Chemical sewage treatment process abnormal working condition biochemical potential assessment method

Abstract

The invention discloses a chemical sewage treatment process abnormal working condition biochemical potential assessment method. After a sample data set is constructed based on water quality parameters of a chemical sewage treatment plant, a sewage treatment process ATP and BOD prediction model is established based on a multiple-input multiple-output fuzzy neural network, and chemical sewage water quality parameter samples in the sample data set are utilized to train and test the sewage treatment process ATP and BOD prediction model. The ATP concentration value and the BOD concentration value are predicted in real time by using the tested ATP and BOD prediction model in the sewage treatment process, an abnormal working condition discrimination matrix based on ATP and BOD real-time prediction is established, the abnormal working condition type is accurately identified, and the sludge condition and the biochemical treatment potential of a chemical sewage treatment plant are evaluated by combining the ATP concentration value and the BOD concentration value. And adjusting the load operation scheme of the chemical sewage treatment plant. Accurate recognition of abnormal working conditions such as sludge poisoning and carbon source shortage is achieved, the abnormal working conditions of sludge activity in the chemical sewage treatment process can be mastered in time, and the operation efficiency of the sewage treatment process is improved.

Description

Technical Field

[0001] The invention relates to the technical field of sewage treatment, and in particular to a method for evaluating the biochemical potential of abnormal working conditions in a chemical sewage treatment process. Background Art

[0002] The pollutant components of chemical wastewater are relatively stable, and there are treatment bottlenecks such as high organic concentration, poor biodegradability, and biological toxicity. Since the microorganisms in the biochemical unit of sewage treatment are extremely vulnerable to shocks, if they are not controlled in time, it is easy to cause abnormal operating conditions in the sewage treatment plant, and even excessive discharge of wastewater. Common abnormal operating conditions in the chemical sewage treatment process include sludge bulking, sludge poisoning, etc. Failure to discover and deal with abnormal conditions in a timely manner and poor safety control of the sewage treatment process will directly affect the effluent quality of chemical sewage treatment, endanger the operation of the entire sewage treatment system, and in severe cases even cause the entire chemical sewage treatment process to collapse.

[0003] The influent flow, composition, types of pollutants and organic matter concentration in most chemical wastewater treatment processes are complex, which can easily lead to large fluctuations in the microbial content in the mud-water mixture in the biochemical reaction zone. At the same time, there are many types of activated sludge microorganisms, and there are differences between the microorganisms used under different process operating conditions. The impact of the operating environment on the microorganisms is large and uncertain.

[0004] In order to timely grasp the abnormal working conditions of chemical wastewater treatment process, technicians in this field have conducted relevant research. Chinese invention patent CN 112363391 B discloses a sludge bulking suppression method based on adaptive segmented sliding mode control. The invention uses fuzzy neural network to obtain water quality parameters of the sewage treatment operation process, predict the sludge volume index in real time to judge the working condition of the operation process, provide reference signals for the switching of the controller, and design a segmented sliding mode controller with adaptive switching mechanism to regulate the dissolved oxygen concentration and nitrate nitrogen concentration, coordinate the biochemical reaction process, improve the sludge sedimentation characteristics, and ensure that the sewage treatment process returns to normal working conditions when sludge bulking occurs. Chinese invention patent CN 115578015 B discloses a method, system and storage medium for monitoring the whole process of sewage treatment based on the Internet of Things, which specifically includes: obtaining the current sewage treatment process information, setting the key water quality monitoring indicators of the water quality monitoring points; obtaining the water quality monitoring results of each point through the key water quality monitoring indicators, building a sewage treatment whole process model, and using the sewage treatment whole process model for real-time monitoring and control according to the water quality monitoring results; judging and generating fault warning information of the sewage treatment process, adding a fault detection model through deep learning for real-time learning, tracing the fault of the sewage treatment process according to the water quality monitoring results and multivariate working condition data, and visualizing it through the sewage treatment whole process model. The invention uses the Internet of Things technology to remotely monitor the whole process of sewage treatment, realizes the dynamic monitoring of water quality changes and the identification of abnormal working conditions in the sewage treatment process, and significantly improves the efficiency and accuracy of water quality monitoring.

[0005] However, the research on the abnormal condition identification method of the chemical wastewater treatment process in the existing technology mainly focuses on deep learning and other aspects. The key variables describing the activated sludge in the biochemical unit in the chemical wastewater treatment process have not been fully analyzed. It is impossible to effectively identify the abnormal conditions of the chemical wastewater treatment process, accurately evaluate the treatment potential of the biochemical unit, and ensure that the wastewater treatment facilities reach a stable operation state.

[0006] The evaluation of the biological activity treatment capacity of sludge in the biochemical unit of the chemical wastewater treatment process is mainly based on judging the degradation capacity of activated sludge on pollutants. Among them, the activity of sludge and the effect of sewage treatment are the two main evaluation bases. On the one hand, adenosine triphosphate (ATP) is not only the most direct energy source in the organism, but also one of the important means to reflect the life activities of microorganisms, and can be used to master the activity of sludge. When the influent water quality changes or the concentration of toxic substances changes sharply, the microorganisms in the sludge will be dormant and their biological activity will also decrease. However, ATP is a sign of the life and vitality of microbial cells. Using ATP can indirectly reflect the activity of microorganisms and then obtain the metabolic status of microorganisms. Therefore, ATP is also used as one of the important parameters reflecting the activity of microorganisms in the sewage treatment process. On the other hand, biochemical oxygen demand (BOD) is an important water quality indicator in the sewage treatment process. The level of BOD can reflect the degree of degradation of organic matter in the water body by microbial biochemical action. Real-time monitoring of BOD is an important means for sewage treatment. Therefore, how to achieve real-time prediction of ATP and BOD is the key to mastering the biochemical treatment capacity of sludge in the biochemical unit.

[0007] Therefore, it is urgent to propose a biochemical potential assessment method for abnormal conditions in chemical wastewater treatment processes based on real-time prediction of ATP and BOD, so as to accurately identify abnormal conditions in chemical wastewater treatment processes and ensure that the effluent quality meets the standards under abnormal conditions in chemical wastewater treatment processes. Summary of the invention

[0008] In view of the problem that the activity of sludge in biochemical units cannot be directly and accurately judged at the current stage, the present invention proposes a method for evaluating the biochemical potential of abnormal conditions in chemical wastewater treatment processes. By combining chemical wastewater management and control data with artificial intelligence algorithms, an abnormal condition discrimination matrix for evaluating abnormalities in petrochemical wastewater treatment processes is established, thereby achieving an accurate evaluation of the biochemical treatment potential of biochemical units in petrochemical wastewater treatment processes and effectively ensuring the effluent quality of petrochemical wastewater treatment processes.

[0009] The present invention specifically adopts the following technical solutions:

[0010] A method for evaluating the biochemical potential of abnormal working conditions in a chemical wastewater treatment process, which evaluates the biochemical treatment potential of abnormal working conditions of chemical wastewater by real-time prediction of adenosine triphosphate ATP and biochemical oxygen demand BOD, specifically comprises the following steps:

[0011] Step 1, obtaining water quality parameters of a chemical wastewater treatment plant, performing pretreatment, normalization and principal component analysis on the water quality parameters to obtain multiple chemical wastewater water quality parameter samples, and randomly assigning each chemical wastewater water quality parameter sample to a training sample set and a test sample set, which together constitute a sample data set for training an ATP and BOD prediction model for a wastewater treatment process;

[0012] Step 2, establishing a prediction model of ATP and BOD in the sewage treatment process based on a multi-input multi-output fuzzy neural network;

[0013] Step 3, using the training set to train the ATP and BOD prediction model of the sewage treatment process to obtain the trained ATP and BOD prediction model of the sewage treatment process;

[0014] Step 4, using the test set to verify the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training. If the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training has reached the best, then proceed to step 5; otherwise, return to step 3 and continue to train the ATP and BOD prediction model of the sewage treatment process using the training set;

[0015] Step 5, using the tested ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate (ATP) concentration and the biochemical oxygen demand (BOD) concentration in real time according to the water quality parameters of the chemical sewage, generate an abnormal operating condition discrimination matrix, use the abnormal operating condition discrimination matrix to discriminate the type of abnormal operating condition, and evaluate the sludge condition and biochemical treatment potential of the chemical sewage treatment plant according to the adenosine triphosphate (ATP) concentration value and the biochemical oxygen demand (BOD) concentration value, and adjust the load operation plan of the chemical sewage treatment plant.

[0016] Preferably, the step 1 specifically includes the following steps:

[0017] Step 1.1, obtaining chemical wastewater from a chemical wastewater treatment plant, and measuring the influent volume, water quality parameters, adenosine triphosphate (ATP), and biochemical oxygen demand (BOD) of the chemical wastewater;

[0018] Step 1.2, pre-processing each water quality parameter of the chemical wastewater, eliminating abnormal data in each water quality parameter, and then normalizing each pre-processed water quality parameter, and using the principal component analysis method to analyze the obtained water quality parameters, based on the correlation coefficient and contribution rate of each water quality parameter for the prediction of ATP and BOD in the sewage treatment process, screening is performed, and the screened water quality parameters and water inflow are used as input variables, and the output variables are determined in combination with the measured adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value of the chemical wastewater, and the input variables and output variables are used to form a chemical wastewater water quality parameter sample;

[0019] In step 1.3, each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set to generate a training sample set and a test sample set, and a sample data set is constructed for training the ATP and BOD prediction model of the wastewater treatment process.

[0020] Preferably, when the chemical wastewater is propylene oxide device wastewater and ethylene alkali slag wastewater, the water quality parameters include inlet chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, inlet flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, inlet ammonia nitrogen concentration, effluent ammonia nitrogen concentration, biochemical pool sulfide concentration, biochemical pool volatile phenol concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and redox potential ORP;

[0021] When the chemical wastewater is purified terephthalic acid wastewater and ethylene alkali residue wastewater, the water quality parameters include influent chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, formaldehyde concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and oxidation-reduction potential ORP;

[0022] When the chemical wastewater is propylene oxide wastewater, purified terephthalic acid wastewater and ethylene alkali residue wastewater, the water quality parameters include influent chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and redox potential ORP;

[0023] When the chemical wastewater is propylene oxide wastewater, aromatic hydrocarbon wastewater, purified terephthalic acid wastewater and ethylene alkali slag wastewater, the water quality parameters include inlet chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, cyanide concentration, polycyclic aromatic hydrocarbons, volatile phenol concentration, sulfide concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, inlet flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, inlet ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and redox potential ORP.

[0024] Preferably, when the chemical wastewater is propylene oxide device wastewater and ethylene alkali residue wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, propylene oxide concentration, formaldehyde concentration and influent flow rate, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value;

[0025] When the chemical wastewater is purified terephthalic acid wastewater and ethylene alkali residue wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, terephthalic acid concentration, influent flow rate and isophthalic acid concentration, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value;

[0026] When the chemical wastewater is propylene oxide wastewater, purified terephthalic acid wastewater and ethylene alkali residue wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, terephthalic acid concentration and influent flow rate, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value;

[0027] When the chemical wastewater is propylene oxide wastewater, aromatic hydrocarbon wastewater, purified terephthalic acid wastewater and ethylene alkali slag wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, polycyclic aromatic hydrocarbons, volatile phenol concentration, sulfide concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, isopropylbenzene concentration, propylene oxide concentration, terephthalic acid concentration and influent flow rate, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value.

[0028] Preferably, in step 2, the multi-input multi-output fuzzy neural network includes an input layer, an RBF layer, a rule layer and an output layer;

[0029] The input layer of the multi-input multi-output fuzzy neural network is set as:

[0030] x i =u i (1)

[0031] x=[x1,x2,...,x k ] (2)

[0032] Where i is the serial number of the input variable, i = 1, 2, ..., k, k is the total number of input variables; x i is the value of the i-th input variable, u i is the input value of the i-th input neuron, and x is the input vector;

[0033] The RBF layer is provided with a plurality of neurons, and the output function value of the neurons in the RBF layer is:

[0034]

[0035] In the formula, φ j is the output value of the jth neuron in the RBF layer, j is the number of the neuron, j = 1, 2, ..., Q, Q is the total number of neurons in the RBF layer; c ij is the center of the jth neuron, σ ij is the width of the jth neuron;

[0036] A plurality of neurons are arranged in the rule layer, and the output function value of the neurons in the rule layer is:

[0037]

[0038] Where l is the inner layer number of the rule layer, l = 1, 2, ..., Q; v l is the output value of the lth layer in the regular layer, φ l is the output value of the lth layer in the RBF layer;

[0039] The output value of the output layer is:

[0040] W=[w 1 ,w 2 ,…,w M ] T (5)

[0041] y=Wv (6)

[0042] Where W is the weight matrix, w Mis the output value of the Mth neuron in the output layer, M is the total number of neurons in the output layer, and T is the transposed matrix; y is the abnormal working condition discrimination matrix, y=[y1,y2], y1 is the predicted value of adenosine triphosphate ATP concentration, y2 is the predicted value of biochemical oxygen demand BOD concentration; v is the output matrix of the rule layer, v=[v1,v2,···,v Q ] T , v Q is the output value of the Qth layer in the regular layer.

[0043] Preferably, the number of neurons arranged in the regular layer is equal to that in the RBF layer.

[0044] Preferably, in step 2, a multi-input multi-output fuzzy neural network is trained to construct an ATP and BOD prediction model for a sewage treatment process using a Levenberg-Marquardt algorithm based on an adaptive learning rate;

[0045] The update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate is set as:

[0046] Θ(t+1)=Θ(t)+(Ψ(t)+λ(t)×I) -1 Ω(t) (7)

[0047] Where t is time, Θ(t+1) is the variable vector at time t+1, Θ(t) is the variable vector at time t, Ψ(t) is the quasi-Hessian matrix at time t, λ(t) is the adaptive learning rate at time t, 0<λ(t)<1, I is the identity matrix, and Ω(t) is the gradient vector at time t;

[0048] In the update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate, the calculation formula of the adaptive learning rate λ(t) is:

[0049] λ(t)=μ(t)λ(t-1) (8)

[0050]

[0051] Where μ(t) is the adaptive factor at time t, λ(t-1) is the adaptive learning rate at time t-1; τ min (t) is the minimum eigenvalue of the quasi-Hessian matrix at time t, τ max (t) is the maximum eigenvalue of the quasi-Hessian matrix at time t, 0<τ min (t)<τ max (t);

[0052] The variable vector Θ(t) includes a connection weight vector w(t), a center vector c(t) and a width vector σ(t). The calculation formula of the variable vector is:

[0053] Θ(t)=[w1(t),...,w J (t),c1(t),...,c J (t),...,σ1(t),...,σ J (t)] (10)

[0055] In the formula, w J (t) is the value of the J-th connection weight vector at time t, c J (t) is the J-th center vector value at time t, σ J (t) is the J-th width vector value at time t;

[0056] The calculation formula of the quasi-Hessian matrix Ψ(t) and the gradient vector Ω(t) is:

[0057] Ψ(t)=j T (t)j(t) (10)

[0058] Ω(t)=j T (t)e(t) (11)

[0059] e(t)=y d (t)-y(t) (12)

[0060] in,

[0061] The Jacobian vector j(t) is:

[0062]

[0063] Where j(t) is the Jacobian vector at time t, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, and y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0064] Preferably, the step 3 specifically includes the following steps:

[0065] Step 3.1, set the training accuracy value;

[0066] Step 3.2, inputting the water quality parameter samples of the chemical wastewater in the training set into the ATP and BOD prediction model of the wastewater treatment process constructed in step 2, using the ATP and BOD prediction model of the wastewater treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtaining the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater;

[0067] Step 3.3, comparing the predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value in the chemical sewage water quality parameter sample, and calculating the accuracy value of the ATP and BOD prediction model of the sewage treatment process. If the accuracy value of the ATP and BOD prediction model of the sewage treatment process is less than the preset accuracy value, proceed to step 3.4; otherwise, proceed to step 3.5;

[0068] Step 3.4, based on the Levenberg-Marquardt algorithm, the adaptive learning rate of the ATP and BOD prediction model of the sewage treatment process is updated to obtain the updated ATP and BOD prediction model of the sewage treatment process, and the chemical sewage water quality parameter samples are continuously randomly selected in the training set and input into the updated ATP and BOD prediction model of the sewage treatment process, and the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical sewage water quality parameter samples are predicted by using the updated ATP and BOD prediction model of the sewage treatment process, and after obtaining the predicted values ​​of the adenosine triphosphate ATP concentration and the biochemical oxygen demand BOD concentration of the chemical sewage water quality parameter samples, return to step 3.3;

[0069] Step 3.5, complete the training of the ATP and BOD prediction model for the sewage treatment process, and obtain the trained ATP and BOD prediction model for the sewage treatment process.

[0070] Preferably, the step 4 specifically includes the following steps:

[0071] Step 4.1, inputting the water quality parameter samples of the chemical wastewater in the test set into the trained ATP and BOD prediction model of the wastewater treatment process, using the ATP and BOD prediction model of the wastewater treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtaining the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater;

[0072] Step 4.2, compare the predicted adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical sewage water quality parameter sample, and calculate the accuracy of the ATP and BOD prediction model of the sewage treatment process. If the accuracy of the ATP and BOD prediction model of the sewage treatment process is not less than the preset accuracy, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process.

[0073] Preferably, the error values ​​of the ATP and BOD prediction models for the sewage treatment process are:

[0074]

[0075] The accuracy of the ATP and BOD prediction model for sewage treatment process is:

[0076]

[0077] in,

[0078] e(t)=y d (t)-y(t) (16)

[0079] Where RMSE(t) is the error value of the ATP and BOD prediction model of the sewage treatment process, t is time, t = 1, 2, ..., N, N is the total time, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0080] Preferably, in step 5, the abnormal operating condition discrimination matrix is:

[0081]

[0082] In the formula, ATP is the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process, BOD is the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process; C1 is the first type of abnormal operating condition, C2 is the second type of abnormal operating condition, C3 is the third type of abnormal operating condition, and C4 is the fourth type of abnormal operating condition.

[0083] Preferably, in step 5, based on the abnormal operating condition discrimination matrix, when the abnormal operating condition type is discriminated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value:

[0084] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are high, the current operation condition of the chemical wastewater treatment plant is determined to be the first abnormal condition type C1, the influent concentration of the chemical wastewater treatment plant exceeds the treatment capacity, the BOD concentration is abnormal, and the biochemical treatment capacity is high;

[0085] When adenosine triphosphate ATP is expressed as high and biochemical oxygen demand BOD is expressed as low, it is determined that the current operating condition of the chemical wastewater treatment plant is the second abnormal condition type C2, and the biochemical treatment capacity of the chemical wastewater treatment plant is surplus;

[0086] When the adenosine triphosphate (ATP) is low and the biochemical oxygen demand (BOD) is high, the current operation condition of the chemical wastewater treatment plant is determined to be the third abnormal condition type C3, and the sludge in the current chemical wastewater treatment plant is determined to be impacted by water flow. It is necessary to measure the adenosine triphosphate (ATP) in the sludge, and analyze the abnormal condition in combination with the adenosine triphosphate (ATP) concentration value of the sludge;

[0087] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are expressed as low, it is determined that the current operating condition of the chemical wastewater treatment plant is the fourth abnormal condition type C4, and the biochemical treatment capacity of the chemical wastewater treatment plant is low.

[0088] Preferably, when the operating condition of the chemical sewage treatment plant is determined to be the fourth abnormal condition type C4, the sludge disposal capacity of the chemical sewage treatment plant is improved by reducing the sewage treatment volume of the chemical sewage treatment plant, increasing the aeration volume or supplementing the carbon source.

[0089] The present invention has the following beneficial effects:

[0090] This patent provides a method for evaluating the biochemical potential of abnormal conditions in the chemical wastewater treatment process. By real-time predicting the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical wastewater treatment process, an abnormal condition discrimination matrix for evaluating abnormalities in the petrochemical wastewater treatment process is established, and real-time judgment is made on whether abnormal conditions occur in the activated sludge of the biochemical unit in the chemical wastewater treatment process.

[0091] The present invention solves the problem that the prior art cannot directly and accurately judge the activity of biochemical unit sludge in the sewage treatment process. By applying a multi-input and multi-output fuzzy neural network to the prediction process of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, an abnormal condition discrimination matrix is ​​established using the real-time predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, thereby achieving accurate identification of abnormal conditions such as sludge poisoning and carbon source deficiency. By automatically pushing process adjustment plans and emergency remedial measures, it is effectively guaranteed that the effluent water quality of chemical sewage treatment under abnormal conditions is stable and meets the standards.

[0092] The present invention is helpful for timely grasping abnormal working conditions of sludge activity in the process of chemical sewage treatment, effectively avoiding the occurrence of working conditions such as sludge poisoning and sludge swelling, and improving the operation efficiency of the sewage treatment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is a schematic diagram of the structure of the multi-input multi-output fuzzy neural network of the present invention.

[0094] Figure 2 The present invention is a flow chart of using an abnormal operating condition discrimination matrix to identify abnormal operating conditions. DETAILED DESCRIPTION

[0095] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0096] Example 1

[0097] This embodiment adopts a method for evaluating the biochemical potential of abnormal working conditions in a chemical wastewater treatment process proposed by the present invention, and evaluates the biochemical treatment potential of abnormal working conditions of chemical wastewater by real-time prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, which specifically includes the following steps:

[0098] Step 1, obtaining water quality parameters of a chemical wastewater treatment plant. The chemical wastewater obtained in this embodiment is mainly wastewater from a propylene oxide device and ethylene alkali residue wastewater. After pretreatment, normalization and principal component analysis of the water quality parameters, multiple chemical wastewater water quality parameter samples are obtained. Each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set. The training sample set and the test sample set together constitute a sample data set for training an ATP and BOD prediction model for a wastewater treatment process, specifically comprising the following steps:

[0099] Step 1.1, obtain chemical wastewater from a chemical wastewater treatment plant, and determine the influent volume, water quality parameters, adenosine triphosphate (ATP), and biochemical oxygen demand (BOD) of the chemical wastewater by online instrumentation or laboratory testing and measurement, wherein the water quality parameters include influent chemical oxygen demand (COD) concentration, effluent chemical oxygen demand (COD) concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen (DO) concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, biochemical pool sulfide concentration, biochemical pool volatile phenol concentration, sludge settling ratio, sludge volume index, effluent BOD concentration, biochemical pool adenosine triphosphate (ATP), temperature, pH, and redox potential (ORP).

[0100] Step 1.2, pre-process the water quality parameters of chemical wastewater, eliminate the abnormal data in each water quality parameter, and then normalize the pre-treated water quality parameters to eliminate the impact of dimension and order of magnitude differences on the water quality parameters, and use the principal component analysis method to analyze the obtained water quality parameters, based on the correlation coefficient and contribution rate of each water quality parameter for the prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the sewage treatment process, screen them, use the screened water quality parameters and water inflow as input variables, combine the measured adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value of the chemical wastewater to determine the output variable, and use the input variables and output variables to form a chemical wastewater water quality parameter sample.

[0101] In this embodiment, the input variables include the influent chemical oxygen demand (COD) concentration, the biochemical pool sludge MLSS concentration, the biochemical pool dissolved oxygen (DO) concentration, the sludge volume index, the temperature, the influent ammonia nitrogen concentration, the propylene oxide concentration, the formaldehyde concentration and the influent flow rate, and the output variables include the adenosine triphosphate (ATP) concentration value and the biochemical oxygen demand (BOD) concentration value.

[0102] In step 1.3, each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set to generate a training sample set and a test sample set, and a sample data set is constructed for training the ATP and BOD prediction model of the wastewater treatment process.

[0103] In this embodiment, a total of 350 groups of chemical wastewater water quality parameter samples are obtained, wherein the training sample set includes 200 groups of chemical wastewater water quality parameter samples, and the test sample set includes 150 groups of test sample sets.

[0104] Step 2: Establish the ATP and BOD prediction model of the sewage treatment process based on the multi-input multi-output fuzzy neural network.

[0105] In this embodiment, the multi-input multi-output fuzzy neural network includes an input layer, an RBF layer, a rule layer and an output layer. Figure 1 shown.

[0106] The input layer of the multi-input multi-output fuzzy neural network is set as:

[0107] x i =u i (1)

[0108] x=[x1,x2,...,x k ] (2)

[0109] Wherein, i is the serial number of the input variable, i=1, 2, ..., k, k is the total number of input variables, in this embodiment, k=9; x i is the value of the i-th input variable, u iis the input value of the i-th input neuron, and x is the input vector.

[0110] The RBF layer is provided with a plurality of neurons, and the output function value of the neurons in the RBF layer is:

[0111]

[0112] In the formula, φ j is the output value of the jth neuron in the RBF layer, j is the number of the neuron, j=1, 2, ..., Q, Q is the total number of neurons in the RBF layer, in this embodiment, Q=6; c ij is the center of the jth neuron, σ ij is the width of the jth neuron;

[0113] The number of neurons set in the rule layer is equal to that in the RBF layer, and the output function value of the neurons in the rule layer is:

[0114]

[0115] Where l is the inner layer number of the rule layer, l = 1, 2, ..., Q; v l is the output value of the lth layer in the regular layer, φ l is the output value of the lth layer in the RBF layer;

[0116] The output value of the output layer is:

[0117] W=[w 1 ,w 2 ,…,w M ] T (5)

[0118] y=Wv (6)

[0119] Where W is the weight matrix, w M is the output value of the Mth neuron in the output layer, M is the total number of neurons in the output layer, and T is the transposed matrix; y is the abnormal working condition discrimination matrix, y=[y1,y2], y1 is the predicted value of adenosine triphosphate ATP concentration, y2 is the predicted value of biochemical oxygen demand BOD concentration; v is the output matrix of the rule layer, v=[v1,v2,···,v Q ] T , v Q is the output value of the Qth layer in the regular layer.

[0120] In this embodiment, a multi-input multi-output fuzzy neural network based on an adaptive learning rate is trained to construct an ATP and BOD prediction model for a sewage treatment process. By combining a second-order algorithm with an adaptive learning rate, the convergence speed of the Levenberg-Marquardt algorithm is effectively improved by utilizing an adaptive learning rate.

[0121] The update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate is set as:

[0122] Θ(t+1)=Θ(t)+(Ψ(t)+λ(t)×I) -1 Ω(t) (7)

[0123] Wherein, t is time, Θ(t+1) is the variable vector at time t+1, Θ(t) is the variable vector at time t, Ψ(t) is the quasi-Hessian matrix at time t, λ(t) is the adaptive learning rate at time t, 0<λ(t)<1, I is the unit matrix, and Ω(t) is the gradient vector at time t.

[0124] In the update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate, the calculation formula of the adaptive learning rate λ(t) is:

[0125] λ(t)=μ(t)λ(t-1) (8)

[0126]

[0127] Where μ(t) is the adaptive factor at time t, λ(t-1) is the adaptive learning rate at time t-1; τ min (t) is the minimum eigenvalue of the quasi-Hessian matrix at time t, τ max (t) is the maximum eigenvalue of the quasi-Hessian matrix at time t, 0<τ min (t)<τ max (t).

[0128] The variable vector Θ(t) includes a connection weight vector w(t), a center vector c(t) and a width vector σ(t). The calculation formula of the variable vector is:

[0129] Θ(t)=[w1(t),...,w J (t),c1(t),...,c J (t),...,σ1(t),...,σ J (t)] (10)

[0131] In the formula, w J(t) is the value of the J-th connection weight vector at time t, c J (t) is the J-th center vector value at time t, σ J (t) is the J-th width vector value at time t.

[0132] The calculation formula of the quasi-Hessian matrix Ψ(t) and the gradient vector Ω(t) is:

[0133] Ψ(t)=j T (t)j(t) (10)

[0134] Ω(t)=j T (t)e(t) (11)

[0135] e(t)=y d (t)-y(t) (12)

[0136] in,

[0137] The Jacobian vector j(t) is:

[0138]

[0139] Where j(t) is the Jacobian vector at time t, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, and y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0140] Step 3, using the training set to train the ATP and BOD prediction model of the sewage treatment process to obtain the trained ATP and BOD prediction model of the sewage treatment process, specifically including the following steps:

[0141] Step 3.1, set the training accuracy value.

[0142] Step 3.2, input the chemical wastewater water quality parameter samples in the training set into the ATP and BOD prediction model of the sewage treatment process constructed in step 2, and use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater to obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0143] Step 3.3, compare the predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value in the chemical sewage water quality parameter sample, and calculate the accuracy value of the ATP and BOD prediction model of the sewage treatment process. If the accuracy value of the ATP and BOD prediction model of the sewage treatment process is less than the preset accuracy value, proceed to step 3.4; otherwise, proceed to step 3.5.

[0144] Step 3.4, based on the Levenberg-Marquardt algorithm, the adaptive learning rate of the ATP and BOD prediction model of the sewage treatment process is updated to obtain the updated ATP and BOD prediction model of the sewage treatment process, and chemical wastewater water quality parameter samples are continuously randomly selected in the training set and input into the updated ATP and BOD prediction model of the sewage treatment process. The adenosine triphosphate (ATP) concentration and biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples are predicted using the updated ATP and BOD prediction model of the sewage treatment process. After obtaining the predicted values ​​of the adenosine triphosphate (ATP) concentration and the biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples, return to step 3.3.

[0145] Step 3.5, complete the training of the ATP and BOD prediction model for the sewage treatment process, and obtain the trained ATP and BOD prediction model for the sewage treatment process.

[0146] Step 4: Use the test set to verify the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training. If the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training has reached the best, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process, which specifically includes the following steps:

[0147] Step 4.1, input the chemical wastewater water quality parameter samples in the test set into the trained ATP and BOD prediction model of the sewage treatment process, use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0148] Step 4.2, compare the predicted adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical sewage water quality parameter sample, and calculate the accuracy of the ATP and BOD prediction model of the sewage treatment process. If the accuracy of the ATP and BOD prediction model of the sewage treatment process is not less than the preset accuracy, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process.

[0149] In this embodiment, the error values ​​of the ATP and BOD prediction models in the sewage treatment process are:

[0150]

[0151] The accuracy of the ATP and BOD prediction model for sewage treatment process is:

[0152]

[0153] in,

[0154] e(t)=y d (t)-y(t) (16)

[0155] Where RMSE(t) is the error value of the ATP and BOD prediction model of the sewage treatment process, t is time, t = 1, 2, ..., N, N is the total time, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0156] Step 5, using the tested ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in real time according to the water quality parameters of the chemical sewage, and generate an abnormal condition discrimination matrix, such as Figure 2 As shown in the figure, the abnormal operating condition discrimination matrix is ​​used to discriminate the type of abnormal operating condition, and the sludge condition and biochemical treatment potential of the chemical wastewater treatment plant are evaluated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value, and the load operation plan of the chemical wastewater treatment plant is adjusted.

[0157] The abnormal operating condition discrimination matrix is:

[0158]

[0159] In the formula, ATP is the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process, BOD is the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process; C1 is the first type of abnormal operating condition, C2 is the second type of abnormal operating condition, C3 is the third type of abnormal operating condition, and C4 is the fourth type of abnormal operating condition.

[0160] In this embodiment, when the adenosine triphosphate (ATP) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1.2, it is characterized by high ATP; when the adenosine triphosphate (ATP) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1.2, it is characterized by low ATP; when the biochemical oxygen demand (BOD) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1500, it is characterized by high BOD; when the biochemical oxygen demand (BOD) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1500, it is characterized by low BOD.

[0161] Based on the abnormal operating condition discrimination matrix, when the abnormal operating condition type is discriminated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value:

[0162] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are high, the current operating condition of the chemical wastewater treatment plant is determined to be the first abnormal condition type C1. The influent concentration of the chemical wastewater treatment plant exceeds the treatment capacity, the BOD concentration is abnormal, but the sludge activity is relatively high and the biochemical treatment capacity is high.

[0163] When adenosine triphosphate ATP is high and biochemical oxygen demand BOD is low, it is determined that the current operating condition of the chemical wastewater treatment plant is the second abnormal condition type C2. The operating condition is good, and the biochemical treatment capacity of the chemical wastewater treatment plant is surplus. The aeration volume can be appropriately reduced or the treatment load can be increased.

[0164] When adenosine triphosphate (ATP) is expressed as low and biochemical oxygen demand (BOD) is expressed as high, it is determined that the current operating condition of the chemical wastewater treatment plant is the third abnormal condition type C3, and the sludge in the current chemical wastewater treatment plant is determined to be impacted by water flow. It is necessary to measure the adenosine triphosphate (ATP) in the sludge, and analyze the abnormal condition in combination with the adenosine triphosphate (ATP) concentration value of the sludge to determine the abnormal state that the sludge may be in.

[0165] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are low, the current operating condition of the chemical wastewater treatment plant is determined to be the fourth abnormal condition type C4, with low biochemical treatment potential and low biochemical treatment capacity of the chemical wastewater treatment plant. The sludge treatment capacity of the chemical wastewater treatment plant can be improved by reducing the wastewater treatment volume of the chemical wastewater treatment plant, increasing the aeration volume or supplementing the carbon source.

[0166] It can be seen that the biochemical potential assessment method for abnormal conditions in the chemical wastewater treatment process in this embodiment uses the real-time predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration to establish an abnormal condition discrimination matrix, which realizes the accurate identification of abnormal conditions such as sludge poisoning and carbon source deficiency, and effectively ensures that the water quality of the effluent from chemical wastewater treatment under abnormal conditions is stable and meets the standards by automatically pushing process adjustment plans and emergency remedial measures.

[0167] Example 2

[0168] This embodiment adopts a method for evaluating the biochemical potential of abnormal working conditions in a chemical wastewater treatment process proposed by the present invention, and evaluates the biochemical treatment potential of abnormal working conditions of chemical wastewater by real-time prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, which specifically includes the following steps:

[0169] Step 1, obtaining water quality parameters of a chemical wastewater treatment plant. The chemical wastewater obtained in this embodiment is mainly purified terephthalic acid wastewater and ethylene alkali residue wastewater. After pre-processing, normalization and principal component analysis of the water quality parameters, a plurality of chemical wastewater water quality parameter samples are obtained. Each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set. The training sample set and the test sample set together constitute a sample data set for training an ATP and BOD prediction model for a wastewater treatment process. Specifically, the following steps are included:

[0170] Step 1.1, obtain chemical wastewater from the chemical wastewater treatment plant, and determine the influent volume, water quality parameters, adenosine triphosphate (ATP) and biochemical oxygen demand (BOD) of the chemical wastewater by online instrumentation or laboratory testing and measurement, wherein the water quality parameters include influent chemical oxygen demand (COD) concentration, effluent chemical oxygen demand (COD) concentration, formaldehyde concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen (DO) concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand (BOD) concentration, biochemical pool adenosine triphosphate (ATP), temperature, pH and redox potential (ORP).

[0171] Step 1.2, pre-process the water quality parameters of chemical wastewater, eliminate the abnormal data in each water quality parameter, and then normalize the pre-treated water quality parameters to eliminate the impact of dimension and order of magnitude differences on the water quality parameters, and use the principal component analysis method to analyze the obtained water quality parameters, based on the correlation coefficient and contribution rate of each water quality parameter for the prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the sewage treatment process, screen them, use the screened water quality parameters and water inflow as input variables, combine the measured adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value of the chemical wastewater to determine the output variable, and use the input variables and output variables to form a chemical wastewater water quality parameter sample.

[0172] In this embodiment, the input variables include the influent chemical oxygen demand (COD) concentration, the biochemical pool sludge MLSS concentration, the biochemical pool dissolved oxygen (DO) concentration, the sludge volume index, the temperature, the influent ammonia nitrogen concentration, the formaldehyde concentration, the terephthalic acid concentration, the influent flow rate and the isophthalic acid concentration, and the output variables include the adenosine triphosphate (ATP) concentration value and the biochemical oxygen demand (BOD) concentration value.

[0173] In step 1.3, each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set to generate a training sample set and a test sample set, and a sample data set is constructed for training the ATP and BOD prediction model of the wastewater treatment process.

[0174] In this embodiment, a total of 300 groups of chemical wastewater water quality parameter samples are obtained, wherein the training sample set includes 150 groups of chemical wastewater water quality parameter samples, and the test sample set includes 150 groups of test sample sets.

[0175] Step 2: Establish the ATP and BOD prediction model of the sewage treatment process based on the multi-input multi-output fuzzy neural network.

[0176] In this embodiment, the multi-input multi-output fuzzy neural network includes an input layer, an RBF layer, a rule layer and an output layer. Figure 1 shown.

[0177] The input layer of the multi-input multi-output fuzzy neural network is set as:

[0178] x i =u i (1)

[0179] x=[x1,x2,...,x k ] (2)

[0180] Wherein, i is the serial number of the input variable, i=1, 2, ..., k, k is the total number of input variables, in this embodiment, k=10; x i is the value of the i-th input variable, u iis the input value of the i-th input neuron, and x is the input vector.

[0181] The RBF layer is provided with a plurality of neurons, and the output function value of the neurons in the RBF layer is:

[0182]

[0183] In the formula, φ j is the output value of the jth neuron in the RBF layer, j is the number of the neuron, j=1, 2, ..., Q, Q is the total number of neurons in the RBF layer, in this embodiment, Q=7; c ij is the center of the jth neuron, σ ij is the width of the jth neuron.

[0184] The number of neurons set in the rule layer is equal to that in the RBF layer, and the output function value of the neurons in the rule layer is:

[0185]

[0186] Where l is the inner layer number of the rule layer, l = 1, 2, ..., Q; v l is the output value of the lth layer in the regular layer, φ l is the output value of the lth layer in the RBF layer.

[0187] The output value of the output layer is:

[0188] W=[w 1 ,w 2 ,…,w M ] T (5)

[0189] y=Wv (6)

[0190] Where W is the weight matrix, w M is the output value of the Mth neuron in the output layer, M is the total number of neurons in the output layer, and T is the transposed matrix; y is the abnormal working condition discrimination matrix, y=[y1,y2], y1 is the predicted value of adenosine triphosphate ATP concentration, y2 is the predicted value of biochemical oxygen demand BOD concentration; v is the output matrix of the rule layer, v=[v1,v2,···,v Q ] T , v Q is the output value of the Qth layer in the regular layer.

[0191] In this embodiment, a multi-input multi-output fuzzy neural network based on an adaptive learning rate is trained to construct an ATP and BOD prediction model for a sewage treatment process. By combining a second-order algorithm with an adaptive learning rate, the convergence speed of the Levenberg-Marquardt algorithm is effectively improved by utilizing an adaptive learning rate.

[0192] The update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate is set as:

[0193] Θ(t+1)=Θ(t)+(Ψ(t)+λ(t)×I) -1 Ω(t) (7)

[0194] Wherein, t is time, Θ(t+1) is the variable vector at time t+1, Θ(t) is the variable vector at time t, Ψ(t) is the quasi-Hessian matrix at time t, λ(t) is the adaptive learning rate at time t, 0<λ(t)<1, I is the unit matrix, and Ω(t) is the gradient vector at time t.

[0195] In the update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate, the calculation formula of the adaptive learning rate λ(t) is:

[0196] λ(t)=μ(t)λ(t-1) (8)

[0197]

[0198] Where μ(t) is the adaptive factor at time t, λ(t-1) is the adaptive learning rate at time t-1; τ min (t) is the minimum eigenvalue of the quasi-Hessian matrix at time t, τ max (t) is the maximum eigenvalue of the quasi-Hessian matrix at time t, 0<τ min (t)<τ max (t).

[0199] The variable vector Θ(t) includes a connection weight vector w(t), a center vector c(t) and a width vector σ(t). The calculation formula of the variable vector is:

[0200] Θ(t)=[w1(t),...,w J (t),c1(t),...,c J (t),...,σ1(t),...,σ J (t)] (10)

[0202] In the formula, w J(t) is the value of the J-th connection weight vector at time t, c J (t) is the J-th center vector value at time t, σ J (t) is the J-th width vector value at time t.

[0203] The calculation formula of the quasi-Hessian matrix Ψ(t) and the gradient vector Ω(t) is:

[0204] Ψ(t)=j T (t)j(t) (10)

[0205] Ω(t)=j T (t)e(t) (11)

[0206] e(t)=y d (t)-y(t) (12)

[0207] in,

[0208] The Jacobian vector j(t) is:

[0209]

[0210] Where j(t) is the Jacobian vector at time t, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, and y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0211] Step 3, using the training set to train the ATP and BOD prediction model of the sewage treatment process to obtain the trained ATP and BOD prediction model of the sewage treatment process, specifically including the following steps:

[0212] Step 3.1, set the training accuracy value.

[0213] Step 3.2, input the chemical wastewater water quality parameter samples in the training set into the ATP and BOD prediction model of the sewage treatment process constructed in step 2, and use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater to obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0214] Step 3.3, compare the predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value in the chemical sewage water quality parameter sample, and calculate the accuracy value of the ATP and BOD prediction model of the sewage treatment process. If the accuracy value of the ATP and BOD prediction model of the sewage treatment process is less than the preset accuracy value, proceed to step 3.4; otherwise, proceed to step 3.5.

[0215] Step 3.4, based on the Levenberg-Marquardt algorithm, the adaptive learning rate of the ATP and BOD prediction model of the sewage treatment process is updated to obtain the updated ATP and BOD prediction model of the sewage treatment process, and chemical wastewater water quality parameter samples are continuously randomly selected in the training set and input into the updated ATP and BOD prediction model of the sewage treatment process. The adenosine triphosphate (ATP) concentration and biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples are predicted using the updated ATP and BOD prediction model of the sewage treatment process. After obtaining the predicted values ​​of the adenosine triphosphate (ATP) concentration and the biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples, return to step 3.3.

[0216] Step 3.5, complete the training of the ATP and BOD prediction model for the sewage treatment process, and obtain the trained ATP and BOD prediction model for the sewage treatment process.

[0217] Step 4: Use the test set to verify the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training. If the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training has reached the best, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process, which specifically includes the following steps:

[0218] Step 4.1, input the chemical wastewater water quality parameter samples in the test set into the trained ATP and BOD prediction model of the sewage treatment process, use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0219] Step 4.2, compare the predicted adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical sewage water quality parameter sample, and calculate the accuracy of the ATP and BOD prediction model of the sewage treatment process. If the accuracy of the ATP and BOD prediction model of the sewage treatment process is not less than the preset accuracy, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process.

[0220] In this embodiment, the error values ​​of the ATP and BOD prediction models in the sewage treatment process are:

[0221]

[0222] The accuracy of the ATP and BOD prediction model for sewage treatment process is:

[0223]

[0224] in,

[0225] e(t)=y d (t)-y(t) (16)

[0226] Where RMSE(t) is the error value of the ATP and BOD prediction model of the sewage treatment process, t is time, t = 1, 2, ..., N, N is the total time, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0227] Step 5, using the tested ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in real time according to the water quality parameters of the chemical sewage, and generate an abnormal condition discrimination matrix, such as Figure 2 As shown in the figure, the abnormal operating condition discrimination matrix is ​​used to discriminate the type of abnormal operating condition, and the sludge condition and biochemical treatment potential of the chemical wastewater treatment plant are evaluated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value, and the load operation plan of the chemical wastewater treatment plant is adjusted.

[0228] The abnormal operating condition discrimination matrix is:

[0229]

[0230] In the formula, ATP is the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process, BOD is the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process; C1 is the first type of abnormal operating condition, C2 is the second type of abnormal operating condition, C3 is the third type of abnormal operating condition, and C4 is the fourth type of abnormal operating condition.

[0231] In this embodiment, when the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1.3, it is characterized by high ATP; when the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1.3, it is characterized by low ATP; when the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1300, it is characterized by high BOD; when the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1300, it is characterized by low BOD.

[0232] Based on the abnormal operating condition discrimination matrix, when the abnormal operating condition type is discriminated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value:

[0233] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are high, the current operating condition of the chemical wastewater treatment plant is determined to be the first abnormal condition type C1. The influent concentration of the chemical wastewater treatment plant exceeds the treatment capacity, the BOD concentration is abnormal, but the sludge activity is relatively high and the biochemical treatment capacity is high.

[0234] When adenosine triphosphate ATP is high and biochemical oxygen demand BOD is low, it is determined that the current operating condition of the chemical wastewater treatment plant is the second abnormal condition type C2. The operating condition is good, and the biochemical treatment capacity of the chemical wastewater treatment plant is surplus. The aeration volume can be appropriately reduced or the treatment load can be increased.

[0235] When adenosine triphosphate (ATP) is expressed as low and biochemical oxygen demand (BOD) is expressed as high, it is determined that the current operating condition of the chemical wastewater treatment plant is the third abnormal condition type C3, and the sludge in the current chemical wastewater treatment plant is determined to be impacted by water flow. It is necessary to measure the adenosine triphosphate (ATP) in the sludge, and analyze the abnormal condition in combination with the adenosine triphosphate (ATP) concentration value of the sludge to determine the abnormal state that the sludge may be in.

[0236] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are low, the current operating condition of the chemical wastewater treatment plant is determined to be the fourth abnormal condition type C4, with low biochemical treatment potential and low biochemical treatment capacity of the chemical wastewater treatment plant. The sludge treatment capacity of the chemical wastewater treatment plant can be improved by reducing the wastewater treatment volume of the chemical wastewater treatment plant, increasing the aeration volume or supplementing the carbon source.

[0237] It can be seen that the biochemical potential assessment method for abnormal conditions in the chemical wastewater treatment process in this embodiment uses the real-time predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration to establish an abnormal condition discrimination matrix, which realizes the accurate identification of abnormal conditions such as sludge poisoning and carbon source deficiency, and effectively ensures that the water quality of the effluent from chemical wastewater treatment under abnormal conditions is stable and meets the standards by automatically pushing process adjustment plans and emergency remedial measures.

[0238] Example 3

[0239] This embodiment adopts a method for evaluating the biochemical potential of abnormal working conditions in a chemical wastewater treatment process proposed by the present invention, and evaluates the biochemical treatment potential of abnormal working conditions of chemical wastewater by real-time prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, which specifically includes the following steps:

[0240] Step 1, obtaining water quality parameters of a chemical wastewater treatment plant. The chemical wastewater obtained in this embodiment is mainly propylene oxide wastewater, purified terephthalic acid wastewater and ethylene alkali residue wastewater. After pre-treatment, normalization and principal component analysis of the water quality parameters, a plurality of chemical wastewater water quality parameter samples are obtained. Each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set. The training sample set and the test sample set together constitute a sample data set for training an ATP and BOD prediction model for a wastewater treatment process, specifically comprising the following steps:

[0241] Step 1.1, obtain chemical wastewater from a chemical wastewater treatment plant, and determine the influent volume, water quality parameters, adenosine triphosphate (ATP), and biochemical oxygen demand (BOD) of the chemical wastewater by online instrumentation or laboratory testing and measurement, wherein the water quality parameters include influent chemical oxygen demand (COD) concentration, effluent chemical oxygen demand (COD) concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen (DO) concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand (BOD) concentration, biochemical pool adenosine triphosphate (ATP), temperature, pH, and redox potential (ORP).

[0242] Step 1.2, pre-process the water quality parameters of chemical wastewater, eliminate the abnormal data in each water quality parameter, and then normalize the pre-treated water quality parameters to eliminate the impact of dimension and order of magnitude differences on the water quality parameters, and use the principal component analysis method to analyze the obtained water quality parameters, based on the correlation coefficient and contribution rate of each water quality parameter for the prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the sewage treatment process, screen them, use the screened water quality parameters and water inflow as input variables, combine the measured adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value of the chemical wastewater to determine the output variable, and use the input variables and output variables to form a chemical wastewater water quality parameter sample.

[0243] In this embodiment, the input variables include influent chemical oxygen demand (COD) concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen (DO) concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, terephthalic acid concentration and influent flow rate, and the output variables include adenosine triphosphate (ATP) concentration value and biochemical oxygen demand (BOD) concentration value.

[0244] In step 1.3, each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set to generate a training sample set and a test sample set, and a sample data set is constructed for training the ATP and BOD prediction model of the wastewater treatment process.

[0245] In this embodiment, a total of 400 groups of chemical wastewater water quality parameter samples are obtained, wherein the training sample set includes 200 groups of chemical wastewater water quality parameter samples, and the test sample set includes 200 groups of test sample sets.

[0246] Step 2: Establish the ATP and BOD prediction model of the sewage treatment process based on the multi-input multi-output fuzzy neural network.

[0247] In this embodiment, the multi-input multi-output fuzzy neural network includes an input layer, an RBF layer, a rule layer and an output layer. Figure 1 shown.

[0248] The input layer of the multi-input multi-output fuzzy neural network is set as:

[0249] x i =u i (1)

[0250] x=[x1,x2,...,x k ] (2)

[0251] Wherein, i is the serial number of the input variable, i=1, 2, ..., k, k is the total number of input variables, in this embodiment, k=12; x i is the value of the i-th input variable, ui is the input value of the i-th input neuron, and x is the input vector.

[0252] The RBF layer is provided with a plurality of neurons, and the output function value of the neurons in the RBF layer is:

[0253]

[0254] In the formula, φ j is the output value of the jth neuron in the RBF layer, j is the number of the neuron, j=1, 2, ..., Q, Q is the total number of neurons in the RBF layer, in this embodiment, Q=7; c ij is the center of the jth neuron, σ ij is the width of the jth neuron.

[0255] The number of neurons set in the rule layer is equal to that in the RBF layer, and the output function value of the neurons in the rule layer is:

[0256]

[0257] Where l is the inner layer number of the rule layer, l = 1, 2, ..., Q; v l is the output value of the lth layer in the regular layer, φ l is the output value of the lth layer in the RBF layer.

[0258] The output value of the output layer is:

[0259] W=[w 1 ,w 2 ,…,w M ] T (5)

[0260] y=Wv (6)

[0261] Where W is the weight matrix, w M is the output value of the Mth neuron in the output layer, M is the total number of neurons in the output layer, and T is the transposed matrix; y is the abnormal working condition discrimination matrix, y=[y1,y2], y1 is the predicted value of adenosine triphosphate ATP concentration, y2 is the predicted value of biochemical oxygen demand BOD concentration; v is the output matrix of the rule layer, v=[v1,v2,···,v Q ] T , v Q is the output value of the Qth layer in the regular layer.

[0262] In this embodiment, a multi-input multi-output fuzzy neural network based on an adaptive learning rate is trained to construct an ATP and BOD prediction model for a sewage treatment process. By combining a second-order algorithm with an adaptive learning rate, the convergence speed of the Levenberg-Marquardt algorithm is effectively improved by utilizing an adaptive learning rate.

[0263] The update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate is set as:

[0264] Θ(t+1)=Θ(t)+(Ψ(t)+λ(t)×I) -1 Ω(t) (7)

[0265] Wherein, t is time, Θ(t+1) is the variable vector at time t+1, Θ(t) is the variable vector at time t, Ψ(t) is the quasi-Hessian matrix at time t, λ(t) is the adaptive learning rate at time t, 0<λ(t)<1, I is the unit matrix, and Ω(t) is the gradient vector at time t.

[0266] In the update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate, the calculation formula of the adaptive learning rate λ(t) is:

[0267] λ(t)=μ(t)λ(t-1) (8)

[0268]

[0269] Where μ(t) is the adaptive factor at time t, λ(t-1) is the adaptive learning rate at time t-1; τ min (t) is the minimum eigenvalue of the quasi-Hessian matrix at time t, τ max (t) is the maximum eigenvalue of the quasi-Hessian matrix at time t, 0<τ min (t)<τ max (t).

[0270] The variable vector Θ(t) includes a connection weight vector w(t), a center vector c(t) and a width vector σ(t). The calculation formula of the variable vector is:

[0271] Θ(t)=[w1(t),...,w J (t),c1(t),...,c J (t),...,σ1(t),...,σ J (t)] (10)

[0273] In the formula, w J(t) is the value of the J-th connection weight vector at time t, c J (t) is the J-th center vector value at time t, σ J (t) is the J-th width vector value at time t.

[0274] The calculation formula of the quasi-Hessian matrix Ψ(t) and the gradient vector Ω(t) is:

[0275] Ψ(t)=j T (t)j(t) (10)

[0276] Ω(t)=j T (t)e(t) (11)

[0277] e(t)=y d (t)-y(t) (12)

[0278] in,

[0279] The Jacobian vector j(t) is:

[0280]

[0281] Where j(t) is the Jacobian vector at time t, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, and y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0282] Step 3, using the training set to train the ATP and BOD prediction model of the sewage treatment process to obtain the trained ATP and BOD prediction model of the sewage treatment process, specifically including the following steps:

[0283] Step 3.1, set the training accuracy value.

[0284] Step 3.2, input the chemical wastewater water quality parameter samples in the training set into the ATP and BOD prediction model of the sewage treatment process constructed in step 2, and use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater to obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0285] Step 3.3, compare the predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value in the chemical sewage water quality parameter sample, and calculate the accuracy value of the ATP and BOD prediction model of the sewage treatment process. If the accuracy value of the ATP and BOD prediction model of the sewage treatment process is less than the preset accuracy value, proceed to step 3.4; otherwise, proceed to step 3.5.

[0286] Step 3.4, based on the Levenberg-Marquardt algorithm, the adaptive learning rate of the ATP and BOD prediction model of the sewage treatment process is updated to obtain the updated ATP and BOD prediction model of the sewage treatment process, and chemical wastewater water quality parameter samples are continuously randomly selected in the training set and input into the updated ATP and BOD prediction model of the sewage treatment process. The adenosine triphosphate (ATP) concentration and biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples are predicted using the updated ATP and BOD prediction model of the sewage treatment process. After obtaining the predicted values ​​of the adenosine triphosphate (ATP) concentration and the biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples, return to step 3.3.

[0287] Step 3.5, complete the training of the ATP and BOD prediction model for the sewage treatment process, and obtain the trained ATP and BOD prediction model for the sewage treatment process.

[0288] Step 4: Use the test set to verify the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training. If the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training has reached the best, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process, which specifically includes the following steps:

[0289] Step 4.1, input the chemical wastewater water quality parameter samples in the test set into the trained ATP and BOD prediction model of the sewage treatment process, use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0290] Step 4.2, compare the predicted adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical sewage water quality parameter sample, and calculate the accuracy of the ATP and BOD prediction model of the sewage treatment process. If the accuracy of the ATP and BOD prediction model of the sewage treatment process is not less than the preset accuracy, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process.

[0291] In this embodiment, the error values ​​of the ATP and BOD prediction models in the sewage treatment process are:

[0292]

[0293] The accuracy of the ATP and BOD prediction model for sewage treatment process is:

[0294]

[0295] in,

[0296] e(t)=y d (t)-y(t) (16)

[0297] Where RMSE(t) is the error value of the ATP and BOD prediction model of the sewage treatment process, t is time, t = 1, 2, ..., N, N is the total time, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0298] Step 5, using the tested ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in real time according to the water quality parameters of the chemical sewage, and generate an abnormal condition discrimination matrix, such as Figure 2 As shown in the figure, the abnormal operating condition discrimination matrix is ​​used to discriminate the type of abnormal operating condition, and the sludge condition and biochemical treatment potential of the chemical wastewater treatment plant are evaluated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value, and the load operation plan of the chemical wastewater treatment plant is adjusted.

[0299] The abnormal operating condition discrimination matrix is:

[0300]

[0301] In the formula, ATP is the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process, BOD is the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process; C1 is the first type of abnormal operating condition, C2 is the second type of abnormal operating condition, C3 is the third type of abnormal operating condition, and C4 is the fourth type of abnormal operating condition.

[0302] In this embodiment, when the adenosine triphosphate (ATP) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1.1, it is characterized by high ATP; when the adenosine triphosphate (ATP) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1.1, it is characterized by low ATP; when the biochemical oxygen demand (BOD) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1200, it is characterized by high BOD; when the biochemical oxygen demand (BOD) concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1200, it is characterized by low BOD.

[0303] Based on the abnormal operating condition discrimination matrix, when the abnormal operating condition type is discriminated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value:

[0304] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are high, the current operating condition of the chemical wastewater treatment plant is determined to be the first abnormal condition type C1. The influent concentration of the chemical wastewater treatment plant exceeds the treatment capacity, the BOD concentration is abnormal, but the sludge activity is relatively high and the biochemical treatment capacity is high.

[0305] When adenosine triphosphate ATP is high and biochemical oxygen demand BOD is low, it is determined that the current operating condition of the chemical wastewater treatment plant is the second abnormal condition type C2. The operating condition is good, and the biochemical treatment capacity of the chemical wastewater treatment plant is surplus. The aeration volume can be appropriately reduced or the treatment load can be increased.

[0306] When adenosine triphosphate (ATP) is expressed as low and biochemical oxygen demand (BOD) is expressed as high, it is determined that the current operating condition of the chemical wastewater treatment plant is the third abnormal condition type C3, and the sludge in the current chemical wastewater treatment plant is determined to be impacted by water flow. It is necessary to measure the adenosine triphosphate (ATP) in the sludge, and analyze the abnormal condition in combination with the adenosine triphosphate (ATP) concentration value of the sludge to determine the abnormal state that the sludge may be in.

[0307] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are low, the current operating condition of the chemical wastewater treatment plant is determined to be the fourth abnormal condition type C4, with low biochemical treatment potential and low biochemical treatment capacity of the chemical wastewater treatment plant. The sludge treatment capacity of the chemical wastewater treatment plant can be improved by reducing the wastewater treatment volume of the chemical wastewater treatment plant, increasing the aeration volume or supplementing the carbon source.

[0308] It can be seen that the biochemical potential assessment method for abnormal conditions in the chemical wastewater treatment process in this embodiment uses the real-time predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration to establish an abnormal condition discrimination matrix, which realizes the accurate identification of abnormal conditions such as sludge poisoning and carbon source deficiency, and effectively ensures that the water quality of the effluent from chemical wastewater treatment under abnormal conditions is stable and meets the standards by automatically pushing process adjustment plans and emergency remedial measures.

[0309] Example 4

[0310] This embodiment adopts a method for evaluating the biochemical potential of abnormal working conditions in a chemical wastewater treatment process proposed by the present invention, and evaluates the biochemical treatment potential of abnormal working conditions of chemical wastewater by real-time prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, which specifically includes the following steps:

[0311] Step 1, obtaining water quality parameters of a chemical wastewater treatment plant. The chemical wastewater obtained in this embodiment is mainly propylene oxide wastewater, aromatic hydrocarbon wastewater, purified terephthalic acid wastewater and ethylene alkali residue wastewater. After pre-treatment, normalization and principal component analysis of the water quality parameters, a plurality of chemical wastewater water quality parameter samples are obtained. Each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set. The training sample set and the test sample set together constitute a sample data set for training an ATP and BOD prediction model for a wastewater treatment process, specifically comprising the following steps:

[0312] Step 1.1, obtain chemical sewage from the chemical sewage treatment plant, and determine the influent volume, water quality parameters, adenosine triphosphate (ATP) and biochemical oxygen demand (BOD) of the chemical sewage by online instrumentation or laboratory testing and measurement, wherein the water quality parameters include influent chemical oxygen demand (COD) concentration, effluent chemical oxygen demand (COD) concentration, cyanide concentration, polycyclic aromatic hydrocarbons, volatile phenol concentration, sulfide concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen (DO) concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand (BOD) concentration, biochemical pool adenosine triphosphate (ATP), temperature, pH and redox potential (ORP).

[0313] Step 1.2, pre-process the water quality parameters of chemical wastewater, eliminate the abnormal data in each water quality parameter, and then normalize the pre-treated water quality parameters to eliminate the impact of dimension and order of magnitude differences on the water quality parameters, and use the principal component analysis method to analyze the obtained water quality parameters, based on the correlation coefficient and contribution rate of each water quality parameter for the prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the sewage treatment process, screen them, use the screened water quality parameters and water inflow as input variables, combine the measured adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value of the chemical wastewater to determine the output variable, and use the input variables and output variables to form a chemical wastewater water quality parameter sample.

[0314] In this embodiment, the input variables include influent chemical oxygen demand (COD) concentration, biochemical pool sludge MLSS concentration, polycyclic aromatic hydrocarbons, volatile phenol concentration, sulfide concentration, biochemical pool dissolved oxygen (DO) concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, isopropylbenzene concentration, propylene oxide concentration, terephthalic acid concentration and influent flow rate, and the output variables include adenosine triphosphate (ATP) concentration value and biochemical oxygen demand (BOD) concentration value.

[0315] In step 1.3, each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set to generate a training sample set and a test sample set, and a sample data set is constructed for training the ATP and BOD prediction model of the wastewater treatment process.

[0316] In this embodiment, a total of 350 groups of chemical wastewater water quality parameter samples are obtained, wherein the training sample set includes 200 groups of chemical wastewater water quality parameter samples, and the test sample set includes 150 groups of test sample sets.

[0317] Step 2: Establish the ATP and BOD prediction model of the sewage treatment process based on the multi-input multi-output fuzzy neural network.

[0318] In this embodiment, the multi-input multi-output fuzzy neural network includes an input layer, an RBF layer, a rule layer and an output layer. Figure 1 shown.

[0319] The input layer of the multi-input multi-output fuzzy neural network is set as:

[0320] x i =u i (1)

[0321] x=[x1,x2,...,x k ] (2)

[0322] Wherein, i is the serial number of the input variable, i=1, 2, ..., k, k is the total number of input variables, in this embodiment, k=14; x iis the value of the i-th input variable, u i is the input value of the i-th input neuron, and x is the input vector.

[0323] The RBF layer is provided with a plurality of neurons, and the output function value of the neurons in the RBF layer is:

[0324]

[0325] In the formula, φ j is the output value of the jth neuron in the RBF layer, j is the number of the neuron, j=1, 2, ..., Q, Q is the total number of neurons in the RBF layer, in this embodiment, Q=8; c ij is the center of the jth neuron, σ ij is the width of the jth neuron.

[0326] The number of neurons set in the rule layer is equal to that in the RBF layer, and the output function value of the neurons in the rule layer is:

[0327]

[0328] Where l is the inner layer number of the rule layer, l = 1, 2, ..., Q; v l is the output value of the lth layer in the regular layer, φ l is the output value of the lth layer in the RBF layer.

[0329] The output value of the output layer is:

[0330] W=[w 1 ,w 2 ,…,w M ] T (5)

[0331] y=Wv (6)

[0332] Where W is the weight matrix, w M is the output value of the Mth neuron in the output layer, M is the total number of neurons in the output layer, and T is the transposed matrix; y is the abnormal working condition discrimination matrix, y=[y1,y2], y1 is the predicted value of adenosine triphosphate ATP concentration, y2 is the predicted value of biochemical oxygen demand BOD concentration; v is the output matrix of the rule layer, v=[v1,v2,···,v Q ] T , v Q is the output value of the Qth layer in the regular layer.

[0333] In this embodiment, a multi-input multi-output fuzzy neural network based on an adaptive learning rate is trained to construct an ATP and BOD prediction model for a sewage treatment process. By combining a second-order algorithm with an adaptive learning rate, the convergence speed of the Levenberg-Marquardt algorithm is effectively improved by utilizing an adaptive learning rate.

[0334] The update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate is set as:

[0335] Θ(t+1)=Θ(t)+(Ψ(t)+λ(t)×I) -1 Ω(t) (7)

[0336] Wherein, t is time, Θ(t+1) is the variable vector at time t+1, Θ(t) is the variable vector at time t, Ψ(t) is the quasi-Hessian matrix at time t, λ(t) is the adaptive learning rate at time t, 0<λ(t)<1, I is the unit matrix, and Ω(t) is the gradient vector at time t.

[0337] In the update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate, the calculation formula of the adaptive learning rate λ(t) is:

[0338] λ(t)=μ(t)λ(t-1) (8)

[0339]

[0340] Where μ(t) is the adaptive factor at time t, λ(t-1) is the adaptive learning rate at time t-1; τ min (t) is the minimum eigenvalue of the quasi-Hessian matrix at time t, τ max (t) is the maximum eigenvalue of the quasi-Hessian matrix at time t, 0<τ min (t)<τ max (t).

[0341] The variable vector Θ(t) includes a connection weight vector w(t), a center vector c(t) and a width vector σ(t). The calculation formula of the variable vector is:

[0342] Θ(t)=[w1(t),...,w J (t),c1(t),...,c J (t),...,σ1(t),...,σ J (t)] (10)

[0344] In the formula, w J(t) is the value of the J-th connection weight vector at time t, c J (t) is the J-th center vector value at time t, σ J (t) is the J-th width vector value at time t.

[0345] The calculation formula of the quasi-Hessian matrix Ψ(t) and the gradient vector Ω(t) is:

[0346] Ψ(t)=j T (t)j(t) (10)

[0347] Ω(t)=j T (t)e(t) (11)

[0348] e(t)=y d (t)-y(t) (12)

[0349] in,

[0350] The Jacobian vector j(t) is:

[0351]

[0352] Where j(t) is the Jacobian vector at time t, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, and y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0353] Step 3, using the training set to train the ATP and BOD prediction model of the sewage treatment process to obtain the trained ATP and BOD prediction model of the sewage treatment process, specifically including the following steps:

[0354] Step 3.1, set the training accuracy value.

[0355] Step 3.2, input the chemical wastewater water quality parameter samples in the training set into the ATP and BOD prediction model of the sewage treatment process constructed in step 2, and use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater to obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0356] Step 3.3, compare the predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value in the chemical sewage water quality parameter sample, and calculate the accuracy value of the ATP and BOD prediction model of the sewage treatment process. If the accuracy value of the ATP and BOD prediction model of the sewage treatment process is less than the preset accuracy value, proceed to step 3.4; otherwise, proceed to step 3.5.

[0357] Step 3.4, based on the Levenberg-Marquardt algorithm, the adaptive learning rate of the ATP and BOD prediction model of the sewage treatment process is updated to obtain the updated ATP and BOD prediction model of the sewage treatment process, and chemical wastewater water quality parameter samples are continuously randomly selected in the training set and input into the updated ATP and BOD prediction model of the sewage treatment process. The adenosine triphosphate (ATP) concentration and biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples are predicted using the updated ATP and BOD prediction model of the sewage treatment process. After obtaining the predicted values ​​of the adenosine triphosphate (ATP) concentration and the biochemical oxygen demand (BOD) concentration of the chemical wastewater water quality parameter samples, return to step 3.3.

[0358] Step 3.5, complete the training of the ATP and BOD prediction model for the sewage treatment process, and obtain the trained ATP and BOD prediction model for the sewage treatment process.

[0359] Step 4: Use the test set to verify the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training. If the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training has reached the best, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process, which specifically includes the following steps:

[0360] Step 4.1, input the chemical wastewater water quality parameter samples in the test set into the trained ATP and BOD prediction model of the sewage treatment process, use the ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtain the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater.

[0361] Step 4.2, compare the predicted adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical sewage water quality parameter sample, and calculate the accuracy of the ATP and BOD prediction model of the sewage treatment process. If the accuracy of the ATP and BOD prediction model of the sewage treatment process is not less than the preset accuracy, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process.

[0362] In this embodiment, the error values ​​of the ATP and BOD prediction models in the sewage treatment process are:

[0363]

[0364] The accuracy of the ATP and BOD prediction model for sewage treatment process is:

[0365]

[0366] in,

[0367] e(t)=y d (t)-y(t) (16)

[0368] Where RMSE(t) is the error value of the ATP and BOD prediction model of the sewage treatment process, t is time, t = 1, 2, ..., N, N is the total time, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

[0369] Step 5, using the tested ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in real time according to the water quality parameters of the chemical sewage, and generate an abnormal condition discrimination matrix, such as Figure 2 As shown in the figure, the abnormal operating condition discrimination matrix is ​​used to discriminate the type of abnormal operating condition, and the sludge condition and biochemical treatment potential of the chemical wastewater treatment plant are evaluated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value, and the load operation plan of the chemical wastewater treatment plant is adjusted.

[0370] The abnormal operating condition discrimination matrix is:

[0371]

[0372] In the formula, ATP is the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process, BOD is the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process; C1 is the first type of abnormal operating condition, C2 is the second type of abnormal operating condition, C3 is the third type of abnormal operating condition, and C4 is the fourth type of abnormal operating condition.

[0373] In this embodiment, when the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1.1, it is characterized by high ATP; when the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1.1, it is characterized by low ATP; when the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process is greater than 1100, it is characterized by high BOD; when the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process does not exceed 1100, it is characterized by low BOD.

[0374] Based on the abnormal operating condition discrimination matrix, when the abnormal operating condition type is discriminated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value:

[0375] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are high, the current operating condition of the chemical wastewater treatment plant is determined to be the first abnormal condition type C1. The influent concentration of the chemical wastewater treatment plant exceeds the treatment capacity, the BOD concentration is abnormal, but the sludge activity is relatively high and the biochemical treatment capacity is high.

[0376] When adenosine triphosphate ATP is high and biochemical oxygen demand BOD is low, it is determined that the current operating condition of the chemical wastewater treatment plant is the second abnormal condition type C2. The operating condition is good, and the biochemical treatment capacity of the chemical wastewater treatment plant is surplus. The aeration volume can be appropriately reduced or the treatment load can be increased.

[0377] When adenosine triphosphate (ATP) is expressed as low and biochemical oxygen demand (BOD) is expressed as high, it is determined that the current operating condition of the chemical wastewater treatment plant is the third abnormal condition type C3, and the sludge in the current chemical wastewater treatment plant is determined to be impacted by water flow. It is necessary to measure the adenosine triphosphate (ATP) in the sludge, and analyze the abnormal condition in combination with the adenosine triphosphate (ATP) concentration value of the sludge to determine the abnormal state that the sludge may be in.

[0378] When both adenosine triphosphate ATP and biochemical oxygen demand BOD are low, the current operating condition of the chemical wastewater treatment plant is determined to be the fourth abnormal condition type C4, with low biochemical treatment potential and low biochemical treatment capacity of the chemical wastewater treatment plant. The sludge treatment capacity of the chemical wastewater treatment plant can be improved by reducing the wastewater treatment volume of the chemical wastewater treatment plant, increasing the aeration volume or supplementing the carbon source.

[0379] It can be seen that the biochemical potential assessment method for abnormal conditions in the chemical wastewater treatment process in this embodiment uses the real-time predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration to establish an abnormal condition discrimination matrix, which realizes the accurate identification of abnormal conditions such as sludge poisoning and carbon source deficiency, and effectively ensures that the water quality of the effluent from chemical wastewater treatment under abnormal conditions is stable and meets the standards by automatically pushing process adjustment plans and emergency remedial measures.

[0380] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms used herein are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0381] In the present invention, the directions or positional relationships indicated by terms such as "upper", "lower", "bottom", "top", etc. are based on the directions or positional relationships shown in the accompanying drawings. They are relational words determined only for the convenience of describing the structural relationships of the various parts or elements of the present invention. They do not specifically refer to any part or element in the present invention and cannot be understood as limitations on the present invention.

[0382] In the present invention, terms such as "connected" and "connection" should be understood in a broad sense, indicating that the connection can be fixed, integral or detachable; it can be directly connected or indirectly connected through an intermediate medium. Relevant scientific research or technical personnel in this field can determine the specific meaning of the above terms in the present invention according to specific circumstances, and they should not be understood as limiting the present invention.

[0383] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process, characterized in that: The biochemical treatment potential of abnormal conditions of chemical wastewater is evaluated by real-time prediction of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration, which specifically includes the following steps: Step 1, obtaining water quality parameters of a chemical wastewater treatment plant, performing pretreatment, normalization and principal component analysis on the water quality parameters to obtain multiple chemical wastewater water quality parameter samples, and randomly assigning each chemical wastewater water quality parameter sample to a training sample set and a test sample set, which together constitute a sample data set for training an ATP and BOD prediction model for a wastewater treatment process; Step 2, establishing a prediction model of ATP and BOD in the sewage treatment process based on a multi-input multi-output fuzzy neural network; Step 3, using the training set to train the ATP and BOD prediction model of the sewage treatment process to obtain the trained ATP and BOD prediction model of the sewage treatment process; Step 4, using the test set to verify the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training. If the prediction effect of the ATP and BOD prediction model of the sewage treatment process after training has reached the best, then proceed to step 5; otherwise, return to step 3 and continue to train the ATP and BOD prediction model of the sewage treatment process using the training set; Step 5, using the tested ATP and BOD prediction model of the sewage treatment process to predict the adenosine triphosphate (ATP) concentration and the biochemical oxygen demand (BOD) concentration in real time according to the water quality parameters of the chemical sewage, generate an abnormal operating condition discrimination matrix, use the abnormal operating condition discrimination matrix to discriminate the type of abnormal operating condition, and evaluate the sludge condition and biochemical treatment potential of the chemical sewage treatment plant according to the adenosine triphosphate (ATP) concentration value and the biochemical oxygen demand (BOD) concentration value, and adjust the load operation plan of the chemical sewage treatment plant.

2. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 1, characterized in that: The step 1 specifically includes the following steps: Step 1.1, obtaining chemical wastewater from a chemical wastewater treatment plant, and measuring the influent volume, water quality parameters, adenosine triphosphate (ATP), and biochemical oxygen demand (BOD) of the chemical wastewater; Step 1.2, pre-processing each water quality parameter of the chemical wastewater, eliminating abnormal data in each water quality parameter, and then normalizing each pre-processed water quality parameter, and using the principal component analysis method to analyze the obtained water quality parameters, based on the correlation coefficient and contribution rate of each water quality parameter for the prediction of ATP and BOD in the sewage treatment process, screening is performed, and the screened water quality parameters and water inflow are used as input variables, and the output variables are determined in combination with the measured adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value of the chemical wastewater, and the input variables and output variables are used to form a chemical wastewater water quality parameter sample; In step 1.3, each chemical wastewater water quality parameter sample is randomly assigned to a training sample set and a test sample set to generate a training sample set and a test sample set, and a sample data set is constructed for training the ATP and BOD prediction model of the wastewater treatment process.

3. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 2, characterized in that: When the chemical wastewater is propylene oxide device wastewater and ethylene alkali slag wastewater, the water quality parameters include inlet chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, inlet flow, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, inlet ammonia nitrogen concentration, effluent ammonia nitrogen concentration, biochemical pool sulfide concentration, biochemical pool volatile phenol concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and redox potential ORP; When the chemical wastewater is purified terephthalic acid wastewater and ethylene alkali residue wastewater, the water quality parameters include influent chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, formaldehyde concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and oxidation-reduction potential ORP; When the chemical wastewater is propylene oxide wastewater, purified terephthalic acid wastewater and ethylene alkali residue wastewater, the water quality parameters include influent chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, influent flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, influent ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and redox potential ORP; When the chemical wastewater is propylene oxide wastewater, aromatic hydrocarbon wastewater, purified terephthalic acid wastewater and ethylene alkali slag wastewater, the water quality parameters include inlet chemical oxygen demand COD concentration, effluent chemical oxygen demand COD concentration, cyanide concentration, polycyclic aromatic hydrocarbons, volatile phenol concentration, sulfide concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, propylene concentration, formic acid concentration, terephthalic acid concentration, isophthalic acid concentration, acetic acid concentration, inlet flow rate, sludge retention time, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, inlet ammonia nitrogen concentration, effluent ammonia nitrogen concentration, sludge settling ratio, sludge volume index, effluent biochemical oxygen demand BOD concentration, biochemical pool adenosine triphosphate ATP, temperature, pH and redox potential ORP.

4. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 3, characterized in that: When the chemical wastewater is propylene oxide device wastewater and ethylene alkali residue wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, propylene oxide concentration, formaldehyde concentration and influent flow rate, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value; When the chemical wastewater is purified terephthalic acid wastewater and ethylene alkali residue wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, terephthalic acid concentration, influent flow rate and isophthalic acid concentration, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value; When the chemical wastewater is propylene oxide wastewater, purified terephthalic acid wastewater and ethylene alkali residue wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, isopropylbenzene concentration, methanol concentration, propylene oxide concentration, terephthalic acid concentration and influent flow rate, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value; When the chemical wastewater is propylene oxide wastewater, aromatic hydrocarbon wastewater, purified terephthalic acid wastewater and ethylene alkali slag wastewater, the input variables include influent chemical oxygen demand COD concentration, biochemical pool sludge MLSS concentration, polycyclic aromatic hydrocarbons, volatile phenol concentration, sulfide concentration, biochemical pool dissolved oxygen DO concentration, sludge volume index, temperature, influent ammonia nitrogen concentration, formaldehyde concentration, isopropylbenzene concentration, propylene oxide concentration, terephthalic acid concentration and influent flow rate, and the output variables include adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value.

5. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 2, characterized in that: In step 2, the multi-input multi-output fuzzy neural network includes an input layer, an RBF layer, a rule layer and an output layer; The input layer of the multi-input multi-output fuzzy neural network is set as: x i =u i (1) x=[x1,x2,...,x k ] (2) Where i is the serial number of the input variable, i = 1, 2, ..., k, k is the total number of input variables; x i is the value of the i-th input variable, u i is the input value of the i-th input neuron, and x is the input vector; The RBF layer is provided with a plurality of neurons, and the output function value of the neurons in the RBF layer is: In the formula, φ j is the output value of the jth neuron in the RBF layer, j is the number of the neuron, j = 1, 2, ..., Q, Q is the total number of neurons in the RBF layer; c ij is the center of the jth neuron, σ ij is the width of the jth neuron; A plurality of neurons are arranged in the rule layer, and the output function value of the neurons in the rule layer is: Where l is the inner layer number of the rule layer, l = 1, 2, ..., Q; v l is the output value of the lth layer in the regular layer, φ l is the output value of the lth layer in the RBF layer; The output value of the output layer is: In=[in 1 ,In 2 ,…,In M ] T (5) y=Wv (6) Where W is the weight matrix, w M is the output value of the Mth neuron in the output layer, M is the total number of neurons in the output layer, and T is the transposed matrix; y is the abnormal working condition discrimination matrix, y=[y1,y2], y1 is the predicted value of adenosine triphosphate ATP concentration, y2 is the predicted value of biochemical oxygen demand BOD concentration; v is the output matrix of the rule layer, v=[v1,v2,···,v Q ] T , v Q is the output value of the Qth layer in the regular layer.

6. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 5, characterized in that: The number of neurons arranged in the regular layer is equal to that in the RBF layer.

7. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 6, characterized in that: In the step 2, a multi-input multi-output fuzzy neural network is trained to construct an ATP and BOD prediction model for a sewage treatment process using a Levenberg-Marquardt algorithm based on an adaptive learning rate; The update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate is set as: Θ(t+1)=Θ(t)+(Ψ(t)+λ(t)×I) -1 Ω(t) (7) Where t is time, Θ(t+1) is the variable vector at time t+1, Θ(t) is the variable vector at time t, Ψ(t) is the quasi-Hessian matrix at time t, λ(t) is the adaptive learning rate at time t, 0<λ(t)<1, I is the identity matrix, and Ω(t) is the gradient vector at time t; In the update rule of the Levenberg-Marquardt algorithm based on adaptive learning rate, the calculation formula of the adaptive learning rate λ(t) is: λ(t)=μ(t)λ(t-1) (8) Where μ(t) is the adaptive factor at time t, λ(t-1) is the adaptive learning rate at time t-1; τ min (t) is the minimum eigenvalue of the quasi-Hessian matrix at time t, τ max (t) is the maximum eigenvalue of the quasi-Hessian matrix at time t, 0<τ min (t)<τ max (t); The variable vector Θ(t) includes a connection weight vector w(t), a center vector c(t) and a width vector σ(t). The calculation formula of the variable vector is: Θ(t)=[w1(t),...,w J (t),c1(t),...,c J (t),...,σ1(t),...,σ J (t)] (10) In the formula, w J (t) is the value of the J-th connection weight vector at time t, c J (t) is the J-th center vector value at time t, σ J (t) is the J-th width vector value at time t; The calculation formula of the quasi-Hessian matrix Ψ(t) and the gradient vector Ω(t) is: Ψ(t)=j T (t)j(t) (10) Ω(t)=j T (t)e(t) (11) e(t)=y d (t)-y(t) (12) in, The Jacobian vector j(t) is: Where j(t) is the Jacobian vector at time t, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, and y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

8. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 1, characterized in that: The step 3 specifically includes the following steps: Step 3.1, set the training accuracy value; Step 3.2, inputting the water quality parameter samples of the chemical wastewater in the training set into the ATP and BOD prediction model of the wastewater treatment process constructed in step 2, using the ATP and BOD prediction model of the wastewater treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtaining the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater; Step 3.3, comparing the predicted values ​​of adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration value and biochemical oxygen demand BOD concentration value in the chemical sewage water quality parameter sample, and calculating the accuracy value of the ATP and BOD prediction model of the sewage treatment process. If the accuracy value of the ATP and BOD prediction model of the sewage treatment process is less than the preset accuracy value, proceed to step 3.4; otherwise, proceed to step 3.5; Step 3.4, based on the Levenberg-Marquardt algorithm, the adaptive learning rate of the ATP and BOD prediction model of the sewage treatment process is updated to obtain the updated ATP and BOD prediction model of the sewage treatment process, and the chemical sewage water quality parameter samples are continuously randomly selected in the training set and input into the updated ATP and BOD prediction model of the sewage treatment process, and the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical sewage water quality parameter samples are predicted by using the updated ATP and BOD prediction model of the sewage treatment process, and after obtaining the predicted values ​​of the adenosine triphosphate ATP concentration and the biochemical oxygen demand BOD concentration of the chemical sewage water quality parameter samples, return to step 3.3; Step 3.5, complete the training of the ATP and BOD prediction model for the sewage treatment process, and obtain the trained ATP and BOD prediction model for the sewage treatment process.

9. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 1, characterized in that: The step 4 specifically includes the following steps: Step 4.1, inputting the water quality parameter samples of the chemical wastewater in the test set into the trained ATP and BOD prediction model of the wastewater treatment process, using the ATP and BOD prediction model of the wastewater treatment process to predict the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater, and obtaining the predicted values ​​of the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration of the chemical wastewater; Step 4.2, compare the predicted adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration predicted by the ATP and BOD prediction model of the sewage treatment process with the adenosine triphosphate ATP concentration and biochemical oxygen demand BOD concentration in the chemical sewage water quality parameter sample, and calculate the accuracy of the ATP and BOD prediction model of the sewage treatment process. If the accuracy of the ATP and BOD prediction model of the sewage treatment process is not less than the preset accuracy, proceed to step 5; otherwise, return to step 3 and continue to use the training set to train the ATP and BOD prediction model of the sewage treatment process.

10. The method for evaluating biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 8 or 9, characterized in that: The error value of the ATP and BOD prediction model for the sewage treatment process is: The accuracy of the ATP and BOD prediction model for sewage treatment process is: in, e(t)=y d (t)-y(t) (16) Where RMSE(t) is the error value of the ATP and BOD prediction model of the sewage treatment process, t is time, t = 1, 2, ..., N, N is the total time, e(t) is the error value of the multi-input multi-output fuzzy neural network at time t, y d (t) is the predicted value of the multi-input multi-output fuzzy neural network at time t, and y(t) is the output variable value of the chemical wastewater quality parameter sample at time t.

11. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 1, characterized in that: In step 5, the abnormal operating condition discrimination matrix is: In the formula, ATP is the adenosine triphosphate ATP concentration value predicted by the ATP and BOD prediction model of the sewage treatment process, BOD is the biochemical oxygen demand BOD concentration value predicted by the ATP and BOD prediction model of the sewage treatment process; C1 is the first type of abnormal operating condition, C2 is the second type of abnormal operating condition, C3 is the third type of abnormal operating condition, and C4 is the fourth type of abnormal operating condition.

12. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 11, characterized in that: In step 5, based on the abnormal operating condition discrimination matrix, when the abnormal operating condition type is discriminated according to the adenosine triphosphate ATP concentration value and the biochemical oxygen demand BOD concentration value: When both adenosine triphosphate ATP and biochemical oxygen demand BOD are high, the current operation condition of the chemical wastewater treatment plant is determined to be the first abnormal condition type C1, the influent concentration of the chemical wastewater treatment plant exceeds the treatment capacity, the BOD concentration is abnormal, and the biochemical treatment capacity is high; When adenosine triphosphate ATP is expressed as high and biochemical oxygen demand BOD is expressed as low, it is determined that the current operating condition of the chemical wastewater treatment plant is the second abnormal condition type C2, and the biochemical treatment capacity of the chemical wastewater treatment plant is surplus; When the adenosine triphosphate (ATP) is low and the biochemical oxygen demand (BOD) is high, the current operation condition of the chemical wastewater treatment plant is determined to be the third abnormal condition type C3, and the sludge in the current chemical wastewater treatment plant is determined to be impacted by water flow. It is necessary to measure the adenosine triphosphate (ATP) in the sludge, and analyze the abnormal condition in combination with the adenosine triphosphate (ATP) concentration value of the sludge; When both adenosine triphosphate ATP and biochemical oxygen demand BOD are expressed as low, it is determined that the current operating condition of the chemical wastewater treatment plant is the fourth abnormal condition type C4, and the biochemical treatment capacity of the chemical wastewater treatment plant is low.

13. The method for evaluating the biochemical potential of abnormal conditions in a chemical wastewater treatment process according to claim 11, characterized in that: When the operating condition of the chemical wastewater treatment plant is determined to be the fourth abnormal condition type C4, the sludge disposal capacity of the chemical wastewater treatment plant is improved by reducing the wastewater treatment volume of the chemical wastewater treatment plant, increasing the aeration volume or supplementing the carbon source.

Images