Intelligent prediction and feedback ammonia nitrogen concentration prediction method

Abstract

The invention discloses an intelligent prediction and feedback ammonia nitrogen concentration prediction method, which belongs to the field of inflow ammonia nitrogen concentration prediction in water treatment, and comprises the following steps: constructing an offset data set between ammonia nitrogen concentration value historical data and total nitrogen concentration value historical data; constructing a relationship between the bias data set and historical data of the auxiliary variables, and forming an ammonia nitrogen concentration prediction model; inputting the auxiliary variable at a to-be-predicted moment into the ammonia nitrogen concentration prediction model to obtain predicted bias data, and obtaining a theoretical prediction value of the ammonia nitrogen concentration according to the predicted bias data and the total nitrogen concentration value at the to-be-predicted moment; and determining a compensation coefficient according to the check variable to obtain a final ammonia nitrogen concentration prediction value. By introducing a bias data set and various auxiliary variables, a finer prediction model is constructed, errors caused by single variable prediction are reduced, a compensation coefficient is set, and the numerical value of the compensation coefficient is determined by checking variables, so that the ammonia nitrogen concentration prediction value is further modified, and the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of prediction of ammonia nitrogen concentration in water treatment, and in particular to a method for predicting ammonia nitrogen concentration with intelligent prediction and feedback. Background Art

[0002] Existing related technologies include:

[0003] CN107664682A: discloses a water quality soft measurement prediction method for ammonia nitrogen. The method forms a data sample set by acquiring data values ​​of multiple water quality monitoring indicators including ammonia nitrogen in the water environment to be tested, and further divides the data into a training set and a test set, and trains the training set using a fuzzy neural network algorithm to obtain an ammonia nitrogen soft measurement model; the model is tested using the test set, and the model establishment and testing are repeatedly performed until the obtained test results meet the preset conditions, and the ammonia nitrogen soft measurement model that meets the preset conditions is used as the final soft measurement model.

[0004] CN118298962A: discloses a method for predicting influent ammonia nitrogen concentration based on long short-term memory step sequence (LSTM). The method first constructs prediction input data and output data, and preprocesses the data, then uses the LSTM model to predict the ammonia nitrogen concentration, sets the model parameters to improve the prediction accuracy, and finally simulates the ammonia nitrogen concentration of the sewage plant influent and evaluates the prediction performance of the model.

[0005] CN117711521A: discloses a method for predicting the ammonia nitrogen content of compost using an artificial intelligence model. The method obtains environmental parameters and ammonia nitrogen content information during the composting process of organic solid waste, inputs this information into an elastic network to select the optimal environmental parameter combination, and then inputs it into an artificial neural network to predict the ammonia nitrogen content.

[0006] CN118886529A: Provides a method for predicting water quality in a sewage treatment plant based on an improved LSTM neural network model. The method first obtains water quality monitoring data from the sewage treatment plant, performs outlier cleaning, smoothing and normalization on the data, and then optimizes the LSTM model using a genetic algorithm, performs training and testing to obtain a water quality prediction model.

[0007] The problem with this type of patent is:

[0008] ① The amount of data required is large and the investment cost is high. The ammonia nitrogen prediction model is completely dependent on neural network training and requires massive data support, which increases the requirements for operating data; CN107664682A mentioned that in the implementation method, 24 water quality indicators were selected for modeling through three modelings; CN 117711521 A mentioned that the environmental information to be predicted in the composting process of organic solid waste includes many parameters such as number of days, temperature, total carbon, total nitrogen, compost volume, pH value, moisture content, carbon-nitrogen ratio, seed germination rate, conductivity, etc.; the influencing factors of ammonia nitrogen concentration in CN118298962A include PH, suspended solids (SS), water temperature (WT), air temperature, air pressure, relative humidity, rainfall, wind speed, average total cloud cover and visibility, etc. Too many indicators are required, which requires the support of instrument probes, and thus increases investment costs;

[0009] ② Low data utilization. CN107664682A optimizes the model through repeated training and testing, which increases development costs and may also lead to a long model training cycle, affecting the speed and efficiency of practical applications. Excessive data demand also complicates data collection and processing, reducing the response speed and flexibility of the system.

[0010] ③ Poor prediction accuracy. Its input data is completely based on instrument measurement data, and it is impossible to rule out the confusion of correlation caused by poor instrument measurement accuracy, which in turn affects the accuracy of the model. In addition, a single model is difficult to handle complex and changeable water quality conditions, especially when facing different types of sewage, which may lead to a decrease in prediction accuracy.

[0011] ④ Lack of feedback mechanism. Most existing technologies lack an effective feedback mechanism and are unable to self-correct and optimize based on the difference between actual operating data and predicted results. This defect causes the model to gradually lose accuracy in long-term operation, especially when faced with sudden changes in water quality, and the prediction strategy cannot be adjusted in time. Although CN107664682A and CN118298962A have conducted multiple model training and testing, they did not introduce a feedback loop of actual operating data, making it difficult to achieve continuous improvement.

[0012] In summary, although the existing ammonia nitrogen concentration prediction methods have achieved monitoring and prediction functions to a certain extent, they still have problems such as large data demand, low data utilization, poor prediction accuracy and no feedback mechanism in practical applications. These problems limit the widespread application and effect improvement of existing technologies in the field of sewage treatment. Summary of the invention

[0013] In view of the above problems existing in the prior art, the present invention provides an ammonia nitrogen concentration prediction method with intelligent prediction and feedback.

[0014] The present invention adopts the following technical solutions:

[0015] An ammonia nitrogen concentration prediction method with intelligent prediction and feedback comprises the following steps:

[0016] Step 1: Construct a biased data set between the historical data of ammonia nitrogen concentration values ​​and the historical data of total nitrogen concentration values. The biased data b in the biased data set i for:

[0017] b i =TN i -kAN i ;

[0018] Where k is the calibration coefficient, TN i is the total nitrogen concentration at time i, AN i is the ammonia nitrogen concentration value at time i;

[0019] Step 2: Construct the relationship between the biased data set and the historical data of the auxiliary variables; select the auxiliary variables as chemical oxygen demand, suspended solids concentration, pH value and water temperature, construct the functional relationship between the biased data set and the historical data of the auxiliary variables, and form an ammonia nitrogen concentration prediction model;

[0020] Step 3: Input the auxiliary variables at the time to be predicted into the ammonia nitrogen concentration prediction model to obtain the predicted bias data, and obtain the theoretical prediction value AN1 of the ammonia nitrogen concentration according to the predicted bias data and the total nitrogen concentration value at the time to be predicted;

[0021] Step 4: Determine the calibration variable as the ratio of the dissolved oxygen content in the biochemical aerobic tank to the ammonia nitrogen influent load. Determine the compensation coefficient a based on the calibration variable, and the final predicted ammonia nitrogen concentration is a*AN1.

[0022] Preferably, the value of k is set according to the inlet water carbon-nitrogen ratio C / N:

[0023] When C / N≥5, k=0.7; when 3<C / N<5, k=0.8; when C / N≤3, k=0.9.

[0024] Preferably, the historical data of ammonia nitrogen concentration values ​​refers to 24 historical ammonia nitrogen concentration values ​​collected for 30 consecutive days every day; the historical data of total nitrogen concentration values ​​refers to 24 historical total nitrogen concentration values ​​collected for 30 consecutive days every day; the historical data of auxiliary variables refers to 24 historical chemical oxygen demand, suspended solids concentration, pH value and water temperature values ​​collected for 30 consecutive days every day.

[0025] Preferably, a multivariate polynomial fitting method is used to bias the functional relationship between the data set and the auxiliary variable historical data.

[0026] Preferably, the inlet water carbon-nitrogen ratio is the ratio of the total chemical oxygen demand and the total nitrogen concentration of the inlet water.

[0027] Preferably, the process of determining the compensation coefficient a is:

[0028] After obtaining the theoretical predicted value of ammonia nitrogen concentration, 25%-50% of the actual hydraulic retention time of the biochemical pool is used as the calibration interval, the calibration variable value corresponding to 25% of the actual hydraulic retention time of the biochemical pool is C1, and the calibration variable value corresponding to 50% of the actual hydraulic retention time of the biochemical pool is C2. If |C1-C2|≤10%C1, the compensation coefficient a is set to 1; when C2-C1>10%C1, the compensation coefficient a is set to 0.8; when C1-C2>10%C1, the compensation coefficient a is set to 1.2.

[0029] When |C1-C2|≤10%C1, it means that the system outlet water fluctuation is small at this time, and there is no need to correct the prediction result; when C2-C1>10%C1, the ratio of the dissolved oxygen content in the aerobic tank to the influent ammonia nitrogen load increases rapidly, indicating that the aeration air volume is too high, and further indicating that the predicted ammonia nitrogen concentration is larger than the actual value, so it is necessary to set the compensation coefficient to 0.8 to correct the prediction result; when C1-C2>10%C1, the ratio of the dissolved oxygen content in the aerobic tank to the influent ammonia nitrogen load decreases rapidly, indicating that the aeration air volume is insufficient, and further indicating that the predicted ammonia nitrogen concentration is smaller than the actual value, so it is necessary to set the compensation coefficient to 1.2 to correct the prediction result.

[0030] The present invention has the following beneficial effects:

[0031] 1. High prediction accuracy. By introducing biased data sets and multiple auxiliary variables, a more sophisticated prediction model is constructed, the error caused by single variable prediction is reduced, and a compensation coefficient is set. The value of the compensation coefficient is determined by checking the variable, thereby further modifying the predicted value of ammonia nitrogen concentration and improving the prediction accuracy. In practical applications, this method can ensure that the deviation between the prediction result and the true value is less than 5% within 6 months, which is significantly better than the common error range in the prior art.

[0032] 2. Reduce the complexity and workload of prediction. Through artificial experience, a dual-mode prediction method of "experience + data fitting" was constructed based on the strong correlation between ammonia nitrogen and total nitrogen. The required data volume is reduced to only 30 days of continuous data, which greatly reduces the workload of data collection and processing, while maintaining the prediction accuracy.

[0033] 3. Enhance the adaptability and robustness of the model. The value range of the calibration coefficient k is clearly specified according to the influent carbon-nitrogen ratio C / N, making the model more suitable for sewage of different types and conditions, and improving its versatility and applicability.

[0034] 4. Optimize operational convenience and automation level. By building the relationship between biased data sets and auxiliary variables, the process of building the prediction model is simplified, the complex data processing steps are reduced, and the convenience and efficiency of operation are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the process of the present invention.

[0036] Figure 2 This is a schematic diagram of the predicted ammonia nitrogen concentration and the actual ammonia nitrogen concentration obtained in Example 1.

[0037] Figure 3 This is a schematic diagram of the predicted ammonia nitrogen concentration obtained in Comparative Example 1 and the actual value of ammonia nitrogen concentration.

[0038] Figure 4 This is a schematic diagram of the predicted ammonia nitrogen concentration obtained in Comparative Example 2 and the actual value of ammonia nitrogen concentration.

[0039] Figure 5 This is a schematic diagram of the predicted ammonia nitrogen concentration obtained in Comparative Example 3 and the actual value of ammonia nitrogen concentration. DETAILED DESCRIPTION

[0040] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments:

[0041] Example 1

[0042] Combination Figure 1 and Figure 2 The actual measured value of the carbon-nitrogen ratio C / N of a sewage treatment plant’s influent is 5.1, and the calibration coefficient k is 0.7.

[0043] Collect 24 historical ammonia nitrogen concentration values, historical total nitrogen concentration values, historical chemical oxygen demand, suspended solids concentration, pH value and water temperature values ​​every day for 30 consecutive days.

[0044] Step 1: Construct a biased data set between the historical data of ammonia nitrogen concentration values ​​and the historical data of ammonia nitrogen concentration values. The biased data b in the biased data set i for:

[0045] b i =TN i -0.7AN i .

[0046] Step 2: Construct the relationship between the biased data set and the historical data of auxiliary variables; select the auxiliary variables as chemical oxygen demand, suspended solids concentration, pH value and water temperature, where the biased data at time i corresponds to the auxiliary variables at time i, and use the multivariate polynomial fitting method to construct the functional relationship between the biased data set and the auxiliary variables to form an ammonia nitrogen concentration prediction model.

[0047] Step 3: Input the auxiliary variables at the time to be predicted into the ammonia nitrogen concentration prediction model to obtain the predicted bias data, and obtain the theoretical predicted value AN1 of the ammonia nitrogen concentration based on the predicted bias data and the ammonia nitrogen concentration value at the time to be predicted.

[0048] Step 4: Determine the calibration variable as the ratio of the dissolved oxygen content in the biochemical aerobic tank to the ammonia nitrogen influent load, and determine the compensation coefficient a based on the calibration variable. The final predicted value of ammonia nitrogen concentration is a*AN1.

[0049] After obtaining the theoretical predicted value of ammonia nitrogen concentration, 25%-50% of the actual hydraulic retention time of the biochemical pool is used as the calibration interval, the calibration variable value corresponding to 25% of the actual hydraulic retention time of the biochemical pool is C1, and the calibration variable value corresponding to 50% of the actual hydraulic retention time of the biochemical pool is C2. If |C1-C2|≤10%C1, the compensation coefficient a is set to 1; when C2-C1>10%C1, the compensation coefficient a is set to 0.8; when C1-C2>10%C1, the compensation coefficient a is set to 1.2.

[0050] In this embodiment, the predicted value of ammonia nitrogen concentration obtained by this method from January 1, 2023 to June 30, 2023 is as follows: Figure 2 As shown, from Figure 2 It can be seen that in 6 months, the true value of ammonia nitrogen concentration was 43.93±13.83 mg / L, the predicted value of ammonia nitrogen concentration was 44.91±13.99 mg / L, and the deviation was 4.71±3.82%.

[0051] Comparative Example 1

[0052] The measured value of the carbon-nitrogen ratio C / N of a sewage treatment plant’s influent is 4.2, and the calibration coefficient k is 0.8.

[0053] In this comparative example, only the process from step 1 to step 3 is used to obtain the predicted value of ammonia nitrogen concentration.

[0054] The predicted and actual values ​​of ammonia nitrogen concentration are as follows Figure 3 As shown, from Figure 3 It can be concluded that from January 1, 2023 to June 30, 2023, in the six months, the actual value of ammonia nitrogen concentration in January-March was 44.99±14.34mg / L, the predicted value of ammonia nitrogen concentration was 44.69±14.65mg / L, and the deviation was 3.9±3.15mg / L. However, in April-June, the actual value of ammonia nitrogen concentration was 42.88±13.22mg / L, and the predicted value of ammonia nitrogen concentration was 45.27±13.36mg / L, with a deviation of 8.82±7.52%, which has deviated significantly from the actual value.

[0055] Comparative Example 2

[0056] The measured value of the carbon-nitrogen ratio C / N of a sewage treatment plant’s influent is 5.3, and the calibration coefficient k is 0.8.

[0057] In this embodiment, the predicted value of ammonia nitrogen concentration obtained by this method from July 1, 2023 to December 31, 2023 is as follows: Figure 4 As shown, from Figure 4 It can be concluded that the actual value of the ammonia nitrogen concentration in 6 months is 47.86±17.00 mg / L, the predicted value of the ammonia nitrogen concentration is 49.95±17.59 mg / L, and the deviation is 10.09±8.44%. Therefore, the value of the calibration coefficient has a great influence on the accuracy of the prediction of the present invention.

[0058] Comparative Example 3

[0059] The measured value of the carbon-nitrogen ratio C / N of a sewage treatment plant’s influent is 5.3, and the calibration coefficient k is 0.6.

[0060] In this embodiment, the predicted value of ammonia nitrogen concentration obtained by this method from July 1, 2023 to December 31, 2023 is as follows: Figure 5 As shown, from Figure 5 It can be concluded that the actual value of the ammonia nitrogen concentration in 6 months is 47.86±17.00 mg / L, the predicted value of the ammonia nitrogen concentration is 49.47±17.45 mg / L, and the deviation is 10.07±8.29%. Therefore, the value of the calibration coefficient has a great influence on the accuracy of the prediction of the present invention.

[0061] 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 ammonia nitrogen concentration with intelligent prediction and feedback, characterized in that: The following steps are involved: Step 1: Construct a biased data set between the historical data of ammonia nitrogen concentration values ​​and the historical data of total nitrogen concentration values. The biased data b in the biased data set i for: b i =TN i -can i ; Where k is the calibration coefficient, TN i is the total nitrogen concentration at time i, AN i is the ammonia nitrogen concentration value at time i; Step 2: Construct the relationship between the biased data set and the historical data of the auxiliary variables; select the auxiliary variables as chemical oxygen demand, suspended solids concentration, pH value and water temperature, construct the functional relationship between the biased data set and the historical data of the auxiliary variables, and form an ammonia nitrogen concentration prediction model; Step 3: Input the auxiliary variables at the time to be predicted into the ammonia nitrogen concentration prediction model to obtain the predicted bias data, and obtain the theoretical prediction value AN1 of the ammonia nitrogen concentration according to the predicted bias data and the total nitrogen concentration value at the time to be predicted; Step 4: Determine the calibration variable as the ratio of the dissolved oxygen content in the biochemical aerobic tank to the ammonia nitrogen influent load. Determine the compensation coefficient a based on the calibration variable, and the final predicted ammonia nitrogen concentration is a*AN1.

2. The method for predicting ammonia nitrogen concentration by intelligent prediction and feedback according to claim 1, characterized in that: The value of k is set according to the inlet water carbon-nitrogen ratio C / N: When C / N≥5, k=0.7; when 3<C / N<5, k=0.8; when C / N≤3, k=0.

9.

3. The method for predicting ammonia nitrogen concentration with intelligent prediction and feedback according to claim 1, characterized in that: The historical data of ammonia nitrogen concentration values ​​refers to the historical ammonia nitrogen concentration values ​​collected for 30 consecutive days, 24 values ​​per day; the historical data of total nitrogen concentration values ​​refers to the historical total nitrogen concentration values ​​collected for 30 consecutive days, 24 values ​​per day; the historical data of auxiliary variables refers to the historical chemical oxygen demand, suspended solids concentration, pH value and water temperature values ​​collected for 30 consecutive days, 24 values ​​per day.

4. The method for predicting ammonia nitrogen concentration with intelligent prediction and feedback according to claim 1, characterized in that: The multivariate polynomial fitting method is used to bias the functional relationship between the data set and the historical data of the auxiliary variables.

5. The method for predicting ammonia nitrogen concentration with intelligent prediction and feedback according to claim 1, characterized in that: The inlet carbon-nitrogen ratio is the ratio of the total chemical oxygen demand and the total nitrogen concentration of the inlet water.

6. The method for predicting ammonia nitrogen concentration with intelligent prediction and feedback according to claim 1, characterized in that: The process of determining the compensation coefficient a is: After obtaining the theoretical predicted value of ammonia nitrogen concentration, 25%-50% of the actual hydraulic retention time of the biochemical pool is used as the calibration interval, the calibration variable value corresponding to 25% of the actual hydraulic retention time of the biochemical pool is C1, and the calibration variable value corresponding to 50% of the actual hydraulic retention time of the biochemical pool is C2. If |C1-C2|≤10%C1, the compensation coefficient a is set to 1; when C2-C1>10%C1, the compensation coefficient a is set to 0.8; when C1-C2>10%C1, the compensation coefficient a is set to 1.2.

Images