Water quality prediction method, parameter recommendation, optimization method, and model training method

Abstract

The present disclosure provides a water quality prediction method, parameter recommendation, optimization method and model training method, relates to the field of artificial intelligence, in particular to the technical field of intelligent industry and deep learning, and can be applied to sewage treatment and the like. The specific implementation scheme of the water quality prediction method is as follows: a main network of a water quality prediction model is used to process water quality parameters of inflow water to obtain water quality feature data; a first branch network of the water quality prediction model is used to process working condition parameters of a sewage treatment system to obtain working condition feature data; a second branch network of the water quality prediction model is used to process model parameters of a water quality mechanism model to obtain parameter feature data; and the water quality feature data is weighted processed according to the working condition feature data and the parameter feature data to obtain water quality parameters of outflow water, wherein the water quality prediction model is constructed based on a deep operator network.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of artificial intelligence, in particular to the technical field of intelligent industry and deep learning, and can be applied to scenarios such as sewage treatment. BACKGROUND

[0002] With the development of electronic technology, deep learning technology has been widely applied in many fields. In order to improve the intelligent level of industry, deep learning technology can be combined with traditional industry to predict the operation results of working conditions based on deep learning technology, and to provide reference information for the operation of working conditions. SUMMARY

[0003] The present disclosure aims to provide a water quality prediction method, a parameter recommendation method, a parameter optimization method, a model training method, a device, an apparatus, and a medium, and aims to improve the learning ability of deep learning on water treatment mechanism, improve the water quality prediction accuracy, and further guide the operation of working conditions.

[0004] According to a first aspect of the present disclosure, a water quality prediction method is provided, comprising: processing the water quality parameters of the influent by using the backbone network of a water quality prediction model to obtain water quality feature data; processing the working condition parameters of the sewage treatment system by using the first branch network of the water quality prediction model to obtain working condition feature data; processing the model parameters of the water quality mechanism model by using the second branch network of the water quality prediction model to obtain parameter feature data; and weighting the water quality feature data according to the working condition feature data and the parameter feature data to obtain the water quality parameters of the effluent, wherein the water quality prediction model is constructed based on a deep operator network.

[0005] According to a second aspect of the present disclosure, a parameter recommendation method is provided, comprising: predicting the water quality parameters of the effluent by using the water quality prediction method described above; determining the value of the water quality mechanism model according to the water quality parameters of the influent, the working condition parameters of the sewage treatment system, the model parameters, and the water quality parameters of the effluent; and adjusting the working condition parameters to obtain recommended working condition parameters, with the goal of minimizing the value of the water quality mechanism model.

[0006] According to a third aspect of the present disclosure, a parameter optimization method is provided, comprising: predicting the water quality parameters of the effluent by using the water quality prediction method described above; determining the actual water quality parameters of the effluent obtained by processing the influent with the water quality parameters of the influent according to the working condition parameters of the sewage treatment system; and adjusting the value of the model parameters according to the difference between the actual water quality parameters and the predicted water quality parameters to obtain optimized model parameters.

[0007] According to a fourth aspect of the present disclosure, a training method of a water quality prediction model is provided, wherein the water quality prediction model comprises a backbone network, a first branch network, a second branch network and a fusion network arranged in parallel; the water quality prediction model is constructed based on a deep operator network; the training method comprises: processing water quality parameters of influent in sample data by using the backbone network to obtain water quality feature data; processing working condition parameters of a sewage treatment system in the sample data by using the first branch network to obtain working condition feature data; processing model parameters of a water quality mechanism model in the sample data by using the second branch network to obtain parameter feature data; performing weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data by using the fusion network to obtain predicted water quality parameters of effluent; and training the water quality prediction model according to differences between the water quality parameters of the effluent in the sample data and the predicted water quality parameters.

[0008] According to a fifth aspect of the present disclosure, a water quality prediction device is provided, comprising: a water quality parameter processing module configured to process water quality parameters of influent by using a backbone network of a water quality prediction model to obtain water quality feature data; a working condition parameter processing module configured to process working condition parameters of a sewage treatment system by using a first branch network of the water quality prediction model to obtain working condition feature data; a model parameter processing module configured to process model parameters of a water quality mechanism model by using a second branch network of the water quality prediction model to obtain parameter feature data; and a water quality parameter prediction module configured to perform weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data to obtain water quality parameters of effluent, wherein the water quality prediction model is constructed based on a deep operator network.

[0009] According to a sixth aspect of the present disclosure, a parameter recommendation device is provided, comprising: a water quality parameter prediction module configured to predict water quality parameters of effluent by using the water quality prediction device provided in the fifth aspect of the present disclosure; a model value determination module configured to determine values of a water quality mechanism model according to the water quality parameters of influent, the working condition parameters of the sewage treatment system, the model parameters and the water quality parameters of effluent; and a recommended parameter determination module configured to adjust the working condition parameters to obtain recommended working condition parameters with the objective of minimizing the values of the water quality mechanism model.

[0010] According to a seventh aspect of the present disclosure, a parameter optimization device is provided, comprising: a water quality parameter prediction module configured to predict water quality parameters of effluent by using the water quality prediction device provided in the fifth aspect of the present disclosure; an actual parameter determination module configured to determine actual water quality parameters of effluent obtained by processing influent with the water quality parameters of influent by the sewage treatment system according to the working condition parameters; and a parameter optimization module configured to adjust values of model parameters according to differences between the actual water quality parameters and the predicted water quality parameters to obtain optimized model parameters.

[0011] According to an eighth aspect of the present disclosure, a training device of a water quality prediction model is provided, wherein the water quality prediction model comprises a backbone network, a first branch network, a second branch network and a fusion network arranged side by side; the water quality prediction model is constructed based on a deep operator network; the training device comprises: a water quality parameter processing module configured to process water quality parameters of influent in sample data by using the backbone network to obtain water quality feature data; a working condition parameter processing module configured to process working condition parameters of a sewage treatment system in the sample data by using the first branch network to obtain working condition feature data; a model parameter processing module configured to process model parameters of a water quality mechanism model in the sample data by using the second branch network to obtain parameter feature data; a water quality parameter prediction module configured to perform weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data by using the fusion network to obtain predicted water quality parameters of effluent; and a model training module configured to train the water quality prediction model according to differences between the water quality parameters of the effluent in the sample data and the predicted water quality parameters.

[0012] According to a ninth aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform at least one of the following methods provided by the present disclosure: a water quality prediction method, a parameter recommendation method, a parameter optimization method and a training method of a water quality prediction model.

[0013] According to a tenth aspect of the present disclosure, a non-transitory computer readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to perform at least one of the following methods provided by the present disclosure: a water quality prediction method, a parameter recommendation method, a parameter optimization method and a training method of a water quality prediction model.

[0014] According to an eleventh aspect of the present disclosure, a computer program product is provided, comprising computer programs / instructions stored on at least one of a readable storage medium and an electronic device, and the computer programs / instructions, when executed by a processor, implement at least one of the following methods provided by the present disclosure: a water quality prediction method, a parameter recommendation method, a parameter optimization method and a training method of a water quality prediction model.

[0015] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are used to better understand the present scheme, and do not limit the present disclosure. Among them:

[0017] Figure 1 is an application scenario schematic diagram of a water quality prediction method, a parameter recommendation method, a parameter optimization method, and a training method of a water quality prediction model according to an embodiment of the present disclosure;

[0018] Figure 2 is a flow schematic diagram of a water quality prediction method according to an embodiment of the present disclosure;

[0019] Figure 3 is an implementation principle diagram of a water quality prediction method according to an embodiment of the present disclosure;

[0020] Figure 4 is a flow schematic diagram of a parameter recommendation method according to an embodiment of the present disclosure;

[0021] Figure 5 is a flow schematic diagram of a parameter optimization method according to an embodiment of the present disclosure;

[0022] Figure 6 is a flow schematic diagram of a training method of a water quality prediction model according to an embodiment of the present disclosure;

[0023] Figure 7 is a structural block diagram of a water quality prediction device according to an embodiment of the present disclosure;

[0024] Figure 8 is a structural block diagram of a parameter recommendation device according to an embodiment of the present disclosure;

[0025] Figure 9 is a structural block diagram of a parameter optimization device according to an embodiment of the present disclosure;

[0026] Figure 10 is a structural block diagram of a training device of a water quality prediction model according to an embodiment of the present disclosure;

[0027] Figure 11 is a block diagram of an electronic device for implementing the method of the present disclosure. DETAILED DESCRIPTION

[0028] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the present disclosure to assist in understanding. They should be considered in their context only as exemplary. Thus, those of ordinary skill in the art will recognize various changes and modifications to the embodiments described herein, without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and structures are omitted in the following description for clarity and conciseness.

[0029] The professional terms related to the present disclosure are described in detail as follows:

[0030] Mechanism model (white box) : a model that exists a physical relationship, which is expressed by physical or mathematical formula. In the present disclosure, the water quality mechanism model refers to a physical relationship between the working condition parameters, the water quality parameters of the inlet water and the water quality parameters of the outlet water, wherein the physical relationship between the working condition parameters, the water quality parameters of the inlet water and the water quality parameters of the outlet water is associated via model parameters.

[0031] Deep Operator Network, DeepONet for short. Shallow networks can be seen as operators, and since they involve neural networks, they are also called neural operators. DeepONet is a deep neural network architecture that can learn neural operators and solve multiple partial differential equations at once. DeepONet is a bifurcated architecture that processes data in parallel networks. Parallel networks include a trunk network and a branch network, the former learns to estimate many functions on the input side, and the latter learns to estimate functions on the output side. DeepONet combines the outputs of parallel networks to learn the operators required for partial differential equations. The process of training DeepONet involves repeatedly presenting input, output data generated using a set of partial differential equations with a numerical solver, and adjusting the weights in the branch network and the trunk network in each iteration until the amount of error that appears throughout the network can be accepted.

[0032] Water system is the blood of the city, and water is a valuable resource for human survival. With the development of economy, the amount of sewage discharge and treatment is increasing, and the cost of sewage treatment is high. How to improve the sewage treatment capacity, control the sewage treatment compliance rate, and achieve economic energy saving and reduce operation cost has become the top priority of sewage treatment.

[0033] Anaerobic-Anoxic-Oxic (A2 / O) process is widely used in urban sewage treatment due to its good nitrogen and phosphorus removal effect.

[0034] In the project of improving sewage treatment capacity and controlling sewage treatment compliance rate, the process of modeling the sewage treatment process is essential. For example, the following three ways can be used to model the sewage treatment process.

[0035] The first one is based on mechanism model. Since 1950s, the understanding of activated sludge and its purification function has been deepened and the theory of microorganism has been developed. Some scholars have proposed semi-empirical and semi-theoretical dynamic models to describe the relationship between the degradation rate of organic substrate, the proliferation rate of microorganisms and the oxygen consumption rate in activated sludge system and the relevant parameters. Among them, Eckenfelder, Lawrence-McCarty and McKinney models are representative. These three models all adopt the "growth-decay" mechanism, which greatly simplifies the actual biochemical reaction system and does not effectively describe the dynamic characteristics of the activated sludge system. The International Water Association (IWA) summarized the previous research results and proposed three sets of ASM models, which take into account the dynamic phenomena of "rapid removal of organic matter" and the change of microorganism growth rate, and can predict the degradation of nitrogen and phosphorus. Most of the models used in practice are improved and simplified on the basis of IWA models.

[0036] The second one is based on intelligent model. For example, a series of intelligent modeling methods such as neural network are suitable for systems whose mechanism is not well understood or cannot be represented by mechanism model. Through learning a large amount of raw data, the required nonlinear shape is achieved to simulate the object model, which has great flexibility and adaptability. Neural network itself is nonlinear, which can play an important role in the complex and nonlinear wastewater treatment process of activated sludge method. The main modeling methods include fuzzy modeling, fuzzy neural network modeling, time delay neural network modeling, recurrent neural network modeling, BP neural network modeling and RBF radial basis function neural network modeling, among which the literature of BP network wastewater treatment process model is the most.

[0037] The third one is based on fusion model. The idea of fusion model is to use a simplified mechanism model to preliminarily model the wastewater treatment process, and then use intelligent models such as neural network to predict the error between the simulation results of mechanism model and experimental test results. According to the predicted error, the simulation results of mechanism model are compensated in time and effectively to improve the prediction results. This method combines the advantages of mechanism model and neural network model, uses the good ability of neural network to handle nonlinear complex process to solve the problem of low prediction accuracy of simplified mechanism model for actual activated sludge system. At the same time, the objective mechanism of mechanism model improves the generalization ability and robustness of the model, making it suitable for general problems.

[0038] But the sewage tank process is a complex biochemical reaction process, with high nonlinearity, time-varying, uncertainty and time delay characteristics, process modeling and control optimization is more difficult than other industrial processes. Although the existing mechanism model can describe the dynamic characteristics of biochemical process, many quantitative relationships in the model are obtained by experience, there are many unknown parameters, and there are uncertainties in the changing conditions, which leads to low calculation accuracy of the model, and is not suitable for prediction and control of variable water quality and working conditions. The accuracy of the method using intelligent model depends largely on the number, comprehensiveness and reliability of the historical data, and the adaptability to variable water quality and different sewage plants cannot be guaranteed.

[0039] In order to solve this problem, the present disclosure provides a water quality prediction method, a parameter recommendation method, a parameter optimization method and a model training method, device, equipment and medium. The following will be combined with Figure 1 The application scenario of the method and device provided by the present disclosure is described.

[0040] Figure 1 The application scenario of the water quality prediction method, the parameter recommendation method, the parameter optimization method and the model training method, device according to the embodiments of the present disclosure is shown in the figure.

[0041] As Figure 1 The application scenario 100 of the embodiments can include an electronic device 110, which can be various electronic devices with processing functions, including but not limited to smartphones, tablet computers, laptop computers, desktop computers and servers, etc.

[0042] The electronic device 110 can be installed with various client applications, such as water quality prediction applications, working condition monitoring applications of sewage treatment systems, instant messaging applications, etc., which are not limited by the present disclosure.

[0043] The electronic device 110 can process the working condition parameters 131 used by the sewage treatment system 120, the water quality parameters 132 of the inflow, and the model parameters 133 of the water quality mechanism model, for example, to predict the water quality parameters 140 of the effluent. The water quality parameters 140 of the effluent can be used as a basis for determining whether the operation of the sewage treatment system 120 can meet the sewage treatment standard.

[0044] In an embodiment, the electronic device 110 can employ the water quality prediction model 160 pre-trained by the server 150 to process the working condition parameter 131, the water quality parameter 132 of the incoming water, and the model parameter 133 of the water quality mechanism model, to obtain the water quality parameter 140 of the outgoing water. The electronic device 110 can be communicatively connected to the server 150 through a network, for example. The network can include wired or wireless communication links. For example, the server 150 can be a background management server that provides support for the running of a client application installed in the electronic device 110, or can be a cloud server or a blockchain server, etc., without limitation in the present disclosure. The water quality prediction model 160 can be a model trained by the training method provided in the present disclosure.

[0045] In an embodiment, the electronic device 110 can also send the working condition parameter 131, the water quality parameter 132 of the incoming water, and the model parameter 133 of the water quality mechanism model to the server 150 via the network, and obtain the water quality parameter 140 of the outgoing water by processing by the server 150.

[0046] It should be noted that the water quality prediction method, the parameter recommendation method, and the parameter optimization method provided in the present disclosure can be executed by the electronic device 110 or the server 150. Correspondingly, the water quality prediction apparatus, the parameter recommendation apparatus, and the parameter optimization apparatus provided in the present disclosure can be arranged in the electronic device 110 or the server 150. The training method of the water quality prediction model provided in the present disclosure can be executed by the server 150. Correspondingly, the training apparatus of the water quality prediction model provided in the present disclosure can be arranged in the server 150.

[0047] It should be understood that Figure 1 The number and type of the electronic device 110 and the server 150 in the present disclosure are only illustrative. Any number and type of electronic device 110 and server 150 can be provided according to the needs of implementation.

[0048] The water quality prediction method provided in the present disclosure will be described in detail below. Figures 2-3 The water quality prediction method provided in the present disclosure will be described in detail below.

[0049] Figure 2 is a flowchart of the water quality prediction method according to an embodiment of the present disclosure.

[0050] As shown in Figure 2 , the water quality prediction method 200 of the embodiment can include operation S210 to operation S240.

[0051] In operation S210, the water quality parameter of the incoming water is processed by employing the backbone network of the water quality prediction model to obtain water quality feature data.

[0052] According to an embodiment of the present disclosure, the water quality prediction model is constructed based on a deep operator network. The water quality prediction model comprises a trunk network and a branch network. The trunk network may, for example, be a feedforward neural network. Specifically, the trunk network may, for example, be constituted by a backpropagation neural network or a convolutional neural network and a fully connected feedforward neural network, and the present disclosure does not limit the same.

[0053] According to an embodiment of the present disclosure, the water quality parameters of the influent may, for example, comprise at least one of the following parameters: chemical oxygen demand, biochemical oxygen demand, total nitrogen content, total phosphorus content, and the like. The chemical oxygen demand refers to the amount of oxidizing agent consumed when a water sample is treated with a certain strong oxidizing agent under certain conditions. The biochemical oxygen demand refers to the amount of dissolved oxygen consumed in the biochemical reaction process of microbial decomposition of biodegradable organic matter present in the water under certain conditions. The total nitrogen content of the influent is the total amount of inorganic and organic nitrogen in various forms in the influent. The total phosphorus content of the influent is the sum of inorganic and organic phosphorus present in the influent.

[0054] In this embodiment, the trunk network is used to encode the water quality parameters of the influent, thereby obtaining water quality feature data. Specifically, the water quality parameters of the influent are input into the trunk network, and the trunk network outputs the water quality feature data.

[0055] In operation S220, the first branch network of the water quality prediction model is used to process the operating condition parameters of the sewage treatment system, thereby obtaining operating condition feature data.

[0056] According to an embodiment of the present disclosure, the first branch network may, for example, be a feedforward neural network, and the network structure adopted by the first branch network may be similar to that of the trunk network, except that the trunk network is used to encode the output function domain, while the first branch network is used to encode the discrete input function space. When the operating condition parameters are multiple, the first branch network may, for example, be provided with sub-networks for encoding the multiple operating condition parameters, respectively, to learn the features of each operating condition parameter distributed discretely.

[0057] According to an embodiment of the present disclosure, the sewage treatment system may, for example, be constituted by multiple reaction tanks. The multiple reaction tanks may, for example, comprise an initial sedimentation tank, an anaerobic tank, an aerobic tank, a secondary sedimentation tank, a filter tank, and the like connected in sequence, and the present disclosure does not limit the same. The operating condition parameters may, for example, comprise at least one of the following parameters: internal reflux ratio, external reflux ratio, oxidation-reduction potential of the anaerobic tank, oxidation-reduction potential of the anoxic tank, and the like, and the present disclosure does not limit the same.

[0058] In this embodiment, the operating condition parameters are input into the first branch network, and the first branch network encodes the operating condition parameters, thereby outputting operating condition feature data after encoding.

[0059] At operation S230, the second branch network adopting the water quality prediction model processes the model parameters of the water quality mechanism model to obtain parameter feature data.

[0060] According to an embodiment of the present disclosure, the second branch network is similar to the first branch network, and the two branch networks are arranged side by side.

[0061] The water quality mechanism model associates the working condition parameters, the water quality parameters of the influent and the water quality parameters of the effluent via the model parameters. In essence, the model parameters can be understood as undetermined coefficients involved in expressing the relationship between the working condition parameters, the water quality parameters of the influent and the water quality parameters of the effluent changing over time. The model parameters are usually determined according to the actual wastewater treatment process. For example, the A2 / O process based on the combination of engineering principles and microbial reflection mechanism includes processes such as carbon oxidation, nitrification and denitrification, biological phosphorus removal and solid-liquid separation. The water quality mechanism model usually takes the influent sludge load of a certain substrate as the core parameter, considers that the reaction of microorganisms and the substrate basically follows first-order reaction kinetics, and focuses on the relationship between the sludge load and the substrate removal rate. For example, the water quality mechanism model can be expressed by the algebraic differential equation shown in the following formula (1).

[0062] N(t, x, y; θ, K φ ) = 0 Formula (1)

[0063] Where t is time, x is the water quality parameter of the influent, y is the water quality parameter of the effluent, θ is the working condition parameter of the wastewater treatment system, and K φ is the undetermined coefficient, i.e., the model parameter. The water quality parameters of the effluent and the water quality parameters of the influent in the water quality mechanism model can use the conventional parameters in wastewater treatment tests, and the model parameters can be biological reaction kinetics parameters and engineering experience parameters. The water quality mechanism model can be used to simulate the biological reaction process of the wastewater treatment process unit. Under the condition that the water quality parameters of the influent and the working condition parameters are known, the water quality parameters of the effluent can be predicted, so as to simulate the change of water quality in the reaction process in the treatment unit. Based on the water quality mechanism model, the process operation conditions of the wastewater treatment system can also be optimized under the condition that the water quality parameters of the influent and the water quality requirements of the effluent are known.

[0064] In this embodiment, the model parameters set according to artificial experience can be input into the second branch network, and the model parameters are encoded by the second branch network, and the parameter feature data is output after encoding.

[0065] At operation S240, the water quality feature data is weighted processed according to the working condition feature data and the parameter feature data to obtain the water quality parameters of the effluent.

[0066] In this embodiment, both the working condition characteristic data and the parameter characteristic data can be used as adjustment factors to adjust the water quality characteristic data. For example, the working condition characteristic data, the parameter characteristic data, and the water quality characteristic data can all be represented by vectors. In this embodiment, the working condition characteristic data and the parameter characteristic data can be multiplied element by element, that is, the Hadamard product of the working condition characteristic data and the parameter characteristic data can be calculated, and then the inner product of the calculated Hadamard product and the water quality characteristic data can be calculated, so as to obtain the water quality parameter of the effluent. It can be understood that the vectors representing the working condition characteristic data, the parameter characteristic data, and the water quality characteristic data have the same dimension.

[0067] The water quality prediction method of the embodiments of the present disclosure can improve the accuracy of the predicted water quality parameter of the effluent by using the water quality prediction model constructed based on the deep operator network. This is because the deep operator network can fully learn the algebraic differential expression of the water quality mechanism model to establish an accurate water quality prediction model. Furthermore, by combining the idea of deep learning, the nonlinear relationship between the water quality parameter and the working condition parameter can be fully learned.

[0068] In an embodiment, the water quality parameter of the influent can include a plurality of first water quality parameters, for example, N first water quality parameters, and correspondingly, the predicted water quality parameter of the effluent is N water quality parameters of the effluent corresponding to the N first water quality parameters, which can be represented by N second water quality parameters.

[0069] In this embodiment, the backbone network can process the N first water quality parameters to obtain N characteristic data corresponding to the N first water quality parameters. Each of the N characteristic data is obtained by processing the N first water quality parameters. For example, the backbone network can encode each first water quality parameter according to the correlation between the remaining (N-1) first water quality parameters and the each first water quality parameter, so as to obtain the characteristic data corresponding to the each first water quality parameter.

[0070] For example, the backbone network is a multi-input and multi-output feedforward neural network, the input is the N first water quality parameters, and the output can include N P-dimensional vectors. When weighting the water quality characteristic data, each characteristic data (i.e., each P-dimensional vector) in the N characteristic data can be weighted according to the working condition characteristic data and the parameter characteristic data, so as to obtain a second water quality parameter.

[0071] For example, the working condition characteristic data and the parameter characteristic data can both be P-dimensional vectors. In this embodiment, the Hadamard product of the working condition characteristic data and the parameter characteristic data can be calculated, and then the inner product of the Hadamard product and each characteristic data can be calculated, so as to obtain a second water quality parameter.

[0072] In this way, the simultaneous prediction of multiple different water quality parameters can be achieved, thereby improving the efficiency of water quality prediction. Moreover, the correlation between the water quality parameters of the multiple influents is considered in the prediction of the water quality parameters of each effluent, thereby improving the accuracy of the predicted water quality parameters of the effluent.

[0073] Figure 3 FIG. 1 is a schematic diagram of a water quality prediction method according to an embodiment of the present disclosure.

[0074] In an embodiment, the operating condition parameters can be multiple, so as to more realistically simulate the sewage treatment process. The model parameters can be multiple, so as to better describe the nonlinear relationship between the parameters in the sewage treatment process.

[0075] For example, the operating condition parameters can be M, and generally, the multiple operating condition parameters are independent of each other. In order to better learn the features of different operating condition parameters, in this embodiment, M first sub-networks can be arranged in the first branch network to encode the M operating condition parameters in parallel and one-to-one correspondence. Wherein, M is a natural number greater than 1.

[0076] For example, the model parameters can be L, and generally, the multiple model parameters are independent of each other. In order to better learn the features of different model parameters, in this embodiment, L second sub-networks can be arranged in the second branch network to encode the L model parameters in parallel and one-to-one correspondence. Wherein, L is a natural number greater than 1. It can be understood that according to actual needs, the value of L and the value of M can be the same or different, and the present disclosure does not limit this.

[0077] In the case where the water quality parameters of the influent include N first water quality parameters, the principle of water quality prediction in this embodiment 300 can be as shown in Figure 3 .

[0078] In this embodiment 300, the water quality prediction model includes a trunk network 310, a first branch network 320 and a second branch network 330, and the three networks are arranged in parallel. Wherein, the first branch network 320 includes M first sub-networks, and each first sub-network is set to adopt a feedforward neural network, then the first branch network 320 can include FNN1~FNNM arranged side by side. M The second branch network 330 includes L second sub-networks, and each second sub-network is set to adopt a feedforward neural network, then the second branch network 330 can include FNN1~FNNL arranged side by side. L .

[0079] In this embodiment 300, the N first water quality parameters x1~xN included in the water quality parameters 301 of the influent can be encoded by the first branch network 320, and the L model parameters y1~yL included in the water quality parameters 302 of the effluent can be encoded by the second branch network 330.N The parameters are input into the backbone network 310 in the form of a sequence of parameters. The backbone network 310 outputs a vector sequence containing N vectors. Each of the N vectors can be a P-dimensional vector, and the N vectors represent N feature data.

[0080] In this embodiment 300, M first subnetworks FNN1 to FNN can be used. M A one-to-one correspondence for M operating condition parameters θ1~θ M The data is processed so that each first sub-network outputs one working condition feature data, resulting in a total of M working condition feature data. Specifically, the M parameters θ1 to θ2 in the working condition parameter 302 can be processed. M The inputs are fed one-to-one into M first subnetworks FNN1 to FNN2. M Thus, the first branch network 320 outputs M operating condition feature data. Each operating condition feature data can be, for example, a P-dimensional vector.

[0081] In this embodiment 300, L second subnetworks FNN1 to FNN can be used. L One-to-one correspondence for L model parameters K Φ1 ~K ΦL The data is processed so that each second sub-network outputs one parameter feature data, resulting in a total of L parameter feature data. Specifically, the L parameters K in model parameter 303 can be... Φ1 ~K ΦL The inputs are fed one-to-one into L second subnetworks FNN1 to FNN2. L The second branch network 330 then outputs L parameter feature data. Each parameter feature data can be, for example, a P-dimensional vector.

[0082] After obtaining N feature data, M operating condition feature data, and L parameter feature data, the water quality prediction model can, for example, determine the dot product of each feature data, the M operating condition feature data, and the L parameter feature data for each feature data to obtain a second water quality parameter. For instance, this embodiment can calculate the product of the i-th element in each feature data with the M i-th elements in the M operating condition feature data and the L i-th elements in the L parameter feature data, calculating a total of P products for each feature data. The sum of these P products is then used as the i-th water quality parameter y among the N second water quality parameters. pre，i Similarly, we can obtain N effluent water quality parameters 304, which is the second water quality parameter y. pre，1 ~y pre，N .

[0083] In one embodiment, if t is used ik (x1, x2, ..., x Nrepresents the kth element of the ith characteristic data in N characteristic data, using b jk (θ j represents the kth element of the jth working condition characteristic data in M working condition characteristic data, using q jk (K Φj represents the kth element of the jth parameter characteristic data in L parameter characteristic data, the ith second water quality parameter can be calculated using the following formula (2), for example.

[0084]

[0085] Wherein, N can be a natural number, M, L is a natural number greater than 1, P is a natural number greater than 1. The present disclosure does not limit the value of N, M, L, P.

[0086] Based on the water quality prediction method provided by the present disclosure, the present disclosure also provides a parameter recommendation method to recommend more reasonable working condition parameters. The following will be combined with Figure 4 to describe the parameter recommendation method in detail.

[0087] Figure 4 is a flowchart of the parameter recommendation method according to the embodiments of the present disclosure.

[0088] As Figure 4 shown, the parameter recommendation method 400 of this embodiment can include operation S410 to operation S430.

[0089] In operation S410, the water quality parameter of the effluent is predicted using the water quality prediction method.

[0090] The operation S410 can include the operation S210 to operation S240 described above, which will not be repeated here. Wherein, the water quality parameter of the influent can be the real-time water quality parameter of the influent of the sewage treatment system, the working condition parameter is the real-time working condition parameter of the sewage treatment system. The model parameter is a pre-set parameter.

[0091] In operation S420, the value of the water quality mechanism model is determined according to the water quality parameter of the influent, the working condition parameter of the sewage treatment system, the model parameter and the water quality parameter of the effluent.

[0092] According to the embodiments of the present disclosure, the water quality parameter x of the influent, the working condition parameter θ of the sewage treatment system, the model parameter K Φ , the current time t and the predicted water quality parameter y of the effluent in operation S410 can be brought into the left side of the formula (1) described above, and the value of the function on the left side of the formula (1) can be calculated. The value is taken as the value of the water quality mechanism model.

[0093] In operation S430, the working condition parameters are adjusted to minimize the value of the water quality mechanism model, and recommended working condition parameters are obtained.

[0094] In this embodiment, the gradient of the water quality mechanism model with respect to the working condition parameters can be calculated, and the working condition parameters θ can be adjusted using formula (3) as shown below. Wherein, α represents the rate of gradient descent, and the value of α can be set according to actual needs, for example, it can be 10, 20, etc., and the present disclosure does not limit it. It should be noted that the value of the water quality mechanism model is greater than or equal to 0. The operation S430 adjusts the working condition parameters by minimizing the value of the water quality mechanism model, which can make the value of the water quality mechanism model tend to 0 under the action of the adjusted working condition parameters. Thus, the operation of the sewage treatment system is more in line with the constraints of the water quality mechanism model.

[0095]

[0096] In an embodiment, during the adjustment of the working condition parameters, the water quality parameters of the effluent can also be constrained, so that the water quality of the effluent obtained when the sewage treatment system operates according to the adjusted working condition parameters can meet the standards. For example, during the adjustment of the working condition parameters, the adjustment is performed under the condition that the water quality parameters of the effluent are less than a predetermined value. In this way, the processing capacity of the sewage treatment system can be ensured while the operation of the sewage treatment system is more in line with the constraints of the water quality mechanism model, which is beneficial to improve the rationality of the operation of the sewage treatment system.

[0097] In an embodiment, when adjusting the working condition parameters, the resource consumption of the sewage treatment system can also be considered, and the independent variable of the function of the resource consumption is the working condition parameters. Therefore, in this embodiment, when determining the recommended working condition parameters, the resource consumption of the sewage treatment system can also be determined according to the working condition parameters first. For example, the function ε(θ) can be used to calculate the resource consumption of the sewage treatment system. In this embodiment, any function ε(θ) can be selected to calculate the resource consumption according to actual needs, and the present disclosure does not limit it.

[0098] In this embodiment, the working condition parameters can be adjusted to minimize the weighted sum of the value of the water quality mechanism model and the resource consumption. Wherein, the weights used when calculating the weighted sum can be set according to actual needs, and the present disclosure does not limit it. For example, the following formula (4) can be used to determine the recommended working condition parameters θ * .

[0099]

[0100] Wherein, is the value of the water quality mechanism model, that is, Wherein, y preThis refers to the water quality parameters of the effluent obtained by operating S410, which are the predicted water quality parameters.

[0101] In this embodiment, by simultaneously considering the resource consumption of the wastewater treatment system when adjusting the operating parameters, the recommended operating parameters can make the operation of the wastewater treatment system more in line with the constraints of the water quality mechanism model (i.e., more in line with the dynamic model), and also make the operation of the wastewater treatment system consume less resources, thus achieving the effect of energy saving and environmental protection.

[0102] Based on the water quality prediction method provided in this disclosure, this disclosure also provides a parameter optimization method to make the model parameter values ​​more consistent with actual needs and further improve the accuracy of water quality prediction. The following will combine... Figure 5 The method for optimizing this parameter is described in detail.

[0103] Figure 5 This is a flowchart illustrating a parameter optimization method according to an embodiment of the present disclosure.

[0104] like Figure 5 As shown, the parameter optimization method 500 of this embodiment may include operations S510 to S530.

[0105] In operation S510, the water quality parameters of the effluent are predicted using a water quality prediction method and used as the predicted water quality parameters.

[0106] The implementation principle of operation S510 is similar to that of operation S410, and will not be repeated here.

[0107] In operation S520, the wastewater treatment system is determined to process the influent with influent water quality parameters according to the operating parameters, and the resulting effluent water quality parameters are obtained.

[0108] This embodiment can acquire the water quality parameters of the influent flowing into the sewage treatment system, the operating parameters of the sewage treatment system, and the water quality parameters of the effluent obtained after the influent has been treated by the sewage treatment system. By operating step S510, the acquired influent water quality parameters, operating parameters, and pre-set model parameters are processed to obtain the predicted water quality parameters of the effluent. The acquired effluent water quality parameters are then used as the actual water quality parameters.

[0109] When operating S530, the model parameters are adjusted based on the difference between the actual and predicted water quality parameters to obtain the optimized model parameters.

[0110] According to an embodiment of the present disclosure, the values of the model parameters can be adjusted to minimize the difference between the actual water quality parameters and the predicted water quality parameters. For example, the difference between the water quality parameters can be represented by the difference between the water quality parameters, the mean square error, the minimum mean square error, etc., which are not limited by the present disclosure. This embodiment can first calculate the gradient of the difference with respect to the model parameters, and then adjust the model parameters using a formula similar to formula (3) described above. In an embodiment, when the water quality parameters are N, this embodiment can determine the difference between the actual water quality parameters and the predicted water quality parameters using, for example, the following formula (5) The difference can be understood as the loss value of the predicted water quality parameters.

[0111]

[0112] where y i is the i-th actual water quality parameter of the N actual water quality parameters, and y pre，i is the i-th predicted water quality parameter of the N predicted water quality parameters.

[0113] Through this embodiment, the predicted water quality parameters obtained based on the optimized model parameters can be consistent with the values of the actual water quality parameters obtained by the wastewater treatment system, so that the settings of the model parameters can be more in line with the actual needs, and the accuracy of the subsequent predicted effluent water quality parameters can be improved.

[0114] In an embodiment, when optimizing the model parameters, the values of the water quality mechanism model determined according to the predicted water quality parameters can also be considered. The values of the water quality mechanism model are adjusted to tend to 0. Through this embodiment, the settings of the determined model parameters can make the settings of the model parameters more in line with the constraints of the water quality mechanism model, so as to further improve the rationality of the settings of the model parameters, and further improve the accuracy of the subsequent predicted effluent water quality parameters.

[0115] For example, the embodiment can also determine the actual value of the water quality mechanism model according to the water quality parameter of the water inlet, the working condition parameter of the sewage treatment system, the model parameter and the predicted water quality parameter when optimizing the model parameter. The operation can be similar to the operation S420 described above, and will not be described here again. After obtaining the actual value, the embodiment can adjust the value of the model parameter according to the difference between the actual water quality parameter and the predicted water quality parameter, and the weighted sum of the actual value. For example, the value of the model parameter can be adjusted to minimize the weighted sum, so as to obtain the optimized model parameter. Wherein, the actual value can be taken as the loss value of the reaction mechanism, and the difference between the actual water quality parameter and the predicted water quality parameter can be taken as the prediction loss value. The weighted sum of the two parts of the loss value is taken as the total loss value, and the value of the model parameter is adjusted to minimize the total loss value. Wherein, the loss value of the reaction mechanism can be determined by the formula (5) described above. Wherein, the weight used when calculating the weighted sum can be set according to actual needs, for example, the weight can be 1, and the present disclosure does not limit this.

[0116] For example, the embodiment can also adjust the value of the model parameter to minimize the expected value of the weighted sum For example, the following formula (6) can be used to determine the optimized model parameter

[0117]

[0118] In an embodiment, the termination condition of adjusting the model parameter can be, for example, that the relative error of iteration is less than 0.01. For example, the termination condition can use the formula (7) as shown below.

[0119]

[0120] Wherein, K φ (n+1) is the value of the model parameter obtained in the (n+1) th iteration, K φ (n) is the value of the model parameter obtained in the n th iteration.

[0121] In an embodiment, the difference between the actual water quality parameter and the predicted water quality parameter can be determined in real time during operation of the wastewater treatment system. An average value of the difference in a predetermined period is calculated. The value of the model parameter is adjusted according to the difference only when the average value of the difference is greater than a predetermined threshold. The predetermined period can be one day, one week, or any period set according to actual needs. The predetermined threshold can be set according to actual needs, for example, 10% if the difference is a relative difference. The present disclosure does not limit this. In this way, the model parameter can be optimized according to actual needs without frequent optimization, thereby reducing the optimization cost while ensuring the prediction accuracy of the water quality parameter.

[0122] In an embodiment, the value of the model parameter can also be optimized according to the difference periodically, which is not limited by the present disclosure.

[0123] In order to facilitate the implementation of the water quality prediction method of the embodiments of the present disclosure, the present disclosure further provides a training method of a water quality prediction model, which will be described below in combination with Figure 6 The training method will be described in detail.

[0124] Figure 6 is a flowchart of the training method of the water quality prediction model according to the embodiments of the present disclosure.

[0125] As Figure 6 shown, the training method 600 of the water quality prediction model of the embodiment can include operation S610 to operation S650. The water quality prediction model can include the backbone network, the first branch network, the second branch network, and the fusion network described above.

[0126] In operation S610, the water quality parameters of the influent in the sample data are processed by using the backbone network to obtain water quality feature data.

[0127] According to the embodiments of the present disclosure, the sample data can be data collected according to the operation of the wastewater treatment system, which includes the water quality parameters of the influent, the working condition parameters of the wastewater treatment system, the model parameters of the water quality mechanism model, and the water quality parameters of the effluent. The water quality parameters of the effluent in the sample data are the supervision signals for model training. The implementation principle of operation S610 is similar to that of operation S210 described above, which will not be described here.

[0128] In operation S620, the working condition parameters of the wastewater treatment system in the sample data are processed by using the first branch network to obtain working condition feature data.

[0129] In operation S630, the model parameters of the water quality mechanism model in the sample data are processed by using the second branch network to obtain parameter feature data.

[0130] The implementation principle of the operation S620 is similar to that of the operation S220 described above, and the implementation principle of the operation S630 is similar to that of the operation S230 described above, which will not be described here again.

[0131] In operation S640, the fusion network is adopted to perform weighted processing on the water quality characteristic data according to the working condition characteristic data and the parameter characteristic data, to obtain the predicted water quality parameter of the effluent.

[0132] According to an embodiment of the present disclosure, the fusion network can perform weighted processing on the water quality characteristic data by using a principle similar to that of the operation S240 described above, to obtain the water quality parameter of the effluent as the predicted water quality parameter.

[0133] In operation S650, the water quality prediction model is trained according to the difference between the water quality parameter of the effluent in the sample data and the predicted water quality parameter.

[0134] The operation S650 can use a principle similar to that of the operation S530 described above to adjust the network parameters of the water quality prediction model, so as to realize the training of the water quality prediction model. For example, the loss value of the predicted water quality parameter can be calculated by using the formula (5) described above, and the water quality prediction model is trained with the objective of minimizing the loss value.

[0135] In an embodiment, when training the water quality prediction model, the loss value of the reaction mechanism described above can also be considered Specifically, the actual value of the water quality mechanism model can be determined according to the water quality parameter of the influent, the working condition parameter, the model parameter and the predicted water quality parameter. Subsequently, the water quality prediction model is trained according to the weighted sum of the difference and the actual value of the water quality mechanism model. Specifically, the value of can be taken as the loss value of the water quality prediction model, and the water quality prediction model is trained with the objective of minimizing the loss value. It should be noted that in this embodiment, the loss value of the predicted water quality parameter is calculated by using the formula (6) described above. i is the i th water quality parameter in the water quality parameter of the effluent included in the sample data.

[0136] For example, the water quality prediction model can be trained with the objective of minimizing the expected value of the weighted sum. For example, a formula similar to the formula (6) described above can be used to optimize the value of the network parameters of the water quality prediction model.

[0137] In an embodiment, when the sample data is acquired, the data of the operation of the sewage treatment system can be collected in real time, specifically the water quality parameters of the influent are collected, the working condition parameters of the sewage treatment system are acquired, and the model parameters of the water quality mechanism model are acquired, then the collected water quality parameters of the influent, the acquired working condition parameters of the sewage treatment system, and the model parameters of the water quality mechanism model are substituted into the formula (1) described above, to determine the target water quality parameters of the effluent (i.e. the water quality parameters of the effluent when the formula (1) is established) that make the value of the water quality mechanism model a predetermined value, as the water quality parameters of the influent. Then, the embodiment can combine the collected water quality parameters of the influent, the acquired working condition parameters of the sewage treatment system, the model parameters of the water quality mechanism model, and the target water quality parameters to form a sample data. That is, the target water quality parameters are used as the supervision signal of the water quality prediction model. In this way, when the water quality prediction model is trained according to the sample data, the water quality prediction model can learn the algebraic differential expression of the water quality mechanism model, and an accurate water quality prediction model is trained.

[0138] In an embodiment, the water quality parameters of the influent, the working condition parameters, and the model parameters can be acquired by sampling. Specifically, the sampling method can be used to sample the water quality parameter library, the working condition parameter library, and the model parameter library respectively, so as to obtain the water quality parameters of the influent, the working condition parameters, and the model parameters included in the sample data. The water quality parameter library, the working condition parameter library, and the model parameter library can be parameter libraries maintained in advance according to actual needs, which are not limited in the present disclosure. For example, the sampling method can be a low-discrepancy sequence sampling method to sample, so that the sample data obtained by sampling is uniformly distributed in the sample space, so as to facilitate improving the universality of the water quality prediction model trained.

[0139] The low-discrepancy sequence sampling method can include, for example, a Latin hypercube sampling method and a Sobol sequence sampling method, which are not limited in the present disclosure.

[0140] In an embodiment, as described above, the water quality parameters of the influent can include N first water quality parameters. The operation S610 can be implemented by using the following operation: using the backbone network to process the N first water quality parameters to obtain N feature data corresponding to the N first water quality parameters. Accordingly, the operation S640 can be implemented by using the following operation: processing each of the N feature data according to the working condition feature data and the parameter feature data to obtain N second water quality parameters corresponding to the N first water quality parameters.

[0141] In an embodiment, the first branch network includes M first sub-networks, and the second branch network includes L second sub-networks, as described above. The operation S620 can be implemented by, for example, processing the M working condition parameters one by one using the M first sub-networks to obtain M working condition feature data. The operation S630 can be implemented by, for example, processing the L model parameters one by one using the L second sub-networks to obtain L parameter feature data.

[0142] In an embodiment, the water quality feature data includes N feature data corresponding to N first water quality parameters included in the water quality parameters of the influent. The operation S640 can be implemented by, for example, for each of the N feature data, determining the dot product of the M working condition feature data, the L parameter feature data and the feature data using the fusion network to obtain a second water quality parameter. The dimensions of the M working condition feature data, the L parameter feature data and the N feature data are equal. The predicted water quality parameters of the effluent include N second water quality parameters obtained for the N feature data.

[0143] Based on the water quality prediction method provided in the present disclosure, the present disclosure further provides a water quality prediction device, which will be described below in combination with Figure 7 The device will be described in detail.

[0144] Figure 7 is a structural block diagram of a water quality prediction device according to an embodiment of the present disclosure.

[0145] As Figure 7 shown, the water quality prediction device 700 of this embodiment can include a water quality parameter processing module 710, a working condition parameter processing module 720, a model parameter processing module 730 and a water quality parameter prediction module 740.

[0146] The water quality parameter processing module 710 is configured to process the water quality parameters of the influent using the backbone network of the water quality prediction model to obtain water quality feature data. The water quality prediction model is constructed based on a deep operator network. In an embodiment, the water quality parameter processing module 710 can be configured to perform the operation S210 described above, which will not be described here again.

[0147] The working condition parameter processing module 720 is configured to process the working condition parameters of the sewage treatment system using the first branch network of the water quality prediction model to obtain working condition feature data. In an embodiment, the working condition parameter processing module 720 can be configured to perform the operation S220 described above, which will not be described here again.

[0148] The model parameter processing module 730 is configured to process the model parameters of the water quality mechanism model by using the second branch network of the water quality prediction model to obtain parameter feature data. In an embodiment, the model parameter processing module 730 can be configured to perform the operation S230 described above, and details are not described herein again.

[0149] The water quality parameter prediction module 740 is configured to perform weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data to obtain the water quality parameters of the effluent. In an embodiment, the water quality parameter prediction module 740 can be configured to perform the operation S240 described above, and details are not described herein again.

[0150] According to an embodiment of the present disclosure, the water quality parameters of the influent include N first water quality parameters. The water quality parameter processing module 710 can be specifically configured to process the N first water quality parameters by using the backbone network to obtain N feature data corresponding to the N first water quality parameters. Correspondingly, the water quality parameter prediction module 740 is specifically configured to perform weighted processing on the N feature data respectively according to the working condition feature data and the parameter feature data to obtain N second water quality parameters corresponding to the N first water quality parameters, wherein the water quality parameters of the effluent include the N second water quality parameters; and N is a natural number.

[0151] According to an embodiment of the present disclosure, the first branch network includes M first sub-networks, and the second branch network includes L second sub-networks. The working condition parameter processing module 720 can be specifically configured to process the M working condition parameters by using the M first sub-networks to obtain M working condition feature data. The model parameter processing module 730 can be specifically configured to process the L model parameters by using the L second sub-networks to obtain L parameter feature data. Wherein, M and L are both natural numbers greater than 1, the M first sub-networks correspond to the M working condition parameters, and the L second sub-networks correspond to the L model parameters.

[0152] According to an embodiment of the present disclosure, the water quality feature data includes N feature data corresponding to N first water quality parameters included in the water quality parameters of the influent. The water quality parameter prediction module 740 can be specifically configured to, for a feature data in the N feature data, determine a dot product of the M working condition feature data, the L parameter feature data and the feature data to obtain a second water quality parameter, wherein the dimensions of the M working condition feature data, the L parameter feature data and the N feature data are equal; and the water quality parameters of the effluent include N second water quality parameters obtained for the N feature data.

[0153] Based on the parameter recommendation method provided by the present disclosure, the present disclosure further provides a parameter recommendation device, which will be described below in combination with Figure 8 The device will be described in detail.

[0154] Figure 8is a structural block diagram of a parameter recommendation device according to an embodiment of the present disclosure.

[0155] As shown in Figure 8 , the parameter recommendation device 800 of this embodiment can include a water quality parameter prediction module 810, a model value determination module 820, and a recommended parameter determination module 830.

[0156] The water quality parameter prediction module 810 is configured to predict the water quality parameter of the effluent by using the water quality prediction device 700 described above. In an embodiment, the water quality parameter prediction module 810 can be configured to perform the operation S410 described above, which will not be repeated here.

[0157] The model value determination module 820 is configured to determine the value of the water quality mechanism model according to the water quality parameter of the influent, the working condition parameter of the sewage treatment system, the model parameter, and the water quality parameter of the effluent. In an embodiment, the model value determination module 820 can be configured to perform the operation S420 described above, which will not be repeated here.

[0158] The recommended parameter determination module 830 is configured to adjust the working condition parameter to obtain the recommended working condition parameter, with the objective of minimizing the value of the water quality mechanism model. In an embodiment, the recommended parameter determination module 830 can be configured to perform the operation S430 described above, which will not be repeated here.

[0159] According to an embodiment of the present disclosure, the parameter recommendation device 800 described above can further include a resource consumption determination module configured to determine the resource consumption amount of the sewage treatment system according to the working condition parameter of the sewage treatment system. Specifically, the recommended parameter determination module 830 described above can be configured to adjust the working condition parameter to obtain the recommended working condition parameter, with the objective of minimizing the weighted sum of the value of the water quality mechanism model and the resource consumption amount.

[0160] According to an embodiment of the present disclosure, the recommended parameter determination module 830 described above can be configured to adjust the working condition parameter to obtain the recommended working condition parameter, with the condition that the water quality parameter of the effluent is less than a predetermined value.

[0161] Based on the parameter optimization method provided by the present disclosure, the present disclosure further provides a parameter optimization device, which will be described in detail below in combination with Figure 9 .

[0162] Figure 9 is a structural block diagram of a parameter optimization device according to an embodiment of the present disclosure.

[0163] As shown in Figure 9 , the parameter optimization device 900 of this embodiment can include a water quality parameter prediction module 910, an actual parameter determination module 920, and a recommended parameter optimization module 930.

[0164] The water quality parameter prediction module 910 is configured to predict the water quality parameter of the effluent by using the water quality prediction device 700 described above, as the predicted water quality parameter. In an embodiment, the water quality parameter prediction module 910 can be configured to perform the operation S510 described above, which will not be repeated here.

[0165] The actual parameter determination module 920 is configured to determine the actual water quality parameter of the effluent obtained by the wastewater treatment system processing the influent with the water quality parameter of the influent according to the working condition parameter. In an embodiment, the actual parameter determination module 920 can be configured to perform the operation S520 described above, which will not be repeated here.

[0166] The parameter optimization module 930 is configured to adjust the value of the model parameter according to the difference between the actual water quality parameter and the predicted water quality parameter, to obtain the optimized model parameter. In an embodiment, the parameter optimization module 930 can be configured to perform the operation S530 described above, which will not be repeated here.

[0167] According to an embodiment of the present disclosure, the parameter optimization device 900 described above may, for example, further include a model value determination module configured to determine the actual value of the water quality mechanism model according to the water quality parameter of the influent, the working condition parameter of the wastewater treatment system, the model parameter, and the predicted water quality parameter. The parameter optimization module 930 described above may, for example, be specifically configured to adjust the value of the model parameter according to the weighted sum of the difference and the actual value, to obtain the optimized model parameter.

[0168] According to an embodiment of the present disclosure, the parameter optimization module 930 described above may, for example, be specifically configured to adjust the value of the model parameter to minimize the expected value of the weighted sum, to obtain the optimized model parameter.

[0169] According to an embodiment of the present disclosure, the parameter optimization device 900 described above may, for example, further include a difference statistics module configured to statistically determine the average value of the difference between the actual water quality parameter and the predicted water quality parameter within a predetermined period. The parameter optimization module 930 described above may, for example, be specifically configured to adjust the value of the model parameter according to the difference in response to the average value being greater than a predetermined threshold, to obtain the optimized model parameter.

[0170] Based on the training method of the water quality prediction model provided by the present disclosure, the present disclosure further provides a training device of a water quality prediction model, which will be described below in combination with Figure 10 The device will be described in detail.

[0171] Figure 10 is a structural block diagram of the training device of the water quality prediction model according to an embodiment of the present disclosure.

[0172] As Figure 10As shown, the training apparatus 1000 of the water quality prediction model of the embodiment can include a water quality parameter processing module 1010, an operating condition parameter processing module 1020, a model parameter processing module 1030, and a water quality parameter prediction module 1040. The water quality prediction model includes a backbone network, a first branch network, a second branch network, and a fusion network arranged side by side. The water quality prediction model is constructed based on a deep operator network.

[0173] The water quality parameter processing module 1010 is configured to process the water quality parameters of the influent in the sample data by using the backbone network to obtain water quality feature data. In an embodiment, the water quality parameter processing module 1010 can be configured to perform the operation S610 described above, and details are not repeated here.

[0174] The operating condition parameter processing module 1020 is configured to process the operating condition parameters of the wastewater treatment system in the sample data by using the first branch network to obtain operating condition feature data. In an embodiment, the operating condition parameter processing module 1020 can be configured to perform the operation S620 described above, and details are not repeated here.

[0175] The model parameter processing module 1030 is configured to process the model parameters of the water quality mechanism model in the sample data by using the second branch network to obtain parameter feature data. In an embodiment, the model parameter processing module 1030 can be configured to perform the operation S630 described above, and details are not repeated here.

[0176] The water quality parameter prediction module 1040 is configured to perform weighted processing on the water quality feature data according to the operating condition feature data and the parameter feature data by using the fusion network to obtain the predicted water quality parameters of the effluent. In an embodiment, the water quality parameter prediction module 1040 can be configured to perform the operation S640 described above, and details are not repeated here.

[0177] The model training module 1050 is configured to train the water quality prediction model according to the difference between the water quality parameters of the effluent and the predicted water quality parameters in the sample data. In an embodiment, the model training module 1050 can be configured to perform the operation S650 described above, and details are not repeated here.

[0178] According to an embodiment of the present disclosure, the training apparatus 1000 of the water quality prediction model described above may, for example, further include a model value determination module configured to determine the actual value of the water quality mechanism model according to the water quality parameters of the influent, the operating condition parameters, the model parameters, and the predicted water quality parameters. The model training module 1050 described above may, for example, be configured to train the water quality prediction model according to the weighted sum of the difference and the actual value of the water quality mechanism model.

[0179] According to an embodiment of the present disclosure, the model training module 1050 described above may, for example, be configured to train the water quality prediction model with the expectation value of the weighted sum being minimized as a target.

[0180] According to an embodiment of the present disclosure, the training device 1000 of the water quality prediction model can further include a target parameter determination module configured to determine a target water quality parameter that makes the value of the water quality mechanism model a predetermined value as the water quality parameter of the effluent included in the sample data according to the water quality parameter of the influent, the working condition parameter and the model parameter.

[0181] According to an embodiment of the present disclosure, the training device 1000 of the water quality prediction model can further include a parameter sampling module configured to sample the water quality parameter library of the influent, the working condition parameter library and the model parameter library respectively by using a low difference sequence sampling method to obtain the water quality parameter of the influent, the working condition parameter and the model parameter included in the sample data.

[0182] According to an embodiment of the present disclosure, the water quality parameter of the influent includes N first water quality parameters. The water quality parameter processing module 1010 can be specifically configured to process the N first water quality parameters by using a backbone network to obtain N feature data corresponding to the N first water quality parameters. The water quality parameter prediction module 1040 can be specifically configured to process the N feature data respectively according to the working condition feature data and the parameter feature data to obtain N second water quality parameters corresponding to the N first water quality parameters. The water quality parameter of the effluent includes the N second water quality parameters, and N is a natural number.

[0183] According to an embodiment of the present disclosure, the first branch network includes M first sub-networks, and the second branch network includes L second sub-networks. The working condition parameter processing module 1020 can be specifically configured to process the M working condition parameters by using the M first sub-networks to obtain M working condition feature data. The model parameter processing module 1030 can be specifically configured to process the L model parameters by using the L second sub-networks to obtain L parameter feature data, wherein M and L are both natural numbers greater than 1, the M first sub-networks correspond to the M working condition parameters, and the L second sub-networks correspond to the L model parameters.

[0184] According to an embodiment of the present disclosure, the water quality feature data includes N feature data corresponding to the N first water quality parameters included in the water quality parameter of the influent. The water quality parameter prediction module 1040 can be specifically configured to determine the dot product of the M working condition feature data, the L parameter feature data and the feature data of the N feature data by using a fusion network to obtain one second water quality parameter for the feature data in the N feature data, wherein the dimensions of the M working condition feature data, the L parameter feature data and the N feature data are equal; and the predicted water quality parameter of the effluent includes N second water quality parameters obtained for the N feature data.

[0185] It should be noted that in the technical solutions of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure and application of user personal information meet the relevant legal regulations, necessary security measures are taken, and the public order and good customs are not violated. In the technical solutions of the present disclosure, the authorization or consent of the user is obtained before the user personal information is acquired or collected.

[0186] According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0187] Figure 11 A schematic block diagram of an example electronic device 1100 that can be used to implement methods of embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit implementations of the present disclosure described and / or claimed in this document.

[0188] As shown in Figure 11 The device 1100 includes a computing unit 1101 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded into a random access memory (RAM) 1103 from a storage unit 1108. Various programs and data required for the operation of the device 1100 can also be stored in the RAM 1103. The computing unit 1101, the ROM 1102, and the RAM 1103 are connected to each other through a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0189] Various components in the device 1100 are connected to the I / O interface 1105, including an input unit 1106, such as a keyboard, a mouse, etc., an output unit 1107, such as various types of displays, speakers, etc., a storage unit 1108, such as a magnetic disk, an optical disk, etc., and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1109 allows the device 1100 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunications networks.

[0190] The computing unit 1101 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, and the like. The computing unit 1101 performs various methods and processes described above, such as at least one of the following methods: a water quality prediction method, a parameter recommendation method, a parameter optimization method, and a training method of a water quality prediction model. For example, in some embodiments, at least one of the following methods: a water quality prediction method, a parameter recommendation method, a parameter optimization method, and a training method of a water quality prediction model can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded onto the RAM 1103 and executed by the computing unit 1101, one or more steps of at least one of the following methods: a water quality prediction method, a parameter recommendation method, a parameter optimization method, and a training method of a water quality prediction model described above can be performed. Alternatively, in other embodiments, the computing unit 1101 can be configured to perform at least one of the following methods: a water quality prediction method, a parameter recommendation method, a parameter optimization method, and a training method of a water quality prediction model provided by the present disclosure by any other appropriate means, such as by means of firmware.

[0191] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0192] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a standalone software package, or entirely on a remote machine or server.

[0193] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0194] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0195] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0196] The computer system can include clients and servers. This relationship can be. The servers are typically remote from the clients with the interactions between them occurring over a communication network. The relationship between clients and servers arises by interplay between programs running on the respective computers and having a client-server relationship. Among other things, the server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS (Virtual Private Server, or VPS for short) services. The server can also be a server of a distributed system, or a server combined with a blockchain.

[0197] It should be understood that the various forms of flow shown above can be reordered, steps added or removed. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technical solutions of the present disclosure are achieved, which are not limited herein.

[0198] The above detailed description does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A water quality prediction method, comprising: processing, by a main network of a water quality prediction model, water quality parameters of influent water to obtain water quality feature data; processing, by a first branch network of the water quality prediction model, working condition parameters of a sewage treatment system to obtain working condition feature data; processing, by a second branch network of the water quality prediction model, model parameters of a water quality mechanism model to obtain parameter feature data; performing weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data to obtain water quality parameters of effluent water, wherein the water quality prediction model is constructed based on a deep operator network; determining values of the water quality mechanism model according to the water quality parameters of the influent water, the working condition parameters of the sewage treatment system, the model parameters, and the water quality parameters of the effluent water; and adjusting the working condition parameters to obtain recommended working condition parameters, with the aim of minimizing the values of the water quality mechanism model. The water quality parameters of the influent water include N first water quality parameters.

2. The method of claim 1, wherein, The processing, by the main network of the water quality prediction model, of the water quality parameters of the influent water to obtain the water quality feature data includes processing, by the main network, of the N first water quality parameters to obtain N feature data corresponding to the N first water quality parameters. The performing of the weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data to obtain the water quality parameters of the effluent water includes performing the weighted processing on the N feature data respectively according to the working condition feature data and the parameter feature data to obtain N second water quality parameters corresponding to the N first water quality parameters. The water quality parameters of the effluent water include the N second water quality parameters; and N is a natural number. The first branch network includes M first sub-networks; and the second branch network includes L second sub-networks.

3. The method of claim 1 or 2, wherein, The processing, by the first branch network of the water quality prediction model, of the working condition parameters of the sewage treatment system to obtain the working condition feature data includes processing, by the M first sub-networks, of M working condition parameters to obtain M working condition feature data. The processing, by the second branch network of the water quality prediction model, of the model parameters of the water quality mechanism model to obtain the parameter feature data includes processing, by the L second sub-networks, of L model parameters to obtain L parameter feature data. M and L are both natural numbers greater than 1; the M first sub-networks correspond to the M working condition parameters; and the L second sub-networks correspond to the L model parameters. The water quality feature data includes N feature data corresponding to the N first water quality parameters included in the water quality parameters of the influent water.

4. The method of claim 3, wherein, For a feature data in the N feature data, a dot product of the M working condition feature data, the L parameter feature data, and the feature data is determined to obtain a second water quality parameter. ​ The M working condition characteristic data, the L parameter characteristic data, and the N characteristic data have equal dimensions; and the water quality parameters of the effluent include N second water quality parameters obtained for the N characteristic data.

5. The method of claim 1, further comprising: determining a resource consumption amount of the wastewater treatment system according to the working condition parameters of the wastewater treatment system; wherein the adjusting the working condition parameters to minimize the value of the water quality mechanism model to obtain recommended working condition parameters comprises: adjusting the working condition parameters to minimize a weighted sum of the value of the water quality mechanism model and the resource consumption amount to obtain the recommended working condition parameters.

6. The method of claim 1 or 5, wherein, The adjusting the working condition parameters to minimize the value of the water quality mechanism model to obtain recommended working condition parameters comprises: adjusting the working condition parameters to obtain recommended working condition parameters under the condition that the water quality parameters of the effluent are less than predetermined values.

7. A parameter optimization method, comprising: predicting water quality parameters of effluent by the method of any one of claims 1-6 as predicted water quality parameters; determining actual water quality parameters of effluent obtained by the wastewater treatment system processing influent having the water quality parameters of the influent according to the working condition parameters; and adjusting the values of the model parameters according to differences between the actual water quality parameters and the predicted water quality parameters to obtain optimized model parameters.

8. The method of claim 7, further comprising: determining actual values of the water quality mechanism model according to the water quality parameters of the influent, the working condition parameters of the wastewater treatment system, the model parameters, and the predicted water quality parameters; and The adjusting the values of the model parameters according to the differences between the actual water quality parameters and the predicted water quality parameters to obtain optimized model parameters comprises: adjusting the values of the model parameters according to a weighted sum of the differences and the actual values to obtain the optimized model parameters.

9. The method of claim 8, wherein, The adjusting the values of the model parameters according to the differences and the weighted sum of the actual values to obtain the optimized model parameters comprises: adjusting the values of the model parameters to minimize an expected value of the weighted sum to obtain the optimized model parameters.

10. The method of claim 7, further comprising: statistically determining an average value of the differences between the actual water quality parameters and the predicted water quality parameters within a predetermined period of time; and The adjusting the values of the model parameters according to the differences between the actual water quality parameters and the predicted water quality parameters to obtain optimized model parameters comprises: in response to the average value being greater than a predetermined threshold, adjusting the values of the model parameters according to the differences to obtain the optimized model parameters.

11. A method of training a water quality prediction model, wherein, The water quality prediction model comprises a trunk network, a first branch network, a second branch network, and a fusion network arranged in parallel; The water quality prediction model is constructed based on a deep operator network; and the training method comprises: processing water quality parameters of influent in sample data by the trunk network to obtain water quality characteristic data; The first branch network is used to process the working condition parameters of the sewage treatment system in the sample data, to obtain working condition characteristic data; The second branch network is used to process the model parameters of the water quality mechanism model in the sample data, to obtain parameter characteristic data; The fusion network is used to perform weighted processing on the water quality characteristic data according to the working condition characteristic data and the parameter characteristic data, to obtain predicted water quality parameters of effluent; and The water quality prediction model is trained according to differences between the water quality parameters of effluent in the sample data and the predicted water quality parameters. It also includes: According to the water quality parameters of the influent, the working condition parameters, the model parameters and the predicted water quality parameters, the actual values of the water quality mechanism model are determined, Wherein, the water quality prediction model is trained according to the differences between the water quality parameters of effluent in the sample data and the predicted water quality parameters, including: the water quality prediction model is trained according to the weighted sum of the differences and the actual values of the water quality mechanism model.

12. The method of claim 11, wherein, The water quality prediction model is trained according to the differences and the weighted sum of the values of the water quality mechanism model, including: The water quality prediction model is trained to minimize the expected value of the weighted sum.

13. The method of claim 11, further comprising: According to the water quality parameters of the influent, the working condition parameters and the model parameters, the target water quality parameters are determined, which make the values of the water quality mechanism model be predetermined values, as the water quality parameters of effluent included in the sample data.

14. The method of claim 13, further comprising: The water quality parameter library, the working condition parameter library and the model parameter library of influent are respectively sampled by using a low difference sequence sampling method, to obtain the water quality parameters of influent, the working condition parameters and the model parameters included in the sample data.

15. The method of claim 11, wherein, The water quality parameters of the influent include N first water quality parameters; The water quality characteristic data is obtained by processing the water quality parameters of the influent in the sample data by using the backbone network, including: the N first water quality parameters are processed by using the backbone network, to obtain N characteristic data corresponding to the N first water quality parameters; The predicted water quality parameters of effluent are obtained by performing weighted processing on the water quality characteristic data according to the working condition characteristic data and the parameter characteristic data by using the fusion network, including: the N characteristic data are respectively processed according to the working condition characteristic data and the parameter characteristic data, to obtain N second water quality parameters corresponding to the N first water quality parameters, Wherein, the water quality parameters of the effluent include the N second water quality parameters; N is a natural number.

16. The method of claim 11 or 15, wherein, The first branch network includes M first sub-networks; the second branch network includes L second sub-networks; The working condition characteristic data is obtained by processing the working condition parameters of the sewage treatment system in the sample data by using the first branch network, including: the M working condition parameters are processed by using the M first sub-networks, to obtain M working condition characteristic data; The processing of the model parameters of the water quality mechanism model in the sample data by the second branch network includes processing L model parameters by the L second sub-networks to obtain L parameter feature data, wherein M and L are natural numbers greater than 1; the M first sub-networks correspond to the M working condition parameters, and the L second sub-networks correspond to the L model parameters.

17. The method of claim 16, wherein, The water quality feature data includes N feature data; the N feature data correspond to N first water quality parameters included in the water quality parameters of the influent; the weighted processing of the water quality feature data by the fusion network according to the working condition feature data and the parameter feature data to obtain the predicted water quality parameters of the effluent includes: For a feature data in the N feature data, the fusion network is used to determine the dot product of the M working condition feature data, the L parameter feature data and the feature data to obtain a second water quality parameter, wherein the dimensions of the M working condition feature data, the L parameter feature data and the N feature data are equal; the predicted water quality parameters of the effluent include N second water quality parameters obtained for N feature data.

18. A water quality prediction device, comprising: a water quality parameter processing module configured to process water quality parameters of an influent by a backbone network of a water quality prediction model to obtain water quality feature data; a working condition parameter processing module configured to process working condition parameters of a sewage treatment system by a first branch network of the water quality prediction model to obtain working condition feature data; a model parameter processing module configured to process model parameters of a water quality mechanism model by a second branch network of the water quality prediction model to obtain parameter feature data; a water quality parameter prediction module configured to perform weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data to obtain water quality parameters of an effluent, wherein the water quality prediction model is constructed based on a deep operator network; a model value determination module configured to determine a value of the water quality mechanism model according to the water quality parameters of the influent, the working condition parameters of the sewage treatment system, the model parameters and the water quality parameters of the effluent; and a recommended parameter determination module configured to adjust the working condition parameters to obtain recommended working condition parameters with the aim of minimizing the value of the water quality mechanism model.

19. A parameter optimization device, comprising: a water quality parameter prediction module configured to predict water quality parameters of an effluent by the water quality prediction device of claim 18 as predicted water quality parameters; an actual parameter determination module configured to determine actual water quality parameters of an effluent obtained by processing an influent having the water quality parameters of the influent by the sewage treatment system according to the working condition parameters; and a parameter optimization module configured to adjust the value of the model parameters according to the difference between the actual water quality parameters and the predicted water quality parameters to obtain optimized model parameters. The water quality prediction model includes a backbone network, a first branch network, a second branch network and a fusion network arranged side by side; 20. An apparatus for training a water quality prediction model, wherein, ​ The water quality prediction model is constructed based on a deep operator network; and the device comprises: a water quality parameter processing module configured to process water quality parameters of the influent in the sample data using the backbone network to obtain water quality feature data; a working condition parameter processing module configured to process working condition parameters of the wastewater treatment system in the sample data using the first branch network to obtain working condition feature data; a model parameter processing module configured to process model parameters of the water quality mechanism model in the sample data using the second branch network to obtain parameter feature data; a water quality parameter prediction module configured to perform weighted processing on the water quality feature data according to the working condition feature data and the parameter feature data using the fusion network to obtain predicted water quality parameters of the effluent; and a model training module configured to train the water quality prediction model according to differences between the water quality parameters of the effluent in the sample data and the predicted water quality parameters; The device further comprises: a model value determination module configured to determine actual values of the water quality mechanism model according to the water quality parameters of the influent, the working condition parameters, the model parameters and the predicted water quality parameters, The model training module is further configured to train the water quality prediction model according to the differences and a weighted sum of the actual values of the water quality mechanism model. 21.An electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-17.

22. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-17. 23.A computer program product comprising computer programs / instructions stored on at least one of a readable storage medium and an electronic device, the computer programs / instructions, when executed by a processor, implement the steps of the method of any one of claims 1-17.

Images