Sewage quality prediction method based on double-model combined application

Abstract

The invention provides a sewage quality prediction method based on double-model combined application, and relates to the technical field of environmental protection. The invention discloses a sewage quality prediction method based on double-model combined application. The method comprises the following steps: constructing a mechanism model of a sewage treatment system by utilizing modeling software; collecting data of the sewage treatment system, and constructing a data model of the sewage treatment system by adopting a machine learning method based on python software; the collected data comprises water quality data; the mechanism model and the data model are jointly applied, and the final output value is the water quality prediction condition of the sewage treatment system. According to the sewage quality prediction system and method based on double-model combined application, the defect of single-model use of a mechanism model or a data model is overcome, and application and development of an intelligent management and control technology of a sewage treatment system are promoted.

Description

Technical Field

[0001] This application belongs to the field of environmental protection technologies. Specifically, it relates to a sewage water quality prediction method based on the combined application of dual models. Background Art

[0002] Sewage biological treatment methods represented by the activated sludge process are widely used worldwide. The sewage treatment process mainly relies on the action of microorganisms, and the treatment process of microorganisms is affected by various factors such as complex mechanisms, a large number of reactions, and significant external influencing factors. Currently, in the operation and management practice of sewage treatment facilities, mechanism models or data models are usually adopted.

[0003] Mechanism model: To improve the management level of sewage treatment facilities, reduce sewage treatment energy consumption, and reduce the risk of excessive sewage discharge, the International Water Association (IWA) established a research group for the mathematical model of the activated sludge process design and operation in 1987 and successfully launched ASM1. This model mainly describes the biochemical reaction process in the activated sludge process, including the degradation of organic matter, nitrogen transformation, and phosphorus release. In 1995, the ASM2 model was launched, further refining the biochemical reaction process in ASM1. Subsequently, the ASM2d model containing denitrification and phosphorus-accumulating bacteria was launched, enabling the model to more accurately describe the nitrogen and phosphorus transformation processes in the sewage biological treatment process. In 1998, the ASM3 model was successfully developed. This model has a deeper understanding of the activated sludge process and particularly focuses on the important role of intracellular storage substances in biological metabolism. The ASMS series of models provide important support for the operation and management of sewage treatment facilities.

[0004] Data model: With the development of AI technology, the construction of sewage treatment models using machine learning methods has gradually emerged. By installing sensors and monitoring devices, the AI system can collect data on key indicators such as water quality, flow rate, and pollutant concentration in real time, and through the processing and analysis of AI algorithms, reveal the operation status and potential problems of sewage treatment. This intelligent management method not only improves the efficiency and accuracy of sewage treatment but also optimizes the decision-making process and reduces operating costs.

[0005] However, in the operation and management practice of sewage treatment facilities using the model construction method, the process of building a mechanism model is complex and the modeling cost is high, while building a data model has problems such as low model accuracy and high data quality requirements, which limits the practical application of water quality models in the operation and management of sewage treatment facilities. Summary of the Invention

[0006] The purpose of the present invention is to propose a sewage water quality prediction method based on the joint application of dual models, overcome the disadvantages of using a single mechanism model or a data model, and promote the application and development of intelligent management and control technology for sewage treatment systems.

[0007] The purpose of the present invention is achieved through the following technical solutions:

[0008] A sewage quality prediction method based on the combined application of dual models comprises the following steps:

[0009] S1. Use modeling software to construct a mechanism model of the sewage treatment system;

[0010] S2. Collect data of the sewage treatment system and construct a data model of the sewage treatment system using a machine learning method based on Python software; the collected data includes water quality data;

[0011] S3. The above mechanism model and data model are jointly applied, and the final output value is the water quality prediction or system operation status of the sewage treatment system.

[0012] The water quality prediction logic of this application is: first divide the data into 3 categories, which we can name A (influent water quality indicators, including influent COD / ammonia nitrogen / TN / TP, etc.), B (process operation parameters, including DO / reflux ratio, etc.), and C (effluent water quality indicators, including influent COD / ammonia nitrogen / TN / TP, etc.).

[0013] Data model: Take ABC as a data group, import it into the model for training, and build a preliminary model. After the training is completed, we will test the accuracy. When the accuracy meets the requirements, the model can be considered to have been built. When applying, we can output the remaining indicators by inputting one or two of A / B / C (generally using AC to derive B, or AB to derive C). In this way, we can use it as a prediction of effluent water quality and as a control of operating parameters.

[0014] Mechanism model: Based on the existing commercial software, input the parameters of the sewage treatment facility (including but not limited to tank capacity, DO, process flow, etc.) to build a preliminary model. After the model is built, input A (same as above) and B (same as above) to get C (same as above).

[0015] In some embodiments of the present application, the modeling software in the above-mentioned S1 step includes Biowin, AQUASIM, GPS-X or WEST, preferably the Biowin platform, which is a professional modeling software for sewage treatment facilities on sale. The modeling steps and processes can refer to the modeling guide provided by the software; the machine learning methods in the S2 step include SVR, KNN, GBDT and BP neural network.

[0016] In some embodiments of the present application, the data collected in the above S2 step further includes operating parameters and energy consumption index data.

[0017] In some embodiments of the present application, the above water quality data includes the COD, ammonia nitrogen content, total nitrogen content and pH value of the sewage; the operating parameters include DO, reflux ratio and MLSS; the energy consumption index includes power consumption and chemical consumption. The water quality data is an essential item, and the operating parameters, energy consumption index, etc. are non-essential items, which can be selectively added according to the functions required by the model.

[0018] In some embodiments of the present application, the combined application of the above S3 step includes a conventional application mode and a rapid application mode.

[0019] In some embodiments of the present application, the above conventional application mode is specifically as follows: Input the data into the mechanism model for data verification. If the requirements are not met, the program terminates. If the requirements are met, the prediction result A is obtained. At the same time, the data is transmitted to the data model, and the data model verifies the data. If the requirements are met, the prediction result B is obtained. The prediction result A and the prediction result B are mutually verified A and the result is output.

[0020] In some embodiments of the present application, the above rapid application mode is specifically as follows: Input the data into the mechanism model and the data model simultaneously. The data model obtains the prediction result D, and the mechanism model conducts data verification. If the requirements are not met, the program terminates. If the requirements are met, the prediction result C is obtained. The prediction result C and the prediction result D are mutually verified B and the result is output. When in use, the data input into the mechanism model and the data model is the influent water quality index, including influent CAD / ammonia nitrogen / TN / TP, etc.

[0021] In some embodiments of the present application, the above data verification is to judge whether the data conforms to the general water quality law through the constructed mechanism model, such as the influent water quality index cannot be a negative value or an abnormally high value, etc.; the data transmission is to transmit the data to the data model through software or manual means. The prediction results A and C are the values output after prediction by the mechanism model. The prediction results B and D are the values output after prediction by the data model.

[0022] In some embodiments of the present application, the specific method of the above mutual verification A is as follows: For the indicators that both the prediction result A and the prediction result B have, take 75%A + 25%B as the output result; for the indicators unique to the prediction result B, take 100%B as the output result.

[0023] In some embodiments of the present application, the specific method of the above mutual verification B is: for the indicators that both the prediction results C and the prediction results D have, the deviation value between the prediction results D and the prediction results C is determined. If the deviation value is greater than 20%, 100% C is used as the output result; if the deviation value is less than or equal to 20%, 75% C + 25% D is used as the output result. The deviation value calculation formula is: For the unique index of the prediction result D, 100% D is taken as the output result after judgment. The judgment is realized manually or by introducing boundary conditions in advance.

[0024] If the deviation between the mechanism model and the data model reaches more than 20%, it means that unreliable results appear in the two.

[0025] The advantage of the mechanism model is that the data is relatively accurate, but it has complex uses. The data model is simple and diverse to use, but the model accuracy is somewhat poor (sometimes due to problems in model establishment, it will lead to rather outrageous data).

[0026] Based on this, when there is a large deviation between the prediction results of the mechanism model and the data model, the output value of the mechanism model is preferred.

[0027] When the deviation between the two data values ​​is not large, the value can be taken by taking the output value of the mechanism model as the main value and the output value of the data model as the auxiliary value. After some data simulation and comparison, the present application found that 75%-25% is a better value taking method.

[0028] In another case, as mentioned above, the use of data models is more flexible, so more types of related data can be artificially embedded to obtain the desired prediction results, but the mechanism model is limited by the mechanism and cannot do this. In some cases, data types unique to the data model will appear. In this case, we compare the data types shared by the two models. If the numerical deviation is within 20%, we will use 100% D's result as the final output result. If the numerical deviation is greater than 20%, we will not be able to output the data types unique to D. This is what is described in the article as "after judgment."

[0029] Beneficial effects of the present invention:

[0030] 1. The method of this application uses two models in combination, avoiding the disadvantages of using a single model as a prediction method, namely, the mechanism model is inconvenient to use and the data model is prone to inaccuracy.

[0031] 2. This method can provide a new idea for the development of sewage treatment system models.

[0032] The main solution of the present application and its various further alternative solutions can be freely combined to form multiple solutions, all of which are solutions that can be adopted and claimed in the present application; moreover, in the present application, (each non-conflicting alternative) alternatives can be freely combined with each other and with other alternatives. Those skilled in the art can understand that there are various combinations according to the prior art and common general knowledge after understanding the solutions of the present application, and all of them are the technical solutions to be protected by the present application, which will not be enumerated here. Brief Description of the Drawings

[0033] Figure 1 It is a schematic diagram of the conventional application mode of the joint application of the dual models of the present application.

[0034] Figure 2 It is a schematic diagram of the fast application mode of the joint application of the dual models of the present application.

[0035] Figure 3 It is a sewage process flow chart for the process flow modeling of the Biowin software interface in Embodiment 1 of the present application.

[0036] Figure 4 It is the setting of the structure parameters for the process flow modeling of the Biowin software interface in Embodiment 1 of the present application.

[0037] Figure 5 It is the influent water quality input interface of the mechanism model in Embodiment 1 of the present application.

[0038] Figure 6 It is the operation interface of the machine learning model based on python in Embodiment 1 of the present application.

[0039] Figure 7 It is the result output of the data model in Embodiment 1 of the present application. Detailed Description of the Preferred Embodiments

[0040] The following non-limiting embodiments are used to illustrate the present application.

[0041] Embodiment 1:

[0042] Based on the need for intelligent operation and regulation of the sewage treatment system, a dual-model joint modeling application is carried out on a sewage treatment plant with a scale of 20,000 m 3 / day in Sichuan. The specific implementation steps are as follows:

[0043] S1: Collect the main process design parameters of the sewage treatment plant, and carry out process flow modeling on the Biowin software interface, including but not limited to the selection of treatment structures, the connection of sewage pipelines, the connection of sludge pipelines, the input of structure parameters, etc. (where the sewage process flow chart is as Figure 3 , and the structure parameter setting is as Figure 4 shown);

[0044] S2: On the basis of completing the construction of the S1 model, by inputting the influent water quality indicators, the effluent water quality is predicted relying on the constructed model (as Figure 5 shown);

[0045] S3: Based on the Python software, a data model of the sewage treatment system is constructed using machine learning methods. The main steps include but are not limited to the preprocessing of the data set, the selection of machine learning algorithms, the calibration of the model, etc. The modeling data for this case are: influent water quality data (influent COD / NH 3 -N / TN / TP), operating parameters (DO / reflux), effluent water quality data (effluent COD / NH 3 -N / TN / TP), as Figure 6 and Figure 7 shown.

[0046] S4: Based on several scenarios of the combined application of the dual models in this case.

[0047] The influent water quality indicator A of this case (as shown in Table 1), the operating parameter indicator B (as shown in Table 2), and the effluent water quality data C (as shown in Table 3)

[0048] Table 1 Influent Water Quality Indicator A of the Sewage Treatment Plant (from January to December 2023, only taking the example from 0:00 to 24:00 on July 14th as an example)

[0049] Time COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 0 150 17.6 25.1 2.0 2 140 18.4 26.3 2.1 4 145 16.8 24.5 2.2 6 128 16.1 24.6 2.2 8 160 16.2 23.1 2.5 10 181 14.9 20.3 2.6 12 159 14.4 21.8 2.1 14 166 14.1 20.8 1.9 16 158 15.9 22.5 1.9 18 154 16.1 22.9 2.1 20 160 16.8 23.6 2.3 22 120 18.2 25.8 2.5 24 114 17.3 24.4 2.3

[0050] Table 2 Operating Parameter Indicator B of the Sewage Treatment Plant (from January to December 2023, only taking the example from 0:00 to 24:00 on July 14th as an example)

[0051]

[0052]

[0053] Table 3 Effluent Water Quality Indicator C of the Sewage Treatment Plant (from January to December 2023, only taking the example from 0:00 to 24:00 on July 14th as an example)

[0054] Time COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 0 18 1.3 6.1 0.2 2 20 1.1 6.9 0.1 4 21 1.2 5.8 0.1 6 25 1.2 5.9 0.1 8 23 1.1 6.3 0.1 10 24 0.9 7.1 0.2 12 29 0.9 7.8 0.1 14 22 0.7 7.9 0.1 16 24 0.8 8.1 0.1 18 21 0.9 8.3 0.1 20 19 1.0 7.5 0.2 22 15 1.0 7.3 0.2 24 10 0.9 7.6 0.1

[0055] Usage Scenario 1:

[0056] Obtain the influent water quality indicator D at 10:00 am on May 10, 2024, and the operating parameter indicator E of the sewage treatment plant (as shown in Table 5) through the on-line detection equipment of the sewage treatment plant. The indicator values are as follows (as shown in Table 4).

[0057] Table 4 Influent Water Quality Indicator D of the Sewage Treatment Plant (at 10:00 am on May 10, 2024)

[0058]

[0059]

[0060] Table 5 Operating Parameter Indexes of Sewage Treatment Plant E (at 10:00 am on May 10, 2024)

[0061]

[0062] Input data D and E into the mechanism model and the data model respectively to obtain the output result F (as shown in Table 6);

[0063] Table 6 Effluent Indexes F of Sewage Treatment Plant

[0064] Model used COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) Mechanistic model 20 1.2 6.2 0.1 Data model 18 1.2 5.8 0.1 Deviation 11.11％ 0％ 6.89％ 0％

[0065] As shown in Table 6, the deviation of the effluent quality results predicted by the two models is less than 20%, which corroborates each other (for the indexes that both prediction result A and prediction result B have, take 75% A + 25% B as the output result; for the indexes that prediction result B uniquely has, take 100% B as the output result). After that, the final output result is shown in Table 7 below.

[0066] Table 7 Output Results of Effluent Quality Indexes of Sewage Treatment Plant Based on the Joint Application of Dual Models (at 10:00 am on May 10, 2024)

[0067] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 19.5 1.2 6.1 0.1

[0068] The effluent indexes measured by the online detection equipment of the sewage treatment facilities are as shown in Table 8 below.

[0069] Table 8 Indexes Read by the Online Detection Instruments for the Effluent of Sewage Treatment Facilities (at 10:00 am on May 10, 2024)

[0070] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 19.0 1.2 6.0 0.1

[0071] Comparing Table 6, Table 7, and Table 8, it can be seen that the predicted effluent quality values obtained based on the dual - model prediction are closer to the true measured values than those predicted by the individual mechanism model and data model.

[0072] Usage Scenario 2:

[0073] Obtain the influent water quality index G at 10:00 am on June 1, 2024 through the online detection equipment of the sewage treatment plant. The index values are as follows (as shown in Table 9), and the operating parameter indexes H of the sewage treatment plant (as shown in Table 10).

[0074] Table 9 Influent Water Quality Indexes G of Sewage Treatment Plant (at 10:00 am on June 1, 2024)

[0075] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 212 30.8 37.9 2.5

[0076] Table 10 Operating Parameter Indexes of the Sewage Treatment Plant H (at 10:00 am on June 1, 2024)

[0077]

[0078] Input data G and H into the mechanism model and the data model respectively to obtain the output result I (as shown in Table 11);

[0079] Table 11 Effluent Index I of the Sewage Treatment Plant

[0080] Model used COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) Mechanistic model 30 1.4 8.0 0.2 Data model 22 1.1 6.2 0.1 Deviation 36.4％ 27.3％ 29％ 100％

[0081] As shown in Table 11, the deviation of the effluent water quality results predicted by the two models is greater than 20%, which corroborates each other (for the indexes that both prediction result C and prediction result D have, judge the deviation value between prediction result D and prediction result C. If the deviation value is greater than 20%, then use 100% C as the output result; if the deviation value is less than or equal to 20%, then use 75% C + 25% D as the output result. The formula for the deviation value is: For the indexes unique to prediction result D, take 100% D as the output result), and the final output result is shown in Table 12 below.

[0082] Table 12 Output Results of the Effluent Water Quality Indexes of the Sewage Treatment Plant Based on the Joint Application of Dual Models (at 10:00 am on June 1, 2024)

[0083] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 30 1.4 8.0 0.2

[0084] The effluent indexes measured by the online detection equipment for the sewage treatment facilities are as shown in Table 13 below.

[0085] Table 13 Indexes Read by the Online Detection Instruments for the Effluent of the Sewage Treatment Facilities (at 10:00 am on June 1, 2024)

[0086] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 28 1.3 8.1 0.2

[0087] Comparing Table 11, Table 12, and Table 13, it can be seen that the predicted values of the effluent water quality obtained based on the dual-model prediction are closer to the true measured values than the values predicted by the individual mechanism model and data model.

[0088] Usage Scenario 3:

[0089] When predicting indicators other than the effluent quality, the mechanism model is limited by the model and cannot be used inversely. In this case, to predict the optimal operating parameters of the sewage treatment facility (which can actually guide the operation and regulation of the sewage treatment plant and has great practical significance), the influent quality indicators at 10:00 am on June 15, 2024, and the effluent quality indicators that meet the discharge requirements of the sewage treatment facility (obtained from the current discharge standards of urban sewage treatment plants, and the "Discharge Standards for Water Pollutants in the Minjiang River and Tuojiang River Basins of Sichuan Province" (DB51 / 2311—2016) is cited in this project) are adopted. The specific input data are shown in Tables 14 and 15 below.

[0090] Table 14 Influent Quality Indicators J of the Sewage Treatment Plant (at 10:00 am on June 15, 2024)

[0091] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 157 22.6 30.9 2.0

[0092] Table 15 Effluent Quality Indicators K of the Sewage Treatment Plant (Effluent Discharge Standards of Urban Sewage Treatment Plants)

[0093] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 30 1.5 10 0.3

[0094] Input data J and K into the data model respectively to obtain the output result L (as shown in Table 16);

[0095] Table 16 Output Indicators L of the Operating Parameters of the Sewage Treatment Plant Based on the Application of the Mechanism Model (at 10:00 am on June 15, 2024)

[0096]

[0097] The actual effluent quality indicators and operating parameters obtained by the sewage treatment plant are as shown in the following table:

[0098] Table 17 Indicators Read by the On-line Detection Instrument for the Effluent of the Sewage Treatment Facility (at 10:00 am on June 15, 2024)

[0099] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 18 1.0 6.4 0.1

[0100] Table 18 Operating Parameter Indicators of the Sewage Treatment Plant (at 10:00 am on June 15, 2024)

[0101]

[0102] It can be seen from the comparison of Table 15 - Table 18 that the effluent quality indicators of the actual operation of the sewage treatment plant are better than the discharge standards (compare Table 15 and Table 17), but the relative energy consumption is greater (compare Table 16 and Table 18). For the actual operation and maintenance of the sewage treatment plant, keeping the effluent indicators close to the discharge limit and reducing energy consumption is the best operation strategy. Based on the results of the model simulation, the operation parameters are adjusted accordingly (adjusted according to the model output results in Table 16). After the adjustment, the operation is carried out for 8 hours, and the effluent quality is shown in Table 19.

[0103] Table 19 Indicators Read by the On - line Detection Instrument for the Effluent of the Sewage Treatment Facility (18:00 p.m., June 15, 2024)

[0104] COD (mg / L) <![CDATA[NH 3 -N (mg / L)]]> TN (mg / L) TP (mg / L) 28 1.3 9.2 0.2

[0105] It can be seen that the effluent of the sewage treatment plant is closer to the discharge standard than the original operation parameters and is within the sewage effluent discharge limit range.

[0106] The foregoing basic examples of the present application and its various further alternative examples can be freely combined to form multiple embodiments, all of which are embodiments that can be adopted and claimed in the present application. In the solution of the present application, any alternative example can be arbitrarily combined with any other basic example and alternative example.

[0107] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A sewage quality prediction method based on the joint application of dual models, characterized in that: The following steps are involved: S1. Use modeling software to construct a mechanism model of the sewage treatment system; S2. Collect data of the sewage treatment system and construct a data model of the sewage treatment system using a machine learning method based on Python software; the collected data includes water quality data; S3. The above mechanism model and data model are jointly applied, and the final output value is the water quality prediction or system operation status of the sewage treatment system.

2. A sewage quality prediction method based on the combined application of dual models according to claim 1, characterized in that: The modeling software in step S1 includes Biowin, AQUASIM, GPS-X or WEST; the machine learning method in step S2 includes SVR, KNN, GBDT and BP neural network.

3. A sewage quality prediction method based on the combined application of dual models according to claim 1, characterized in that: The data collected in step S2 also include operating parameters and energy consumption index data.

4. A sewage quality prediction method based on the combined application of dual models according to claim 3, characterized in that: The water quality data include COD, ammonia nitrogen content, total nitrogen content and pH value of the sewage; the operating parameters include DO, reflux ratio and MLSS; the energy consumption indicators include electricity consumption and drug consumption.

5. A sewage quality prediction method based on the combined application of dual models according to any one of claims 1 to 4, characterized in that: The combined application of the S3 step includes a regular application mode and a fast application mode.

6. A sewage quality prediction method based on the combined application of dual models according to claim 5, characterized in that: The conventional application mode is specifically as follows: inputting data into the mechanism model and performing data verification. If the data does not meet the requirements, the program terminates. If the data meets the requirements, a prediction result A is obtained. At the same time, the data is transmitted to the data model, and the data model verifies the data. If the data meets the requirements, a prediction result B is obtained. The prediction results A and B are mutually verified. If the results differ by less than 10%, they can be considered to be credible data and the results are output.

7. The method for predicting sewage quality based on the combined application of dual models according to claim 5, characterized in that: The rapid application mode is specifically as follows: data is input into the mechanism model and the data model at the same time, and the data model obtains the prediction result D; the mechanism model verifies the data, and if it does not meet the requirements, the program terminates, and if it meets the requirements, the prediction result C is obtained, and the prediction result C and the prediction result D are mutually verified B and the result is output.

8. A sewage quality prediction method based on the combined application of dual models according to claim 6 or 7, characterized in that: The data verification is to determine whether the data conforms to the general water quality law through the constructed mechanism model; the data transmission is to transfer the data to the data model through software or manual methods.

9. The method for predicting sewage quality based on the combined application of dual models according to claim 6, characterized in that: The specific method of mutually verifying A is: for indicators that are common to both prediction result A and prediction result B, 75% A+25% B is taken as the output result; for indicators that are unique to prediction result B, 100% B is taken as the output result.

10. The method for predicting sewage quality based on the combined application of dual models according to claim 7, characterized in that: The specific method of mutual verification B is: for the indicators that both prediction results C and prediction results D have, determine the deviation value between prediction results D and prediction results C. If the deviation value is greater than 20%, 100% C is used as the output result; If the deviation value is less than or equal to 20%, 75%C+25%D is used as the output result. The deviation value calculation formula is: For indicators unique to the prediction result D, 100% D is taken as the output result.

Images