Water treatment optimization system based on multispectral technology combination and regulation and control method

Abstract

The invention discloses a water treatment optimization system based on multispectral technology combination and a regulation and control method, and belongs to the technical field of software. The invention aims to solve the problems of serious redundancy, noise amplification or information idleness caused by lack of a systematic decision-making process in state vector construction in a multispectral combination scene in the prior art. Therefore, the invention provides a water treatment optimization system and a regulation and control method, and the method comprises the steps: constructing a candidate state set containing multispectral analysis indexes and operation parameters; establishing a prediction contribution degree, signal stability and control sensitivity three-dimensional evaluation system; key state quantities are dynamically selected through a layered screening strategy; and generating a refined and information-rich target state vector for a control decision. Compared with the prior art, the method has the advantages that refining and information enrichment of state vectors are realized through dynamic screening and derivation construction mechanisms, the control precision and the calculation efficiency are remarkably improved, and the noise robustness is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water treatment, more particularly, the present application relates to a water treatment optimization system based on multi-spectral technology combination and a regulation and control method. BACKGROUND

[0002] In existing sewage treatment plants, automatic control or intelligent regulation and control systems usually rely on a small amount of online sensor data and indirectly estimated COD, ammonia nitrogen and other indicators to form a process state vector. Although this traditional method has a lower dimension, it is easy to use simple PID control or empirical rule-based control, but its ability to finely describe pollutant components and bacterial function is limited.

[0003] In recent years, multi-spectral combination technologies such as near-infrared, ultraviolet-visible, three-dimensional fluorescence, and Raman have been gradually introduced into sewage treatment processes, enabling operators to obtain spectral prediction values of conventional indicators such as COD, TN, and TP at a high time resolution, and to analyze protein components, humic acid components, and their aromaticity of dissolved organic matter (DOM) in real time. Even indirectly reflect the metabolic activity of functional bacteria such as nitrifying bacteria through specific Raman or fluorescence signals. This significantly increases the dimensionality of the process observables and greatly improves the state perception capability. However, this also presents new problems: at each control time step, the system obtains a large number of candidate information from the multi-spectral analysis module and the process operation module, including various spectral features, DOM sub-component proportions, NADH activity indices, DO current values and their historical changes, influent flow Q and its load fluctuations, aeration intensity and historical adjustment trajectories, forming a candidate state set with high dimensionality and complex correlation between variables.

[0004] The prior art has the following disadvantages: In the prior art, there are mainly two types of utilization methods for these candidate state quantities: the first type is basically not to screen, and all available spectral analysis results and operation data are used as inputs to some prediction model or control model. This approach leads to dimensionality expansion of the state vector, with a large amount of redundancy and strong correlation between variables, making model training difficult, increasing real-time operation computation burden, and being very sensitive to noise and outliers. For example, when 50 or more spectral features and operating parameters are all input into a neural network model, the model converges slowly and is prone to overfitting, resulting in unstable control strategies.

[0005] The second type still uses the empirical selection method, and a process operator subjectively selects a few important indicators from a large number of candidate indicators according to experience, for example, only DO, Q, a few spectrally predicted CODs, ammonia nitrogen (NH4-N) and the like are reserved as control inputs, and information such as DOM structure evolution and microbial activity is ignored. Although this method can maintain the simplicity of the model, a large amount of potential useful information is idle, and the monitoring advantage brought by multi-spectral technology cannot be fully utilized. For example, when the carbon source is insufficient in the anoxic section, the change of DOM structure may indicate that the denitrification efficiency is decreasing, but if this information is ignored, the carbon source dosage cannot be adjusted in time, resulting in fluctuations in the TN removal rate.

[0006] To solve the above problems, the present application provides a solution. SUMMARY

[0007] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide a water treatment optimization system and control method based on the combination of multi-spectral technology to solve the problems raised in the background art.

[0008] To achieve the above-mentioned purpose, the present application provides the following technical scheme: A water treatment optimization system and control method based on the combination of multi-spectral technology, comprising the following steps: Obtaining multi-spectral analysis results output by a multi-spectral analysis module and operation data provided by a process control parameter module to form a candidate state set containing the multi-spectral analysis results and the operation data; Evaluating each candidate variable in the candidate state set from three dimensions of prediction contribution, signal stability and control sensitivity to form multi-criteria evaluation results; According to the multi-criteria evaluation results, a hierarchical screening and sorting method is used to select a limited number of key state variables from the candidate state set to form a basic state subset; According to the historical time series and the change trend of the spectral characteristics of the state variables in the basic state subset, a derived state variable reflecting the dynamic characteristics of the process is constructed; The derived state variable is combined with the basic state subset to generate a target state vector for the current control target, which is moderate in dimension and rich in information, and is used as the input of an operation decision module.

[0009] In a preferred embodiment, the multi-spectral analysis results include conventional water quality indicators and spectral characteristic variables obtained or predicted by near-infrared and ultraviolet-visible light, dissolved organic matter component information analyzed by three-dimensional fluorescence, and NADH activity index calculated by Raman spectroscopy.

[0010] In a preferred embodiment, the operation data includes online dissolved oxygen, influent flow rate, historical aeration intensity, valve opening, and carbon source dosage.

[0011] In a preferred embodiment, the evaluating step of the prediction contribution includes analyzing the influence of introducing or removing a certain candidate quantity on the prediction error of the target such as short-term dissolved oxygen or total nitrogen, to determine the contribution of the candidate quantity to the current control target.

[0012] In a preferred embodiment, the evaluating step of the signal stability includes calculating the variance, abnormal point proportion, and drift trend of the candidate quantity in the recent period of time, to evaluate its measurement noise and stability.

[0013] In a preferred embodiment, the evaluating step of the control sensitivity includes reviewing the relationship between historical control actions and process responses, and analyzing the sensitivity correlation between the change of the candidate quantity and the change of the control object.

[0014] In a preferred embodiment, the hierarchical screening and sorting method includes: First, removing the candidate quantity that has little effect on the target according to the prediction contribution; Second, removing the candidate quantity that is susceptible to noise interference according to the signal stability; Finally, sorting the remaining set according to the control sensitivity, and selecting a limited number of key state quantities to form the basic state subset.

[0015] In a preferred embodiment, the derived state quantity includes: Dissolved oxygen recovery time or dissolved oxygen response intensity caused by unit aeration adjustment, which is constructed according to the decay trajectory of dissolved oxygen deviation over time; Dissolved organic matter structure mutation amplitude, which is constructed according to the short-term change of dissolved organic matter fluorescence peak intensity and E2 / E3 ratio; wherein, E2 represents the absorbance of ultraviolet-visible spectrum at 254 nm wavelength, and E3 represents the absorbance of ultraviolet-visible spectrum at 365 nm wavelength; Active inhibition duration, which is constructed according to the length of time that the NADH activity index continuously remains below the reference value.

[0016] A water treatment optimization control system based on the combination of multi-spectral technology, comprising: A multi-spectral analysis module for outputting multi-spectral analysis results; A process control parameter module for providing operation data; An operation decision module, which is internally added with a state vector screening and derivation construction unit, is connected with the multi-spectrum analysis module and the process control parameter module, used to receive the multi-spectrum analysis result and the operation data, and execute the above method.

[0017] In a preferred embodiment, the operation decision module further comprises a rule layer and a model layer, and the target state vector is used as the unified input of the rule layer and the model layer, so as to realize the synergistic optimization control of the aeration fan frequency, the aeration valve opening degree and the carbon source dosage.

[0018] The water treatment optimization system and the control method based on the multi-spectrum technology have the following technical effects and advantages: Through the dynamic screening mechanism, the dimension of the state vector is reduced from the original 50+ to 8-12 core variables, and the dimension is reduced by 70%. This greatly reduces the model reasoning time and improves the calculation efficiency, so as to meet the requirements of real-time control and reduce the calculation burden of the operation decision module.

[0019] Through the construction of the refined and information-rich target state vector, the DO control deviation is reduced, and the TN removal rate fluctuation range is narrowed. This shows that the present application can more accurately reflect the process state, so as to realize more accurate control and improve the efficiency and stability of water treatment.

[0020] Through the introduction of the derived state quantity construction mechanism, the key information reflecting the process dynamic characteristics and the functional state of the microbial community is included in the state vector, and the problem that a large amount of potential useful information is idle in the traditional method is avoided. For example, through the derived state quantities such as DO recovery time, dissolved organic matter DOM structure mutation amplitude and active inhibition duration, the system can more comprehensively perceive the process dynamic response, and provide more abundant information for control decision.

[0021] The present application establishes a three-dimensional evaluation system of prediction contribution, signal stability and control sensitivity, provides a systematic decision framework for control target, and solves the problem of lack of clear selection mechanism in the construction of state vector in the prior art. This makes the utilization of multi-spectrum data in the control level more scientific and reasonable, and realizes the integrated technical path from deeper and finer observation to more accurate and more economical adjustment. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 The structure diagram of the water treatment optimization system and the control method based on the multi-spectrum technology of the present application; Figure 2 The flowchart of the water treatment optimization control method based on the multi-spectrum technology of the present application; Reference signs: 110, multi-spectral analysis module; 120, process control parameter module; 130, operation decision module; 131, state vector screening and derivation construction unit; 132, rule layer; 133, model layer. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments.

[0024] Embodiment one, the embodiment provides a water treatment optimization method based on multi-spectral technology combination, the core of which is to extract refined and information-rich state vectors from massive multi-spectral data and operation data through dynamic screening and derivation construction mechanism, to support the optimization control of water treatment process. In each control time step, the specific control target at present is taken as the core, it is decided comprehensively which state quantity should be included in the comprehensive state vector, which state quantity should be temporarily discarded, and whether the derived state quantity reflecting the dynamic characteristics of the process needs to be dynamically supplemented.

[0025] Reference Figure 2 The method includes the following steps: Step 210: Constructing a candidate state set. In each control time step, the system first obtains multi-spectral analysis results from the multi-spectral analysis module (110), and obtains operation data from the process control parameter module (120). These data together constitute the current candidate state set.

[0026] Specifically, the multi-spectral analysis results include: Routine water quality indicators predicted by near-infrared (NIR) and ultraviolet-visible (UV-Vis) spectroscopy, such as chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), etc. These indicators provide concentration information of main pollutants in water.

[0027] Dissolved organic matter (DOM) component information analyzed by three-dimensional fluorescence spectroscopy, such as humic acid substances, protein substances, soluble microbial metabolites, etc. These component information reflects the source, structure and biodegradability of organic matter.

[0028] NADH activity index calculated by Raman spectroscopy. NADH is an important coenzyme in microbial metabolic process, and its activity index can indirectly reflect the metabolic intensity of microorganisms and the activity of functional flora. Specifically, the reduced form of nicotinamide adenine dinucleotide is a key electron transfer coenzyme in microbial respiratory metabolism, and its fluorescence or Raman signal intensity can be used as an indicator of microbial metabolic activity, especially the activity of functional flora such as nitrifying bacteria. The NADH activity index in the present application is a dimensionless activity representation quantity based on this signal.

[0029] The operation data includes: The current value of online dissolved oxygen (DO) and its historical trend. DO is a key indicator for aeration control; The influent flow (Q) and its load fluctuation. The change of influent flow directly affects the treatment load; The historical aeration intensity and adjustment trajectory. Aeration intensity is an important operating parameter that affects DO and microbial activity; The valve opening.

[0030] The carbon source dosage.

[0031] These data are gathered together to form a high-dimensional candidate state set. For example, in a typical wastewater treatment plant, there may be more than 50 original spectral features, more than 10 DOM sub-component ratios, 3 NADH activity indices, and about 10 operating parameters, totaling more than 70 candidate quantities.

[0032] Step 220: Perform multi-criteria evaluation. For each candidate quantity in the candidate state set constructed in step 210, quantitative or semi-quantitative evaluation is performed from the three dimensions of prediction contribution, signal stability, and control sensitivity, forming a multi-criteria evaluation result. Specifically, it includes: 221. Prediction contribution evaluation: This evaluation aims to judge the prediction ability or explanation ability of a certain candidate quantity for the current control target, for example, maintaining stable DO in the aerobic section, ensuring that the TN removal rate does not decrease, and avoiding short-term carbon deficiency in the anoxic section. The specific method can use Ridge Regression analysis to evaluate the influence of introducing or excluding a certain candidate quantity on the prediction error of short-term DO, TN, etc. Assume we have a prediction model M, whose input is the state vector and the output is the control target Y. For each candidate quantity , we can calculate its marginal contribution degree .

[0033] One way to calculate is to compare the prediction error changes when including and not including . For example, using mean squared error (MSE) as the prediction error indicator: ; ; then . If is larger, it means has a greater contribution to prediction.

[0034] In the formula, is the total number of samples; j is the index of the jth sample; Real observation value of the jth sample, such as pollutant concentration, dissolved oxygen value, carbon source demand, etc. Complete input feature vector of the jth sample, containing all features from multispectral data, process operation data, microbial activity indicators, etc. The ith feature in the feature vector, such as the intensity of a certain waveband, the ratio of certain spectra, or the operating value of a certain sensor. The feature vector of the jth sample, but the ith feature xi is removed or masked. Masking methods usually include: setting to mean; setting to zero; replacing with noise, etc. M(·) is the prediction model function, which can be a linear model, a tree model, or a neural network, which will not be described here. The mean square error of prediction when using the complete feature set. The mean square error of prediction after removing the ith feature. Feature contribution score, greater than 0 indicates that the feature contributes.

[0035] In addition, the prediction contribution can also be measured by calculating the mutual information (Mutual Information) between the feature and the target.

[0036] ; Where, Mutual information between random variables X and Y, is their joint probability distribution function, and are their marginal probability distribution functions. The greater the mutual information, the greater the contribution of the candidate quantity to the target prediction; Input feature variable, which can be a certain feature, such as Ratio, or a set of discrete feature values; Discretized interval of target variable, such as DO, NH4-N, TN, carbon source dosage, etc. A value of random variable X, such as a certain absorbance interval; A value of random variable Y, such as a certain DO interval.

[0037] 222. Signal stability evaluation: This evaluation aims to measure the measurement noise level, frequency of abnormal points and drift trend of the candidate quantity, in order to filter out quantities with too much noise or frequent abnormalities. Specific methods include: Variance or standard deviation: Calculate the variance or standard deviation of the candidate quantity in the recent period, such as the past 24 hours or 100 sampling points. The greater the variance, the greater the signal fluctuation and the worse the stability.

[0038] Outlier Ratio: Identify outliers by statistical methods, such as 3σ criterion or boxplot method, and calculate the proportion of outliers in the statistical period. The higher the proportion, the lower the signal reliability.

[0039] Trend: Analyze whether there is a significant trend in the signal by linear regression or moving average method. For example, for a candidate variable , its stability score can be calculated as follows: where is a decreasing function, for example , where is the weight.

[0040] In the formula, is the standard deviation of the feature , used to describe the volatility; is the mean of the feature ; is the coefficient of variation; is the number of samples judged as outliers; is the total number of samples; is the outlier ratio; is the comprehensive stability evaluation function.

[0041] Control Sensitivity Evaluation: This evaluation aims to analyze the sensitivity between the candidate variable changes and the control object changes, such as DO, TN, to exclude variables that are difficult to effectively influence or have unstable influence direction through control means. Specific methods can review the relationship between historical control actions, such as aeration intensity adjustment, carbon source dosage changes, and process responses, such as DO, TN changes. For example, if a certain spectral feature is highly correlated with the change trend of DO, and this correlation shows a clear causal relationship after aeration intensity adjustment, the control sensitivity of this feature is high. Granger Causality Test or Cross-correlation Analysis can be used to quantify this sensitivity. For candidate variables and control objects , calculate their cross-correlation coefficients at different lag times, which describe the correlation between the feature changes lagging or leading the target variable.

[0042] ; where, ​is the lag time. Higher cross-correlation coefficient indicates stronger sensitivity. is the ith feature of time series type, such as the intensity of a certain spectral channel; is the time series of target variable, such as D0, effluent NH4-N, etc. and is the mean of each time series; and is the standard deviation of each time series; is the mathematical expectation, calculated by sliding window average.

[0043] It should be noted that in this specification, all t represents discrete sampling time t = n·Δt, n is a non-negative integer.

[0044] Step 230: hierarchical screening and sorting. According to the multi-criteria evaluation results formed in step 220, a hierarchical screening and sorting method is used to select a limited number of key state quantities from the candidate state set to form a basic state subset. This hierarchical screening avoids the information loss that may be caused by simple weighted scoring. The specific screening process is as follows: a. Prediction contribution screening: first, eliminate candidate quantities whose prediction contribution to the target is lower than the preset threshold, for example, prediction contribution These quantities have little substantial effect on the current control target and can be safely discarded to reduce redundancy.

[0045] b. Signal stability screening: among the remaining candidate quantities, further eliminate candidate quantities whose signal stability score is lower than the preset threshold, for example, stability score , or coefficient of variation , or the proportion of abnormal points These quantities are susceptible to noise interference or frequent abnormality, and including them in the state vector will introduce uncertainty and affect the stability of the control strategy.

[0046] c. Control sensitivity sorting and selection: after the first two rounds of screening, the remaining candidate quantities contribute to the control target and have stable signals. At this time, according to their control sensitivity, the top N candidate quantities are selected to form the basic state subset, for example, N = 8-12. Specifically, the value of N can be adjusted according to the complexity of the actual control system and the computing resources.

[0047] Step 240: constructing derived state quantities. After obtaining the basic state subset, the present application further introduces a derived state quantity construction mechanism. These derived state quantities are characteristic quantities that can reflect the dynamic characteristics of the process and make up for the shortcomings of static measurement values. Specific derived state quantities include: DO recovery time or DO response strength caused by unit aeration adjustment: constructed according to the decay trajectory of DO deviation over time. For example, the time required for DO to recover from the current value to the target interval after the aeration intensity changes, or the DO change rate caused by unit aeration intensity adjustment. This can depict the dynamic response capability of the current aeration system. For example, define the DO recovery time as: ; wherein, is the aeration adjustment time, is the target DO value, is the allowed deviation. is the sampling time step, i.e. the system time resolution, such as 1 min / 5 min; is the dissolved oxygen monitoring data.

[0048] DOM structure mutation amplitude: constructed according to the short-term changes of DOM fluorescence peak intensity and ratio.

[0049] It should be noted that, represents the absorbance of the ultraviolet-visible spectrum at a wavelength of 254 nm, represents the absorbance of the ultraviolet-visible spectrum at a wavelength of 365 nm, The ratio is used to characterize the average molecular weight and aromaticity of dissolved organic matter, reflecting the degree of humification.

[0050] For example, the maximum change rate of DOM fluorescence peak intensity or ratio in the past 1 hour. This can be used to represent the degree of change in organic matter properties and biodegradability, indicating possible carbon source deficiency or microbial activity changes.

[0051] For example, define the DOM structure mutation amplitude as:

[0052] ; wherein, is the time window, the length of the time window for detecting DOM structure mutation, such as 1 hour or one day. is the current analysis time.

[0053] Activity inhibition duration: constructed according to the time length that the NADH activity index is continuously below the baseline value. For example, if the NADH activity index is continuously below the average value of normal operation for N sampling periods, record its duration. This can be used to reflect the degree of inhibition of functional flora such as nitrifying bacteria, and to warn potential nitrification efficiency decline.

[0054] For example, define the activity inhibition duration is: ; wherein, is an indicator function, 1 if the condition is satisfied, otherwise 0, is the baseline NADH activity value, L is the number of historical sampling points; is also the length of time during which the microbial activity is continuously lower than the baseline value.

[0055] Step 250: generating a target state vector. The derived state quantities constructed in step 240 are combined with the filtered base state subset in step 230 to generate a target state vector that is appropriate for the current control target, moderate in dimension, and rich in information. The target state vector is subsequently used as the unified input of the rule layer (132) and the model layer (133) in the operation and decision module (130) to realize the subsequent coordinated optimization control of the aeration blower frequency, the aeration valve opening degree, and the carbon source dosage.

[0056] Through the above method, the present application no longer simply feeds all the multispectral analysis results to the model or only uses a few empirical variables, but instead uses the internal relationship between the multispectral analysis data and the operation data to complete the selective and structured comprehensive analysis and construction of the state quantities around a clear small target in each control cycle: not only is high-dimensional redundancy and noise amplification avoided, but also key information reflecting the dynamic characteristics and functional state of the microbial community is fully introduced. The target state vector constructed in this way is matched with the subsequent interpretable decision model, so that the water treatment optimization system and the regulation and control method based on the combination of multispectral technology truly realize the integrated technical path from seeing deeper and finer to adjusting more accurate and more economical.

[0057] In Example Two, a water treatment optimization system based on the combination of multispectral technology is provided for implementing the above method. Referring to Figure 1 , the system includes a multispectral analysis module (110), a process regulation and control parameter module (120), and an operation and decision module (130); the operation and decision module (130) is internally provided with a state vector filtering and derivation construction unit (131), a rule layer (132), and a model layer (133). In engineering implementation, the multispectral analysis module (110) can collect and input the original spectral data through external spectral sensors; the process regulation and control parameter module (120) can access the online instruments or control systems on site to obtain operation data. The operation and decision module (130) can receive the multispectral analysis results and operation data through a communication interface, and issue control instructions to the on-site actuators through a control instruction output interface; historical data can be saved by an external storage device and called by the operation and decision module (130). The above external devices / interfaces are engineering configurations and do not constitute Figure 1 the system components shown in the figure.

[0058] Specifically, the system is composed as follows: The multi-spectral analysis module (110) is responsible for receiving raw spectral data collected by external spectral sensors, such as NIR, UV-Vis, three-dimensional fluorescence, and Raman spectrometers, and processing and analyzing the raw spectral data through built-in chemometric models or machine learning models to output conventional water quality indicators, DOM component information, such as the fluorescence intensity and proportion of humic acid and protein substances, and NADH activity index, etc.

[0059] The conventional water quality indicators include chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), and nitrate nitrogen, which are used to represent the load level of organic pollutants and nitrogen-containing pollutants in wastewater.

[0060] The process control parameter module (120) is responsible for collecting various operation data in the wastewater treatment process, including but not limited to online DO sensor data, influent flow meter data, aeration blower frequency, aeration valve opening, carbon source dosage, pH value, temperature, etc. This module transmits these data to the operation decision module in real time.

[0061] The operation decision module (130) is the core control unit of the system, which is responsible for receiving all input data, constructing state vectors, generating control strategies, and outputting instructions. The module internally adds a state vector screening and derivative construction unit (131).

[0062] The state vector screening and derivative construction unit (131) is the innovation point of the invention, which receives multi-spectral analysis results output by the multi-spectral analysis module (110) and operation data provided by the process control parameter module (120). At each control time step, the unit executes the steps of constructing a candidate state set, performing multi-criteria evaluation, hierarchical screening and sorting, and constructing derivative state quantities according to the current control task, as described in Embodiment I, to finally generate a refined and information-rich target state vector.

[0063] The rule layer (132) receives the target state vector output by the state vector screening and derivative construction unit (131) and makes preliminary judgments or corrections on the control actions according to the pre-set expert experience rules or fuzzy control rules. For example, when the DO is below the threshold and the DO recovery time is too long, the rule layer may suggest increasing the aeration intensity.

[0064] Model layer (133): receives the target state vector filtered by the state vector filtering and derivation unit (131) and generates the optimal control instruction using the control model based on machine learning or optimization algorithm, such as reinforcement learning, predictive control model. For example, according to the target state vector, the DO and TN changes in the future period of time are predicted, and the optimal aeration blower frequency and carbon source dosage are calculated. The historical operation data, spectral data, control instruction and state vector can be saved by an external storage device for calling by the operation decision module (130).

[0065] In an optional example, the system workflow is as follows: The external spectral sensor continuously collects water quality spectral information and inputs it to the multispectral analysis module (110) for analysis to generate multispectral analysis results.

[0066] The process control parameter module (120) collects real-time operation data such as online DO and influent flow.

[0067] The multispectral analysis results and operation data are simultaneously transmitted to the state vector filtering and derivation unit (131) in the operation decision module (130).

[0068] The state vector filtering and derivation unit (131) performs the following operations according to the current control target: a. Construct a candidate state set containing multispectral analysis indicators and operation data.

[0069] b. Perform three-dimensional evaluation of prediction contribution, signal stability and control sensitivity for each candidate in the set.

[0070] c. Adopt a hierarchical screening and sorting strategy to select key state quantities from the candidate set to form a basic state subset.

[0071] d. According to the historical time series and spectral characteristics of the basic state quantities, construct derived state quantities reflecting the dynamic characteristics of the process.

[0072] e. Combine the derived state quantities with the basic state subset to generate the target state vector.

[0073] The generated target state vector is the unified input of the rule layer (132) and the model layer (133) in the operation decision module (130).

[0074] The rule layer (132) and the model layer (133) work together to calculate the optimal control instructions such as aeration blower frequency, aeration valve opening and carbon source dosage according to the target state vector and the preset control logic.

[0075] The operation decision module (130) sends the control instructions to the field execution mechanism through a control instruction output interface, so as to realize real-time optimization control of the water treatment process.

[0076] Through the system, the application realizes effective landing of the multispectral technology in the optimization control of the water treatment. The state vector screening and derivation construction unit (131) ensures that the state vector input to the control model is neither excessively redundant nor contains key dynamic process information, so that the intelligent level and operation efficiency of the control system are significantly improved.

[0077] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the application, but not to limit them; although the application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A water treatment optimization control method based on the combination of multispectral technology, characterized in that, The method comprises the following steps: obtaining multi-spectral analysis results output by a multi-spectral analysis module and operation data provided by a process control parameter module to form a candidate state set containing the multi-spectral analysis results and the operation data; evaluating each candidate state in the candidate state set from three dimensions of prediction contribution, signal stability and control sensitivity to form multi-criteria evaluation results; selecting a limited number of key state variables from the candidate state set to form a basic state subset according to the multi-criteria evaluation results by using a hierarchical screening and sorting method; constructing derived state variables reflecting dynamic characteristics of the process according to historical time series and variation trends of spectral characteristics of state variables in the basic state subset; combining the derived state variables with the basic state subset to generate a target state vector for a current control target, which is moderate in dimension and rich in information, and taking the target state vector as an input of an operation decision module.

2. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 1, characterized in that: The multi-spectral analysis results include conventional water quality indexes and spectral characteristic variables obtained or predicted by near-infrared, ultraviolet-visible light, dissolved organic matter component information obtained by three-dimensional fluorescence analysis, and NADH activity index calculated by Raman spectrum.

3. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 1, characterized in that: The operation data include online dissolved oxygen, influent flow, historical aeration intensity, valve opening and carbon source dosage.

4. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 1, characterized in that: The evaluation step of the prediction contribution includes analyzing the influence of introducing or removing a certain candidate variable on the prediction error of short-term dissolved oxygen or total nitrogen to determine the contribution of the candidate variable to the current control target.

5. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 1, characterized in that: The evaluation step of the signal stability includes calculating the variance, proportion of abnormal points and drift trend of the candidate variable in a recent period of time to evaluate the measurement noise and stability of the candidate variable.

6. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 1, characterized in that: The evaluation step of the control sensitivity includes reviewing the relationship between historical control actions and process responses and analyzing the sensitivity correlation between the change of the candidate variable and the change of the control object.

7. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 6, characterized in that: The hierarchical screening and sorting method includes: firstly, removing candidate variables having little effect on the target according to the prediction contribution; secondly, removing candidate variables susceptible to noise interference according to the signal stability; finally, selecting a limited number of key state variables to form the basic state subset according to the control sensitivity in the remaining set.

8. The water treatment optimization and regulation method based on the combination of multispectral technology according to claim 2, characterized in that: The derived state variables include: dissolved oxygen recovery time or dissolved oxygen response intensity caused by unit aeration adjustment according to the decay trajectory of dissolved oxygen deviation over time; dissolved organic matter structure mutation amplitude constructed according to short-term changes of dissolved organic matter fluorescence peak intensity and E2 / E3 ratio; wherein, E2 represents the absorbance of ultraviolet-visible spectrum at 254 nm wavelength, and E3 represents the absorbance of ultraviolet-visible spectrum at 365 nm wavelength; active inhibition duration constructed according to the length of time during which the NADH activity index is continuously lower than the reference value.

9. A water treatment optimization control system based on the combination of multispectral technologies, for implementing the method according to any one of claims 1 to 8, characterized in that, The method comprises: a multi-spectral analysis module for outputting multi-spectral analysis results; a process control parameter module for providing operation data; An operation decision module, which is internally added with a state vector screening and derivation construction unit, is connected with the multi-spectrum analysis module and the process control parameter module, used to receive the multi-spectrum analysis result and the operation data, and execute the method of any one of claims 1 to 8.

10. The water treatment optimization and regulation system based on the combination of multispectral technology according to claim 9, characterized in that: The operation decision module further comprises a rule layer and a model layer, and the target state vector is used as the unified input of the rule layer and the model layer, so as to realize the synergistic optimization control of the aeration fan frequency, the aeration valve opening degree and the carbon source dosage.

Images