Sewage treatment plant full-process simulation system based on digital twinning

Abstract

The invention discloses a sewage treatment plant full-process simulation system based on digital twinning, and relates to the technical field of digital twinning, a three-dimensional digital twinning entity is constructed and integrated with an ASM model, a dynamic parameter calibration module is utilized to calculate biochemical reaction sensitivity factors, temperature, water inlet load and water outlet component characteristics are comprehensively considered, and a simulation result is obtained. The self-adaptive updating of the highly coupled parameters is realized, and the limitations of high calibration difficulty and weak model robustness under complex working conditions are solved; a whole-process feedback deviation index is calculated through a feedback loop error correction module, an Euler constant power function is utilized to quantitatively characterize and eliminate a cascade amplification effect of errors in a backflow loop, and a simulation robustness score and a multi-stage early warning scheduling mechanism are combined, so that nonlinear drift and whole-process prediction distortion caused by chain reaction are effectively inhibited, and the prediction accuracy is improved. And the reliability of simulation prediction in the whole sewage treatment process and the safety of automatic control decision are greatly improved.

Description

A full-process simulation system for wastewater treatment plants based on digital twins Technical Field

[0001] This invention relates to the field of digital twin technology, specifically to a full-process simulation system for a wastewater treatment plant based on digital twins. Background Technology

[0002] The full-process simulation system for wastewater treatment plants originated in the 1980s, with the activated sludge model (ASM series) proposed by the International Water Association (IWA) laying its theoretical foundation. Early simulations mainly focused on static water quality prediction for single biological reactors, used to assist in process design and selection. With the improvement of computing power and the deepening of research on water treatment kinetics, the system gradually evolved into a dynamic simulation tool covering the entire process of pretreatment, biochemical treatment, advanced treatment, and sludge treatment. In the 21st century, driven by "smart water management" and the dual-carbon strategy, full-process simulation began to deeply integrate automatic control theory, big data analysis, and digital twin technology, transforming from a simple research tool into an intelligent management platform integrating real-time monitoring, energy consumption optimization, and process prediction, achieving a leap from "static mechanism description" to "dynamic predictive control."

[0003] In existing technologies, full-process simulation systems for wastewater treatment plants often suffer from the following technical drawbacks: Existing simulation systems generally exhibit limitations such as high difficulty in calibrating physicochemical parameters and weak model robustness. Because the activated sludge process involves complex biochemical reaction kinetics, core models such as ASM contain a large number of highly coupled parameters (such as heterotrophic bacteria yield and decay coefficient), which show significant specificity with changes in influent composition and ambient temperature. Currently, parameter tuning mostly relies on manual experience or offline experiments, leading to a significant drop in prediction accuracy when facing complex and variable influent shocks, making it difficult to achieve low-cost, high-precision automated migration and rapid deployment.

[0004] Building upon the aforementioned calibration deviations, the simulation system also suffers from a deeper hidden danger: "nonlinear error accumulation in the recursive feedback loop throughout the entire process." Due to the extensive sludge recirculation, nitrification liquor recirculation, and side-flow treatment within the wastewater treatment process, a strongly coupled closed-loop structure is formed between the various units. When the initial parameters (as mentioned in the first drawback) have even minor deviations, these errors will amplify repeatedly within the system as materials circulate, generating a nonlinear drift similar to a chain reaction. This cascading effect leads to model distortion in the simulation results over long-term operation. Even if individual units simulate correctly, the overall system prediction of the entire process can experience unpredictable, precipitous failures due to the accumulation of errors, severely impacting the safety of automatic control decisions based on simulation results. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a full-process simulation system for wastewater treatment plants based on digital twins, thereby resolving the technical deficiencies mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a full-process simulation system for a wastewater treatment plant based on digital twins, comprising a virtual model construction module, a real-time data interaction module, a dynamic parameter calibration module, a feedback loop error correction module, and an intelligent operation scheduling module; the virtual model construction module is used to construct a three-dimensional digital twin entity of the wastewater treatment plant, and integrates an activated sludge dynamic model (ASM) based on the three-dimensional digital twin entity to map the physical entity's process structures to the virtual space; the real-time data interaction module is used to collect real-time operational status data of the wastewater treatment plant through IoT sensors, including influent flow rate related data, etc. The system includes data on concentration, ambient temperature, and process; the dynamic parameter calibration module is used to construct a biochemical reaction simulation algorithm model; based on the operating status data, a biochemical reaction sensitivity factor Kdsc is constructed to identify and update the physicochemical parameters coupled within the biochemical reaction simulation algorithm model; the feedback loop error correction module is used to monitor the nonlinear drift during sludge return and nitrification liquor return processes, and calculate the full-process feedback deviation index FDI to eliminate error accumulation in the full-process recursive feedback loop; the intelligent operation scheduling module is used to output process control strategies based on the simulation prediction results after deviation calibration and a preset robustness threshold H.

[0007] Preferably, the virtual model construction module includes a spatial modeling unit and a logical mapping unit; the spatial modeling unit is used to obtain the geometric topology of the sewage treatment plant through point cloud scanning and generate a 3D rendering model with precision texture and spatial coordinates; the logical mapping unit is used to embed the dynamic equations of the biological tank, sedimentation tank, and return pump station into the 3D model nodes to establish a real-time mapping between the physical space and the virtual parameter space.

[0008] Preferably, in the real-time data interaction module, distributed IoT sensor groups are deployed at key process points in the wastewater treatment plant, including the main influent channel, bar screen, corridors of the biochemical reaction tank, secondary sedimentation tank, and return pump station. Electromagnetic flow meters are used to sense influent flow rate data in real time. Online water quality analyzers, based on spectral or electrochemical analysis, acquire real-time data on component concentrations, including chemical oxygen demand (COD), ammonia nitrogen, total phosphorus, and total nitrogen. Thermistor arrays acquire water and atmospheric temperature data. Dissolved oxygen (DO) meters, sludge concentration MLSS meters, and level sensors acquire process-related data. An edge computing gateway is used to filter, reduce noise, and perform analog-to-digital conversion on the acquired raw analog signals, assigning an absolute timestamp to each set of digital signals to ensure linear alignment of physical entity data with the virtual space simulation step size in the time dimension. A standard communication protocol is used to encapsulate the timestamped operational status data into a real-time data stream, which is transmitted to the virtual model construction module via a bidirectional communication link, driving the dynamic updating of attribute values ​​of each node within the three-dimensional digital twin entity.

[0009] Preferably, the dynamic parameter calibration module includes a sensitivity analysis unit and an automatic evolution unit; the sensitivity analysis unit is used to obtain the sensitivity weight of parameters to effluent indicators using orthogonal experimental methods for coupled parameters such as heterotrophic bacteria yield and attenuation coefficient; the construction process of the biochemical reaction simulation algorithm model is as follows: taking the activated sludge kinetic model ASM as the core, combined with the topology of the sewage treatment plant structure, a set of differential equations covering mass conservation equation, energy conservation equation and biodegradation kinetic equation is established, and the set of differential equations is mapped to each functional voxel node of the three-dimensional digital twin entity to form a virtual reaction environment with spatiotemporal dynamic characteristics; preferably, the automatic evolution unit is used to calculate the biochemical reaction sensitivity factor Kdsc, and capture the fluctuations of ambient temperature and influent load in real time, and adjust the sensitivity parameters through a neural network algorithm to realize the adaptive evolution of the simulation model under varying operating conditions; wherein, the specific process of identifying and updating physicochemical parameters is as follows: using the operating status data obtained by the real-time data interaction module, the contribution of each physicochemical parameter to the fluctuation of effluent components is calculated by the sensitivity analysis method, thereby constructing the biochemical reaction sensitivity factor Kdsc; based on The parameters are sorted in descending order based on the magnitude of the biochemical reaction sensitivity factor Kdsc, and a subset of parameters with high sensitivity and strong coupling characteristics are selected as the targets to be updated. A cost function is established with the objective of minimizing the residual between the measured operating data and the model prediction data. A heuristic search algorithm or gradient descent algorithm is used to iteratively optimize the targets to be updated within the preset range of physicochemical parameter values. The optimized parameters are then fed back to the biochemical reaction simulation algorithm model in real time for calibration. The specific calculation logic of the biochemical reaction sensitivity factor Kdsc is as follows: obtain the j-th... The ratio between the change in effluent concentration and the parameter perturbation within each biochemical unit is used to characterize the sensitivity of the model output to fluctuations in specific parameters. The ratio of the current ambient temperature to the preset standard reference temperature is calculated as a temperature correction coefficient to compensate for the shift in kinetic rate with temperature changes. The natural logarithm of the influent load impact coefficient is performed to smooth the nonlinear effects of high load impacts on the biochemical system. Finally, the effluent concentration ratio, temperature correction coefficient, and the logarithmic influent load result are multiplied to obtain the biochemical reaction sensitivity factor Kdsc for the j-th biochemical unit. j .

[0010] Preferably, the feedback loop error correction module includes a loop identification unit and a nonlinear compensation unit; the loop identification unit is used to identify the internal backflow, external backflow and sideflow processing loops in the system and mark them as nonlinear feedback nodes; the nonlinear compensation unit is used to calculate the residual distribution of each node in the iterative process and suppress the cascade amplification of errors by introducing a compensation operator.

[0011] Preferably, the nonlinear compensation unit is used to calculate the full-process feedback deviation index FDI. The specific calculation logic is as follows: for multiple feedback loops involved in the system, the weighted deviation value of each loop is calculated one by one and summed. The deviation calculation method of a single loop is as follows: calculate the absolute value of the difference between the simulated predicted value of the loop and the measured value monitored by the physical sensor, and compare it with the monitored measured value to obtain the relative error ratio; multiply the relative error ratio by the corresponding loop influence weight factor; use the product of the nonlinear accumulation constant and the current loop return ratio as the exponent to calculate the exponent of Euler's constant e to characterize the cumulative effect of error generated by the return cycle; multiply the weighted relative error ratio by the exponent, and sum the product results of all loops to finally obtain the full-process feedback deviation index FDI.

[0012] Preferably, the feedback loop error correction module further includes a linkage early warning subunit, used to set a safety threshold Cerror and determine the degree of distortion of the simulation system; specifically: if the full-process feedback deviation index FDI exceeds the set safety threshold Cerror, the system determines that a chain reaction drift has occurred, triggering a model reset and virtual-real forced alignment command; if the full-process feedback deviation index FDI does not exceed the safety threshold Cerror, the system corrects the slight deviation through moving average filtering.

[0013] Preferably, the intelligent operation scheduling module obtains the simulation robustness score Srs through fitting. The calculation logic is as follows: it is obtained by constructing a fractional model that reflects the stability of the system; the logic of the numerator is: multiplying the biochemical reaction sensitivity factor Kdsc with the preset model convergence speed coefficient q to characterize the model's self-healing and response capabilities under parameter fluctuations; the logic of the denominator is: adding the full-process feedback deviation index FDI to a small correction constant b, where the correction constant b is used to avoid the abnormal zero denominator in mathematical calculations, and the denominator as a whole reflects the current cumulative error level of the system; the simulation robustness score Srs is obtained by the ratio of the numerator and the denominator.

[0014] Preferably, the intelligent operation scheduling module is used to preset a first threshold Qhigh and a second threshold Qlow, wherein the first threshold Qhigh is greater than the second threshold Qlow; and to perform the following management measures: if the simulation robustness score Srs is higher than the first threshold Qhigh, the system maintains the current fully automatic optimization control mode and executes the energy consumption minimization command; if the simulation robustness score Srs is lower than the second threshold Qlow, the system determines that the current digital twin model has failed, automatically switches to the safe operation manual takeover mode, and starts a global parameter rescan; if the simulation robustness score Srs is between the first threshold Qhigh and the second threshold Qlow, the system starts enhanced aeration redundancy control to prevent effluent exceedances caused by predicted fluctuations; at the same time, through a visualization interface, the fluctuation of Kdsc and the error heat map of FDI are superimposed on the three-dimensional twin of the wastewater treatment plant in real time to monitor and predict the risk of process distortion throughout the entire process.

[0015] This invention provides a full-process simulation system for wastewater treatment plants based on digital twins. It has the following beneficial effects: (1) This full-process simulation system for wastewater treatment plants based on digital twins constructs a biochemical reaction simulation algorithm model based on a three-dimensional digital twin entity, uses a dynamic parameter calibration module to calculate the biochemical reaction sensitivity factor in real time, and comprehensively considers the effluent concentration change, parameter perturbation, ratio of current ambient temperature to standard reference temperature, and logarithmic characteristics of influent load impact coefficient of the j-th biochemical unit to achieve adaptive updating of highly coupled parameters in the activated sludge dynamics model ASM. This mechanism effectively solves the problem of significant decline in prediction accuracy caused by the lag of manual experience calibration or the limitations of offline experiments when traditional simulation systems face complex and variable influent impacts, and significantly improves the robustness of the model under different operating conditions and the ability to deploy at low cost.

[0016] (2) This full-process simulation system for wastewater treatment plants based on digital twins accurately identifies nonlinear feedback nodes such as sludge return and nitrification liquid return through the feedback loop error correction module. It uses the full-process feedback deviation index to weight and sum the relative errors between the simulation prediction values ​​and the actual monitoring values ​​involving multiple feedback loops, and introduces the exponent of the Euler constant composed of the return ratio and the nonlinear accumulation constant to quantitatively characterize the cascading amplification effect of the error in the iterative process. Combined with the simulation robustness score generated by fitting the model convergence speed coefficient and the correction constant b, as well as the linkage early warning and scheduling mechanism for the safety threshold, the first threshold and the second threshold, the system eliminates the chain reaction deviation in the recursive feedback loop from the global level, avoids the risk of the whole-process system prediction failure due to local initial deviation, and greatly ensures the safety and stability of the intelligent automatic control decision of the wastewater treatment plant. Attached Figure Description

[0017] Figure 1 is a schematic diagram of the system framework structure of the present invention. Detailed Implementation

[0018] 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.

[0019] Example 1 (Referring to Figure 1) This invention provides a full-process simulation system for a wastewater treatment plant based on digital twins, including a virtual model construction module, a real-time data interaction module, a dynamic parameter calibration module, a feedback loop error correction module, and a smart operation scheduling module. The virtual model construction module is used to construct a three-dimensional digital twin entity of the wastewater treatment plant and integrate the activated sludge dynamic model (ASM) based on the three-dimensional digital twin entity to map the physical entity's process structures to the virtual space. The real-time data interaction module is used to collect real-time operating status data of the wastewater treatment plant through IoT sensors, including influent flow rate data and component concentration data. The system includes data on ambient temperature and process-related data; the dynamic parameter calibration module is used to construct a biochemical reaction simulation algorithm model; based on the operating status data, a biochemical reaction sensitivity factor Kdsc is constructed to identify and update the physicochemical parameters coupled within the biochemical reaction simulation algorithm model; the feedback loop error correction module is used to monitor the nonlinear drift during sludge return and nitrification liquor return processes, and calculate the full-process feedback deviation index FDI to eliminate error accumulation in the full-process recursive feedback loop; the intelligent operation scheduling module is used to output process control strategies based on the simulation prediction results after deviation calibration, combined with a preset robustness threshold H.

[0020] In this embodiment, the virtual model construction module constructs a three-dimensional digital twin entity and integrates the activated sludge dynamic model (ASM), achieving a high-precision mapping of physical process structures to the virtual space. This lays an accurate geometric topological and biochemical logical foundation for the full-process simulation. The real-time data interaction module utilizes IoT sensors to dynamically collect operational status data, ensuring real-time synchronization and timestamp alignment between the virtual model and the physical entity in the data dimension, providing a reliable data stream to drive the dynamic updates of the twin. The dynamic parameter calibration module identifies and adjusts the physicochemical parameters within the biochemical reaction simulation algorithm model by constructing a biochemical reaction sensitivity factor (Kdsc). Reverse iterative updates significantly enhance the system's adaptive evolution capability and parameter characterization accuracy when facing complex influent impacts; the feedback loop error correction module effectively monitors and eliminates nonlinear drift and error cascade amplification effects in the sludge return and nitrification liquor return recursive processes by calculating the full-process feedback deviation index FDI, fundamentally preventing systematic distortion of the simulation model caused by chain reactions; the intelligent operation scheduling module, based on the high-precision simulation prediction results after deviation calibration and combined with the preset robust threshold HH output process control strategy, provides wastewater treatment plants with highly reliable fully automatic operation decision support that balances energy consumption optimization and effluent compliance.

[0021] The virtual model construction module described in Example 2 includes a spatial modeling unit and a logical mapping unit. The spatial modeling unit is used to obtain the geometric topology of the sewage treatment plant through point cloud scanning and generate a 3D rendering model with precision texture and spatial coordinates. The logical mapping unit is used to embed the dynamic equations of the biological tank, sedimentation tank, and return pump station into the 3D model nodes to establish a real-time mapping between the physical space and the virtual parameter space.

[0022] In the real-time data interaction module, distributed IoT sensor groups are deployed at key process points in the wastewater treatment plant, including the main influent channel, bar screen, corridors of the biochemical reaction tank, secondary sedimentation tank, and return pump station. Electromagnetic flow meters are used to sense influent flow rate data in real time. Online water quality analyzers, based on spectral or electrochemical analysis, acquire real-time data on component concentrations, including chemical oxygen demand (COD), ammonia nitrogen, total phosphorus, and total nitrogen. A thermistor array acquires water and atmospheric temperature data. Dissolved oxygen (DO) meters, sludge concentration MLSS meters, and level sensors acquire process-related data. An edge computing gateway filters and denoises the acquired raw analog signals and performs analog-to-digital conversion, assigning an absolute timestamp to each digital signal to ensure linear alignment of physical entity data with the virtual space simulation step size in the time dimension. A standard communication protocol is used to encapsulate the timestamped operational status data into a real-time data stream, which is transmitted to the virtual model construction module via a bidirectional communication link, driving the dynamic updating of attribute values ​​for each node within the three-dimensional digital twin entity.

[0023] The dynamic parameter calibration module includes a sensitivity analysis unit and an automatic evolution unit. The sensitivity analysis unit is used to obtain the sensitivity weights of coupled parameters such as heterotrophic bacteria yield and attenuation coefficient to effluent indicators using orthogonal experimental design. The construction process of the biochemical reaction simulation algorithm model is as follows: based on the activated sludge dynamic model (ASM) and combined with the topology of the wastewater treatment plant structures, a set of differential equations covering mass conservation equations, energy conservation equations, and biodegradation kinetic equations is established. These differential equations are then mapped to the functional voxel nodes of a three-dimensional digital twin entity, forming a virtual reaction environment with spatiotemporal dynamic characteristics. The automatic evolution unit is used to calculate the biochemical reaction sensitivity factor Kdsc and capture fluctuations in ambient temperature and influent load in real time. It adjusts the sensitivity parameters through a neural network algorithm to achieve adaptive evolution of the simulation model under varying operating conditions. Specifically, the process of identifying and updating physicochemical parameters involves using the operating status data obtained from the real-time data interaction module to calculate the contribution of each physicochemical parameter to effluent component fluctuations using sensitivity analysis, thereby constructing the biochemical reaction sensitivity factor Kdsc. Based on the biochemical reaction... The parameters are sorted in descending order based on the sensitivity factor Kdsc, and a subset of parameters with high sensitivity and strong coupling characteristics are selected as the targets to be updated. A cost function is established with the goal of minimizing the residual between the measured data and the model prediction data. A heuristic search algorithm or gradient descent algorithm is used to iteratively optimize the targets to be updated within the preset range of physicochemical parameter values. The optimized parameters are then fed back to the biochemical reaction simulation algorithm model in real time for calibration. The specific calculation logic of the biochemical reaction sensitivity factor Kdsc is as follows: obtain the j-th biochemical reaction sensitivity factor Kdsc... The ratio between the change in effluent concentration and the parameter perturbation within the biochemical unit is used to characterize the sensitivity of the model output to fluctuations in specific parameters. The ratio of the current ambient temperature to the preset standard reference temperature is calculated as a temperature correction coefficient to compensate for the shift in kinetic rate with temperature changes. The natural logarithm of the influent load impact coefficient is performed to smooth the nonlinear effects of high load impacts on the biochemical system. Finally, the effluent concentration ratio, temperature correction coefficient, and the logarithmic influent load result are multiplied to obtain the biochemical reaction sensitivity factor Kdsc for the j-th biochemical unit. j .

[0024] The biochemical reaction sensitivity factor Kdsc of the j-th biochemical unit j Specifically, it is obtained through the following calculation formula: In the formula, Let ΔPj represent the change in effluent concentration of the j-th biochemical unit, ΔPj represent the parameter perturbation of the j-th biochemical unit, Tj represent the current ambient temperature, and Tstd represent the standard reference temperature. Let be the influent load impact coefficient of the j-th biochemical unit.

[0025] The feedback loop error correction module includes a loop identification unit and a nonlinear compensation unit. The loop identification unit is used to identify the internal backflow, external backflow, and sideflow processing loops in the system and mark them as nonlinear feedback nodes. The nonlinear compensation unit is used to calculate the residual distribution of each node in the iterative process and suppress the cascade amplification of errors by introducing a compensation operator.

[0026] The nonlinear compensation unit is used to calculate the full-process feedback deviation index (FDI). The specific calculation logic is as follows: for each of the multiple feedback loops involved in the system, the weighted deviation value of each loop is calculated and summed. The deviation calculation method for a single loop is as follows: calculate the absolute value of the difference between the simulated predicted value and the measured value monitored by the physical sensor, and compare it with the measured value to obtain the relative error ratio; multiply the relative error ratio by the corresponding loop influence weighting factor; use the product of the nonlinear accumulation constant and the current loop return ratio as the exponent to calculate the power of Euler's constant e, which characterizes the cumulative effect of error generated by the return cycle; multiply the weighted relative error ratio by the exponent, and sum the product results of all loops to finally obtain the full-process feedback deviation index (FDI).

[0027] Specifically, obtain it in the following way: In the formula, n is the total number of feedback loops, Vsim,i is the simulated predicted value of the i-th loop, Vreal,i is the measured value monitored by the physical sensor, wi is the loop influence weighting factor, λ is the nonlinear accumulation constant, and Ri is the current return ratio of the current loop.

[0028] The feedback loop error correction module also includes a linkage early warning subunit, which is used to set a safety threshold Cerror and determine the degree of distortion of the simulation system. Specifically: if the full-process feedback deviation index FDI exceeds the set safety threshold Cerror, the system determines that a chain reaction drift has occurred, triggering a model reset and virtual-real forced alignment command; if the full-process feedback deviation index FDI does not exceed the safety threshold Cerror, the system corrects the slight deviation through moving average filtering.

[0029] The intelligent operation scheduling module obtains the simulation robustness score Srs through fitting. The specific calculation logic is as follows: it is obtained by constructing a fractional model that reflects the stability of the system; the logic of the numerator is: multiplying the biochemical reaction sensitivity factor Kdsc with the preset model convergence speed coefficient q to characterize the model's self-healing and response capabilities under parameter fluctuations; the logic of the denominator is: adding the full-process feedback deviation index FDI to a small correction constant b, where the correction constant b is used to avoid the abnormal zero denominator in mathematical calculations, and the denominator as a whole reflects the current cumulative error level of the system; the simulation robustness score Srs is obtained by the ratio of the numerator and the denominator.

[0030] The specific calculation formula is as follows: A higher simulation robustness score Srs indicates stronger predictive reliability of the simulation system in resisting the accumulation of feedback errors and influent shocks, reflecting the predictive stability of the simulation system in response to external influent shocks after eliminating error accumulation. The intelligent operation scheduling module is used to preset a first threshold Qhigh and a second threshold Qlow, with the first threshold Qhigh being greater than the second threshold Qlow. It then executes the following management measures: if the simulation robustness score Srs is higher than the first threshold Qhigh, the system maintains the current fully automatic optimized control mode and executes the energy consumption minimization command; if the simulation robustness score Srs is lower than the second threshold Qlow, the system determines that the current digital twin model has failed, automatically switches to the safe operation manual takeover mode, and starts a global parameter rescan; if the simulation robustness score Srs is between the first threshold Qhigh and the second threshold Qlow, the system starts enhanced aeration redundancy control to prevent effluent exceedances caused by predictive fluctuations. Simultaneously, through a visualization interface, the fluctuation of Kdsc and the error heat map of FDI are superimposed in real time on the three-dimensional twin of the wastewater treatment plant to monitor and predict the risk of process distortion throughout the entire process.

[0031] In this embodiment, the ASM equation of the activated sludge kinetic model is embedded into 3D voxel nodes through spatial modeling units and logical mapping units, realizing high-fidelity mapping and spatiotemporal dynamic simulation between physical structures and virtual space. Distributed IoT sensor arrays collect in real-time influent flow data, component concentration data (including chemical oxygen demand (COD), ammonia nitrogen, total phosphorus, and total nitrogen), water and atmospheric ambient temperature data (Tj), and process-related data (including dissolved oxygen (DO), sludge concentration (MLSS), and liquid level). An edge computing gateway assigns absolute timestamps, ensuring linear alignment and precise driving of physical entity data and virtual simulation step sizes in the time dimension. By calculating the biochemical reaction sensitivity factor and comprehensively considering the logarithmic effects of effluent concentration changes, parameter perturbations, temperature correction coefficients, and influent load impact coefficients, highly coupled heterotrophic bacteria yield and decay coefficients can be identified and updated in real-time. The parameters enable the model to adaptively evolve under varying operating conditions. By calculating the full-process feedback deviation index, utilizing the loop influence weighting factor, the relative error between the simulation prediction and the measured value, and introducing an exponential power composed of the nonlinear accumulation constant and the reflux ratio, the error cascade amplification effect in the reflux cycle process is quantitatively characterized and eliminated. The virtual-real forced alignment command triggered by the safety threshold effectively eliminates the risk of nonlinear drift in the full-process recursive feedback. By fitting the biochemical reaction sensitivity factor, convergence speed coefficient, feedback deviation index, and correction constant b, the simulation robustness score is obtained. Combined with the first and second thresholds, intelligent switching is achieved from fully automatic control with minimal energy consumption to enhanced aeration redundancy control, and then to manual takeover mode. This not only provides the system with quantitative predictive reliability evaluation indicators, but also ensures the operational stability and decision-making safety of the wastewater treatment process under extreme shock loads.

[0032] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A full-process simulation system for a wastewater treatment plant based on digital twins, characterized in that: It includes a virtual model construction module, a real-time data interaction module, a dynamic parameter calibration module, a feedback loop error correction module, and an intelligent operation scheduling module. The virtual model construction module is used to construct a three-dimensional digital twin entity of the wastewater treatment plant and integrate the activated sludge dynamic model (ASM) based on the three-dimensional digital twin entity to map the physical entity's process structures to the virtual space. The real-time data interaction module is used to collect real-time operating status data of the wastewater treatment plant through IoT sensors, including influent flow rate data, component concentration data, ambient temperature data, and process-related data. The dynamic parameter calibration module is used to construct a biochemical reaction simulation algorithm model; based on the operating status data, a biochemical reaction sensitivity factor Kdsc is constructed to identify and update the physicochemical parameters coupled within the biochemical reaction simulation algorithm model; the feedback loop error correction module is used to monitor the nonlinear drift during sludge return and nitrification liquor return processes, and calculate the full-process feedback deviation index FDI to eliminate error accumulation in the full-process recursive feedback loop. The intelligent operation scheduling module is used to output process control strategies based on the simulation prediction results after deviation calibration and in combination with the preset robustness threshold H.

2. The wastewater treatment plant full-process simulation system based on digital twins according to claim 1, characterized in that: The virtual model construction module includes a spatial modeling unit and a logical mapping unit. The spatial modeling unit is used to obtain the geometric topology of the sewage treatment plant through point cloud scanning and generate a 3D rendering model with accurate texture and spatial coordinates. The logical mapping unit is used to embed the dynamic equations of the biological tank, sedimentation tank, and return pump station into the 3D model nodes to establish a real-time mapping between the physical space and the virtual parameter space.

3. The wastewater treatment plant full-process simulation system based on digital twins according to claim 1, characterized in that: In the real-time data interaction module, distributed IoT sensor groups are deployed at key process points in the wastewater treatment plant, including the main influent channel, bar screen, corridors of the biochemical reaction tank, secondary sedimentation tank, and return pump station. Electromagnetic flow meters are used to sense influent flow rate data in real time. Online water quality analyzers, based on spectral or electrochemical analysis methods, acquire real-time data on component concentrations, including chemical oxygen demand (COD), ammonia nitrogen, total phosphorus, and total nitrogen. A thermistor array is used to acquire water and atmospheric temperature data. Dissolved oxygen (DO) meters, sludge concentration MLSS meters, and level sensors are used to acquire process-related data. The edge computing gateway is used to filter and reduce noise and perform analog-to-digital conversion on the acquired raw analog signals, and an absolute timestamp is assigned to each set of digital signals to ensure linear alignment of physical entity data and virtual space simulation step size in the time dimension. The time-stamped running status data is encapsulated into a real-time data stream using a standard communication protocol and transmitted to the virtual model construction module through a bidirectional communication link to drive the dynamic update of the attribute values ​​of each node inside the three-dimensional digital twin entity.

4. The wastewater treatment plant full-process simulation system based on digital twins according to claim 1, characterized in that: The dynamic parameter calibration module includes a sensitivity analysis unit and an automatic evolution unit. The sensitivity analysis unit is used to obtain the sensitivity weight of parameters to effluent indicators using orthogonal experimental methods for coupled parameters such as heterotrophic bacteria yield and attenuation coefficient. The construction process of the biochemical reaction simulation algorithm model is as follows: taking the activated sludge kinetic model ASM as the core and combining it with the topology of the sewage treatment plant structure, a set of differential equations covering the mass conservation equation, energy conservation equation and biodegradation kinetic equation is established, and the set of differential equations is mapped to each functional voxel node of the three-dimensional digital twin entity to form a virtual reaction environment with spatiotemporal dynamic characteristics.

5. The wastewater treatment plant full-process simulation system based on digital twins according to claim 1, characterized in that: The automatic evolution unit is used to calculate the biochemical reaction sensitivity factor Kdsc and capture fluctuations in ambient temperature and influent load in real time. It adjusts the sensitivity parameters through a neural network algorithm to achieve adaptive evolution of the simulation model under varying operating conditions. The specific process of identifying and updating physicochemical parameters is as follows: using operational status data acquired by the real-time data interaction module, the contribution of each physicochemical parameter to the fluctuations in effluent components is calculated through sensitivity analysis, thereby constructing the biochemical reaction sensitivity factor Kdsc; based on the magnitude of the biochemical reaction sensitivity factor Kdsc, the parameters are sorted in descending order, and a subset of parameters in the high-sensitivity range with high coupling characteristics is selected as the targets to be updated; a cost function is established with the objective of minimizing the residual between measured operational data and model predicted data, and a heuristic search algorithm or gradient descent algorithm is used to... Within the preset range of physicochemical parameters, the target to be updated is iteratively optimized in reverse, and the optimized parameters are fed back to the biochemical reaction simulation algorithm model in real time for calibration. The specific calculation logic of the biochemical reaction sensitivity factor Kdsc is as follows: the ratio between the change in effluent concentration and the parameter perturbation in the j-th biochemical unit is obtained to characterize the sensitivity of the model output to fluctuations in specific parameters; the ratio between the current ambient temperature and the preset standard reference temperature is calculated as a temperature correction coefficient to compensate for the shift in kinetic rate with temperature; the natural logarithm of the influent load impact coefficient is performed to smooth the nonlinear effect of high load impact on the biochemical system; finally, the above effluent concentration ratio, temperature correction coefficient, and influent load logarithmic processing result are multiplied together to obtain the biochemical reaction sensitivity factor Kdsc of the j-th biochemical unit. j .

6. The wastewater treatment plant full-process simulation system based on digital twins according to claim 5, characterized in that: The feedback loop error correction module includes a loop identification unit and a nonlinear compensation unit. The loop identification unit is used to identify the internal backflow, external backflow, and sideflow processing loops in the system and mark them as nonlinear feedback nodes. The nonlinear compensation unit is used to calculate the residual distribution of each node in the iterative process and suppress the cascade amplification of errors by introducing a compensation operator.

7. The wastewater treatment plant full-process simulation system based on digital twins according to claim 6, characterized in that: The nonlinear compensation unit is used to calculate the full-process feedback deviation index (FDI). The specific calculation logic is as follows: for each of the multiple feedback loops involved in the system, the weighted deviation value of each loop is calculated and summed. The deviation calculation method for a single loop is as follows: calculate the absolute value of the difference between the simulated predicted value and the measured value monitored by the physical sensor, and compare it with the measured value to obtain the relative error ratio; multiply the relative error ratio by the corresponding loop influence weighting factor; use the product of the nonlinear accumulation constant and the current loop return ratio as the exponent to calculate the power of Euler's constant e, which characterizes the cumulative effect of error generated by the return cycle; multiply the weighted relative error ratio by the exponent, and sum the product results of all loops to finally obtain the full-process feedback deviation index (FDI).

8. The wastewater treatment plant full-process simulation system based on digital twins according to claim 1, characterized in that: The feedback loop error correction module also includes a linkage early warning subunit, which is used to set a safety threshold Cerror and determine the degree of distortion of the simulation system. Specifically: if the full-process feedback deviation index FDI exceeds the set safety threshold Cerror, the system determines that a chain reaction drift has occurred, triggering a model reset and virtual-real forced alignment command; if the full-process feedback deviation index FDI does not exceed the safety threshold Cerror, the system corrects the slight deviation through moving average filtering.

9. A full-process simulation system for a wastewater treatment plant based on digital twins as described in claim 8, characterized in that: The intelligent operation scheduling module obtains the simulation robustness score Srs through fitting. The specific calculation logic is as follows: it is obtained by constructing a fractional model that reflects the stability of the system; the logic of the numerator is: multiplying the biochemical reaction sensitivity factor Kdsc with the preset model convergence speed coefficient q to characterize the model's self-healing and response capabilities under parameter fluctuations; the logic of the denominator is: adding the full-process feedback deviation index FDI to a small correction constant b, where the correction constant b is used to avoid the abnormal zero denominator in mathematical calculations, and the denominator as a whole reflects the current cumulative error level of the system; the simulation robustness score Srs is obtained by the ratio of the numerator and the denominator.

10. A full-process simulation system for a wastewater treatment plant based on digital twins as described in claim 1, characterized in that: The intelligent operation scheduling module is used to preset a first threshold Qhigh and a second threshold Qlow, wherein the first threshold Qhigh is greater than the second threshold Qlow; The system implements the following management measures: If the simulation robustness score Srs is higher than the first threshold Qhigh, the system maintains the current fully automatic optimization control mode and executes the energy minimization command; if the simulation robustness score Srs is lower than the second threshold Qlow, the system determines that the current digital twin model has failed, automatically switches to the safe operation manual takeover mode, and starts a global parameter rescan; if the simulation robustness score Srs is between the first threshold Qhigh and the second threshold Qlow, the system starts enhanced aeration redundancy control to prevent effluent exceedances caused by predicted fluctuations; at the same time, through a visualization interface, the fluctuation of Kdsc and the error heat map of FDI are superimposed on the three-dimensional twin of the wastewater treatment plant in real time to monitor and predict the risk of process distortion throughout the entire process.

Images