Device monitoring system based on digital twinning

Abstract

This invention relates to the field of wastewater treatment and intelligent control technology, specifically a monitoring system for biochemical fluid transport equipment based on digital twins. It includes a multi-source heterogeneous data fusion unit: collecting equipment operation and environmental response parameters, and aligning low-frequency data to a high-frequency time axis using a time-series calibration algorithm; a digital twin energy efficiency coupling unit: loading a biochemical mechanism model to calculate theoretical load, and using a neural network to calculate medium transmission efficiency; a load demand prediction and matching unit: calculating the supply-demand matching degree index, and generating a collaborative control strategy if it exceeds the steady-state range; and an adaptive closed-loop optimization unit: executing the strategy and monitoring deviations, and correcting model weights and dynamic constants in reverse. This invention solves the problem of spatiotemporal decoupling between physical output and biochemical demand, achieving predictive on-demand supply, and ensuring energy saving, consumption reduction, and process stability.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment and intelligent control technology, specifically to a digital twin-based equipment monitoring system. Background Technology

[0002] With the development of industrial internet and intelligent manufacturing technologies, monitoring of fluid transport equipment based on digital twins has become an important research direction in the process industry. In order to achieve effective management and energy efficiency optimization of complex biochemical treatment systems, deep coupling analysis of equipment operating status and environmental response parameters has become particularly important. In digital twin modeling of wastewater treatment, biochemical mechanism models have shown great potential due to their analytical capabilities of reaction processes. However, the operating parameters of physical equipment are typically high-frequency sampled volumetric flow rates, while the demands of biochemical reactions are low-frequency sampled mass loads. This decoupling in the spatiotemporal dimensions makes precise control extremely difficult. Furthermore, existing control systems often neglect media transport efficiency. The dynamic changes of the system rely solely on a single lagging indicator such as dissolved oxygen for feedback, making it difficult to accurately assess how much oxygen in the air is actually utilized by microorganisms. This results in the system being unable to make accurate predictive adjustments under complex operating conditions such as aeration head blockage or changes in sludge properties. Therefore, how to perform time-series calibration on multi-source heterogeneous monitoring data to obtain a standardized state dataset is crucial for ensuring the safe and efficient operation of biochemical treatment systems. This will facilitate the subsequent integration of biochemical mechanisms and neural networks to establish a more accurate energy efficiency coupling model and construct corresponding equipment collaborative control strategies. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a device monitoring system based on digital twins. Specifically, the technical solution of this invention includes: The multi-source heterogeneous data fusion unit is configured to collect the equipment operation parameters of the fluid transport equipment and the environmental response parameters of the controlled environment. It uses a time-series calibration algorithm to align the low-frequency sampled environmental response parameters to the time axis of the high-frequency sampled equipment operation parameters, thereby constructing an equipment-environment status dataset. The digital twin energy efficiency coupling unit is configured to load a preset biochemical mechanism model, analyze the environmental response parameters to calculate the theoretical biochemical demand load; at the same time, it constructs a transmission efficiency dynamic model based on a neural network, inputs the device-environment state dataset into the transmission efficiency dynamic model, and calculates the medium transmission efficiency under the current operating conditions. The load demand prediction and matching unit is configured to calculate the actual effective supply based on the product of the equipment operating parameters and the medium transmission efficiency; calculate the ratio of the actual effective supply to the theoretical biochemical demand load to generate a supply-demand matching index; if the supply-demand matching index is greater than the upper limit of the preset steady-state range or less than the lower limit of the preset steady-state range, then generate an equipment collaborative control strategy. An adaptive closed-loop optimization unit is configured to execute the device collaborative control strategy and continuously monitor the changing trend of the environmental response parameters; calculate the deviation between the changing trend and the model prediction value; if the deviation exceeds a preset tolerance range, then reversely correct the weight parameters of the transmission efficiency dynamic model and update the kinetic constants of the biochemical mechanism model.

[0004] Preferably, the method for data acquisition and processing by the multi-source heterogeneous data fusion unit is configured as follows: The current, voltage, operating frequency, and outlet pressure of the fluid conveying equipment are obtained from the frequency converter of the fluid conveying equipment as operating parameters. Data on biochemical reaction concentration, redox potential, and temperature are obtained from online monitoring instruments in the controlled environment and used as environmental response parameters. The high-frequency sampling period of the device operating parameters and the low-frequency sampling period of the environmental response parameters are identified, and an interpolation algorithm is used to map the low-frequency data to the high-frequency time axis to complete the timing alignment.

[0005] Preferably, the method by which the digital twin energy efficiency coupling unit constructs the dynamic model of transmission efficiency is configured as follows: The exhaust flow rate of the fluid transport equipment, the liquid level depth of the controlled environment, and the sludge concentration are selected as input feature vectors. The neural network model was trained using oxygen transfer efficiency as the output target label. During training, a physical constraint layer is introduced, which uses activation functions or penalty terms to limit the numerical range of the output target label to between zero and one, thus obtaining a dynamic model of transmission efficiency. The medium transport efficiency characterizes the proportion of the physical medium output by the fluid transport equipment that is converted into components usable in a controlled environment.

[0006] Preferably, the method for calculating the theoretical biochemical demand load is configured as follows: A biochemical mechanism model was constructed based on the mathematical model of activated sludge to simulate the growth, decay and substrate degradation processes of microorganisms in digital space. The pollutant concentration and temperature in the environmental response parameters are input into the biochemical mechanism model to calculate the instantaneous oxygen consumption rate required for microbial metabolism at the current moment. Based on the instantaneous oxygen consumption rate and combined with the volume parameters of the controlled environment, the theoretical biochemical load is calculated through time integration.

[0007] Preferably, the load demand forecasting and matching unit is configured to generate the supply-demand matching degree index and determine the status in the following way: Extract the air flow rate value from the equipment operating parameters, multiply it by the medium transmission efficiency, and obtain the actual effective oxygen supply. The ratio of the actual effective oxygen supply to the theoretical biochemical load is calculated and used as the supply-demand matching index. A first threshold and a second threshold are preset, wherein the first threshold is greater than the second threshold, the upper limit of the steady-state interval is the first threshold, and the lower limit is the second threshold; When the supply-demand matching index is greater than the first threshold, a digital marker for the energy waste state is generated. When the supply and demand matching index is less than the second threshold, a digital label for the biochemical treatment risk status is generated.

[0008] Preferably, the collaborative control strategy for the generating device is configured as follows: Based on the time lag characteristics of the biochemical mechanism model, the theoretical biochemical demand load change curve is predicted within a future preset time window. Based on the change curve, the target frequency of the fluid transport equipment and the target opening degree of the regulating valve required to satisfy the steady-state range are calculated using a reverse deduction algorithm. A surge boundary check is performed on the target frequency. If the check passes, the target frequency and the target opening are combined to generate the equipment collaborative control strategy.

[0009] Preferably, the adaptive closed-loop optimization unit corrects the model in the following way: Record the measured values ​​of environmental response parameters after executing the device collaborative control strategy; Residual analysis was performed between the measured values ​​and the predicted values ​​from the biochemical mechanism model. If the residual exceeds the preset tolerance limit, the neuron connection weights of the transmission efficiency dynamic model are adjusted using the backpropagation algorithm. Meanwhile, based on long-term operating data, the yield coefficient and decay coefficient in the biochemical mechanism model are calibrated using the parameter estimation method.

[0010] Preferably, the system further includes: An abnormal operating condition diagnostic unit is configured to monitor the time change rate of the medium transmission efficiency in real time. If it is detected that the decrease in the medium transmission efficiency exceeds a safety threshold within a preset time window when the change in the operating parameters of the equipment is within a preset tolerance range, a fault diagnosis signal is generated to indicate that there is a physical abnormality in the end effector of the fluid transport equipment. In response to the fault diagnosis signal, a maintenance early warning data packet is generated and a blocking command is triggered. The blocking command is used to prevent the adaptive closed-loop optimization unit from executing the parameter correction process, so as to prevent abnormal data from updating the model parameters.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. This system achieves high-precision time alignment of second-level equipment operating parameters with minute-level environmental response parameters through the application of multi-source heterogeneous data fusion units and cubic spline interpolation algorithms. This eliminates the data step effect caused by asynchronous sampling frequencies and provides smooth and time-synchronized feature inputs for model training. On this basis, a dynamic model of transmission efficiency is constructed using a neural network with embedded physical constraint layers. This successfully maps the hard index of volumetric flow rate in the physical world to the soft index of quality demand in the biochemical world. This solves the calculation error caused by ignoring the dynamic changes in medium transmission efficiency in traditional control. It can keenly capture the decline in mass transfer efficiency caused by changes in sludge concentration or aging of aeration heads, thereby realizing real-time soft measurement of oxygen utilization and avoiding the lag and illusion caused by feedback adjustment based solely on dissolved oxygen. 2. This system achieves quantitative assessment of microbial metabolic oxygen demand by constructing a refined biochemical mechanism model. This model not only distinguishes the kinetic differences between heterotrophic and autotrophic bacteria, but also introduces active biomass fraction correction and dynamic temperature compensation based on the Arrhenius equation, significantly improving the load prediction accuracy under different seasons and water quality conditions. Combined with the calculation of the supply-demand matching index and the setting of the dual-threshold steady-state interval, this invention transforms the complex biochemical control problem into a standardized interval control problem, which can intuitively quantify the current supply-demand balance state, clearly distinguish between energy waste and process risks, and thus guide the system to achieve maximum energy saving and consumption reduction while ensuring the quality of biochemical treatment. 3. This system adopts a feedforward predictive control strategy based on time lag characteristics, effectively overcoming the large inertia of biochemical reactions. By predicting the theoretical biochemical demand load change curve within a future time window, and using a back-calculation algorithm and pipeline resistance characteristic equation, it accurately calculates the target frequency of fluid transport equipment and the target opening of regulating valves required to meet the steady-state range. At the same time, the system's built-in surge boundary verification logic ensures that the equipment always operates in a safe area, avoiding equipment failures caused by blind adjustments. This shift from passive response to predictive on-demand supply ensures the timeliness and accuracy of the control strategy in dealing with influent load fluctuations, achieving the dual goals of process stability and equipment safety. 4. This system establishes a dual guarantee mechanism of adaptive closed-loop optimization and abnormal operating condition diagnosis. On the one hand, by monitoring the residual changes of environmental response parameters, the system can use the backpropagation algorithm to fine-tune the weights of the transmission efficiency model to cope with short-term fluctuations, and use the parameter estimation method to calibrate the kinetic constants of the biochemical mechanism model to adapt to long-term changes in sludge properties, ensuring the model's continued viability. On the other hand, the abnormal operating condition diagnosis unit can accurately identify physical structural faults and model algorithm errors by monitoring the rate of change of medium transmission efficiency and the stability of equipment operating parameters in real time. When a physical anomaly is detected, it triggers a blocking command to prevent erroneous data from contaminating the core model parameters, thereby ensuring the robustness and reliability of the digital twin system. Attached Figure Description

[0012] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0014] Example 1: Please see Figure 1 The equipment monitoring system based on digital twins includes: a multi-source heterogeneous data fusion unit, configured to collect equipment operating parameters of fluid conveying equipment and environmental response parameters of the controlled environment, and use a time-series calibration algorithm to align the low-frequency sampled environmental response parameters to the time axis of the high-frequency sampled equipment operating parameters to construct an equipment-environment status dataset; The digital twin energy efficiency coupling unit is configured to load a preset biochemical mechanism model, analyze environmental response parameters to calculate theoretical biochemical demand load; at the same time, it constructs a transmission efficiency dynamic model based on a neural network, inputs the equipment-environment status dataset into the transmission efficiency dynamic model, and calculates the medium transmission efficiency under the current operating conditions. The load demand forecasting and matching unit is configured to calculate the actual effective supply based on the product of equipment operating parameters and medium transmission efficiency; calculate the ratio of the actual effective supply to the theoretical biochemical demand load to generate a supply-demand matching index; if the supply-demand matching index is greater than the upper limit of the preset steady-state range or less than the lower limit of the preset steady-state range, a coordinated equipment control strategy is generated. The adaptive closed-loop optimization unit is configured to execute the equipment collaborative control strategy and continuously monitor the changing trend of environmental response parameters; calculate the deviation between the changing trend and the model prediction value; if the deviation exceeds the preset tolerance range, it reversely corrects the weight parameters of the transmission efficiency dynamic model and updates the kinetic constants of the biochemical mechanism model.

[0015] This embodiment elaborates on the overall architecture and core logic of the above system. The system aims to solve the problem of spatiotemporal decoupling between the output of physical equipment and the demand for biochemical reactions by constructing a biochemical-mechanical coupled digital twin. The multi-source heterogeneous data fusion unit performs data acquisition and alignment tasks. This unit collects the operating parameters of fluid transport equipment, such as magnetic levitation blowers or centrifugal blowers, in real time at a sampling frequency of seconds. At the same time, it collects the environmental response parameters of the controlled environment, namely the biological aeration tank, at a sampling frequency of minutes. This unit utilizes a built-in time-series calibration algorithm to interpolate and fit low-frequency sampled environmental response parameters, aligning them to the time axis of high-frequency sampled equipment operating parameters, thereby constructing a time-synchronized equipment-environment status dataset. Based on this, the digital twin energy efficiency coupling unit runs two models in parallel: on one hand, the biochemical mechanism model loads a pre-defined activated sludge mathematical model to analyze environmental response parameters such as chemical oxygen demand (COD), ammonia nitrogen, and temperature, calculating the theoretical biochemical load required for microbial metabolism at the current moment; on the other hand, the transmission efficiency dynamic model, built on a neural network, uses the equipment-environment status dataset as input to calculate the media transmission efficiency under the current operating conditions. This efficiency characterizes the proportion of air output by the equipment that is actually absorbed by the mixed liquor. The load demand forecasting and matching unit quantifies the supply-demand balance and calculates the supply-demand matching degree index using the following formula:

[0016] in, The supply-demand matching index, derived from calculations, physically represents the degree of balance between current oxygen supply and demand; it is dimensionless. Actual effective supply, derived from model calculations, is expressed in kilograms per hour. Theoretical biochemical load, derived from biochemical mechanism models, is expressed in kilograms per hour. The equipment's medium flow rate, i.e., air flow rate, is obtained from sensor data and is measured in cubic meters per hour. This symbol is used here for consistency with subsequent implementation examples. Air density, derived from standard constants or actual measurements, is expressed in kilograms per cubic meter. The oxygen mass fraction is derived from a constant setting such as 0.23, which is dimensionless. Medium transmission efficiency is derived from a dynamic model of transmission efficiency; in response to A value greater than the upper limit of the preset steady-state range indicates energy waste, or a response to... If the value is less than the lower limit, it indicates that there is a process risk. This unit generates a collaborative control strategy for the equipment, which includes adjusting the fan frequency or valve opening. The adaptive closed-loop optimization unit executes the above strategy and forms a closed loop, continuously monitoring the changing trend of environmental response parameters after execution, and calculating the deviation between this trend and the model prediction. ; in response If the current model is found to be inaccurate, the system will correct the weight parameters of the dynamic model of transmission efficiency in reverse to cope with short-term fluctuations, and update the kinetic constants of the biochemical mechanism model to cope with long-term changes in sludge properties. This embodiment introduces media transmission efficiency. As an intermediate variable, and combined with density and mass fraction corrections, this scheme successfully deeply couples the hard indicator of volumetric flow rate in the physical world with the soft indicator of mass demand in the biochemical world. This approach corrects the calculation errors caused by considering only volumetric flow rate while neglecting mass conversion, and avoids the lag and illusions caused by traditional control methods that rely solely on dissolved oxygen for feedback adjustment. For example, this can occur when aeration head blockage leads to… When pressure is reduced, the system can identify the efficiency decline and issue a maintenance warning instead of blindly increasing the pressure. This realizes the transformation from passive response to predictive on-demand supply in the wastewater treatment scenario, ensuring the dual goals of energy saving and consumption reduction as well as process stability.

[0017] Example 2: The configuration for data acquisition and processing by the multi-source heterogeneous data fusion unit is as follows: The current, voltage, operating frequency, and outlet pressure of the fluid conveying equipment are obtained from the frequency converter of the fluid conveying equipment as operating parameters. Data on biochemical reaction concentration, redox potential, and temperature are obtained from online monitoring instruments in the controlled environment and used as environmental response parameters. The high-frequency sampling period of the equipment operating parameters and the low-frequency sampling period of the environmental response parameters are identified. An interpolation algorithm is used to map the low-frequency data to the high-frequency time axis to complete the time alignment.

[0018] This embodiment details the data acquisition and timing alignment process; the system directly reads the equipment operating parameter vector from the frequency converter of the fluid conveying equipment via Modbus or Profibus communication protocols. Specifically, including current ,Voltage Operating frequency and export pressure Simultaneously, the system acquires environmental response parameter vectors from online monitoring instruments installed in the biochemical pool. Specifically, this includes biochemical reaction concentration, i.e., dissolved oxygen concentration. Oxidation-reduction potential and the temperature of the mixture ; The system identifies the high-frequency sampling period of the equipment operating parameters. Low-frequency sampling period of environmental response parameters There are differences; to distinguish it from the symbolic definition of the physical quantity temperature, the Greek letter is used here. The sampling period, representing the time dimension, is used to align the data. For the time axis, the system employs a cubic spline interpolation algorithm; for any two low-frequency sampling points and High frequency moments between The interpolation calculation is as follows:

[0019] in, :time The estimated environmental parameters are derived from interpolation calculations. Adjacent low-frequency sampling times are derived from the original data timestamps. The coefficients of the cubic spline interpolation polynomial are, in physical terms, the weights of the cubic, quadratic, linear, and constant terms used to describe the shape of the parameter variation curve in the k-th time sub-interval. These coefficients are not arbitrarily set, but are obtained by solving a system of linear equations based on the boundary conditions of continuous function values, continuous first derivatives, and continuous second derivatives at the interpolation nodes. To ensure that those skilled in the art can implement this algorithm, the specific coefficient calculation logic is disclosed here: defining the step size. and the second derivative at the node Then the formula for calculating each coefficient is: , , , ;in, Let be the second derivative of the curve at the nodes, i.e., the curvature, which is obtained by solving the following tridiagonal linear equations: ,for Simultaneously, the natural boundary conditions are set. Solve using the chasing method The interpolation coefficients can be determined by back-substituting the sequence; this method, rather than simple statistical regression, ensures the smoothness of the interpolation curve at the second derivative level and avoids high-frequency noise introduced by data alignment. Through this algorithm, the system fills the data gaps during the sampling interval of the biochemical sensor. This embodiment utilizes a high-precision interpolation algorithm to eliminate the data step effect caused by asynchronous sampling frequencies. This enables the subsequent neural network model to obtain smooth, continuous, and time-aligned input features, significantly improving the convergence speed and prediction accuracy of model training. In particular, when dealing with biochemical reaction processes with large inertia, it ensures the control system's ability to keenly capture environmental changes.

[0020] Example 3: The method for constructing a dynamic model of transmission efficiency using a digital twin energy efficiency coupling unit is configured as follows: The exhaust flow rate of the fluid transport equipment, the liquid level depth of the controlled environment, and the sludge concentration are selected as input feature vectors. The neural network model was trained using oxygen transfer efficiency as the output target label. During training, a physical constraint layer is introduced, which uses activation functions or penalty terms to limit the numerical range of the output target label to between zero and one, thus obtaining a dynamic model of transmission efficiency. Medium transport efficiency characterizes the proportion of physical medium output from a fluid transport device that is converted into components usable in a controlled environment.

[0021] This embodiment details the construction process of the dynamic model for oxygen transfer efficiency, which serves as a bridge connecting physical devices and the biochemical environment. The system performs feature engineering to select the three key variables that have the greatest impact on the oxygen mass transfer process to form the input feature vector. This includes exhaust flow rate, which reflects the degree of gas phase turbulence. Liquid level depth reflects the residence time of air bubbles and hydrostatic pressure. And sludge concentration, which reflects the viscosity of the liquid phase and its effect on hindering air bubbles. ; A multilayer perceptron neural network was constructed. To balance online computational efficiency and nonlinear fitting accuracy, the network was configured with two hidden layers: the first hidden layer had 16 neurons, and the second hidden layer had 8 neurons, both using the ReLU activation function to accelerate convergence. The model was trained using the Adam optimizer with a learning rate of 0.001 and mean squared error as the loss function. The target label in the training set, historical oxygen transfer efficiency, was denoted as... To distinguish it from the model's predicted output Its calculations are derived from the inversion of historical data, specifically based on the oxygen mass balance equation: at steady state, using the formula... Mark, where, Defined as the effective volume of a controlled environment, expressed in cubic meters; Defined as the specific oxygen consumption rate of activated sludge, based on volume, in grams per cubic meter per hour. ; Special note here regarding dimensional consistency: molecular part The original calculation results are in grams per hour (g / h), and the coefficients are... This is used to convert the grams per hour to kilograms per hour (kg / h), thus relating it to the denominator. To ensure that the kilograms per hour produced are kept consistent in unit, A dimensionless ratio; for the training dataset To overcome the loop logic defects caused by relying on the model to be calibrated and to construct an effective supervised learning dataset, this embodiment adopts an offline physicochemical analysis anchoring method independent of the online model: It prioritizes selecting time windows containing laboratory respiratory spectrum measurement records in historical time series, and directly extracts the measured oxygen consumption rate (OUR) as... The absolute truth value; in periods lacking direct OUR data, calculations are performed based on the mass conservation principle of measured pollutant concentrations in the influent and effluent, i.e., using the formula

[0022] Obtain certainty Data, among which, To observe the yield coefficient, thereby ensuring that the training labels are derived from objective physical facts rather than simulation estimates; Defined as the rate of change of dissolved oxygen concentration, its physical meaning is the instantaneous derivative of dissolved oxygen concentration in the mixture with time, characterizing the rate of accumulation or consumption of oxygen in the liquid phase, with units of milligrams per liter per hour. Regarding the specific method for obtaining this parameter, to overcome the interference of sensor signal noise on differential calculations, this embodiment is configured as follows: extracting the historical dissolved oxygen concentration data sequence within a preset time window (e.g., 5 minutes) before and after the current moment, performing linear fitting on the sequence using the least squares method, and extracting the slope of the fitted line as the value of dissolved oxygen at the current moment. The numerical calculation method ensures the robustness of training label generation and avoids numerical oscillations caused by direct differencing. To resolve the conflict between training objectives and model capabilities, and to prevent purely data-driven models from producing outputs that violate physical principles, the system introduces a physical constraint layer at the output layer, specifically employing a variant of the Sigmoid activation function:

[0023] in, Medium transport efficiency, or model prediction, is derived from the model output and its physical meaning is oxygen utilization rate. The upper limit scaling factor for model output is set to 1.0 (100%) in this embodiment; it is not rigidly set to the physical limit of 0.55 here because of the following considerations: historical training labels. Due to sensor noise or non-steady-state fluctuations, there may be instantaneous calculated values ​​exceeding 0.55, such as 0.6. If the model output is rigidly truncated to 0.55, the model will be unable to fit these high-value samples, causing gradient vanishing or training bias. Therefore, this embodiment is configured to: set... To cover the full probability space, but in the subsequent control policy generation stage, as in the example, the effective efficiency value is logically clamped at 0.55, thereby ensuring the convergence of model training while satisfying physical constraints. The linearly weighted output value of the last layer of the neural network is derived from the calculations of the previous layers and is dimensionless. This embodiment achieves high transmission efficiency for media that is difficult to measure directly online by combining soft measurement technology with a physically constrained neural network. The model provides real-time estimation; the introduction of a physical constraint layer eliminates outliers caused by model illusions, ensuring the safety of the control strategy. The model can keenly capture the decline in mass transfer efficiency caused by changes in sludge concentration or aging of aerator heads, providing a reliable basis for accurate energy efficiency assessment.

[0024] Example 4: The method for calculating theoretical biochemical load is configured as follows: A biochemical mechanism model was constructed based on the mathematical model of activated sludge to simulate the growth, decay and substrate degradation processes of microorganisms in digital space. The pollutant concentration and temperature in the environmental response parameters are input into the biochemical mechanism model to calculate the instantaneous oxygen consumption rate required for microbial metabolism at the current moment. Based on the instantaneous oxygen consumption rate and combined with the volume parameters of the controlled environment, the theoretical biochemical load is calculated through time integration.

[0025] This embodiment constructs a biochemical mechanism model based on the International Water Association's activated sludge model, aiming to quantify oxygen demand at the microscopic level; the system specifically distinguishes pollutant concentrations in environmental response parameters as carbon source organic matter concentrations. Corresponding COD and ammonia nitrogen concentrations Corresponding to NH3-N, and temperature Input Model; To address the issue that a single total reaction cannot distinguish the differences in oxygen demand among different bacterial communities, the system employs a Monod kinetic equation with dual substrates to calculate the instantaneous oxygen consumption rate at the current moment. The complete calculation logic is as follows:

[0026] Specifically, it can be elaborated as follows:

[0027]

[0028]

[0029] in, Total instantaneous oxygen consumption rate, expressed in milligrams of oxygen per liter per hour. : Yield coefficients of 0.67 for heterotrophic bacteria and 0.24 for autotrophic bacteria, respectively, derived from the parameters of the ASM1 standard model. The oxygen equivalent constant for nitration, i.e., 4.57g of oxygen is required to oxidize 1g of ammonia nitrogen. Real-time pollutant concentration, derived from sensors. The half-saturation constant, to ensure the reproducibility of the model, is adopted directly from the standard value recommended by the International Water Association's ASM1 model in this embodiment: The half-saturation constant for heterotrophic bacteria was set to 20.0. , That is, the half-saturation constant of nitrifying bacteria is set to 1.0. Mixed liquor sludge concentration (MLSS) is obtained from online instruments and is expressed in mg / L. The active biomass fraction, set to 0.65 in this embodiment, is used to adjust the total sludge concentration measured by physical analysis. Conversion into the concentration of active microorganisms participating in biochemical reactions ,Right now This clarifies the mapping relationship between sensor physical quantities and model state variables, thus resolving the technical defect that directly using MLSS leads to an overestimation of the reaction rate calculation. Temperature The modified maximum specific growth rate of heterotrophic to autotrophic bacteria; To meet the model's reproducibility requirements across different seasons, this embodiment explicitly employs the Arrhenius equation for dynamic temperature compensation, with the specific calculation formula configured as follows: The system's built-in parameter set includes: heterotrophic bacteria temperature coefficient. Temperature coefficient of autotrophic bacteria Characterizing nitrifying bacteria to be more sensitive to low temperatures, baseline rate Take 6.0 , Take 0.8 The definition of this explicit function ensures that the input parameters... Able to be actually consumed by mathematical models The temperature-corrected attenuation coefficient also follows the same principle. Correction logic Based on instantaneous oxygen consumption rate Combined with the effective volume of the controlled environment, i.e., the biological treatment tank The time integral calculation for the implementation example: defining the integral control period. For example, over 15 minutes, the cumulative oxygen demand is calculated using the trapezoidal numerical integral formula and converted into an equivalent average load. :

[0030] This embodiment eliminates the load estimation bias caused by the general calculation of traditional models by refining the bacterial community classification calculation and introducing activity fraction correction, and in particular improves the prediction accuracy of nitrification oxygen demand under low C / N ratio water quality conditions.

[0031] Example 5: The method for the load demand forecasting and matching unit to generate the supply-demand matching degree index and determine the status is configured as follows: Extract the air flow rate value from the equipment operating parameters, multiply it by the medium transmission efficiency, and obtain the actual effective oxygen supply. The ratio of actual effective oxygen supply to theoretical biochemical demand load is calculated and used as the supply-demand matching index. A first threshold and a second threshold are preset, wherein the first threshold is greater than the second threshold, the upper limit of the steady-state interval is the first threshold, and the lower limit is the second threshold; When the supply and demand matching index is greater than the first threshold, a digital marker for the state of energy waste is generated. When the supply and demand matching index is less than the second threshold, a digital label for the risk status of biochemical treatment is generated.

[0032] This embodiment details the generation and status determination process of the supply-demand matching index; the system extracts airflow values ​​from equipment operating parameters. Multiply it by the medium transmission efficiency calculated from the model in Example 3. And introduce the mass fraction constant of oxygen in the air. and air density To obtain the actual effective oxygen supply :

[0033] System calculates ratio As a supply-demand matching index, it corresponds to the symbol in the implementation of Example 1. Maintain uniformity; based on this, the system presets a steady-state range. ,in, The first threshold is set at 1.2, which allows for 20% excess redundancy. The second threshold is set to 1.05, ensuring a safety margin of at least 5%; the system executes the status determination logic: in response to... The system generates an energy waste status flag, indicating that the fan output is too high and oxygen is not fully utilized before escaping to the water surface; in response to The system generates a biochemical treatment risk status marker, indicating insufficient oxygen supply, which may lead to inhibition of nitrification. This embodiment defines a dimensionless matching degree index. The dual threshold interval simplifies complex biochemical control problems into standardized interval control problems. This quantitative evaluation method intuitively reflects the operational health of the system and clarifies whether the optimization direction is energy saving or quality maintenance, greatly reducing the decision-making difficulty for operation and maintenance personnel.

[0034] Example 6: The method for generating device collaborative control strategies is configured as follows: Based on the time lag characteristics of the biochemical mechanism model, the theoretical biochemical demand load change curve is predicted within a future preset time window. Based on the change curve, the target frequency of the fluid transport equipment and the target opening of the regulating valve required to meet the steady-state range are calculated using the reverse deduction algorithm. A surge boundary check is performed on the target frequency. If the check passes, the target frequency and target opening are combined to generate a collaborative control strategy for the equipment.

[0035] This embodiment employs a feedforward predictive control strategy to address the time lag characteristics of biochemical reactions. Using a biochemical mechanism model, the current pollutant concentration gradient is input to predict future preset time windows. Theoretical biochemical demand load variation curve within During the prediction process, the system employs a linear trend extrapolation method for the input boundary conditions within the future time window, thereby providing the necessary input parameters for solving the differential equation system; based on the predicted... Set target matching degree Given the median of the steady-state range, the required target airflow is calculated using a backpropagation algorithm. :

[0036] This problem is solved using a fixed-point iterative algorithm. and The coupling problem, until the flow converges; in order to To accurately map this to equipment control parameters, the system must determine the operating pressure point at that flow rate to cope with changes in pipeline characteristics under varying operating conditions. The system introduces the pipeline resistance characteristic equation to calculate the expected discharge pressure at the target flow rate. :

[0037] in, The current liquid level depth in the controlled environment, derived from a sensor, represents the hydrostatic pressure. Pipeline resistance coefficient, derived from the system commissioning phase... Quadratic fitting of data to characterize friction resistance Local resistance loss constant of aerator and valve Based on the determined operating point The system performs precise analysis of the target frequency: To meet the patent law's requirement of sufficient disclosure, this embodiment clearly defines the specific mapping path from flow rate and pressure to frequency, rather than merely stating the purpose of table lookup; the system pre-loads a frequency response surface function fitted based on variable operating condition data collected during the factory standard performance test or on-site commissioning phase of the fluid transport equipment. The function is specifically constructed using a bivariate quadratic polynomial form:

[0038] in, The equipment characteristic coefficients were obtained by least squares regression using the above test data as a sample set. Calculated and Substituting into this analytical expression, the uniquely determined target frequency can be obtained. This avoids the control uncertainty caused by interpolation errors in traditional table lookup methods; the specific collaborative logic follows an energy-saving strategy of frequency conversion priority and full valve opening; regarding the calculation of the target opening of the regulating valve, in order to overcome the black-box defect of back-reasoning, this embodiment constructs a clear equal percentage flow characteristic model: calculating the required flow coefficient ,in, This is the system back pressure; The solution is obtained using the aperture equation, and the specific formula is as follows:

[0039] in, This is a typical value for the valve's adjustable ratio, such as... , The coefficients are all open; this system of algebraic equations guarantees a unique solution for the control strategy. Based on this, the system must... Safety checks are performed; the system stores the surge boundary equations for the wind turbine. ; in response to Current exhaust pressure Approaching surge pressure When the safety margin is less than 10%, the system triggers a safety correction logic: based on the surge boundary equation, the pressure is solved inversely. Minimum safe flow rate And apply a safety factor of 1.1; target traffic Force reset to Then, by substituting the valve flow characteristic equation back into the equation, the corrected target opening of the regulating valve is obtained through reverse calculation. This ensures that the fluid transport equipment always operates in the safe zone on the right, ultimately generating a collaborative control strategy for the equipment.

[0040] Example 7: The adaptive closed-loop optimization unit is configured to correct the model as follows: Record the measured values ​​of environmental response parameters after the execution of the equipment collaborative control strategy; Residual analysis was performed between the measured values ​​and the predicted values ​​from the biochemical mechanism model; If the residual exceeds the preset tolerance limit, the neuron connection weights of the transmission efficiency dynamic model are adjusted using the backpropagation algorithm. Meanwhile, based on long-term operating data, the yield coefficient and decay coefficient in the biochemical mechanism model were calibrated using the parameter estimation method.

[0041] This embodiment constructs a two-layer correction mechanism to combat model drift caused by environmental changes; the system records the measured values ​​of environmental response parameters after the control strategy is executed. For example, the actual dissolved oxygen value, and compare it with the predicted value of the biochemical mechanism model. Compare and calculate the residuals. Perform short-term weight adjustments in response to If the preset tolerance limit is exceeded and sensor malfunction has been ruled out, the system determines that the transmission efficiency model is inaccurate and uses the backpropagation algorithm to... The loss function represents the neuron connection weights in the dynamic model of transmission efficiency. Make fine adjustments to adjust its output. This better reflects current working conditions; to ensure mathematical reproducibility, the gradient propagation path for backpropagation is explicitly defined here: since the dissolved oxygen concentration is directly observed... Rather than efficiency The system constructs a gradient conversion factor based on the oxygen mass balance equation, namely...

[0042] in, Based on the oxygen mass balance equation at discrete time step The incremental form is derived from the above. Specifically, it will be... Approximately After sorting, we get about The partial derivatives are:

[0043] This conversion factor maps the dissolved oxygen error gradient to the output gradient of the efficiency model, and then updates the neural network weights through the chain rule; simultaneously, it performs long-term parameter calibration: to concretize the parameter estimation method, this embodiment constructs a nonlinear least squares optimization problem; the system defines the objective function. The model simulates pollutant concentrations over the past 7 days. Compared with sensor measured values The sum of the mean square errors between them:

[0044] The system uses the Levenberg-Marquardt algorithm (LMA) as the solver, within a preset biological constraint range, for example... Internal iterative search for the optimal yield coefficient With attenuation coefficient , so that the objective function Minimization; this process transforms abstract calibration into a computable mathematical optimization process, ensuring that model parameters are adaptively updated as seasonal environmental changes occur.

[0045] Example 8: The system also includes: The abnormal operating condition diagnostic unit is configured to monitor the time change rate of media transmission efficiency in real time. If it is detected that the decrease in medium transmission efficiency exceeds the safety threshold within a preset time window when the change in equipment operating parameters is within the preset tolerance range, a fault diagnosis signal is generated to indicate that there is a physical abnormality in the end actuator of the fluid conveying equipment. In response to a fault diagnosis signal, a maintenance warning data packet is generated and a blocking command is triggered. The blocking command is used to prevent the adaptive closed-loop optimization unit from executing the parameter correction process to prevent abnormal data from updating the model parameters.

[0046] This embodiment introduces an abnormal operating condition diagnostic unit to distinguish between model errors and physical faults. It should be specifically noted here that, to avoid logical paradoxes, the media transmission efficiency monitored by this unit specifically refers to the real-time inversion efficiency. This is to distinguish it from the predicted transmission efficiency generated by a neural network in Example 3. This value is derived from the oxygen mass balance formula described in Example 3. It is obtained by reverse calculation based on real-time sensor data; It is worth noting that, in order to avoid the risk of misjudgment caused by circular dependencies, i.e., model estimation errors leading to false reports of physical faults, the parameters in this formula... Instead of directly calling the instantaneous output of the biochemical mechanism model, it calls the steady-state SOUR value after low-pass filtering, such as the moving average of the past 4 hours. This processing is based on the time-scale separation principle that the mutation rate of biological response characteristics is on a daily scale, while the mutation rate of physical structural failures is on a second scale. This ensures that even if the model has high-frequency oscillation errors, it will not affect the diagnostic accuracy of sudden efficiency drops in physical facilities. Similarly, the formula contains... Conversion factors are used to ensure that the calculation result is a dimensionless ratio; The system calculates the rate of change of the inversion efficiency in real time. The system executes fault diagnosis logic, simultaneously monitoring two conditions: Condition A is that the variation range of equipment operating parameters such as flow rate and frequency is within the preset tolerance range, indicating that the fan itself is operating smoothly and the input characteristics of the neural network have not undergone abrupt changes, theoretically the model prediction value should remain constant; Condition B is the inversion efficiency. If the decrease exceeds a safety threshold within a preset time window, for example, a decrease of 30%; in response to the simultaneous fulfillment of the above combined conditions, the system determines that there is a physical structural abnormality, that is, the input remains unchanged but the physical response of the output deteriorates drastically, and generates a fault diagnosis signal characterizing that there is a physical abnormality in the end actuator of the fluid conveying equipment, such as a rupture of the aeration pipe network; in response to the signal, the system immediately generates a maintenance warning data packet and pushes it to the operation and maintenance personnel, and triggers a blocking command; the blocking command forcibly prohibits the adaptive closed-loop optimization unit of Embodiment 7 from executing the parameter correction process; This embodiment serves as an anti-virus measure, preventing data contamination caused by physical faults. When a physical fault occurs, efficiency drops drastically. If learning is not interrupted at this time, the closed-loop optimization unit will mistakenly believe that the model parameters are deviated, and thus forcibly adjust the weights to adapt to the incorrect physical phenomenon. This solution protects the purity and reliability of the core algorithm model by identifying physical anomalies and cutting off the learning path.

[0047] 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.

Claims

1. A device monitoring system based on digital twins, characterized in that, include: The multi-source heterogeneous data fusion unit is configured to collect the equipment operation parameters of the fluid transport equipment and the environmental response parameters of the controlled environment. It uses a time-series calibration algorithm to align the low-frequency sampled environmental response parameters to the time axis of the high-frequency sampled equipment operation parameters, thereby constructing an equipment-environment status dataset. The digital twin energy efficiency coupling unit is configured to load a preset biochemical mechanism model, analyze the environmental response parameters to calculate the theoretical biochemical demand load; at the same time, it constructs a transmission efficiency dynamic model based on a neural network, inputs the device-environment state dataset into the transmission efficiency dynamic model, and calculates the medium transmission efficiency under the current operating conditions. The load demand forecasting and matching unit is configured to calculate the actual effective supply based on the product of the equipment operating parameters and the medium transmission efficiency. Calculate the ratio of the actual effective supply to the theoretical biochemical demand load to generate a supply-demand matching index; If the supply and demand matching index is greater than the upper limit of the preset steady-state range or less than the lower limit of the preset steady-state range, a device collaborative control strategy is generated. An adaptive closed-loop optimization unit is configured to execute the device collaborative control strategy and continuously monitor the changing trend of the environmental response parameters; The deviation between the changing trend and the model prediction is calculated. If the deviation exceeds a preset tolerance range, the weight parameters of the transmission efficiency dynamic model are corrected in reverse, and the kinetic constants of the biochemical mechanism model are updated.

2. The equipment monitoring system based on digital twins according to claim 1, characterized in that, The multi-source heterogeneous data fusion unit is configured to collect and process data in the following way: The current, voltage, operating frequency, and outlet pressure of the fluid conveying equipment are obtained from the frequency converter of the fluid conveying equipment as operating parameters. Data on biochemical reaction concentration, redox potential, and temperature are obtained from online monitoring instruments in the controlled environment and used as environmental response parameters. The high-frequency sampling period of the device operating parameters and the low-frequency sampling period of the environmental response parameters are identified, and an interpolation algorithm is used to map the low-frequency data to the high-frequency time axis to complete the timing alignment.

3. The equipment monitoring system based on digital twins according to claim 1, characterized in that, The method for constructing the dynamic model of transmission efficiency using the digital twin energy efficiency coupling unit is configured as follows: The exhaust flow rate of the fluid transport equipment, the liquid level depth of the controlled environment, and the sludge concentration are selected as input feature vectors. The neural network model was trained using oxygen transfer efficiency as the output target label. During training, a physical constraint layer is introduced, which uses activation functions or penalty terms to limit the numerical range of the output target label to between zero and one, thus obtaining a dynamic model of transmission efficiency. The medium transport efficiency characterizes the proportion of the physical medium output by the fluid transport equipment that is converted into components usable in a controlled environment.

4. The equipment monitoring system based on digital twins according to claim 1, characterized in that, The method for calculating the theoretical biochemical demand load is configured as follows: A biochemical mechanism model was constructed based on the mathematical model of activated sludge to simulate the growth, decay and substrate degradation processes of microorganisms in digital space. The pollutant concentration and temperature in the environmental response parameters are input into the biochemical mechanism model to calculate the instantaneous oxygen consumption rate required for microbial metabolism at the current moment. Based on the instantaneous oxygen consumption rate and combined with the volume parameters of the controlled environment, the theoretical biochemical load is calculated through time integration.

5. The equipment monitoring system based on digital twins according to claim 1, characterized in that, The method for the load demand forecasting and matching unit to generate the supply-demand matching degree index and determine the status is configured as follows: Extract the air flow rate value from the equipment operating parameters, multiply it by the medium transmission efficiency, and obtain the actual effective oxygen supply. The ratio of the actual effective oxygen supply to the theoretical biochemical load is calculated and used as the supply-demand matching index. A first threshold and a second threshold are preset, wherein the first threshold is greater than the second threshold, the upper limit of the steady-state interval is the first threshold, and the lower limit is the second threshold; When the supply-demand matching index is greater than the first threshold, a digital marker for the energy waste state is generated. When the supply and demand matching index is less than the second threshold, a digital label for the biochemical treatment risk status is generated.

6. The equipment monitoring system based on digital twins according to claim 5, characterized in that, The method for configuring the collaborative control strategy of the generating device is as follows: Based on the time lag characteristics of the biochemical mechanism model, the theoretical biochemical demand load change curve is predicted within a future preset time window. Based on the change curve, the target frequency of the fluid transport equipment and the target opening degree of the regulating valve required to satisfy the steady-state range are calculated using a reverse deduction algorithm. A surge boundary check is performed on the target frequency. If the check passes, the target frequency and the target opening are combined to generate the equipment collaborative control strategy.

7. The equipment monitoring system based on digital twins according to claim 1, characterized in that, The adaptive closed-loop optimization unit is configured to correct the model as follows: Record the measured values ​​of environmental response parameters after executing the device collaborative control strategy; Residual analysis was performed between the measured values ​​and the predicted values ​​from the biochemical mechanism model. If the residual exceeds the preset tolerance limit, the neuron connection weights of the transmission efficiency dynamic model are adjusted using the backpropagation algorithm. Meanwhile, based on long-term operating data, the yield coefficient and decay coefficient in the biochemical mechanism model are calibrated using the parameter estimation method.

8. The equipment monitoring system based on digital twins according to claim 1, characterized in that, The system also includes: An abnormal operating condition diagnostic unit is configured to monitor the time change rate of the medium transmission efficiency in real time. If it is detected that the decrease in the medium transmission efficiency exceeds a safety threshold within a preset time window when the change in the operating parameters of the equipment is within a preset tolerance range, a fault diagnosis signal is generated to indicate that there is a physical abnormality in the end effector of the fluid transport equipment. In response to the fault diagnosis signal, a maintenance early warning data packet is generated and a blocking command is triggered. The blocking command is used to prevent the adaptive closed-loop optimization unit from executing the parameter correction process, so as to prevent abnormal data from updating the model parameters.

Images