Multi-objective optimization control method for wastewater treatment process based on knowledge selection

Abstract

The invention relates to the technical field of chemical or physical variable control, and discloses a wastewater treatment process multi-objective optimization control method based on knowledge selection. According to the invention, a series of on-line detection devices are installed in the to-be-treated wastewater storage tank, and monitoring of substance data in wastewater in the wastewater storage tank is realized through cooperation of the on-line detection devices; meanwhile, on the basis of collection of different waste water data by each online detection device, the collected waste water data is preprocessed, so that the redundancy of the data is reduced; the feature extraction of the wastewater data is realized by inputting the processed data into the Elman neural network for processing, and the processing precision of the Elman neural network is improved by optimizing the Elman neural network through the non-dominated genetic algorithm. Meanwhile, the wastewater is treated and adjusted according to a result obtained after Elman neural network treatment, and the precision and efficiency of the wastewater treatment process are improved.

Description

Technical Field

[0001] The invention relates to the technical field of control of chemical or physical variables, and in particular to a multi-objective optimization control method for a wastewater treatment process based on knowledge selection. Background Art

[0002] The urban wastewater treatment process is a typical energy-intensive industry. With the increasing scale of urban wastewater treatment in my country and the continuous improvement of wastewater treatment efficiency, its operating energy consumption is also increasing.

[0003] The prior art CN103597418B uses a nonlinear model of the pH of the material to be controlled in the process, uses standard orthogonal basis functions and ordinal spline basis functions to monitor the pH of the material and controls the pH of the material by adding neutralizing substances. However, this technology has great limitations and is insufficient to meet actual needs.

[0004] In addition, the urban wastewater treatment process passively accepts the water quality and quantity of the influent, the operating environment, and is also affected by weather changes, temperature, precipitation, and pH, which results in the treatment process showing dynamic nonlinear characteristics. Therefore, urban wastewater treatment is an operation process with strong nonlinearity, dynamic uncertainty, and strong coupling characteristics. And due to the lack of effective optimization control strategies, it is difficult for the urban wastewater treatment process to obtain satisfactory operating results. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the shortcomings of the prior art, the present invention provides a multi-objective optimization control method for wastewater treatment process based on knowledge selection, which has the advantages of high efficiency and stability, and solves the problem that it is difficult to obtain satisfactory operating results in urban wastewater treatment process due to the lack of effective optimization control strategies.

[0007] (II) Technical solution

[0008] In order to solve the above technical problem that it is difficult to obtain satisfactory operation effect in the urban wastewater treatment process due to the lack of effective optimization control strategy, the present invention provides the following technical solutions:

[0009] This embodiment discloses a multi-objective optimization control method for a wastewater treatment process based on knowledge selection, which specifically includes the following steps:

[0010] S1. Collect data on major substances in wastewater in real time by installing a series of detection equipment in the wastewater storage tank;

[0011] S2, based on installing a series of detection equipment in the wastewater storage tank, collecting the classified wastewater data in real time and performing data preprocessing to obtain preprocessed wastewater data;

[0012] S3, optimizing the initial weights and thresholds of the Elman neural network by a non-dominated genetic algorithm to obtain an optimized Elman neural network; processing the pre-treated wastewater data by the optimized Elman neural network to obtain wastewater data processed by the Elman neural network, the wastewater data including content data of each component in the wastewater;

[0013] S4, analyzing the wastewater data after the Elman neural network treatment in combination with the weights and the set thresholds;

[0014] S5. Based on the comparison result, the wastewater is directed to a corresponding wastewater treatment center for treatment.

[0015] Preferably, the real-time collection of data on major substances in wastewater by installing a series of detection equipment in the wastewater storage tank includes:

[0016] The installed series of testing equipment includes: online pH meter, online dissolved oxygen meter, online component detector;

[0017] The online pH meter is used to monitor the pH value in the wastewater in real time;

[0018] The online component detector is used to monitor the concentration changes of various components in wastewater in real time;

[0019] The online dissolved oxygen meter is used to monitor the concentration of dissolved oxygen in wastewater in real time;

[0020] Based on the concentration of dissolved oxygen in the wastewater measured by an online dissolved oxygen meter, the collected wastewater is differentiated by setting the concentration threshold of dissolved oxygen in the wastewater;

[0021] Set the dissolved oxygen concentration under aerobic conditions to >2.0mg / L, under anaerobic conditions to <0.2mg / L, under anoxic conditions to 0.2-0.5mg / L, and under normal conditions to 0.5-2.0mg / L;

[0022] The wastewater is initially classified based on anaerobic, anoxic and aerobic conditions.

[0023] Preferably, the method of installing a series of detection equipment in the wastewater storage tank to collect the classified wastewater data in real time and perform data preprocessing to obtain the preprocessed wastewater data includes:

[0024] The wastewater data collected and classified in real time is stored in a matrix form;

[0025] According to formula 1, we can get the auxiliary variable data sample matrix X m×n The mean and variance of :

[0026]

[0027] According to formula 2, the sample matrix is ​​normalized to zero mean, and the normalized matrix is ​​calculated;

[0028]

[0029] Among them, m is the number of samples, n is the sample weight, and x ij represents the jth component of the i-th sample; X j is the mean of the i-th sample component, S j represents the standard deviation of the i-th sample;

[0030] The standardized matrix Z is calculated according to formula 3 and formula 4 m×n The covariance matrix R m×n ;

[0031]

[0032] Among them, r represents covariance, T represents matrix transpose;

[0033] According to formula 5, the different eigenvalues ​​λ of R are solved j (j=1,2,...,n), and arrange the n eigenvalues ​​of R in descending order;

[0034]

[0035] Among them, λJ represents the different eigenvalues ​​λ j The matrix composed of, R represents the covariance matrix;

[0036] The obtained result is calculated by formula 6 to obtain the unit eigenvector b corresponding to the corresponding eigenvalue j (j=1,2,...,n), b j =(b 1j , b 2j , ..., b nj );

[0037] Rb=λ J b (6)

[0038] Calculate the cumulative variance contribution rate of the principal component based on the calculated eigenvalues, and determine the number of principal components k according to the cumulative variance contribution rate;

[0039]

[0040] According to formula 8, the normalized matrix Z m×n Projection on k-dimensional coordinates:

[0041]

[0042] Among them, U 1 is the first principal component, U 2 is the second principal component, U k is the kth principal component.

[0043] Preferably, the method of optimizing the initial weights and thresholds of the Elman neural network by a non-dominated genetic algorithm to obtain the optimized Elman neural network specifically comprises the following steps:

[0044] The weighted objective function uses the following formula:

[0045]

[0046] Among them, w max , w min are the maximum and minimum values ​​of the weight, respectively, max =0.9, w min =0.4; k is the number of iterations, k max is the maximum number of iterations;

[0047] By setting the initial basic parameters of the Elman neural network and encoding the weights and thresholds of the Elman neural network;

[0048] S31, initialize the population, set basic parameters such as the maximum number of iterations, population size, crossover probability, mutation probability, and set appropriate constraints on input and output variables;

[0049] S32, perform fast non-dominated sorting on the population and calculate the congestion degree, and enter the algorithm iteration process;

[0050] S33, performing a selection operation according to the non-dominated layer ranking and crowding degree of the individuals, individuals with the same non-dominated ranking are selected with smaller crowding degree; individuals with different non-dominated rankings are selected with higher crowding degree;

[0051] S34, crossover and mutation operations;

[0052] S35, merging populations, iterating and screening, merging the parent and child populations in step S34 to generate a new merged population, using the method described in step S33 to screen and generate a new population, and stopping when the maximum number of iterations is reached, and the algorithm terminates.

[0053] Preferably, the fast non-dominated sorting specifically comprises the following steps:

[0054] S321, traverse all solutions in the set, calculate the corresponding n and S sets respectively, n p Indicates the number of dominant solutions p, which is a numerical value; S prepresents the solutions dominated by solution p, which is a set;

[0055] S322. Find the solution of n=0 and divide it into set F 1 , and assign the corresponding non-dominated layer ranking value to 1;

[0056] S323, traverse set F 1 For each individual j in the set, find the solution set dominated by individual j as S j , and S j Each individual h in n corresponds to h Subtract 1;

[0057] S324, find S j The corresponding n h The solution equal to 0 is divided into the set F 2 , and assign the corresponding non-dominated layer ranking value to 2, and repeat steps S322-S324 to complete the non-dominated sorting operation of all individuals in the set.

[0058] Preferably, the congestion degree calculation method is as follows:

[0059] Set the population size to N, initialize the crowding distance of each individual in the population, and set n d =0, n∈1,2,...,N;

[0060] Set each objective function f m Perform fast non-dominated sorting and set f m max is the individual objective function f m The maximum value, f m min is the individual objective function f m The minimum value of

[0061] The crowding degree of the solution on the boundary of each objective function is set to infinity, as follows:

[0062] I(d 1 )=∞,I(d n )=∞ (10)

[0063] The crowding degree of the remaining individuals is:

[0064]

[0065] Among them, I(d 1 ) represents the solution of the first individual of the objective function, I represents the crowding degree, and k represents the kth individual.

[0066] Preferably, the binary crossover process is as follows:

[0067]

[0068] Among them, p i，k For its parent individual, c i，k is the offspring individual produced by its crossover, i,k represents the kth individual of the i-th generation, β k ≥0 indicates the individual that performs the crossover operation, and its probability density function is:

[0069]

[0070] Among them, η c represents the cth offspring individual, and β represents the crossover operation;

[0071] The process of generating offspring by polynomial mutation operation is generated by polynomial mutation operator. The specific formula is as follows:

[0072]

[0073] In which, set u∈[0,1], c k represents the offspring individuals produced by mutation, p k represents its parent individual, δ k It is expressed by the following formula:

[0074]

[0075]

[0076] Among them, r k represents the non-dominated ordering of individuals, η m is the variation distribution index.

[0077] Preferably, the pre-treated wastewater data is processed by the optimized Elman neural network to obtain

[0078] The wastewater data after Elman neural network processing includes:

[0079] The structure of the Elman neural network includes input layer, hidden layer, receiving layer and output layer;

[0080] Furthermore, the nonlinear state expression of the structure of the Elman neural network is as follows:

[0081] x(t)=f(w 1 x c (t)+w 2 u(t-1)) (19)

[0082] x c (t) = x(t-1) (20)

[0083] yt =g(w 3 x(t)) (21)

[0084] Here, r, n, and m are set to represent the number of nodes in the input layer, hidden layer, and output layer, respectively. The number of nodes in the receiving layer is the same as the number of nodes in the hidden layer, which is n, x(t), x c (t), y(t) represent the outputs of the hidden layer, the receiving layer and the output layer at time t respectively; w 1 represents the connection weight from the receiving layer to the hidden layer; w 2 represents the connection weight from the input layer to the hidden layer; w 3 represents the connection weight from the hidden layer to the output layer; g() represents the transfer function of the output layer; f() represents the activation function of the hidden layer, and u represents the input vector of the input layer;

[0085] Furthermore, the activation function of the hidden layer is the Sigmoid function:

[0086]

[0087] g() takes a linear function:

[0088] y t =w 3 x(t) (23)

[0089] Furthermore, the learning algorithm of the Elman neural network adopts the gradient descent algorithm, and its error indicator function expression is as follows:

[0090]

[0091] Among them, y d (q) represents the actual output of the Elman neural network at the qth step, and y(q) represents the input of the Elman neural network at the qth step.

[0092] Preferably, the analysis of the wastewater data processed by the Elman neural network in combination with weights and set thresholds includes:

[0093] According to the comparison results after combining the weights and the set thresholds, the first n principal components with cumulative variance contribution rates greater than 85% are selected, and the first n principal components are sorted according to the contribution rates, and the main components in the wastewater are processed in sequence based on the sorting;

[0094] After the treatment is completed, the components with a cumulative variance contribution rate of not less than 15% are sorted according to the contribution rate and the wastewater with the same ranking of components is mixed.

[0095] Preferably, directing the wastewater to a corresponding wastewater treatment center for treatment based on the comparison result includes:

[0096] After the first n principal components in the wastewater with cumulative variance contribution rates greater than 85% are processed, the treated wastewater is subjected to the operations of S2 to S5 repeatedly until the wastewater treatment standard is reached.

[0097] (III) Beneficial effects

[0098] Compared with the prior art, the present invention provides a multi-objective optimization control method for wastewater treatment process based on knowledge selection, which has the following beneficial effects:

[0099] 1. The present invention monitors the data of each component in the wastewater in real time by installing a series of online detection equipment. First, based on the online dissolved oxygen meter, the collected wastewater is distinguished by setting the concentration threshold of dissolved oxygen in the wastewater to determine whether the collected wastewater is aerobic or anaerobic; at the same time, the wastewater is transferred to the corresponding environment according to the judgment standard of the online dissolved oxygen meter, and then the changes of each component in the wastewater are detected in real time by the online detection equipment; the collected wastewater is distinguished and transferred to reduce the influence of environmental parameters on the wastewater components, and the accuracy of the real-time monitoring equipment is improved. The present invention designs a wastewater treatment process optimization control strategy, and realizing efficient and stable operation of the treatment process is a challenging problem currently faced by urban wastewater treatment plants. Strive to improve the operation effect of the treatment process and enhance the competitiveness of the wastewater treatment industry by studying the optimization control theory and technology of urban wastewater treatment process. The present invention distinguishes wastewater from multiple angles, and improves the safety and effectiveness of wastewater treatment by measuring the content of each component in the wastewater and making corresponding treatments.

[0100] 2. The present invention processes the data collected by the online detection equipment, stores the data collected by the online detection equipment by constructing a matrix, solves the variance of the constructed matrix to obtain the eigenvalue of the matrix, calculates the cumulative variance contribution rate of the principal components based on the eigenvalue, determines the number of principal components according to the cumulative variance contribution rate, determines the order of treatment in the wastewater according to the contribution of the principal components, and ensures the effectiveness and integrity of the wastewater treatment process.

[0101] 3. The present invention transmits the processed wastewater matrix data to the Elman neural network, and further processes the wastewater data through the input layer, hidden layer, receiving layer and output layer in the Elman neural network and by setting weights and thresholds, thereby reducing the error of wastewater treatment.

[0102] 4. The present invention optimizes the initial weights and thresholds of the Elman neural network by using a non-dominated genetic algorithm, and transforms the multi-objective optimization problem into a problem of optimizing a fitness function by cross-mutation and iteration of the input weight and threshold encoding of the Elman neural network, thereby reducing the complexity of the algorithm and improving the running speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 It is a schematic diagram of the multi-objective optimization process structure of the wastewater treatment process based on knowledge selection of the present invention;

[0104] Figure 2 The present invention is a schematic diagram of the process structure of optimizing the initial weights and thresholds of the Elman neural network based on the non-dominated genetic algorithm. DETAILED DESCRIPTION

[0105] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0106] This embodiment discloses a multi-objective optimization control method for a wastewater treatment process based on knowledge selection, which specifically includes the following steps:

[0107] S1. Install a series of detection equipment in the wastewater storage tank to collect real-time data on the main substances in the wastewater and conduct preliminary classification of the wastewater;

[0108] The installed series of testing equipment includes: online pH meter, online dissolved oxygen meter, online component detector;

[0109] The online pH meter is used to monitor the pH value in the wastewater in real time;

[0110] The online component detector is used to monitor the concentration changes of various components in wastewater in real time;

[0111] The online dissolved oxygen meter is used to monitor the concentration of dissolved oxygen in wastewater in real time;

[0112] Based on the concentration of dissolved oxygen in the wastewater measured by an online dissolved oxygen meter, the collected wastewater is differentiated by setting the concentration threshold of dissolved oxygen in the wastewater;

[0113] Set the dissolved oxygen concentration under aerobic conditions to >2.0mg / L, under anaerobic conditions to <0.2mg / L, under anoxic conditions to 0.2-0.5mg / L, and under normal conditions to 0.5-2.0mg / L;

[0114] Initial classification of wastewater based on anaerobic, anoxic and aerobic conditions;

[0115] S2, based on installing a series of detection equipment in the wastewater storage tank to collect the classified wastewater data in real time and perform data preprocessing to obtain preprocessed wastewater data, the wastewater data including the content data of each component in the wastewater;

[0116] The pre-treated wastewater data is stored in a matrix form;

[0117] According to formula 1, we can get the auxiliary variable data sample matrix X m×n The mean and variance of :

[0118]

[0119] Furthermore, the sample matrix is ​​subjected to zero-mean normalization according to Formula 2, and a normalized matrix is ​​calculated;

[0120]

[0121] Among them, m is the number of samples, n is the sample weight, and x ij represents the jth component of the i-th sample; X j is the mean of the i-th sample component, S j represents the standard deviation of the i-th sample;

[0122] Further, the standardized matrix Z is calculated according to Formula 3 and Formula 4 m×n The covariance matrix R m×n ;

[0123]

[0124] Among them, r represents covariance, T represents matrix transpose;

[0125] Further, according to Formula 5, different eigenvalues ​​λ of R are solved j (j=1,2,...,n), and arrange the n eigenvalues ​​of R in descending order;

[0126] R-λ J E|=0 (5)

[0127] Among them, λJ represents the different eigenvalues ​​λ j The matrix composed of, R represents the covariance matrix;

[0128] Furthermore, the obtained result is calculated by formula 6 to obtain the unit eigenvector b corresponding to the corresponding eigenvalue j (j=1,2,...,n), b j =(b 1j , b 2j , ..., b nj );

[0129] Rb=λ J b (6)

[0130] Further, the cumulative variance contribution rate of the principal component is calculated based on the calculated eigenvalue, and the number k of the principal components is determined according to the cumulative variance contribution rate;

[0131]

[0132] Further, according to Formula 8, the normalized matrix Z m×n Projection on k-dimensional coordinates:

[0133] U i,j =Z i T b j ,i=1,2...,m,j=1,2...,k (8)

[0134] Among them, U 1 is the first principal component, U 2 is the second principal component, U k is the kth principal component;

[0135] Furthermore, the principal components in the wastewater are arranged in descending order according to their cumulative variance contribution rates, and the components with the highest cumulative variance contribution rates in the wastewater are processed first;

[0136] S3, optimizing the initial weights and thresholds of the Elman neural network by a non-dominated genetic algorithm to obtain an optimized Elman neural network; processing the pre-treated wastewater data by the optimized Elman neural network to obtain wastewater data processed by the Elman neural network;

[0137] Furthermore, the initial weights and thresholds of the Elman neural network are optimized by a genetic algorithm, which specifically includes the following steps:

[0138] The weighted objective function uses the following formula:

[0139]

[0140] Among them, w max , w min are the maximum and minimum values ​​of the weight, respectively, max =0.9, w min =0.4; k is the number of iterations, k max is the maximum number of iterations;

[0141] Furthermore, by setting the initial basic parameters of the Elman neural network, and encoding the weights and thresholds of the Elman neural network;

[0142] S31, initialize the population, set basic parameters such as the maximum number of iterations, population size, crossover probability, mutation probability, and set appropriate constraints on input and output variables;

[0143] S32, perform fast non-dominated sorting on the population and calculate the congestion degree, and enter the algorithm iteration process;

[0144] The fast non-dominated sort specifically includes the following steps:

[0145] S321, traverse all solutions in the set, calculate the corresponding n and S sets respectively, n p Indicates the number of dominant solutions p, which is a numerical value; S p represents the solutions dominated by solution p, which is a set;

[0146] S322. Find the solution of n=0 and divide it into set F 1 , and assign the corresponding non-dominated layer ranking value to 1;

[0147] S323, traverse set F 1 For each individual j in the set, find the solution set dominated by individual j as S j , and S j Each individual h in n corresponds to h Subtract 1;

[0148] S324, find S j The corresponding n h The solution equal to 0 is divided into the set F 2 , and assign the corresponding non-dominated layer ranking value to 2, and repeat steps S422-S424 to complete the non-dominated sorting operation of all individuals in the set;

[0149] S33, performing a selection operation according to the non-dominated layer ranking and crowding degree of the individuals, individuals with the same non-dominated ranking are selected with a smaller crowding degree; individuals with different non-dominated rankings are selected with a higher crowding degree;

[0150] The congestion calculation method is as follows:

[0151] Set the population size to N, initialize the crowding distance of each individual in the population, and set n d =0, n∈1,2,...,N;

[0152] Set each objective function f m Perform fast non-dominated sorting and set f m max is the individual objective function f m The maximum value, f m min is the individual objective function fm The minimum value of

[0153] The crowding degree of the solution on the boundary of each objective function is set to infinity, as follows:

[0154] I(d 1 )=∞,I(d n )=∞ (10)

[0155] The crowding degree of the remaining individuals is:

[0156]

[0157] Among them, I(d 1 ) represents the solution of the first individual of the objective function, I represents the crowding degree, and k represents the kth individual;

[0158] S34, crossover and mutation operations;

[0159] Recombination and mutation of data by simulating binary crossover and polynomial mutation;

[0160] Further, the binary crossover process is as follows:

[0161]

[0162] Among them, p i，k For its parent individual, c i，k is the offspring individual produced by its crossover, i,k represents the kth individual of the i-th generation, β k ≥0 indicates the individual that performs the crossover operation, and its probability density function is:

[0163]

[0164] Among them, η c represents the cth offspring individual, and β represents the crossover operation;

[0165] Furthermore, the process of generating offspring by polynomial mutation operation is generated by polynomial mutation operator, and the specific formula is as follows:

[0166]

[0167] In which, set u∈[0,1], c k represents the offspring individuals produced by mutation, p k represents its parent individual, δ k It is expressed by the following formula:

[0168]

[0169] Among them, r k represents the non-dominated ordering of individuals, ηm is the mutation distribution index;

[0170] S35. Merge the populations, iterate, and screen. Merge the parental and offspring populations in step S34 to generate a new merged population, and use the method described in step S33 for screening to generate a new population. Stop when the maximum number of iterations is reached, and the algorithm terminates;

[0171] Furthermore, the optimized Elman neural network is used to process the wastewater data;

[0172] The structure of the Elman neural network includes an input layer, a hidden layer, a context layer, and an output layer;

[0173] Furthermore, the non-linear state expression of the structure of the Elman neural network is as follows:

[0174] x(t) = f(w 1 x c (t) + w 2 u(t - 1)) (19)

[0175] x c (t) = x(t - 1) (20)

[0176] y t = g(w 3 x(t)) (21)

[0177] Among them, it is set that r, n, and m respectively represent the number of nodes in the input layer, the hidden layer, and the output layer. The number of nodes in the context layer is the same as that in the hidden layer, which is n. x(t), x c (t), y t respectively represent the outputs of the hidden layer, the context layer, and the output layer at time t; w 1 represents the connection weight from the context layer to the hidden layer; w 2 represents the connection weight from the input layer to the hidden layer; w 3 represents the connection weight from the hidden layer to the output layer; g() represents the transfer function of the output layer; f() represents the activation function of the hidden layer, and u represents the input vector of the input layer;

[0178] Furthermore, the activation function of the hidden layer takes the Sigmoid function:

[0179]

[0180] g() takes a linear function:

[0181] y t = w 3 x(t) (23)

[0182] Furthermore, the learning algorithm of the Elman neural network adopts the gradient descent algorithm, and its error indicator function expression is as follows:

[0183]

[0184] Among them, y d (q) represents the actual output of the Elman neural network at step q, and y(q) represents the input of the Elman neural network at step q;

[0185] S4, analyzing the wastewater data after the Elman neural network treatment in combination with the weights and the set thresholds;

[0186] According to the comparison results after combining the weights and the set thresholds, the first n principal components with cumulative variance contribution rates greater than 85% are selected, and the first n principal components are sorted according to the contribution rates, and the main components in the wastewater are processed in sequence based on the sorting;

[0187] After the treatment is completed, the components with a cumulative variance contribution rate of not less than 15% are sorted according to the contribution rate and the wastewater with the same ranking of components is mixed;

[0188] S5. Based on the comparison result, the wastewater is directed to a corresponding wastewater treatment center for treatment.

[0189] Furthermore, after the first n principal components in the wastewater whose cumulative variance contribution rate is greater than 85% are processed, the processed wastewater is repeatedly subjected to the operations of S2 to S5 until the wastewater treatment standard is reached.

[0190] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-objective optimization control method for wastewater treatment process based on knowledge selection, characterized in that: The following steps are involved: S1. Collect data on major substances in wastewater in real time by installing a series of detection equipment in the wastewater storage tank; S2, based on installing a series of detection equipment in the wastewater storage tank to collect the classified wastewater data in real time and perform data preprocessing to obtain preprocessed wastewater data, the wastewater data including the content data of each component in the wastewater; S3, optimizing the initial weights and thresholds of the Elman neural network by a non-dominated genetic algorithm to obtain an optimized Elman neural network; processing the pre-treated wastewater data by the optimized Elman neural network to obtain wastewater data processed by the Elman neural network; S4, analyzing the wastewater data after the Elman neural network treatment in combination with the weights and the set thresholds; S5. Based on the comparison result, the wastewater is directed to a corresponding wastewater treatment center for treatment.

2. According to claim 1, a multi-objective optimization control method for wastewater treatment process based on knowledge selection is characterized in that: The real-time collection of data on major substances in wastewater by installing a series of detection equipment in the wastewater storage tank includes: The installed series of testing equipment includes: online pH meter, online dissolved oxygen meter, online component detector; The online pH meter is used to monitor the pH value in the wastewater in real time; The online component detector is used to monitor the concentration changes of various components in wastewater in real time; The online dissolved oxygen meter is used to monitor the concentration of dissolved oxygen in wastewater in real time; Based on the concentration of dissolved oxygen in the wastewater measured by an online dissolved oxygen meter, the collected wastewater is differentiated by setting the concentration threshold of dissolved oxygen in the wastewater; Set the dissolved oxygen concentration under aerobic conditions to >2.0mg / L, under anaerobic conditions to <0.2mg / L, under anoxic conditions to 0.2-0.5mg / L, and under normal conditions to 0.5-2.0mg / L; The wastewater is initially classified based on anaerobic, anoxic and aerobic conditions.

3. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 1 is characterized in that: The wastewater data after classification is collected in real time and preprocessed based on a series of detection equipment installed in the wastewater storage tank, and the preprocessed wastewater data includes: The wastewater data collected and classified in real time is stored in a matrix form; According to formula 1, we can get the auxiliary variable data sample matrix X m×n The mean and variance of : According to formula 2, the sample matrix is ​​normalized to zero mean, and the normalized matrix is ​​calculated; Among them, m is the number of samples, n is the sample weight, and x ij represents the jth component of the i-th sample; is the mean of the i-th sample component, S j represents the standard deviation of the i-th sample; The standardized matrix Z is calculated according to formula 3 and formula 4 m×n The covariance matrix R m×n ; Among them, r represents covariance, T represents matrix transpose; According to formula 5, the different eigenvalues ​​λ of R are solved j (j=1,2,...,n), and arrange the n eigenvalues ​​of R in descending order; |R-λ J E|=0 (5) Among them, λJ represents the different eigenvalues ​​λ j The matrix composed of, R represents the covariance matrix; The obtained result is calculated by formula 6 to obtain the unit eigenvector b corresponding to the corresponding eigenvalue j (j=1,2,...,n), b j =(b 1j , b 2j , ..., b nj ); Rb=λ J b (6) Calculate the cumulative variance contribution rate of the principal component based on the calculated eigenvalues, and determine the number of principal components k according to the cumulative variance contribution rate; According to formula 8, the normalized matrix Z m×n Projection on k-dimensional coordinates: Among them, U1 is the first principal component, U2 is the second principal component, and U k is the kth principal component.

4. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 1 is characterized in that: The method of optimizing the initial weights and thresholds of the Elman neural network by a non-dominated genetic algorithm to obtain the optimized Elman neural network specifically includes the following steps: The weighted objective function uses the following formula: Among them, w max , w min are the maximum and minimum values ​​of the weight, respectively, max =0.9, w min =0.4; k is the number of iterations, k max is the maximum number of iterations; By setting the initial basic parameters of the Elman neural network and encoding the weights and thresholds of the Elman neural network; S31, initialize the population, set basic parameters such as the maximum number of iterations, population size, crossover probability, mutation probability, and set appropriate constraints on input and output variables; S32, quickly sort the population and calculate the congestion degree, and enter the algorithm iteration process; S33, performing a selection operation according to the non-dominated layer ranking and crowding degree of the individuals, individuals with the same non-dominated ranking are selected with a smaller crowding degree; individuals with different non-dominated rankings are selected with a higher crowding degree; S34, crossover and mutation operations; S35, merging populations, iterating and screening, merging the parent and child populations in step S34 to generate a new merged population, using the method described in step S33 to screen and generate a new population, and stopping when the maximum number of iterations is reached, and the algorithm terminates.

5. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 4 is characterized in that: The fast non-dominated sorting specifically comprises the following steps: S321, traverse all solutions in the set, calculate the corresponding n and S sets respectively, n p Indicates the number of dominant solutions p, which is a numerical value; S p represents the solutions dominated by solution p, which is a set; S322, find the solution of n=0, divide it into set F1, and assign the corresponding non-dominated layer ranking value to 1; S323, traverse each individual j in the set F1, and find the solution set dominated by individual j as S j , and S j Each individual h in n corresponds to h Subtract 1; S324, find S j The corresponding n h The solution equal to 0 is divided into the set F2, and its corresponding non-dominated layer ranking is assigned a value of 2. Repeating steps S322-S324 can complete the non-dominated sorting operation for all individuals in the set.

6. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 4 is characterized in that: The congestion degree calculation method is as follows: Set the population size to N, initialize the crowding distance of each individual in the population, and set n d =0, n∈1,2,...,N; Set each objective function f m Perform fast non-dominated sorting and set f m max is the individual objective function f m The maximum value, f m min is the individual objective function f m The minimum value of The crowding degree of the solution on the boundary of each objective function is set to infinity, as follows: I(d1)=∞,I(d n )=∞ (10) The crowding degree of the remaining individuals is: Among them, I(d1) represents the solution of the first individual of the objective function, I represents the crowding degree, and k represents the kth individual.

7. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 4 is characterized in that: The binary crossover process is as follows: Among them, p i，k For its parent individual, c i，k is the offspring individual produced by its crossover, i,k represents the kth individual of the i-th generation, β k ≥0 indicates the individual that performs the crossover operation, and its probability density function is: Among them, η c represents the cth offspring individual, and β represents the crossover operation; The process of generating offspring by polynomial mutation operation is generated by polynomial mutation operator. The specific formula is as follows: In which, set u∈[0,1], c k represents the offspring individuals produced by mutation, p k represents its parent individual, δ k It is expressed by the following formula: Among them, r k represents the non-dominated ordering of individuals, η m is the variation distribution index.

8. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 1 is characterized in that: The wastewater data after pretreatment is processed by the optimized Elman neural network to obtain the wastewater data after Elman neural network processing, including: The structure of the Elman neural network includes input layer, hidden layer, receiving layer and output layer; The nonlinear state expression of the structure of the Elman neural network is as follows: x(t)=f(w1x c (t)+w2u(t-1)) (19) x c (t)=x(t-1) (20) y t =g(w3x(t)) (21) Here, r, n, and m are set to represent the number of nodes in the input layer, hidden layer, and output layer, respectively. The number of nodes in the receiving layer is the same as the number of nodes in the hidden layer, which is n, x(t), x c (t), y t Respectively represent the output of the hidden layer, the receiving layer and the output layer at time t; w1 represents the connection weight from the receiving layer to the hidden layer; w2 represents the connection weight from the input layer to the hidden layer; w3 represents the connection weight from the hidden layer to the output layer; g() represents the transfer function of the output layer; f() represents the activation function of the hidden layer, and u represents the input vector of the input layer; The activation function of the hidden layer is the Sigmoid function: g() takes a linear function: y t =w3x(t) (23) The learning algorithm of the Elman neural network adopts the gradient descent algorithm, and its error indicator function expression is as follows: Among them, y d (q) represents the actual output of the Elman neural network at the qth step, and y(q) represents the input of the Elman neural network at the qth step.

9. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 1 is characterized in that: The analysis of the wastewater data processed by the Elman neural network combined with the weights and the set thresholds includes: According to the comparison results after combining the weights and the set thresholds, the first n principal components with cumulative variance contribution rates greater than 85% are selected, and the first n principal components are sorted according to the contribution rates, and the main components in the wastewater are processed in sequence based on the sorting; After the treatment is completed, the components with a cumulative variance contribution rate of not less than 15% are sorted according to the contribution rate and the wastewater with the same ranking of components is mixed.

10. The multi-objective optimization control method for wastewater treatment process based on knowledge selection according to claim 1 is characterized in that: The directing the wastewater to the corresponding wastewater treatment center for treatment based on the comparison result includes: After the first n principal components in the wastewater with cumulative variance contribution rates greater than 85% are processed, the treated wastewater is subjected to the operations of S2 to S5 repeatedly until the wastewater treatment standard is reached.

Images