty factor to their basic values.

[0045] Step S404: Perform variational mode decomposition on the DO sequence in the current window using the updated parameters, and output multi-scale subsequences.

[0046] The specific values ​​of the threshold, modal layer number, and penalty factor can be determined based on the historical operating data of the target wastewater treatment plant by statistically analyzing the Rh distribution characteristics of the physical residual sequence, or updated online using an adaptive statistical method. Example 3

[0047] This invention also provides another method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition; this method is implemented based on the method in the above embodiments; this method focuses on describing the specific structure of the time-series prediction model.

[0048] Figure 3 This is a schematic diagram of the structure of the multi-scale feature enhancement and temporal prediction model provided in an embodiment of the present invention. Figure 3 As shown, in a preferred embodiment, the prediction model employs a channel-temporal dual attention network, and its forward propagation process includes: Step S501, Channel Attention Layer. Global average pooling is performed on the multi-scale enhanced feature matrix in the time dimension. The weight vectors of each feature channel are output through a fully connected layer and a sigmoid activation function, adaptively recalibrating the response intensity of each process variable and each intrinsic mode function subsequence.

[0049] Step S502, Bidirectional LSTM encoding layer. The recalibrated features are input into the bidirectional long short-term memory network for temporal encoding to obtain the hidden state sequence containing forward and backward historical information.

[0050] Step S503, Temporal Attention Layer. The attention weights for each historical time step in the hidden state sequence for the current prediction task are calculated, and the hidden states at each time step are weighted and summed to generate a context vector. This attention mechanism can identify key time steps such as the physical response delay period after the aeration adjustment command is issued.

[0051] Step S504, Output Layer. The context vector is mapped through a fully connected layer to output the predicted dissolved oxygen concentration at the target time.

[0052] It should be noted that the aforementioned channel-temporal dual attention network is a preferred implementation of the prediction model. In other implementations, the prediction model may also employ any of the following neural network models with temporal feature extraction capabilities: recurrent neural networks, temporal convolutional networks, Transformer-like networks, or combinations thereof. Example 4

[0053] Corresponding to the above method embodiments, this invention provides a dissolved oxygen prediction system for wastewater aeration tanks. Figure 4 This is a schematic diagram of a dissolved oxygen prediction system for a wastewater aeration tank provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the system may include: The data acquisition module 100 is used to collect operational data of the wastewater aeration tank; wherein the operational data includes at least dissolved oxygen concentration, influent flow rate, aeration rate, and one or more process variables characterizing influent load or sludge state.

[0054] The physical residual construction module 200 is used to calculate the theoretical dissolved oxygen change rate based on the dissolved oxygen mass balance relationship in the aeration tank, and to obtain the physical residual sequence by subtracting the theoretical dissolved oxygen change rate from the measured dissolved oxygen change rate.

[0055] The parameter adaptive update module 300 is used to perform frequency domain analysis on the physical residual sequence and update the mode layer number K or penalty factor of the variational mode decomposition according to the proportion of high-frequency residual energy.

[0056] The mode decomposition module 400 is used to perform variational mode decomposition on the dissolved oxygen sequence using the updated parameters to obtain multi-scale subsequences.

[0057] The prediction module 500 is used to output the predicted dissolved oxygen value at the target time based on the multi-scale subsequence and running data.

[0058] The dissolved oxygen prediction system for wastewater aeration tanks provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0059] In some embodiments, the parameter adaptive update module 300 is further configured to set at least two threshold levels: when the proportion of high-frequency residual energy exceeds the first threshold but does not exceed the second threshold, the modal layer number K is increased; when the proportion of high-frequency residual energy exceeds the second threshold, the penalty factor is adjusted or the decomposition window length is shortened while increasing the modal layer number K.

[0060] In some embodiments, the parameter adaptive update module 300 is also used to perform a rollback mechanism, which restores the modal layer number K and the penalty factor to their base values ​​when the proportion of high-frequency residual energy is lower than the first threshold within a consecutive preset number of sliding windows.

[0061] In some embodiments, the prediction model in the prediction module 500 includes a channel attention mechanism and a temporal attention mechanism. The channel attention mechanism is used to adaptively weigh the contributions of different process variables and different modal subsequences, while the temporal attention mechanism is used to identify the key time steps in the historical sequence that are most strongly associated with the current prediction task.

[0062] The system provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment. Example 5

[0063] This invention also provides an electronic device for running the above-described method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition; see also Figure 5 The diagram shows the structure of an electronic device, which includes a memory 602 and a processor 601. The memory 602 stores one or more computer instructions, which are executed by the processor 601 to implement the above-mentioned method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition.

[0064] Furthermore, Figure 5 The electronic device shown also includes a communication interface 603 and a bus 604, with the processor 601, communication interface 603 and memory 602 connected via the bus 604.

[0065] The memory 602 may include high-speed random access memory (RAM), or it may also include unstable memory, such as at least one disk storage device. Communication between the electronic device and an external network is achieved through at least one communication interface 603 (which may be wired or wireless).

[0066] Processor 601 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed through integrated logic circuits in the hardware of processor 601 or through software instructions. Processor 601 can be a general-purpose processor, including a central processing unit (CPU), a network processor, etc.; it can also be a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0067] This invention also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are called and executed by a processor, they cause the processor to implement the above-described method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0068] The computer program product of the wastewater aeration tank dissolved oxygen prediction method based on adaptive mode decomposition provided in this embodiment of the invention includes a computer-readable storage medium storing non-volatile program code executable by a processor. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0069] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0070] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0071] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0072] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0073] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0074] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, 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 the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition, characterized in that, Includes the following steps: Collect operational data of the wastewater aeration tank. The operational data includes at least dissolved oxygen concentration, saturated dissolved oxygen concentration, oxygen consumption rate, influent flow rate, effective volume of the aeration tank, influent dissolved oxygen concentration, aeration volume, and process variables characterizing influent load or sludge state. The theoretical dissolved oxygen change rate for the current period is calculated based on the dissolved oxygen mass balance relationship in the aeration tank. The physical residual sequence is obtained by subtracting the theoretical dissolved oxygen change rate from the measured dissolved oxygen change rate. Frequency domain analysis was performed on the physical residual sequence to obtain the proportion of high-frequency residual energy; The mode number or penalty factor of variational mode decomposition is adaptively updated based on the proportion of high-frequency residual energy. The updated parameters are then used to perform variational mode decomposition on the dissolved oxygen sequence to obtain multi-scale subsequences. The multi-scale subsequences are aligned and fused with the running data and then input into the prediction model to output the predicted dissolved oxygen value at the target time. The high-frequency residual energy ratio is the ratio of the spectral energy in the preset high-frequency range to the total energy of the entire spectrum after the physical residual sequence has undergone frequency domain transformation. The adaptive update of the modal layer number or penalty factor of variational mode decomposition based on the proportion of high-frequency residual energy includes: setting at least two judgment thresholds; increasing the modal layer number when the proportion of high-frequency residual energy exceeds the first threshold but does not exceed the second threshold; and simultaneously increasing the modal layer number and adjusting the penalty factor or shortening the decomposition window length when the proportion of high-frequency residual energy exceeds the second threshold. The adaptive update also includes a parameter rollback mechanism: when the proportion of high-frequency residual energy is lower than the first threshold within a consecutive preset number of sliding windows, the modal layer number and penalty factor are restored to the basic configuration. The multi-scale subsequence includes at least two of the following: trend subsequence, periodic subsequence, and high-frequency perturbation subsequence.

2. The method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition according to claim 1, characterized in that, The dissolved oxygen mass balance relationship is as follows: ;in, The rate of change of dissolved oxygen concentration. Oxygen transfer coefficient, This represents the saturated dissolved oxygen concentration. Dissolved oxygen concentration, Oxygen consumption rate, This refers to the inlet water flow rate. The effective volume of the aeration tank, This refers to the dissolved oxygen concentration in the influent.

3. The method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition according to claim 1, characterized in that, The prediction model is a neural network model with temporal feature extraction capability. The prediction model integrates channel attention mechanism and temporal attention mechanism. The channel attention mechanism is used to adaptively weigh the contribution of different process variables and modal subsequences, and the temporal attention mechanism is used to locate the time step in the historical sequence that has the highest correlation with the current prediction. The forward propagation process of the prediction model includes: performing global average pooling on the multi-scale feature matrix through a channel attention mechanism, outputting the feature channel weight vector and completing feature recalibration; The recalibrated features are input into a bidirectional long short-term memory network to complete temporal encoding, resulting in a hidden state sequence. The weights of historical time steps are calculated through a temporal attention mechanism, and the hidden states are weighted and summed to obtain a context vector. The context vector is then mapped through a fully connected layer to output the dissolved oxygen prediction value.

4. The method for predicting dissolved oxygen in wastewater aeration tanks based on adaptive mode decomposition according to claim 1, characterized in that, Also includes: Anomaly alerts are generated based on the deviation between the predicted and measured dissolved oxygen values, combined with dynamic statistical thresholds. Alternatively, it can output auxiliary decision-making information for aeration adjustment based on the trend of dissolved oxygen prediction values.

5. A dissolved oxygen prediction system for wastewater aeration tanks, characterized in that, include: The data acquisition module is used to collect operational data of the wastewater aeration tank. The operational data includes at least dissolved oxygen concentration, saturated dissolved oxygen concentration, oxygen consumption rate, influent flow rate, effective volume of the aeration tank, influent dissolved oxygen concentration, aeration rate, and process variables characterizing influent load or sludge state. The physical residual construction module is used to calculate the theoretical dissolved oxygen change rate based on the dissolved oxygen mass balance relationship, and subtract the measured dissolved oxygen change rate from the theoretical dissolved oxygen change rate to obtain the physical residual sequence. The parameter adaptive update module is used to perform frequency domain analysis on the physical residual sequence to obtain the high-frequency residual energy ratio, and update the mode layer number or penalty factor of the variational mode decomposition according to the high-frequency residual energy ratio. The mode decomposition module is used to perform variational mode decomposition on the dissolved oxygen sequence using updated parameters to obtain multi-scale subsequences; the prediction module is used to fuse the multi-scale subsequences with the running data and input them into the prediction model to output the predicted dissolved oxygen value at the target time. The high-frequency residual energy ratio is the ratio of the spectral energy in the preset high-frequency range to the total energy of the entire spectrum after the physical residual sequence has undergone frequency domain transformation. The adaptive update of the modal layer number or penalty factor of variational mode decomposition based on the proportion of high-frequency residual energy includes: setting at least two judgment thresholds; increasing the modal layer number when the proportion of high-frequency residual energy exceeds the first threshold but does not exceed the second threshold; and simultaneously increasing the modal layer number and adjusting the penalty factor or shortening the decomposition window length when the proportion of high-frequency residual energy exceeds the second threshold. The adaptive update also includes a parameter rollback mechanism: when the proportion of high-frequency residual energy is lower than the first threshold within a consecutive preset number of sliding windows, the modal layer number and penalty factor are restored to the basic configuration. The multi-scale subsequence includes at least two of the following: trend subsequence, periodic subsequence, and high-frequency perturbation subsequence.

6. An electronic device comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor, characterized in that, When the processor executes the computer-executable instructions, it implements the method for predicting dissolved oxygen in a wastewater aeration tank based on adaptive mode decomposition as described in any one of claims 1-4.

7. A computer-readable storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed by the processor, they implement the method for predicting dissolved oxygen in a wastewater aeration tank based on adaptive mode decomposition as described in any one of claims 1-4.

Images