An in-situ monitoring system and method for greenhouse gas fluxes at water-air interface

Abstract

The embodiment of the present application discloses a kind of water-gas interface greenhouse gas flux in situ monitoring system and monitoring method.Therein, system includes: sampling device module, for alternately collecting water gas and water surface gas;Weather sensor and temperature sensor are used for respectively collecting the air speed, air pressure, gas Kelvin temperature of monitoring point and water Celsius temperature;Optical sensor is used to cooperate with the sampling device module, and the content of CO2, CH4 and N2O in water and the content of CO2, CH4 and N2O in water surface air is alternately detected;Industrial computer module is used to control the sampling device module and each sensor operation;Calculation module is used to determine the greenhouse gas flux of water-gas interface according to the collected data.The embodiment can accurately monitor the water-gas interface flux of various greenhouse gases under actual conditions.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the technical field of gas flux monitoring, in particular to a water-gas interface greenhouse gas flux in-situ monitoring system and a monitoring method. BACKGROUND

[0002] The monitoring and estimation of long-term changes in water greenhouse gas fluxes not only have great significance for studying the biogeochemical cycle of carbon and nitrogen and global climate change, but also provide effective data support for predicting water carbon sequestration capacity on a longer time scale and promoting water environmental ecological governance.

[0003] Currently, the observation equipment for water greenhouse gas emission flux mainly uses water surface float box technology. It places a sealed flux box on the water surface, collects the greenhouse gases emitted by the surface water in a diffusion manner through the middle of the bottom of the flux box, measures the concentration of the measured gas in the flux box every certain period of time, and calculates the emission flux of the measured gas in the covered water area according to the concentration change rate over time. Both static float box method and dynamic float box method are difficult to distinguish the influence of wind field on the gas exchange rate of water-gas interface, especially for water areas with high water flow velocity, which is easy to cause gas flux monitoring error. For example, patent application CN221350771U discloses a greenhouse gas emission sampling device using static float box technology; patent application CN113358426A discloses a collection device and a determination method for determining the contribution of in-situ greenhouse gas emission in shallow water, which uses fixed sealed box technology; both of them collect the greenhouse gases emitted by the water, but the concentration gradient of the actual water-gas interface is greatly changed due to the continuous accumulation of water-released gas in the box, which leads to the change of Henry balance of gas-liquid exchange, especially when the water gas release is high, which is easy to cause the underestimation of gas exchange flux. Patent application CN114354290A discloses a device and method for dynamically measuring water greenhouse gas emission flux, which overcomes the defects of water greenhouse gas emission flux measurement technology represented by static float box technology, but the sealed float box space destroys the original water-gas interface balance mode, which will cause a huge error between the measured data and the actual situation. Patent application CN114689794A discloses a water area greenhouse gas flux monitoring device and monitoring method, which uses a disturbance assembly extending into the water to simulate the wind field signal to interfere with the water fluctuation within the range covered by the float box. However, this disturbance destroys the original water environment and water dynamics, and destroys the balance mode of dissolved gas in the water. The above methods only measure the greenhouse gas flux in the water-gas balance device, while the greenhouse gas flux in the environment is changing over time. These methods ignore the interference of the external environment (such as the artificial turbulence generated by the friction between the contact water and the water surface, and the influence of the water flow velocity, which will cause a huge error in the measured parameters of the water surface atmospheric pressure, saturated water vapor pressure, and balanced gas in the balance device), which has certain limitations.

[0004] Therefore, the existing observation equipment, technology and method cannot accurately observe the greenhouse gas flux change of the water-gas two-phase medium of the water body. SUMMARY

[0005] Embodiments of the present application provide a water-gas interface greenhouse gas flux in-situ monitoring system and a monitoring method to solve the above technical problems.

[0006] In a first aspect, embodiments of the present application provide a water-gas interface greenhouse gas flux in-situ monitoring system, comprising:

[0007] A sample introduction device module is configured to alternately collect water gas and water surface gas; wherein the sample introduction device module comprises a gas-liquid separator, a waterproof and air-permeable membrane in the gas-liquid separator divides the gas-liquid separator into a first part for storing liquid and a second part for separating gas; the first part is connected with a water body sampling pipe and a water inlet of a peristaltic pump; the second part is connected with a first gas collection vacuum pump, gas in the first gas collection vacuum pump enters a gas collector after passing through a gas booster valve and a gas one-way valve; the gas collector is connected with a second gas collection vacuum pump, a gas outlet of the second gas collection vacuum pump is connected with one end of a three-way electromagnetic valve, the other two ends of the three-way electromagnetic valve are respectively connected with an air inlet and a sample outlet, and the three-way electromagnetic valve realizes the alternate collection of water gas and water surface gas by switching a communication port;

[0008] A meteorological sensor and a temperature sensor are respectively configured to collect air speed, air pressure, gas Kelvin temperature and water Celsius temperature of a monitoring point;

[0009] An optical sensor is configured to cooperate with the sample introduction device module to alternately detect the content of CO2, CH4 and N2O in water and the content of CO2, CH4 and N2O in water surface air;

[0010] An industrial computer module is configured to control the operation of the sample introduction device module and the sensors;

[0011] A calculation module is configured to determine the greenhouse gas flux of the water-gas interface according to the collected data.

[0012] In a second aspect, embodiments of the present application provide a water-gas interface greenhouse gas flux in-situ monitoring method applied to the above water-gas interface greenhouse gas flux in-situ monitoring system.

[0013] The method comprises:

[0014] The sample introduction device module is floated and fixed at a water body point to be monitored, and a water outlet of the water body sampling pipe and the peristaltic pump is inserted into the water body;

[0015] The industrial computer module controls the three-way electromagnetic valve to connect the air inlet and the air outlet of the second gas collection vacuum pump, closes the sample outlet, and controls each gas collection vacuum pump to open to extract the gas in the closed pipeline to form a negative pressure vacuum;

[0016] After the extraction is completed, the industrial computer module controls the second gas collection vacuum pump to close and the peristaltic pump to open to collect the greenhouse gas in the water;

[0017] After the collection is completed, the industrial computer module controls the second gas collection vacuum pump to open in sequence, controls the three-way electromagnetic valve to connect the air outlet of the second gas collection vacuum pump and the sample outlet, closes the air inlet, and makes the greenhouse gas in the water enter the optical sensor for detection;

[0018] After the detection is completed, the industrial computer module controls the three-way electromagnetic valve to connect the air inlet and the sample outlet, closes the air outlet of the second gas collection vacuum pump, controls the second gas collection vacuum pump to close, and makes the greenhouse gas in the air on the water surface enter the optical sensor for detection;

[0019] The industrial computer module synchronously collects data of the optical sensor, the wind speed sensor, and the temperature sensor and transmits the data to a calculation module to determine the greenhouse gas flux of the water-air interface.

[0020] In a third aspect, an embodiment of the present application provides a water-air interface greenhouse gas flux distribution prediction method, comprising:

[0021] S210, dividing a water area plane to be monitored into a plurality of grids;

[0022] S220, taking each grid as a monitoring point, respectively installing the above water-air interface greenhouse gas flux in-situ monitoring system, and monitoring the flux of any one greenhouse gas of each grid in real time; and taking the flux of each grid at the same time as a sample to construct a limited sample set;

[0023] S230, averaging the flux distribution monitored in a period of time, taking a plurality of extreme grids and large gradient grids in the average flux distribution as a plurality of initial grids; and constructing a neural network model with the flux of the plurality of initial grids at the same time as input and the flux distribution of all grids at the same time as output;

[0024] S240, training and testing the neural network model by using the limited sample set, if the model accuracy does not meet the requirement, clustering the grids with a flux prediction value and a monitoring value difference exceeding a set threshold in the test result according to geographical positions, and taking each cluster center into the model input;

[0025] S250, return to S240 according to the new neural network model, and thus circulate until the model accuracy meets the requirement;

[0026] S260, using a genetic algorithm, selecting the optimal input grid combination from the input grid set of the final model, which can make the model accuracy meet the requirement and the grid number is the least, and using the optimal model corresponding to the optimal input grid combination to predict the water-gas interface flux distribution.

[0027] In summary, the embodiment of the present application discloses a water-gas interface greenhouse gas flux in-situ monitoring system and monitoring method, which can realize the following beneficial effects:

[0028] 1. By the specific sampling device module, the water gas and the water surface gas are alternately collected, and in cooperation with the optical sensor, the content of the two greenhouse gases, methane and nitrous oxide, in the water and air is alternately detected, so that comprehensive data support is provided for long-time monitoring of the greenhouse gas flux of inland water bodies;

[0029] 2. When calculating the gas flux, by introducing the correction coefficient of the ideal gas state equation, the influence of the actual gas pressure and temperature of the water body interface on the gas flux is considered, so that more accurate gas flux data is obtained;

[0030] 3. The whole system does not collect water or gas samples, does not change the original water environment, and does not use expensive and complex equipment, so that the gas-liquid exchange balance mode of the water body itself is maintained, and the gas flux data is further ensured to be consistent with the actual situation;

[0031] 4. The flux distribution prediction method based on the water-gas interface greenhouse gas flux in-situ monitoring system, by combining partial point monitoring and overall water area prediction, using as few water-gas interface greenhouse gas flux in-situ monitoring devices as possible, the greenhouse gas flux distribution of the whole 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 the greenhouse gas flux is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0033] Figure 1 is a schematic diagram of a water-gas interface greenhouse gas flux in-situ monitoring system provided by the embodiment of the present application;

[0034] Figure 2is a mounting schematic diagram of an in-situ monitoring device provided by an embodiment of the present application;

[0035] Figure 3 is a structural schematic diagram of an in-situ monitoring device provided by an embodiment of the present application;

[0036] Figure 4 is a flowchart of a water-gas interface greenhouse gas flux distribution prediction method provided by an embodiment of the present application;

[0037] Figure 5 is a grid division schematic diagram of a water area to be monitored provided by an embodiment of the present application; DETAILED DESCRIPTION

[0038] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0039] In the description of the present application, it should be noted that the orientations or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first", "second", "third" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0040] In the description of the present application, it should be noted that, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0041] Figure 1 is a schematic diagram of a water-gas interface greenhouse gas flux in-situ monitoring system provided by an embodiment of the present application. As shown in the figure, the system comprises an in-situ monitoring device and a calculation module. Figure 1

[0042] Figure 2 is a mounting schematic diagram of an in-situ monitoring device provided by an embodiment of the present application.​Figure 2 The in-situ monitoring device mainly comprises a floating box, a power supply device module, a sensor combination, a sample introduction device module and an industrial computer module.

[0043] The power supply device module mainly provides power energy for the in-situ monitoring device through the photovoltaic panel on the top and the battery under the panel, and the warning device is used to prevent the collision of water vehicles.

[0044] The sample introduction device module is used to alternately collect the gas in the water and the gas on the surface of the water body. Figure 3 The sample introduction device module comprises a gas-liquid separator, a waterproof and air-permeable membrane is arranged in the gas-liquid separator, and the waterproof and air-permeable membrane divides the gas-liquid separator into a first part for storing liquid and a second part for separating gas. The first part is connected with a water body sampling pipe and a water inlet of a peristaltic pump, and the end of the water body sampling pipe is provided with a filter funnel; the second part is connected with a gas inlet of a first gas collection vacuum pump, a gas outlet of the first gas collection vacuum pump is connected with a gas booster valve and a gas one-way valve, and the gas output from the gas outlet enters a gas collector after passing through the gas booster valve and the gas one-way valve; the gas collector is connected with a gas inlet of a second gas collection vacuum pump, a gas outlet of the second gas collection vacuum pump is connected with one end of a three-way electromagnetic valve, the other two ends of the three-way electromagnetic valve are respectively connected with an air inlet and a sample outlet, and the three-way electromagnetic valve realizes the alternate collection of the gas in the water body and the gas on the surface of the water body by switching the communication ports.

[0045] The sensor combination comprises a meteorological sensor and a temperature sensor. Figure 2 The meteorological sensor is used to collect the air speed, air pressure and gas Kelvin temperature of the monitoring point, and the temperature sensor is used to collect the water body Celsius temperature.

[0046] The optical sensor is used to alternately detect the content of CO2, CH4 and N2O in the water and the content of CO2, CH4 and N2O in the air on the surface of the water body .

[0047] The industrial computer module is used to control the operation of the sample introduction device module and each sensor, and the calculation module is used to determine the greenhouse gas flux of the water-gas interface according to the collected data.

[0048] In a specific embodiment, before the system is used, as shown in Figure 2 , the floating box can be fixed at the position of the water body point to be monitored by using a reinforced concrete anchor point in the field, and the temperature sensor probe of the filter funnel and the water body is adjusted to be located at a position 5-10 cm deep below the water surface by fixing the floating ball and the cable, so as to obtain the physical and chemical parameters of the surface water body. The transmission signal of the temperature sensor (model PT100) The transmission signal of the meteorological sensor (model Lufft WS600) for monitoring the water temperature (unit, °C) of the point The transmission signal of the optical sensor (model Picarro G2508 or Gasboard-3000GHG) for monitoring the air speed (unit, m / s), air pressure (unit, Pa), and gas Kelvin temperature (unit, °K) of the point And The CO2, CH4, and N2O content (unit, ppm) in the surface water body and the surface air of the water body, respectively.

[0049] Further, the filter funnel in the sampling device module is to prevent the pipeline from being blocked by impurities in the water. The gas-liquid separator and the waterproof air permeable membrane are to separate the gas in the water body. The outlet pipeline of the peristaltic pump extends below the water surface to maintain the original water environment state of the water body. The first and second gas collection vacuum pumps collect and transport gas by forming a vacuum negative pressure under the drive of the power supply device. The gas booster valve is to avoid the small amount of gas affecting the collection of the gas collector, and the one-way gas valve is to prevent backflow of the gas collected by the gas collector. The three ports of the three-way electromagnetic valve are respectively connected to the water gas transported by the gas outlet of the second gas collection vacuum pump, the surface air of the water body transported by the air inlet, and the sample outlet connected to the optical sensor. By controlling the communication of the three ports, the water gas or the surface air of the water body can be controlled to input the optical sensor. The function of the desiccator 6 is to prevent water vapor from entering the air inlet of the optical sensor and affecting the test data quality of the greenhouse gas.

[0050] During the use of the system, the industrial computer model controls the operation of the sampling device module and the optical sensor in the following way:

[0051] First, ensure that the filter funnel and the outlet of the peristaltic pump are 5-10 cm below the water surface.

[0052] Then, control the three-way electromagnetic valve to make the outlet channel of the second gas collection vacuum pump communicate with the air inlet channel, and the sample outlet channel is closed, while the first and second gas collection vacuum pumps are turned on to form a negative pressure vacuum by extracting the existing gas in the closed pipeline.

[0053] After the extraction is completed, the second gas collection vacuum pump is turned off, and the peristaltic pump is turned on to start collecting the greenhouse gas in the surface water body. When the pressure gauge on the second gas collection vacuum pump shows that the gas pressure in the gas collector is close to atmospheric pressure, the optical sensor and the second gas collection vacuum pump are turned on in turn, and the three-way electromagnetic valve is controlled to make the outlet channel of the second gas collection vacuum pump communicate with the sample outlet channel, and the air inlet channel is closed, so that the greenhouse gas in the surface water body enters the optical sensor to generate stable test data.

[0054] Then, the three-way electromagnetic valve is controlled to make the air inlet channel communicate with the sample outlet channel, the outlet channel of the second gas collection vacuum pump is closed, and the second gas collection vacuum pump is closed at the same time, so that the greenhouse gas in the surface layer of the water surface enters the optical sensor to generate stable test data.

[0055] At the same time, the analysis test data of the optical sensor, the meteorological sensor data of the water surface, and the temperature sensor data of the surface water body are collected in real time, the data are stored and transmitted to the computing module through the signal transmitter. The above control flow steps are automatically repeated to realize long-time sequence continuous in-situ monitoring of water-gas interface greenhouse gas flux.

[0056] Further, after the computing module (such as a user computer end) receives the real-time in-situ monitoring data sent by the communication base station, the flux of the water-gas interface greenhouse gas per unit time per unit area of the water body is determined by the following method:

[0057] S110, calculating the solubility of CO2, CH4 and N2O in the surface water (unit, mole·L -1 ·atm -1 ), and the water-gas exchange rate (unit, cm·h -1 ).

[0058] Optionally, the data collected by the meteorological sensor and the temperature sensor are calculated to obtain the change amount of the water-gas interface CO2, CH4 and N2O flux caused by different physical and chemical actions, and the calculation formula is:

[0059]

[0060] In the formula, is the Kelvin temperature (unit, °K), which is calculated by ; s is the salinity of the water body (unit, ‰), which is 0 for inland water bodies, and for oceans or highland salt lakes, a sensor for monitoring salt in the field can be added to collect the salinity signal value s; is the wind speed of the water surface, .

[0061] S120, calculating the water-gas interface greenhouse gas flux per unit area per unit time of the water area at the monitoring point (unit, mg·m -2 ·h -1 ):

[0062] Optionally, the data collected by the optical sensor can be calculated The water-gas interface CO2, CH4 and N2O fluxes of the monitoring water area are calculated by calculating the water-gas interface CO2, CH4 and N2O fluxes respectively, and the calculation formula is:

[0063]

[0064] wherein, is a correction coefficient of the ideal gas state equation, which is used for correcting the ideal gas equation to the actual gas equation, so as to obtain the accurate gas flux of the water body interface under the actual pressure and actual temperature.

[0065] In summary, the embodiment provides a water-gas interface greenhouse gas flux in-situ monitoring system, which can achieve the following beneficial effects:

[0066] 1. The specific sampling device module is used for alternately collecting water gas and water surface gas, and cooperates with the optical sensor to alternately detect the contents of methane and nitrous oxide, two kinds of greenhouse gases in water and air, so as to provide comprehensive data support for the monitoring of inland water greenhouse gas flux;

[0067] 2. In the calculation of gas flux, the correction coefficient of the ideal gas state equation is introduced to consider the influence of the actual pressure and temperature of the water body interface on the gas flux, so that more accurate gas flux data are obtained;

[0068] 3. The whole system does not collect water or gas samples, does not change the original water environment condition, and does not use expensive and complex equipment, so that the gas-liquid exchange balance mode of the water body itself is maintained, and the gas flux data is consistent with the actual situation.

[0069] Based on the above water-gas interface greenhouse gas flux in-situ monitoring system, Figure 4 It is a flow chart of a water-gas interface greenhouse gas flux distribution prediction method, which can predict the greenhouse gas flux distribution of the whole water area by as few monitoring points as possible, and reduce the number of in-service equipment in long-term monitoring. As shown in Figure 4 The method specifically includes:

[0070] S210, dividing the plane of the water area to be monitored into a plurality of grids.

[0071] Figure 4 A regular rectangle is taken as an example to show a planar top view of a piece of water area to be monitored, as shown in the figure, the plane can be divided into a plurality of grids for partition monitoring. For the plane of the water area of irregular shape, the division method is similar.

[0072] S220, respectively install the above water-gas interface greenhouse gas flux in-situ monitoring system as the monitoring point of each grid, and monitor the flux of any greenhouse gas at the water-gas interface of each grid in real time; and take the flux of each grid at the same time as a sample to construct a limited sample set.

[0073] In combination Figure 2 , in each grid, install an in-situ monitoring device in the manner shown in Figure 2 , and transmit the collected sensor data in real time. The in-situ monitoring devices can share a computing module to calculate the greenhouse gas flux of each grid at each time in real time.

[0074] After a period of monitoring, for any greenhouse gas, take the flux of the greenhouse gas of each grid at the same time in the monitoring data as a sample, and multiple samples can be obtained from the data at multiple historical times to form a limited sample set. The finiteness of the sample set is emphasized here because the cost of installing an in-situ monitoring device for each grid in a water area with a large area is very high, so this embodiment can only realize sample collection for a limited time length while ensuring that all grids have devices. In the subsequent steps, a prediction model of the greenhouse gas flux distribution will be trained using the limited samples to predict the flux distribution of the entire water area through the monitored flux of part of the grids, thereby reducing the number of devices in service for a long time and reducing the flux monitoring cost of the entire water area.

[0075] S230, average the flux distribution monitored in a period of time, take multiple extreme grids and large gradient grids in the average flux distribution as multiple initial grids; and construct a neural network model with the flux of the multiple initial grids at the same time as input and the flux distribution of all grids at the same time as output.

[0076] The time period can be selected as a period of time with relatively stable wind speed and temperature, the flux distribution at each time in the period of time is taken as a two-dimensional matrix arranged according to the geographical position of the grid (in an irregular water area, the grid element corresponding to the land position is permanently set to 0), and the two-dimensional matrices at multiple times are added and then divided by the number of times in the period of time to obtain the average flux distribution in the period of time. From the average flux distribution, select the grids where the extreme points are located and the grids with a change gradient exceeding a set threshold to form an initial set of neural network model input grids. For ease of description, the grids in these initial sets are referred to as initial grids in this embodiment. These grids are selected because these grids often represent key positions where the flux distribution of the entire water area changes or mutates, and can provide key information for predicting the flux distribution of the entire water area in the later stage.

[0077] Based on the initial set, a neural network model is constructed. In this embodiment, the model will be trained so that after inputting the initial grid greenhouse gas flux at the same time, it can output the greenhouse gas flux distribution of all grids at that time. Optionally, the model can use a multi-layer convolutional neural network, a convolutional neural network connected to a fully connected layer for generating a specific dimension result; 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 7, and the convolution kernel can use 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.

[0078] S240, training and testing the neural network model using the limited sample set, if the model accuracy does not meet the requirements, clustering the grids with flux prediction values and monitoring value differences exceeding the set threshold in the test results according to geographical location, and including each cluster center into the model input.

[0079] As described above, the model needs to learn the data relationship between part of the grid flux and all the grid flux. Due to the existence of a large amount of missing grid data, the learning difficulty of the model is very large. In the case of limited sample quantity, the accuracy of the trained model may not meet the ideal requirements. Therefore, in the case of poor model accuracy, this embodiment gradually reduces the learning difficulty by increasing the number of input grids to provide more data information for the model.

[0080] In a specific embodiment, in order to make full use of limited sample data, model training can be completed through the following steps:

[0081] Step one, divide the limited 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 and distinguish the subsequent two training stages, and the samples in the two training sets can be the same, different or partially the same, and this embodiment does not make specific restrictions.

[0082] Step two, use the first training set to train the neural network model once. In one training, the model parameters are updated by minimizing the flux change of adjacent grids. This step corresponds to the first training stage of the model. In this stage, the model parameters are updated by using the rules followed by the input data itself, so that the model output has a certain rule. As can be seen from the above system embodiment, the gas concentration, temperature environment and other parameters used to calculate the flux are continuously changed in space, so the change of greenhouse gas flux along the water-air interface should also be relatively uniform. Therefore, this embodiment updates the model parameters by constraining the flux change of adjacent grids to be as small as possible to reduce the random changes as much as possible. Optionally, the following loss function can be constructed:

[0083]

[0084] wherein, represents the loss function value, and represent the grid index along the horizontal and vertical direction respectively, represents the flux prediction value of the grid at the row and the column, represents the summation over all grids of the water-air interface (excluding the land grids with permanent flux of 0), represents taking the absolute value. Through the minimization of the loss function, the model parameters are updated.

[0085] Step three, the once trained model is retrained using the second training set. In the retraining, the model parameters are updated by minimizing the difference between the flux prediction value of each grid and the monitoring value. This step corresponds to the second training phase of the model. In this phase, the actual monitoring flux data is used as the true value, and the once trained model is supervised learning by constraining the model output to continuously approach the true value. Optionally, the following loss function can be constructed:

[0086]

[0087] wherein, represents the loss function value, represents the flux true value of the grid at the row and the column.

[0088] Further, the present embodiment also considers the case where the true value data is empty due to equipment failure or communication failure, etc. In this case, although the sample data is missing part of the grid data, it can still be used for the retraining of the model. Only the minimization of the flux change of the part of the grid and the adjacent grid, and the minimization of the difference between the flux prediction value and the monitoring value of the other grid are required. Optionally, for this kind of sample, the following loss function can be constructed:

[0089]

[0090] wherein, represents the loss function value, represents the two-dimensional index of the data complete grid represents the two-dimensional index of the data missing grid, represents the flux prediction value of the corresponding grid, represents the flux monitoring value of the corresponding grid, respectively represent the weight coefficient. In this way, for the sample with complete data, the network parameters can be updated using and for the sample with missing data, the network parameters can be updated using The network parameters are updated, and waste of sample data is further avoided.

[0091] The two training stages first enable the model to learn the rules 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 the embodiment can accelerate the convergence speed of the model and achieve the best training effect with a limited sample set.

[0092] Step four, after the training is completed, the model after the secondary training is tested by using the test set. If the model accuracy still does not meet the requirements, the grids whose flux prediction value and monitoring value difference exceeds the set threshold in the test results are clustered according to the geographical position, and each cluster center is included in the model input. This step selects the grid with large prediction error from the prediction result of each test sample as the input grid of the model for the case that the model accuracy does not reach a good level after two training stages.

[0093] Optionally, if some grids have large errors in multiple test samples, these grids are clustered according to the geographical position to obtain multiple spatial clusters. Each spatial cluster represents a region with close distance, and the grids in the region have large measurement errors multiple times, which indicates that the model cannot effectively learn the data rules of the region from the existing input grids, so it is necessary to select one or more grids in the region as the model input to provide effective information of the region for the model. Optionally, the DBSCAN clustering method can be used, and finally the cluster center of each cluster is selected as the model input.

[0094] Further, when a new model input is added, a convolution kernel connected with the new network and the subsequent calculation layer can be added in the first layer of the convolution layer based on the original model; or a fully connected layer can be added between the input layer and the first layer of the convolution layer to transform the dimension of the new input layer to the dimension of the original hidden layer; the remaining structure in the model remains unchanged, and the original parameters are inherited as initial parameters.

[0095] S250, return to S240 according to the new neural network model, and the cycle is repeated until the model accuracy meets the requirements.

[0096] Specifically, the new neural network model is trained and tested again using the limited sample set; if the accuracy of the new model still does not meet the requirements, the grid clusters with flux prediction values and monitoring values that differ by more than a set threshold are clustered by geographical location, and each cluster center is included in the model input to obtain an updated neural network, and the process returns to S240. This cycle is repeated until the final model accuracy meets the required indicators. At this time, the neural network model has sufficient effective grid data, and according to the variation law of these data and the flux itself, the flux distribution of the entire water area plane can be accurately predicted. In order to facilitate differentiation and description, the neural network model obtained at this time is referred to as a full-amount model.

[0097] S260, using a genetic algorithm to select an optimal input grid combination from the input grid set of the full-amount model that can make the model accuracy meet the requirements and has the least number of grids, and using the optimal model corresponding to the optimal input grid combination to predict the water-gas interface flux distribution.

[0098] As described above, in the multiple cycles of S240-S250, the number of input grids is increased to improve the model accuracy. In order to prevent redundancy in the final input grid, this step uses a genetic algorithm to filter the input grid of the full-amount model to remove redundant grids, and to achieve flux distribution prediction of the entire water area plane with the least model input and in-situ greenhouse gas monitoring system.

[0099] In a specific implementation, the above filtering process can be completed in the following way:

[0100] Step one, according to the number of grids in the input grid set of the full-amount model, construct an equal-length chromosome sequence structure. For example, if the number of input grids of the full-amount model is 15, a vector with a length of 15 can be constructed as the chromosome sequence structure in the genetic algorithm, wherein each bit in the vector corresponds to a grid in the input grid set; a numerical bit takes 1, representing retaining the corresponding grid as the input of the neural network model, and taking 0 represents deleting the corresponding grid from the input grid of the neural network model.

[0101] Step two, generate the current population of the genetic algorithm according to the chromosome sequence structure. For example, a random initialization method can be used to generate multiple chromosomes to form the current population.

[0102] Step three, pruning the full model according to each chromosome in the current population, and training the pruned model three times using the limited sample set. Specifically, for a certain chromosome, the grid corresponding to the bit with a value of 0 in the chromosome is deleted from the input grid of the full model, and the branches connecting the first hidden layer of the model to the input grid are all deleted, obtaining a new model; then, the new model is trained using the limited sample set, which is referred to as three-time training in this embodiment.

[0103] Step four, judging whether the model after three-time training meets the optimal target, if not, selecting, crossing and mutating the current population according to the genetic algorithm, selecting excellent chromosomes from the mutated current population to form a new current population, and returning to step three to prune the full model according to each chromosome in the new current population and train the pruned model three times using the limited sample set. This cycle is iterated until the model after the final three-time training meets the optimal target. The optimal target is that the model after three-time training meets the accuracy requirement and has the least number of input grids among the models obtained in the current cumulative three-time training.

[0104] Optionally, after obtaining a chromosome meeting the model accuracy requirement each time, the chromosome with the least number of input grids is selected from all the chromosomes meeting the model accuracy requirement obtained in the current cumulative training as the optimal chromosome; if the optimal chromosome remains unchanged after continuous multiple three-time training, it is considered that the model with the least number of input grids has been obtained. Of course, the iteration can also be directly iterated to the maximum iteration number, and the model with the least number of input grids and meeting the accuracy requirement is selected as the optimal model after iteration termination.

[0105] Further, when selecting excellent chromosomes, an fitness function can be constructed according to the change pattern of the predicted values of the air flux among the input grids in the chromosome; the excellent chromosomes are selected from the mutated current population according to the fitness function to construct a new current population. Optionally, the following fitness function can be constructed:

[0106]

[0107] wherein, is the fitness value, represents the number of extreme grids in the predicted flux distribution of the input grid with a value of 1 in the current chromosome after three-time training of the model, is the two-dimensional index of the input grid with a value of 1 in the current chromosome, represents the predicted value of the flux of the corresponding grid, represents the number of grids with a value of 0 in the current chromosome, respectively, are weight coefficients. The larger the fitness value is, the better the chromosome is.

[0108] By the above fitness function, the chromosomes with less input grid number and covering more extreme value grids and large gradient grids (which represent the key shape change of flux distribution) can be selected preferentially to evolve new current population, which helps to obtain the optimal chromosome and optimal model as soon as possible.

[0109] Finally, after obtaining the optimal input grid combination, only the water-air interface greenhouse gas flux in-situ monitoring system of each grid in the combination is retained, and the water-air interface greenhouse gas flux in-situ monitoring system of the remaining grids is removed; the air flux of each target grid is monitored in real time by using the remaining system, and the real-time flux distribution of all grids can be obtained by inputting the optimal model. The removed equipment system can be used for flux monitoring and prediction of other water areas, improving the utilization rate of equipment and reducing the monitoring cost of greenhouse gas flux.

[0110] Further, the above method takes a kind of greenhouse gas as an example to illustrate the specific method of predicting the flux distribution of the entire water area by using as few monitoring points as possible. In actual application, the operations of S220-S260 can be performed for each type of greenhouse gas; in S260, when predicting the flux distribution at the water-air interface by using the optimal model corresponding to the optimal input grid combination, the optimal input grid combinations corresponding to various greenhouse gases are taken as a set, the water-air interface greenhouse gas flux in-situ monitoring system of each target grid in the set is retained, and the water-air interface greenhouse gas flux in-situ monitoring system of the remaining grids is removed; the greenhouse gas flux of each target grid is monitored in real time by using the remaining system, and the optimal model corresponding to each type of greenhouse gas is inputted respectively to obtain the real-time greenhouse gas flux distribution of all grids. In this way, the flux distribution of three types of greenhouse gases in the entire water area can be predicted by using as few monitoring points as possible.

[0111] It should be noted that, in the above water-air interface greenhouse gas flux distribution prediction method, the steps of installing and removing the water-air interface greenhouse gas flux in-situ monitoring system can be automatically executed by an electronic device.

[0112] In summary, the embodiment provides a water-air interface greenhouse gas flux distribution prediction method, which can achieve the following beneficial effects:

[0113] 1. By combining partial point monitoring and overall water area prediction, the flux distribution of the entire water area can be obtained by using as few water-air interface greenhouse gas flux in-situ monitoring devices as possible, the number of equipment used for long-term monitoring is reduced, the utilization rate of equipment is improved, and the monitoring cost of greenhouse gas flux is reduced.

[0114] 2. Since the neural network model needs to learn the complex flux distribution law according to the partial grid, the learning difficulty is large and the sample quantity is limited, and the model accuracy after training is difficult to achieve ideal requirements at one time. In view of this difficulty, in the case of poor model accuracy, the embodiment selects representative grids from the area with large prediction error as model input, and provides more effective information for the model by increasing the number of input grids, and gradually reduces the learning difficulty;

[0115] 3. In view of the limited sample of the full grid data, the embodiment trains the model in multiple stages to learn the law of the input data itself first, and then carries out one-to-one supervised training through the monitoring value of each grid. Compared with directly using the monitoring data for supervised training, this method can accelerate the convergence speed of the model and achieve the best training effect with limited sample set;

[0116] 4. In view of the data missing sample caused by partial grid device or communication failure, the embodiment fuses the uniform change of flux and the monitoring value, adopts different loss value calculation methods for different grids, fully utilizes the data information of the remaining samples, and further avoids sample waste;

[0117] 5. After obtaining the full amount model including sufficient effective input grids, in order to prevent the existence of redundancy in the input grids of the full amount model, the embodiment uses genetic algorithm to screen the input grids and remove the redundant grids, so as to realize the prediction of the flux distribution of the entire water area plane with the least model input and greenhouse gas flux in-situ monitoring system. In the case of large water area and prediction of flux distribution of each greenhouse gas, the final equipment quantity can be greatly reduced, and the monitoring cost can be reduced.

[0118] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solution deviate from the technical solutions of the embodiments of the present application.

Claims

1. A water-air interface greenhouse gas flux distribution prediction method, comprising: S210, dividing the water area to be monitored into a plurality of grids; S220, installing a water-air interface greenhouse gas flux in-situ monitoring system at each grid as a monitoring point to monitor the flux of any greenhouse gas at the water-air interface of each grid in real time; and taking the flux of each grid at the same time as a sample to construct a limited sample set; S230, averaging the flux distribution monitored in a period of time, and taking a plurality of extreme value grids and large gradient grids in the average flux distribution as a plurality of initial grids; and constructing a neural network model taking the flux of the plurality of initial grids at the same time as input and the flux distribution of all grids at the same time as output; the plurality of extreme value grids and large gradient grids represent key positions where the flux distribution of the entire water area changes or mutates; S240, dividing the limited sample set into a first training set, a second training set and a test set; training the neural network model once using the first training set, updating the model parameters in the once training by minimizing the flux change of adjacent grids; training the model after the once training twice using the second training set, updating the model parameters in the twice training by minimizing the difference between the flux prediction value and the monitoring value of each grid; testing the model after the twice training using the test set; if the model accuracy does not meet the requirements, clustering the grids whose flux prediction value and monitoring value difference exceeds the set threshold in the test results according to the geographical position, and taking each cluster center into the model input; each spatial cluster represents an area; S250, returning to S240 according to the new neural network model, and repeating the cycle until the model accuracy meets the requirements; S260, selecting the optimal input grid combination that can make the model accuracy meet the requirements and has the least number of grids from the input grid set of the final model using a genetic algorithm, and predicting the water-air interface flux distribution using the optimal model corresponding to the optimal input grid combination; The in-situ monitoring system comprises: a sample introduction device module that alternately collects water gas and water surface gas; a weather sensor and a temperature sensor that collect air speed, air pressure, gas Kelvin temperature and water Celsius temperature of the monitoring point, respectively; an optical sensor that cooperates with the sample introduction device module to alternately detect the content of CO2, CH4 and N2O in water and the content of CO2, CH4 and N2O in water surface air; an industrial computer module that controls the operation of the sample introduction device module and the sensors; a calculation module that determines the greenhouse gas flux at the water-air interface according to the collected data.

2. The prediction method of claim 1, wherein, The twice training of the model after the once training using the second training set, updating the model parameters by minimizing the difference between the flux prediction value and the monitoring value of each grid in the twice training, comprises: If there are some grids missing monitoring data in the samples in the second training set, the model parameters are updated by minimizing the flux change of the some grids and adjacent grids and minimizing the difference between the flux prediction value and the monitoring value of other grids in the twice training.

3. The prediction method of claim 1, wherein, The optimal input grid combination capable of satisfying the model precision requirement and having the least number of grids is selected from the input grid set of the final model by using a genetic algorithm, and the method comprises the following steps: According to the size of the input grid set of the final model, an equal-length chromosome sequence structure is constructed, wherein 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 the corresponding grid is used as the input grid of the neural network model or not; According to the chromosome sequence structure, a current population of the genetic algorithm is generated; The final model is pruned according to each chromosome in the current population, and the pruned model is trained three times by using the limited sample set; It is judged whether the model after three times of training meets the optimal target or not, and if not, a new current population is generated according to the genetic algorithm, and the model pruning operation is returned until the final model after three times of training meets the optimal target; The optimal target is that the model after three times of training meets the precision requirement and has the least number of input grids in the three times of training.

4. The prediction method of claim 3, wherein, The new current population is generated according to the genetic algorithm, and the method comprises the following steps: According to the change pattern of the flux prediction value between each input grid in the chromosome, an adaptability function is constructed; According to the adaptability function, excellent chromosomes are selected from the mutated current population to construct a new current population; The adaptability function meets the following requirements: the chromosomes with the least number of input grids and the most extreme grids and large gradient grids covered by the input grids are preferentially selected to evolve the new current population, which helps to quickly obtain the optimal chromosome and the optimal model.

5. The prediction method of claim 1, wherein, The optimal model corresponding to the optimal input grid combination is used to predict the water-air interface flux distribution, and the method comprises the following steps: Only the water-air interface greenhouse gas flux in-situ monitoring system of each target grid in the optimal input grid combination is retained, and the water-air interface greenhouse gas flux in-situ monitoring systems of the remaining grids are removed; The gas flux of each target grid is monitored in real time by using the remaining systems, and the optimal model is input to obtain the real-time flux distribution of all grids.

6. The prediction method of claim 1, wherein, The sampling device module comprises a gas-liquid separator, a waterproof and air-permeable membrane in the gas-liquid separator divides the gas-liquid separator into a first part for storing liquid and a second part for separating gas; the first part is connected with a water body sampling pipe and a water inlet of a peristaltic pump; the second part is connected with a first gas collection vacuum pump, the gas in the first gas collection vacuum pump passes through a gas booster valve and a gas one-way valve and then enters a gas collector; the gas collector is connected with a second gas collection vacuum pump, a gas outlet of the second gas collection vacuum pump is connected with one end of a three-way electromagnetic valve, the other two ends of the three-way electromagnetic valve are respectively connected with an air inlet and a sample outlet, and the three-way electromagnetic valve realizes the alternate collection of the gas in water and the gas on the surface of the water body by switching the communication ports; Each gas collection vacuum pump is used to form a vacuum negative pressure under the drive of electricity to drive gas collection; The gas booster valve is used to boost the gas to avoid the influence of the small amount of gas on the collection of the gas collector. The gas one-way valve is used to realize one-way flow of the gas, and prevent backflow of the gas collected by the gas collector.

7. The prediction method of claim 6, wherein, The method for using each water-gas interface greenhouse gas flux in-situ monitoring system comprises: Floatingly fixing the sample device module at a water body point to be monitored, inserting the water body sample pipe and the water outlet of the peristaltic pump into the water body; The industrial computer module controls the three-way electromagnetic valve to connect the gas outlet of the second gas collection vacuum pump and the air inlet, close the sample outlet, and controls each gas collection vacuum pump to be opened to extract the gas in the closed pipeline to form a negative pressure vacuum; After the extraction is completed, the industrial computer module controls the second gas collection vacuum pump to be closed and the peristaltic pump to be opened to collect the greenhouse gas in the water; After the collection is completed, the industrial computer module controls the second gas collection vacuum pump to be sequentially opened, controls the three-way electromagnetic valve to connect the gas outlet of the second gas collection vacuum pump and the sample outlet, closes the air inlet, and makes the greenhouse gas in the water enter the optical sensor for detection; After the detection is completed, the industrial computer module controls the three-way electromagnetic valve to connect the air inlet and the sample outlet, closes the gas outlet of the second gas collection vacuum pump, controls the second gas collection vacuum pump to be closed, and makes the greenhouse gas in the air on the surface of the water body enter the optical sensor for detection; The industrial computer module synchronously collects data of the optical sensor, the wind speed sensor and the temperature sensor and transmits the data to a calculation module, so that the calculation module determines the greenhouse gas flux of the water-gas interface.

8. The prediction method of claim 7, wherein, The calculation module determines the greenhouse gas flux of the water-gas interface by the following way: S110, calculating the solubility of CO2, CH4 and N2O in water according to the air wind speed and the water temperature in Celsius 、 and , and the water-gas exchange rate 、 and ; S120, calculate the water-gas interface greenhouse gas flux of the monitoring point water area per unit area per unit time according to the following formula 、 and : wherein , and are the contents of CO2, CH4 and N2O in water, , and are the contents of CO2, CH4 and N2O in the air above the water surface, is the gas temperature in Kelvin, is the air pressure; is a correction factor for the ideal gas state equation for correcting the ideal gas equation to the actual gas equation.

Images