Methods and devices for predicting the generation of flue gas pollutants

Abstract

This invention relates to the field of pollution control technology and discloses a method and apparatus for predicting the generation of flue gas pollutants. The invention first acquires the current flue gas concentration data of the target detection area, then constructs a corresponding three-dimensional feature vector based on the current flue gas concentration data, and inputs the three-dimensional feature vector into a preset recurrent neural network model to obtain the corresponding first predicted concentration data. Next, the current flue gas concentration data and historical flue gas concentration data are input into the preset recurrent neural network model to obtain second predicted concentration data. Finally, based on the weight values ​​of the first and second predicted concentration data, the predicted flue gas generation data for the target detection area over a future period is obtained. Thus, by combining the current first predicted concentration data with the historical trend of the second predicted concentration data, relatively accurate predicted flue gas generation data can be obtained. Furthermore, in conjunction with an automated control program, this allows for proactive control of flue gas pollutant emissions, reducing emissions.

Description

Technical Field

[0001] This invention relates to the field of pollution control technology, and in particular to a method and apparatus for predicting the generation of flue gas pollutants. Background Technology

[0002] With societal development and the continuous increase in factories and motor vehicles, numerous environmental problems have arisen, with air pollution becoming increasingly serious. Air pollutants can enter the human body and cause harm, severely impacting human health. Industry is generally a major source of air pollution, with most pollutants originating from coal combustion. Coal combustion in power plants generates large amounts of sulfur dioxide, nitrogen oxides, particulate matter, mercury, and other flue gas pollutants. To effectively protect atmospheric health, prevention and control at the source are necessary. Achieving ultra-low emissions from power plants generally requires highly sophisticated flue gas treatment equipment, resulting in significant costs. Therefore, how to further control flue gas pollutant emissions from power plants is an urgent problem to be solved.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method and apparatus for predicting the generation of flue gas pollutants, aiming to solve the technical problem of how to further reduce the emission of flue gas pollutants from thermal power plants in the prior art.

[0005] To achieve the above objectives, the present invention provides a method for predicting the generation of flue gas pollutants, the method comprising the following steps:

[0006] Obtain the current flue gas concentration data within the target detection area;

[0007] The current flue gas concentration data is preprocessed, and a three-dimensional feature vector corresponding to the current flue gas concentration data is constructed based on the preprocessing results;

[0008] The three-dimensional feature vector is input into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area;

[0009] The current flue gas concentration data and historical flue gas concentration data are input into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area;

[0010] The first predicted concentration data and the second predicted concentration data are weighted according to preset weight values ​​to obtain the predicted flue gas generation data of the target detection area within a certain period of time in the future.

[0011] Optionally, the step of obtaining the current flue gas concentration data within the target detection area includes:

[0012] Acquire the first flue gas concentration data at the first preset time point within the target detection area;

[0013] Obtain the second flue gas concentration data at a second preset time point within the target detection area;

[0014] The first flue gas concentration data and the second flue gas concentration data are normalized according to the normalization algorithm to obtain the current flue gas concentration data corresponding to the target detection area.

[0015] Optionally, the step of preprocessing the current flue gas concentration data and constructing a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing result includes:

[0016] Obtain the temperature data at each detection time point corresponding to the current flue gas concentration data;

[0017] Obtain the air pressure data of the space at each detection time point corresponding to the current flue gas concentration data;

[0018] Based on the temperature data, the air pressure data, and the current flue gas concentration data, a three-dimensional feature vector corresponding to the current flue gas concentration data is constructed.

[0019] Optionally, before the step of acquiring the current flue gas concentration data within the target detection area, the method further includes:

[0020] Obtain historical flue gas concentration data corresponding to the target detection area;

[0021] Obtain the historical temperature and historical air pressure corresponding to the historical flue gas concentration data;

[0022] The historical flue gas concentration data, historical temperature, and historical air pressure are input into the initial neural network model for iterative training to obtain a recurrent neural network model.

[0023] Optionally, the step of inputting the historical flue gas concentration data, the historical temperature, and the historical air pressure into the initial neural network model for iterative training to obtain a recurrent neural network model includes:

[0024] The historical flue gas concentration data, the historical temperature, and the historical air pressure are input into the initial neural network model for iterative training to obtain the trained model.

[0025] The detection data is input into the training model, and it is determined whether the output result reaches the prediction range.

[0026] When the output result does not reach the prediction range, historical flue gas concentration data is input into the training model for reverse training until the output result reaches the prediction range, and the training model corresponding to the output result reaching the prediction range is used as a recurrent neural network model.

[0027] Furthermore, to achieve the above objectives, the present invention also proposes a flue gas pollutant generation prediction device, the device comprising:

[0028] The data acquisition module is used to acquire the current flue gas concentration data within the target detection area;

[0029] The data processing module is used to preprocess the current flue gas concentration data and construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing results.

[0030] The concentration prediction module is used to input the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area;

[0031] The concentration prediction module is also used to input the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area;

[0032] The generation quantity display module is used to perform weight processing on the first predicted concentration data and the second predicted concentration data according to a preset weight value to obtain the predicted flue gas generation quantity data of the target detection area within a certain period of time in the future.

[0033] Optionally, the data acquisition module is further configured to acquire the first flue gas concentration data at a first preset time point within the target detection area;

[0034] The data acquisition module is also used to acquire the second flue gas concentration data at a second preset time point within the target detection area;

[0035] The data acquisition module is further configured to perform normalization processing on the first flue gas concentration data and the second flue gas concentration data according to a normalization algorithm to obtain the current flue gas concentration data corresponding to the target detection area.

[0036] Optionally, the data processing module is further configured to acquire temperature data at each detection time point corresponding to the current flue gas concentration data;

[0037] The data processing module is also used to obtain the air pressure data of the space at each detection time point corresponding to the current flue gas concentration data;

[0038] The data processing module is further configured to construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, the air pressure data, and the current flue gas concentration data.

[0039] Optionally, the device further includes: a model building module;

[0040] The model building module is used to obtain historical flue gas concentration data corresponding to the target detection area;

[0041] The model building module is also used to obtain the historical temperature and historical pressure corresponding to the historical flue gas concentration data;

[0042] The model building module is also used to input the historical flue gas concentration data, the historical temperature and the historical air pressure into the initial neural network model for iterative training to obtain a recurrent neural network model.

[0043] Optionally, the model building module is further configured to input the historical flue gas concentration data, the historical temperature and the historical air pressure into the initial neural network model for iterative training to obtain a trained model;

[0044] The model building module is also used to input the detection data into the training model and determine whether the output result reaches the prediction range;

[0045] The model building module is further configured to input historical flue gas concentration data into the training model for reverse training when the output result does not reach the prediction range, until the output result reaches the prediction range, and use the training model corresponding to the output result reaching the prediction range as a recurrent neural network model.

[0046] This invention acquires current flue gas concentration data within a target detection area; preprocesses the current flue gas concentration data, constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing result; then inputs the three-dimensional feature vector into a preset recurrent neural network model to obtain first predicted concentration data corresponding to the target detection area; inputs the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain second predicted concentration data corresponding to the target detection area; finally, weights the first predicted concentration data and the second predicted concentration data according to preset weight values ​​to obtain predicted flue gas generation data for a certain period of time in the future for the target detection area. This invention constructs a preset recurrent neural network model, inputting current and historical flue gas concentration data into the model to obtain first and second predicted concentration data. Based on these data, it predicts the amount of flue gas generated in the future. Thus, by combining the current first predicted concentration data with the historical trend of the second predicted concentration data, a relatively accurate predicted amount of flue gas generated can be obtained. This, in conjunction with an automated control program, enables proactive control of flue gas pollutant emissions, reducing not only emissions but also material consumption while ensuring emissions meet standards. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the first embodiment of the flue gas pollutant generation prediction method of the present invention.

[0048] Figure 2 This is a flowchart illustrating the second embodiment of the flue gas pollutant generation prediction method of the present invention.

[0049] Figure 3 This is a structural block diagram of the first embodiment of the flue gas pollutant generation prediction device of the present invention.

[0050] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0051] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0052] This invention provides a method for predicting the generation of flue gas pollutants, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the flue gas pollutant generation prediction method of the present invention.

[0053] In this embodiment, the method for predicting the amount of flue gas pollutants generated includes the following steps:

[0054] Step S10: Obtain the current flue gas concentration data within the target detection area.

[0055] It should be noted that the executing entity of the method in this embodiment can be a computing service device with data processing and program execution functions, such as a tablet computer or a server, or other electronic devices capable of performing the same or similar functions, such as the aforementioned flue gas pollutant generation prediction device. This embodiment does not limit this. Here, the flue gas pollutant generation prediction method provided in this embodiment and the following embodiments will be specifically described using the aforementioned flue gas pollutant generation prediction device (hereinafter referred to as the prediction device).

[0056] Understandably, the target detection area is the area selected to predict the concentration of flue gas pollutants generated over a future period. Since the measurement involves flue gas pollutant concentrations from thermal power plants, and because flue gas pollutants are mobile, the amount generated in each area varies. To ensure the accuracy of the predicted flue gas pollutant generation data, the selected target detection area should generally not be too large and should be located in a confined space. This can be determined based on actual conditions, and this embodiment does not impose any restrictions on this.

[0057] It should be understood that the current flue gas concentration data refers to the current concentration of flue gas pollutants in the target detection area. It is important to note that "current" is a relative concept. If only flue gas concentration data is measured at a specific point in time, then the current flue gas concentration data is the flue gas concentration data for that target detection area at that specific point in time. If it is the average of multiple flue gas concentration measurements taken over a certain time period (e.g., several minutes, half an hour, etc.), then the current flue gas concentration data is the flue gas concentration data for that target detection area during that time period.

[0058] It should be noted that since flue gas pollutants generated by thermal power plants are continuously produced, the concentration of flue gas pollutants changes every moment. Generally, multiple measurements can be taken, or measurements can be taken only when the flue gas density in the space reaches a certain level, to ensure the accuracy of the current flue gas concentration measurement. For example, detection devices can be installed at multiple locations within the target detection area to obtain the average value of the flue gas concentration at multiple locations as the current flue gas concentration data; multiple measurements can also be taken within a certain period of time, and the average value of the flue gas concentration within that period of time can be used as the current flue gas concentration data; other methods are also possible, and this embodiment does not limit them.

[0059] In practical applications, current flue gas concentration data can be obtained using sensors (such as gas sensors). When flue gas pollutants are present in the sensor's environment, the sensor's conductivity increases with the concentration of pollutants in the air. A simple circuit can be used to convert the change in conductivity into an output signal corresponding to the gas concentration. Alternatively, other measurement methods can be used; this embodiment does not limit this approach.

[0060] In practice, users can use a prediction device to obtain the current flue gas concentration data within the target detection area based on the actual conditions of the detection area.

[0061] Step S20: Preprocess the current flue gas concentration data, and construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing results.

[0062] It should be noted that the three-dimensional feature vector reflects the current flue gas concentration data of the target detection area. It is important to note that since the three-dimensional feature vector is input into the preset recurrent neural network model, the constructed three-dimensional feature vector should be compatible with the preset recurrent neural network model.

[0063] Understandably, the acquired current flue gas concentration data includes not only the concentration data of various pollutants, but also data such as the detection time, detection temperature, and air pressure of the detection space. The current flue gas concentration data can be preprocessed, for example, by obtaining the detection time point and the corresponding detection temperature, and constructing the corresponding three-dimensional feature vector based on the selected data.

[0064] In its implementation, the prediction device preprocesses the current flue gas concentration data and constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing results.

[0065] Step S30: Input the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area.

[0066] Step S40: Input the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area.

[0067] It should be noted that the preset recurrent neural network model is a recurrent neural network model pre-set in the prediction device. This preset recurrent neural network model is trained in advance based on relevant flue gas pollutant concentration data, and the constructed three-dimensional feature vector is also matched with this preset recurrent neural network model.

[0068] Understandably, the first predicted concentration data is based on the three-dimensional eigenvector corresponding to the current flue gas concentration data. It's worth noting that since the current flue gas concentration data represents the pollutant concentration measured at the current moment, the first predicted concentration data is also based on the pollutant concentration measured at the current moment. It predicts the flue gas concentration for the next moment, clearly reflecting the current concentration change and closely correlated with the current concentration. However, due to the limited data, predicting data over a longer period can easily lose accuracy. Therefore, a second prediction can be made by combining historical flue gas concentration data.

[0069] It should be understood that historical flue gas concentration data refers to the concentration data of flue gas pollutants obtained from historical monitoring of the target detection area. During the generation of flue gas pollutants in a thermal power plant, this historical flue gas concentration data can be the concentration data of flue gas pollutants generated on the current day, or it can be the historical flue gas concentration data of all equipment under the same power conditions. It can indicate the change in the amount of flue gas pollutants generated in the target detection area based on the characteristics of the equipment, or it can represent the trend of the generated flue gas pollutant concentration based on the flue gas pollutants generated on the current day. This embodiment does not impose any limitations on this.

[0070] It should be noted that the second predicted concentration data is based on current and historical flue gas concentration data. It is important to note that since historical flue gas concentration data clearly indicates changes in the amount of pollutants generated in the target detection area, the second predicted concentration data, derived from historical data, can also reflect these changes. However, because it reflects a trend, it is difficult to effectively reflect the current concentration of pollutants. In thermal power plants, flue gas pollutant concentrations generally change rapidly; therefore, combining the second predicted concentration data with the first predicted concentration data yields a more accurate prediction.

[0071] In its specific implementation, the prediction device inputs the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area, and inputs the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area.

[0072] Step S50: Perform weighting processing on the first predicted concentration data and the second predicted concentration data according to the preset weight value to obtain the predicted flue gas generation data of the target detection area within a certain period of time in the future.

[0073] It should be noted that the preset weight values ​​are weight values ​​set in advance in the prediction device. The weight values ​​of the first and second predicted concentration data can be set according to the actual situation of the target detection area. For example, if the flue gas pollutants generated by a thermal power plant change significantly at any given moment, the weight value of the first predicted concentration data can be increased, thereby improving the accuracy of the prediction data.

[0074] Understandably, the predicted flue gas generation data is a prediction of the amount of flue gas generated over a future period relative to the current flue gas concentration data. Generally, the predicted flue gas generation data can be from the last few minutes or the last half hour, and can be adjusted according to changes in the current flue gas concentration data. If the flue gas pollutant generation in a thermal power plant is unstable, flue gas generation data for a shorter period can be predicted. Generally, the longer the prediction period, the greater the error.

[0075] In practice, the prediction device performs weighted processing on the first and second predicted concentration data according to preset weight values, and finally obtains the predicted flue gas generation data of the target detection area within a certain period of time in the future.

[0076] Furthermore, in order to make the obtained current flue gas concentration data more accurate, this embodiment includes step S10 as follows: acquiring first flue gas concentration data at a first preset time point within the target detection area; acquiring second flue gas concentration data at a second preset time point within the target detection area; and performing normalization processing on the first flue gas concentration data and the second flue gas concentration data according to a normalization algorithm to obtain the current flue gas concentration data corresponding to the target detection area.

[0077] It should be noted that the first and second preset time points are pre-set time points within the current detection period. To ensure more accurate current flue gas concentration data, measurements can also be taken when the power plant equipment is stably generating flue gas pollutants.

[0078] Understandably, the first flue gas concentration data and the second flue gas concentration data are concentration data obtained by detecting flue gas pollutants at the first preset time point and the second preset time point.

[0079] It should be understood that "first" and "second" are merely names for different detection time points and do not represent the quality of the detection time points. This embodiment does not impose any limitations on this.

[0080] It should be noted that by applying a normalization algorithm to the first and second flue gas concentration data, a more average current flue gas concentration data for the current time period can be obtained.

[0081] In a specific implementation, the prediction device acquires first flue gas concentration data at a first preset time point within the target detection area; acquires second flue gas concentration data at the first preset time point within the target detection area; and performs normalization processing on the first flue gas concentration data and the second flue gas concentration data according to a normalization algorithm, thereby making the current flue gas concentration data corresponding to the target detection area more accurate.

[0082] Furthermore, to make the predicted flue gas generation data more accurate, step S20 in this embodiment includes: acquiring temperature data at each detection time point corresponding to the current flue gas concentration data; acquiring air pressure data in the space where each detection time point corresponding to the current flue gas concentration data is located; and constructing a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, the air pressure data, and the current flue gas concentration data.

[0083] It should be noted that the detection time point is the time point at which flue gas pollutants are detected in the target detection area, which corresponds to the preset time point mentioned above.

[0084] Understandably, the temperature data is measured when detecting pollutants in flue gas produced by thermal power plants. Generally speaking, gases are significantly affected by temperature changes, and flue gas pollutants are also gases; the higher the temperature, the greater the concentration change. Therefore, the temperature data at the current moment can reflect the accuracy of the current flue gas concentration data quite well.

[0085] It should be understood that the air pressure data is the gas pressure measured when detecting pollutants in flue gas produced by thermal power plants. Generally speaking, gases are significantly affected by changes in air pressure and temperature. Flue gas pollutants are also gases, and the higher the air pressure, the greater the concentration change. Therefore, the air pressure data at the current moment can also reflect the accuracy of the current flue gas concentration data.

[0086] Understandably, a three-dimensional feature vector corresponding to the current flue gas concentration can be constructed based on temperature data, air pressure data, and current flue gas concentration data. The corresponding predicted concentration data can be obtained by inputting the three-dimensional feature vector into a preset recurrent neural network model.

[0087] In a specific implementation, the prediction device acquires the temperature data at each detection time point of the current flue gas concentration data; acquires the air pressure data of the space where each detection time point of the current flue gas concentration data is located; and constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, air pressure data, and current flue gas concentration data, so that the predicted flue gas generation data generated by the preset recurrent neural network model is more accurate.

[0088] In this embodiment, the user can use a prediction device to obtain the current flue gas concentration data within the target detection area based on the actual conditions of the detection area. Then, the user preprocesses the current flue gas concentration data and constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing results. Next, the user can use the prediction device to input the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area. The user can then input the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area. Finally, the user performs weight processing on the first and second predicted concentration data according to preset weight values ​​to obtain the predicted flue gas generation data for the target detection area within a certain future time. This invention constructs a preset recurrent neural network model, inputting current and historical flue gas concentration data into the model to obtain first and second predicted concentration data. Based on these data, it predicts the amount of flue gas generated in the future. Thus, by combining the current first predicted concentration data with the historical trend of the second predicted concentration data, a relatively accurate predicted amount of flue gas generated can be obtained. This, in conjunction with an automated control program, enables proactive control of flue gas pollutant emissions, reducing not only emissions but also material consumption while ensuring emissions meet standards.

[0089] refer to Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the flue gas pollutant generation prediction method of the present invention.

[0090] Based on the first embodiment described above, in this embodiment, the construction of the preset recurrent neural network model is consistent with the construction of the three-dimensional feature vector. The obtained preset recurrent neural network model can quickly output predicted concentration data based on the three-dimensional feature vector, thereby improving the operating speed of the prediction device. Before step S10, the following steps are also included:

[0091] Step S01: Obtain historical flue gas concentration data corresponding to the target detection area.

[0092] Step S02: Obtain the historical temperature and historical air pressure corresponding to the historical flue gas concentration data.

[0093] It should be noted that historical temperature refers to the temperature at which flue gas pollutants were detected in the target detection area over a past period, and corresponds to the temperature data mentioned above. Historical air pressure refers to the air pressure at which flue gas pollutants were detected in the target detection area over a past period, and corresponds to the air pressure data mentioned above.

[0094] In practice, the prediction device can acquire historical flue gas concentration data corresponding to the target detection area; then it can acquire historical temperature and historical air pressure corresponding to the historical flue gas concentration data.

[0095] Step S03: Input the historical flue gas concentration data, the historical temperature and the historical air pressure into the initial neural network model for iterative training to obtain a recurrent neural network model.

[0096] It should be noted that the initial convolutional network model is the initial model built based on the convolutional network for iterative training. Historical flue gas concentration data, historical temperature, and historical air pressure are input into the initial convolutional network model to define the initial data of the initial convolutional network model. The model is trained based on the changes in historical flue gas concentration data to predict the changing trends of flue gas pollutants. Finally, the recurrent neural network model can be determined from the trained model.

[0097] Understandably, the resulting recurrent neural network model has long-term memory capabilities, such as the LSTM neural network.

[0098] In the specific implementation, historical flue gas concentration data, historical temperature, and historical air pressure are input into the initial neural network model for iterative training to obtain a recurrent neural network model.

[0099] Furthermore, to improve the accuracy of the preset recurrent neural network model, step S03 in this embodiment includes: inputting the historical flue gas concentration data, the historical temperature, and the historical air pressure into the initial neural network model for iterative training to obtain a training model; inputting the detection data into the training model to determine whether the output result reaches the prediction range; when the output result does not reach the prediction range, inputting the historical flue gas concentration data into the training model for reverse training until the output result reaches the prediction range, and using the training model corresponding to when the output result reaches the prediction range as the recurrent neural network model.

[0100] It should be noted that the trained model is the one obtained after initial iterations of training the initial convolutional network model. At this point, the accuracy of the trained model has not been verified, and the accuracy level is uncontrollable. Therefore, further training is needed to improve the accuracy of the trained model.

[0101] Understandably, the output result can be a set accuracy rate. For example, the detection data is input into the training model, and the output is the concentration data predicted by the training model based on the detection data. At this point, the predicted concentration data is used to determine whether it meets the prediction range based on historical flue gas concentration data. Finally, the accuracy rate is used to determine whether further training of the training model is needed. It is worth noting that the accuracy rate can be determined manually or by other configured accuracy rate determination programs; this embodiment does not impose any restrictions on this.

[0102] It should be understood that the prediction range is a pre-set condition used to determine whether further training of the model is needed based on the output results. For example, if the accuracy of the final prediction range is relatively low, a preset condition can be set for an accuracy greater than a certain value, such as 90%. If the accuracy is greater than 90%, no further training is needed; if the accuracy is less than 90%, the model should be trained further. Backpropagation can be used for reverse training to improve the accuracy of the trained model.

[0103] It should be noted that reverse training can be performed using the backpropagation algorithm. The backpropagation algorithm is a suitable learning algorithm for multi-layer neural networks, and its principles can also be applied to convolutional networks. The input-output relationship of a network using the backpropagation algorithm is essentially a mapping relationship. For example, an initial convolutional network with n inputs and m outputs performs a continuous mapping from n-dimensional space to a finite field in m-dimensional space. This mapping is highly nonlinear.

[0104] In the specific implementation, historical flue gas concentration data, historical temperature, and historical air pressure are first input into the initial neural network model for iterative training to obtain a training model. Then, the detection data is input into the training model to determine whether the output result reaches the prediction range. If the output result does not reach the prediction range, historical flue gas concentration data is input into the training model for reverse training until the output result reaches the prediction range. The training model corresponding to the output result reaching the prediction range is used as the recurrent neural network model, thereby improving the accuracy of the preset recurrent neural network model.

[0105] In practical applications, a pre-trained recurrent neural network model can be integrated into existing flue gas pollutant concentration detection devices to form a predictive device. This device can then make an advanced prediction of pollutant generation based on existing flue gas pollutant concentration data. The device may include instruments, sensors, and range conversion and display instruments, and possesses its own computing power to compute the pre-trained recurrent neural network model. The sensors can collect data on pollutants in the flue gas, including oxygen, carbon dioxide, sulfur dioxide, and particulate matter concentrations. After collecting the data, a prediction of pollutant concentrations for a given period can be made.

[0106] This embodiment acquires historical flue gas concentration data corresponding to the target detection area; then, it acquires the historical temperature and historical pressure corresponding to the historical flue gas concentration data. Finally, it inputs the historical flue gas concentration data, historical temperature, and historical pressure into an initial neural network model for iterative training to obtain a recurrent neural network model. The resulting pre-set recurrent neural network model can quickly output predicted concentration data based on three-dimensional feature vectors, thereby improving the operating speed of the prediction device. Furthermore, it can also first input the historical flue gas concentration data, historical temperature, and historical pressure into the initial neural network model for iterative training to obtain a training model; then, it inputs the detection data into the training model to determine whether the output result reaches the prediction range; if the output result does not reach the prediction range, it inputs the historical flue gas concentration data into the training model for reverse training until the output result reaches the prediction range, and uses the training model corresponding to the output result reaching the prediction range as the recurrent neural network model, thereby improving the accuracy of the pre-set recurrent neural network model.

[0107] Furthermore, this invention also proposes a flue gas pollutant generation prediction device, which implements the steps of the flue gas pollutant generation prediction method described above when executed.

[0108] refer to Figure 3 , Figure 3 This is a structural block diagram of the first embodiment of the flue gas pollutant generation prediction device of the present invention.

[0109] In this embodiment, the flue gas pollutant generation prediction device includes:

[0110] Data acquisition module 301: Used to acquire the current flue gas concentration data within the target detection area.

[0111] It should be noted that the executing entity of the method in this embodiment can be a computing service device with data processing and program execution functions, such as a tablet computer or a server, or other electronic devices capable of performing the same or similar functions, such as the aforementioned flue gas pollutant generation prediction device. This embodiment does not limit this. Here, the flue gas pollutant generation prediction method provided in this embodiment and the following embodiments will be specifically described using the aforementioned flue gas pollutant generation prediction device (hereinafter referred to as the prediction device).

[0112] Understandably, the target detection area is the area selected to predict the concentration of flue gas pollutants generated over a future period. Since the measurement involves flue gas pollutant concentrations from thermal power plants, and because flue gas pollutants are mobile, the amount generated in each area varies. To ensure the accuracy of the predicted flue gas pollutant generation data, the selected target detection area should generally not be too large and should be located in a confined space. This can be determined based on actual conditions, and this embodiment does not impose any restrictions on this.

[0113] It should be understood that the current flue gas concentration data refers to the current concentration of flue gas pollutants in the target detection area. It is important to note that "current" is a relative concept. If only flue gas concentration data is measured at a specific point in time, then the current flue gas concentration data is the flue gas concentration data for that target detection area at that specific point in time. If it is the average of multiple flue gas concentration measurements taken over a certain time period (e.g., several minutes, half an hour, etc.), then the current flue gas concentration data is the flue gas concentration data for that target detection area during that time period.

[0114] It should be noted that since flue gas pollutants generated by thermal power plants are continuously produced, the concentration of flue gas pollutants changes every moment. Generally, multiple measurements can be taken, or measurements can be taken only when the flue gas density in the space reaches a certain level, to ensure the accuracy of the current flue gas concentration measurement. For example, detection devices can be installed at multiple locations within the target detection area to obtain the average value of the flue gas concentration at multiple locations as the current flue gas concentration data; multiple measurements can also be taken within a certain period of time, and the average value of the flue gas concentration within that period of time can be used as the current flue gas concentration data; other methods are also possible, and this embodiment does not limit them.

[0115] In practical applications, current flue gas concentration data can be obtained using sensors (such as gas sensors). When flue gas pollutants are present in the sensor's environment, the sensor's conductivity increases with the concentration of pollutants in the air. A simple circuit can be used to convert the change in conductivity into an output signal corresponding to the gas concentration. Alternatively, other measurement methods can be used; this embodiment does not limit this approach.

[0116] In practice, users can use a prediction device to obtain the current flue gas concentration data within the target detection area based on the actual conditions of the detection area.

[0117] The data processing module 302 is used to preprocess the current flue gas concentration data and construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing result.

[0118] It should be noted that the three-dimensional feature vector reflects the current flue gas concentration data of the target detection area. It is important to note that since the three-dimensional feature vector is input into the preset recurrent neural network model, the constructed three-dimensional feature vector should be compatible with the preset recurrent neural network model.

[0119] Understandably, the acquired current flue gas concentration data includes not only the concentration data of various pollutants, but also data such as the detection time, detection temperature, and air pressure of the detection space. The current flue gas concentration data can be preprocessed, for example, by obtaining the detection time point and the corresponding detection temperature, and constructing the corresponding three-dimensional feature vector based on the selected data.

[0120] In its implementation, the prediction device preprocesses the current flue gas concentration data and constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing results.

[0121] The concentration prediction module 303 is used to input the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area;

[0122] The concentration prediction module 303 is also used to input the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area.

[0123] It should be noted that the preset recurrent neural network model is a recurrent neural network model pre-set in the prediction device. This preset recurrent neural network model is trained in advance based on relevant flue gas pollutant concentration data, and the constructed three-dimensional feature vector is also matched with this preset recurrent neural network model.

[0124] Understandably, the first predicted concentration data is based on the three-dimensional eigenvector corresponding to the current flue gas concentration data. It's worth noting that since the current flue gas concentration data represents the pollutant concentration measured at the current moment, the first predicted concentration data is also based on the pollutant concentration measured at the current moment. It predicts the flue gas concentration for the next moment, clearly reflecting the current concentration change and closely correlated with the current concentration. However, due to the limited data, predicting data over a longer period can easily lose accuracy. Therefore, a second prediction can be made by combining historical flue gas concentration data.

[0125] It should be understood that historical flue gas concentration data refers to the concentration data of flue gas pollutants obtained from historical monitoring of the target detection area. During the generation of flue gas pollutants in a thermal power plant, this historical flue gas concentration data can be the concentration data of flue gas pollutants generated on the current day, or it can be the historical flue gas concentration data of all equipment under the same power conditions. It can indicate the change in the amount of flue gas pollutants generated in the target detection area based on the characteristics of the equipment, or it can represent the trend of the generated flue gas pollutant concentration based on the flue gas pollutants generated on the current day. This embodiment does not impose any limitations on this.

[0126] It should be noted that the second predicted concentration data is based on current and historical flue gas concentration data. It is important to note that since historical flue gas concentration data clearly indicates changes in the amount of pollutants generated in the target detection area, the second predicted concentration data, derived from historical data, can also reflect these changes. However, because it reflects a trend, it is difficult to effectively reflect the current concentration of pollutants. In thermal power plants, flue gas pollutant concentrations generally change rapidly; therefore, combining the second predicted concentration data with the first predicted concentration data yields a more accurate prediction.

[0127] In its specific implementation, the prediction device inputs the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area, and inputs the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area.

[0128] The generation quantity display module 304 is used to perform weight processing on the first predicted concentration data and the second predicted concentration data according to a preset weight value to obtain the predicted flue gas generation quantity data of the target detection area within a certain period of time in the future.

[0129] It should be noted that the preset weight values ​​are weight values ​​set in advance in the prediction device. The weight values ​​of the first and second predicted concentration data can be set according to the actual situation of the target detection area. For example, if the flue gas pollutants generated by a thermal power plant change significantly at any given moment, the weight value of the first predicted concentration data can be increased, thereby improving the accuracy of the prediction data.

[0130] Understandably, the predicted flue gas generation data is a prediction of the amount of flue gas generated over a future period relative to the current flue gas concentration data. Generally, the predicted flue gas generation data can be from the last few minutes or the last half hour, and can be adjusted according to changes in the current flue gas concentration data. If the flue gas pollutant generation in a thermal power plant is unstable, flue gas generation data for a shorter period can be predicted. Generally, the longer the prediction period, the greater the error.

[0131] In practice, the prediction device performs weighted processing on the first and second predicted concentration data according to preset weight values, and finally obtains the predicted flue gas generation data of the target detection area within a certain period of time in the future.

[0132] Furthermore, to make the obtained current flue gas concentration data more accurate, the data acquisition module 301 in this embodiment is also used to acquire the first flue gas concentration data at a first preset time point within the target detection area; the data acquisition module 301 is also used to acquire the second flue gas concentration data at a second preset time point within the target detection area; the data acquisition module 301 is also used to perform normalization processing on the first flue gas concentration data and the second flue gas concentration data according to a normalization algorithm to obtain the current flue gas concentration data corresponding to the target detection area.

[0133] It should be noted that the first and second preset time points are pre-set time points within the current detection period. To ensure more accurate current flue gas concentration data, measurements can also be taken when the power plant equipment is stably generating flue gas pollutants.

[0134] Understandably, the first flue gas concentration data and the second flue gas concentration data are concentration data obtained by detecting flue gas pollutants at the first preset time point and the second preset time point.

[0135] It should be understood that "first" and "second" are merely names for different detection time points and do not represent the quality of the detection time points. This embodiment does not impose any limitations on this.

[0136] It should be noted that by applying a normalization algorithm to the first and second flue gas concentration data, a more average current flue gas concentration data for the current time period can be obtained.

[0137] In a specific implementation, the prediction device acquires first flue gas concentration data at a first preset time point within the target detection area; acquires second flue gas concentration data at the first preset time point within the target detection area; and performs normalization processing on the first flue gas concentration data and the second flue gas concentration data according to a normalization algorithm, thereby making the current flue gas concentration data corresponding to the target detection area more accurate.

[0138] Furthermore, to make the predicted flue gas generation data more accurate, the data processing module 302 in this embodiment is also used to acquire the temperature data at each detection time point corresponding to the current flue gas concentration data; the data processing module 302 is also used to acquire the air pressure data of the space where each detection time point corresponding to the current flue gas concentration data is located; the data processing module 302 is also used to construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, the air pressure data and the current flue gas concentration data.

[0139] It should be noted that the detection time point is the time point at which flue gas pollutants are detected in the target detection area, which corresponds to the preset time point mentioned above.

[0140] Understandably, the temperature data is measured when detecting pollutants in flue gas produced by thermal power plants. Generally speaking, gases are significantly affected by temperature changes, and flue gas pollutants are also gases; the higher the temperature, the greater the concentration change. Therefore, the temperature data at the current moment can reflect the accuracy of the current flue gas concentration data quite well.

[0141] It should be understood that the air pressure data is the gas pressure measured when detecting pollutants in flue gas produced by thermal power plants. Generally speaking, gases are significantly affected by changes in air pressure and temperature. Flue gas pollutants are also gases, and the higher the air pressure, the greater the concentration change. Therefore, the air pressure data at the current moment can also reflect the accuracy of the current flue gas concentration data.

[0142] Understandably, a three-dimensional feature vector corresponding to the current flue gas concentration can be constructed based on temperature data, air pressure data, and current flue gas concentration data. The corresponding predicted concentration data can be obtained by inputting the three-dimensional feature vector into a preset recurrent neural network model.

[0143] In a specific implementation, the prediction device acquires the temperature data at each detection time point of the current flue gas concentration data; acquires the air pressure data of the space where each detection time point of the current flue gas concentration data is located; and constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, air pressure data, and current flue gas concentration data, so that the predicted flue gas generation data generated by the preset recurrent neural network model is more accurate.

[0144] In this embodiment, the user can use a prediction device to obtain the current flue gas concentration data within the target detection area based on the actual conditions of the detection area. Then, the user preprocesses the current flue gas concentration data and constructs a three-dimensional feature vector corresponding to the current flue gas concentration data based on the preprocessing results. Next, the user can use the prediction device to input the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area. The user can then input the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area. Finally, the user performs weight processing on the first and second predicted concentration data according to preset weight values ​​to obtain the predicted flue gas generation data for the target detection area within a certain future time. This invention constructs a preset recurrent neural network model, inputting current and historical flue gas concentration data into the model to obtain first and second predicted concentration data. Based on these data, it predicts the amount of flue gas generated in the future. Thus, by combining the current first predicted concentration data with the historical trend of the second predicted concentration data, a relatively accurate predicted amount of flue gas generated can be obtained. This, in conjunction with an automated control program, enables proactive control of flue gas pollutant emissions, reducing not only emissions but also material consumption while ensuring emissions meet standards.

[0145] Based on the first embodiment of the above-mentioned device, in this embodiment, the construction of the preset recurrent neural network model is consistent with the construction of the three-dimensional feature vector. The obtained preset recurrent neural network model can quickly output the predicted concentration data according to the three-dimensional feature vector, thereby improving the running speed of the prediction device. The device also includes: a model construction module 305.

[0146] The model building module 305 is used to obtain historical flue gas concentration data corresponding to the target detection area;

[0147] The model building module 305 is also used to obtain the historical temperature and historical pressure corresponding to the historical flue gas concentration data;

[0148] It should be noted that historical temperature refers to the temperature at which flue gas pollutants were detected in the target detection area over a past period, and corresponds to the temperature data mentioned above. Historical air pressure refers to the air pressure at which flue gas pollutants were detected in the target detection area over a past period, and corresponds to the air pressure data mentioned above.

[0149] In practice, the prediction device can acquire historical flue gas concentration data corresponding to the target detection area; then it can acquire historical temperature and historical air pressure corresponding to the historical flue gas concentration data.

[0150] The model building module 305 is also used to input the historical flue gas concentration data, the historical temperature and the historical air pressure into the initial neural network model for iterative training to obtain a recurrent neural network model.

[0151] It should be noted that the initial convolutional network model is the initial model built based on the convolutional network for iterative training. Historical flue gas concentration data, historical temperature, and historical air pressure are input into the initial convolutional network model to define the initial data of the initial convolutional network model. The model is trained based on the changes in historical flue gas concentration data to predict the changing trends of flue gas pollutants. Finally, the recurrent neural network model can be determined from the trained model.

[0152] Understandably, the resulting recurrent neural network model has long-term memory capabilities, such as the LSTM neural network.

[0153] In the specific implementation, historical flue gas concentration data, historical temperature, and historical air pressure are input into the initial neural network model for iterative training to obtain a recurrent neural network model.

[0154] Furthermore, to improve the accuracy of the preset recurrent neural network model, the model building module 305 in this embodiment is also used to input the historical flue gas concentration data, the historical temperature, and the historical air pressure into the initial neural network model for iterative training to obtain a training model; the model building module 305 is also used to input the detection data into the training model and determine whether the output result reaches the prediction range; the model building module 305 is also used to input the historical flue gas concentration data into the training model for reverse training when the output result does not reach the prediction range, until the output result reaches the prediction range, and use the training model corresponding to the output result reaching the prediction range as the recurrent neural network model.

[0155] It should be noted that the trained model is the one obtained after initial iterations of training the initial convolutional network model. At this point, the accuracy of the trained model has not been verified, and the accuracy level is uncontrollable. Therefore, further training is needed to improve the accuracy of the trained model.

[0156] Understandably, the output result can be a set accuracy rate. For example, the detection data is input into the training model, and the output is the concentration data predicted by the training model based on the detection data. At this point, the predicted concentration data is used to determine whether it meets the prediction range based on historical flue gas concentration data. Finally, the accuracy rate is used to determine whether further training of the training model is needed. It is worth noting that the accuracy rate can be determined manually or by other configured accuracy rate determination programs; this embodiment does not impose any restrictions on this.

[0157] It should be understood that the prediction range is a pre-set condition used to determine whether further training of the model is needed based on the output results. For example, if the accuracy of the final prediction range is relatively low, a preset condition can be set for an accuracy greater than a certain value, such as 90%. If the accuracy is greater than 90%, no further training is needed; if the accuracy is less than 90%, the model should be trained further. Backpropagation can be used for reverse training to improve the accuracy of the trained model.

[0158] It should be noted that reverse training can be performed using the backpropagation algorithm. The backpropagation algorithm is a suitable learning algorithm for multi-layer neural networks, and its principles can also be applied to convolutional networks. The input-output relationship of a network using the backpropagation algorithm is essentially a mapping relationship. For example, an initial convolutional network with n inputs and m outputs performs a continuous mapping from n-dimensional space to a finite field in m-dimensional space. This mapping is highly nonlinear.

[0159] In the specific implementation, historical flue gas concentration data, historical temperature, and historical air pressure are first input into the initial neural network model for iterative training to obtain a training model. Then, the detection data is input into the training model to determine whether the output result reaches the prediction range. If the output result does not reach the prediction range, historical flue gas concentration data is input into the training model for reverse training until the output result reaches the prediction range. The training model corresponding to the output result reaching the prediction range is used as the recurrent neural network model, thereby improving the accuracy of the preset recurrent neural network model.

[0160] In practical applications, a pre-trained recurrent neural network model can be integrated into existing flue gas pollutant concentration detection devices to form a predictive device. This device can then make advanced predictions about pollutant generation based on existing flue gas pollutant concentration data. The device may include instruments, sensors, and range conversion and display instruments, and possesses its own computing power to compute the pre-trained recurrent neural network model. The sensors can collect data on pollutants in the flue gas, including oxygen, carbon dioxide, sulfur dioxide, and particulate matter concentrations.

[0161] After obtaining the data, a prediction of pollutant concentrations for the next period of time can be made.

[0162] This embodiment obtains historical flue gas concentration data corresponding to the target detection area; then it obtains historical temperature and historical air pressure corresponding to the historical flue gas concentration data; finally, it combines the historical flue gas concentration data with historical data...

[0163] Historical temperature and historical air pressure are input into an initial neural network model for iterative training to obtain a recurrent neural network model. The resulting pre-defined recurrent neural network model can quickly output predictions based on three-dimensional feature vectors.

[0164] Concentration data can improve the operating speed of the prediction device.

[0165] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to...

[0166] This encompasses non-exclusive inclusion, such that a process, method, article, or system that includes a list of elements includes not only those elements but also other elements not explicitly listed, or also includes...

[0167] An element inherent to such a process, method, article, or system. Without further limitations, an element specified by the phrase "including one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0168] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0169] 0. Through the above description of the embodiments, those skilled in the art can clearly understand the above.

[0170] The implementation methods can be achieved using software and necessary general-purpose hardware platforms, or hardware, but in many cases the former is a better implementation. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product.

[0171] The computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone) to...

[0172] Computers, servers, or network devices, etc., may execute the methods described in the various embodiments of the present invention.

[0173] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for predicting the generation of pollutants in flue gas, characterized in that, The method includes: Acquire current flue gas concentration data within the target detection area in the confined space of a thermal power plant; Obtain the temperature data and air pressure data of the space at each detection time point corresponding to the current flue gas concentration data; Construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, the air pressure data, and the current flue gas concentration data; The three-dimensional feature vector is input into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area; The current flue gas concentration data and historical flue gas concentration data are input into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area. The historical flue gas concentration data is the concentration data generated in the target detection area on the same day or generated by the same device under the same power condition. The first predicted concentration data and the second predicted concentration data are weighted according to preset weight values ​​to obtain the predicted flue gas generation data of the target detection area within a certain period of time in the future.

2. The method for predicting the generation of flue gas pollutants as described in claim 1, characterized in that, The step of acquiring the current flue gas concentration data within the target detection area in the confined space of a thermal power plant includes: Acquire the first flue gas concentration data at the first preset time point within the target detection area; Obtain the second flue gas concentration data at a second preset time point within the target detection area; The first flue gas concentration data and the second flue gas concentration data are normalized according to the normalization algorithm to obtain the current flue gas concentration data corresponding to the target detection area.

3. The method for predicting the generation of flue gas pollutants as described in claim 1, characterized in that, Before the step of acquiring the current flue gas concentration data within the target detection area in the confined space of a thermal power plant, the method further includes: Obtain historical flue gas concentration data corresponding to the target detection area; Obtain the historical temperature and historical air pressure corresponding to the historical flue gas concentration data; The historical flue gas concentration data, historical temperature, and historical air pressure are input into the initial neural network model for iterative training to obtain a recurrent neural network model.

4. The method for predicting the generation of flue gas pollutants as described in claim 3, characterized in that, The step of inputting the historical flue gas concentration data, the historical temperature, and the historical air pressure into the initial neural network model for iterative training to obtain the recurrent neural network model includes: The historical flue gas concentration data, the historical temperature, and the historical air pressure are input into the initial neural network model for iterative training to obtain the trained model. The detection data is input into the training model, and it is determined whether the output result reaches the prediction range. When the output result does not reach the prediction range, historical flue gas concentration data is input into the training model for reverse training until the output result reaches the prediction range, and the training model corresponding to the output result reaching the prediction range is used as a recurrent neural network model.

5. A device for predicting the generation of flue gas pollutants, characterized in that, The device includes: The data acquisition module is used to acquire the current flue gas concentration data within the target detection area in the enclosed space of a thermal power plant; The data processing module is used to acquire the temperature data and air pressure data of the space at each detection time point corresponding to the current flue gas concentration data; and to construct a three-dimensional feature vector corresponding to the current flue gas concentration data based on the temperature data, the air pressure data and the current flue gas concentration data. The concentration prediction module is used to input the three-dimensional feature vector into a preset recurrent neural network model to obtain the first predicted concentration data corresponding to the target detection area; The concentration prediction module is also used to input the current flue gas concentration data and historical flue gas concentration data into the preset recurrent neural network model to obtain the second predicted concentration data corresponding to the target detection area. The historical flue gas concentration data is the concentration data generated in the target detection area on the same day or generated by the same device under the same power condition. The generation quantity display module is used to perform weight processing on the first predicted concentration data and the second predicted concentration data according to a preset weight value to obtain the predicted flue gas generation quantity data of the target detection area within a certain period of time in the future.

6. The flue gas pollutant generation prediction device as described in claim 5, characterized in that, The data acquisition module is also used to acquire the first flue gas concentration data at a first preset time point within the target detection area; The data acquisition module is also used to acquire the second flue gas concentration data at a second preset time point within the target detection area; The data acquisition module is further configured to perform normalization processing on the first flue gas concentration data and the second flue gas concentration data according to a normalization algorithm to obtain the current flue gas concentration data corresponding to the target detection area.

7. The flue gas pollutant generation prediction device as described in claim 5, characterized in that, The device further includes: a model building module; The model building module is used to obtain historical flue gas concentration data corresponding to the target detection area; The model building module is also used to obtain the historical temperature and historical pressure corresponding to the historical flue gas concentration data; The model building module is also used to input the historical flue gas concentration data, the historical temperature and the historical air pressure into the initial neural network model for iterative training to obtain a recurrent neural network model.

8. The flue gas pollutant generation prediction device as described in claim 7, characterized in that, The model building module is also used to input the historical flue gas concentration data, the historical temperature and the historical air pressure into the initial neural network model for iterative training to obtain a trained model; The model building module is also used to input the detection data into the training model and determine whether the output result reaches the prediction range; The model building module is further configured to input historical flue gas concentration data into the training model for reverse training when the output result does not reach the prediction range, until the output result reaches the prediction range, and use the training model corresponding to the output result reaching the prediction range as a recurrent neural network model.

Images