Ammonia nitrogen concentration prediction method, computing device and computer readable storage medium

Abstract

The invention discloses an ammonia nitrogen concentration prediction method, a computing device and a computer readable storage medium wherein the ammonia nitrogen concentration prediction method comprises: determining a plurality of characteristic factors affecting the ammonia nitrogen concentration, and collecting historical data corresponding to each characteristic factor wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration in the sewage treatment process; determining a correlation coefficient between each characteristic factor and the ammonia nitrogen concentration, and determining training data for a variable structure echo state network model from the historical data based on the correlation coefficient; constructing the variable structure echo state network model; and optimizing the variable structure echo state network model by using the training data, and predicting the ammonia nitrogen concentration in the sewage treatment process by using the optimized variable structure echo state network model. Therefore, the ammonia nitrogen concentration can be accurately determined in the sewage treatment process.

Description

Technical Field

[0001] This specification relates to the field of wastewater treatment technology, and in particular to a method for predicting ammonia nitrogen concentration. One or more embodiments of this specification also relate to a computing device and a computer-readable storage medium. Background Technology

[0002] With the acceleration of urbanization, the operational efficiency and effluent quality stability of urban wastewater treatment plants are becoming increasingly important. Ammonia nitrogen is a key indicator of effluent quality from wastewater treatment plants; excessively high concentrations can lead to eutrophication of water bodies and harm the ecological environment. Therefore, accurately determining the ammonia nitrogen concentration has become an urgent technical problem to be solved. Summary of the Invention

[0003] In view of this, embodiments of this specification provide a method for predicting ammonia nitrogen concentration. One or more embodiments of this specification also relate to an ammonia nitrogen concentration prediction device, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this specification, a method for predicting ammonia nitrogen concentration is provided, comprising:

[0005] Multiple characteristic factors affecting ammonia nitrogen concentration were identified, and historical data corresponding to each characteristic factor were collected, wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration during the wastewater treatment process;

[0006] Determine the correlation coefficient between each of the aforementioned characteristic factors and the ammonia nitrogen concentration, and based on the correlation coefficient, determine the training data for the variable structure echo state network model from the historical data;

[0007] The variable structure echo state network model is constructed, wherein the variable structure echo state network model includes a fractional-order reservoir and an integer-order reservoir. The fractional-order reservoir is used for feature extraction, and the integer-order reservoir is used for feature fusion. The number of fractional-order reservoirs is determined according to the amount of training data.

[0008] The training data is used to optimize the variable structure echo state network model, and the optimized variable structure echo state network model is used to predict ammonia nitrogen concentration in the wastewater treatment process.

[0009] According to a second aspect of the embodiments of this specification, an ammonia nitrogen concentration prediction device is provided, comprising:

[0010] The historical data determination module is configured to determine multiple characteristic factors affecting ammonia nitrogen concentration and collect historical data corresponding to each characteristic factor, wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration during the wastewater treatment process;

[0011] The training data determination module is configured to determine the correlation coefficient between each feature factor and the ammonia nitrogen concentration, and based on the correlation coefficient, determine training data for the variable structure echo state network model from the historical data.

[0012] The model building module is configured to build the variable structure echo state network model, wherein the variable structure echo state network model includes a fractional-order reservoir and an integer-order reservoir, the fractional-order reservoir is used for feature extraction, the integer-order reservoir is used for feature fusion, and the number of fractional-order reservoirs is determined according to the amount of training data.

[0013] The model optimization module is configured to optimize the variable structure echo state network model using the training data, and to use the optimized variable structure echo state network model to predict ammonia nitrogen concentration during wastewater treatment.

[0014] According to a third aspect of the embodiments of this specification, a computing device is provided, comprising:

[0015] Memory and processor;

[0016] The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which, when executed by the processor, implement the steps of the above-described ammonia nitrogen concentration prediction method.

[0017] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores a computer program / instructions that, when executed by a processor, implement the steps of the above-described ammonia nitrogen concentration prediction method.

[0018] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described ammonia nitrogen concentration prediction method.

[0019] This specification provides one or more embodiments of a method for predicting ammonia nitrogen concentration, comprising: identifying multiple characteristic factors affecting ammonia nitrogen concentration and collecting historical data corresponding to each characteristic factor, wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration during wastewater treatment; determining the correlation coefficient between each characteristic factor and the ammonia nitrogen concentration, and based on the correlation coefficient, determining training data for a variable structure echo state network model from the historical data; constructing the variable structure echo state network model, wherein the variable structure echo state network model includes fractional-order reservoirs and integer-order reservoirs, the fractional-order reservoirs are used for feature extraction, the integer-order reservoirs are used for feature fusion, and the number of fractional-order reservoirs is determined according to the amount of training data; optimizing the variable structure echo state network model using the training data, and using the optimized variable structure echo state network model to predict ammonia nitrogen concentration during wastewater treatment.

[0020] Specifically, the ammonia nitrogen concentration prediction method in the embodiments of this specification can construct a variable structure echo state network model including fractional-order and integer-order reservoirs. Furthermore, based on various characteristic factors affecting ammonia nitrogen concentration and the correlation coefficient between each characteristic factor and ammonia nitrogen concentration, training data for the variable structure echo state network model is determined from historical data. The variable structure echo state network model is then optimized using the training data, thereby obtaining a variable structure echo state network model capable of accurately predicting ammonia nitrogen concentration. This ensures accurate determination of ammonia nitrogen concentration during wastewater treatment, thereby optimizing reagent dosage, reducing treatment costs, and ensuring effluent meets standards. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an ammonia nitrogen concentration prediction method provided in one embodiment of this specification;

[0022] Figure 2 This is a flowchart illustrating the processing steps of an ammonia nitrogen concentration prediction method provided in one embodiment of this specification.

[0023] Figure 3 This is a schematic diagram of the structure of a variable structure echo state network model in an ammonia nitrogen concentration prediction method provided in one embodiment of this specification;

[0024] Figure 4 This is a schematic diagram of the structure of an ammonia nitrogen concentration prediction device provided in one embodiment of this specification;

[0025] Figure 5 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0026] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0027] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0028] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0029] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0030] With the acceleration of urbanization, the operational efficiency and effluent quality stability of urban wastewater treatment plants are becoming increasingly important. Ammonia nitrogen is a key indicator of effluent quality from wastewater treatment plants; excessively high concentrations can lead to eutrophication of water bodies, harming the ecological environment. Therefore, accurate prediction of ammonia nitrogen concentration is necessary to optimize reagent dosage, reduce treatment costs, and ensure that effluent meets standards.

[0031] The ammonia nitrogen concentration prediction scheme provided in this specification mainly includes a mechanistic model and a data-driven model.

[0032] Mechanism models are based on the biochemical reaction principles of wastewater treatment, but they suffer from problems such as complex parameters and poor adaptability.

[0033] Data-driven models (such as neural networks and support vector machines) are widely used due to their strong nonlinear fitting capabilities. Among them, echo state networks (ESNs) have attracted attention due to their simple training and strong ability to process time series data. However, ESNs use integer-order reservoirs, making it difficult to fully capture the complex dynamic characteristics and long-term memory effects in wastewater treatment. While some fractional-order ESNs introduce fractional derivatives to improve memory capabilities, they have limitations: either they cannot effectively decompose the independent features of input variables, or they ignore the coupling relationships between variables, resulting in insufficient prediction accuracy. Furthermore, the aforementioned mechanistic models and data-driven models lack clear stability criteria, and parameter selection relies on experience, further affecting prediction reliability. Therefore, there is an urgent need for a method for predicting ammonia nitrogen concentration that can simultaneously consider variable decomposition and coupling, has stability guarantees, and optimizes parameters.

[0034] Based on this, this specification provides a method for predicting ammonia nitrogen concentration. This specification also relates to an ammonia nitrogen concentration prediction device, a computing device, and a computer-readable storage medium, which will be described in detail in the following embodiments.

[0035] See Figure 1 , Figure 1 A flowchart of an ammonia nitrogen concentration prediction method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0036] Step 102: Identify the various characteristic factors affecting ammonia nitrogen concentration and collect historical data corresponding to each characteristic factor, wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration during the wastewater treatment process.

[0037] Here, the characteristic factor can be understood as the factor that can affect the ammonia nitrogen concentration; historical data can be understood as the index data or values ​​generated by each characteristic factor in the past.

[0038] In one or more embodiments provided in this specification, the characteristic factors include pH value (i.e., hydrogen ion concentration index), water temperature, dissolved oxygen concentration, total phosphorus content, and total suspended solids content. Specifically, the method provided in this specification can collect historical data on various characteristic factors affecting ammonia nitrogen concentration during wastewater treatment during the training of the variable structure echo state network model; that is, it is necessary to determine which characteristic factors affect ammonia nitrogen concentration during wastewater treatment and collect historical data on these characteristic factors. Specifically, the method provided in this specification selects five key water quality indicators as characteristic factors: pH value, water temperature, dissolved oxygen concentration, total phosphorus content, and total suspended solids content. By collecting historical data on these indicators, training data can be selected from the historical data to effectively train the variable structure echo state network model, enabling the model to accurately predict ammonia nitrogen concentration in wastewater, thereby providing data support for subsequent process optimization. In other words, the ammonia nitrogen concentration prediction method provided in this specification can establish training and testing datasets for the model based on the historical data, for training and testing the variable structure echo state network model.

[0039] Step 104: Determine the correlation coefficient between each feature factor and the ammonia nitrogen concentration, and based on the correlation coefficient, determine the training data for the variable structure echo state network model from the historical data;

[0040] The correlation coefficient can be understood as a coefficient that characterizes the degree of correlation between a characteristic factor and ammonia nitrogen concentration. The influence of the characteristic factor on ammonia nitrogen concentration can be determined through this correlation coefficient. For example, the correlation coefficient can be numerical values, indicators, or other data.

[0041] Specifically, after acquiring historical data, the correlation coefficients between each characteristic factor and ammonia nitrogen concentration are analyzed to calculate the impact of historical data at past moments on the current ammonia nitrogen concentration. Then, based on the correlation coefficients, the input characteristic variables (also known as input feature variables) and the length of historical time are determined from the historical data. Based on the characteristic variables and the length of historical time, the training data for the variable structure echo state network model can be determined.

[0042] In one or more embodiments provided in this specification, determining the correlation coefficient between each feature factor and the ammonia nitrogen concentration, and determining training data for the variable structure echo state network model from the historical data based on the correlation coefficient, includes:

[0043] Based on the historical data, a training dataset for the variable structure echo state network model is constructed;

[0044] The correlation coefficient between each characteristic factor and the ammonia nitrogen concentration was calculated using sequence cross-correlation analysis.

[0045] Analyze the time correlation coefficient between the historical data and the ammonia nitrogen concentration;

[0046] Based on the correlation coefficient, multiple input features are selected from the training dataset, and based on the temporal correlation coefficient, the length of the historical time corresponding to each input feature is determined.

[0047] The multiple input features and the length of the historical time corresponding to each input feature are determined as the training data of the variable structure echo state network model.

[0048] The time correlation coefficient can be understood as a coefficient that represents the degree of correlation between historical data and ammonia nitrogen concentration in the time dimension. This time correlation coefficient can be time series data (such as indicators or values ​​arranged in chronological order).

[0049] Specifically, the ammonia nitrogen concentration prediction method provided in this specification performs correlation analysis between various characteristic factors and ammonia nitrogen concentration, calculates the impact of historical data from past moments on the current ammonia nitrogen concentration, and thus determines the input characteristic variables and the length of historical time. The correlation analysis can employ serial cross-correlation analysis. By calculating the correlation coefficient between characteristic factors and ammonia nitrogen concentration, input features with significant correlations can be screened out. Simultaneously, the temporal correlation between historical data and the current ammonia nitrogen concentration can be analyzed to determine the length of the historical time window for the input features.

[0050] Taking the application of the ammonia nitrogen concentration prediction method provided in this manual in a wastewater treatment scenario as an example, this manual explains the ammonia nitrogen concentration prediction method. Specifically, the ammonia nitrogen concentration prediction method provided in this manual employs a data preprocessing scheme during the construction of training data to determine the optimal input features and key parameters of the model. This method includes two steps:

[0051] First, input features are selected. The ammonia nitrogen concentration prediction method provided in this manual does not use all water quality indicators (i.e., historical data) as training data. Instead, it employs "serial cross-correlation analysis" to calculate the correlation coefficient between the historical data sequence of each characteristic factor (such as pH value, water temperature, etc.) and the ammonia nitrogen concentration data sequence (i.e., the aforementioned ammonia nitrogen concentration) at different time shifts. By analyzing these correlation coefficients, characteristic variables that are statistically significantly associated with changes in ammonia nitrogen concentration can be selected and used as input features of the model. This eliminates interference from irrelevant or redundant information, improving the effectiveness of the model input.

[0052] Secondly, the length of the historical time period needs to be determined. The ammonia nitrogen concentration prediction method provided in this manual takes into account that changes in ammonia nitrogen concentration are often not instantaneous, but rather continuously influenced by previous water quality conditions. Based on this, the correlation strength between data from different historical time periods (e.g., one day ago, two days ago, etc.) and the current ammonia nitrogen concentration can be analyzed. By analyzing this temporal correlation, an optimal length of historical time period (also known as the historical time window length) can be determined to indicate how long the model needs to process data from the past, facilitating the model's effective prediction of the current ammonia nitrogen concentration. This ensures that the model can capture key temporal dependencies.

[0053] As can be seen from the above embodiments, this method determines the amount of input data by the strength of the correlation between data points. Many factors influence ammonia nitrogen concentration; if all were treated as input variables, the model would require extremely powerful training capabilities. Therefore, it is necessary to analyze the correlation between various factors to select feature inputs, thereby reducing the computational complexity of the model and improving prediction efficiency and accuracy.

[0054] Step 106: Construct the variable structure echo state network model, wherein the variable structure echo state network model includes a fractional-order reservoir and an integer-order reservoir, the fractional-order reservoir is used for feature extraction, the integer-order reservoir is used for feature fusion, and the number of fractional-order reservoirs is determined according to the amount of training data;

[0055] Specifically, the ammonia nitrogen concentration prediction method provided in this specification can construct a variable structure echo state network model. The variable structure echo state network model can include multiple fractional-order reservoirs, and it can include one or at least one integer-order reservoir. It should be noted that after initializing the variable structure echo state network model, the model (i.e., the variable structure echo state network model) can be trained using a training dataset.

[0056] In one or more embodiments provided in this specification, the fractional-order reservoir is a plurality of parallel fractional-order reservoirs used to extract features from the training data to obtain data features, wherein the number of the plurality of fractional-order reservoirs is determined according to the number of input features in the training data;

[0057] The integer-order reservoir is a series of single integer-order reservoirs used for feature fusion of data features to obtain fused features; and for predicting ammonia nitrogen concentration; wherein, the method for predicting ammonia nitrogen concentration is: using the fused features to predict ammonia nitrogen concentration to obtain predicted ammonia nitrogen concentration.

[0058] Specifically, the ammonia nitrogen concentration prediction method provided in this specification can construct a variable structure echo state network model and initialize the variable structure echo state network model. The fractional-order reservoirs included in the variable structure echo state network model can be multiple parallel fractional-order reservoirs (also known as multi-fractional-order reservoirs), and the integer-order reservoirs included in the variable structure echo state network model can be a single integer-order reservoir connected in series (also known as a single integer-order reservoir).

[0059] During the training of the variable structure echo state network model using the training dataset, the multi-fractional-order reservoir is used to extract independent features of the input variables (i.e., the training data), while the integer-order reservoir is used to fuse the features output by the multi-fractional-order reservoir and capture the coupling relationship between variables, thereby predicting the ammonia nitrogen concentration.

[0060] It should be noted that there can be multiple input features (or various types). In this case, the variable structure echo state network model needs to construct fractional-order reservoirs that can process each type of input feature separately. Therefore, the number of fractional-order reservoirs in the variable structure echo state network model can be determined based on the number of multiple input features or the number of types of multiple input features, thus ensuring that the input features can be processed effectively and accurately.

[0061] Step 108: Optimize the variable structure echo state network model using the training data, and use the optimized variable structure echo state network model to predict ammonia nitrogen concentration during wastewater treatment.

[0062] Optimizing the variable structure echo state network model can be understood as training the model or adjusting its parameters; these parameters can be model weights, activation functions, etc.

[0063] In some embodiments, optimizing the variable structure echo state network model using the training data and using the optimized variable structure echo state network model to predict ammonia nitrogen concentration during wastewater treatment includes steps one to three:

[0064] Step 1: Using the variable structure echo state network model, perform prediction processing based on the training data to obtain the predicted ammonia nitrogen concentration.

[0065] The ammonia nitrogen concentration prediction method provided in this manual can input training data into a variable structure echo state network model, and use the fractional-order and integer-order reservoirs in the variable structure echo state network model to process the training data to obtain the predicted ammonia nitrogen concentration.

[0066] In some embodiments, the training data includes multiple input features and the length of the historical time corresponding to each input feature;

[0067] The step of using the variable structure echo state network model to perform prediction processing based on the training data to obtain the predicted ammonia nitrogen concentration includes:

[0068] From multiple parallel fractional-order reserve pools, a target fractional-order reserve pool corresponding to the target input feature is determined, wherein the target input feature is any one of the multiple input features, and the target fractional-order reserve pool is a reserve pool among the multiple fractional-order reserve pools used for feature extraction of the target input feature;

[0069] The target input features and the corresponding historical time lengths are input into the target fractional-order reserve pool for feature extraction, and the data features are output.

[0070] The data features output from the multiple fractional-order reservoirs are input into the integer-order reservoir for feature fusion to obtain fused features, and the predicted ammonia nitrogen concentration is obtained using the fused features.

[0071] Continuing with the previous example, during the training of the model using training data, a variable structure echo state network model is needed to predict the training data and obtain the predicted ammonia nitrogen concentration. Subsequently, a loss function can be calculated based on this predicted ammonia nitrogen concentration, and the model parameters can be adjusted based on this loss function. The steps for using the variable structure echo state network model to predict the training data are as follows:

[0072] Step 1: Initialize the multi-fractional-order reservoir. Each fractional-order reservoir corresponds to one input feature; the state update equation for this multi-fractional-order reservoir is:

[0073]

[0074] Where j = 1, 2, ..., J (J is the input feature dimension), x j (r) represents the state of the j-th fractional-order reserve pool at time r, a ij Here, a represents the historical state coefficient. fj f is the activation term coefficient. j =tanh is the activation function. For the input weights, u j (r) is the j-th input variable, W j This refers to the weights within the reserve pool.

[0075] Step 2: For the initialization of the integer-order reserve pool, receive the output (i.e., data features) of all fractional-order reserve pools as input and perform feature fusion processing to obtain fused features.

[0076] The state update equation for this integer-order reserve pool is:

[0077] x(r) = bx(r-1) + f(W) in u(r)+Wx(r-1)+W fb y(r-1))

[0078] The equation corresponding to the output y(r) of the integer-order reservoir (i.e., the predicted ammonia nitrogen concentration) is as follows:

[0079] y(r)=W out x(r)

[0080] Where x(r) is the state of the integer-order reserve pool at time r, b is the leakage rate, f = tanh is the activation function, and W in Input weights for the integer-order reservoir, u(r) = [x1(r), x2(r), ..., x J [r] represents the set of states in the fractional-order reserve pool, and W represents the internal weights of the integer-order reserve pool. fb For feedback weights, y(r-1) represents historical output, and W... out For output weights.

[0081] Step 3: Based on the stability criterion, specifically, this method can set a model stability criterion based on the principle of inequality compression, and the criterion satisfies the following equation;

[0082]

[0083] Where, δ max , These are the maximum singular values ​​of the corresponding weight matrices, ensuring that the initial parameters of the model are within a stable range.

[0084] Based on the above formula, it can be seen that this method can verify whether the initial parameters meet the stability requirements. If they do not meet the stability requirements, they will be re-initialized. If they do meet the stability requirements, step 4 below will be executed.

[0085] Step 4: Input the training data into the model, process the training data using a multi-fractional-order reservoir and an integer-order reservoir, and then use the formula... Calculate the output weight W out , where X(r) is the set of integer-order reserve pool states, and Y(r) is the desired output matrix.

[0086] Based on the above, it can be seen that the output weight W out It can be calculated using Tikhonov regularized regression, with the specific formula as follows:

[0087]

[0088] Where X(r) is the set of integer-order reservoir states, and Y(r) is the desired output matrix. It is the pseudoinverse of X(r).

[0089] Based on the above, this method determines the number of fractional-order reservoirs by the number of input features. Standard echo state networks may encounter problems such as insufficient data feature extraction and gradually decreasing storage capacity in ammonia nitrogen concentration prediction applications. Employing multiple fractional-order reservoirs in parallel to process input features can fully decompose the independent features of each variable. Combined with the infinite memory capability of fractional derivatives, this enhances the ability to capture long-term dynamic features of the wastewater treatment process. Subsequently, features are fused through serial connection of single-integer-order reservoirs, effectively considering the coupling relationships between input variables.

[0090] Step 2: Using the actual ammonia nitrogen concentration corresponding to the training data and the predicted ammonia nitrogen concentration, determine the loss function, and train the variable structure echo state network model based on the loss function to obtain the trained variable structure echo state network model.

[0091] The training data mentioned above can be understood as training samples, and the actual ammonia nitrogen concentration can be understood as the training label corresponding to the training sample; the actual ammonia nitrogen concentration is the actual ammonia nitrogen concentration in the wastewater treatment process.

[0092] Specifically, the ammonia nitrogen concentration prediction method provided in this specification can optimize the reservoir parameters of a variable structure echo state network. During the optimization process, the model parameters and search range can be initialized, and the optimal parameters can be determined by minimizing the prediction error objective function through the gradient descent algorithm.

[0093] In some embodiments, determining a loss function using the actual ammonia nitrogen concentration corresponding to the training data and the predicted ammonia nitrogen concentration, and training the variable structure echo state network model based on the loss function to obtain the trained variable structure echo state network model includes:

[0094] The objective function is calculated using the actual ammonia nitrogen concentration corresponding to the training data and the predicted ammonia nitrogen concentration.

[0095] The gradient descent algorithm is used to calculate the derivative of the objective function with respect to the parameters of the reservoir, wherein the reservoir includes a fractional-order reservoir and an integer-order reservoir.

[0096] The parameters of the reservoir are updated based on the derivative until the model training stopping condition is met, thereby obtaining the trained variable structure echo state network model.

[0097] In some embodiments, the parameters of the reservoir include the parameters of the fractional-order reservoir and the parameters of the integer-order reservoir; the parameters of the fractional-order reservoir include the scaling factor, spectral radius, and input scaling factor of the fractional-order reservoir; the parameters of the integer-order reservoir include the leakage rate, spectral radius, and feedback scaling factor of the integer-order reservoir.

[0098] The model training stopping condition can be set according to the actual application scenario. This manual does not impose specific restrictions on it. For example, the model training stopping condition can be that the model has reached a preset number of iterations or that the objective function (i.e., the loss function) has converged.

[0099] Following the example above, the ammonia nitrogen concentration prediction method provided in this manual requires optimization of the reservoir parameters during model training. Once optimization is complete, the trained variable-structure echo state network model can be obtained. Based on this, the optimization of the reservoir parameters includes the following steps:

[0100] Step 1: Determine the parameters to be optimized (i.e., the parameters of the reservoir), including but not limited to: the scaling factor of the fractional-order reservoir. Input scaling factor Spectral radius ρ j Leakage rate b of integer-order reserve pool, scaling factor a f s in s fb (i.e., feedback scaling factor), spectral radius ρ.

[0101] Step 2: Define the objective function for prediction error as follows:

[0102] Step 3: Gradient Descent Optimization: Calculate the derivative of the objective function with respect to each parameter. The formula for gradient descent optimization is as follows:

[0103]

[0104] Where e(r) = d(r) - y(r).

[0105] Step 4: Update the parameters based on the derivative, using the following formula:

[0106]

[0107] Where q is the parameter to be optimized, and η∈[0.00001,0.0001] is the learning rate. Based on this, it can be seen that the ammonia nitrogen concentration prediction method provided in this specification can be used to predict ammonia nitrogen concentration according to... Update the reservoir parameters until the objective function converges (i.e., until the error converges).

[0108] It should be noted that after training the variable-structure echo state network model using the training dataset, it is also necessary to test the model using the test dataset. If the test is passed, the training of the variable-structure echo state network model can be considered complete. The processing method for the test dataset is the same as that for the training dataset. For the processing procedure for the test dataset, please refer to the above description of the processing procedure for the training dataset; this specification will not elaborate further.

[0109] Step 3: During the wastewater treatment process, the trained variable structure echo state network model is used to predict ammonia nitrogen concentration.

[0110] Specifically, after the model training is completed, the optimized variable structure echo state network can be used to predict ammonia nitrogen concentration in wastewater treatment and put into practical application.

[0111] Based on the above embodiments, it can be seen that this method can optimize reservoir parameters and inter-reservoir connectivity coefficients using gradient descent; finally, a variable structure echo state network is trained and used for predicting ammonia nitrogen concentration in wastewater treatment.

[0112] Based on the above embodiments, it is clear that the ammonia nitrogen concentration prediction method in this specification can construct a variable structure echo state network model including fractional-order and integer-order reservoirs. Furthermore, based on various characteristic factors affecting ammonia nitrogen concentration and the correlation coefficient between each characteristic factor and ammonia nitrogen concentration, training data for the variable structure echo state network model is determined from historical data. The variable structure echo state network model is then optimized using the training data, thereby obtaining a variable structure echo state network model capable of accurately predicting ammonia nitrogen concentration. This ensures accurate determination of ammonia nitrogen concentration during wastewater treatment, thereby optimizing reagent dosage, reducing treatment costs, and ensuring effluent meets standards.

[0113] The following is in conjunction with the appendix Figure 2 Taking the application of the ammonia nitrogen concentration prediction method provided in this specification in wastewater treatment as an example, the ammonia nitrogen concentration prediction method will be further explained. Among other things, Figure 2 The present specification shows a flowchart of a method for predicting ammonia nitrogen concentration according to an embodiment of the present specification, which specifically includes the following steps.

[0114] Step 202: Use correlation analysis to determine the relationships between feature factors to identify the input.

[0115] Specifically, the method provided in this manual can collect historical data on various characteristic factors affecting ammonia nitrogen concentration during wastewater treatment in the process of training the variable structure echo state network model. That is, it is first necessary to determine which characteristic factors affect ammonia nitrogen concentration during wastewater treatment and collect historical data on these factors. Specifically, the method provided in this manual selects five key water quality indicators as characteristic factors: pH value, water temperature, dissolved oxygen concentration, total phosphorus content, and total suspended solids content.

[0116] Step 204: Construct the training dataset and the test dataset.

[0117] Specifically, after acquiring historical data, the correlation coefficients between each characteristic factor and ammonia nitrogen concentration are analyzed to calculate the impact of historical data at past moments on the current ammonia nitrogen concentration. Then, based on the correlation coefficients, the input characteristic variables and the length of historical time are determined from the historical data. Based on the characteristic variables and the length of historical time, the training dataset and test dataset for the variable structure echo state network model can be determined.

[0118] Step 206: Construct a prediction model using a variable structure echo state network.

[0119] Specifically, the ammonia nitrogen concentration prediction method provided in this specification can construct a variable structure echo state network model (i.e., a prediction model used to predict ammonia nitrogen concentration) and initialize the variable structure echo state network model. The fractional-order reservoirs included in the variable structure echo state network model can be multiple parallel fractional-order reservoirs (also known as multi-fractional-order reservoirs), and the integer-order reservoirs included in the variable structure echo state network model can be a single integer-order reservoir connected in series (also known as a single integer-order reservoir).

[0120] During the training of the variable structure echo state network model using the training dataset, the multi-fractional-order reservoir is used to extract independent features of the input variables (i.e., the training data), while the integer-order reservoir is used to fuse the features output by the multi-fractional-order reservoir and capture the coupling relationship between variables, thereby predicting the ammonia nitrogen concentration.

[0121] The model structure for this variable structure echo state network model can be found in [reference needed]. Figure 3 , Figure 3 This is a schematic diagram of the structure of a variable structure echo state network model in an ammonia nitrogen concentration prediction method provided in one embodiment of this specification; based on Figure 3It is understood that the variable structure echo state network model includes multiple fractional-order reservoirs and one integer-order reservoir. The fractional-order reservoirs are used to extract features from the input data (e.g., training data), and then the extracted features are aggregated and input into the integer-order reservoir for processing; the feature aggregation can be achieved through feature concatenation or feature fusion. The integer-order reservoir can perform predictive processing on the input features to obtain the predicted ammonia nitrogen concentration.

[0122] Step 208: Put the trained variable structure echo state network into use.

[0123] Specifically, based on steps 106-108 above, this method can optimize the reservoir parameters of the variable structure echo state network. The optimization method involves initializing the model parameters and search range, minimizing the prediction error objective function using the gradient descent algorithm, determining the optimal parameters, and then using the optimized variable structure echo state network for ammonia nitrogen concentration prediction in wastewater treatment processes, thus enabling practical application. Therefore, the explanation of step 208 can be found in the corresponding or relevant content of steps 106-108 in the above embodiments, and this specification will not elaborate further on this.

[0124] Based on the above, the ammonia nitrogen concentration prediction method provided in this specification is a wastewater treatment ammonia nitrogen concentration prediction method based on a variable structure echo state network. By employing multiple fractional-order reservoirs to process input features in parallel, it can fully decompose the independent features of each variable. Combined with the infinite memory capability of fractional-order differentials, it enhances the ability to capture long-term dynamic features of the wastewater treatment process. By using single-integer-order reservoirs to fuse features in series, it effectively considers the coupling relationship between input variables, solving the problem that existing models either ignore coupling or have insufficient decomposition. It provides clear stability criteria to ensure that the initial parameters of the model are within a stable range, improving prediction reliability. It uses a gradient descent algorithm to optimize the reservoir parameters, further reducing prediction error. Experimental verification shows that compared with existing methods, the prediction RMSE of this invention is reduced by more than 5 times, and the MAPE is reduced to 0.83%, which can significantly reduce reagent waste and treatment costs.

[0125] To address the shortcomings of existing technologies in predicting ammonia nitrogen concentration in wastewater treatment, such as insufficient accuracy, poor model stability, and inadequate consideration of variable decomposition and coupling relationships, this invention provides an ammonia nitrogen concentration prediction method that improves prediction accuracy and stability, offering a reliable basis for optimized control of wastewater treatment processes. In other words, the ammonia nitrogen concentration prediction method provided in this specification aims to improve the prediction accuracy of ammonia nitrogen concentration in wastewater treatment processes, reducing reagent waste and increased treatment costs due to inaccurate predictions. This method constructs a variable-structure echo state network consisting of parallel multi-fractional-order reservoirs and cascaded single-integer-order reservoirs. It determines input features through variable correlation analysis, initializes model parameters based on stability criteria, and optimizes reservoir parameters using a gradient descent algorithm to achieve accurate ammonia nitrogen concentration prediction. This invention simultaneously considers the decomposition and coupling relationships of input variables, improving prediction stability and accuracy, and is suitable for real-time optimized control of wastewater treatment processes.

[0126] Corresponding to the above method embodiments, this specification also provides another embodiment of an ammonia nitrogen concentration prediction method, which includes:

[0127] Wastewater data from the wastewater treatment process is acquired, and an optimized variable structure echo state network model is used to predict the ammonia nitrogen concentration based on the wastewater data. The optimized variable structure echo state network model is determined according to the aforementioned ammonia nitrogen concentration prediction method. The wastewater data is data that can affect the ammonia nitrogen concentration, and the wastewater data can refer to the aforementioned historical data or characteristic factors.

[0128] Based on the aforementioned alternative ammonia nitrogen concentration prediction method, the ammonia nitrogen concentration can be accurately determined during wastewater treatment. Consequently, based on the accurately predicted ammonia nitrogen concentration, the dosage of reagents can be optimized, treatment costs can be reduced, and effluent can be guaranteed to meet standards.

[0129] The above is an illustrative scheme of another ammonia nitrogen concentration prediction method in this embodiment. It should be noted that the technical solution of this other ammonia nitrogen concentration prediction method belongs to the same concept as the technical solution of the ammonia nitrogen concentration prediction method described above. For details not described in detail in the technical solution of the other ammonia nitrogen concentration prediction method, please refer to the description of the technical solution of the ammonia nitrogen concentration prediction method described above.

[0130] Corresponding to the above method embodiments, this specification also provides embodiments of an ammonia nitrogen concentration prediction device. Figure 4 A schematic diagram of an ammonia nitrogen concentration prediction device according to one embodiment of this specification is shown. Figure 4 As shown, the device includes:

[0131] The historical data determination module 402 is configured to determine multiple characteristic factors affecting ammonia nitrogen concentration and collect historical data corresponding to each characteristic factor, wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration during the wastewater treatment process;

[0132] The training data determination module 404 is configured to determine the correlation coefficient between each feature factor and the ammonia nitrogen concentration, and based on the correlation coefficient, determine training data for the variable structure echo state network model from the historical data.

[0133] The model building module 406 is configured to build the variable structure echo state network model, wherein the variable structure echo state network model includes a fractional-order reservoir and an integer-order reservoir, the fractional-order reservoir is used for feature extraction, the integer-order reservoir is used for feature fusion, and the number of fractional-order reservoirs is determined according to the amount of training data.

[0134] The model optimization module 408 is configured to optimize the variable structure echo state network model using the training data, and to use the optimized variable structure echo state network model to predict ammonia nitrogen concentration during wastewater treatment.

[0135] Optionally, the training data determination module 404 is further configured to:

[0136] Based on the historical data, a training dataset for the variable structure echo state network model is constructed;

[0137] The correlation coefficient between each characteristic factor and the ammonia nitrogen concentration was calculated using sequence cross-correlation analysis.

[0138] Analyze the time correlation coefficient between the historical data and the ammonia nitrogen concentration;

[0139] Based on the correlation coefficient, multiple input features are selected from the training dataset, and based on the temporal correlation coefficient, the length of the historical time corresponding to each input feature is determined.

[0140] The multiple input features and the length of the historical time corresponding to each input feature are determined as the training data of the variable structure echo state network model.

[0141] Optionally, the fractional-order reserve pool consists of multiple parallel fractional-order reserve pools used to extract features from the training data and obtain data features. The number of the multiple fractional-order reserve pools is determined based on the number of input features in the training data.

[0142] The integer-order reserve pool is a single integer-order reserve pool connected in series, used for feature fusion of data features.

[0143] Optionally, the characteristic factors include pH value, water temperature, dissolved oxygen concentration, total phosphorus content, and total suspended solids content.

[0144] Optionally, the model optimization module 408 is further configured to:

[0145] Using the variable structure echo state network model, prediction processing is performed based on the training data to obtain the predicted ammonia nitrogen concentration.

[0146] Using the actual ammonia nitrogen concentration corresponding to the training data and the predicted ammonia nitrogen concentration, a loss function is determined, and the variable structure echo state network model is trained based on the loss function to obtain the trained variable structure echo state network model.

[0147] In the wastewater treatment process, the trained variable structure echo state network model is used to predict ammonia nitrogen concentration.

[0148] Optionally, the model optimization module 408 is further configured to:

[0149] The objective function is calculated using the actual ammonia nitrogen concentration corresponding to the training data and the predicted ammonia nitrogen concentration.

[0150] The gradient descent algorithm is used to calculate the derivative of the objective function with respect to the parameters of the reservoir, wherein the reservoir includes a fractional-order reservoir and an integer-order reservoir.

[0151] The parameters of the reservoir are updated based on the derivative until the model training stopping condition is met, thereby obtaining the trained variable structure echo state network model.

[0152] Optionally, the parameters of the reserve pool include the parameters of the fractional-order reserve pool and the parameters of the integer-order reserve pool;

[0153] The parameters of the fractional-order reservoir include the scaling factor, spectral radius, and input scaling factor of the fractional-order reservoir.

[0154] The parameters of the integer-order reservoir include the leakage rate, spectral radius, and feedback scaling factor of the integer-order reservoir.

[0155] Optionally, the training data includes multiple input features and the length of the historical time corresponding to each input feature;

[0156] The model optimization module 408 is further configured as follows:

[0157] From multiple parallel fractional-order reserve pools, a target fractional-order reserve pool corresponding to the target input feature is determined, wherein the target input feature is any one of the multiple input features, and the target fractional-order reserve pool is a reserve pool among the multiple fractional-order reserve pools used for feature extraction of the target input feature;

[0158] The target input features and the corresponding historical time lengths are input into the target fractional-order reserve pool for feature extraction, and the data features are output.

[0159] The data features output from the multiple fractional-order reservoirs are input into the integer-order reservoir for feature fusion to obtain fused features, and the predicted ammonia nitrogen concentration is obtained using the fused features.

[0160] The ammonia nitrogen concentration prediction device in the embodiments of this specification can construct a variable structure echo state network model including fractional-order and integer-order reservoirs. Furthermore, based on various characteristic factors affecting ammonia nitrogen concentration and the correlation coefficient between each characteristic factor and ammonia nitrogen concentration, training data for the variable structure echo state network model is determined from historical data. The variable structure echo state network model is optimized using the training data, thereby obtaining a variable structure echo state network model capable of accurately predicting ammonia nitrogen concentration. This ensures accurate determination of ammonia nitrogen concentration during wastewater treatment, thereby optimizing reagent dosage, reducing treatment costs, and ensuring effluent meets standards.

[0161] The above is a schematic scheme of an ammonia nitrogen concentration prediction device according to this embodiment. It should be noted that the technical solution of this ammonia nitrogen concentration prediction device and the technical solution of the ammonia nitrogen concentration prediction method described above belong to the same concept. For details not described in detail in the technical solution of the ammonia nitrogen concentration prediction device, please refer to the description of the technical solution of the ammonia nitrogen concentration prediction method described above.

[0162] Figure 5 A structural block diagram of a computing device 500 according to one embodiment of this specification is shown. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and a database 550 is used to store data.

[0163] The computing device 500 also includes an access device 540, which enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 540 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0164] In one embodiment of this specification, the above-described components of the computing device 500 and Figure 5 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 5 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0165] Computing device 500 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). Computing device 500 can also be a mobile or stationary server.

[0166] The processor 520 is used to execute the following computer program / instructions, which, when executed by the processor, implement the steps of the above-mentioned ammonia nitrogen concentration prediction method.

[0167] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the computing device embodiments are basically similar to the ammonia nitrogen concentration prediction method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the ammonia nitrogen concentration prediction method embodiments.

[0168] An embodiment of this specification also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the above-described ammonia nitrogen concentration prediction method.

[0169] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the computer-readable storage medium embodiments are described simply because they are substantially similar to the ammonia nitrogen concentration prediction method embodiments; relevant parts can be found in the descriptions of the ammonia nitrogen concentration prediction method embodiments.

[0170] An embodiment of this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described ammonia nitrogen concentration prediction method.

[0171] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the above-described ammonia nitrogen concentration prediction method belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the above-described ammonia nitrogen concentration prediction method.

[0172] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

[0173] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0174] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0175] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0176] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0177] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method of predicting ammonia nitrogen concentration, characterized by, The method comprises the following steps: determining a plurality of characteristic factors affecting the ammonia nitrogen concentration, and collecting historical data corresponding to each characteristic factor, wherein the ammonia nitrogen concentration is the ammonia nitrogen concentration in the sewage treatment process; determining the correlation coefficient between each characteristic factor and the ammonia nitrogen concentration, and determining the training data for the variable structure echo state network model from the historical data based on the correlation coefficient; constructing the variable structure echo state network model, wherein the variable structure echo state network model comprises fractional order reservoirs and integer order reservoirs, the fractional order reservoirs are used for feature extraction, the integer order reservoirs are used for feature fusion, and the number of fractional order reservoirs is determined according to the data quantity of the training data; optimizing the variable structure echo state network model by using the training data, and predicting the ammonia nitrogen concentration in the sewage treatment process by using the optimized variable structure echo state network model.

2. The method of claim 1, wherein, The method comprises the following steps: According to the historical data, a training data set of the variable structure echo state network model is constructed; The correlation coefficient between each characteristic factor and the ammonia nitrogen concentration is calculated by using sequence cross-correlation analysis; The time correlation coefficient between the historical data and the ammonia nitrogen concentration is analyzed; According to the correlation coefficient, a plurality of input features are selected from the training data set, and the historical time length corresponding to each input feature is determined according to the time correlation coefficient; The plurality of input features and the historical time length corresponding to each input feature are determined as the training data of the variable structure echo state network model.

3. The method of claim 1, wherein, The fractional order reservoirs are a plurality of fractional order reservoirs in parallel, which are used for feature extraction on the training data to obtain data features, wherein the number of fractional order reservoirs is determined according to the number of input features in the training data; The integer order reservoir is a single integer order reservoir in series, which is used for feature fusion on the data features.

4. The method according to any one of claims 1 to 3, characterized in that, The characteristic factors include pH value, water temperature, dissolved oxygen concentration, total phosphorus content and total suspended solid content.

5. The method of claim 1, wherein, The method comprises the following steps: The variable structure echo state network model is used to perform prediction processing according to the training data to obtain a predicted ammonia nitrogen concentration; The real ammonia nitrogen concentration corresponding to the training data and the predicted ammonia nitrogen concentration are used to determine a loss function, and the variable structure echo state network model is trained based on the loss function to obtain a trained variable structure echo state network model; In the sewage treatment process, the trained variable structure echo state network model is used to predict the ammonia nitrogen concentration.

6. The method of claim 5, wherein, The corresponding real ammonia nitrogen concentration and the predicted ammonia nitrogen concentration of the training data are used to determine a loss function, and the model training of the variable structure echo state network model is performed based on the loss function, and a trained variable structure echo state network model is obtained, comprising: The corresponding real ammonia nitrogen concentration and the predicted ammonia nitrogen concentration of the training data are used to calculate an objective function; The gradient descent algorithm is used to calculate the derivative of the objective function with respect to the parameters of the reservoir pool, wherein the reservoir pool includes a fractional order reservoir pool and an integer order reservoir pool; Based on the derivative, the parameters of the reservoir pool are updated until the model training stopping condition is reached, and the trained variable structure echo state network model is obtained.

7. The method of claim 6, wherein, The parameters of the reservoir pool include the parameters of the fractional order reservoir pool and the parameters of the integer order reservoir pool; The parameters of the fractional order reservoir pool include the scaling factor, the spectral radius, and the input scaling factor of the fractional order reservoir pool; The parameters of the integer order reservoir pool include the leakage rate, the spectral radius, and the feedback scaling factor of the integer order reservoir pool.

8. The method of claim 5, wherein, The training data includes a plurality of input features and a corresponding historical time length of each input feature; The variable structure echo state network model is used to perform prediction processing based on the training data to obtain a predicted ammonia nitrogen concentration, comprising: From the parallel multiple fractional order reservoir pools, a target fractional order reservoir pool corresponding to a target input feature is determined, wherein the target input feature is any one of the multiple input features, and the target fractional order reservoir pool is a reservoir pool in the multiple fractional order reservoir pools used for feature extraction of the target input feature; The target input feature and the historical time length corresponding to the target input feature are input into the target fractional order reservoir pool for feature extraction to obtain data features output by the target fractional order reservoir pool; The data features output by the multiple fractional order reservoir pools are input into the integer order reservoir pool for feature fusion to obtain fused features, and the fused features are used to obtain a predicted ammonia nitrogen concentration.

9. A computing device, comprising: Comprising: a memory and a processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which realize the steps of the method of any one of claims 1 to 8 when executed by the processor.

10. A computer-readable storage medium, characterized in that, It stores computer programs / instructions, which realize the steps of the method of any one of claims 1 to 8 when executed by the processor.

Images