Air sampling strategy optimization method and system based on artificial intelligence

Abstract

The invention discloses an air sampling strategy optimization method and system based on artificial intelligence, and relates to the technical field of environment monitoring, an air quality prediction model is constructed based on a neural network, historical environment monitoring data is taken as training data, the air quality prediction model is trained and optimized, and the air sampling strategy optimization method and system based on artificial intelligence are obtained. Predicting and outputting an air quality change trend in a period of time in the future; formulating an optimal sampling strategy based on an output result of the air quality prediction model in combination with air quality data, meteorological data and pollution source data collected in real time; and automatically controlling the operation of the air sampling equipment according to the optimal sampling strategy. According to the method, the multi-source environment data is collected in real time, the high-precision air quality prediction model is constructed, the optimal sampling strategy is made in combination with the multi-target optimization method, operation of the air sampling equipment is automatically controlled, the sampling strategy can be dynamically adjusted according to real-time environment changes, and the pertinence and effectiveness of air sampling are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and more particularly to an optimization method and system for air sampling strategies based on artificial intelligence. Background Art

[0002] Air sampling is a key link in obtaining air quality data and evaluating the atmospheric environmental quality. At present, traditional air sampling methods are widely used in various environmental monitoring scenarios. However, with the increasing complexity of environmental problems and the continuous improvement of people's requirements for the accuracy of air quality monitoring, the limitations of traditional air sampling methods have gradually emerged.

[0003] Most traditional air sampling work relies on fixed sampling frequencies and pre-set point layouts to carry out. In the actual operation process, staff collect air samples at pre-planned fixed sampling points at established time intervals, such as every few hours or at fixed time periods every day. This sampling method can obtain a certain amount of air samples under relatively stable and slowly changing environmental conditions, and conduct a preliminary evaluation and analysis of air quality.

[0004] However, in the face of complex and changing air quality conditions, this traditional sampling mode has obvious deficiencies. Today's atmospheric environment is affected by a variety of factors, including but not limited to fluctuations in industrial production activities, changes in traffic flow, drastic changes in meteorological conditions, and sudden environmental pollution events. In the case of sudden pollution, such as the leakage of toxic gases from a chemical plant or a large-scale haze weather in a city, the concentration and distribution of pollutants will change rapidly in a short period of time. At this time, fixed sampling frequencies and point layouts cannot capture these rapidly changing information in time, resulting in the collected data being seriously lagging behind the actual pollution situation. And because the sampling point positions are fixed, it may not cover the key areas of pollution diffusion, making the collected data unable to truly reflect the actual situation of pollution and lacking representativeness.

[0005] Similarly, in other scenarios where air quality changes rapidly, such as the rapid diffusion or aggregation of pollutants caused by severe convective weather, traditional sampling strategies cannot respond in time. Since the sampling frequency and points cannot be adjusted according to real-time environmental changes, the collected data cannot accurately reflect the dynamic change process of air quality, which brings great difficulties to subsequent environmental quality assessment, pollution source tracing, and the formulation of treatment decisions.

[0006] Therefore, how to provide an optimization method and system for air sampling strategies based on artificial intelligence, which can dynamically adjust sampling strategies according to real-time environmental changes to obtain more representative air sample data and improve the accuracy and timeliness of air quality monitoring, is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides an optimization method and system for air sampling strategies based on artificial intelligence, which fully integrates multi-source data, comprehensively considers the impacts of meteorology, pollution sources and other factors on air quality, and improves the accuracy and reliability of air quality prediction; by introducing artificial intelligence to optimize the sampling strategy, according to real-time air quality data, meteorological data and pollution source data, dynamically adjust the sampling frequency, sampling points and sampling duration, realize the intelligent and adaptive optimization of the sampling strategy, and when it is predicted that a pollution peak will occur in a certain area, automatically increase the sampling frequency and duration of that area to ensure obtaining more comprehensive and accurate data.

[0008] To achieve the above object, the present invention adopts the following technical solutions: An optimization method for air sampling strategies based on artificial intelligence, comprising:

[0009] Real-time collect air quality data, meteorological data, and pollution source data, and preprocess the real-time collected air quality data, meteorological data, and pollution source data;

[0010] Build an air quality prediction model based on a neural network, use historical environmental monitoring data as training data, train and optimize the air quality prediction model, and predict and output the air quality change trend in a future period of time;

[0011] Based on the output result of the air quality prediction model, and at the same time combining the real-time collected air quality data, meteorological data and pollution source data, formulate an optimal sampling strategy;

[0012] According to the optimal sampling strategy, automatically control the operation of the air sampling equipment.

[0013] Preferably, according to historical environmental monitoring data, obtain environmental monitoring matrices at different sampling times based on the same time interval, and generate a training data set and a test data set;

[0014] Fuse the ability of a convolutional neural network to extract spatial features and the ability of a recurrent neural network to process time series features to build an air quality prediction model;

[0015] Train the air quality prediction model through the training data set, and test the air quality prediction model through the test data set.

[0016] Preferably, combine the output result of the air quality prediction model with the real-time collected air quality data, meteorological data, and pollution source data to build a sampling strategy optimization model, and the sampling strategy optimization model includes multiple sampling strategies; solve the sampling strategy optimization model according to the multiple sampling strategies to obtain the optimal sampling strategy.

[0017] Preferably, the output result of the air quality prediction model is fused with the real-time collected air quality data, meteorological data, and pollution source data to obtain a fused data set, wherein the fusion process includes aligning and synchronizing time series data;

[0018] Based on the fused data set, determine the structure of the sampling strategy optimization model;

[0019] Incorporate multiple sampling strategies into the sampling strategy optimization model to solve the sampling strategy optimization model incorporating multiple sampling strategies.

[0020] Preferably, a convolutional neural network is constructed. One matrix input to the model corresponds to a corresponding convolutional neural network. If q historical environmental monitoring matrices are input, then q corresponding convolutional neural networks are used. The convolutional neural network consists of several convolutional modules and fully connected layers. When used, the network parameters are randomly initialized. Each convolutional module consists of one convolutional layer and zero or one pooling layer. Each independent convolutional neural network outputs a spatial feature vector through a flattening layer;

[0021] Construct a recurrent neural network, which consists of several layers of neurons, and the number of neurons in each layer is k;

[0022] After the recurrent neural network is constructed, a flattening layer is added to this network;

[0023] The output of the flattening layer neurons is connected through a fully connected layer, and finally the value of the future environmental monitoring matrix at the next time step is output through a regression prediction layer.

[0024] Preferably, the air quality prediction model is trained with a training data set, including:

[0025] Obtain the data input to the model from the training data set with a sliding window of time width q, which can be represented by a matrix as follows:

[0026] X input =[X t-q ,X t-q+1 ,…,X t

[0027] Where the input to the model is the environmental monitoring matrix of the q historical time slots before time t, and the goal is to predict the value of the environmental monitoring matrix at the (t + 1)-th time slot in the future;

[0028] Input the continuous training historical environmental monitoring data obtained using the sliding window into the air quality prediction model, and perform forward propagation on the air quality prediction model;

[0029] ​Among them, the outputs of q convolutional neural networks are input into the recurrent neural network. Combining the time dependence of air quality, the predicted air vector eigenvalues are output. Then the vector representation output by the q-th MGU unit is h t , and the output of the k-th MGU unit is the output h of the hidden gating unit at the corresponding sampling moment t , h t The vector is flattened by a flattening layer and used as the input of the fully connected layer;

[0030] By calculating the loss between the predicted value and the true value of the model, backpropagation is performed on the air quality prediction model to adjust the corresponding parameters in the model;

[0031] The Adam optimization method is used to optimize the loss L of the overall network, and its formula is as follows:

[0032]

[0033] Among them, p ij represents the predicted value corresponding to the model target matrix at the corresponding sampling time slot, and x ij represents the true flow value at the corresponding sampling time slot;

[0034] Repeat the above steps. When the number of iterations reaches the preset number, stop the model training;

[0035] Continue to use Adam to optimize the loss function, adjust the parameters, and obtain the final air quality prediction model.

[0036] Preferably, the historical environmental monitoring data includes air quality data, meteorological data, and pollution source data.

[0037] Preferably, an air sampling strategy optimization system based on artificial intelligence includes:

[0038] A data acquisition and preprocessing module, which is used to collect air quality data, meteorological data, and pollution source data in real time, and preprocess the collected data;

[0039] An air quality prediction module, which is used to construct an air quality prediction model based on a neural network, use historical air quality data, meteorological data, and pollution source data as training data, train and optimize the air quality prediction model, and predict and output the air quality change trend in the future for a period of time;

[0040] A sampling strategy formulation module, which is used to formulate an optimal sampling strategy based on the output results of the air quality prediction model, and at the same time combine the air quality data, meteorological data, and pollution source data collected in real time;

[0041] An automatic control module, which is used to automatically control the operation of the air sampling device according to the optimal sampling strategy.

[0042] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses an optimization method and system for air sampling strategies based on artificial intelligence, including: collecting air quality data, meteorological data, and pollution source data in real time, and preprocessing the collected air quality data, meteorological data, and pollution source data in real time; constructing an air quality prediction model based on a neural network, using historical environmental monitoring data as training data to train and optimize the air quality prediction model, and predicting and outputting the air quality change trend in a future period of time; formulating an optimal sampling strategy based on the output result of the air quality prediction model, and simultaneously combining the collected air quality data, meteorological data, and pollution source data in real time; and automatically controlling the operation of the air sampling device according to the optimal sampling strategy. By collecting multi-source environmental data in real time, using a neural network integrating CNN and RNN to construct a high-precision air quality prediction model, formulating an optimal sampling strategy by combining a multi-objective optimization method, and automatically controlling the operation of the air sampling device, the present invention can dynamically adjust the sampling strategy according to real-time environmental changes, improve the pertinence and effectiveness of air sampling. The obtained air sample data is more representative, can accurately reflect the dynamic change process of air quality, provides reliable data support for environmental quality assessment, pollution source tracing, and treatment decision-making, helps improve the accuracy and timeliness of air quality monitoring, and promotes the development and application of environmental monitoring technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0044] Figure 1 It is a schematic flow chart of an optimization method for air sampling strategies based on artificial intelligence provided by an embodiment of the present invention.

[0045] Figure 2 It is a schematic structural diagram of an optimization system for air sampling strategies based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0047] An embodiment of the present invention discloses an optimization method for air sampling strategy based on artificial intelligence, as Figure 1 shown, including:

[0048] Collect real-time air quality data, meteorological data, and pollution source data, and preprocess the real-time collected air quality data, meteorological data, and pollution source data;

[0049] Build an air quality prediction model based on a neural network, use historical environmental monitoring data as training data, train and optimize the air quality prediction model, and predict and output the air quality change trend in a future period of time;

[0050] Based on the output result of the air quality prediction model, and combining the real-time collected air quality data, meteorological data, and pollution source data at the same time, formulate an optimal sampling strategy; ensure that the sampling strategy always adapts to the complex and changeable air quality conditions;

[0051] According to the optimal sampling strategy, automatically control the operation of the air sampling equipment.

[0052] Automatically adjust the sampling frequency, sampling points, and sampling duration; ensure that the air sampling equipment can sample according to the predetermined strategy and obtain high-quality data.

[0053] Specifically, the air quality data includes but is not limited to the concentrations of pollutants such as PM2.5, PM10, sulfur dioxide, nitrogen oxides, ozone, etc.; the meteorological data covers meteorological elements such as wind direction, wind speed, temperature, humidity, air pressure, etc.; the pollution source data includes information such as the emission intensity and emission time of industrial pollution sources, and the traffic flow and vehicle type distribution of traffic pollution sources.

[0054] Specifically, the preprocessing of the collected air quality data, meteorological data, and pollution source data includes: data cleaning, removing outliers and noise data; data normalization, unifying data in different ranges to the same scale; and data interpolation, filling in missing data points to ensure the accuracy and integrity of the data.

[0055] Specifically, according to the historical environmental monitoring data, obtain the environmental monitoring matrix at different sampling times based on the same time interval, and generate a training data set and a test data set;

[0056] Fuse the ability of the convolutional neural network to extract spatial features and the ability of the recurrent neural network to process time series features to build an air quality prediction model;

[0057] Train the air quality prediction model through the training data set, and test the air quality prediction model through the test data set.

[0058] Specifically, the output result of the air quality prediction model is combined with the real-time collected air quality data, meteorological data, and pollution source data to construct a sampling strategy optimization model, and the sampling strategy optimization model includes multiple sampling strategies; the optimal sampling strategy is obtained by solving the sampling strategy optimization model according to the multiple sampling strategies.

[0059] Specifically, the output result of the air quality prediction model is fused with the real-time collected air quality data, meteorological data, and pollution source data to obtain a fused data set, where the fusion process includes the alignment and synchronization of time series data.

[0060] Based on the fused data set, the structure of the sampling strategy optimization model is determined.

[0061] Multiple sampling strategies are incorporated into the sampling strategy optimization model to solve the sampling strategy optimization model incorporating multiple sampling strategies.

[0062] Specifically, incorporating multiple sampling strategies into the sampling strategy optimization model includes:

[0063] Determine the type of sampling strategy incorporated into the sampling strategy optimization model; comprehensively consider factors such as the trend of air quality change, pollution characteristics in different regions, and limitations of monitoring resources, etc., to determine the type of sampling strategy incorporated into the model, such as a differential sampling strategy based on regional pollution risk, a sampling strategy dynamically adjusted according to meteorological conditions, etc.

[0064] Set corresponding decision variables for each sampling strategy, where each of the decision variables will become the direct manipulation object of the multi-objective optimization model; for each determined sampling strategy, set corresponding decision variables, and these decision variables will directly affect the implementation of the sampling strategy and become the direct manipulation object of the multi-objective optimization model. For example, the sampling frequency, sampling duration, selection of sampling points, etc. can all be used as decision variables.

[0065] Set variable constraint conditions for the sampling strategy variables; to ensure the feasibility and effectiveness of the sampling strategy, set variable constraint conditions for the sampling strategy variables, including the working time limit of air sampling equipment, the total amount limit of monitoring resources, the geographical condition limit of sampling points, etc., which can all be incorporated into the model as constraint conditions.

[0066] Construct at least two mutually independent and conflicting objective functions; one objective function can be to maximize the accuracy and representativeness of the sampling data, and the other objective function can be to minimize the sampling cost and resource consumption. There is a trade-off relationship between these objective functions, and the optimal balance is sought through an optimization algorithm.

[0067] Integrate the decision variables, variable constraint conditions, and objective function corresponding to the sampling strategy into the sampling strategy optimization model framework to form a complete sampling strategy optimization model for seeking the best sampling strategy. By solving this model, the optimal sampling strategy under the current data and conditions can be obtained.

[0068] Specifically, a convolutional neural network is constructed. A matrix input to the model corresponds to a corresponding convolutional neural network. If q historical environmental monitoring matrices are input, then q corresponding convolutional neural networks are involved. The convolutional neural network consists of several convolutional modules and fully connected layers. When used, the network parameters are randomly initialized. Each convolutional module consists of a convolutional layer and zero or one pooling layer. Each independent convolutional neural network outputs a spatial feature vector through a flattening layer.

[0069] Construct a recurrent neural network, which consists of several layers of neurons, and the number of neurons in each layer is k. To alleviate the problems of gradient explosion and complex parameters in the recurrent neural network, the recurrent neural network adopts a recurrent neural network based on MGU units.

[0070] After the recurrent neural network is constructed, a flattening layer is added to this network.

[0071] The output of the neurons in the flattening layer is connected through a fully connected layer, and finally the value of the environmental monitoring matrix at the future moment is output through a regression prediction layer.

[0072] Specifically, training the air quality prediction model with a training dataset includes:

[0073] Obtain the data input to the model from the training dataset with a sliding window of time width q, which can be represented by a matrix as follows:

[0074] X input =[X t-q ,X t-q+1 ,…,X t

[0075] Where the input to the model is the environmental monitoring matrix of q historical time slots before time t, and the goal is to predict the value of the environmental monitoring matrix at the (t + 1)-th time slot in the future.

[0076] Input the continuous training historical environmental monitoring data obtained using the sliding window into the air quality prediction model for forward propagation of the air quality prediction model.

[0077] Among them, the outputs of q convolutional neural networks are input into the recurrent neural network. Combining the time dependence of air quality, the predicted air vector feature values are output. The vector output by the q-th MGU unit is denoted as h t ​, the output of the k-th MGU unit is the output h of the hidden gating unit at the corresponding sampling moment. t , h t The h vector is flattened by a flattening layer and used as the input to the fully connected layer;

[0078] By calculating the loss between the predicted value and the true value of the model, backpropagation is performed on the air quality prediction model to adjust the corresponding parameters in the model;

[0079] The Adam optimization method is used to optimize the loss L of the overall network, and its formula is as follows:

[0080]

[0081] where p ij represents the predicted value corresponding to the model target matrix at the corresponding sampling time slot, and x ij represents the true traffic value at the corresponding sampling time slot;

[0082] Repeat the above steps. When the number of iterations reaches the preset number, stop the model training;

[0083] Continue to use Adam to optimize the loss function, adjust the parameters, and obtain the final air quality prediction model.

[0084] Specifically, the historical environmental monitoring data includes air quality data, meteorological data, and pollution source data.

[0085] In a specific embodiment of the present invention, an air sampling strategy optimization system based on artificial intelligence, as Figure 2 shown, includes:

[0086] A data collection and preprocessing module, which is used to collect air quality data, meteorological data, and pollution source data in real time, and preprocess the collected air quality data, meteorological data, and pollution source data;

[0087] An air quality prediction module, which is used to construct an air quality prediction model based on a neural network, use historical air quality data, meteorological data, and pollution source data as training data, train and optimize the air quality prediction model, and predict and output the air quality change trend in the future for a period of time; this module uses a model structure that combines CNN and RNN, trains through a large amount of historical data, and continuously optimizes the model parameters to improve the prediction accuracy.

[0088] A sampling strategy formulation module, which is used to formulate an optimal sampling strategy based on the output result of the air quality prediction model and in combination with the real-time collected air quality data, meteorological data, and pollution source data;

[0089] An automatic control module, which is used to automatically control the operation of the air sampling device according to the optimal sampling strategy.

[0090] In a specific embodiment of the present invention, an optimization method for air sampling strategy based on artificial intelligence includes: S1. Data collection and preprocessing:

[0091] In practical applications, air quality data, meteorological data, and pollution source data are collected in real time through air quality monitoring stations, meteorological stations, and pollution source monitoring devices distributed in different regions of the city. After the collected data is transmitted to the data processing center, data cleaning is first performed. By setting a threshold range, obvious abnormal data points are removed, such as abnormally high or low values caused by sensor failures. Then, data normalization processing is carried out to unify the data of different pollutant concentrations and meteorological elements into the [0,1] interval, eliminating the influence of data scales. For missing data points, linear interpolation or interpolation methods based on time series are used to fill them, ensuring the continuity and integrity of the data.

[0092] S2. Construction and training of the air quality prediction model

[0093] S201. Data preparation

[0094] Collect historical environmental monitoring data for a past period (such as one year), and combine the air quality, meteorological, and pollution source data at each sampling moment into an environmental monitoring matrix at an hourly time interval. These matrices are divided into a training data set and a test data set in a ratio of 8:2. The training data set is used for model training, and the test data set is used to evaluate the performance of the model.

[0095] S202. Model construction

[0096] Construct a convolutional neural network and a recurrent neural network, and fuse them into an air quality prediction model. The specific implementation is as follows:

[0097] Convolutional neural network: Assume that the input is an environmental monitoring matrix of 12 historical moments, and each matrix corresponds to a convolutional neural network. Each convolutional neural network contains 3 convolutional modules and 1 fully connected layer. The convolutional layer in the convolutional module uses a 3×3 convolutional kernel, the stride is 1, the pooling layer uses max pooling, and the pooling window size is 2×2. The output of the convolutional neural network is converted into a one-dimensional spatial feature vector through a flattening layer.

[0098] Recurrent neural network: The recurrent neural network contains 2 layers of neurons, and the number of neurons in each layer is 12. A flattening layer is added at the output end of the recurrent neural network to convert the output into a one-dimensional vector.

[0099] Model integration: Connect the flattened outputs of the convolutional neural network and the recurrent neural network through a fully connected layer, and finally output the environmental monitoring matrix values at future times through a regression prediction layer.

[0100] S203. Model training

[0101] Use the Adam optimizer to train the model, set the learning rate to 0.001, and the number of iterations to 100. In each iteration, obtain the input data of the model from the training dataset with a sliding window of time width 12, perform forward propagation and backward propagation, and adjust the model parameters. When the number of iterations reaches 100, stop training, continue to use the Adam optimizer to optimize the loss function, and fine-tune the model parameters to obtain the final air quality prediction model.

[0102] S204. Model testing

[0103] Use the test dataset to test the trained air quality prediction model, calculate evaluation metrics such as the mean squared error (MSE) and mean absolute error (MAE) between the model prediction values and the true values, and evaluate the prediction accuracy and generalization ability of the model.

[0104] S3. Formulation of the optimal sampling strategy

[0105] S301. Data fusion

[0106] Fuse the output results of the air quality prediction model with the real-time collected air quality data, meteorological data, and pollution source data. During the fusion process, align the data from different sources through timestamps to ensure the consistency of the data in the time dimension. For time series data, use the method of linear interpolation for synchronization to accurately match the data with different time intervals.

[0107] S302. Construction of the sampling strategy optimization model

[0108] Determine the types of sampling strategies included in the sampling strategy optimization model, including sampling strategies based on pollution risk, sampling strategies adjusted according to meteorological conditions, and sampling strategies based on regional functions. Set the corresponding decision variables for each sampling strategy:

[0109] Sampling strategy based on pollution risk: The decision variables are the sampling frequency and sampling duration in high-pollution areas; Sampling strategy adjusted according to meteorological conditions: The decision variables are the sampling points under different wind directions and wind speeds; Sampling strategy based on regional functions: The decision variables are the sampling proportions in commercial areas, industrial areas, and residential areas.

[0110] Set variable constraints for the sampling strategy variables, such as the maximum working time of the sampling device being 24 hours per day, and the total limit of monitoring resources being 100 samples collected per day. Construct the objective function, including maximizing the accuracy and representativeness of the sampling data, and minimizing the sampling cost and resource consumption. Integrate the decision variables, variable constraints, and objective function into the sampling strategy optimization model.

[0111] S303. Model solution

[0112] Use the genetic algorithm to solve the sampling strategy optimization model, set the population size to 50, and the number of iterations to 50. In each iteration, generate a new population through selection, crossover, and mutation operations, calculate the fitness value of each individual, and select the individual with the optimal fitness value as part of the next generation population. When the number of iterations reaches 50, output the optimal sampling strategy.

[0113] S4. Control of the air sampling device

[0114] Automatically control the operation of the air sampling device according to the optimal sampling strategy. Send parameters such as the sampling frequency, sampling location, and sampling duration to the device through the communication interface with the air sampling device. For example, for a highly polluted area, increase the sampling frequency from once per hour to once every 30 minutes, and extend the sampling duration from 10 minutes to 20 minutes; adjust the sampling location according to the changes in wind direction and wind speed to ensure that the most representative air samples can be collected.

[0115] During the actual operation process, monitor the operation status of the air sampling device in real time to ensure that the device samples according to the predetermined strategy. If the device fails or an abnormal situation occurs, issue an alarm in a timely manner and perform corresponding processing to ensure the normal progress of the air sampling work.

[0116] In the embodiment of the present invention, by collecting multi-source environmental data in real time, using a neural network integrating CNN and RNN to construct a high-precision air quality prediction model, formulating an optimal sampling strategy in combination with a multi-objective optimization method, and automatically controlling the operation of the air sampling device, it is possible to dynamically adjust the sampling strategy according to real-time environmental changes, improve the pertinence and effectiveness of air sampling. The obtained air sample data is more representative, can accurately reflect the dynamic change process of air quality, provide reliable data support for environmental quality assessment, pollution source tracing, and treatment decision-making, help improve the accuracy and timeliness of air quality monitoring, and promote the development and application of environmental monitoring technologies.

[0117] In the present specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0118] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An air sampling strategy optimization method based on artificial intelligence, characterized in that: include: Collect air quality data, meteorological data, and pollution source data in real time, and pre-process the collected air quality data, meteorological data, and pollution source data; Build an air quality prediction model based on a neural network, use historical environmental monitoring data as training data, train and optimize the air quality prediction model, and predict and output the air quality change trend in the future; Based on the output of the air quality prediction model, combined with real-time collected air quality data, meteorological data and pollution source data, the optimal sampling strategy is formulated; Automatically control the operation of air sampling equipment according to the optimal sampling strategy.

2. The method for optimizing air sampling strategy based on artificial intelligence according to claim 1, characterized in that: According to the historical environmental monitoring data, the environmental monitoring matrix at different sampling times is obtained based on the same time interval to generate training data sets and test data sets; The air quality prediction model is constructed by integrating the convolutional neural network's ability to extract spatial features and the recurrent neural network's ability to process time series features; The air quality prediction model is trained using a training data set, and the air quality prediction model is tested using a test data set.

3. The method for optimizing air sampling strategy based on artificial intelligence according to claim 1, characterized in that: The output results of the air quality prediction model are combined with the real-time collected air quality data, meteorological data, and pollution source data to construct a sampling strategy optimization model, which includes multiple sampling strategies; the sampling strategy optimization model is solved based on the multiple sampling strategies to obtain the optimal sampling strategy.

4. The method for optimizing air sampling strategy based on artificial intelligence according to claim 3, characterized in that: The output of the air quality prediction model is fused with the real-time collected air quality data, meteorological data, and pollution source data to obtain a fused data set. The fusion process includes the alignment and synchronization of time series data. Based on the fused data set, determining the structure of the sampling strategy optimization model; A plurality of sampling strategies are incorporated into a sampling strategy optimization model to solve the sampling strategy optimization model incorporating the plurality of sampling strategies.

5. The method for optimizing air sampling strategy based on artificial intelligence according to claim 2, characterized in that: Construct a convolutional neural network. A matrix input to the model corresponds to a corresponding convolutional neural network. Inputting the environmental monitoring matrix of q historical moments corresponds to q convolutional neural networks. The convolutional neural network consists of several layers of convolutional modules and fully connected layers. The network parameters are randomly initialized when used. Each layer of convolutional module consists of a convolutional layer and zero or one pooling layer. Each independent convolutional neural network outputs a spatial feature vector through a flattening layer. Constructing a recurrent neural network, wherein the recurrent neural network is composed of several layers of neurons, and the number of neurons in each layer is q; After the recurrent neural network is built, add a flattening layer to the network; The output of the neurons in the flattening layer is connected through the fully connected layer, and finally the environment monitoring matrix value at the future moment is output through the regression prediction layer.

6. The method for optimizing air sampling strategy based on artificial intelligence according to claim 5, characterized in that: The air quality prediction model is trained using a training data set, including: Obtain model input data from the training data set using a sliding window with a time width of q; The continuous training historical environmental monitoring data obtained by using a sliding window is input into the air quality prediction model, and the air quality prediction model is forward propagated; By calculating the loss between the model's predicted value and the true value, the air quality prediction model is back-propagated to adjust the corresponding parameters in the model; Use the Adam optimization method to optimize the loss of the overall network; Repeat the above steps, and when the number of iterations reaches the preset number, stop model training; Continue to use Adam to optimize the loss function and adjust the parameters to obtain the final air quality prediction model.

7. The method for optimizing air sampling strategy based on artificial intelligence according to claim 1, characterized in that: The historical environmental monitoring data includes air quality data, meteorological data and pollution source data.

8. An air sampling strategy optimization system based on artificial intelligence, characterized in that: include: Data collection and preprocessing module, used to collect air quality data, meteorological data and pollution source data in real time, and preprocess the collected data; The air quality prediction module is used to build an air quality prediction model based on a neural network. It uses historical air quality data, meteorological data, and pollution source data as training data to train and optimize the air quality prediction model, and predict and output the air quality change trend in the future. The sampling strategy formulation module is used to formulate the optimal sampling strategy based on the output of the air quality prediction model and the real-time collected air quality data, meteorological data and pollution source data; The automatic control module is used to automatically control the operation of the air sampling equipment according to the optimal sampling strategy.

Images