Automatic monitoring system and distribution prediction method for concentration of greenhouse gas in inland water body

Abstract

The embodiment of the invention discloses an automatic monitoring system and a distribution prediction method for greenhouse gas concentration in an inland water body. The system comprises an acidity sensor and a temperature sensor which are respectively used for collecting the pH value and the temperature of water at a monitoring point; the carbon dioxide sensor is used for collecting the concentration of bicarbonate in water at a monitoring point; the methane sensor and the nitrous oxide sensor are respectively used for collecting the contents of greenhouse gases CH4 and N2O dissolved in the monitoring point water; and the operator device is used for calculating the concentration of the greenhouse gas in the water according to the sensor data. According to the embodiment, the concentration of various greenhouse gases in the water body under the actual condition can be accurately monitored.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of detecting gas concentration in water, and in particular to an automatic monitoring system, a monitoring method, and a distribution prediction method for greenhouse gas concentration in inland water bodies. Background Art

[0002] Under the dual influence of global climate change and human activities, the challenges faced by the water environment are becoming increasingly severe. Greenhouse gases dissolved in water, such as carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O), not only enter the atmosphere through the air-water interface exchange, enhancing the greenhouse effect and further exacerbating global warming; the change in their concentration may also affect the metabolic processes of aquatic plants and microorganisms, thereby changing the structure and function of the entire ecosystem. These parameters are crucial for water environments such as wetlands, rivers, lakes, and reservoirs, as well as for microbial activities.

[0003] Currently, for the quantitative technology of greenhouse gas concentration in water bodies, on-site gas collection using a closed chamber or gas extraction using the headspace method is mostly adopted. The basic principle is the two-phase equilibrium of the measured substance in the gas-liquid or gas-solid phase, and then an experimental instrument device (such as a gas chromatograph) is used to test and analyze the gas. First of all, these two technologies not only require specific scientific experimental instruments to detect and quantitatively analyze gases, but the instrument equipment is complex, the testing cost is expensive, it is not convenient for field operation, and in-situ continuous automatic detection cannot be achieved; secondly, both of these two technologies ignore the influence of the wind field on the gas exchange rate at the water-air interface. Especially for waters with a relatively large water flow velocity, it will cause a significant difference between the equilibrium gas in the closed space and the actual water environment conditions, resulting in a huge error in the detection data of greenhouse gas concentration in water bodies.

[0004] Patent application CN112255385A discloses a system, method, and application for in-situ continuous determination of greenhouse gas concentration in highly turbid water bodies. It uses a diaphragm pump to pump water through an atomizing nozzle at the top of the equilibrium chamber to atomize and exchange gas with the gas in the equilibrium chamber, and a greenhouse gas detector to measure the greenhouse gas concentration in the equilibrium chamber. Although it overcomes the problems of the greenhouse gas measurement technology represented by the floating box technology and the headspace technology, the extraction of water samples destroys the original water environment state, resulting in huge errors in the water pressure, saturated water vapor pressure, and equilibrium gas in the measuring parameters in the equilibrium device; in addition, since the greenhouse gas concentration in the water environment changes in real time and over a long period, maintaining the Henry equilibrium mode of water-gas in the equilibrium chamber requires a specific long time, making this technology have certain limitations in the time-frequency domain.

[0005] Patent application CN117347456A discloses an on-line automatic detection device for carbon dioxide partial pressure in inland water bodies and its detection method. Three types of sensors are used to on-line detect the carbon dioxide partial pressure in water bodies, overcoming the water body greenhouse gas measurement technologies represented by the floating box technology and the headspace technology. However, it is still necessary to further quantify the concentration of carbon dioxide in water bodies using water environment parameters, and there is a lack of monitoring of the concentrations of the other two important greenhouse gases in water bodies, methane and nitrous oxide, which have a much greater global warming potential, 25 times and 298 times that of carbon dioxide respectively. Summary of the Invention

[0006] An embodiment of the present invention provides an automatic monitoring system and distribution prediction method for greenhouse gas concentrations in inland water bodies to solve the above technical problems.

[0007] In a first aspect, an embodiment of the present invention provides an automatic monitoring system for greenhouse gas concentrations in inland water bodies, which is applied to inland water bodies. The system includes:

[0008] An acidity sensor and a temperature sensor, which are respectively used to collect the pH value S1 and temperature S2 of the water at the monitoring point;

[0009] A carbon dioxide sensor, which is used to collect the bicarbonate concentration S in the water at the monitoring point CO2 ;

[0010] A methane sensor and a nitrous oxide sensor, which are respectively used to collect the contents S of dissolved greenhouse gases CH4 and N2O in the water at the monitoring point CH4 and S N2O ;

[0011] An operator device, which is used to calculate the concentration of greenhouse gases in water in the following manner:

[0012] S110. According to the following formula, calculate the solubilities K CO2 , K CH4 and K N2O of CO2, CH4 and N2O in the surface water, and the Henry's equilibrium constant k of the bicarbonate concentration in the water body h :

[0013]

[0014] where A, B, C, D, E, F, G, H, I, J, K, L, M, N, P, Q, R are all fixed constants calibrated or derived;

[0015] S120. According to the following formula, calculate the concentrations C CO2 , C CH4 and C N2O of dissolved CO2, CH4 and N2O in the water body at the monitoring point:

[0016] C CO2 = S CO2 ×K CO2 ×k h

[0017] C CH4 = S CH4 ×K CH4

[0018] C N2O = S N2O ×K N2O

[0019] In a second aspect, an embodiment of the present invention provides a method for predicting the concentration distribution of greenhouse gases in inland waters, including:

[0020] S210. Divide the water area plane to be monitored into multiple grids;

[0021] S220. Take each grid as a monitoring point, and respectively install the above-mentioned automatic monitoring system for the concentration of greenhouse gases in inland waters to monitor the concentration of any one greenhouse gas in the water body of each grid in real time; and take the concentrations of each grid at the same moment as samples to construct a finite sample set;

[0022] S230. Average the concentration distributions monitored over a period of time, and take multiple extreme value grids and large gradient grids in the averaged concentration distribution as multiple initial grids; and construct a neural network model with the concentrations of the multiple initial grids at the same moment as the input and the concentration distribution of all grids at the same moment as the output;

[0023] S240. Use the finite sample set to train and test the neural network model. If the model accuracy does not meet the requirements, cluster the grids with the difference between the predicted concentration value and the monitored value exceeding the set threshold in the test results according to the geographical location, and incorporate the cluster centers into the model input;

[0024] S250. Return to S240 according to the new neural network model, and repeat this process until the model accuracy meets the requirements;

[0025] S260. Use the genetic algorithm to select the optimal input grid combination from the input grid set of the final model that can meet the model accuracy requirements and has the smallest number of grids, and use the optimal model corresponding to the optimal input grid combination to predict the concentration distribution in the water body.

[0026] In summary, an automatic monitoring system and a distribution prediction method for the concentration of greenhouse gases in inland waters can achieve the following beneficial effects:

[0027] 1. In the prior art, the observation and sampling extraction change the original environmental conditions. Affected by the gas phase equilibrium of the basic principle of the technology, there is a large error between the greenhouse gas concentration and the actual value. In contrast, this embodiment does not change the environmental conditions of the original water body, and the automatically on-line monitored data is more accurate.

[0028] 2. In the prior art, specific gas analysis instruments such as gas chromatographs are needed for test analysis. The test is costly, the equipment is complex, and it is not convenient for field detection operations. In contrast, the monitoring system of this embodiment is convenient to use, and the monitored data can be wirelessly transmitted and received by the PC side.

[0029] 3. In the prior art, the monitoring of greenhouse gases has spatio-temporal limitations, and it is impossible to simultaneously conduct fixed-point, real-time, and long-term series on-line monitoring of three greenhouse gases, namely carbon dioxide, methane, and nitrous oxide. In contrast, this embodiment uses on-line water temperature to calibrate the measurement of dissolved CO2, CH4, and N2O in water, and can automatically realize the concentration monitoring of CO2, CH4, and N2O in the water body in real time, at fixed points, and in long-term series.

[0030] 4. The concentration distribution prediction method of the greenhouse gas concentration automatic monitoring system in inland water bodies of this embodiment combines partial point monitoring and overall water area prediction. By using as few greenhouse gas concentration automatic monitoring systems in inland water bodies as possible, the greenhouse gas concentration distribution of the entire water area is obtained, the number of devices used for long-term monitoring is reduced, the device utilization rate is improved, and the monitoring cost of greenhouse gas concentration is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 is a schematic structural diagram of an automatic monitoring system for greenhouse gas concentration in inland water bodies provided by an embodiment of the present invention;

[0033] Figure 2 is an installation schematic diagram of an automatic monitoring system for greenhouse gas concentration provided by an embodiment of the present invention;

[0034] Figure 3 is a box plot statistical value and linear correlation comparison diagram of the data results of two methods provided by an embodiment of the present invention;

[0035] Figure 4 is a flowchart of a method for predicting the concentration distribution of greenhouse gas in inland water bodies provided by an embodiment of the present invention;

[0036] Figure 5 It is a schematic diagram of grid division for a water area to be monitored provided by an embodiment of the present invention;

[0037] Reference numerals:

[0038] 1. Power supply device;

[0039] 2. Warning light;

[0040] 3. Signal transmitter;

[0041] 4. Industrial control computer;

[0042] 5. Acidity sensor;

[0043] 6. Temperature sensor;

[0044] 7. Carbon dioxide sensor;

[0045] 8. Methane sensor;

[0046] 9. Nitrous oxide sensor. Detailed implementation manners

[0047] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be described clearly and completely below. Apparently, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope protected by the present invention.

[0048] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0049] In the description of the present invention, it should also be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0050] Figure 1 and Figure 2 are respectively the structural schematic diagram and the installation schematic diagram of an automatic monitoring system for greenhouse gas concentration in inland water bodies provided by an embodiment of the present invention. Combining Figure 1 and Figure 2 , the system mainly includes a floating box, a power supply device 1, a series of sensors and an industrial control computer 4. The power supply device 1 mainly provides electrical energy for the system through the upper photovoltaic panel and the lower battery. The warning light 2 is used to prevent collisions with water vehicles.

[0051] When in use, the floating box can be fixed at the position of the water body to be monitored in the wild by using reinforced concrete anchor points. Through the float-type automatic regulator of the floating box, the probes of the 5 sensors are located at a depth of 5-10 cm below the water surface. The photovoltaic solar panel is used to supply power to the entire system. The monitoring principle of greenhouse gas concentration is the water chemical ion balance relationship and Henry's law in the water body. The data acquisition module of the industrial control computer 4 collects and calculates the sensor signals, and then the signal transmitter 3 of the remote communication module outputs the result data (wired RS485 communication interface and wireless GPRS transmission).

[0052] Before installing the system on-site in the wild, first use pH buffer solutions (2.00, 4.01, 7.00, 9.21, and 10.00) to perform pH electrode standard calibration on the acidity sensor 5 to accurately measure the pH value (unit, dimensionless) of the water environment parameters of the surface water body; then use NaHCO3 standard solutions (0.1 mM, 0.5 mM, 1.0 mM, 5.0 mM, and 10.0 mM, mM represents millimoles per liter) to perform ion electrode standard calibration on the carbon dioxide sensor 7 to accurately measure the content of bicarbonate in the surface water body (unit, mM).

[0053] The system automatically compensates and corrects the signals of the acidity sensor 5 and the carbon dioxide sensor 7 by using the accurate signals (±0.01 °) transmitted by the temperature sensor 6; collects the signals of the 5 sensors. Among them, the signal S1 of the acidity sensor 5 is the pH value of the water at the monitoring point (unit, dimensionless), the signal S2 of the temperature sensor 6 is the Kelvin temperature of the water body at the monitoring point (unit, °K), and the signal S CO2 of the carbon dioxide sensor 7 is the concentration of bicarbonate in the water at the monitoring point (unit, mM), the signal S CH4 of the methane sensor 8 is the content of dissolved greenhouse gas CH4 in the water at the monitoring point (unit, ppm), and the signal S N2OFor monitoring the content of dissolved greenhouse gas N2O in water at the monitoring point (unit: ppm); and based on the water chemical equilibrium equation and Henry's law, the operator device embedded in the acquisition module is used to calculate the signals collected by the sensor to obtain the concentrations of dissolved CO2, CH4, and N2O in water.

[0054] In a specific embodiment, the calculation method may include the following steps:

[0055] S110. Calculate the solubility K of CO2, CH4, and N2O in surface water according to the following formula: CO2 K CH4 and K N2O (unit: mole·L -1 ·atm -1 ), and the Henry equilibrium constant k of the bicarbonate concentration in the water body (unit: atm·L·mol h ): -1 )

[0056]

[0057]

[0058] S120. Calculate the concentrations C of dissolved CO2, CH4, and N2O in the water body at the monitoring point according to the following formula: CO2 C CH4 and C N2O (unit: μM, micromole):

[0059] C CO2 = S CO2 ×K CO2 ×k h (5)

[0060] C CH4 = S CH4 ×K CH4 (6)

[0061] C N2O = S N2O ×K N2O (7)

[0062] Optionally, the data acquisition module may store the signals C CO2 C CH4 and C N2O or wirelessly transmit them to the office PC for storage.

[0063] Further, in order to verify the reliability of the automatic monitoring system for greenhouse gas concentration in inland water bodies in this embodiment, this embodiment also comparatively analyzed the differences between the gas concentrations measured by the system of the present application and the traditional headspace method technology. In a specific implementation manner, 78 sampling points along the river surface water body can be sampled in the field, and the bicarbonate signal S CO2 value, the signal S1 value obtained by the pH sensor of the pH value, and the signal S2 value obtained by the temperature sensor are obtained, and the carbon dioxide concentration signal C CO2 value in the water body is obtained through the above formula (5). At the same time, water samples are collected at the sensor detection position, and the headspace method technology and the indoor instrument gas chromatograph are used for analysis and testing to obtain the partial pressure P CO2 value (μatm) of carbon dioxide in the sample. Combining the signal S2 value obtained by the temperature sensor in the field, the dissolved carbon dioxide concentration C CO2 (unit, mM) in the water body is calculated. The calculation formula is:

[0064] C CO2-sample =P CO2 ×K CO2 (8)

[0065] One-way analysis of variance is performed on the data obtained by the two technical methods. F>Fcrit, and the p-value is 0.0356, which is obviously less than 0.05, indicating that there are significant differences between the two groups of data. The comparison results of the two methods are as Figure 3 shown. The data obtained by the method of this patent has a very strong correlation with the data obtained by the sample headspace-gas chromatograph analysis and testing method. The correlation coefficient R 2 reaches 0.83, indicating that the overall data change trends are consistent. However, since the sample headspace method changes the original water environment, the environmental temperatures tend to be the same due to the indoor operation of the water sample container being brought back, and in addition, the headspace method cannot well separate the dissolved gases in the water and it is very difficult to completely extract all the gases in the headspace after gas equilibrium for gas volume measurement. Therefore, compared with the method of this patent, the data change range and detection values obtained by the sample headspace-gas chromatograph analysis and testing method are both smaller; it shows that the method adopted in this patent obtains data in the original water environment and has better representativeness and accuracy.

[0066] In summary, this embodiment provides an automatic monitoring system for greenhouse gas concentration in inland water bodies, which can achieve the following beneficial effects:

[0067] 1. The existing technology observes, extracts and samples, which changes the original environmental conditions. Affected by the basic principle of gas phase equilibrium of the technology, the greenhouse gas concentration has a large error compared with the actual situation; while this embodiment does not change the environmental conditions of the original water body, and the automatically online monitored data is more accurate;

[0068] 2. In the prior art, specific gas analysis instruments such as gas chromatographs are required for test analysis. The tests are costly, the equipment is complex, and it is not convenient for field detection operations. However, the monitoring system of this embodiment is convenient to use, and the monitoring data can be wirelessly transmitted and received on the PC side.

[0069] 3. In the prior art, the monitoring of greenhouse gases has spatio-temporal limitations and cannot simultaneously perform fixed-point, real-time, and long-term series on-line monitoring of three greenhouse gases, namely carbon dioxide, methane, and nitrous oxide. However, this embodiment uses on-line water temperature calibration to measure dissolved CO2, CH4, and N2O in water, and can automatically achieve real-time, fixed-point, and long-term series monitoring of the concentrations of CO2, CH4, and N2O in the water body.

[0070] Based on the above-mentioned automatic monitoring system for greenhouse gas concentrations in inland water bodies, Figure 4 is a flowchart of a method for predicting the distribution of greenhouse gas concentrations in inland water bodies. This method can predict the distribution of greenhouse gas concentrations in the entire water area through as few monitoring points as possible, reducing the number of in-service devices in long-term monitoring. As Figure 4 shown, this method specifically includes:

[0071] S210. Divide the plane of the water area to be monitored into multiple grids.

[0072] Figure 5 Taking a regular rectangle as an example, a top view of a water area to be monitored is shown. As Figure 5 shown, this plane can be divided into multiple grids for zonal monitoring. For water area planes with irregular shapes, the division method is similar.

[0073] S220. Take each grid as a monitoring point, and respectively install the above-mentioned automatic monitoring system for greenhouse gas concentrations in inland water bodies to monitor the concentration of any one greenhouse gas in the water body of each grid in real time; and take the concentrations of each grid at the same moment as samples to construct a finite sample set.

[0074] Combined with Figure 2 In each grid, in the manner shown in Figure 2 shown, respectively install the automatic monitoring system for greenhouse gas concentrations, collect sensor data in real time, and calculate the greenhouse gas concentrations of each grid at each moment.

[0075] After monitoring for a period of time, for any greenhouse gas, the concentrations of the greenhouse gas in each grid at the same moment in the monitoring data are taken as a sample. Multiple samples can be obtained from the data at multiple historical moments, forming a finite sample set. The finiteness of the sample set is emphasized here because for a large water area, the cost of installing an automatic monitoring system for the concentration of each greenhouse gas in each grid is very high. Therefore, in this embodiment, only a finite duration of sample collection can be achieved while ensuring that there are devices in all grids. In the subsequent steps, a prediction model for the concentration distribution of the greenhouse gas will be trained using the finite samples, and the concentration distribution of the entire water area will be predicted based on the monitored concentrations of some grids, thereby reducing the number of long-term in-service devices and the concentration monitoring cost of the entire water area.

[0076] S230. Average the concentration distribution monitored over a period of time, and take multiple extreme value grids and large gradient grids in the averaged concentration distribution as multiple initial grids; and construct a neural network model with the concentrations of the multiple initial grids at the same moment as the input and the concentration distribution of all grids at the same moment as the output.

[0077] This period of time can be selected as a period when the water flow, temperature, etc. are relatively stable. The concentration distribution at each moment during this period is regarded as a two-dimensional matrix arranged according to the geographical location of the grids (in an irregular water area, the grid elements corresponding to the land positions are permanently set to 0). By adding the two-dimensional matrices at multiple moments and dividing by the number of moments in this period, the average concentration distribution in this period can be obtained. From the average concentration distribution, select the grids where the extreme value points are located and the grids where the change gradient exceeds the set threshold, and jointly form the initial set of the input grids of the neural network model. For the convenience of description, the grids in these initial sets are all referred to as initial grids in this embodiment. The reason for selecting these grids is that these grids often represent the key positions where the concentration distribution of the entire water area turns or mutates, and can provide key information for predicting the concentration distribution of the entire water area in the later stage.

[0078] Based on this initial set, a neural network model is constructed. In this embodiment, the model will be trained so that after inputting the concentrations of the greenhouse gas in these initial grids at the same moment, it can output the concentration distribution of the greenhouse gas in the water of all grids at this moment. Optionally, the model can adopt a multi-layer convolutional neural network, and a fully connected layer is connected after the convolutional neural network to generate results of a specific dimension; the dimensions of the input layer and the output layer can be adjusted according to the number of input data and output data; the number of convolutional layers is greater than 5 layers, and the convolutional kernel can adopt 3×3 to expand or extract data features layer by layer; the fully connected layer is used to connect the dimensions of the last convolutional layer and the output layer.

[0079] S240. Use the finite sample set to train and test the neural network model. If the model accuracy does not meet the requirements, cluster the grids where the difference between the predicted concentration value and the monitored value in the test result exceeds the set threshold according to the geographical location, and incorporate the cluster centers into the model input.

[0080] As described above, the model needs to learn the data relationship between the concentrations of some grids and all grids. Due to a large amount of missing grid data, the learning difficulty of the model is very high. In the case of limited sample size, the accuracy of the trained model is likely to fall short of the ideal requirements. Therefore, in this embodiment, when the model accuracy is poor, by increasing the number of input grids, more data information is provided for the model, and the learning difficulty is gradually reduced.

[0081] In a specific embodiment, to make full use of the limited sample data, the model training can be completed through the following steps:

[0082] Step 1. Divide the finite sample set into a first training set, a second training set, and a test set. Among them, the first training set and the second training set are used to correspond to and distinguish the subsequent two training stages. The samples in the two training sets can be the same, different, or partially the same, and this embodiment does not make specific restrictions.

[0083] Step 2. Use the first training set to perform a first training on the neural network model. In the first training, update the model parameters by minimizing the concentration change of adjacent grids. This step corresponds to the first training stage of the model. In this stage, the model parameters are updated using the law followed by the input data itself, so that the model output has a certain law. Due to the continuity of water bodies, the change in the concentration of greenhouse gases in water should also be relatively uniform. Therefore, in this embodiment, the model parameters are updated by constraining the concentration change of adjacent grids to be minimized, so as to minimize the random changes without rules as much as possible. Optionally, the following loss function can be constructed:

[0084]

[0085] where L1 represents the loss function value, i and j respectively represent the grid indices along the horizontal and vertical directions, F(i,j) represents the predicted concentration value of the grid in the i-th row and j-th column, represents the sum over all grids in the water body (excluding land grids with a permanent concentration of 0), and abs represents taking the absolute value. By minimizing this loss function, the model parameters are updated.

[0086] Step 3: Use the second training set to perform secondary training on the model after the first training. During the secondary training, update the model parameters by minimizing the difference between the predicted concentration values and the monitored values of each grid. This step corresponds to the second training phase of the model. In this phase, the actually monitored concentration data is used as the true value, and the model output is constrained to continuously approach the true value to perform supervised learning on the model after the first training. Optionally, the following loss function can be constructed:

[0087]

[0088] where \(L_2\) represents the value of the loss function, represents the true concentration value of the grid in the \(i\)-th row and \(j\)-th column.

[0089] Furthermore, in this embodiment, it is also considered that due to equipment failures or communication failures in some grids, the true value data is empty. In this case, although the sample data lacks the data of some grids, it can still be used for the secondary training of the model. It only needs to simultaneously constrain the minimum change in the concentration between the partial grids and the adjacent grids, and the minimum difference between the predicted concentration value and the monitored value of other grids in the loss function. Optionally, the following loss function can be constructed for such samples:

[0090]

[0091] where \(L_3\) represents the value of the loss function, \(i_1\) and \(j_1\) represent the two-dimensional indices of the grids with complete data, \(i_2\) and \(j_2\) represent the two-dimensional indices of the grids with missing data, \(F(,)\) represents the predicted concentration value of the corresponding grid, represents the monitored concentration value of the corresponding grid, and \(\alpha\) and \(\beta\) respectively represent the weight coefficients. In this way, for the samples with complete data, the network parameters can be updated using \(L_2\), and for the samples with missing data, the network parameters can be updated using \(L_3\), further avoiding the waste of sample data.

[0092] The above two training phases first let the model learn the laws of the input data itself, and then perform one-to-one supervised training through the true value of each grid. Compared with directly using the true value data for supervised training, the training method of this embodiment can accelerate the model convergence speed and achieve as good a training effect as possible using a limited sample set.

[0093] Step 4: After the training is completed, use the test set to test the model after the secondary training. If the model accuracy still does not meet the requirements, cluster the grids whose difference between the predicted concentration value and the monitored value in the test results exceeds the set threshold according to the geographical location, and include each cluster center in the model input. This step is for the situation where the model accuracy still has not reached a good level after two training phases, and selects the grids with large prediction errors from the prediction results of each test sample as the input grids of the model.

[0094] Optionally, if large errors occur in some grids in multiple test samples, cluster these grids according to their geographical locations to obtain multiple spatial clustering clusters. Each spatial clustering cluster represents a nearby area where the grids in this area have large measurement errors multiple times, indicating that the model cannot effectively learn the data patterns of this area from the existing input grids. Therefore, it is necessary to select one or more grids from this area as model inputs to provide effective information of this area for the model. Optionally, the DBSCAN clustering method can be used, and finally select the clustering center of each clustering cluster as the model input.

[0095] Furthermore, when adding new model inputs, on the basis of the original model, convolutional kernels connected to the new network and subsequent calculation layers can be added in the first convolutional layer; or a fully connected layer can be added between the input layer and the first convolutional layer to transform the dimension of the new input layer to the dimension of the original hidden layer; the remaining structures in the model remain unchanged and inherit the original parameters as initial parameters.

[0096] S250. Return to S240 according to the new neural network model, and so on until the model accuracy meets the requirements.

[0097] Specifically, use the finite sample set again to train and test the new neural network model; if the accuracy of the new model still does not meet the requirements, cluster the grids whose difference between the predicted concentration value and the monitored value in the test results exceeds the set threshold according to their geographical locations again, and include the clustering centers in the model inputs to obtain an updated neural network, and then return to S240. Repeat this process until the accuracy of the final model reaches the required index. At this time, the neural network model has sufficient effective grid data, and according to these data and the change law of the concentration itself, the concentration distribution of the entire water plane can be predicted more accurately. For the convenience of distinction and description, the neural network model obtained at this time is called the full-scale model.

[0098] S260. Use the genetic algorithm to select the optimal input grid combination from the input grid set of the full-scale model that can meet the model accuracy requirements and has the least number of grids, and use the optimal model corresponding to the optimal input grid combination to predict the concentration distribution in the water body.

[0099] As described above, in multiple cycles of S240 - S250, the model accuracy is improved by continuously increasing the number of input grids. To prevent redundancy in the final input grids, this step uses the genetic algorithm to screen the input grids of the full-scale model to remove redundant grids, and predicts the concentration distribution of the entire water plane with the least model inputs and the greenhouse gas concentration automatic monitoring system.

[0100] In a specific implementation, the above screening process can be completed in the following way:

[0101] Step 1. Construct a chromosome sequence structure of equal length according to the number of grids in the input grid set of the full model. Exemplarily, if the number of input grids of the full model is 15, a vector of length 15 can be constructed as the chromosome sequence structure in the genetic algorithm. Each bit in the vector corresponds to a grid in the input grid set. When the numerical bit takes 1, it means to retain the corresponding grid as the input of the neural network model, and when it takes 0, it means to delete the corresponding grid from the input grids of the neural network model.

[0102] Step 2. Generate the current population of the genetic algorithm according to the chromosome sequence structure. Exemplarily, multiple chromosomes can be generated by random initialization to form the current population.

[0103] Step 3. Prune the full model according to each chromosome in the current population, and use the finite sample set to train the pruned model three times. Specifically, for a certain chromosome, delete the grids corresponding to the bits with a value of 0 in the chromosome from the input grids of the full model, and delete all branches connecting the first hidden layer in the model to the input grids, obtaining a new model. Then, use the finite sample set to train this new model. In this embodiment, this training is called three - time training.

[0104] Step 4. Determine whether the model after three - time training meets the optimal goal. If not, perform selection, crossover, and mutation on the current population according to the genetic algorithm, select excellent chromosomes from the mutated current population to form a new current population, and return to Step 3. Prune the full model according to each chromosome in the new current population, and use the finite sample set to train the pruned model three times again. Repeat this cycle until the model after the final three - time training meets the optimal goal. Among them, the optimal goal is that the accuracy of the model after three - time training meets the requirements, and the model with the least number of input grids has been obtained in the three - time training carried out so far.

[0105] Optionally, after obtaining a chromosome that meets the model accuracy requirements each time, select the one with the least number of input grids from all the chromosomes that meet the model accuracy requirements accumulated so far as the optimal chromosome. If the optimal chromosome remains unchanged after continuous three - time trainings, it is considered that the model with the least number of input grids has been obtained. Of course, it is also possible to directly iterate to the maximum number of iterations, and after the iteration ends, select the model with the model accuracy meeting the requirements and the least number of input grids as the optimal model.

[0106] Furthermore, when selecting excellent chromosomes, a fitness function can be constructed according to the change patterns of the predicted gas concentration values between the input grids in the chromosome; according to the fitness function, excellent chromosomes are selected from the current population after mutation to construct a new current population. Optionally, the following fitness function can be constructed:

[0107]

[0108] where fit is the fitness value, num(G) represents the number of extreme value grids in the concentration distribution predicted by the model after three trainings for the input grids with a value of 1 in the current chromosome, i3 and j3 are the two-dimensional indices of the input grids with a value of 1 in the current chromosome, num(0) represents the number of grids with a value of 0 in the current chromosome, and γ, δ, and ε respectively represent weight coefficients. The larger the fitness value, the more excellent the chromosome.

[0109] Through the above fitness function, chromosomes with fewer input grids and more extreme value grids and large gradient grids (these grids represent the key morphological changes in the concentration distribution) covered by the input grids can be preferentially selected to evolve the new current population, which helps to obtain the optimal chromosome and the optimal model as soon as possible.

[0110] Finally, after obtaining the optimal input grid combination, only the automatic monitoring systems for greenhouse gas concentrations in the inland waters of each grid in this combination are retained, and the automatic monitoring systems for greenhouse gas concentrations in the inland waters of the remaining grids are removed; the remaining systems are used to monitor the gas concentration of each target grid in real time and input it into the optimal model, and the real-time concentration distribution of all grids can be obtained. The removed equipment systems can be used for the concentration monitoring set prediction of other waters, improving the equipment utilization rate and reducing the monitoring cost of greenhouse gas concentrations.

[0111] Furthermore, the above method takes a certain greenhouse gas as an example to illustrate the specific method of predicting the greenhouse gas concentration distribution of the entire water area with as few monitoring points as possible. In practical applications, for each greenhouse gas, the operations of S220 - S260 can be performed; among them, when predicting the concentration distribution in the water body using the optimal model corresponding to the optimal input grid combination in S260, the union of the optimal input grid combinations corresponding to various greenhouse gases is taken, the automatic monitoring systems for greenhouse gas concentrations in the inland waters of each target grid in the union are retained, and the automatic monitoring systems for greenhouse gas concentrations in the inland waters of the remaining grids are removed; the remaining systems are used to monitor the greenhouse gas concentration of each target grid in real time and are respectively input into the optimal models corresponding to various greenhouse gases to obtain the real-time greenhouse gas concentration distribution of all grids. In this way, the concentration distributions of three greenhouse gases in the entire water area can be predicted with as few monitoring points as possible.

[0112] It should be noted that in the above method for predicting the greenhouse gas concentration distribution at the water-air interface, except for the steps of installing and removing the automatic monitoring system for greenhouse gas concentration at the water-air interface, the remaining steps can be automatically executed by an electronic device.

[0113] In summary, this embodiment also provides a method for predicting the greenhouse gas concentration distribution in inland waters, which can achieve the following beneficial effects:

[0114] 1. By combining partial point monitoring and overall water area prediction, using as few automatic monitoring systems for greenhouse gas concentration in inland waters as possible, the greenhouse gas concentration distribution of the entire water area is obtained, reducing the number of devices for long-term monitoring, improving the device utilization rate, and reducing the monitoring cost of greenhouse gas concentration;

[0115] 2. Since the neural network model needs to learn complex concentration distribution rules based on partial grids, the learning difficulty is large and the number of samples is limited, and it is difficult for the trained model accuracy to reach the ideal requirements at one time. To address this difficulty, in this embodiment, when the model accuracy is not good, representative grids are selected from the areas with large prediction errors as the model input, and by increasing the number of input grids, more effective information is provided for the model, gradually reducing the learning difficulty;

[0116] 3. For the situation where the full-grid data samples are limited, this embodiment enables the model to first learn the rules of the input data itself through multi-stage training, and then perform one-to-one supervised training through the monitoring values of each grid. Compared with directly using the monitoring data for supervised training, this method can accelerate the model convergence speed and achieve the best possible training effect with a limited sample set;

[0117] 4. For the missing data samples caused by partial grid device or communication failures, this embodiment combines the uniform change of concentration and the monitoring values, adopts different loss value calculation methods for different grids, makes full use of the remaining data information in the samples, and further avoids sample waste;

[0118] 5. After obtaining the full-scale model including sufficient effective input grids, in order to prevent redundancy in the input grids of the full-scale model, this embodiment uses a genetic algorithm to screen the input grids, removes the redundant grids, and realizes the concentration distribution prediction of the entire water area plane with the least model input and automatic monitoring system for greenhouse gas concentration. This can greatly reduce the final number of devices and the monitoring cost in the case of a very large water area and predicting the concentration distribution of each greenhouse gas.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. An automatic monitoring system for greenhouse gas concentration in inland water bodies, characterized in that, Comprising: An acidity sensor and a temperature sensor, which are respectively used to collect the pH value S1 and the temperature S2 of the water at the monitoring point; A carbon dioxide sensor for collecting the concentration S of bicarbonate in the water at the monitoring point CO2 ; A methane sensor and a nitrous oxide sensor are respectively used to collect the contents S of dissolved greenhouse gases CH4 and N2O in the water at the monitoring point CH4 and S N2O ; An operator device, which is used to calculate the concentration of greenhouse gases in water in the following manner: S110. Calculate the solubilities \(K\) of \(CO_2\), \(CH_4\) and \(N_2O\) in surface water, the Henry's equilibrium constant \(k\) of the bicarbonate concentration in the water body according to the following formula: CO2 , \(K\) CH4 , and \(K\) N2O , and h : Wherein, A, B, C, D, E, F, G, H, I, J, K, L, M, N, P, Q, R are all fixed constants calibrated or derived; S120. Calculate the concentrations C CO2 , C CH4 , and C N2O of dissolved CO2, CH4, and N2O in the water body at the monitoring point according to the following formula: CO2 , C CH4 , and C N2O : C CO2 = S CO2 × K CO2 × k h C CH4 = S CH4 × K CH4 C N2O = S N2O × K N2O 2. A method for predicting the concentration distribution of greenhouse gases in inland waters, characterized in that, Comprising: S210. Divide the water area plane to be monitored into multiple grids; S220. Respectively use each grid as a monitoring point, and install the automatic monitoring system for the concentration of greenhouse gases in inland water bodies as described in claim 1 to monitor the concentration of any greenhouse gas in the water of each grid in real time; and use the concentrations of each grid at the same moment as samples to construct a finite sample set; S230. Average the concentration distributions monitored over a period of time, and take multiple extreme value grids and large gradient grids in the averaged concentration distribution as multiple initial grids; And construct a neural network model with the concentrations of the multiple initial grids at the same moment as the input and the concentration distribution of all grids at the same moment as the output; S240. Use the finite sample set to train and test the neural network model. If the model accuracy does not meet the requirements, cluster the grids with the difference between the concentration predicted value and the monitored value exceeding the set threshold in the test results according to the geographical location, and incorporate the cluster centers into the model input; S250. Return to S240 according to the new neural network model, and repeat this process until the model accuracy meets the requirements; S260. Use the genetic algorithm to select the optimal input grid combination from the input grid set of the final model that can meet the model accuracy requirements and has the smallest number of grids, and use the optimal model corresponding to the optimal input grid combination to predict the concentration distribution in the water body.

3. The method according to claim 2, wherein The training and testing of the neural network model using the finite sample set includes: Dividing the finite sample set into a first training set, a second training set and a test set; Using the first training set to perform one-time training on the neural network model, and updating the model parameters by minimizing the concentration change between adjacent grids during one-time training; Using the second training set to perform secondary training on the model after one-time training, and updating the model parameters by minimizing the difference between the concentration predicted value and the monitored value of each grid during secondary training; Using the test set to test the model after secondary training.

4. The method according to claim 3, wherein The secondary training of the model after one-time training using the second training set, and updating the model parameters by minimizing the difference between the concentration predicted value and the monitored value of each grid during secondary training includes: If there is missing monitoring data for some grids in the samples in the second training set, during secondary training, update the model parameters by minimizing the concentration change between the part of the grids and adjacent grids, and minimizing the difference between the concentration predicted values and the monitored values of other grids.

5. The method according to claim 4, wherein The updating of the model parameters by minimizing the concentration change between the part of the grids and adjacent grids, and minimizing the difference between the concentration predicted values and the monitored values of other grids includes: Construct the following loss function: Among them, L3 represents the loss function value, i1 and j1 represent the two-dimensional indices of the complete data grid, i2 and j2 represent the two-dimensional indices of the missing data grid, F(,) represents the concentration prediction value of the corresponding grid, represents the concentration monitoring value of the corresponding grid, abs represents taking the absolute value, and α and β represent the weight coefficients respectively.

6. The method according to claim 2, wherein Using the genetic algorithm to select the optimal input grid combination from the input grid set of the final model, which can meet the requirements of the model accuracy and has the least number of grids, includes: Construct an equal-length chromosome sequence structure according to the size of the input grid set of the final model, where each numerical bit in the sequence structure corresponds to each grid in the set, and the 1 / 0 value of the numerical bit represents whether to use the corresponding grid as the input grid of the neural network model; Generate the current population of the genetic algorithm according to the chromosome sequence structure; Prune the final model according to each chromosome in the current population, and use the finite sample set to train the pruned model three times; Judge whether the model after three trainings meets the optimal goal. If not, generate a new current population and return to the operation of model pruning until the model after the final three trainings meets the optimal goal; Wherein, the optimal goal is that the accuracy of the model after three trainings meets the requirements, and the model with the least number of input grids has been obtained in the three trainings carried out so far.

7. The method according to claim 6, characterized in that The generating of the new current population includes: Construct a fitness function according to the change form of the concentration prediction values between the input grids in the chromosome; Select excellent chromosomes from the mutated current population according to the fitness function to construct a new current population.

8. The method according to claim 7, characterized in that The constructing of the fitness function according to the change form of the concentration prediction values between the input grids in the chromosome includes: Construct the following fitness function: Where fit is the fitness value, num(G) represents the number of extreme value grids in the concentration distribution predicted by the model after three trainings for the input grids with a value of 1 in the chromosome, i3 and j3 are the two-dimensional indexes of the input grids with a value of 1 in the chromosome, F(,) represents the concentration prediction value of the corresponding grid, abs represents taking the absolute value, num(0) represents the number of grids with a value of 0 in the chromosome, and γ, δ, and ε respectively represent weight coefficients; the larger the fitness value, the better the chromosome.

9. The method according to claim 2, characterized in that, Using the optimal model corresponding to the optimal input grid combination to predict the concentration distribution in the water body includes: Only keep the inland water body greenhouse gas concentration automatic monitoring system of each target grid in the optimal input grid combination, and remove the inland water body greenhouse gas concentration automatic monitoring system of the remaining grids; Use the remaining system to monitor the gas concentration of each target grid in real time and input it into the optimal model to obtain the real-time concentration distribution of all grids.

10. The method according to claim 2, characterized in that, Includes: Perform operations S220 - S260 for each greenhouse gas; Wherein, the using of the optimal model corresponding to the optimal input grid combination to predict the concentration distribution in the water body includes: Take the union of the optimal input grid combinations corresponding to various greenhouse gases; Keep the inland water body greenhouse gas concentration automatic monitoring system of each target grid in the union, and remove the inland water body greenhouse gas concentration automatic monitoring system of the remaining grids; Use the remaining system to monitor the greenhouse gas concentration of each target grid in real time and input it into the optimal models corresponding to various greenhouse gases respectively to obtain the real-time greenhouse gas concentration distribution of all grids.

Images