A shipborne wastewater monitoring and treatment evaluation method based on high-resolution full-spectrum soft sensing

Abstract

This invention discloses a method for monitoring and evaluating shipborne wastewater treatment based on high-resolution full-spectrum soft sensing, relating to the fields of water quality monitoring and marine environmental protection. It aims to address the technical problems of existing shipborne water quality monitoring systems, which rely on hardware sensors, have weak anti-interference capabilities, poor model generalization, and cannot accurately analyze the status of each stage of the wastewater treatment system in real time, resulting in an incomplete evaluation system. This method uses high-resolution full-spectrum data as a foundation, integrates soft sensing technology, and constructs a water quality parameter prediction model based on machine learning and self-learning to achieve real-time and accurate measurement of key water quality parameters of shipborne wastewater. Simultaneously, it establishes a comprehensive evaluation index system for the shipborne wastewater treatment system, combining monitored water quality parameters and operational data from each stage to complete status analysis and fault prediction for core stages such as sedimentation, oxidation, and biological treatment.

Description

Technical Field

[0001] This invention relates to the fields of water quality monitoring and ship environmental protection technology, specifically to a method for monitoring and evaluating shipborne wastewater treatment based on high-resolution full-spectrum soft sensing. Background Technology

[0002] With the development of the global shipping industry, the impact of shipboard wastewater discharge on the marine ecological environment has become increasingly prominent. The IMO's "Global Ship Pollution Control Report 2024" shows that the annual discharge of sewage from ships worldwide has reached 1.2 billion tons, and is projected to increase to 1.35 billion tons by 2025. Of this, the proportion of untreated direct discharge remains as high as 15%. Therefore, accurate monitoring and efficient treatment of shipboard wastewater have become core requirements for ship environmental protection. Shipboard wastewater is complex in composition and highly volatile, and is affected by environmental factors such as ship turbulence and temperature / salinity fluctuations. Traditional water quality monitoring methods mainly rely on hardware sensors, which have many drawbacks: First, hardware sensors are costly and difficult to maintain, and are susceptible to malfunctions due to the harsh shipboard environment. For example, traditional COD sensors have long response times (≤10 minutes) and short lifespans in high-salinity environments. Second, the monitoring parameters are limited, making simultaneous monitoring of multiple parameters impossible, and the monitoring accuracy is easily affected by interference; different analysts can produce results of up to 12% different values ​​for the same sample. Third, they lack adaptive capabilities; when wastewater quality changes abruptly, the monitoring accuracy drops significantly, failing to meet the needs of real-time monitoring.

[0003] Soft sensing technology, as a novel monitoring technology that "replaces some hardware with software," achieves indirect estimation of difficult-to-measure parameters through the mathematical relationship between easily measurable auxiliary variables and difficult-to-measure dominant variables. It can effectively overcome the limitations of hardware sensors and has been widely used in industrial process monitoring. However, its application in shipborne water quality monitoring still has shortcomings: existing soft sensing models mostly use traditional machine learning algorithms, which have poor generalization ability and cannot adapt to the complex and variable characteristics of shipborne wastewater quality; they lack self-learning ability, and when water quality fluctuates or the environment changes, the model parameters cannot be automatically updated, requiring frequent manual calibration, which increases operation and maintenance costs; and they are not deeply integrated with shipborne wastewater treatment systems, only able to monitor water quality parameters, and unable to accurately analyze and evaluate the operating status of each link of the wastewater treatment system.

[0004] Shipboard wastewater treatment systems mainly consist of three core stages: sedimentation, oxidation, and biological treatment. The operational status of each stage directly determines the wastewater treatment effect. Existing evaluation methods often use single indicators, lacking a comprehensive evaluation system. Furthermore, the evaluation process relies on manual sampling and testing, resulting in poor real-time performance and an inability to promptly identify potential faults in each stage. For example, sludge accumulation in the sedimentation stage, insufficient oxidant dosage in the oxidation stage, and microbial imbalance in the biological stage can lead to decreased wastewater treatment efficiency and even exceed emission standards. In addition, traditional ship wastewater treatment effect assessment methods have a low testing frequency (results are typically available every 24 hours), failing to provide real-time feedback on treatment effectiveness and hindering the achievement of a closed-loop linkage between monitoring, evaluation, and control.

[0005] High-resolution full-spectrum technology can acquire spectral information of wastewater across the entire wavelength range, containing rich water quality component characteristics and reflecting the overall water quality status of the wastewater. Combined with soft sensing technology and machine learning algorithms, it can achieve simultaneous and accurate monitoring of multiple parameters. However, there is currently no technical solution that combines high-resolution full-spectrum technology, soft sensing technology, machine learning, and self-learning models for shipboard wastewater monitoring, and establishes a comprehensive evaluation index system for shipboard wastewater treatment systems to achieve state analysis and closed-loop control of each stage. Therefore, developing a shipboard wastewater treatment monitoring and evaluation method that is adaptable to the shipboard environment, highly accurate, highly adaptive, and can achieve integrated monitoring and evaluation has significant practical significance and application value. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a shipborne wastewater monitoring and treatment evaluation method based on high-resolution full-spectrum soft sensing. This method solves the technical problems of existing shipborne water quality monitoring relying on hardware sensors, having weak anti-interference capabilities, poor model generalization, and being unable to accurately analyze the status of each link in the wastewater treatment system in real time, as well as having an imperfect evaluation system. This invention enables real-time and accurate monitoring of shipborne wastewater quality parameters, accurate analysis and comprehensive evaluation of the status of each link in the wastewater treatment system, and provides reliable support for the optimization and control of shipborne wastewater treatment systems.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for monitoring and evaluating shipborne wastewater treatment based on high-resolution full-spectrum soft sensing includes the following steps: Step 1: Build a shipborne high-resolution full-spectrum soft sensing acquisition system to collect high-resolution full-spectrum data of shipborne wastewater, simultaneously collect easily measurable auxiliary parameters, and complete data preprocessing.

[0008] The high-resolution full-spectrum soft sensing acquisition system is designed specifically for shipboard environments. It adopts a waterproof and vibration-resistant structure and is suitable for sea states 3 to 6. It includes a high-resolution full-spectrum instrument, an auxiliary parameter acquisition module, a data transmission module, and a preprocessing module. The high-resolution full-spectrum instrument uses a spectral range of 200~1100nm and a spectral resolution of ≤1nm. The acquisition frequency is set to 1~5 times / minute, and the acquisition points cover the inlet, sedimentation outlet, oxidation outlet, biological treatment outlet, and main outlet of the wastewater treatment system to ensure comprehensive capture of the spectral characteristics of wastewater at each stage. The auxiliary parameter acquisition module is used to acquire easily measurable parameters such as water temperature, pH value, turbidity, and conductivity, with acquisition accuracies of ±0.1℃, ±0.01pH, ±0.1NTU, and ±1μS / cm, respectively, providing rich auxiliary input variables for the soft sensing model. The data transmission module adopts a redundant design of fiber optic network and LoRa technology. The fiber optic network is used for high-speed transmission of core data, and LoRa technology is used for backup transmission to ensure the stability and real-time performance of data transmission, with a transmission delay of ≤10ms. The preprocessing module is used to preprocess the acquired spectral data and auxiliary parameters to eliminate interference factors and improve data quality.

[0009] The spectral data preprocessing includes dark current correction, baseline correction, spectral smoothing, and feature extraction. Dark current correction is used to eliminate the influence of electronic noise from the full spectrometer itself. Baseline correction is used to eliminate spectral baseline drift (caused by shipboard vibration and temperature fluctuations). Spectral smoothing uses the Savitzky-Golay filtering algorithm to reduce random noise interference. Feature extraction uses principal component analysis (PCA) combined with continuous projection algorithm (SPA). PCA is used to reduce the dimensionality of spectral data and remove redundant information. SPA is used to select the characteristic spectral bands that are most correlated with water quality parameters, reduce the interference of irrelevant spectral information on the model, and improve the model's computational efficiency and accuracy.

[0010] Auxiliary parameter preprocessing includes outlier removal, data normalization, and missing value completion: Outlier removal uses the ±3σ rule to remove abnormal data caused by equipment failure or severe ship turbulence; data normalization uses the min-max normalization method to normalize the auxiliary parameters to the [0,1] interval, eliminating the influence of differences in the dimensions of different parameters on the model; missing value completion uses the K-nearest neighbor interpolation method to complete the missing auxiliary parameter values ​​based on valid data at adjacent times, ensuring data integrity.

[0011] Step 2: Construct a soft-sensor water quality parameter prediction model based on machine learning and self-learning. Using preprocessed full-spectrum data and easily measurable auxiliary parameters as input, train the model and achieve self-learning optimization to output real-time measured values ​​of key water quality parameters of shipborne wastewater.

[0012] The core of this step is to build a soft sensing model that combines high accuracy and adaptability, solving the problems of poor generalization and frequent manual calibration required by traditional models. Specifically, it includes two parts: building a basic machine learning model and designing a self-learning optimization module.

[0013] The basic machine learning model adopts a CNN-LSTM hybrid model, combining the advantages of convolutional neural networks (CNN) and long short-term memory networks (LSTM): the CNN module has a powerful feature extraction capability, which can deeply mine the deep spectral features related to water quality parameters from the preprocessed feature spectral data, and capture the correlation between the intensity and position changes of spectral peaks and water quality components; the LSTM module has a good time series prediction capability, which can capture the time series variation pattern of water quality parameters. At the same time, combined with easily measurable auxiliary parameters, a dynamic mapping relationship between spectral features, auxiliary parameters and water quality parameters is established, which improves the prediction accuracy and generalization ability of the model.

[0014] The key water quality parameters include chemical oxygen demand (COD), biochemical oxygen demand (BOD), ammonia nitrogen (NH3-N), total phosphorus (TP), total nitrogen (TN), and suspended solids (SS). These parameters are the core indicators for evaluating the treatment effect of shipboard wastewater. The measurement accuracy of the model output meets the requirements of ±3%, ±5%, ±2%, ±2%, ±3%, and ±1%, respectively, which are in line with the accuracy requirements for monitoring ship sewage discharge.

[0015] The self-learning optimization module is used to automatically update model parameters, adapting to scenarios with complex and variable shipboard wastewater quality and significant environmental interference. Its implementation process is as follows: S21: Set the model prediction error thresholds and self-learning trigger conditions. Based on the measurement accuracy requirements of each water quality parameter, the prediction error thresholds for COD, BOD, NH3-N, TP, TN, and SS are set to 3%, 5%, 2%, 2%, 3%, and 1%, respectively. The self-learning trigger conditions include three categories: first, the model prediction error exceeds the corresponding threshold three times consecutively; second, there is a sudden change in the influent wastewater quality (COD change ≥ 50 mg / L); and third, there is a drastic fluctuation in the shipborne environmental parameters (water temperature change ≥ 5℃ or salinity change ≥ 1‰). This ensures that the model can promptly initiate self-learning when there are significant changes in water quality or the environment, thus avoiding a decrease in prediction accuracy.

[0016] S22: Construction of Incremental Training Dataset. When the self-learning trigger condition is met, the model automatically collects the latest spectral data, auxiliary parameters, and corresponding measured values ​​of water quality parameters. The measured values ​​of water quality parameters are obtained through periodic calibration using national standard methods (mandatory calibration every 30 days, and supplementary calibration when self-learning is triggered daily). This constructs an incremental training dataset to ensure that the new dataset accurately reflects the current water quality characteristics and environmental status.

[0017] S23: Incremental Update of Model Parameters. An incremental learning algorithm is used to update the parameters of the CNN-LSTM hybrid model. Unlike traditional full retraining, incremental learning only retains the model's original effective parameters and fine-tunes only those parameters that do not match the new data, reducing the computational load and shortening the update time, thus adapting to the limited computing power of shipboard equipment. During the update process, 10-fold cross-validation is used to ensure that the model still has good generalization ability after the update, avoiding overfitting. After the update is completed, the model automatically verifies the prediction accuracy until the prediction error is lower than the corresponding threshold, completing self-learning optimization.

[0018] S24: Model Performance Log Recording. Establish a model parameter update log to record in detail the triggering reason for each self-learning, the updated parameters, the size of the incremental dataset, and changes in model accuracy. This facilitates subsequent model maintenance and performance analysis, and provides data support for the operation optimization of the shipborne wastewater treatment system.

[0019] In addition, the model deployment adopts an edge computing architecture, deploying the model on the ship's edge computing nodes, reducing dependence on the cloud, improving the real-time performance and security of data processing, avoiding monitoring failure due to network interruption, and reducing energy consumption for data transmission, which is suitable for the limited energy resources on board ships.

[0020] Step 3: Establish a comprehensive evaluation index system for shipborne wastewater treatment systems. The evaluation index system includes overall evaluation indexes and sub-evaluation indexes for the three core links of sedimentation, oxidation, and biological treatment.

[0021] Based on the operational characteristics of shipborne wastewater treatment systems (limited space, energy consumption constraints, and high anti-interference requirements) and IMO ship wastewater discharge standards, a comprehensive evaluation index system is established that balances scientific rigor, practicality, and operability. This system can evaluate the overall operational effectiveness of the system and accurately reflect the operational status of each core component. Specific indicators are as follows: 3.1 Overall Evaluation Indicators The overall evaluation index is used to measure the comprehensive operational effectiveness of the shipboard wastewater treatment system. It includes five core indicators, each with clearly defined threshold requirements to ensure the objectivity and accuracy of the evaluation: (1) Treatment efficiency (η): The core evaluation index is defined as the ratio of the difference between the influent water quality parameter value and the effluent water quality parameter value to the influent water quality parameter value. The calculation formula is η = (influent water quality parameter value - effluent water quality parameter value) / influent water quality parameter value × 100%. Different threshold requirements are set for different water quality parameters: the treatment efficiency η of COD and BOD is ≥ 90%, the treatment efficiency η of NH3-N is ≥ 95%, the treatment efficiency η of TP and TN is ≥ 90%, and the treatment efficiency η of SS is ≥ 98%, to ensure that the treated wastewater can meet the ship sewage discharge standards.

[0022] (2) Energy intensity (E): measures the economic efficiency of system operation, defined as the power consumption per unit volume of wastewater treated, with the unit being kWh / m³; considering the limited energy on board, E is set to ≤1.5 kWh / m³, and the system energy consumption is reduced by optimizing the operating parameters of each link.

[0023] (3) Operational stability (S): This measures the system’s ability to adapt to fluctuations in the shipboard environment. It is defined as the ratio of continuous stable operating time to total operating time. The calculation formula is S = (continuous stable operating time / total operating time) × 100%. S is set to ≥ 95% to ensure that the system can still operate stably in environments such as ship turbulence and temperature and salinity fluctuations, thus avoiding frequent failures.

[0024] (4) Emission Compliance Rate (C): This measures the compliance of the system's emissions. It is defined as the ratio of the number of compliant emissions to the total number of emissions. The calculation formula is C = (number of compliant emissions / total number of emissions) × 100%. C = 100% is set to strictly comply with the IMO's "2024 Global Ship Pollution Control Report" and relevant standards for ship wastewater discharge, so as to avoid pollution of the marine environment caused by illegal wastewater discharge.

[0025] (5) Operation and maintenance cost (M): measures the economic efficiency of system operation, defined as the total cost of equipment maintenance, chemical consumption and consumable replacement per unit time, in yuan / day; set reasonable cost thresholds according to the actual needs of shipboard operation and maintenance to ensure that the system operation cost is controllable.

[0026] 3.2 Sub-item evaluation indicators The sub-evaluation indicators are used to accurately reflect the operational status of the three core processes: precipitation, oxidation, and biological treatment, providing a basis for fault diagnosis and optimization control of each process: (1) Evaluation indicators for the sedimentation process: The core function of the sedimentation process is to remove suspended solids (SS) and some colloidal impurities from wastewater. The evaluation indicators include four items: ① Suspended solids removal rate (η1): A core indicator that measures the treatment effectiveness of the sedimentation process; η1 ≥ 90%; ② Sludge settling ratio (SV30): measures the settling performance of sludge and reflects the sludge accumulation in the sedimentation tank. It should be controlled between 15% and 30%. If SV30 is too high, it indicates poor sludge settling performance, which can easily lead to clogging of the sedimentation tank. If SV30 is too low, it indicates insufficient sludge and poor sedimentation effect. ③ Settling time (t1): This measures the operational efficiency of the settling process and should be controlled within 30-60 minutes. Considering the limited space on board the ship, excessively long settling times should be avoided to prevent the equipment from becoming too bulky. ④ Sludge moisture content (W1): measures the dewatering performance of sludge after sedimentation. W1≤98%. Excessive moisture content will increase the difficulty and cost of subsequent sludge treatment.

[0027] (2) Evaluation indicators for the oxidation process: The core function of the oxidation process is to remove recalcitrant organic matter from wastewater and reduce COD concentration. The evaluation indicators include four items: ① COD removal rate (η2): A core indicator that measures the treatment effect of the oxidation process; η2 ≥ 85%; ② Oxidant utilization rate (η3): measures the efficiency of oxidant use. η3≥80% avoids oxidant waste and reduces operating costs; ③ Oxidation reaction time (t2): This measures the operating efficiency of the oxidation process and should be controlled within 20-40 minutes, balancing treatment effectiveness and operating efficiency. ④ Residual chlorine concentration in effluent (ρ): This measures the sufficiency of the oxidation reaction. It should be controlled between 0.5 and 1.0 mg / L. Too low a residual chlorine concentration indicates insufficient oxidation, while too high a concentration will inhibit the activity of microorganisms in subsequent biological treatment processes.

[0028] (3) Evaluation indicators for biological treatment process: The core function of biological treatment is to remove pollutants such as BOD, NH3-N, and TN from wastewater by utilizing the metabolic activity of microorganisms. The evaluation indicators include four items: ① BOD removal rate (η4): A core indicator that measures the treatment effectiveness of the biological treatment process; η4 ≥ 90%; ② Microbial activity (A): measures the metabolic capacity of the microbial community, measured by dehydrogenase activity. A≥0.8 indicates that the microbial community is unbalanced or the growth environment is poor, which will lead to a decrease in the treatment effect. ③ Dissolved oxygen concentration (DO): This measures the oxygen supply in the bioreactor and should be controlled at 2~4 mg / L. Too low a DO will lead to microbial hypoxia and decreased metabolic capacity, while too high a DO will increase aeration energy consumption and may inhibit denitrification. ④ Sludge Retention Time (SRT): This measures the retention time of biological sludge, which should be controlled between 8 and 12 days. If the sludge retention time is too long, it will lead to sludge aging and decreased activity. If the sludge retention time is too short, it will lead to insufficient microbial biomass and poor treatment effect.

[0029] Step 4: Collect the operating parameters of each link of the shipborne wastewater treatment system, combine them with the water quality parameter measurement values ​​output in Step 2, substitute them into the evaluation index system, and complete the status analysis of each link and the overall evaluation of the system.

[0030] First, the operating parameters of each stage of the shipborne wastewater treatment system were collected, including sludge settling ratio, settling time, and sludge moisture content in the sedimentation stage; oxidant dosage, oxidation reaction time, and residual chlorine concentration in the effluent in the oxidation stage; and microbial activity, dissolved oxygen concentration, and sludge age in the biological treatment stage. The collection frequency was consistent with that of the spectral data collection frequency (1~5 times / minute) to ensure data synchronization.

[0031] Then, combining the water quality parameter measurements from each point output in step 2 (COD, BOD, NH3-N, TP, TN, SS at the inlet and outlet of each stage), the actual values ​​of each indicator in the evaluation index system are calculated, compared with the preset thresholds, and the status analysis of each stage and the overall system evaluation are completed. 4.1 Status Analysis of Each Stage (1) Settling process state analysis: Combining suspended solids (SS) measurement, suspended solids removal rate (η1), sludge settling ratio (SV30), and sludge moisture content (W1), a state determination rule was constructed: ① Abnormal state: When η1<90%, SV30>30%, or W1>98%, the sedimentation process is determined to be in an abnormal state, and the fault is located according to the specific abnormal indicators: η1<90% and SV30>30%, warning of sludge accumulation and sedimentation tank blockage; η1<90% and SV30<15%, warning of insufficient flocculant addition; W1>98%, warning of sludge dewatering equipment failure; ② Normal state: When η1≥90%, SV30 is between 15% and 30% and W1≤98%, the precipitation process is judged to be in a normal and stable operating state; Meanwhile, by combining the turbidity correlation characteristics (absorbance changes in specific spectral bands) in high-resolution full-spectrum data, the precipitation effect can be judged. When the spectral absorbance increases abnormally, it can help to warn of malfunctions in the precipitation process, thereby improving the accuracy and timeliness of the status analysis.

[0032] (2) Oxidation process status analysis: Combining COD measurement value, COD removal rate (η2), oxidant utilization rate (η3), and effluent residual chlorine concentration (ρ), a status judgment rule is constructed: ① Abnormal state: When η2<85%, η3<80%, or ρ is not within the range of 0.5~1.0mg / L, the oxidation process is determined to be in an abnormal state, and fault location is performed: η2<85% and η3<80%, warning of insufficient oxidant dosage; η2<85% and ρ>1.0mg / L, warning of insufficient oxidation reaction time; ρ<0.5mg / L, warning of incomplete oxidation reaction; ρ>1.0mg / L, warning of excessive oxidant dosage; ② Normal state: When η2≥85%, η3≥80% and ρ is in the range of 0.5~1.0mg / L, the oxidation process is judged to be in a normal and stable operating state; By utilizing the variation range of characteristic peaks of organic compounds in high-resolution full-spectrum (such as the ultraviolet absorption peak near 254 nm), the sufficiency of oxidation reactions can be analyzed. When the intensity of the characteristic peaks decreases insufficiently, it can help to provide early warning of inadequate processing in the oxidation process.

[0033] (3) Status analysis of biological treatment process: Combining BOD, NH3-N, TN measurements, BOD removal rate (η4), microbial activity (A), dissolved oxygen concentration (DO), and sludge age (SRT), a status determination rule was constructed: ① Abnormal State: When η4 < 90%, A < 0.8, DO is not within the range of 2~4 mg / L, or SRT is not within the range of 8~12d, the biological treatment process is judged to be in an abnormal state, and fault location is performed: η4 < 90% and A < 0.8, warning of microbial community imbalance and poor growth environment; η4 < 90% and DO < 2 mg / L, warning of insufficient aeration; DO > 4 mg / L, warning of excessive aeration and energy waste; SRT > 12d, warning of sludge aging; SRT < 8d, warning of insufficient microbial biomass; ② Normal state: When η4≥90%, A≥0.8, DO is in the range of 2~4mg / L and SRT is in the range of 8~12d, the biological treatment process is judged to be in a normal and stable operating state. By combining the correlation characteristics of biological metabolites in high-resolution full-spectrum data (such as changes in the spectral peaks of organic matter produced by microbial metabolism), we can assist in assessing microbial activity and improve the accuracy of state analysis.

[0034] 4.2 Overall System Evaluation The Analytic Hierarchy Process (AHP) was used to assign weighted scores to the overall evaluation indicators of processing efficiency, energy intensity, operational stability, emission compliance rate, and operation and maintenance cost, with weights set at 0.4, 0.2, 0.2, 0.1, and 0.1 respectively, for a total score of 100. Based on the scoring results, the overall system operation status was divided into three levels: excellent (≥90 points), good (80~89 points), and unqualified (<80 points). When the system was rated as unqualified, the key links affecting the overall system operation effect were identified by combining the status analysis results of each link, providing a basis for subsequent optimization and control.

[0035] Step 5: Based on the status analysis results and evaluation conclusions, generate operation optimization suggestions and fault early warning information for each link to achieve closed-loop linkage of monitoring, evaluation and control.

[0036] The core of this step is to achieve seamless integration of "monitoring-evaluation-control". Based on the status analysis results of step 4 and the overall system evaluation conclusions, targeted operational optimization suggestions and fault early warning information are generated for each link to ensure that the shipborne wastewater treatment system is always in a highly efficient and stable operating state, as detailed below: (1) Optimization and control of the sedimentation process: For abnormal conditions in the sedimentation process, output corresponding optimization suggestions and early warning information: ① Sludge accumulation and sedimentation tank blockage: Issue a sedimentation tank cleaning warning, specifying the warning level (mild accumulation: SV30 = 30%~35%; moderate accumulation: SV30 = 35%~40%; severe accumulation: SV30 > 40%), and provide corresponding optimization and control steps: For mild accumulation, shorten the sludge discharge cycle from the original 72h to 48h, and increase the flocculant dosage by 10%~15% to improve sludge settling performance; for moderate accumulation, shorten the sludge discharge cycle to 2... After 4 hours, the flocculant dosage is increased by 15%~20%, and the bottom aeration and stirring of the sedimentation tank are started at the same time to prevent sludge from caking. In case of severe sludge accumulation, the sedimentation tank shutdown and cleaning warning is triggered immediately. High-pressure spraying combined with manual cleaning is used to clean the sludge accumulated on the tank walls and bottom. After cleaning, the equipment is restarted and first run at 50% load for 2 hours, and then gradually restore the normal load to ensure that after cleaning, η1≥90% and SV30 drops to 15%~30% (in line with the evaluation index requirements of the sedimentation process).

[0037] ② Insufficient flocculant dosage: It is recommended to increase the flocculant dosage by 1.1 to 1.2 times the current dosage, optimize the dosage ratio (if using a combination of polyaluminum chloride and polyacrylamide, adjust the mass ratio to 10:1 to 8:1), and change the dosage method to segmented dosing (70% at the inlet and 30% in the middle of the sedimentation tank). At the same time, monitor the changes in SV30 and η1 in real time, and adjust the dosage every 30 minutes until η1 ≥ 90% and SV30 stabilizes at 15% to 30%. If the standard is still not met after adjustment, trigger the flocculant effectiveness detection warning, check whether the flocculant is damp or ineffective, and replace it with qualified flocculant in time, adapting to the characteristics of limited storage space for shipboard flocculants and the need for precise dosage control.

[0038] ③ Sludge dewatering equipment malfunction: Output equipment maintenance warning and specify the key maintenance steps: First, shut down the sludge dewatering equipment and disconnect the power supply. Check if the equipment filter screen is clogged (if clogged, rinse with clean water or disassemble and clean), check if the filter roller pressure is normal (adjust to 0.3~0.5MPa), and check if the sludge feed pump is malfunctioning (check for pump blockage or motor abnormality). After maintenance, start the equipment and run it unloaded for 10 minutes. After confirming that there are no abnormalities, introduce settled sludge for dewatering treatment, monitor the sludge moisture content in real time, and adjust the equipment operating parameters to ensure that W1 ≤ 98% after dewatering. If the equipment malfunction cannot be repaired on-site, trigger the start command of the backup dewatering equipment and record the fault information for comprehensive maintenance after docking, avoiding affecting the continuous operation of the sedimentation process.

[0039] (2) Optimization and control of oxidation process: For abnormal states in the oxidation process, output corresponding optimization suggestions and early warning information: ① Insufficient oxidant dosage: Considering the limited onboard reagent storage, the dosage should be precisely controlled. It is recommended to gradually increase the dosage by 1.1 to 1.3 times the existing dosage (each adjustment should not exceed 10%) to avoid reagent waste. Optimize the dosing method. To address the problem of uneven reagent mixing caused by shipboard turbulence, adopt the "multi-point segmented dosing + bottom aeration and stirring" mode (adjust the aeration intensity to 0.8~1.2 m³ / (m²·h)) to ensure sufficient contact between the reagent and the wastewater. At the same time, monitor the COD removal rate (η2) and oxidant utilization rate (η3) in real time, testing every 20 minutes, until η2≥85% and η3≥80% (meeting the requirements of the sub-evaluation indicators of the oxidation process). If the standards are still not met after adjustment, trigger the oxidant effectiveness detection warning, check whether the reagent has become damp or ineffective due to shipboard temperature and humidity fluctuations, replace it with a qualified oxidant in time, and readjust the dosage according to the above steps after replacement.

[0040] ② Insufficient oxidation reaction time: Considering the compact space of the shipboard equipment, the oxidation reaction time will be extended by 5-10 minutes (not exceeding the upper limit of 40 minutes after extension, in line with the evaluation indicators of the oxidation process) without affecting subsequent treatment stages. Simultaneously, the water flow rate in the oxidation reactor will be adjusted from 0.8-1.2 m / h to 0.5-0.8 m / h to extend the contact time between wastewater and oxidant. The reactor stirring parameters will be optimized, increasing the stirring frequency (stirring for 5 minutes every 15 minutes) to address the uneven reaction caused by shipboard turbulence. The oxidation sufficiency will be monitored in real time using changes in the 254 nm organic characteristic peak in high-resolution full-spectrum data. When the characteristic peak intensity decreases by ≥80%, the oxidation reaction is confirmed to be sufficient, and the current reaction time will be maintained. If the target is still not met, the oxidant dosage can be appropriately increased (not exceeding 10% of the original amount) to synergistically improve the oxidation effect, ensuring that the final η2 ≥ 85% and ρ stabilizes at 0.5-1.0 mg / L.

[0041] ③ Excessive Oxidant Dosing: Immediately reduce the oxidant dosage by 10%~20% from the existing level and adjust to a segmented, small-volume dosing mode (the total amount of oxidant remains unchanged, but it is added in 3~4 times, with an interval of 15 minutes between each addition) to reduce the local residual chlorine concentration; monitor the residual chlorine concentration (ρ) of the effluent in real time, once every 15 minutes, until ρ drops back to the range of 0.5~1.0 mg / L; at the same time, monitor the microbial activity (A) of the subsequent biological treatment process. If A<0.8, a small amount of reducing agent (such as sodium sulfite, the dosage is calculated according to the residual chlorine concentration ratio of 1:1.5) needs to be added at the front end of the biological reactor to eliminate the inhibitory effect of excessive residual chlorine on microorganisms; subsequently, continuously monitor η2, ρ and microbial activity to ensure that η2≥85% and does not affect the subsequent biological treatment, adapting to the needs of shipboard chemical saving and process linkage.

[0042] ④ Oxidation Reactor Failure: Output equipment maintenance early warning, clarify the key points and steps of on-site shipboard maintenance, and adapt to common failures caused by shipboard turbulence (nozzle blockage, loose agitator, pipeline leakage): First, shut down the reactor and disconnect the power supply and reagent supply pipeline. Check if the reagent nozzles are blocked by wastewater impurities (if blocked, use high-pressure clean water to flush, avoid disassembly damage, adapt to on-site shipboard maintenance conditions); check if the agitator blades are loose and if the speed is normal (adjust the speed to 120~150r / min), and troubleshoot abnormal motor vibration (due to loose fixing bolts caused by shipboard turbulence, tighten them in time); check if there is a leak in the reagent delivery pipeline, and temporarily seal it with sealant (thoroughly inspect after docking); after maintenance, start the reactor and run it unloaded for 15 minutes. After confirming that there are no abnormalities, introduce wastewater and resume reagent addition. Monitor η2, η3, and ρ in real time to ensure that all indicators meet the standards; if the failure cannot be repaired on-site, trigger the start command of the backup oxidation reactor, record the failure details, facilitate comprehensive maintenance after docking, and ensure continuous operation of the oxidation process.

[0043] (3) Optimization and control of biological treatment process: For abnormal states in the biological treatment process, output corresponding optimization suggestions and early warning information: ① Microbial community imbalance and poor growth environment: Based on the status assessment results of the biological treatment process in step 4, an early warning for microbial community regulation is output. Considering the limited space and operational conditions of the shipborne bioreactor, specific regulation steps are given: Prioritize the supplementation of salt-tolerant and temperature-fluctuation-resistant dominant microbial communities (such as nitrifying and denitrifying bacteria compound agents), with a dosage controlled at 5-10 mg / L, added in two doses (12 hours apart) to avoid a single addition leading to microbial community competition imbalance; optimize the nutrient ratio in the bioreactor, supplementing glucose, urea, and other nutrients at a C:N:P ratio of 100:5:1. Addressing the issue of nutrient imbalance in shipborne wastewater; real-time adjustment of water temperature in the reaction tank (controlled at 25~35℃) and pH value (controlled at 7.0~8.0). If the water temperature fluctuates excessively due to the shipborne environment, activate the reaction tank constant temperature auxiliary device (adapted to shipborne energy consumption limits, power controlled at 0.5~1.0kW); simultaneously, based on the correlation characteristics of biological metabolites in high-resolution full-spectrum data, monitor microbial activity (A) every 24 hours until A≥0.8 to ensure stable microbial community recovery; if the microbial community imbalance is severe, 10%~15% of the aged sludge in the reaction tank can be discharged, and fresh compound microbial agent can be added to accelerate microbial community recovery.

[0044] ② Insufficient Aeration: Based on the status assessment results of the biological treatment stage in step 4, an early warning for adjusting the aeration intensity is output. Considering the limited energy resources on board, aeration parameters are precisely controlled: the aeration intensity is gradually adjusted from the original 0.6~0.8 m³ / (m²·h) to 1.0~1.2 m³ / (m²·h), with each adjustment not exceeding 0.2 m³ / (m²·h) to avoid a sudden increase in energy consumption; the operating parameters of the aeration equipment are optimized, and the aeration frequency is adjusted to an intermittent aeration mode (30 minutes of aeration, 10 minutes of aeration pause) to accommodate uneven aeration caused by shipboard turbulence. To address the issue of reduced ineffective energy consumption, check whether the aeration discs are clogged due to biofilm adhesion or sludge blockage (which is easily caused by large fluctuations in sludge content in shipboard wastewater). Use low-pressure clean water backwashing (pressure controlled at 0.2~0.3MPa) without disassembly, which is suitable for shipboard on-site maintenance conditions. Monitor dissolved oxygen concentration (DO) in real time, once every 15 minutes, until DO stabilizes at 2~4mg / L (meeting the requirements of the biological treatment sub-evaluation indicators). At the same time, monitor BOD removal rate (η4) to ensure that η4 gradually increases to ≥90% to avoid insufficient aeration leading to microbial hypoxia and inactivation.

[0045] ③ Over-aeration: Based on the status assessment results of the biological treatment process in step 4, an early warning for optimized aeration energy consumption is immediately issued, balancing treatment effectiveness with onboard energy conservation needs. Specific control steps include: reducing the aeration intensity from the original 1.3~1.5 m³ / (m²·h) to 0.8~1.0 m³ / (m²·h); adjusting the aeration mode to "aeration for 20 minutes, then aeration stop for 15 minutes" to reduce ineffective aeration time; closing some redundant aeration discs (20%~30% of the total number, evenly distributed) to avoid localized over-aeration; and real-time monitoring of DO concentration and microbial activity (A). Detection is performed every 20 minutes to ensure that DO is maintained at 2~4 mg / L and A≥0.8, while monitoring energy intensity (E) to ensure that E≤1.5kWh / m³ (meeting the overall system evaluation index requirements). If excessive aeration inhibits the denitrification reaction (reducing TN removal rate), an anaerobic section can be added at the end of the reaction tank (utilizing existing shipboard space for modification, without the need for new equipment) to extend the denitrification reaction time, ensure that TN treatment efficiency meets the standard, and achieve a dual balance between aeration optimization and treatment effect, consistent with the "reagent saving + process linkage" approach in the oxidation process control.

[0046] ④ Sludge Aging: Based on the status assessment results of the biological treatment stage in step 4, output a sludge renewal early warning, clarify the sludge age adjustment and discharge parameters, and adapt to the limited space of the shipboard sludge treatment: shorten the sludge age (SRT) from the original 13-15 days to 8-10 days (in line with the evaluation index requirements of the biological treatment sub-item), adjust the sludge discharge cycle from once every 24 hours to once every 12 hours, and control the discharge volume each time to 5%-8% of the total sludge in the reaction tank to avoid insufficient microbial biomass due to excessive discharge at one time; the discharged aged sludge is connected to the shipboard sludge dewatering system. The equipment (shared with the sedimentation stage, saving shipboard space, and with consistent sludge treatment logic with the sedimentation stage) is dewatered and stored for centralized processing upon arrival at the shore. Simultaneously, a small amount of fresh sludge is added (3%–5% of the total sludge volume in the reaction tank, which can be taken from qualified sludge in the shipboard sedimentation tank) to accelerate sludge renewal. Sludge age, microbial activity (A), and BOD removal rate (η4) are monitored in real-time, every 24 hours, until the SRT stabilizes at 8–12 days, A ≥ 0.8, and η4 ≥ 90%, thus completing sludge aging control and preventing a continuous decline in treatment efficiency due to sludge aging.

[0047] ⑤ Insufficient Microbial Biomass: Based on the status assessment results of the biological treatment process in step 4, an early warning for microbial cultivation is issued. Considering the characteristics of shipborne operation and maintenance, a mild cultivation plan is formulated to avoid nutrient waste caused by excessively rapid cultivation: the sludge retention time (SRT) is extended from the original 6-7 days to 11-12 days (in line with the requirement of controlling SRT to 8-12 days in the biological treatment sub-evaluation indicators), sludge discharge is reduced to once every 48 hours, and the discharge volume is controlled at 2%-3% of the total sludge in the reaction tank each time, retaining sufficient microorganisms; nutrient supply is optimized, and nutrients are supplemented according to the ratio of C:N:P=100:5:1, giving priority to the use of shipborne solid nutrients that are easy to store and add (such as slow-release urea, grape...). Sugar granules) reduce the storage pressure of liquid nutrients; appropriately increase the aeration intensity to 0.9~1.1 m³ / (m²·h) to ensure sufficient oxygen supply for microorganisms, while adopting a "low-intensity stirring + intermittent aeration" mode (adapted to shipboard turbulence) to promote uniform microbial growth; monitor microbial biomass (measured by MLSS concentration, controlled at 2000~3000 mg / L) and microbial activity (A) in real time, testing every 12 hours until microbial biomass meets the standard, A≥0.8, and η4≥90% (meeting the requirements of the biological treatment sub-evaluation indicators), complete microbial cultivation, adapt to the scenario of large fluctuations in shipboard microbial biomass, and maintain the "precise parameters + shipboard adaptation" style of other link control.

[0048] (4) System overall optimization and control: When the overall evaluation of the system is unqualified, the key influencing links are located by combining the status analysis results of each link, and the overall system optimization plan is output, such as optimizing the matching of operating parameters of each link, adjusting the processing flow, and overhauling core equipment; at the same time, based on the model self-learning log and the changing trend of evaluation indicators, the changes in the system operating status are predicted, and early warning information is output in advance to achieve proactive prevention and control, ensure the long-term stable operation of the system, and ensure that the treated wastewater meets the discharge standards.

[0049] In addition, this method includes a model calibration step: key water quality parameters are measured every 30 days using national standard methods (GB / T 11914-1989, GB / T7488-1987, etc.), and the measured values ​​are compared with the model predictions. If the average error exceeds the corresponding threshold, forced self-learning is triggered to ensure the long-term accuracy of the model. Simultaneously, salinity and temperature interference factors are added during model training to improve the model's resistance to shipboard environmental interference and adapt it to shipboard scenarios with high salinity and large temperature and humidity fluctuations. Beneficial Effects This invention provides a method for monitoring and evaluating shipborne wastewater treatment based on high-resolution full-spectrum soft sensing.

[0050] 1. This invention deeply integrates high-resolution full-spectrum technology with soft sensing technology to replace traditional hardware sensors for monitoring shipboard wastewater quality parameters. It requires no complex hardware maintenance, has strong resistance to interference from shipboard turbulence, temperature and salinity fluctuations, and is suitable for harsh shipboard operating environments. At the same time, the full-spectrum data contains rich water quality features, and combined with feature extraction algorithms, it can achieve simultaneous monitoring of multiple parameters, solving the problems of single parameters and low accuracy in traditional monitoring methods. The monitoring accuracy meets the requirements for ship wastewater discharge monitoring.

[0051] 2. The soft sensing model constructed in this invention integrates the CNN-LSTM hybrid machine learning algorithm with a self-learning optimization module. The CNN-LSTM model can deeply mine the correlation between spectral features and water quality parameters, capture the temporal change pattern of water quality, and improve prediction accuracy. The self-learning module can automatically update model parameters when water quality changes suddenly or the environment fluctuates, without the need for frequent manual calibration. It is suitable for the complex and variable characteristics of shipborne wastewater quality, reduces operation and maintenance costs, and the model adopts edge computing deployment to improve the real-time performance and security of data processing.

[0052] 3. This invention establishes a comprehensive evaluation index system for shipborne wastewater treatment systems, covering the overall evaluation of the system and the sub-evaluations of the three core links of sedimentation, oxidation, and biological treatment. The indexes are scientifically set and operable. Combined with the monitored water quality parameters and the operational data of each link, the operating status of each link can be accurately analyzed, and fault location and early warning can be achieved. This solves the problems of imperfection and poor real-time performance of existing evaluation methods and provides a reliable basis for the optimization and control of each link.

[0053] 4. This invention realizes a closed-loop linkage of "monitoring-evaluation-control". Based on the water quality monitoring results, the status analysis of each link and the overall system evaluation conclusion, it outputs targeted optimization suggestions and fault warnings. It can adjust the operating parameters of each link in a timely manner to ensure the efficient and stable operation of the wastewater treatment system, improve treatment efficiency, reduce energy consumption and operation and maintenance costs, ensure that the treated wastewater meets the discharge standards, complies with IMO ship environmental protection standards, and reduces pollution to the marine ecological environment.

[0054] 5. This invention is adapted to the characteristics of limited space and energy on board ships, and optimizes the operating parameters and equipment design requirements of each link, such as controlling the sedimentation time and oxidation reaction time. It adopts a waterproof and vibration-resistant acquisition system, which takes into account both the treatment effect and shipboard adaptability. It has good engineering application value and promotion prospects. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention. Figure 2 This is a structural block diagram of the high-resolution full-spectrum soft sensing acquisition system of the present invention; Figure 3 This is a structural block diagram of the CNN-LSTM hybrid model and the self-learning optimization module in this invention; Figure 4 This is a logic block diagram of the state analysis and closed-loop control of each stage in this invention. Detailed Implementation

[0056] The present invention will be further described in detail below with reference to specific embodiments. Example

[0057] This embodiment provides a method for monitoring and evaluating the treatment of shipborne wastewater based on high-resolution full-spectrum soft sensing. It is applied to the wastewater treatment system of a large cruise ship with a displacement of 50,000 tons. The designed treatment capacity of the shipborne wastewater treatment system is 5 m³ / h, and the treatment process is "grid-sedimentation-ozone oxidation-biological contact oxidation-disinfection-discharge". The specific implementation steps of this method are as follows: Step 1: Build a shipborne high-resolution full-spectrum soft sensing acquisition system to collect high-resolution full-spectrum data of shipborne wastewater, simultaneously collect easily measurable auxiliary parameters, and complete data preprocessing.

[0058] The high-resolution full-spectrum soft-sensor acquisition system is waterproof and vibration-resistant, suitable for cruise ship operation in sea states 3-6. It includes: a high-resolution full-spectrum instrument (spectral range 200-1100nm, spectral resolution 0.8nm, acquisition frequency 3 times / minute), with acquisition points located at the wastewater treatment system inlet, sedimentation tank outlet, ozone oxidation tank outlet, biological contact oxidation tank outlet, and the main outlet; an auxiliary parameter acquisition module (water temperature acquisition accuracy ±0.1℃, pH value ±0.01pH, turbidity ±0.1NTU, conductivity ±1μS / cm); a data transmission module (fiber optic network + LoRa technology, transmission delay ≤8ms); and a preprocessing module.

[0059] Spectral data preprocessing: Dark current correction (to eliminate instrument electronic noise), baseline correction (to eliminate spectral baseline drift), Savitzky-Golay filtering and smoothing (window size 5, polynomial order 2), PCA+SPA feature extraction (PCA extracts 10 principal components, SPA selects 20 characteristic spectral bands, covering bands with strong correlation to water quality parameters such as 254nm and 410nm).

[0060] Auxiliary parameter preprocessing: outlier data were removed using the ±3σ rule, min-max was normalized to the [0,1] interval, and missing values ​​were filled in using the K nearest neighbor interpolation method (K=5).

[0061] Step 2: Construct a soft-sensor water quality parameter prediction model based on machine learning and self-learning, and output the real-time measurement values ​​of key water quality parameters.

[0062] The basic machine learning model adopts a CNN-LSTM hybrid model: the CNN module contains 2 convolutional layers (3×3 kernel size, ReLU activation function) and 1 pooling layer (2×2 pooling kernel size) to extract deep features of the feature spectral data; the LSTM module contains 1 hidden layer (64 hidden units) and 1 output layer (Linear activation function) to capture the temporal variation pattern of water quality; the input is the preprocessed feature spectral data and easily measurable auxiliary parameters, and the output is the real-time measurement values ​​of COD, BOD, NH3-N, TP, TN, and SS.

[0063] The self-learning optimization module is configured as follows: the prediction error thresholds for COD, BOD, NH3-N, TP, TN, and SS are 3%, 5%, 2%, 2%, 3%, and 1%, respectively; the self-learning trigger conditions are: the prediction error exceeds the threshold three times consecutively, the change in COD in the influent is ≥50mg / L, the change in water temperature is ≥5℃, or the change in salinity is ≥1‰; incremental learning uses the stochastic gradient descent algorithm (learning rate 0.001, number of iterations 100), and 10-fold cross-validation ensures generalization ability; a model parameter update log is established to record the details of each self-learning session.

[0064] Model calibration: Water quality parameters are calibrated every 30 days using national standard methods (COD using potassium dichromate method GB / T 11914-1989, BOD using dilution and inoculation method GB / T 7488-1987, NH3-N using Nessler's reagent spectrophotometry GB / T 7479-1987, TP using ammonium molybdate spectrophotometry GB / T 11893-1989, TN using alkaline potassium persulfate digestion ultraviolet spectrophotometry GB / T 11894-1989, SS using gravimetric method GB / T 11901-1989). The measured values ​​are compared with the model prediction values. If the average error exceeds the corresponding threshold (e.g., average error of COD prediction > 3%), forced self-learning is triggered to update the model parameters and ensure the long-term accuracy of the model. Meanwhile, salinity (salinity of 30-35‰ in the cruise ship's navigation area) and temperature (water temperature of 5-32℃ in the navigation environment) interference factors are added during the model training process to improve the model's ability to resist interference from the ship's environment and adapt to the scenario of temperature and salinity fluctuations during cruise ship navigation.

[0065] Step 3: Establish a comprehensive evaluation index system for shipborne wastewater treatment systems and clarify the practical calculation standards for each index.

[0066] Based on the operational characteristics of cruise ship wastewater treatment systems and the IMO's "Global Ship Pollution Control Report 2024" standard, the evaluation index system mentioned above is used to refine the calculations and threshold controls during the implementation process: 3.1 Practical setting of overall evaluation indicators: Treatment efficiency η: COD, BOD ≥ 90%, NH3-N ≥ 95%, TP, TN ≥ 90%, SS ≥ 98%; Energy consumption intensity E ≤ 1.5 kWh / m³ (calculated based on a cruise ship wastewater treatment system rated power of 8 kW and treatment capacity of 5 m³ / h, E = 8 ÷ 5 = 1.6 kWh / m³, subsequently optimized to the standard range through adjustment); Operational stability S ≥ 95% (statistical continuous operating time, with monthly allowable downtime not exceeding 3.6 hours); Discharge compliance rate C = 100% (testing before each batch of discharge to ensure effluent quality meets the first-class standard for ship wastewater discharge); Operation and maintenance cost M controlled within 800 RMB / day (including reagent consumption, equipment maintenance, and consumable replacement).

[0067] 3.2 Practical setting of sub-item evaluation indicators: Sedimentation stage η1≥90%, SV30=15%~30%, t1=45min (considering both sedimentation effect and cruise ship equipment space), W1≤98%; Oxidation stage (ozone oxidation) η2≥85%, η3≥80% (ozone dosage controlled at 80mg / L, utilization rate calculated as the ratio of actual COD degradation ozone consumption to total dosage), t2=30min, ρ=0.5~1.0mg / L; Biological treatment stage (biological contact oxidation) η4≥90%, A≥0.8 (dehydrogenase activity is determined by TTC-dehydrogenase activity method), DO=2~4mg / L, SRT=10d (adapting to the sludge retention requirements of biological contact oxidation process).

[0068] Step 4: Collect the operating parameters of each stage, and combine them with the water quality parameter measurements to complete the status analysis and overall evaluation.

[0069] 4.1 Parameter Acquisition: Using the same frequency as the spectral data acquisition (3 times / minute), the operating parameters of each stage were collected: sedimentation stage (sludge settling ratio SV30, sedimentation time t1=45min, sludge moisture content W1), oxidation stage (ozone dosage 80mg / L, oxidation reaction time t2=30min, effluent residual chlorine concentration ρ), and biological treatment stage (microbial activity A, dissolved oxygen concentration DO, sludge age SRT=10d). Simultaneously, the water quality parameters at each point output in step 2 were collected (inlet COD=350mg / L, BOD=180mg / L, NH3-N=35mg / L, TP=5mg / L, TN=40mg / L, SS=200mg / L; corresponding parameters at the effluent outlet of each stage were monitored synchronously).

[0070] 4.2 Practical Operation of Status Analysis at Each Stage: (1) Sedimentation stage: Calculate η1=(influent SS-effluent SS) / influent SS×100%=(200-18) / 200×100%=91% (≥90%), measured SV30=22% (15%~30%), W1=97.5% (≤98%), combined with the full spectrum turbidity correlation characteristics (absorbance in 280nm band is stable at 0.3~0.4), it is determined that the sedimentation stage is in a normal and stable operating state and no adjustment is required.

[0071] (2) Oxidation stage: η2 = (COD of effluent from sedimentation - COD of effluent from oxidation) / COD of effluent from sedimentation × 100% = (315 - 42) / 315 × 100% = 86.7% (≥ 85%), η3 = (315 × 5 × 10⁻³ × 8 × 32 / 16) / (80 × 10⁻³ × 5) × 100% = 84% (≥ 80%), the measured ρ = 0.7 mg / L (0.5 ~ 1.0 mg / L), combined with the decrease of 88% (≥ 80%) in the intensity of the characteristic peak of organic matter at 254 nm of the full spectrum, it is determined that the oxidation stage is operating normally; in a later monitoring, ρ = 0.3 mg / L (< 0.5 mg / L) and η2 = 82% (< 85%), it is determined that the oxidation reaction is insufficient, an ozone nozzle blockage warning is issued, and the subsequent control process is initiated.

[0072] (3) Biological treatment process: η4 = (BOD of effluent from oxidation - BOD of effluent from biological treatment) / BOD of effluent from oxidation × 100% = (153-12) / 153 × 100% = 92.2% (≥90%). The measured values ​​were A = 0.85 (≥0.8), DO = 3.2 mg / L (2~4 mg / L), and SRT = 10d (8~12d). Based on the correlation characteristics of biological metabolites across the entire spectrum (stable absorbance in the 350nm band), the biological treatment process was determined to be operating normally. During a certain voyage, uneven aeration was caused by the turbulence of the cruise ship. The measured values ​​were DO = 1.8 mg / L (<2 mg / L) and η4 = 87% (<90%). This was determined to be insufficient aeration, triggering an aeration intensity adjustment warning.

[0073] 4.3 Overall System Evaluation: Using the analytic hierarchy process (AHP), the following factors were weighted and scored: processing efficiency (weight 0.4), energy intensity (0.2), operational stability (0.2), emission compliance rate (0.1), and maintenance cost (0.1). The measured scores were: processing efficiency 92, energy intensity 88, operational stability 96, emission compliance rate 100, and maintenance cost 85. The total score was 92×0.4+88×0.2+96×0.2+100×0.1+85×0.1=91.3, which was rated as excellent. The overall system operation was good.

[0074] Step 5: Based on the status analysis results, implement closed-loop control to ensure stable system operation.

[0075] 5.1 Abnormal Control of Oxidation Process (Ozone Nozzle Clog): For the abnormal situation of ρ=0.3mg / L and η2=82%, implement the optimization control steps described above: Shut down the machine and disconnect the ozone supply pipeline, flush the ozone nozzle with high-pressure clean water (pressure 0.3MPa) to remove impurities (caused by residual suspended solids in cruise ship wastewater) clogging the nozzle. No disassembly is required, which is suitable for on-site maintenance conditions on cruise ships. After flushing, start the oxidation reactor and run it unloaded for 15 minutes. After confirming that the nozzle spray is uniform, resume ozone addition (dosage remains at 80mg / L), monitor ρ and η2 in real time, and check every 15 minutes. After 30 minutes, ρ rises to 0.7mg / L and η2 rises to 86.5%, and normal operation is restored. Record the control details to the system log.

[0076] 5.2 Abnormal Control in Biological Treatment (Insufficient Aeration): For the abnormal situation of DO=1.8mg / L and η4=87%, the following adjustments were implemented: The aeration intensity was gradually adjusted from the original 0.7m³ / (m²·h) to 1.0m³ / (m²·h) (each adjustment was 0.1m³ / (m²·h), meeting the requirement mentioned above that "each adjustment increment should not exceed 0.2m³ / (m²·h)"); the optimized aeration mode was set to "aeration for 30 minutes, then aeration stopped for 10 minutes" to reduce uneven aeration caused by shipboard turbulence; low-pressure clean water (0.25MPa) was used. Backwash the aeration discs to remove the biofilm and sludge adhering to the surface (meeting the requirement of "low-pressure clean water backwashing pressure 0.2~0.3MPa" mentioned above); monitor DO and η4 in real time. After 20 minutes, DO rises to 3.0mg / L (meeting the requirement of DO=2~4mg / L for biological treatment), and after 40 minutes, η4 rises to 91.5% (meeting the requirement of η4≥90%). Normal operation is restored. At the same time, monitor the energy consumption intensity E to drop to 1.45kWh / m³ (≤1.5kWh / m³, meeting the overall evaluation index requirements), thus meeting the overall evaluation index requirements.

[0077] 5.3 Overall System Optimization: Based on monthly operation data, it was found that the ozone utilization rate in the oxidation process fluctuated between 80% and 84%. To further reduce operation and maintenance costs, the ozone addition method was optimized to "multi-point segmented addition" (60% added at the front end of the oxidation tank and 40% added in the middle). After optimization, the ozone utilization rate increased to 86% to 88%, and the operation and maintenance cost was reduced to 780 yuan / day (≤800 yuan / day). At the same time, the energy consumption intensity was stabilized at 1.4 to 1.45 kWh / m³, and the overall operating efficiency of the system was further improved.

[0078] Step 6: Verify the implementation effect and compare the data.

[0079] This embodiment compares the method of the present invention with traditional monitoring and evaluation methods (relying on hardware sensors, lacking self-learning models, and using a single indicator for evaluation), running continuously for 3 months (90 days). The comparison results are as follows: (1) Monitoring accuracy: The average prediction errors of COD, BOD, NH3-N, TP, TN and SS by the method of the present invention are 2.3%, 4.1%, 1.7%, 1.5%, 2.5% and 0.8%, respectively, all of which are lower than the corresponding thresholds and far lower than the traditional method (average errors of 8.5%, 10.2%, 7.8%, 6.9%, 8.1% and 5.3%, respectively). It has stronger resistance to temperature and salinity fluctuations and ship turbulence interference. No monitoring failures occurred within 3 months. The traditional method had 12 monitoring data anomalies (due to sensor failure and interference).

[0080] (2) Status analysis and early warning: The method of the present invention identified a total of 8 abnormal states (3 instances of mild siltation in the sedimentation stage, 2 instances of nozzle blockage in the oxidation stage, and 3 instances of insufficient aeration in the biological treatment stage), with an early warning accuracy of 100% and an average early warning response time of ≤5 min. The control could be initiated in time without any escalation of the fault. The traditional method only identified 3 serious abnormalities, with an early warning response time of ≥30 min and 2 instances of risk of failing to meet emission standards (due to the failure to handle the abnormalities in time, the COD of the effluent temporarily exceeded the standard).

[0081] (3) System operation effect: After adopting the method of the present invention, the system processing efficiency is stable at 91%~94%, the average energy consumption intensity is 1.42kWh / m³ (≤1.5kWh / m³), the operation stability is 98.3% (≥95%), the emission compliance rate is 100%, and the average operation and maintenance cost is 770 yuan / day (≤800 yuan / day); when the traditional method is running, the processing efficiency fluctuates between 82%~88%, the average energy consumption intensity is 1.78kWh / m³, the operation stability is 89.2%, the emission compliance rate is 96.7%, and the average operation and maintenance cost is 920 yuan / day. The method of the present invention has significant advantages in operation efficiency, economy and stability.

[0082] (4) Operation and maintenance difficulty: The method of the present invention does not require frequent maintenance of hardware sensors, only national standard calibration once every 30 days, and the model can automatically learn and optimize. The monthly operation and maintenance time is ≤8h. The traditional method requires daily inspection and maintenance of hardware sensors, with monthly operation and maintenance time ≥30h, and manual calibration of model parameters is required. The operation and maintenance difficulty is much higher than that of the method of the present invention. Example

[0083] This embodiment provides the application of the method of the present invention in a small cargo ship (10,000 tons displacement, onboard wastewater treatment system designed to treat 1 m³ / h, process of "sedimentation-sodium hypochlorite oxidation-activated sludge process-discharge"). The core steps are the same as in Embodiment 1, with only some parameters adjusted according to the limited space and energy of the cargo ship. 1. High-resolution full-spectrum soft sensing acquisition system: The acquisition frequency is adjusted to 1 time / minute, the spectrometer has a spectral resolution of 1nm, and the data transmission adopts LoRa single technology (adapted to small cargo ship equipment configuration), with a transmission delay of ≤10ms; 2. Soft sensing model: The number of hidden units in the CNN-LSTM hybrid model was adjusted to 32 (to adapt to the computing power of edge computing nodes of small cargo ships), and the COD change in the self-learning trigger condition was adjusted to ≥40mg / L (the water quality fluctuation of wastewater from small cargo ships is relatively small). 3. Evaluation indicators: The sedimentation time t1 is adjusted to 35 min (meeting the requirement of t1=30~60 min for the sedimentation stage), the oxidation reaction time t2 is adjusted to 25 min (meeting the requirement of t2=20~40 min for the oxidation stage), the energy consumption intensity E≤1.6kWh / m³ (adapting to the power supply capacity of small cargo ships, and adjusting the threshold specifically in light of the overall requirement of limited shipboard energy, consistent with the adaptability description of the overall evaluation indicators above), and the operation and maintenance cost M is controlled within 500 yuan / day; 4. Control parameters: Aeration intensity adjustment range is 0.8~1.0 m³ / (m²·h) (which conforms to the aeration intensity control range of biological treatment and is suitable for the energy consumption limitations of small cargo ships), flocculant dosage adjustment range is ≤8% (which is in line with the limited agent storage of small cargo ships and is lower than the conventional adjustment range of 10% mentioned above, and is reasonably adapted to the scenario), and microbial agent dosage is adjusted to 3~8 mg / L (which conforms to the dosage range of 5~10 mg / L mentioned above and is suitable for the limited space of the reaction tank of small cargo ships).

[0084] After two months (60 days) of continuous operation, the method of this invention can still be stably applied on small cargo ships. The monitoring accuracy, anomaly warning accuracy, and system operation effect all meet the requirements. The processing efficiency is stable at 90%~93%, the emission compliance rate is 100%, and the average operation and maintenance cost is 480 yuan / day. It is suitable for small cargo ships with limited space and energy, and solves the problems of high operation and maintenance difficulty and weak anti-interference ability of traditional methods on small cargo ships.

[0085] In summary, the method of this invention is adaptable to wastewater treatment systems of ships of different tonnages, requires no complex hardware maintenance, has strong resistance to shipboard environmental interference, and can realize closed-loop linkage of water quality monitoring, status evaluation, and operation control, thereby improving the efficiency of ship wastewater treatment, reducing energy consumption and operation and maintenance costs, ensuring that wastewater is discharged in compliance with IMO ship environmental protection standards, and has broad engineering application value and promotion prospects.

Claims

1. A method for monitoring and evaluating shipborne wastewater treatment based on high-resolution full-spectrum soft sensing, characterized in that, Includes the following steps: Step 1: Build a shipborne high-resolution full-spectrum soft sensing acquisition system to collect high-resolution full-spectrum data of shipborne wastewater, simultaneously collect easily measurable auxiliary parameters, and complete data preprocessing; Step 2: Construct a soft-sensor water quality parameter prediction model based on machine learning and self-learning. Using preprocessed full-spectrum data and easily measurable auxiliary parameters as input, train the model and achieve self-learning optimization to output real-time measured values ​​of key water quality parameters of shipborne wastewater. Step 3: Establish a comprehensive evaluation index system for shipborne wastewater treatment system. The evaluation index system includes overall evaluation indexes and sub-evaluation indexes for the three core links of sedimentation, oxidation and biological treatment. Step 4: Collect the operating parameters of each link of the shipborne wastewater treatment system, combine them with the water quality parameter measurement values ​​output in Step 2, substitute them into the evaluation index system, and complete the status analysis of each link and the overall evaluation of the system. Step 5: Based on the status analysis results and evaluation conclusions, generate operation optimization suggestions and fault early warning information for each link to achieve closed-loop linkage of monitoring, evaluation and control.

2. The method according to claim 1, characterized in that, The high-resolution full-spectrum soft-sensor acquisition system described in step 1 includes a high-resolution full-spectrum instrument, an auxiliary parameter acquisition module, a data transmission module, and a preprocessing module. The high-resolution full-spectrum instrument has a spectral range of 200~1100nm, a spectral resolution of ≤1nm, an acquisition frequency of 1~5 times / minute, and acquisition points covering the wastewater treatment system inlet, sedimentation stage outlet, oxidation stage outlet, biological treatment stage outlet, and total outlet. The easily measurable auxiliary parameters include water temperature, pH value, turbidity, and conductivity, with acquisition accuracies of ±0.1℃, ±0.01pH, ±0.1NTU, and ±1μS / cm, respectively.

3. The method according to claim 1, characterized in that, The data preprocessing in step 1 includes spectral data preprocessing and auxiliary parameter preprocessing. The spectral data preprocessing includes dark current correction, baseline correction, spectral smoothing, and feature extraction. The feature extraction uses principal component analysis (PCA) combined with the continuous projection algorithm (SPA) to remove redundant spectral information and retain the characteristic spectral bands that are most correlated with water quality parameters. The auxiliary parameter preprocessing includes outlier removal (using the ±3σ rule), data normalization (using min-max normalization), and missing value completion (using K-nearest neighbor interpolation).

4. The method according to claim 1, characterized in that, The machine learning and self-learning soft-sensor water quality parameter prediction model mentioned in step 2 includes a basic machine learning model and a self-learning optimization module. The basic machine learning model adopts a CNN-LSTM hybrid model, in which the CNN module is used to extract deep correlation features of full-spectrum feature data, and the LSTM module is used to capture the temporal variation law of water quality parameters and the dynamic mapping relationship between spectral features and water quality parameters. The key water quality parameters include chemical oxygen demand (COD), biochemical oxygen demand (BOD), ammonia nitrogen (NH3-N), total phosphorus (TP), total nitrogen (TN), and suspended solids (SS), with measurement accuracies of ±3%, ±5%, ±2%, ±2%, ±3%, and ±1%, respectively.

5. The method according to claim 4, characterized in that, The implementation process of the self-learning optimization module is as follows: S21: Set the model prediction error threshold (the error thresholds for COD, BOD, NH3-N, TP, TN, and SS are set to 3%, 5%, 2%, 2%, 3%, and 1%, respectively) and self-learning trigger conditions. The trigger conditions include: the prediction error exceeds the corresponding threshold three times consecutively, a sudden change in the influent wastewater quality (COD change ≥ 50 mg / L), and drastic fluctuations in shipborne environmental parameters (water temperature change ≥ 5℃ or salinity change ≥ 1‰). S22: When the self-learning trigger condition is met, the model automatically collects the latest spectral data, auxiliary parameters and corresponding measured values ​​of water quality parameters (obtained periodically through national standard methods) to build an incremental training dataset. S23: The incremental learning algorithm is used to update the parameters of the CNN-LSTM hybrid model, retaining the original effective parameters of the model and only fine-tuning the parameters that do not match the new data. Cross-validation is used during the update process to ensure the generalization ability of the model. After the update is completed, the model accuracy is automatically verified until the prediction error is lower than the corresponding threshold. S24: Establish a model parameter update log to record the triggering reasons, updated parameters and accuracy changes for each self-learning, so as to achieve continuous optimization of model performance, adapt to the complex and variable water quality of shipborne wastewater and the large environmental interference, and solve the problems of poor generalization and frequent manual calibration of traditional models.

6. The method according to claim 1, characterized in that, The comprehensive evaluation index system for the shipborne wastewater treatment system mentioned in step 3 is as follows: The overall evaluation indicators include: treatment efficiency (η), energy intensity (E), operational stability (S), emission compliance rate (C), and operation and maintenance cost (M); where treatment efficiency η = (influent water quality parameter value - effluent water quality parameter value) / influent water quality parameter value × 100%, and must meet the following requirements: η ≥ 90% (COD, BOD), η ≥ 95% (NH3-N), η ≥ 90% (TP, TN), η ≥ 98% (SS); energy intensity E = electricity consumption per unit volume of wastewater treated (kWh / m³). The following conditions must be met: E≤1.5kWh / m³; Operational stability S=(continuous stable operation time / total operation time)×100%, which must meet S≥95%; Emission compliance rate C=(number of compliant emissions / total number of emissions)×100%, which must meet C=100% (complying with the IMO's "2024 Global Ship Pollution Control Report" and relevant ship wastewater discharge standards); Operation and maintenance cost M=total cost of equipment maintenance, reagent consumption and consumable replacement per unit time (yuan / day), which must be controlled within the preset threshold. The evaluation indicators for the sedimentation process include: suspended solids removal rate (η1), sludge settling ratio (SV30), sedimentation time (t1), and sludge moisture content (W1); where η1 ≥ 90%, SV30 is controlled at 15%~30%, t1 is controlled at 30~60 min, and W1 ≤ 98%; The evaluation indicators for the oxidation process include: COD removal rate (η2), oxidant utilization rate (η3), oxidation reaction time (t2), and residual chlorine concentration in the effluent (ρ); where η2 ≥ 85%, η3 ≥ 80%, t2 is controlled at 20~40 min, and ρ is controlled at 0.5~1.0 mg / L. The evaluation indicators for the biological treatment process include: BOD removal rate (η4), microbial activity (A), dissolved oxygen concentration (DO), and sludge time (SRT); where η4 ≥ 90%, microbial activity A ≥ 0.8 (measured by dehydrogenase activity), DO controlled at 2~4 mg / L, and SRT controlled at 8~12 days.

7. The method according to claim 1, characterized in that, The status analysis of each stage described in step 4 specifically includes: Sedimentation process status analysis: Combining suspended solids (SS) measurements, suspended solids removal rate (η1), sludge settling ratio (SV30), and sludge moisture content (W1), when η1 < 90%, SV30 > 30%, or W1 > 98%, the sedimentation process is considered to be in an abnormal state, providing an early warning of sludge accumulation, sedimentation tank blockage, or insufficient flocculant dosage; when η1 ≥ 90%, SV30 is between 15% and 30%, and W1 ≤ 98%, the sedimentation process is considered to be in a normal and stable operating state; simultaneously, the turbidity correlation characteristics in the spectral data are used to assist in judging the sedimentation effect and improve the accuracy of the status analysis; Oxidation process status analysis: Combining COD measurement values, COD removal rate (η2), oxidant utilization rate (η3), and effluent residual chlorine concentration (ρ), when η2 < 85%, η3 < 80%, or ρ is not within the range of 0.5~1.0 mg / L, the oxidation process is determined to be in an abnormal state, providing an early warning of unreasonable oxidant dosage, insufficient reaction time, or oxidation reactor malfunction; when η2 ≥ 85%, η3 ≥ 80%, and ρ is within the range of 0.5~1.0 mg / L, the oxidation process is determined to be in a normal and stable operating state; the variation amplitude of characteristic peaks of organic matter in the full spectrum is used to assist in analyzing the sufficiency of the oxidation reaction; Biological treatment process status analysis: Combining BOD, NH3-N, TN measurements, BOD removal rate (η4), microbial activity (A), dissolved oxygen concentration (DO), and sludge age (SRT), when η4 < 90%, A < 0.8, DO is not within the range of 2-4 mg / L, or SRT is not within the range of 8-12 days, the biological treatment process is considered to be in an abnormal state, providing early warning of microbial community imbalance, insufficient aeration, or unreasonable sludge discharge; when η4 ≥ 90%, A ≥ 0.8, DO is within the range of 2-4 mg / L, and SRT is within the range of 8-12 days, the biological treatment process is considered to be in a normal and stable operating state; combining the correlation characteristics of biological metabolites in spectral data assists in assessing microbial activity and improves the accuracy of status analysis.

8. The method according to claim 1, characterized in that, The closed-loop linkage control mentioned in step 5 specifically refers to: for abnormalities in the sedimentation process, outputting suggestions for adjusting the flocculant dosage, optimizing the sludge discharge cycle, or issuing a sedimentation tank cleaning warning; for abnormalities in the oxidation process, outputting suggestions for adjusting the oxidant dosage, optimizing the oxidation reaction time, or issuing a reactor maintenance warning. For abnormalities in the biological treatment process, suggestions for adjusting aeration intensity, microbial acclimatization, or sludge age optimization are provided. At the same time, based on the overall system evaluation results, when the treatment efficiency η does not meet the standard, the energy consumption intensity E exceeds the standard, or the operational stability S < 95%, an overall system optimization plan is provided to achieve seamless integration of water quality monitoring, status evaluation, and operational control.

9. The method according to claim 1, characterized in that, The method also includes a model calibration step: every 30 days, key water quality parameters are measured using national standard methods (GB / T 11914-1989, GB / T 7488-1987, etc.), and the measured values ​​are compared with the model prediction values. If the average error exceeds the corresponding threshold, forced self-learning is triggered to ensure the long-term accuracy of the model. At the same time, to adapt to the shipborne environment with high salinity and large temperature and humidity fluctuations, salinity and temperature interference factors are added during the model training process to improve the model's anti-interference ability.

10. The method according to claim 1, characterized in that, The high-resolution full-spectrum soft sensing acquisition system adopts a waterproof and vibration-resistant design, and is suitable for shipboard operation in sea states 3 to 6. Data transmission employs a redundant design combining fiber optic networks and LoRa technology to ensure data transmission stability and real-time performance, with a transmission latency of ≤10ms. Model deployment utilizes an edge computing architecture to reduce cloud dependency and enhance data security and real-time processing efficiency.

Images