Carbon dioxide concentration prediction method and system

Abstract

The invention provides a carbon dioxide concentration prediction method and system, and relates to the technical field of gas concentration prediction.The method comprises the steps that feature data of gas to be predicted are obtained; acquiring a stable data interval of gas concentration in the to-be-predicted gas characteristic data, calculating gas concentration increment data corresponding to the stable data interval according to a preset baseline value, and acquiring corresponding gas concentration increment characteristic data; and inputting the to-be-predicted gas characteristic data and the gas concentration increment characteristic data into the carbon dioxide prediction model to obtain a carbon dioxide concentration prediction result. According to the invention, the accuracy of indoor carbon dioxide concentration prediction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas concentration prediction, and particularly to a method and system for predicting carbon dioxide concentration. Background Art

[0002] As a key indicator for measuring indoor air quality, the fluctuation of carbon dioxide concentration will affect the physiological comfort of the human body, and high-concentration carbon dioxide will affect human health. Long-term exposure to a high-concentration carbon dioxide environment may also cause a series of health problems, such as headache, drowsiness, inattention, etc.

[0003] In the data processing link of traditional indoor carbon dioxide concentration prediction methods, a single fixed method is often used for data processing, lacking special processing for the characteristics of indoor carbon dioxide concentration data, resulting in weak ability to identify and process outliers, simple missing value filling methods, and insufficient in-depth mining of the characteristics of low-fluctuation data. In addition, most traditional indoor carbon dioxide concentration prediction models are built with temperature and humidity as a single feature. Although this method is simple, due to the complex interaction of various factors affecting indoor carbon dioxide concentration, only considering temperature and humidity is far from being able to capture the full picture of its dynamic changes. These factors jointly restrict the accuracy of the existing technology for predicting indoor carbon dioxide concentration. Summary of the Invention

[0004] In order to solve at least one deficiency of the existing technology, the purpose of the present invention is to provide a method and system for predicting carbon dioxide concentration to improve the accuracy of predicting indoor carbon dioxide concentration.

[0005] To achieve the above purpose, according to some embodiments, in the first aspect of the present invention, a method for predicting carbon dioxide concentration is provided, including:

[0006] Obtain the characteristic data of the gas to be predicted;

[0007] Obtain the stable data interval of the gas concentration in the characteristic data of the gas to be predicted, calculate the gas concentration increment data corresponding to the stable data interval according to the preset baseline value, and obtain the corresponding gas concentration increment characteristic data;

[0008] Input the characteristic data of the gas to be predicted and the gas concentration increment characteristic data into the carbon dioxide prediction model to obtain the carbon dioxide concentration prediction result;

[0009] Wherein, the carbon dioxide prediction model includes a first prediction model and a second prediction model. The first prediction model is used to obtain a first prediction result according to the characteristic data of the gas to be predicted, and the second prediction model is used to obtain a second prediction result according to the gas concentration increment characteristic data. The carbon dioxide concentration prediction result is a fusion result obtained by fusing the first prediction result and the second prediction result.

[0010] Preferably, the first prediction model is an improved SEQ2SEQ model. In the improved SEQ2SEQ model, both the encoder and the decoder are bidirectional gated recurrent units; the second prediction model is a gated recurrent unit; the predicted carbon dioxide concentration result is obtained by substituting the second prediction result for the first prediction result within the corresponding data interval.

[0011] Preferably, the improved SEQ2SEQ model further includes a self-attention module and a temporal attention module; the input data of the improved SEQ2SEQ model is processed by the self-attention module and then input into the encoder for processing, and the output features of the encoder are processed by the temporal attention module and then input into the decoder for processing.

[0012] Preferably, the gas feature data to be predicted is obtained by fusing the gas concentration to be predicted, temperature, humidity, number of people, ventilation effect, door state, and outdoor air quality.

[0013] Preferably, a stable data interval of the gas concentration in the gas feature data to be predicted is obtained, and according to a preset baseline value, the gas concentration increment data corresponding to the stable data interval is calculated, including determining the stable data interval of the gas concentration according to the preset baseline value, and subtracting the preset baseline value from the gas concentration within the stable data interval to obtain the gas concentration increment data.

[0014] Preferably, it further includes using the Isolation Forest algorithm to detect outliers in the gas concentration data of the gas feature data to be predicted; using the Lagrange interpolation method to supplement missing values and / or outliers.

[0015] In a second aspect of the present invention, a carbon dioxide concentration prediction system is provided, including:

[0016] A data acquisition module configured to acquire gas feature data to be predicted;

[0017] An increment data calculation module configured to obtain a stable data interval of the gas concentration in the gas feature data to be predicted, calculate the gas concentration increment data corresponding to the stable data interval according to a preset baseline value, and obtain the corresponding gas concentration increment feature data;

[0018] A prediction module, configured to input gas feature data to be predicted and gas concentration increment feature data into a carbon dioxide prediction model to obtain a carbon dioxide concentration prediction result; wherein, the carbon dioxide prediction model includes a first prediction model and a second prediction model, the first prediction model is used to obtain a first prediction result according to the gas feature data to be predicted, the second prediction model is used to obtain a second prediction result according to the gas concentration increment feature data, and the carbon dioxide concentration prediction result is a fusion result obtained by fusing the first prediction result and the second prediction result.

[0019] In a third aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to complete the steps of the above-mentioned carbon dioxide concentration prediction method.

[0020] In a fourth aspect of the present invention, a computer-readable storage medium is provided for storing computer instructions, and when the computer instructions are executed by a processor, the steps of the above-mentioned carbon dioxide concentration prediction method are completed.

[0021] In a fifth aspect of the present invention, a computer program product is provided, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned carbon dioxide concentration prediction method are implemented.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] The present invention provides a carbon dioxide concentration prediction method and system. Aiming at the problem that the prediction accuracy of the model for data in the data interval close to the outdoor carbon dioxide concentration is low in the prediction of the carbon dioxide concentration in the indoor environment, the method of calculating incremental data is adopted to enhance the data, thereby effectively improving the prediction accuracy. At the same time, for this part of the data, a gated recurrent unit (GRU) model is used for independent prediction, replacing the prediction result of the SEQ2SEQ model in this data interval, and further improving the overall prediction accuracy of the model.

[0024] Advantages of additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention, and the schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0026] Figure 1 It is a schematic diagram of the improved SEQ2SEQ model in the first embodiment of the present invention;

[0027] Figure 2 This is a comparison chart of the predicted values and the true values in carbon dioxide concentration prediction between the method in Embodiment 1 of the present invention and the existing methods. Among them, (a) is the comparison chart of the predicted values and the true values using the TCN-LSTM model, (b) is the comparison chart of the predicted values and the true values using the CNN-LSTM model, (c) is the comparison chart of the predicted values and the true values using the traditional SEQ2SEQ model, and (d) is the comparison chart of the predicted values and the true values using the method of Embodiment 1 of the present invention. Detailed implementation manners

[0028] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0029] Embodiment 1

[0030] Embodiment 1 of the present invention provides a carbon dioxide concentration prediction method, including:

[0031] Obtain the characteristic data of the gas to be predicted;

[0032] Obtain the stable data interval of the gas concentration in the characteristic data of the gas to be predicted, calculate the gas concentration increment data corresponding to the stable data interval according to the preset baseline value, and obtain the corresponding gas concentration increment characteristic data;

[0033] Input the characteristic data of the gas to be predicted and the gas concentration increment characteristic data into the carbon dioxide prediction model to obtain the carbon dioxide concentration prediction result; wherein, the carbon dioxide prediction model includes a first prediction model and a second prediction model. The first prediction model is used to obtain a first prediction result according to the characteristic data of the gas to be predicted, and the second prediction model is used to obtain a second prediction result according to the gas concentration increment characteristic data. The carbon dioxide concentration prediction result is a fusion result obtained by fusing the first prediction result and the second prediction result.

[0034] Traditional indoor CO 2 concentration prediction methods mainly construct models based on the simple linear relationship between temperature, humidity and CO 2 concentration. Although temperature and humidity are correlated with CO 2 concentration to a certain extent, this correlation is not a decisive factor. Indoor CO 2 concentration is a complex system driven by multiple sources. The increase in the number of people will directly lead to an increase in the CO 2 emissions generated by breathing; good ventilation can dilute the high-concentration CO 2 indoors in time and discharge it outdoors; the opening or closing state of the door affects the air exchange rate between indoors and outdoors, and thus changes the CO 2 concentration; the quality of outdoor air determines the background quality of the air entering the room and indirectly affects the indoor CO 2 concentration.

[0035] In terms of data processing, traditional methods usually use simple statistical threshold methods to detect outliers. This method is prone to missing outliers in complex data distributions and is difficult to effectively identify sudden increases in CO concentration caused by non-personnel factors (such as local CO leakage caused by equipment failures). In dealing with missing values, simple mean filling or neighboring value filling is mostly used, and the time series characteristics of the data and the potential relationships between multi-variables are not fully utilized. For the data interval of 400ppm ± 50ppm with relatively small fluctuations and close to the outdoor average concentration, traditional methods lack effective feature enhancement means, resulting in low prediction accuracy when the model processes this part of the data. 2 In the case of a sudden increase in concentration (such as local CO leakage caused by equipment failure) 2 it is difficult to effectively identify. In dealing with missing values, simple mean filling or neighboring value filling is mostly used, and the time series characteristics of the data and the potential relationships between multi-variables are not fully utilized. For the data interval of 400ppm ± 50ppm with relatively small fluctuations and close to the outdoor average concentration, traditional methods lack effective feature enhancement means, resulting in low prediction accuracy when the model processes this part of the data.

[0036] To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described. 2 To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described. 2 To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described. 2 To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described. 2 To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described. 2 To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described. 2 To comprehensively capture the factors affecting indoor CO concentration, in this embodiment, on the basis of traditional temperature and humidity characteristics, multi-dimensional characteristics such as the number of people, ventilation effect, door state, and outdoor air quality are introduced. The increase in the number of people means an increase in CO emission sources and is a direct factor affecting indoor CO concentration; the ventilation effect plays a key regulatory role in CO concentration by controlling the air exchange volume and frequency between indoors and outdoors; the door state determines the connectivity between the indoor space and the outside world and affects the air circulation path and rate; the outdoor air quality provides the background conditions for the change of indoor CO concentration. By integrating these characteristics into the data set, the complex law of indoor CO concentration change can be more comprehensively described.

[0037] For outliers in CO concentration data, this embodiment uses the Isolation Forest algorithm for detection. Based on the principle of random forest, this algorithm can quickly identify data points deviating from the normal distribution in the high-dimensional data space and has a good detection effect on sudden increases in CO concentration caused by non-personnel factors. 2 For outliers in CO concentration data, this embodiment uses the Isolation Forest algorithm for detection. Based on the principle of random forest, this algorithm can quickly identify data points deviating from the normal distribution in the high-dimensional data space and has a good detection effect on sudden increases in CO concentration caused by non-personnel factors. 2 For outliers in CO concentration data, this embodiment uses the Isolation Forest algorithm for detection. Based on the principle of random forest, this algorithm can quickly identify data points deviating from the normal distribution in the high-dimensional data space and has a good detection effect on sudden increases in CO concentration caused by non-personnel factors.

[0038] Isolation Forest algorithm parameter settings:

[0039] Contamination rate: 0.05 (determined according to the data set and actual requirements);

[0040] Number of trees: 100;

[0041] Maximum number of samples: 'auto' (i.e., all samples);

[0042] Random seed: 42.

[0043] In terms of missing value processing, consistent with the detected outliers, the Lagrange interpolation method is adopted. This method estimates the missing values accurately by constructing a polynomial function based on the known data points. Compared with simple mean or neighboring value filling, the Lagrange interpolation method can better retain the original trend and characteristics of the data, ensure the integrity and continuity of the data, and provide a reliable data basis for subsequent data analysis and modeling.

[0044] For the data interval with relatively stable fluctuations of 400 ppm ± 50 ppm, since its characteristic signal is weak, in order to enhance the model's learning ability for this part of the data, first calculate the concentration increment of each data point relative to the 400 ppm baseline, and then normalize these increments. In this way, the originally indistinguishable low-fluctuation data is transformed into data with obvious characteristic differences, improving the distinguishability of the data in model training.

[0045] On this basis, for this part of the data, a gated recurrent unit (GRU) model is used for independent prediction. Because the addition of feature discrimination amplifies the local change characteristics, and the GRU model can effectively capture short-term dependencies when processing time series data and has better adaptability to low-fluctuation, high-frequency change data. Therefore, the prediction results of the GRU model are used to replace the prediction results of the traditional SEQ2SEQ model in this data interval, further improving the overall prediction accuracy of the model.

[0046] In this embodiment, the characteristic data of the gas to be predicted that integrates multi-dimensional features is used to predict the carbon dioxide concentration. First, use a CRDS (Cavity Ring-Down Spectroscopy) device, a thermometer and hygrometer, and other sensors to obtain indoor CO 2Concentration, temperature and humidity, number of people, ventilation effect, door status, and outdoor air quality and other multi-dimensional features. Secondly, the time information, gas concentration, and other multiple feature values ​​are normalized and merged to obtain a data set as the gas feature data to be predicted for subsequent prediction tasks. The outliers in the data set are detected using the isolation forest algorithm, and the missing values ​​and the outliers detected after anomaly detection are interpolated using the Lagrange interpolation method. Finally, for the data interval that is usually in a stable fluctuation in the detection environment, a baseline value is preset. In this embodiment, the average value of the carbon dioxide concentration in the outdoor environment is 400ppm as the baseline value. The baseline value can also be set according to the carbon dioxide concentration value under normal circumstances indoors and outdoors in the actual prediction scenario, for example, the average value of the carbon dioxide concentration value before the set time (for example, 7 days) in the environment to be tested is used as the baseline value. The stable data interval of the gas concentration is determined according to the preset baseline value. In this embodiment, the gas concentration falling into 400±50ppm is used as the data in the stable data interval, and the data is enhanced, and the difference between the gas concentration and the baseline value is calculated to obtain the gas concentration increment data, which is further normalized and fused with other multi-dimensional features to obtain the gas concentration increment feature data.

[0047] In this embodiment, a bidirectional gated recurrent unit (BiGRU) is introduced into the encoder and decoder of the traditional SEQ2SEQ model, and Bi GRU is used as the encoder and decoder of the model to enhance the model's ability to capture bidirectional information of time series data. 2 In concentration prediction, not only the concentration change and environmental status at the previous moment have an important impact on the current moment, but also the possible change trend in the future cannot be ignored. Bi GRU can more comprehensively understand the complex patterns in time series data by learning the forward and backward information of the sequence at the same time, thereby improving the model's prediction of indoor CO 2 Predictive power of concentration dynamics.

[0048] In order to further improve the efficiency and accuracy of the model in processing sequence data, the self-attention mechanism is introduced. The self-attention mechanism can automatically assign weights to input information at different positions when calculating the output at the current position, thereby better capturing the long-distance dependencies between data. However, the self-attention mechanism has certain limitations in processing the dynamic association of time series data and cannot fully consider the changes in information in the time dimension.

[0049] Therefore, this embodiment further introduces a temporal attention mechanism, which is combined with the self-attention mechanism. The temporal attention mechanism can, according to the characteristics of the time series, weight the information at different time steps, highlighting the information at key time points, so as to better capture the dynamic changes and long-term dependencies in the sequence data. The combination of this dual attention mechanism enables the model to more precisely focus on the key factors affecting the indoor CO 2 concentration change, significantly improving the prediction performance of the model.

[0050] SEQ2SEQ model parameter settings:

[0051] Encoder and decoder: Bi GRU;

[0052] Bi GRU: The number of hidden units is 64, and the activation function is the Tanh function;

[0053] Model learning rate: 0.002;

[0054] Number of iterations: 20;

[0055] Batch size: 32.

[0056] GRU model settings for predicting the concentration in the data interval with relatively stable fluctuations of 400 ppm ± 50 ppm:

[0057] GRU: The number of hidden units is 64, and the activation function is the Tanh function;

[0058] Model learning rate: 0.0001;

[0059] Number of iterations: 20;

[0060] Batch size: 32.

[0061] In the actual prediction process, after setting the parameters for the improved SEQ2SEQ model and the GRU model respectively, the corresponding data is input for concentration prediction, and the prediction result of the GRU model in the feature interval of 400 ppm ± 50 ppm is used to replace the prediction result of the SEQ2SEQ model. The obtained fusion result is the final prediction result.

[0062] Finally, calculate the evaluation index of the prediction result and draw a comparison chart of the predicted value and the true value of the CO 2 concentration.

[0063] This embodiment uses a large amount of indoor data from urban living environments, covering multi-dimensional information such as temperature, humidity, number of people, ventilation effect, door status, outdoor air quality, and CO 2 concentration at different time periods. The collected dataset is divided into a training set and a test set according to a ratio of approximately 75% and 25%.

[0064] In terms of model evaluation metrics, the root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and coefficient of determination (R 2 ) etc. are selected as the main evaluation metrics. RMSE can reflect the deviation degree between the predicted value and the true value, while MAE measures the average absolute value of the prediction error, and R 2 is used to evaluate the goodness of fit of the model to the data. At the same time, the traditional SEQ2SEQ model, TCN-LSTM model, and CNN-LSTM model are selected as comparison models to verify the superiority of the improved model.

[0065] The calculation formulas for the root mean square error (RMSE), mean absolute error (MAE), coefficient of determination (R 2 ) and mean absolute percentage error (MAPE) are as follows:

[0066]

[0067] where, y i and represent the predicted value and the true value respectively.

[0068] The experimental results of the four models are shown in Table 1, where SEQ2SEQ (Bi GRU) represents the improved SEQ2SEQ model in this embodiment.

[0069] Table 1. Comparison of experimental results of four models;

[0070]

[0071] The comparison results of the predicted values and the true values of the four models are as Figure 2 shown. In the figure, the abscissa represents the data point, the ordinate represents the carbon dioxide concentration (unit: ppm), the blue curve represents the actual value of the carbon dioxide concentration, and the orange curve represents the predicted value of the carbon dioxide concentration. The higher the coincidence degree of the two lines, the more accurate the predicted value. From Figure 2 it can be seen that the optimized model has improved in prediction accuracy, and the prediction effect for the peak value is more accurate.

[0072] The experimental results show that the improved SEQ2SEQ model is significantly better than the traditional SEQ2SEQ model and other comparison models in various evaluation metrics, and the prediction result performance of the improved model is better than that of the traditional SEQ2SEQ model (especially in predicting the CO 2 peak value). Through multi-feature value fusion, the model can capture the factors affecting indoor CO 2Complex factors of concentration; the refined data processing flow effectively improves data quality, providing more reliable data support for model training; the introduction of BiGRU and dual attention mechanisms significantly enhances the model's ability to process time series data and feature extraction ability, enabling it to more accurately predict indoor CO 2 concentration change trend.

[0073] In this embodiment, through systematic improvement of the data fusion, processing flow, and model structure of the indoor CO 2 concentration prediction method, the limitations of traditional methods are successfully overcome. Under the synergistic effect of the improved SEQ2SEQ model with multi-feature fusion and dual attention mechanisms, high-precision prediction of indoor CO 2 concentration is achieved, providing a strong technical guarantee for the real-time monitoring and intelligent regulation of indoor air quality.

[0074] Embodiment 2

[0075] This embodiment provides a carbon dioxide concentration prediction system, including:

[0076] A data acquisition module configured to acquire gas characteristic data to be predicted;

[0077] An incremental data calculation module configured to obtain a stable data interval of the gas concentration in the gas characteristic data to be predicted, calculate the gas concentration incremental data corresponding to the stable data interval according to a preset baseline value, and obtain the corresponding gas concentration incremental feature data;

[0078] A prediction module configured to input the gas characteristic data to be predicted and the gas concentration incremental feature data into a carbon dioxide prediction model to obtain a carbon dioxide concentration prediction result; wherein, the carbon dioxide prediction model includes a first prediction model and a second prediction model, the first prediction model is used to obtain a first prediction result according to the gas characteristic data to be predicted, the second prediction model is used to obtain a second prediction result according to the gas concentration incremental feature data, and the carbon dioxide concentration prediction result is a fusion result obtained by fusing the first prediction result and the second prediction result.

[0079] It should be noted here that each module in this embodiment corresponds to the steps of the method in Embodiment 1 one by one, and the specific implementation process is the same, so it will not be repeated here.

[0080] Embodiment 3

[0081] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to complete the steps of the method in Embodiment 1.

[0082] Embodiment 4

[0083] This embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the method in Embodiment 1 are completed.

[0084] Embodiment 5

[0085] This embodiment provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the method in Embodiment 1 are implemented.

[0086] The foregoing are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting carbon dioxide concentration, characterized in that: include: Acquire characteristic data of the gas to be predicted; Obtain a stable data interval of gas concentration in the gas characteristic data to be predicted, calculate the gas concentration increment data corresponding to the stable data interval according to a preset baseline value, and obtain the corresponding gas concentration increment characteristic data; Inputting the characteristic data of the gas to be predicted and the characteristic data of the gas concentration increment into the carbon dioxide prediction model to obtain the carbon dioxide concentration prediction result; Among them, the carbon dioxide prediction model includes a first prediction model and a second prediction model. The first prediction model is used to obtain a first prediction result based on the characteristic data of the gas to be predicted, and the second prediction model is used to obtain a second prediction result based on the gas concentration increment characteristic data. The carbon dioxide concentration prediction result is a fusion result obtained by fusing the first prediction result and the second prediction result.

2. A method for predicting carbon dioxide concentration according to claim 1, characterized in that: The first prediction model is an improved SEQ2SEQ model, in which the encoder and the decoder are both bidirectional gated recurrent units; the second prediction model is a gated recurrent unit; the carbon dioxide concentration prediction result is obtained by replacing the first prediction result in the corresponding data interval with the second prediction result.

3. A method for predicting carbon dioxide concentration according to claim 2, characterized in that: The improved SEQ2SEQ model also includes a self-attention module and a temporal attention module; the input data of the improved SEQ2SEQ model is processed by the self-attention module and then input into the encoder for processing, and the output features of the encoder are processed by the temporal attention module and then input into the decoder for processing.

4. A method for predicting carbon dioxide concentration according to claim 1, characterized in that: The characteristic data of the gas to be predicted are obtained by integrating the concentration of the gas to be predicted, temperature, humidity, number of people, ventilation effect, door status and outdoor air quality.

5. A method for predicting carbon dioxide concentration according to claim 1, characterized in that: Obtain a stable data interval of gas concentration in the gas characteristic data to be predicted, and calculate the incremental gas concentration data corresponding to the stable data interval according to a preset baseline value, including determining the stable data interval of gas concentration according to the preset baseline value, and subtracting the gas concentration in the stable data interval from the preset baseline value to obtain the incremental gas concentration data.

6. A method for predicting carbon dioxide concentration according to claim 1, characterized in that: It also includes: using an isolation forest algorithm to detect outliers in the gas concentration data in the acquired gas characteristic data to be predicted; and using a Lagrange interpolation method to supplement missing values ​​and / or outliers.

7. A carbon dioxide concentration prediction system, characterized in that: include: A data acquisition module is configured to acquire characteristic data of the gas to be predicted; The incremental data calculation module is configured to obtain a stable data interval of gas concentration in the gas characteristic data to be predicted, calculate the incremental data of gas concentration corresponding to the stable data interval according to a preset baseline value, and obtain the corresponding incremental characteristic data of gas concentration; The prediction module is configured to input the characteristic data of the gas to be predicted and the characteristic data of the gas concentration increment into the carbon dioxide prediction model to obtain a carbon dioxide concentration prediction result; wherein the carbon dioxide prediction model includes a first prediction model and a second prediction model, the first prediction model is used to obtain a first prediction result based on the characteristic data of the gas to be predicted, and the second prediction model is used to obtain a second prediction result based on the characteristic data of the gas concentration increment, and the carbon dioxide concentration prediction result is a fusion result obtained by fusing the first prediction result and the second prediction result.

8. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to complete the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps of the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images