Sewage nitrogen concentration prediction method based on artificial intelligence

Abstract

The invention discloses a sewage nitrogen concentration prediction method based on artificial intelligence, and the method specifically comprises the steps: employing the historical data of ammonia nitrogen or total nitrogen, and combining a deep learning algorithm to obtain an ammonia nitrogen or total nitrogen historical data prediction value set; determining an auxiliary variable, constructing a relation model of ammonia nitrogen and the auxiliary variable as an ammonia nitrogen fuzzy prediction model, constructing a relation model of total nitrogen and the auxiliary variable as a total nitrogen fuzzy prediction model, and obtaining an ammonia nitrogen or total nitrogen fuzzy prediction concentration value set; constructing an ammonia nitrogen accurate prediction model or a total nitrogen accurate prediction model; and obtaining an accurate prediction value of ammonia nitrogen or total nitrogen. According to the method, a more comprehensive and accurate prediction model is constructed according to the historical data rule of ammonia nitrogen or total nitrogen, the strong association relationship with ammonia nitrogen or total nitrogen and the association relationship of other water quality parameters. Compared with a single variable or simple combination mode in the prior art, the method can capture the trend of water quality change more accurately, and the accuracy of a prediction result is improved.

Description

Technical Field

[0001] The present invention relates to the field of ammonia nitrogen and total nitrogen concentration prediction, and in particular to a sewage nitrogen concentration prediction method based on artificial intelligence. Background Art

[0002] The accuracy of the sewage inlet water quality data determines the rationality of the operation of the sewage plant, and thus provides a guarantee for the stable compliance of the sewage plant's effluent quality and energy saving and consumption reduction. However, in the actual operation of the sewage plant, due to the accuracy and life of the instrument, there is instability in the water quality measurement data, which brings many unstable factors to the operation control of the sewage plant and the stable compliance of the effluent water quality. Therefore, in addition to improving the accuracy of the inlet water quality instrument, on the other hand, it is necessary to find a short-term alternative solution when individual instruments malfunction, so as to avoid greater losses.

[0003] Soft measurement technology uses computer technology to target variables that are difficult to measure or cannot be measured temporarily, and combines other easily measurable variables to establish models and algorithms to predict variables in the sewage treatment process from limited, discrete data. This can predict potential problems and anomalies in advance, avoid a decline in treatment effects, reduce maintenance costs, and take corresponding measures for preventive maintenance and management. At present, soft measurement technology for sewage has been widely used, and the relevant technologies are as follows:

[0004] CN101625353B discloses a soft measurement method for sewage treatment effluent quality and an online intelligent detection instrument, which selects TOC, DO, ORP and pH values ​​in the influent as inputs, and the soft-measured quantity BOD or COD as outputs, adopts PNN process neural network structure modeling, and based on the improved algorithm of PNN process neural network, the two are stored in an embedded metering instrument processing system in a programmable manner; firstly, offline network training is performed, and multiple sets of input values ​​that have been obtained are input, and the network training accuracy is adjusted by the adaptive network of the instrument processing system until the network training accuracy reaches the set value; online measurement is performed in the sewage treatment site, real-time sampling is performed, and the obtained data is filtered and processed and input to the instrument, and the calculation system quickly calculates the BOD or COD value. Although the patent improves the initial model accuracy through offline network training, it has high requirements for the amount of data, and it is difficult to provide too much data for model training for sewage treatment plants with short operating time; in addition, in practical applications, the sewage treatment process is dynamically changing, and static models are difficult to adapt to such changes, which may lead to increased prediction errors.

[0005] CN112784476B discloses a soft measurement method and device for ammonia nitrogen in the effluent of agricultural sewage treatment facilities of different process types. The prediction device first installs electrodes such as effluent pH, inlet conductivity, effluent conductivity, anaerobic tank ORP, aerobic tank DO, etc. at rural domestic sewage treatment terminals of different process types to collect terminal operating status parameters, and then uses the artificial neural network embedded in the online data platform to predict the terminal effluent ammonia nitrogen concentration. The problem with this patent is that it is only based on relevant control parameters, not water quality parameters for prediction, and the prediction logic is poor, which affects the accuracy. Secondly, although the patent uses artificial neural networks for prediction, this method does not make full use of historical data for modeling, but relies more on real-time data, which limits the learning ability and generalization performance of the model.

[0006] CN102854296A discloses a soft measurement method for sewage treatment based on an integrated neural network, and proposes an integrated neural network model to measure effluent COD, effluent BOD, and effluent TN. The model makes full use of the coupling relationship between the three key effluent water quality parameters, establishes an integrated neural network model containing three sub-feedforward neural networks, and uses the particle swarm algorithm to train each sub-neural network to obtain the optimal structure of each sub-neural network. Finally, the trained neural network is used to predict the effluent COD, effluent BOD, and effluent TN, and the prediction results are accurate. Although the integrated neural network model proposed in this patent can predict multiple key water quality parameters, it does not set up a mutual inspection mechanism, which may lead to the accumulation of errors in a single sub-network and affect the overall prediction results; secondly, the parameter screening is insufficient. In the screening of the response parameters of the prediction data, the mutual relationship and internal logic between the sewage water quality are not fully considered. Considering that too many parameters cause large and frequent measurements, the prediction workload is increased.

[0007] In summary, there are some problems with the current soft measurement methods for sewage water quality. First, the prediction model is single. A single model algorithm is often used for prediction, ignoring the linkage between water quality data in different functional areas, resulting in poor accuracy of the prediction data; secondly, there is a lack of underlying logic based on water quality parameters. In the selection of response parameters, the concept of "the more data, the more accurate the results" is followed, resulting in excessive workload for measurement and analysis. Finally, the method based on data training increases the requirements for data volume, making it difficult to promote and apply it in new sewage treatment plants in the short term. Therefore, it is necessary to improve the sewage soft measurement technology in response to the above problems. Summary of the invention

[0008] In view of the problems existing in the prior art, the present invention provides a method for predicting nitrogen concentration in sewage based on artificial intelligence.

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

[0010] A method for predicting nitrogen concentration in sewage based on artificial intelligence. When the ammonia nitrogen meter has no signal in the current collection period and the total nitrogen meter is normal, the method for predicting ammonia nitrogen concentration includes the following steps:

[0011] Step 1.1: Use the historical data of ammonia nitrogen and combine it with the deep learning algorithm to obtain the historical data prediction model of ammonia nitrogen. Use the actual ammonia nitrogen concentration value of the month before the ammonia nitrogen meter has no signal to input it into the historical data prediction model of ammonia nitrogen to obtain the collection of historical data prediction values ​​of ammonia nitrogen;

[0012] Step 1.2: Determine the auxiliary variables as total nitrogen concentration, chemical oxygen demand, suspended solids concentration and water temperature, obtain historical data of the auxiliary variables and the corresponding real ammonia nitrogen concentration values, and construct a relationship model between ammonia nitrogen and auxiliary variables called an ammonia nitrogen fuzzy prediction model. Use the actual measured values ​​of the auxiliary variables one month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen fuzzy prediction model to obtain a collection of ammonia nitrogen fuzzy prediction concentration values;

[0013] Step 1.3: Use the actual ammonia nitrogen concentration value one month before the ammonia nitrogen meter has no signal, the ammonia nitrogen historical data predicted value collection, and the ammonia nitrogen fuzzy predicted concentration value collection to build an ammonia nitrogen accurate prediction model;

[0014] Step 1.4: Use the ammonia nitrogen historical data prediction model to obtain the ammonia nitrogen historical data prediction value AN1 of the current collection period, obtain the value of the auxiliary variable corresponding to the current collection period and input it into the ammonia nitrogen fuzzy prediction model to obtain the ammonia nitrogen fuzzy prediction concentration value AN2, input the ammonia nitrogen historical data prediction value AN1 and the ammonia nitrogen fuzzy prediction concentration value AN2 into the ammonia nitrogen precise prediction model to obtain the ammonia nitrogen precise prediction value AN3;

[0015] When there is no signal from the total nitrogen meter in the current collection period and the ammonia nitrogen meter is normal, the total nitrogen concentration prediction method includes the following steps:

[0016] Step 2.1: Use the historical data of total nitrogen and combine it with the deep learning algorithm to obtain the historical data prediction model of total nitrogen. Use the actual total nitrogen concentration value of the month before the total nitrogen meter has no signal to input it into the historical data prediction model of total nitrogen to obtain the collection of historical data prediction values ​​of total nitrogen;

[0017] Step 2.2: Determine the auxiliary variables as ammonia nitrogen concentration value, chemical oxygen demand, suspended solids concentration value and water temperature, obtain the historical data of the auxiliary variables and the corresponding true total nitrogen concentration value, and construct a relationship model between total nitrogen and auxiliary variables called the total nitrogen fuzzy prediction model. Use the actual measurement values ​​of the auxiliary variables one month before the total nitrogen instrument has no signal to input into the total nitrogen fuzzy prediction model to obtain the total nitrogen fuzzy prediction concentration value collection;

[0018] Step 2.3: Use the actual total nitrogen concentration value one month before the total nitrogen meter has no signal, the total nitrogen historical data predicted value collection, and the total nitrogen fuzzy predicted concentration value collection to build a total nitrogen accurate prediction model;

[0019] Step 2.4: Use the total nitrogen historical data prediction model to obtain the total nitrogen historical data prediction value TN1 of the current collection period, obtain the value of the auxiliary variable corresponding to the current collection period and input it into the total nitrogen fuzzy prediction model to obtain the total nitrogen fuzzy prediction concentration value TN2, input the total nitrogen historical data prediction value TN1 and the total nitrogen fuzzy prediction concentration value TN2 into the total nitrogen precise prediction model to obtain the total nitrogen precise prediction value TN3.

[0020] Preferably, the specific process of step 1.1 is:

[0021] Step 1.1.1: Continuously collect historical ammonia nitrogen concentration data for m days and n items per day to form a training set; use the long short-term memory network algorithm to build an ammonia nitrogen historical data prediction model;

[0022] Step 1.1.2: Take k consecutive ammonia nitrogen concentration data in the training set as input values, and the k+1th ammonia nitrogen concentration data as output values, and send the training set to the ammonia nitrogen historical data prediction model for training to obtain a trained ammonia nitrogen historical data prediction model;

[0023] Step 1.1.3: Use the actual ammonia nitrogen concentration value of the month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen historical data prediction model to obtain N ammonia nitrogen historical data prediction values ​​to form a collection of ammonia nitrogen historical data prediction values.

[0024] Preferably, the specific process of step 1.2 is:

[0025] Step 1.2.1: Continuously collect historical data of n auxiliary variables and the corresponding actual ammonia nitrogen concentration values ​​for m days each day;

[0026] Step 1.2.2: The data obtained in step 1.2.1 is used to construct a relationship model between ammonia nitrogen concentration value and auxiliary variables through a polynomial regression algorithm, which is called an ammonia nitrogen fuzzy prediction model;

[0027] Step 1.2.3: Use the actual measured values ​​of the auxiliary variables one month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen fuzzy prediction model to obtain N ammonia nitrogen fuzzy prediction concentration values, and form a collection of ammonia nitrogen fuzzy prediction concentration values.

[0028] Preferably, step 1.3 is specifically:

[0029] An ammonia nitrogen accurate prediction model is constructed by using the polynomial regression algorithm based on the actual ammonia nitrogen concentration value one month before the ammonia nitrogen meter had no signal, the corresponding ammonia nitrogen historical data prediction value collection, and the ammonia nitrogen fuzzy prediction concentration value collection.

[0030] Preferably, the specific process of step 2.1 is:

[0031] Step 2.1.1: Continuously collect historical total nitrogen concentration data for n times every day for m days to form a training set; use the long short-term memory network algorithm to build a total nitrogen historical data prediction model;

[0032] Step 2.1.2: Take k consecutive total nitrogen concentration data in the training set as input values, and the k+1th total nitrogen concentration data as output values, and send the training set to the total nitrogen historical data prediction model for training to obtain a trained total nitrogen historical data prediction model;

[0033] Step 2.1.3: Use the actual total nitrogen concentration value of the total nitrogen meter one month before no signal to input into the total nitrogen historical data prediction model, obtain N total nitrogen historical data prediction values, and form a collection of total nitrogen historical data prediction values.

[0034] Preferably, the specific process of step 2.2 is:

[0035] Step 2.2.1: Continuously collect historical data of n auxiliary variables and the corresponding true total nitrogen concentration values ​​for m days each day;

[0036] Step 2.2.2: The data obtained in step 2.2.1 are used to construct a relationship model between total nitrogen concentration value and auxiliary variables through a polynomial regression algorithm, which is called a total nitrogen fuzzy prediction model;

[0037] Step 2.2.3: Use the actual measured values ​​of the auxiliary variables one month before the total nitrogen instrument has no signal to input into the total nitrogen fuzzy prediction model to obtain N total nitrogen fuzzy prediction concentration values, and form a collection of total nitrogen fuzzy prediction concentration values.

[0038] Preferably, step 2.3 is specifically as follows:

[0039] An accurate total nitrogen prediction model was constructed using the polynomial regression algorithm using the actual total nitrogen concentration value one month before the total nitrogen meter had no signal, the corresponding total nitrogen historical data prediction value collection, and the total nitrogen fuzzy prediction concentration value collection.

[0040] The present invention has the following beneficial effects:

[0041] 1) Significantly improve prediction accuracy. This method constructs a more comprehensive and accurate prediction model through the historical data rules of ammonia nitrogen or total nitrogen itself and the strong correlation with ammonia nitrogen or total nitrogen and other water quality parameters (such as chemical oxygen demand, suspended solids concentration and water temperature). Compared with the single variable or simple combination method in the prior art, this method can more accurately capture the trend of water quality changes and improve the accuracy of the prediction results. In practical applications, this method can ensure that the deviation between the prediction results and the true value is less than 5% within 4 months, which is significantly better than the common error range in the prior art.

[0042] 2) Enhance the adaptability and robustness of the model. The data required for water quality concentration prediction in this method are updated in real time, so that the prediction model can be updated in real time according to the actual water quality changes. It ensures that the model can maintain high accuracy under different operating conditions and enhances the stability and reliability of the system. This method is suitable for the prediction of 3-2000mg / L wide-area sewage water quality and shows strong robustness.

[0043] 3) Reduce prediction complexity and workload. Through a strict feature selection process, the problem of large and frequent measurements is reduced, and the workload and computational cost of prediction are reduced. Not only does it improve prediction efficiency, it also simplifies system maintenance and management. This method reduces the number of required monitoring parameters by about 40% through feature selection technology, greatly reducing the workload of data collection and processing, while maintaining prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a logical architecture diagram of the present invention.

[0045] Figure 2 The figure is a comparison chart of the predicted value and the actual value of ammonia nitrogen concentration in Example 1.

[0046] Figure 3 This is a comparison chart of the predicted value and the actual value of ammonia nitrogen concentration in Comparative Example 1.

[0047] Figure 4 This is a comparison chart of the predicted value and the actual value of ammonia nitrogen concentration in Comparative Example 2.

[0048] Figure 5 This is a comparison chart of the predicted value and the actual value of ammonia nitrogen concentration in Comparative Example 3. DETAILED DESCRIPTION

[0049] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments:

[0050] Example 1 is a method for predicting ammonia nitrogen concentration in sewage:

[0051] Combination Figure 1 and Figure 2 In this embodiment, a northern municipal sewage treatment plant has an instrument collection cycle of 1h, and the current collection cycle is from 11:00 to 12:00 on January 31, 2024. The simulated ammonia nitrogen instrument has no signal, and the total nitrogen instrument is normal. The ammonia nitrogen concentration prediction method includes the following steps:

[0052] Step 1.1: Use the historical data of ammonia nitrogen and combine it with the deep learning algorithm to obtain the ammonia nitrogen historical data prediction model. Use the actual ammonia nitrogen concentration value of the ammonia nitrogen meter one month before there is no signal to input into the ammonia nitrogen historical data prediction model to obtain the ammonia nitrogen historical data prediction value collection.

[0053] The specific process is as follows: Step 1.1.1: Collect historical ammonia nitrogen concentration data for 24 hours per day for 60 consecutive days to form a training set, and use the long short-term memory network algorithm to build an ammonia nitrogen historical data prediction model.

[0054] Step 1.1.2: Take 10 consecutive ammonia nitrogen concentration data in the training set as input values, and the 11th ammonia nitrogen concentration data as output values. Send the training set to the ammonia nitrogen historical data prediction model for training to obtain a trained ammonia nitrogen historical data prediction model.

[0055] Step 1.1.3: Use the actual ammonia nitrogen concentration value of the month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen historical data prediction model to obtain 720 ammonia nitrogen historical data prediction values ​​to form a collection of ammonia nitrogen historical data prediction values.

[0056] It should be noted that the historical ammonia nitrogen concentration data in step 1.1.1 do not overlap with the actual ammonia nitrogen concentration value one month before the ammonia nitrogen meter had no signal in step 1.1.3.

[0057] Step 1.2: Determine the auxiliary variables as total nitrogen concentration, chemical oxygen demand, suspended solids concentration and water temperature, obtain the historical data of the auxiliary variables and the corresponding true ammonia nitrogen concentration values, and construct a relationship model between ammonia nitrogen and auxiliary variables called an ammonia nitrogen fuzzy prediction model. Use the actual measured values ​​of the auxiliary variables one month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen fuzzy prediction model to obtain a collection of ammonia nitrogen fuzzy prediction concentration values.

[0058] The specific process is as follows: Step 1.2.1: Continuously collect historical data of 24 auxiliary variables per day for 60 days and the corresponding actual ammonia nitrogen concentration values.

[0059] Step 1.2.2: The data obtained in step 1.2.1 is used to construct a relationship model between ammonia nitrogen concentration values ​​and auxiliary variables through a polynomial regression algorithm, which is called an ammonia nitrogen fuzzy prediction model.

[0060] Step 1.2.3: Use the actual measured values ​​of the auxiliary variables one month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen fuzzy prediction model to obtain 720 ammonia nitrogen fuzzy prediction concentration values, forming a collection of ammonia nitrogen fuzzy prediction concentration values.

[0061] It should be noted that the historical data of the auxiliary variables in step 1.2.1 and the corresponding actual ammonia nitrogen concentration values ​​do not overlap with the actual ammonia nitrogen concentration values ​​of the ammonia nitrogen meter in step 1.2.3 one month before there was no signal.

[0062] Step 1.3: Use the actual ammonia nitrogen concentration value one month before the ammonia nitrogen meter has no signal, the corresponding ammonia nitrogen historical data prediction value collection, and the ammonia nitrogen fuzzy prediction concentration value collection to build an ammonia nitrogen accurate prediction model through a polynomial regression algorithm.

[0063] Step 1.4: Collect the 10 actual ammonia nitrogen concentration values ​​from 1:00 to 11:00 on January 31, 2024 and input them into the ammonia nitrogen historical data prediction model to obtain the ammonia nitrogen historical data prediction value AN1 from 11:00 to 12:00.

[0064] The values ​​of the auxiliary variables in the collection period from 11:00 to 12:00 on January 31, 2024 are obtained and input into the ammonia nitrogen fuzzy prediction model to obtain the ammonia nitrogen fuzzy prediction concentration value AN2.

[0065] The ammonia nitrogen historical data predicted value AN1 and the ammonia nitrogen fuzzy predicted concentration value AN2 are input into the ammonia nitrogen precise prediction model to obtain the ammonia nitrogen precise prediction value AN3. AN3 is used as the ammonia nitrogen predicted concentration value for the collection period from 11:00 to 12:00 on January 31, 2024.

[0066] The comparison chart of the predicted ammonia nitrogen concentration value and the actual value formed by this method for 4 consecutive months is as follows Figure 2 The actual value of ammonia nitrogen in this research period was 47.31±10.76 mg / L, and the average predicted concentration of ammonia nitrogen obtained by this method was 46.74±10.94 mg / L, with a deviation of 4.61±3.40%.

[0067] Comparative Example 1

[0068] In this comparative example, a northern municipal sewage treatment plant has a collection cycle of 1h. The current collection cycle is from 0:00 to 1:00 on August 10, 2024. The simulated ammonia nitrogen meter has no signal, and the total nitrogen meter is normal. The following method is used to predict the ammonia nitrogen concentration value within 123 days after the ammonia nitrogen meter has no signal.

[0069] Only the ammonia nitrogen historical data prediction model obtained in step 1.1 is used to obtain the predicted ammonia nitrogen concentration value for the current collection period, and the predicted ammonia nitrogen concentration value for each collection period is obtained in this way within the next 123 days. The comparison chart between the predicted ammonia nitrogen concentration value and the actual value is shown in the figure below: Figure 3 The mean of the actual value of ammonia nitrogen in this research period is 33.78 mg / L, and the mean of the predicted concentration of ammonia nitrogen obtained by the method of this comparative example is 25.90 mg / L, with a deviation of 23.33%.

[0070] Comparative Example 2

[0071] In this comparative example, a northern municipal sewage treatment plant is used, which is the same as that in comparative example 1. The collection period is 1 hour. The current collection period is from 0:00 to 1:00 on August 10, 2024. The simulated ammonia nitrogen meter has no signal, and the total nitrogen meter is normal. The following method is used to predict the ammonia nitrogen concentration value within 123 days after the ammonia nitrogen meter has no signal.

[0072] Only the ammonia nitrogen fuzzy prediction model obtained in step 1.2 is used to obtain the predicted ammonia nitrogen concentration value of the current collection period, and the ammonia nitrogen predicted concentration value of each collection period is obtained in this way in the following 123 days. The comparison chart between the predicted ammonia nitrogen concentration value and the actual value is shown in the figure below: Figure 4 The mean of the actual value of ammonia nitrogen in the research period is 33.78 mg / L, and the mean of the predicted concentration of ammonia nitrogen obtained by the method of this comparative example is 40.17 mg / L, with a deviation of 18.92%.

[0073] Comparative Example 3

[0074] In this comparative example, a northern municipal sewage treatment plant is used, which is the same as that in comparative example 1. The collection period is 1 hour. The current collection period is from 0:00 to 1:00 on August 10, 2024. The simulated ammonia nitrogen meter has no signal, and the total nitrogen meter is normal. The following method is used to predict the ammonia nitrogen concentration value within 123 days after the ammonia nitrogen meter has no signal.

[0075] The method of the present invention is used to obtain the predicted ammonia nitrogen concentration value for the collection period from 0:00 to 1:00 on August 10, 2024, and the predicted ammonia nitrogen value for each collection period is obtained in this way within the next 123 days. The comparison chart of the predicted ammonia nitrogen concentration value and the actual value is as follows: Figure 5 The mean of the actual value of ammonia nitrogen during the study period was 33.78 mg / L, and the mean of the predicted ammonia nitrogen concentration value predicted by this method was 31.95 mg / L, with a deviation of only 5.42%.

[0076] From the above comparison, it can be seen that the predicted ammonia nitrogen concentration value output by the method of the present invention is closest to the actual value.

[0077] Example 2 is a method for predicting total nitrogen concentration in sewage:

[0078] In this embodiment, a northern municipal sewage treatment plant has an instrument collection cycle of 1 hour, and the current collection cycle is from 11:00 to 12:00 on January 31, 2024. The simulated total nitrogen instrument has no signal, and the ammonia nitrogen instrument is normal. The total nitrogen concentration prediction method includes the following steps:

[0079] Step 1.1: Use the historical data of total nitrogen and combine it with the deep learning algorithm to obtain the total nitrogen historical data prediction model. Use the actual total nitrogen concentration value one month before the total nitrogen meter has no signal to input into the total nitrogen historical data prediction model to obtain the total nitrogen historical data prediction value collection.

[0080] The specific process is as follows: Step 1.1.1: Collect historical total nitrogen concentration data for 24 hours per day for 60 consecutive days to form a training set, and use the long short-term memory network algorithm to build a total nitrogen historical data prediction model.

[0081] Step 1.1.2: The 10 consecutive total nitrogen concentration data in the training set are used as input values, and the 11th total nitrogen concentration data is used as output value. The training set is sent to the total nitrogen historical data prediction model for training to obtain a trained total nitrogen historical data prediction model.

[0082] Step 1.1.3: Use the actual total nitrogen concentration value of the total nitrogen meter one month before no signal to input into the total nitrogen historical data prediction model to obtain 720 total nitrogen historical data prediction values ​​to form a total nitrogen historical data prediction value collection.

[0083] It should be noted that the historical total nitrogen concentration data in step 1.1.1 do not overlap with the actual total nitrogen concentration value one month before the total nitrogen meter had no signal in step 1.1.3.

[0084] Step 1.2: Determine the auxiliary variables as ammonia nitrogen concentration, chemical oxygen demand, suspended solids concentration and water temperature, obtain the historical data of the auxiliary variables and the corresponding true total nitrogen concentration value, and construct a relationship model between total nitrogen and auxiliary variables called the total nitrogen fuzzy prediction model. Use the actual measurement values ​​of the auxiliary variables one month before the total nitrogen instrument has no signal to input into the total nitrogen fuzzy prediction model to obtain the total nitrogen fuzzy prediction concentration value collection.

[0085] The specific process is as follows: Step 1.2.1: Continuously collect historical data of 24 auxiliary variables per day for 60 days and the corresponding true total nitrogen concentration values.

[0086] Step 1.2.2: The data obtained in step 1.2.1 are used to construct a relationship model between total nitrogen concentration values ​​and auxiliary variables through a polynomial regression algorithm, which is called a total nitrogen fuzzy prediction model.

[0087] Step 1.2.3: Use the actual measured values ​​of the auxiliary variables one month before the total nitrogen meter has no signal to input into the total nitrogen fuzzy prediction model to obtain 720 total nitrogen fuzzy prediction concentration values, forming a collection of total nitrogen fuzzy prediction concentration values.

[0088] It should be noted that the historical data of the auxiliary variables in step 1.2.1 and the corresponding true total nitrogen concentration values ​​do not overlap with the true total nitrogen concentration values ​​of the total nitrogen meter in step 1.2.3 one month before no signal.

[0089] Step 1.3: Use the actual total nitrogen concentration value one month before the total nitrogen meter has no signal and the corresponding total nitrogen historical data prediction value collection and total nitrogen fuzzy prediction concentration value collection to build a total nitrogen accurate prediction model through a polynomial regression algorithm.

[0090] Step 1.4: Collect the 10 actual total nitrogen concentration values ​​from 1:00 to 11:00 on January 31, 2024 and input them into the total nitrogen historical data prediction model to obtain the total nitrogen historical data prediction value AN1 from 11:00 to 12:00.

[0091] The values ​​of the auxiliary variables in the collection period from 11:00 to 12:00 on January 31, 2024 are obtained and input into the total nitrogen fuzzy prediction model to obtain the total nitrogen fuzzy prediction concentration value AN2.

[0092] The total nitrogen historical data predicted value AN1 and the total nitrogen fuzzy predicted concentration value AN2 are input into the total nitrogen precise prediction model to obtain the total nitrogen precise prediction value AN3. AN3 is used as the total nitrogen predicted concentration value for the collection period from 11:00 to 12:00 on January 31, 2024.

[0093] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for predicting nitrogen concentration in sewage based on artificial intelligence, characterized in that: When there is no signal from the ammonia nitrogen meter in the current collection period and the total nitrogen meter is normal, the ammonia nitrogen concentration prediction method includes the following steps: Step 1.1: Use the historical data of ammonia nitrogen and combine it with the deep learning algorithm to obtain the historical data prediction model of ammonia nitrogen. Use the actual ammonia nitrogen concentration value of the month before the ammonia nitrogen meter has no signal to input it into the historical data prediction model of ammonia nitrogen to obtain the collection of historical data prediction values ​​of ammonia nitrogen; Step 1.2: Determine the auxiliary variables as total nitrogen concentration, chemical oxygen demand, suspended solids concentration and water temperature, obtain historical data of the auxiliary variables and the corresponding real ammonia nitrogen concentration values, and construct a relationship model between ammonia nitrogen and auxiliary variables called an ammonia nitrogen fuzzy prediction model. Use the actual measured values ​​of the auxiliary variables one month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen fuzzy prediction model to obtain a collection of ammonia nitrogen fuzzy prediction concentration values; Step 1.3: Use the actual ammonia nitrogen concentration value one month before the ammonia nitrogen meter has no signal, the ammonia nitrogen historical data predicted value collection, and the ammonia nitrogen fuzzy predicted concentration value collection to build an ammonia nitrogen accurate prediction model; Step 1.4: Use the ammonia nitrogen historical data prediction model to obtain the ammonia nitrogen historical data prediction value AN1 of the current collection period, obtain the value of the auxiliary variable corresponding to the current collection period and input it into the ammonia nitrogen fuzzy prediction model to obtain the ammonia nitrogen fuzzy prediction concentration value AN2, input the ammonia nitrogen historical data prediction value AN1 and the ammonia nitrogen fuzzy prediction concentration value AN2 into the ammonia nitrogen precise prediction model to obtain the ammonia nitrogen precise prediction value AN3; When there is no signal from the total nitrogen meter in the current collection period and the ammonia nitrogen meter is normal, the total nitrogen concentration prediction method includes the following steps: Step 2.1: Use the historical data of total nitrogen and combine it with the deep learning algorithm to obtain the historical data prediction model of total nitrogen. Use the actual total nitrogen concentration value of the month before the total nitrogen meter has no signal to input it into the historical data prediction model of total nitrogen to obtain the collection of historical data prediction values ​​of total nitrogen; Step 2.2: Determine the auxiliary variables as ammonia nitrogen concentration value, chemical oxygen demand, suspended solids concentration value and water temperature, obtain the historical data of the auxiliary variables and the corresponding true total nitrogen concentration value, and construct a relationship model between total nitrogen and auxiliary variables called the total nitrogen fuzzy prediction model. Use the actual measurement values ​​of the auxiliary variables one month before the total nitrogen instrument has no signal to input into the total nitrogen fuzzy prediction model to obtain the total nitrogen fuzzy prediction concentration value collection; Step 2.3: Use the actual total nitrogen concentration value one month before the total nitrogen meter has no signal, the total nitrogen historical data predicted value collection, and the total nitrogen fuzzy predicted concentration value collection to build a total nitrogen accurate prediction model; Step 2.4: Use the total nitrogen historical data prediction model to obtain the total nitrogen historical data prediction value TN1 of the current collection period, obtain the value of the auxiliary variable corresponding to the current collection period and input it into the total nitrogen fuzzy prediction model to obtain the total nitrogen fuzzy prediction concentration value TN2, input the total nitrogen historical data prediction value TN1 and the total nitrogen fuzzy prediction concentration value TN2 into the total nitrogen precise prediction model to obtain the total nitrogen precise prediction value TN3.

2. The method for predicting nitrogen concentration in sewage based on artificial intelligence according to claim 1, characterized in that: The specific process of step 1.1 is: Step 1.1.1: Continuously collect historical ammonia nitrogen concentration data for m days and n items per day to form a training set; use the long short-term memory network algorithm to build an ammonia nitrogen historical data prediction model; Step 1.1.2: Take k consecutive ammonia nitrogen concentration data in the training set as input values, and the k+1th ammonia nitrogen concentration data as output values, and send the training set to the ammonia nitrogen historical data prediction model for training to obtain a trained ammonia nitrogen historical data prediction model; Step 1.1.3: Use the actual ammonia nitrogen concentration value of the month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen historical data prediction model to obtain N ammonia nitrogen historical data prediction values ​​to form a collection of ammonia nitrogen historical data prediction values.

3. The method for predicting nitrogen concentration in sewage based on artificial intelligence according to claim 1, characterized in that: The specific process of step 1.2 is: Step 1.2.1: Continuously collect historical data of n auxiliary variables and the corresponding actual ammonia nitrogen concentration values ​​for m days each day; Step 1.2.2: The data obtained in step 1.2.1 is used to construct a relationship model between ammonia nitrogen concentration value and auxiliary variables through a polynomial regression algorithm, which is called an ammonia nitrogen fuzzy prediction model; Step 1.2.3: Use the actual measured values ​​of the auxiliary variables one month before the ammonia nitrogen meter has no signal to input into the ammonia nitrogen fuzzy prediction model to obtain N ammonia nitrogen fuzzy prediction concentration values, and form a collection of ammonia nitrogen fuzzy prediction concentration values.

4. The method for predicting nitrogen concentration in sewage based on artificial intelligence according to claim 1, characterized in that: Step 1.3 is as follows: An ammonia nitrogen accurate prediction model is constructed by using the polynomial regression algorithm based on the actual ammonia nitrogen concentration value one month before the ammonia nitrogen meter had no signal, the corresponding ammonia nitrogen historical data prediction value collection, and the ammonia nitrogen fuzzy prediction concentration value collection.

5. The method for predicting nitrogen concentration in sewage based on artificial intelligence according to claim 1, characterized in that: The specific process of step 2.1 is: Step 2.1.1: Continuously collect historical total nitrogen concentration data for n times every day for m days to form a training set; use the long short-term memory network algorithm to build a total nitrogen historical data prediction model; Step 2.1.2: Take k consecutive total nitrogen concentration data in the training set as input values, and the k+1th total nitrogen concentration data as output values, and send the training set to the total nitrogen historical data prediction model for training to obtain a trained total nitrogen historical data prediction model; Step 2.1.3: Use the actual total nitrogen concentration value of the total nitrogen meter one month before no signal to input into the total nitrogen historical data prediction model, obtain N total nitrogen historical data prediction values, and form a collection of total nitrogen historical data prediction values.

6. The method for predicting nitrogen concentration in sewage based on artificial intelligence according to claim 1, characterized in that: The specific process of step 2.2 is: Step 2.2.1: Continuously collect historical data of n auxiliary variables and the corresponding true total nitrogen concentration values ​​for m days each day; Step 2.2.2: The data obtained in step 2.2.1 are used to construct a relationship model between total nitrogen concentration value and auxiliary variables through a polynomial regression algorithm, which is called a total nitrogen fuzzy prediction model; Step 2.2.3: Use the actual measured values ​​of the auxiliary variables one month before the total nitrogen instrument has no signal to input into the total nitrogen fuzzy prediction model to obtain N total nitrogen fuzzy prediction concentration values, and form a collection of total nitrogen fuzzy prediction concentration values.

7. The method for predicting nitrogen concentration in sewage based on artificial intelligence according to claim 1, characterized in that: The details of step 2.3 are as follows: An accurate total nitrogen prediction model was constructed using a polynomial regression algorithm using the actual total nitrogen concentration value one month before the total nitrogen meter had no signal, the corresponding total nitrogen historical data prediction value collection, and the total nitrogen fuzzy prediction concentration value collection.

Images