Soft sensing method of effluent ammonia nitrogen based on ADw-CLPSO radial basis function affective neural network

Abstract

The present invention proposes a soft measurement method for effluent ammonia nitrogen based on ADw-CLPSO radial basis emotion neural network, which belongs to the field of sewage treatment. The main design process of this method is as follows: first, a nonlinear function is constructed to improve the CLPSO algorithm, so that the inertia weight of each particle can be adaptively and dynamically adjusted, and then based on the improved ADw-CLPSO algorithm, a particle variable dimension learning mechanism is used to simultaneously realize parameter updating and structural adjustment in the radial basis emotion network training process. The soft measurement method for effluent ammonia nitrogen based on ADw-CLPSO radial basis emotion neural network composed of the above steps belongs to the protection scope of the present invention. The present invention uses the ADw-CLPSO algorithm to simultaneously optimize the parameters and structure of the radial basis emotion network, which can improve the prediction accuracy of effluent ammonia nitrogen and improve the generalization performance of the model.

Description

Technical field:

[0001] The present invention optimizes the affective neural network by using an improved comprehensive learning particle swarm algorithm, and then establishes a soft measurement model of the effluent ammonia nitrogen concentration in the process of urban sewage treatment, which solves the problem that the ammonia nitrogen concentration in the sewage treatment process is difficult to measure; the effluent ammonia nitrogen concentration is an important indicator for measuring the treatment effect of the sewage treatment process, and achieving accurate measurement is an important basis for reducing the emission of ammonia nitrogen concentration, and is also an important guarantee for achieving water quality discharge standards. The present invention belongs to both the field of water treatment and the field of detection technology. Background technology:

[0002] With the development of urbanization and industrialization in my country, the demand for water in industrial production and urban life is increasing, resulting in an increase in wastewater discharge year by year, which has caused serious water shortage and environmental pollution problems, and has a great impact on human survival and development as well as the ecological balance of the environment. One of the key indicators to measure whether wastewater meets the discharge standards is NH 4 -N concentration, high concentration of NH 4 -N discharge will lead to eutrophication of rivers, lakes and other water bodies, seriously endangering the entire aquatic ecosystem. 4 Accurate prediction of -N concentration is of great significance for ensuring that effluent water quality meets discharge standards.

[0003] Soft measurement technology combines some knowledge and experience in the industrial production process with advanced computer technology, uses some parameters that are easier to measure as indirect measurement variables, and variables that are difficult to detect or difficult to measure, such as ammonia nitrogen concentration, as important variables to be measured, and establishes a certain mathematical model between the two to perform estimation or prediction. In recent years, with the vigorous development of neural network technology and the limitations of traditional methods for determining sewage quality indicators, soft measurement technology has been widely used in the online measurement of water quality parameters in urban sewage treatment processes. The present invention designs a soft measurement method for effluent ammonia nitrogen concentration based on ADw-CLPSO radial basis affective neural network, and realizes the simultaneous adjustment of model parameters and structure through an improved comprehensive learning particle swarm optimization algorithm, which effectively improves the generalization performance of the model in predicting effluent ammonia nitrogen concentration. Summary of the invention:

[0004] A soft measurement method for effluent NH4-N concentration based on ADw-CLPSO emotional neural network has the following main operation process: first, a nonlinear function is constructed to improve the CLPSO algorithm so that the inertia weight of each particle can be adaptively and dynamically adjusted. Then, based on the improved ADw-CLPSO algorithm, a particle variable dimension learning mechanism is used to simultaneously realize parameter updates and structural adjustments during the training process of the radial basis emotional network. This method utilizes the advantages of low cost, strong versatility, and high prediction accuracy of neural networks. Through an evolutionary algorithm, the parameters and structure of the model are optimized and adjusted at the same time, solving the problem of difficulty in real-time measurement of effluent ammonia nitrogen concentration in the process of urban sewage treatment, and improving the level of real-time monitoring of water quality parameters in the wastewater treatment process of sewage treatment plants. It is characterized by comprising the following steps:

[0005] Step 1: Variable selection

[0006] The soft sensor model is based on the outlet NH 4 -N concentration was used as the output variable, with effluent redox potential ORP, dissolved oxygen DO, total suspended solids TSS, effluent nitrate nitrogen NO 3 -N and pH values ​​were used as input variables.

[0007] Step 2: Design the ADw-CLPSO radial basis function neural network topology for effluent ammonia nitrogen concentration prediction

[0008] Determine the topological structure of the radial basis emotion neural network as a 5-p-1 connection mode, that is, the number of input layer neurons is 5, the number of hidden layer neurons is p, p is a positive integer, and the output neuron is 1; assign values ​​to the parameters of the neural network; suppose the neural network input at time t is z(t)=[z 1 (t),z 2 (t),z 3 (t),z 4 (t),z 5 (t)], the actual output is expressed as E(t), and the actual output of the neural network is expressed as:

[0009]

[0010] Among them, ν i (i=1,…,p) and w i (i=1,…,p) are the connection weights of the i-th neuron in the hidden layers “amygdala” and “orbitofrontal cortex” of the radial basis emotion neural network, p is the number of neurons in the hidden layer; b is the bias; is any Gaussian function and u is its corresponding connection weight; is the input of the ith neuron in the hidden layer “thalamus” of the radial basis emotion neural network, and νi The update methods are:

[0011]

[0012] Among them, μ i and σ i represents the center vector and width value of the i-th neuron in the hidden layer "thalamus", σ i >0;||z(t)-μ i || is z(t) and μ i The Euclidean distance between i (t+1) represents the connection weight of the i-th node at time t+1; α is the learning rate of the neurons in the hidden layer “amygdala” of the radial basis emotion neural network;

[0013] Step 3: Train the Neural Network

[0014] Step 3.1: Initialize the parameters of the comprehensive learning particle swarm algorithm

[0015] Determine the initial number of iterations t = 0, the maximum number of iterations max_t = 1000; input vector z l represents the lth group of input samples, represents the expected target value of the lth group of samples, Indicates that the input sample dimension is n, T is the total number of samples; the population is divided into 10 groups to represent the radial basis emotion neural network with 3-12 hidden layer neurons, and the number of particles in each group is 15; the particle acceleration constant c is initialized to a random number (0,1); the initial inertia weight of the particle ω 0 and the minimum inertia weight ω 1 are set to 0.95 and 0.55 respectively; the parameters of the radial basis emotion neural network are represented as particles in the comprehensive learning particle swarm:

[0016]

[0017] Among them, x j represents the position of the jth particle, j = 1, 2, ..., S; S is the total number of particles, S is a positive integer, b is the deviation, w j,p , and σ p are respectively represented as the connection weight, center vector and width value of the neuron in the hidden layer “orbital frontal cortex” of the p-th radial basis emotion neural network in the j-th particle; b, w j,p , and σ p The initial value of is a random number of (-3,3); the connection weights ν of the neurons in the hidden layers "amygdala" and "thalamus" of the radial basis emotion neural network are iThe initial values ​​of and u are random numbers (0,1); the learning rate α for weight update in the "amygdala" neurons in the hidden layer of the radial basis emotion neural network and the search range of the balance factor δ in the particle fitness function are set to [0.1, 0.12, ..., 0.28, 0.3] and [0.001, 0.002, ..., 0.01] respectively, and the optimal parameter combination of α and δ that minimizes the training error is selected by the grid search method; the iteration number flag β for the particle fitness value to be recalculated and the iteration number flag γ for the global optimal position gbest of the population that has not been improved are set to 4; at the same time, the speed of each particle is initialized:

[0018]

[0019] Among them, v j represents the velocity of the jth particle, represents the velocity of the jth particle in the Djth dimension and initializes it to a random number (-2,2). j Indicates the dimension of the jth particle, D j =(2+n)p j +1, n represents the number of input variables, p j represents the number of neurons in the hidden layer of the radial basis emotion neural network represented by the jth particle;

[0020] Step 3.2: Design particle fitness function

[0021] For the input z(t) of the radial basis emotion neural network, determine the dimension D of each particle j =(2+n)p j +1, calculate the fitness value of each particle:

[0022] f(x j (t)) = R j (t)+δD j (t) (6)

[0023] Among them, δ is the balance factor in the particle fitness function, R j (t) is:

[0024]

[0025] Among them, E j (t) and They represent the network output represented by the jth particle and the expected target value at time t; T represents the number of training samples input to the neural network; the initial position x of each particle is j Initialized to the individual optimal position of the particle in is the optimal position of the jth particle in dimension d, f j(d) represents the optimal position of the jth particle, the particle number to be learned in the dth dimension, which is randomly initialized to different particle numbers in the group to which the particle belongs; after calculating the fitness value of each particle, the individual optimal position with the minimum fitness value is taken as the global optimal position gbest of the population;

[0026] Step 3.3: Determine whether the particle dimension has changed

[0027] Record the number of iterations without improvement of the global optimal position gbest of the population. When its value is greater than γ, calculate the average fitness value of each group of particles:

[0028]

[0029] Among them, avg(·) is the function for calculating the average value; Group M represents the Mth group of particles in the population; fit j represents the fitness value of the jth particle; m 1 is the first particle of the Mth group; S M is the number of particles in the Mth group; find the best neural network structure and the corresponding particle dimension D according to the global optimal position gbest best , the optimal number of neurons in the hidden layer of the neural network is (D best -1) / (2+n), n is the number of input variables; according to D best Update the dimension of each particle corresponding to the number of neurons in the hidden layer of the neural network:

[0030]

[0031] Among them, Δd represents the unit length of the particle parameters to be adjusted when the number of neurons in the hidden layer of the neural network changes. In this example, Δd = 2 + n. Then the number of iterations of the recorded global optimal position gbest of the population that has not been improved is cleared, and the inertia weight of each particle at time t + 1 is updated:

[0032]

[0033] in,

[0034]

[0035] ω j (t+1) is the inertia weight of the jth particle at time t+1, ω 0 and ω 1 are the initial inertia weight and minimum inertia weight of the particle respectively; fit(x j(t)) is the fitness value of the jth particle; gbest is the global optimal position of the population; max_t is the maximum number of iterations; tanh{·} is the hyperbolic tangent function; exp{·} is the exponential function;

[0036] Step 3.4: Update particle parameters

[0037] When the individual optimal position of the jth particle stops improving β times, the optimal position of the jth particle is recalculated at the particle number f to be learned in the dth dimension j (d); Then update the position and velocity of each particle:

[0038]

[0039] Among them, v j,d (t+1) and x j,d (t+1) represents the velocity and position of the jth particle in the dth dimension at time t+1; ω j (t) is the inertia weight of the jth particle at time t; r j,d It means that the jth particle takes a random number between [0,1] in the dth dimension; c is the acceleration factor of the particle;

[0040] Step 3.5: Determine whether the training is finished

[0041] Input training sample data z(t+1), repeat steps 3.2-3.4, and stop the calculation if t>max_t. The parameters represented by the global optimal position gbest of the population are used as the optimal network parameters, and then according to D best Determine the optimal network structure size p best =(D best -1) / (2+n), p best The number of hidden layer neurons corresponding to the optimal network structure; execute step 4;

[0042] Step 4: Test the Neural Network

[0043] The test sample data is used as the input of the trained radial basis emotion network based on ADw-CLPSO to test the radial basis emotion neural network. The output of the radial basis emotion neural network is the outflow NH 4 - The predicted value of N.

[0044] The creativity of the present invention is mainly reflected in:

[0045] 1. In view of the current problems of high cost, cumbersome process, inconvenient measurement and low precision in measuring effluent ammonia nitrogen concentration in urban sewage treatment plants, the present invention proposes a soft measurement method for effluent ammonia nitrogen concentration in sewage treatment process. According to the relevant mechanism and experience of sewage treatment process, five easily measurable variables related to effluent ammonia nitrogen concentration are selected in the actual sewage treatment plant work report: effluent redox potential ORP, aerobic terminal dissolved oxygen DO, total solid suspended solids TSS, effluent nitrate nitrogen NO3-N and pH value. The connection between easily measurable variables and effluent ammonia nitrogen concentration is established by adopting radial basis function affective neural network, so as to realize the prediction of effluent ammonia nitrogen concentration. The method has the characteristics of low cost, strong versatility and high prediction accuracy.

[0046] 2. The present invention adopts a nonlinear function to adaptively and dynamically adjust the inertia weight of each particle, so that the particles in the population can be adjusted according to time and diversity at the same time, effectively balancing the global and local search capabilities of the particles; in addition, based on the improved comprehensive learning particle swarm algorithm, a particle variable dimension learning mechanism is designed, which can adjust the dimensions of all particles. Without affecting the update of particle speed and position, it not only expands the search space of particles, but also enables particles to learn from particles of different dimensions. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a structural model diagram of the ADw-CLPSO radial basis emotion neural network of the present invention;

[0048] Figure 2 It is a graph showing the RMSE changes of the ADw-CLPSO radial basis emotion neural network during the training process of the present invention;

[0049] Figure 3 This is a diagram showing the prediction result of the outlet NH4-N concentration using the ADw-CLPSO radial basis affective neural network of the present invention;

[0050] Figure 4 It is the prediction error diagram of the outlet NH4-N concentration of the ADw-CLPSO radial basis affective neural network of the present invention; Specific implementation method:

[0051] The experimental data of the present invention comes from the water quality data report of a sewage treatment plant in August 2014. After removing abnormal experimental samples, 498 sets of available data remain, of which 332 sets of data are used as training data and the remaining 166 sets are used as test data;

[0052] A soft measurement method for effluent NH4-N concentration based on ADw-CLPSO emotional neural network has the following main operation process: first, a nonlinear function is constructed to improve the CLPSO algorithm so that the inertia weight of each particle can be adaptively and dynamically adjusted. Then, based on the improved ADw-CLPSO algorithm, a particle variable dimension learning mechanism is used to simultaneously realize parameter updates and structural adjustments during the training process of the radial basis emotional network. This method utilizes the advantages of low cost, strong versatility, and high prediction accuracy of neural networks. Through an evolutionary algorithm, the parameters and structure of the model are optimized and adjusted at the same time, solving the problem of difficulty in real-time measurement of effluent ammonia nitrogen concentration in the process of urban sewage treatment, and improving the level of real-time monitoring of water quality parameters in the wastewater treatment process of sewage treatment plants. It is characterized by comprising the following steps:

[0053] Step 1: Variable selection

[0054] The soft sensor model is based on the outlet NH 4 -N concentration was used as the output variable, with effluent redox potential ORP, dissolved oxygen DO, total suspended solids TSS, effluent nitrate nitrogen NO 3 -N and pH values ​​were used as input variables.

[0055] Step 2: Design the ADw-CLPSO radial basis function neural network topology for effluent ammonia nitrogen concentration prediction

[0056] Determine the topological structure of the radial basis emotion neural network as a 5-p-1 connection mode, that is, the number of input layer neurons is 5, the number of hidden layer neurons is p, p is a positive integer, and the output neuron is 1; assign values ​​to the parameters of the neural network; suppose the neural network input at time t is z(t)=[z 1 (t),z 2 (t),z 3 (t),z 4 (t),z 5 (t)], the actual output is expressed as E(t), and the actual output of the neural network is expressed as:

[0057]

[0058] Among them, ν i (i=1,…,p) and w i (i=1,...,p) are the connection weights of the i-th neuron in the hidden layers "amygdala" and "orbitofrontal cortex" of the radial basis emotion neural network, p is the number of neurons in the hidden layer; b is the bias; is any Gaussian function and u is its corresponding connection weight; is the input of the ith neuron in the hidden layer “thalamus” of the radial basis emotion neural network, and νi The update methods are:

[0059]

[0060] Among them, μ i and σ i represents the center vector and width value of the i-th neuron in the hidden layer "thalamus", σ i >0;||z(t)-μ i || is z(t) and μ i The Euclidean distance between i (t+1) represents the connection weight of the i-th node at time t+1; α is the learning rate of the neurons in the hidden layer “amygdala” of the radial basis emotion neural network;

[0061] Step 3: Train the Neural Network

[0062] Step 3.1: Initialize the parameters of the comprehensive learning particle swarm algorithm

[0063] Determine the initial number of iterations t = 0, the maximum number of iterations max_t = 1000; input vector z l represents the lth group of input samples, represents the expected target value of the lth group of samples, Indicates that the input sample dimension is n, T is the total number of samples; the population is divided into 10 groups to represent the radial basis emotion neural network with 3-12 hidden layer neurons, and the number of particles in each group is 15; the particle acceleration constant c is initialized to a random number (0,1); the initial inertia weight of the particle ω 0 and the minimum inertia weight ω 1 are set to 0.95 and 0.55 respectively; the parameters of the radial basis emotion neural network are represented as particles in the comprehensive learning particle swarm:

[0064]

[0065] Among them, x j represents the position of the jth particle, j = 1, 2, ..., S; S is the total number of particles, S is a positive integer, b is the deviation, w j,p , and σ p are respectively represented as the connection weight, center vector and width value of the neuron in the hidden layer “orbital frontal cortex” of the p-th radial basis emotion neural network in the j-th particle; b, w j,p , and σ p The initial value of is a random number of (-3,3); the connection weights ν of the neurons in the hidden layers "amygdala" and "thalamus" of the radial basis emotion neural network are iThe initial values ​​of and u are random numbers (0,1); the learning rate α for weight update in the "amygdala" neurons in the hidden layer of the radial basis emotion neural network and the search range of the balance factor δ in the particle fitness function are set to [0.1, 0.12, ..., 0.28, 0.3] and [0.001, 0.002, ..., 0.01] respectively, and the optimal parameter combination of α and δ that minimizes the training error is selected by the grid search method; the iteration number flag β for the particle fitness value to be recalculated and the iteration number flag γ for the global optimal position gbest of the population that has not been improved are set to 4; at the same time, the speed of each particle is initialized:

[0066]

[0067] Among them, v j represents the velocity of the jth particle, Indicates that the jth particle is in the D j dimensional speed and initialize it to a random number of (-2,2), D j Indicates the dimension of the jth particle, D j =(2+n)p j +1, n represents the number of input variables, p j represents the number of neurons in the hidden layer of the radial basis emotion neural network represented by the jth particle;

[0068] Step 3.2: Design particle fitness function

[0069] For the input z(t) of the radial basis emotion neural network, determine the dimension D of each particle j =(2+n)p j +1, calculate the fitness value of each particle:

[0070] f(x j (t)) = R j (t)+δD j (t) (6)

[0071] Among them, δ is the balance factor in the particle fitness function, R j (t) is:

[0072]

[0073] Among them, E j (t) and They represent the network output represented by the jth particle and the expected target value at time t; T represents the number of training samples input to the neural network; the initial position x of each particle is j Initialized to the individual optimal position of the particle in is the optimal position of the jth particle in dimension d, fj (d) represents the optimal position of the jth particle, the particle number to be learned in the dth dimension, which is randomly initialized to different particle numbers in the group to which the particle belongs; after calculating the fitness value of each particle, the individual optimal position with the minimum fitness value is taken as the global optimal position gbest of the population;

[0074] Step 3.3: Determine whether the particle dimension has changed

[0075] Record the number of iterations without improvement of the global optimal position gbest of the population. When its value is greater than γ, calculate the average fitness value of each group of particles:

[0076]

[0077] Among them, avg(·) is the function for calculating the average value; Group M represents the Mth group of particles in the population; fit j represents the fitness value of the jth particle; m 1 is the first particle of the Mth group; S M is the number of particles in the Mth group; find the best neural network structure and the corresponding particle dimension D according to the global optimal position gbest best , the optimal number of neurons in the hidden layer of the neural network is (D best -1) / (2+n), n is the number of input variables; according to D best Update the dimension of each particle corresponding to the number of neurons in the hidden layer of the neural network:

[0078]

[0079] Among them, Δd represents the unit length of the particle parameters to be adjusted when the number of neurons in the hidden layer of the neural network changes. In this example, Δd = 2 + n. Then the number of iterations of the recorded global optimal position gbest of the population that has not been improved is cleared, and the inertia weight of each particle at time t + 1 is updated:

[0080]

[0081] in,

[0082]

[0083] ω j (t+1) is the inertia weight of the jth particle at time t+1, ω 0 and ω 1 are the initial inertia weight and minimum inertia weight of the particle respectively; fit(x j(t)) is the fitness value of the jth particle; gbest is the global optimal position of the population; max_t is the maximum number of iterations; tanh{·} is the hyperbolic tangent function; exp{·} is the exponential function;

[0084] Step 3.4: Update particle parameters

[0085] When the individual optimal position of the jth particle stops improving β times, the optimal position of the jth particle is recalculated at the particle number f to be learned in the dth dimension j (d); Then update the position and velocity of each particle:

[0086]

[0087] Among them, v j,d (t+1) and x j,d (t+1) represents the velocity and position of the jth particle in the dth dimension at time t+1; ω j (t) is the inertia weight of the jth particle at time t; r j,d It means that the jth particle takes a random number between [0,1] in the dth dimension; c is the acceleration factor of the particle;

[0088] Step 3.5: Determine whether the training is finished

[0089] Input training sample data z(t+1), repeat steps 3.2-3.4, and stop the calculation if t>max_t. The parameters represented by the global optimal position gbest of the population are used as the optimal network parameters, and then according to D best Determine the optimal network structure size p best =(D best -1) / (2+n), p best The number of hidden layer neurons corresponding to the optimal network structure; execute step 4;

[0090] Step 4: Test the Neural Network

[0091] The test sample data is used as the input of the trained radial basis emotion network based on ADw-CLPSO to test the radial basis emotion neural network. The output of the radial basis emotion neural network is the outflow NH 4 -N predicted value; the predicted result is as follows Figure 3 As shown, the X-axis is the number of samples, the unit is / sample, the Y-axis is the effluent ammonia nitrogen concentration, the unit is mg / l; the solid line is the actual effluent ammonia nitrogen concentration value, and the dotted line is the tested effluent ammonia nitrogen concentration value; the error between the actual output and the tested output of the effluent ammonia nitrogen concentration is as follows Figure 4As shown, X-axis: number of samples, the unit is / sample, Y-axis: prediction error of effluent ammonia nitrogen concentration, the unit is mg / l; the results show the effectiveness of the soft measurement method of effluent ammonia nitrogen concentration based on ADw-CLPSO radial basis affective neural network.

[0092] Table 1-12 is the experimental data of the present invention, and Table 1-5 is the training input sample: effluent oxidation-reduction potential ORP, aerobic terminal dissolved oxygen DO, total suspended solids TSS, effluent nitrate nitrogen NO 3 -N, pH value, Table 6 is the effluent ammonia nitrogen concentration of the training sample, and Tables 7-11 are the test input samples: effluent redox potential ORP, aerobic terminal dissolved oxygen DO, total suspended solids TSS, effluent nitrate nitrogen NO 3 -N, pH value, Table 12 shows the ammonia nitrogen concentration of the test sample effluent.

[0093] Training samples:

[0094] Table 1. Outlet water oxidation-reduction potential ORP (mV)

[0095] -5.25557 -5.51194 -5.38376 -8.78065 -18.3945 -13.5876 -16.5358 -17.1767 -43.9674 -63.0669 -82.4869 -95.3694 -104.214 -80.6923 -73.8344 -71.1425 -103.573 -115.494 -121.647 -121.199 -123.826 -86.9733 -56.209 -49.2229 -48.0052 -49.6716 -46.2106 -36.4045 -40.0577 -18.4586 -12.498 -6.08877 -21.5991 -11.2803 -22.24 -51.4662 -62.3619 -60.8877 -38.7118 -22.881 -53.3889 -12.498 -3.07643 3.58917 7.94745 9.03702 10.7675 14.3567 18.074 16.664 19.42 19.7404 22.3682 21.3427 18.715 17.5613 18.9713 19.9327 18.6509 27.3674 35.6994 32.1744 29.9952 29.9952 28.9056 31.7898 34.6099 35.0585 37.6863 36.084 32.1744 17.1767 21.0223 15.8949 33.1357 35.9558 39.6091 39.545 37.6863 35.8917 31.0848 26.4701 39.2886 43.7751 46.5951 46.0824 44.9287 46.8515 46.7874 36.0199 -43.2623 -71.8475 -88.8961 -99.279 -108.637 -117.994 -126.839 -135.427 -140.939 -144.849 -147.989 -151.899 -157.283 -160.872 -170.165 -176.703 -181.894 -187.086 -191.764 -194.2 -195.418 -196.187 -196.892 -196.571 -196.635 -199.071 -201.891 -202.211 -202.788 -202.596 -202.788 -202.275 -203.685 -206.441 -206.377 -203.621 -196.443 -190.803 -184.137 -191.38 -187.919 -183.56 -172.857 -163.051 -161.32 -159.269 -159.59 -161.449 -163.628 -171.511 -176.19 -170.486 -166.768 -160.936 -161.961 -163.371 -166.191 -169.076 -206.057 -188.944 -174.075 -165.422 -161.769 -157.603 -169.332 -168.883 -177.408 -172.344 -160.551 -155.168 -154.142 -152.86 -151.002 -153.373 -157.859 -145.49 -5.96059 -5.70422 -5.8324 -8.71656 -20.3814 -16.7281 -16.7922 -28.2006 -51.5302 -71.0143 -87.8065 -98.1895 -97.8049 -77.4876 -68.6429 -80.8204 -108.316 -117.994 -123.121 -119.789 -108.444 -76.334 -49.8638 -46.3388 -54.9912 -47.7488 -45.6337 -37.0454 -33.5203 -15.1258 -11.7289 -39.7373 -15.254 -10.2548 -31.6616 -60.6314 -61.3364 -51.5302 -39.1604 -13.3953 -26.0215 -6.98607 0.064092 4.80693 8.13973 9.80613 10.4471 16.3435 19.6123 18.6509 19.42 21.2787 22.5605 20.5736 18.2663 17.3049 18.9713 19.5482 31.0207 33.1998 36.5967 31.9821 30.7002 29.9952 30.4439 34.3535 34.8662 36.2763 37.1095 35.379 24.2269 16.7281 20.0609 28.7775 34.0971 36.9172 38.7759 37.5581 37.4299 34.4817 30.7643 33.328 40.4423 45.2492 46.467 44.4801 45.3133 47.4283 20.7018 41.2114 -55.9526 -78.8977 -94.2158 -102.42 -112.098 -121.07 -129.21 -137.094 -142.67 -146.515 -149.015 -153.629 -159.077 -163.243 -172.793 -178.433 -183.689 -189.072 -192.854 -194.456 -194.456 -196.956 -196.635 -196.699 -196.635 -200.929 -202.211 -201.891 -202.596 -202.34 -202.852 -202.019 -205.096 -206.441 -206.249 -200.994 -194.2 -188.175 -190.226 -190.547 -186.573 -181.894 -168.627 -162.346 -161.064 -158.564 -160.423 -161.961 -163.435 -177.408 -174.78 -170.614 -163.692 -159.975 -162.602 -164.525 -166.896 -169.396 -201.25 -182.599 -169.781 -164.012 -161.385 -156.001 -170.165 -168.05 -175.741 -158.436 -154.078 -155.744 -152.668 -151.514 -149.784 -156.385 -156.898 -145.297

[0096] Table 2. Aerobic terminal dissolved oxygen DO (mg / l)

[0097] 8.52762 8.59854 8.88176 8.90028 9.20711 8.41457 6.35349 1.84109 1.33207 1.77015 2.40575 2.80003 6.01548 5.91001 5.21696 2.71465 2.59337 3.49289 4.21911 5.5913 8.08898 8.26919 7.63804 6.34748 6.56533 6.82726 7.17285 7.84229 8.87397 8.89269 8.34662 7.35793 8.15413 4.07505 1.86917 4.74672 6.21795 7.74542 9.24433 9.08078 9.51889 9.03586 9.16216 9.72706 10.0582 10.2456 10.527 10.4677 10.2115 10.2056 10.4626 10.8273 10.9962 11.1889 11.6194 11.7616 11.6829 11.8039 11.9374 11.8659 11.9167 8.64676 8.68995 9.25259 9.08549 8.75846 8.91371 9.20022 9.23477 8.79301 9.11265 9.45223 9.5065 9.29965 8.88472 9.04671 9.10067 7.66515 6.09292 6.91959 6.78133 7.62736 8.92776 9.30695 9.58502 9.84497 10.1788 10.3498 10.2875 10.5158 0.808065 0.885007 1.13525 0.648437 0.557665 0.545776 0.488974 0.476872 0.504474 0.540478 0.623293 0.487017 0.522913 0.460929 0.477037 0.482013 0.487936 0.555641 0.572849 0.642597 0.696887 0.586238 0.58138 0.570417 0.606726 0.594872 0.594128 0.547162 0.546108 0.559366 0.585199 0.616554 0.604091 0.560834 2.28828 2.16928 2.1948 2.28619 2.3381 2.55315 2.54024 2.69733 3.41357 3.0879 2.10112 1.96656 2.12771 2.35488 2.38278 2.25447 3.38496 3.48655 3.00395 2.69771 2.62355 2.65284 2.86211 2.79264 3.81622 3.7762 2.47223 2.46525 2.41805 2.63179 2.75857 2.35787 5.78746 5.10408 3.45974 3.623 3.59257 3.67113 3.72357 3.84129 4.77296 3.4398 8.61093 8.71496 8.93187 9.27302 9.04182 8.13102 4.3447 1.5381 1.41326 1.85691 2.76679 3.75526 6.11381 6.1211 4.01761 2.30058 2.7663 3.77435 4.16404 7.48972 7.92836 8.33863 6.91479 6.35385 6.75088 6.87662 7.11462 8.68387 8.78709 8.9123 7.95337 7.95647 7.80432 2.72321 2.10447 5.59238 6.54324 8.82017 9.15986 9.2334 9.62311 8.76673 9.2969 9.81291 10.1406 10.4813 10.5433 10.533 10.1643 10.2589 10.5427 10.8675 11.0707 11.2329 11.672 11.8179 11.7435 11.8878 11.9611 11.8697 11.9997 8.57565 8.93198 9.23704 9.06932 8.72929 9.18021 9.19924 9.23114 8.68224 9.24038 9.45579 9.41315 9.20348 8.76343 9.13291 9.03339 6.54774 6.40309 6.84672 7.04976 7.77474 9.14049 9.34214 9.69204 9.94944 10.1941 10.4058 10.649 10.249 0.837108 0.862799 0.727423 0.68332 0.498502 0.54109 0.477107 0.51313 0.531765 0.535511 0.529054 0.501028 0.45047 0.456457 0.512686 0.488905 0.492448 0.557767 0.62251 0.706762 0.563767 0.581366 0.580635 0.59152 0.565674 0.584325 0.537045 0.504404 0.528814 0.545476 0.576573 0.599285 0.584718 0.585541 2.19963 2.10604 2.36697 2.31623 2.53411 2.63226 2.57705 2.4567 3.31724 3.11042 1.77372 1.94847 2.11053 2.33669 2.33635 2.10643 3.40355 3.56867 2.58079 2.69009 2.573 2.79341 2.85264 2.68034 3.88912 3.55606 2.57127 2.32021 2.5298 2.68567 2.78709 2.13847 5.62294 4.94967 3.62185 3.45491 3.57296 3.6627 3.81267 3.88487 1.25936 3.26749

[0098] Table 3. Total suspended solids TSS (mg / l)

[0099] 2.77245 2.81856 2.82026 2.85089 2.87901 2.78576 2.71757 2.80939 2.7666 2.78225 2.80153 2.76862 2.78831 2.79749 2.85228 2.89848 2.78127 3.17272 2.85361 2.81794 2.79689 2.8099 2.78565 2.87864 2.82253 2.81981 2.76836 2.81652 2.88917 2.85715 2.8119 2.83368 2.80557 2.7277 2.78632 2.73849 2.78147 2.86509 2.9005 2.83804 2.81119 2.99607 2.96987 3.02386 3.01379 3.01508 3.13113 3.00128 3.06402 3.0442 3.1004 3.10693 3.08447 3.09122 3.18922 3.15562 3.22783 3.15001 3.15272 3.0467 3.0319 2.96511 2.94046 2.91218 2.88758 2.87673 2.88092 2.94799 2.96565 2.91944 2.94933 2.92544 2.93288 2.94226 2.95853 2.9781 2.95376 2.86317 2.8895 2.88783 2.91509 2.92752 2.96835 2.97292 2.97797 3.00519 3.00589 3.02333 2.97573 2.99794 2.81638 2.77051 2.83766 2.79417 2.779 2.77735 2.84319 2.90228 2.83573 2.85428 2.7845 2.78206 2.70666 2.77442 2.71305 2.73063 2.62527 2.60049 2.65745 2.66345 2.58687 2.58294 2.53627 2.52786 2.48968 2.46742 2.49157 2.52651 2.45022 2.53969 2.40822 2.49033 2.39318 2.42397 2.48987 2.49064 2.53398 2.38391 2.43197 2.39927 2.53942 2.51395 2.55194 2.46932 2.42449 2.46054 2.46491 2.39188 2.44715 2.47396 2.48181 2.54498 2.5353 2.38985 2.43465 2.3089 2.39469 2.31659 2.27538 2.27676 2.21999 2.33332 2.41552 2.30834 2.2717 2.4388 2.52506 2.47886 2.39382 2.38287 2.44275 2.42549 2.41873 2.41471 2.44402 2.2835 2.77094 2.81509 2.80315 2.90262 2.82948 2.80532 2.83427 2.81511 2.77868 2.78072 2.75388 2.82977 2.78273 2.80633 2.79684 2.80548 2.90435 2.80294 2.79628 2.89356 2.79626 2.79786 2.81081 2.83784 2.81678 2.8262 2.82377 2.85422 2.84818 2.80191 2.80739 2.82596 2.86147 2.77999 2.8132 2.81285 2.87377 2.82906 2.8997 2.8987 2.9148 3.07878 2.94973 3.02499 3.07278 3.0607 3.05322 3.03581 3.05907 3.07264 3.08567 3.21596 3.11782 3.15561 3.22378 3.08956 3.1621 3.1618 3.0722 3.06209 3.01646 2.91957 2.92929 2.92361 2.88682 2.8967 2.91006 2.93386 2.91807 2.94236 2.92132 2.9271 2.94692 3.03488 2.9285 2.96364 2.92758 2.82335 2.87977 2.94521 2.93242 2.90189 2.96712 2.96101 2.94472 2.9889 3.03294 3.02233 3.05353 2.96964 2.80521 2.8212 2.81403 2.7832 2.73162 2.71054 2.86662 2.882 2.80235 2.81731 2.7886 2.75307 2.7616 2.72193 2.74095 2.69412 2.62509 2.554 2.62198 2.53136 2.57646 2.56112 2.51765 2.49713 2.48307 2.46541 2.45635 2.43671 2.45616 2.40679 2.40307 2.51298 2.43763 2.44394 2.43452 2.49819 2.45145 2.44919 2.47248 2.48145 2.52862 2.44291 2.65716 2.48203 2.5025 2.49115 2.42654 2.44811 2.50448 2.43313 2.48658 2.51128 2.43092 2.36546 2.38826 2.28046 2.30775 2.28243 2.26678 2.2297 2.21049 2.41957 2.29354 2.36712 2.30995 2.38208 2.44907 2.57765 2.44396 2.43175 2.40892 2.47839 2.42536 2.4256 2.32425 2.31203

[0100] Table 4. Nitrate nitrogen NO in effluent 3 -N(mg / l)

[0101] 12.059 12.1463 12.1501 12.1923 12.1974 12.7774 13.0879 13.3726 13.1971 13.0552 13.8076 13.3865 12.5341 12.654 12.8412 12.5841 12.156 11.5991 11.5589 11.7893 12.563 13.0676 13.2064 13.232 13.0679 13.0895 13.2195 13.3849 13.0319 13.7622 13.942 13.8699 13.6342 13.4144 13.5353 13.7483 13.9572 14.0803 13.5746 13.3774 13.0971 12.8939 12.5444 12.5014 12.3031 12.2185 11.8753 11.9293 11.8823 11.7222 11.5492 11.6624 11.6112 11.5873 11.5544 11.4553 11.4642 11.5608 11.3813 12.1984 12.2073 14.5598 14.3669 14.2066 13.9853 14.0755 13.9737 13.8414 13.4937 13.1809 13.1574 13.3626 13.4756 13.3289 13.432 13.3333 13.1321 13.018 12.8622 13.2188 13.7822 13.8588 14.4949 14.3036 14.0134 13.906 13.8762 13.7423 13.5202 13.5613 18.4186 17.2616 16.4661 19.0243 18.6134 16.7459 14.8797 13.4827 12.7146 12.5643 12.8184 13.3616 12.4441 10.7352 8.92979 7.93797 7.3449 6.93887 6.84548 7.04185 7.25048 7.91684 7.7814 7.10975 6.05892 6.01141 5.93726 5.83157 5.62513 5.62171 5.55001 5.51265 5.46269 5.3936 5.47897 6.47781 8.16784 9.13357 10.365 10.5668 11.0416 11.4794 12.4688 13.8564 14.7415 14.6572 14.0722 13.7215 13.7288 13.9157 14.4164 15.2791 16.6379 16.8101 15.7517 13.9968 12.6256 11.6286 10.3934 11.5281 13.3324 12.9103 12.0798 11.673 11.494 11.4878 11.4866 12.6792 13.5262 12.9682 12.0995 12.2122 12.4464 12.8156 13.7367 18.6773 12.1031 12.1459 12.1604 12.1844 12.2427 12.8863 13.1319 13.2904 13.1739 12.9795 13.5526 13.1718 11.9977 12.7766 12.7252 12.4493 12.009 11.4738 11.632 11.8512 12.7666 12.9899 13.2197 13.0438 13.0054 13.1803 13.2947 13.2274 13.4429 13.7635 14.07 13.857 13.5616 13.5019 13.5848 13.6969 13.9687 13.9585 13.5054 13.2852 13.0524 12.7982 12.5541 12.4544 12.2274 12.1176 11.8267 11.9295 11.8219 11.4556 11.5804 11.6595 11.6346 11.5833 11.4462 11.3676 11.5513 11.4945 12.0848 12.1361 12.1086 14.5444 14.2971 14.1216 13.9645 13.9882 13.847 13.7627 13.3879 13.2322 13.1344 13.3567 13.1867 13.622 13.246 13.3233 13.0576 12.9751 12.9629 13.4852 13.8805 14.5043 14.4608 14.1925 13.9697 13.8502 13.837 13.7979 13.3354 13.723 18.0473 16.8753 16.7242 19.4747 18.0206 16.063 14.3586 13.1156 12.6086 12.6417 13.1066 13.2954 11.8426 10.2015 8.52338 7.66003 7.17894 6.83982 6.63673 7.28323 7.73316 7.97914 7.61777 6.72507 5.90655 6.09422 5.9726 5.86168 5.60554 5.60416 5.54285 5.4625 5.38136 5.48383 5.52698 6.96314 8.57786 9.88518 10.4496 10.7179 11.1779 11.7328 12.8207 14.2832 14.9089 14.4612 13.9712 13.6408 13.8617 13.9758 14.4829 15.6909 16.9443 16.5498 15.3732 13.5851 12.2428 11.4642 10.4852 12.2201 13.0934 12.5906 11.9742 11.6093 11.5036 11.3927 11.7895 13.04 13.4614 12.7089 12.1315 12.2197 12.4927 12.9392 14.1627 19.197

[0102] Table 5. pH value

[0103] 7.90676 7.90145 7.90551 7.91514 7.92141 7.91908 7.92984 7.91762 7.89074 7.86406 7.8465 7.84475 7.84977 7.85788 7.86311 7.86548 7.86234 7.85805 7.86187 7.86895 7.87459 7.89027 7.91208 7.92162 7.91333 7.90251 7.90189 7.90436 7.90167 7.9309 7.94203 7.95357 7.89802 7.91755 7.91658 7.89533 7.8922 7.89486 7.89149 7.90709 7.91723 7.89338 7.92315 7.92443 7.93603 7.93612 7.9406 7.94151 7.9492 7.94905 7.95412 7.95569 7.95288 7.96016 7.97648 7.98591 7.99768 8.01316 8.01257 7.99985 8.01138 7.96894 7.96581 7.96508 7.96636 7.97149 7.96847 7.96593 7.98463 7.99041 7.99177 7.98638 7.9923 7.99443 8.00039 8.00352 8.00741 8.02379 8.02331 8.01303 8.00901 7.99788 7.9888 7.98844 8.00364 8.01776 8.03339 8.03776 8.04531 8.04811 8.0184 8.01093 7.99022 8.01332 8.02256 7.99395 8.00113 8.00259 7.99269 7.98731 7.9918 8.00472 8.01007 8.00556 7.99557 7.98785 7.99155 7.99076 7.98643 7.99295 7.99592 7.9983 8.01118 8.00451 7.99675 7.99358 7.98662 8.00303 7.99634 8.0069 7.99924 8.00404 8.00334 8.00153 8.00652 8.00904 8.02639 8.02543 8.03728 8.00213 8.02809 8.0288 8.03195 8.03047 8.04305 8.04802 8.03163 8.01837 8.00909 8.01317 7.99659 8.03047 8.06491 8.06171 8.05423 8.02601 8.01396 8.00972 8.01059 8.03392 8.00946 8.03846 8.04124 8.02792 7.97677 8.01394 8.01642 8.01819 8.02767 8.03141 8.01567 8.0092 8.00973 8.01585 8.01664 8.05845 7.90685 7.9005 7.90689 7.91963 7.91987 7.92 7.92984 7.90865 7.88183 7.85857 7.84446 7.84809 7.85169 7.8622 7.86424 7.86667 7.85901 7.85933 7.86425 7.87017 7.8801 7.89719 7.91555 7.92074 7.90786 7.9038 7.90432 7.9031 7.90954 7.93417 7.94663 7.89312 7.90488 7.91989 7.91103 7.89656 7.89006 7.89377 7.89483 7.91391 7.89664 7.90597 7.92203 7.92793 7.93378 7.93952 7.94049 7.94482 7.94918 7.94795 7.95313 7.95474 7.95535 7.95911 7.98345 7.98944 8.00815 8.01449 7.98986 8.00789 8.01145 7.96902 7.96594 7.96633 7.9697 7.96891 7.96384 7.97065 7.99036 7.99161 7.99072 7.98764 7.99117 7.99986 8.00474 8.00471 8.01443 8.02284 8.01799 8.0099 8.00541 7.99404 7.98742 7.98967 8.00947 8.02255 8.03714 8.04034 8.051 8.04542 8.01631 8.011 7.98447 8.02085 8.01445 7.99522 8.00351 7.99587 7.98823 7.98622 8.00077 8.00331 8.01237 8.0022 7.98998 7.98906 7.99168 7.98628 7.98983 7.99622 7.99814 7.99955 8.01124 8.00001 7.99455 7.98561 7.99331 7.99695 8.00094 8.0036 7.99931 8.00645 8.00348 8.00282 8.00316 8.01367 8.02755 8.03016 8.02248 8.01498 8.02719 8.02776 8.03376 8.0334 8.04298 8.0403 8.02609 8.01511 8.0128 7.99819 8.00553 8.0419 8.06268 8.05954 8.04983 8.01581 8.0107 8.01199 8.01954 8.03141 8.0187 8.03977 8.03679 8.0281 7.98501 8.01956 8.0101 8.0212 8.03342 8.02352 8.01233 8.01054 8.01448 8.01241 8.02085 8.07451

[0104] Table 6. Actual effluent ammonia nitrogen concentration (mg / l)

[0105] 3.39707 3.64586 3.56599 3.64233 3.8495 3.73635 3.69221 3.31471 3.37535 3.45851 3.56336 3.67627 3.88143 3.67141 3.7936 3.55853 3.58018 3.71775 3.86844 3.88295 3.8895 3.95768 3.64604 3.77203 3.62062 3.62593 3.68265 3.59831 4.00847 3.632 3.63171 3.58988 3.64583 3.57891 3.612 3.59196 3.61834 3.55199 4.7089 4.449 4.74333 3.5797 3.27963 3.26195 3.14021 3.12424 3.16312 3.59865 3.44005 3.45476 3.5401 3.56502 3.43161 3.55956 3.60204 3.62751 3.64366 3.67739 3.7676 3.63875 3.77204 3.54391 3.61005 3.53152 3.70908 3.62655 3.56697 3.54 3.50345 3.54561 3.53485 3.57594 3.67361 4.00177 3.78297 3.71883 3.65594 3.69507 3.708 3.75552 3.72829 3.87959 3.72742 3.5312 3.74153 3.71005 3.84581 3.73685 3.75309 4.02848 3.8758 4.84293 5.43412 5.77607 5.89499 6.86429 7.65311 8.09222 8.74486 8.87837 9.40336 9.50249 9.94376 10.4556 11.041 11.5016 11.9091 12.087 12.4395 12.4108 12.2645 12.2824 12.3406 12.3668 12.5197 12.6702 12.7935 13.0679 12.878 12.9323 12.9189 13.1193 13.2119 13.1942 13.1843 13.0278 12.5932 12.0214 11.5033 11.1842 10.8915 10.6223 10.0941 9.39166 8.78833 8.52802 8.2748 8.30942 8.18425 8.25038 7.73171 7.37416 6.70381 6.33792 6.32991 6.69471 7.2973 8.11158 8.73827 8.76631 8.83532 9.09669 9.31398 9.10526 9.2865 9.28159 9.21246 8.82795 8.3717 7.65524 9.26747 9.22301 9.36836 9.21729 8.84451 7.11035 3.40234 3.56254 3.80196 3.81045 3.79802 3.74218 3.57606 3.30477 3.41698 3.56789 3.53923 3.85715 3.93417 3.59257 3.70194 3.5754 3.5805 3.72103 3.93938 3.92056 3.86121 3.92259 3.65128 3.71527 3.60187 3.64318 3.62188 3.73903 3.67425 3.65985 3.61747 3.55207 3.59922 3.56815 3.58464 3.62717 3.58884 3.53299 4.77 4.8729 4.04573 3.3696 3.23038 3.23724 3.2063 3.1321 3.18438 3.54827 3.53307 3.5154 3.52233 3.52879 3.55252 3.57894 3.63768 3.61822 3.66097 3.68515 3.73862 3.73218 3.72438 3.55246 3.53322 3.53075 3.73523 3.58937 3.52667 3.64342 3.56072 3.56609 3.45601 3.75822 4.0753 3.72148 3.73234 3.72755 3.69726 3.69504 3.76009 3.76848 3.82179 3.81898 3.49943 3.60834 3.7012 3.67054 3.77596 3.77014 3.99825 3.86241 4.27487 5.18656 5.83202 5.70956 6.27114 7.08371 7.82732 8.29905 8.89435 9.32641 9.40705 9.55395 10.2738 10.782 11.2465 11.53 11.9231 12.1773 12.3372 12.4116 12.4893 12.4335 12.3238 12.5816 12.5137 12.9055 12.8395 13.1354 12.9153 12.9308 13.0146 13.1046 13.0794 13.1832 13.2108 12.8992 12.4099 11.7775 11.3341 11.09 10.781 10.6037 9.95439 9.17682 8.59132 8.42544 8.25713 8.25214 8.19111 8.04273 7.67843 7.1995 6.51722 6.30161 6.37036 6.79368 7.61177 8.3032 8.78245 8.742 8.78927 9.5518 9.21787 9.12662 9.26212 9.20211 9.06548 8.61861 8.271 7.52266 9.31758 9.1937 9.29259 9.08224 8.62818 6.81527

[0106] Test sample:

[0107] Table 7. Outlet water oxidation-reduction potential ORP (mV)

[0108] -5.76831 -4.9992 -6.79379 -18.3945 -15.5103 -17.4331 -15.1258 -35.7635 -57.1704 -76.8467 -91.9084 -101.138 -89.2806 -71.9757 -67.2329 -97.6126 -112.162 -120.045 -122.673 -120.75 -99.7277 -67.297 -48.8384 -51.4021 -52.9403 -45.9542 -45.7619 -38.0068 -24.4192 -12.8185 -12.1135 -30.572 -13.0107 -13.3312 -41.2755 -66.2715 -57.1063 -58.1318 -28.0724 -10.1907 -15.5103 -4.35828 1.15366 6.85788 7.75517 10.4471 9.80613 17.5613 17.8177 19.3559 20.9582 21.5991 21.6632 19.7404 17.5613 18.7791 19.3559 18.9072 31.1489 34.7381 32.0462 31.3412 29.9952 30.508 30.4439 34.9303 34.5458 37.4299 36.0199 33.8408 19.2277 18.9072 16.664 31.3412 35.3149 38.84 38.5195 38.3272 36.6608 33.0717 27.6238 37.494 41.8523 45.9542 46.0183 44.7365 46.4029 48.0052 30.3798 -27.8161 -63.9001 -84.0892 -96.9717 -105.496 -115.558 -124.467 -133.056 -138.952 -144.016 -147.477 -150.617 -155.488 -158.757 -165.935 -174.78 -180.099 -185.291 -190.482 -193.815 -194.52 -194.584 -197.02 -196.571 -196.699 -196.956 -201.378 -202.147 -202.596 -202.34 -202.66 -202.724 -202.596 -205.929 -206.185 -205.352 -198.686 -192.277 -186.06 -190.867 -189.457 -184.971 -174.139 -164.781 -161.641 -159.718 -158.564 -161 -163.179 -163.5 -177.984 -171.96 -169.332 -161.705 -160.551 -163.051 -165.294 -168.307 -199.904 -194.841 -178.497 -164.974 -163.243 -158.5 -165.999 -169.781 -175.549 -173.498 -157.795 -154.334 -153.95 -154.334 -152.86 -149.976 -157.539 -146.323 -144.656

[0109] Table 8. Aerobic terminal dissolved oxygen DO (mg / l)

[0110] 8.62164 8.79934 8.92618 9.32341 8.72147 7.64494 2.63206 1.33762 1.72617 1.97373 2.92381 5.62102 6.03977 5.90297 3.42635 2.35949 3.29965 4.02023 4.28063 7.96556 8.09496 8.25919 6.53687 6.60331 6.84358 7.10001 7.06997 8.95979 8.82943 8.86811 7.6486 8.02626 5.97815 2.04181 3.20333 5.94965 7.12217 9.25386 9.07702 9.40505 9.68643 8.93562 9.52745 9.93701 10.201 10.5732 10.5429 10.5083 10.1724 10.4022 10.7803 10.9162 11.1295 11.5711 11.7333 11.8854 11.7346 11.9131 11.8944 11.8972 5.9678 8.59051 9.07638 9.19526 8.95815 8.69312 9.30462 9.26424 9.06328 8.89382 9.34921 9.51048 9.3812 9.0304 8.96492 9.01655 8.61111 6.08761 6.63068 6.66984 7.34774 8.41383 9.14157 9.50671 9.75259 10.0416 10.2475 10.4039 10.7682 0.81299 0.858016 0.988267 0.68228 0.660338 0.528512 0.491058 0.491551 0.532493 0.571211 0.593547 0.515974 0.513198 0.42863 0.458221 0.468295 0.469763 0.563485 0.567632 0.631162 0.681516 0.584366 0.582236 0.565935 0.573473 0.614534 0.580908 0.567797 0.563439 0.544668 0.555861 0.586122 0.57032 0.584395 0.618609 2.25682 2.14307 2.38559 2.27074 2.66186 2.47066 2.7629 3.48202 3.11333 2.94863 1.88166 2.01599 2.21183 2.30984 2.36159 3.39108 3.39754 3.83651 2.64592 2.61544 2.59893 2.72828 2.83756 3.74977 3.82408 3.42196 2.47916 2.36441 2.60302 2.7205 2.63848 5.76009 5.40471 3.98839 3.55408 3.56215 3.66103 3.69541 3.85411 4.73835 2.13755 3.0876

[0111] Table 9. Total suspended solids TSS (mg / l)

[0112] 2.79965 2.79386 2.78627 2.77584 2.79993 2.82506 2.76997 2.7431 2.77476 2.79975 2.7926 2.7908 2.75555 2.80105 2.81819 2.80891 2.80598 2.92415 2.82021 2.90674 2.81654 2.83801 2.81749 2.82707 2.79572 2.86778 2.816 2.83056 2.87972 2.82129 2.8428 2.86432 2.76964 2.79733 2.79737 2.80613 2.8266 2.82815 2.83306 2.86469 2.86165 3.05081 2.96791 2.99973 3.00772 3.08667 3.04555 3.04938 3.14563 3.02893 3.15554 3.08567 3.10538 3.12656 3.16892 3.14349 3.14252 3.1181 3.04864 3.02205 2.88236 2.88709 2.89729 2.87552 2.94347 2.93794 2.92012 2.95304 2.87526 2.94196 2.93765 2.94192 2.94229 2.92646 2.94162 2.97484 2.87212 2.946 2.90053 2.89887 2.8942 2.9557 2.97338 2.96992 2.97089 3.05264 3.02234 3.01284 3.03694 2.82174 2.79532 2.79901 2.82231 2.85123 2.73648 2.77119 2.8605 2.89791 2.87544 2.78092 2.80868 2.74913 2.74644 2.71467 2.67753 2.7563 2.62515 2.49762 2.60491 2.5817 2.55902 2.56642 2.47093 2.41921 2.52343 2.45012 2.49681 2.47771 2.47763 2.45831 2.44425 2.45045 2.39333 2.46374 2.45729 2.52135 2.37325 2.46015 2.49488 2.56545 2.43302 2.50172 2.45726 2.63053 2.48213 2.41205 2.46999 2.48008 2.47433 2.57774 2.57538 2.56239 2.43037 2.36537 2.3155 2.27395 2.24296 2.26917 2.31572 2.17614 2.13118 2.42613 2.34387 2.3119 2.28225 2.42737 2.41606 2.35143 2.27356 2.38177 2.42869 2.39382 2.51334 2.53214 2.32996 2.40553

[0113] Table 10. Nitrate nitrogen NO in effluent 3 -N(mg / l)

[0114] 12.1225 12.1471 12.1438 12.155 12.5022 12.9956 13.2884 13.2117 13.0797 13.927 13.4202 12.8775 12.6127 12.7819 12.6993 12.1956 11.7575 11.572 11.7068 11.8635 13.0153 13.1411 13.2369 13.0351 13.1148 13.2907 13.3282 13.183 13.634 13.9414 13.8856 13.7639 13.3847 13.4834 13.5821 13.8305 13.9921 13.7059 13.46 13.1795 12.9614 12.7037 12.5334 12.4325 12.2326 11.9899 11.7826 11.9334 11.7879 11.6294 11.5924 11.6376 11.6179 11.5189 11.4142 11.4783 11.5399 11.4543 12.1589 12.1868 14.5787 14.3708 14.2169 14.146 13.9253 13.9486 13.7351 13.6325 13.3241 13.1403 13.1799 13.5794 13.2056 13.4342 13.3163 13.2367 13.0491 12.949 13.0615 13.5936 13.8266 14.5399 14.4021 14.0638 13.936 13.9047 13.8498 13.695 13.4054 19.0773 17.569 16.5405 18.1525 19.2296 17.31 15.334 13.9854 12.8848 12.573 12.5611 13.2709 12.9493 11.3036 9.50539 8.22909 7.51508 6.99893 6.88053 6.69474 7.53834 7.72328 7.87429 7.4206 6.35133 6.04954 6.02727 5.95689 5.8236 5.6425 5.55835 5.51721 5.46329 5.42859 5.45362 5.76626 7.68907 8.88413 10.2344 10.4742 10.8936 11.3214 11.7418 13.4765 14.5953 14.8763 14.2538 13.8325 13.6666 13.8873 14.1119 15.2031 16.1498 16.8975 16.2205 14.5885 12.9808 11.9133 10.7946 10.9491 12.8419 12.8794 12.3108 11.8102 11.4942 11.4617 11.3851 12.3085 13.2389 13.2849 12.2269 12.1361 12.3499 12.7326 13.0438 17.6572 19.9069

[0115] Table 11. pH values

[0116] 7.90242 7.903 7.91051 7.9194 7.91717 7.92658 7.92659 7.89852 7.87181 7.85197 7.84466 7.84898 7.85363 7.86431 7.86428 7.86453 7.85682 7.85947 7.86319 7.87133 7.88456 7.90408 7.91683 7.91535 7.90575 7.90286 7.90325 7.89954 7.92081 7.93864 7.95239 7.89329 7.9095 7.91879 7.9032 7.89208 7.89363 7.89154 7.90152 7.91731 7.88187 7.91856 7.92322 7.93711 7.93153 7.94067 7.93814 7.94469 7.94798 7.95387 7.95554 7.95611 7.95779 7.96721 7.98354 7.99069 8.01065 8.01353 7.99421 8.01248 7.93272 7.96683 7.9638 7.96182 7.9729 7.96636 7.9626 7.97883 7.99151 7.99398 7.9874 7.99 7.99344 8.00182 8.0058 8.00254 8.01909 8.0219 8.01498 8.00668 8.00392 7.99139 7.98845 7.99666 8.01189 8.02739 8.03861 8.04517 8.05309 8.01937 8.01307 7.99942 7.99667 8.02276 7.99958 7.99876 8.00818 7.99371 7.98836 7.98733 8.00734 8.00434 8.00782 8.00002 7.98779 7.99029 7.99179 7.98522 7.99083 7.99491 8.00048 8.00423 8.01016 7.99667 7.99465 7.98444 7.99994 7.99091 8.00562 8.00028 8.00281 8.00434 8.00251 8.0041 8.00438 8.01835 8.02416 8.03374 7.99394 8.02096 8.02744 8.0311 8.02752 8.03982 8.04429 8.03483 8.02165 8.00877 8.01189 7.98652 8.02139 8.05225 8.06162 8.05973 8.03276 8.01372 8.0108 8.0142 8.02959 8.02208 8.03029 8.03992 8.03349 8.01106 8.00009 8.02037 8.01527 8.02206 8.03469 8.02018 8.00903 8.01069 8.01461 8.0146 8.0448 8.0826

[0117] Table 12. Actual effluent ammonia nitrogen concentration (mg / l)

[0118] 3.4096 3.62213 3.72997 3.75166 3.75844 3.72136 3.32112 3.3195 3.42728 3.56972 3.59035 3.88361 3.70863 3.86181 3.67221 3.63954 3.64415 3.80025 3.91885 3.83833 4.06781 3.80538 3.56382 3.64367 3.66317 3.70557 3.73566 3.85178 3.71169 3.6895 3.59665 3.48547 3.56983 3.59459 3.5928 3.65357 3.57038 4.17801 4.3769 4.2244 3.69852 3.32933 3.29505 3.15643 3.16769 3.12063 3.19943 3.54887 3.50727 3.51509 3.47789 3.48063 3.48976 3.61165 3.50616 3.62607 3.68729 3.73111 3.78588 3.7712 3.55908 3.5344 3.61249 3.80938 3.58067 3.65397 3.72711 3.59422 3.50091 3.53588 3.45543 3.6119 4.18721 3.81493 3.68018 3.70755 3.68006 3.73598 3.73521 3.78167 3.93116 3.76504 3.45873 3.62687 3.68509 3.72068 3.68545 3.83061 4.05469 3.33993 4.74973 5.34744 5.88695 5.72592 6.71882 7.56025 8.02043 8.44971 8.86697 9.18126 9.48174 9.64237 10.5585 10.9436 11.3939 11.7327 11.9674 12.2836 12.3155 12.5365 12.2718 12.32 12.3038 12.4523 12.7659 12.7696 12.9716 12.8835 13.0054 12.9644 12.9466 13.0941 13.2232 13.1733 13.2032 12.7643 12.2235 11.5723 11.2749 10.9602 10.7283 10.4957 9.68682 8.99254 8.56815 8.34895 8.29665 8.18503 8.11735 7.96221 7.45065 6.95279 6.3957 6.3166 6.55805 7.09273 7.78201 9.03523 8.74751 8.766 8.84565 9.37012 9.05987 9.24065 9.31572 9.38502 8.95312 8.54612 8.19655 9.34993 9.07057 9.24795 9.37536 9.13056 7.53053 6.56707

Claims

1. A soft sensing method for effluent ammonia nitrogen based on ADw-CLPSO radial basis emotion neural network, It is characterized in that The following steps are involved: Step 1: Variable selection The soft sensor model is based on the outlet NH 4 -N concentration was used as the output variable, with effluent redox potential ORP, dissolved oxygen DO, total suspended solids TSS, effluent nitrate nitrogen NO 3 -N and pH values ​​as input variables; Step 2: Design the ADw-CLPSO radial basis function neural network topology for effluent ammonia nitrogen concentration prediction Determine the topological structure of the radial basis emotion neural network as a 5-p-1 connection mode, that is, the number of input layer neurons is 5, the number of hidden layer neurons is p, p is a positive integer, and the output neuron is 1; assign values ​​to the parameters of the neural network; suppose the neural network input at time t is z(t)=[z 1 (t),z 2 (t),z 3 (t),z 4 (t),z 5 (t)], the actual output is expressed as E(t), and the actual output of the neural network is expressed as: Among them, ν i (i=1,…,p) and w i (i=1,…,p) are the connection weights of the i-th neuron in the hidden layers "amygdala" and "orbitofrontal cortex" of the radial basis emotion neural network, p is the number of neurons in the hidden layer; b is the bias; is any Gaussian function The maximum value among and u is its corresponding connection weight; is the input of the ith neuron in the hidden layer "thalamus" of the radial basis emotion neural network, and ν i The update methods are: Among them, μ i and σ i Represents the center vector and width value of the i-th neuron in the hidden layer "thalamus", σ i >0;||z(t)-μ i || is z(t) and μ i The Euclidean distance between i (t+1) represents the connection weight of the i-th node at time t+1; α is the learning rate of the neurons in the hidden layer "amygdala" of the radial basis emotion neural network; Step 3: Train the Neural Network Step 3.1: Initialize the parameters of the comprehensive learning particle swarm algorithm Determine the initial number of iterations t = 0, the maximum number of iterations max_t = 1000; input vector z l represents the lth group of input samples, represents the expected target value of the lth group of samples, Indicates that the input sample dimension is n, T is the total number of samples; the population is divided into 10 groups to represent the radial basis emotion neural network with 3-12 hidden layer neurons, and the number of particles in each group is 15; the particle acceleration constant c is initialized to a random number (0,1); the initial inertia weight of the particle ω 0 and the minimum inertia weight ω 1 are set to 0.95 and 0.55 respectively; the parameters of the radial basis emotion neural network are represented as particles in the comprehensive learning particle swarm: Among them, x j represents the position of the jth particle, j = 1, 2, ..., S; S is the total number of particles, S is a positive integer, b is the deviation, w j,p , and σ p are respectively represented as the connection weight, center vector and width value of the neuron in the hidden layer "orbital frontal cortex" of the p-th radial basis emotion neural network in the j-th particle; b, w j,p , and σ p The initial value of is a random number (-3,3); the connection weights ν of the neurons in the hidden layers "amygdala" and "thalamus" of the radial basis emotion neural network are i The initial values ​​of and u are random numbers (0,1); the learning rate α for weight update in the "amygdala" neurons in the hidden layer of the radial basis emotion neural network and the search range of the balance factor δ in the particle fitness function are set to [0.1, 0.12, ..., 0.28, 0.3] and [0.001, 0.002, ..., 0.01] respectively, and the optimal parameter combination of α and δ that minimizes the training error is selected by the grid search method; the iteration number flag β for the particle fitness value to be recalculated and the iteration number flag γ for the global optimal position gbest of the population that has not been improved are set to 4; at the same time, the speed of each particle is initialized: Among them, v j represents the velocity of the jth particle, Indicates that the jth particle is in the D j dimensional speed and initialize it to a random number of (-2,2), D j represents the dimension of the jth particle, D j =(2+n)p j +1, n represents the number of input variables, p j represents the number of neurons in the hidden layer of the radial basis emotion neural network represented by the jth particle; Step 3.2: Design particle fitness function For the input z(t) of the radial basis emotion neural network, determine the dimension D of each particle j =(2+n)p j +1, calculate the fitness value of each particle: f(x j (t))=R j (t)+δD j (t) (6) Among them, δ is the balance factor in the particle fitness function, R j (t) is: Among them, E j (t) and They represent the network output represented by the jth particle and the expected target value at time t; T represents the number of training samples input to the neural network; the initial position x of each particle is j Initialized to the individual optimal position of the particle in is the optimal position of the jth particle in dimension d, f j (d) represents the optimal position of the jth particle, the particle number to be learned in the dth dimension, which is randomly initialized to different particle numbers in the group to which the particle belongs; after calculating the fitness value of each particle, the individual optimal position with the minimum fitness value is taken as the global optimal position gbest of the population; Step 3.3: Determine whether the particle dimension has changed Record the number of iterations without improvement of the global optimal position gbest of the population. When its value is greater than γ, calculate the average fitness value of each group of particles: Among them, avg(·) is the function for calculating the average value; Group M represents the Mth group of particles in the population; fit j represents the fitness value of the jth particle; m 1 is the first particle of the Mth group; S M is the number of particles in the Mth group; find the best neural network structure and the corresponding particle dimension D according to the global optimal position gbest best , the optimal number of neurons in the hidden layer of the neural network is (D best -1) / (2+n), n is the number of input variables; according to D best Update the dimension of each particle corresponding to the number of neurons in the hidden layer of the neural network: Among them, △d represents the unit length of the particle parameters to be adjusted when the number of neurons in the hidden layer of the neural network changes. In this example, △d = 2 + n; then the number of iterations of the recorded global optimal position gbest of the population that has not been improved is cleared, and the inertia weight of each particle at time t+1 is updated: in, ω j (t+1) is the inertia weight of the jth particle at time t+1, ω 0 and ω 1 are the initial inertia weight and minimum inertia weight of the particle respectively; fit(x j (t)) is the fitness value of the jth particle; gbest is the global optimal position of the population; max_t is the maximum number of iterations; tanh{·} is the hyperbolic tangent function; exp{·} is the exponential function; Step 3.4: Update particle parameters When the individual optimal position of the jth particle stops improving β times, the optimal position of the jth particle is recalculated at the particle number f to be learned in the dth dimension j (d); Then update the position and velocity of each particle: Among them, v j,d (t+1) and x j,d (t+1) represents the velocity and position of the jth particle in the dth dimension at time t+1; ω j (t) is the inertia weight of the jth particle at time t; r j,d It means that the jth particle takes a random number between [0,1] in the dth dimension; c is the acceleration factor of the particle; Step 3.5: Determine whether the training is finished Input training sample data z(t+1), repeat steps 3.2-3.4, and stop the calculation if t>max_t. The parameters represented by the global optimal position gbest of the population are used as the optimal network parameters, and then according to D best Determine the optimal network structure size p best =(D best -1) / (2+n), p best The number of hidden layer neurons corresponding to the optimal network structure; execute step 4; Step 4: Test the Neural Network The test sample data is used as the input of the trained radial basis emotion network based on ADw-CLPSO to test the radial basis emotion neural network. The output of the radial basis emotion neural network is the outflow NH 4 - The predicted value of N.

Images