A groundwater level prediction method and system based on multi-source data

Abstract

The application provides a groundwater level prediction method and system based on multi-source data, and relates to the technical field of groundwater level prediction.The application integrates meteorological, hydrological, geological and human activity data, constructs three types of key indexes, namely natural recharge parameters, natural loss parameters and artificial disturbance parameters, corrects the parameters based on seasonal, historical data and other data, and establishes a water level prediction model by using a deep learning network.The method realizes high-precision and strong self-adaptability of groundwater level dynamic prediction by collecting data such as precipitation intensity, river level, evaporation intensity, groundwater runoff index and exploitation amount, and significantly improves the prediction accuracy under the coupling action of complex nature and human.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of groundwater level prediction, in particular to a groundwater level prediction method and system based on multi-source data. BACKGROUND

[0002] With the increasing demand for water resource management, groundwater level prediction is of great significance in agricultural irrigation, urban water supply, geological disaster prevention and other fields. Traditional prediction methods rely on single hydrological or meteorological data and do not fully consider the comprehensive influence of multi-source data, resulting in limited prediction accuracy and poor adaptability. Especially in areas with frequent human activities, the groundwater system is coupled by natural and human factors, and traditional models often cannot accurately reflect its dynamic changes, limiting the actual application effect.

[0003] In the prior art, the publication number CN117575085A discloses a tidal flat groundwater level prediction method, system, device and storage medium based on neural network, which collects tidal flat groundwater level time series data and corresponding period tidal flat precipitation; the tidal flat groundwater level time series data and the corresponding period tidal flat precipitation are divided into training set and validation set; the tidal flat groundwater level time series data in the training set is standardized; the standardized tidal flat groundwater level time series data in the training set and the corresponding period tidal flat precipitation are subjected to high-dimensional position coding; the processed training set is used to train the pre-constructed CsL neural network model, the validation set is used to verify the performance of the trained CsL neural network model, and the final available CsL neural network model is obtained; the current tidal flat precipitation and the current tidal flat groundwater level time series data are input into the final available CsL neural network model to obtain the tidal flat groundwater level prediction result at the next time.

[0004] The main problem of the above-mentioned scheme is that the dynamic change of the tidal flat groundwater level is a complex process affected by multiple factors, the rise or fall of the groundwater level is caused by various actual activities, and the effects of different activities are different, which cannot be effectively described by precipitation and historical water level, resulting in inaccurate prediction value and difficulty in reflecting actual data.

[0005] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a groundwater level prediction method and system based on multi-source data to solve the problems raised in the background.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0008] A groundwater level prediction method based on multi-source data, the specific steps comprising:

[0009] Step 1: Collecting water level data, meteorological data, hydrological data, geological data, remote sensing data and human activity data of the target area in the past time period;

[0010] Step 2: Calculate the area proportion of different types of land in the target area, generate the natural recharge parameter of the target area based on meteorological data, hydrological data and the area proportion of different types of land; Calculate the vegetation water consumption evaluation index based on remote sensing data, and then calculate the evaporation loss index of groundwater, and combine the evaporation loss index with the geological data to generate the natural loss parameter;

[0011] Step 3: Assign the impervious area proportion weight to different types of land in the target area based on the water permeability, generate the regional comprehensive impervious area proportion based on the area proportion of different types of land and the impervious area proportion weight, and generate the artificial disturbance parameter based on the human activity data and the regional comprehensive impervious area proportion;

[0012] Step 4: Correct the natural recharge parameter based on the average value of meteorological data and hydrological data, and correct the artificial disturbance parameter based on the cumulative amount of artificial exploitation;

[0013] Step 5: Construct a deep learning network, take the natural loss parameter and the corrected natural recharge parameter and artificial disturbance parameter in the same time period in the historical data as the input, and take the groundwater level in the same time period as the label, train the water level prediction model;

[0014] Step 6: Input the natural loss parameter, the corrected natural recharge parameter and the artificial disturbance parameter of the current time period into the water level prediction model to obtain the groundwater level prediction value of the current time period.

[0015] Further, the time period is one day, and the water level data, meteorological data, hydrological data, geological data and remote sensing data are all taken as the average value collected multiple times in a day, the meteorological data includes precipitation intensity, evaporation intensity and environmental temperature; The hydrological data includes river water level and underground runoff; The geological data includes soil permeability coefficient and soil aquifer thickness; The remote sensing data represents the normalized difference vegetation index of the target area; The human activity data includes groundwater exploitation amount and land use type, the land use type includes forest land, grassland, water area, agricultural land and urban land, and the water level data, meteorological data and hydrological data are standardized by Z-score, and the above data used in subsequent calculation are all the data after Z-score standardization.

[0016] Further, the principle for generating the natural recharge parameter is:

[0017] The land use types of the target area are classified, and the area proportion of different land types is calculated using the following formula:

[0018] ;

[0019] in, Indicates the first The proportion of land area of ​​type 3 to the total area of ​​the target area An index representing land types, and , Indicates the quantity of land types. Indicates the first Land area of ​​type, Indicates the total area of ​​the target region;

[0020] The formula used to calculate the natural replenishment parameter is:

[0021] ;

[0022] in, Indicates the natural supply parameter. Indicates precipitation intensity. The weighting coefficient representing the intensity of precipitation, and , Indicates the precipitation replenishment index. Indicates the river water level. The weighting coefficient represents the river water level, and , Indicates the first Soil permeability coefficient of land-like soil Indicates the first River replenishment index for land-like terrain.

[0023] Furthermore, the principle underlying the generation of natural loss parameters is as follows:

[0024] The vegetation water consumption evaluation index is calculated based on the normalized difference vegetation index, using the following formula:

[0025] ;

[0026] in, This represents the vegetation water consumption evaluation index. Indicates the normalized difference vegetation index. Indicates ambient temperature. This indicates the ambient temperature at which vegetation transpiration is strongest in the target area.

[0027] The evaporation loss index is generated based on the vegetation water consumption evaluation index and evaporation intensity, using the following formula:

[0028] ;

[0029] wherein, Evaporation loss index, Evaporation intensity;

[0030] The seasonal correction coefficient of the evaporation intensity is adjusted based on the season, and the underground runoff is adjusted based on the soil moisture index, to generate the natural loss parameter, and the formula is:

[0031] ;

[0032] wherein, Natural loss parameter, Evaporation loss index, Seasonal correction coefficient, and in summer , in winter , in spring and autumn , Soil water layer thickness of the th land type, Weight coefficients of evaporation loss index and underground runoff index, respectively, And .

[0033] Further, the principle of generating the regional comprehensive impervious area proportion is:

[0034] The land use types of the target region are classified, specifically including forest land, grassland, water area, agricultural land and urban land, and each type of land is allocated an impervious area proportion weight, and the specific weight allocation is 0.1 for forest land, 0.1 for grassland, 0.01 for water area, 0.2 for agricultural land, and 0.8 for urban land; The area proportion of each type of land and its impervious area proportion weight are weighted and averaged to generate the regional comprehensive impervious area proportion, and the formula is:

[0035] ;

[0036] wherein, Regional comprehensive impervious area proportion, Impervious area proportion weight of the th land type.

[0037] Further, the mathematical expression for generating the artificial disturbance parameter is:

[0038] ;

[0039] wherein, Artificial disturbance parameter, Groundwater exploitation, a synergy term representing the groundwater exploitation and the proportion of regional comprehensive impervious area, weight coefficients of the groundwater exploitation, the proportion of regional comprehensive impervious area and the synergy term, respectively, , and .

[0040] Further, the principle for correcting the natural recharge parameter is that:

[0041] For each time period, the average precipitation intensity and the average evaporation intensity of the past 7 days are calculated and Z-score standardized, and the soil moisture index is calculated based on the average precipitation intensity and the average evaporation intensity, and the formula is:

[0042] ;

[0043] wherein, represents the soil moisture index, represents the standardized average precipitation intensity, represents the standardized average evaporation intensity;

[0044] The formula for correcting the natural recharge parameter is:

[0045] ;

[0046] wherein, represents the corrected natural recharge parameter, represents the natural recharge parameter;

[0047] The principle for correcting the artificial disturbance parameter is that:

[0048] For each time period, the net groundwater exploitation of the past 30 days is calculated, and the cumulative effect index is generated, and the formula is:

[0049] ;

[0050] wherein, represents the cumulative effect index, represents the sensitivity coefficient, and , represents the net groundwater exploitation of the past 30 days, represents the exploitation threshold, is a hyperbolic tangent function, which maps the value of to ;

[0051] The corrected artificial disturbance parameter is:

[0052] ;

[0053] wherein, represents the revised artificial disturbance parameter, represents the artificial disturbance parameter.

[0054] The application also provides a groundwater level prediction system based on multi-source data, which is used to implement the above-mentioned groundwater level prediction method based on multi-source data, and specifically comprises:

[0055] A data acquisition module is configured to acquire water level data, meteorological data, hydrological data, geological data, remote sensing data and human activity data of a target region in a past time period.

[0056] A natural activity calculation module is configured to calculate the area proportions of different types of land in the target region, generate a natural recharge parameter of the target region based on the meteorological data, the hydrological data and the area proportions of the different types of land, calculate a vegetation water consumption evaluation index based on the remote sensing data, and further calculate an evaporation loss index of the groundwater, and combine the evaporation loss index with the geological data to generate a natural loss parameter.

[0057] A human activity calculation module is configured to assign an impervious area proportion weight to different types of land in the target region based on the water permeability, generate a regional comprehensive impervious area proportion based on the area proportions of the different types of land and the impervious area proportion weight, and generate an artificial disturbance parameter based on the human activity data and the regional comprehensive impervious area proportion.

[0058] A parameter revision module is configured to revise the natural recharge parameter based on the average values of the meteorological data and the hydrological data, and revise the artificial disturbance parameter based on the cumulative amount of artificial exploitation.

[0059] A model training module is configured to construct a deep learning network, use the natural loss parameter and the revised natural recharge parameter and artificial disturbance parameter in the same time period in the historical data as input, and use the groundwater level in the same time period as a label, to train a water level prediction model.

[0060] A comprehensive output module is configured to input the natural loss parameter, the revised natural recharge parameter and the artificial disturbance parameter in a current time period into the water level prediction model, to obtain a groundwater level prediction value in the current time period.

[0061] Compared with the prior art, the application has the following beneficial effects:

[0062] The factors affecting the groundwater level are divided into natural recharge, natural loss and human activities, the high-order features calculated through pre-processing are input to the subsequent deep learning model, and the learning efficiency and prediction accuracy are improved; the precipitation intensity and river level are combined, and the influence of natural rainfall and river inflow on the groundwater level is fully considered; in the process of calculating the natural loss water quantity, the evaporation and runoff that can be monitored in real time are considered, and for the water absorption of vegetation that is difficult to directly monitor, the growth condition of the vegetation is described, and the actual total water loss is more accurately reflected.

[0063] The application also modifies the calculated natural recharge parameters and artificial disturbance parameters by distinguishing the influence of different seasons on evaporation and the influence of historical data, trains the model based on the modified data, so that the data output by the model is more in line with the actual situation of the groundwater system, and effective fusion and collaborative use of multi-source data are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0064] Fig. 1 The figure is a method flowchart of an embodiment of the application.

[0065] Fig. 2 The figure is a fitting curve diagram of the vegetation water consumption evaluation index of an embodiment of the application.

[0066] Fig. 3 The figure is a system module diagram of an embodiment of the application. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical scheme and advantages of the application clearer and more apparent, the application will be further described in detail below with specific embodiments.

[0068] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the application should be understood as the usual meaning understood by those skilled in the art to which the application belongs. The "first", "second" and similar words used in the application do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. "Up", "down", "left", "right" and the like only represent relative positional relationship, when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0069] Embodiment:

[0070] Please refer to Figs. 1 to 3The application provides a technical scheme:

[0071] A groundwater level prediction method based on multi-source data, and the specific steps include:

[0072] Step 1: Collecting water level data, meteorological data, hydrological data, geological data, remote sensing data and human activity data of the target area in the past time period;

[0073] In this embodiment, the time period is one day, and the water level data, meteorological data, hydrological data, geological data and remote sensing data are all the average values collected in one day. The meteorological data includes precipitation intensity, evaporation intensity and environmental temperature; the hydrological data includes river water level and underground runoff; the geological data includes soil permeability coefficient and soil aquifer thickness; the remote sensing data represents the normalized difference vegetation index of the target area; the human activity data includes groundwater exploitation amount and land use type, and the land use type includes forest land, grassland, water area, agricultural land and urban land; the water level data is collected based on the groundwater monitoring well by recording the water level change through the ultrasonic water level meter; the precipitation intensity is obtained through the meteorological radar, the evaporation intensity is obtained through the evaporation sensor, and the environmental temperature is recorded through the temperature sensor of the meteorological station; the river water level data is obtained through the water level station, and the underground runoff is obtained through the groundwater flowmeter, and the above two kinds of data can be obtained from the hydrological station; the soil permeability coefficient is determined based on different soil types, and the soil aquifer thickness is obtained through geological drilling; the red light and near-infrared band data are obtained through the multispectral remote sensing satellite, and then the normalized difference vegetation index is calculated; the human groundwater exploitation amount is calculated through the water well flowmeter;

[0074] The water level data, meteorological data and hydrological data are subjected to Z-score standardization, and the above data used in subsequent calculation are all the data subjected to Z-score standardization.

[0075] The principle of Z-score standardization of data is as follows:

[0076] Taking the precipitation intensity as an example, when the precipitation intensity of any day is standardized, the precipitation intensity data of the previous 5 days is obtained, the average value of the precipitation intensity of these days is , the standard deviation of the precipitation intensity is , and the standardized precipitation intensity is: , wherein represents the standardized precipitation intensity, represents the precipitation intensity of the current collection time period; the other data that need to be standardized are standardized according to the above method.

[0077] Step 2: Calculate the area ratio of different types of land in the target area, and generate natural replenishment parameters for the target area based on meteorological data, hydrological data, and the area ratio of different types of land; calculate the vegetation water consumption evaluation index based on remote sensing data, and then calculate the groundwater evaporation loss index, and combine the evaporation loss index with geological data to generate natural loss parameters.

[0078] In this embodiment, the principle underlying the generation of natural replenishment parameters is as follows:

[0079] The land use types of the target area are classified, and the area proportion of different land types is calculated using the following formula:

[0080] ;

[0081] in, Indicates the first The proportion of land area of ​​type 3 to the total area of ​​the target area An index representing land types, and , Indicates the quantity of land types. Indicates the first Land area of ​​type, Indicates the total area of ​​the target region;

[0082] Different types of land have significantly different soil types and geological data, so they will be discussed separately.

[0083] The formula used to calculate the natural replenishment parameter is:

[0084] ;

[0085] in, Indicates the natural supply parameter. Indicates precipitation intensity. The weighting coefficient representing the intensity of precipitation, and , Indicates the precipitation replenishment index. Indicates the river water level. The weighting coefficient represents the river water level, and , Indicates the first Soil permeability coefficient of land-like soil Indicates the first River replenishment index for land-like terrain.

[0086] The natural recharge parameter reflects the contribution of natural factors (mainly precipitation and river infiltration) to groundwater recharge in a period. Precipitation is the main source of groundwater recharge, and river level reflects the state of surface water body. The higher the river level, the more it can promote river infiltration to recharge the groundwater along the coast. The infiltration is affected by the soil permeability coefficient. The higher the soil permeability coefficient, the higher the contribution of river infiltration. Wherein is the precipitation recharge index, which reflects the direct contribution of precipitation to groundwater recharge. Precipitation is the most direct factor affecting groundwater recharge. The greater the precipitation, the greater the contribution. The relationship between them is approximately linear. Precipitation intensity always positively promotes groundwater recharge, so the weight coefficient of precipitation intensity is positive. And it cannot completely prevent precipitation infiltration, so the lower limit of is set to 0.5 to avoid the situation that "precipitation has no effect on groundwater level". The upper limit is set to 2, which means that the contribution of precipitation is allowed to be doubled, and at the same time, it prevents a single feature from dominating too much. represents the river recharge index. The influence of river level on groundwater recharge is nonlinear. When the river level is very low, there is no direct relationship between river level and groundwater change. At this time, the contribution of river level is 0. When it exceeds a certain threshold, the river begins to infiltrate and recharge the groundwater. The infiltration effect is strengthened with the increase of river level. And the infiltration of river to groundwater is affected by the soil permeability coefficient. Different soil permeability coefficients are determined for different types of land. Soil permeability coefficient represents the speed of water flowing in soil pores under unit hydraulic gradient. The higher the soil permeability coefficient, the stronger the permeability of the land to water, and the stronger the influence of the river on groundwater recharge. River level has a positive effect on groundwater recharge, and its contribution is very small. The effect of river recharge is affected by distance, geological conditions, etc., and is usually weaker than direct precipitation, so the value range of is ;

[0087] The principle for generating natural loss parameter is:

[0088] The normalized difference vegetation index is calculated based on the formula:

[0089] ;

[0090] Wherein, represents the vegetation water consumption evaluation index, represents the normalized difference vegetation index, represents the environmental temperature, represents the environmental temperature corresponding to the strongest vegetation transpiration of the target area;

[0091] The vegetation water consumption evaluation index reflects the consumption capacity of the vegetation to the groundwater due to the transpiration, represents the normalized difference vegetation index, which is used to measure the vegetation coverage and growth state, The higher the value is, the more lush the vegetation is, the stronger the transpiration of the overall vegetation is, the greater the water consumption is, and the greater the vegetation water consumption evaluation index is, is inversely proportional to the vegetation water consumption evaluation index and the vegetation water consumption; reflects the influence of the temperature on the transpiration of the vegetation, and the temperature at which the transpiration of the vegetation is the strongest is taken as the benchmark. The closer the actual environmental temperature is to this temperature, that is, the greater the actual environmental temperature is, the greater the water consumption of the vegetation is. The vegetation water consumption evaluation index is proportional to the environmental temperature, and the vegetation water consumption evaluation index is inversely proportional to the temperature difference, which represents the temperature difference between the actual environmental temperature and the environmental temperature corresponding to the transpiration of the vegetation being the strongest. Table 1 reflects the change of the vegetation water consumption evaluation index with the normalized difference vegetation index and the environmental temperature. When It can be seen that The smaller the value is, the greater the vegetation water consumption evaluation index is.

[0092] Table 1. Change of the vegetation water consumption evaluation index with NDVI and environmental temperature

[0093]

[0094] The evaporation loss index is generated based on the vegetation water consumption evaluation index and the evaporation intensity, and the formula is:

[0095] ;

[0096] wherein, represents the evaporation loss index, represents the evaporation intensity;

[0097] The evaporation loss index reflects the total evaporation intensity of the target area, including the evaporation from the ground and the evaporation from the plants. The evaporation intensity is adjusted by the transpiration of the vegetation. The water consumption of the evaporation from the ground is evaluated, and the evaporation intensity is adjusted by the transpiration of the vegetation.

[0098] The seasonal correction coefficient of the evaporation intensity is adjusted based on the season, and the groundwater runoff is adjusted based on the soil moisture index to generate the natural loss parameter, and the formula is:

[0099] ;

[0100] wherein, represents the natural loss parameter, represents the evaporation loss index, represents the seasonal correction coefficient, and in summer, and in spring and autumn , denotes the first the soil water layer thickness of the land of the first type, respectively denote the weight coefficient of evaporation loss index and the weight coefficient of underground runoff index, and .

[0101] The natural loss parameter reflects the evaporation loss and runoff loss intensity of the groundwater in the target area, reflects the evaporation of the groundwater, and is adjusted for different seasons on the basis of the evaporation loss index. In summer, the average solar radiation is the strongest in a year, and plants are mostly in the vigorous growth period, resulting in that the evaporation capacity also reaches the annual peak. Therefore, the actual evaporation capacity in summer is much higher than the annual average level, so the basic evaporation loss index is enlarged by 20% through the seasonal correction coefficient to capture the evaporation enhancement effect caused by the season. In winter, the solar radiation is weak, the temperature is low, the vegetation is mostly in the dormant state, and the soil may freeze. The evaporation capacity in winter is much lower than the annual average level, so the basic evaporation loss index is weakened by 20% to reflect the inhibition effect caused by the season. In spring and autumn, the temperature and sunshine are relatively moderate, and the evaporation capacity is close to the average level of the whole year, so no additional adjustment is made; reflects the reduction of the groundwater caused by the underground runoff loss. This process is mainly affected by the soil water layer thickness. The thicker the soil water layer is, the stronger the water storage capacity is, the greater the water flow resistance is, and the slower the runoff loss speed is. The thinner the soil water layer is, the weaker the water storage capacity is, the easier the water flow is, and the faster the runoff loss speed is. The runoff loss is inversely proportional to the thickness of the water layer. For different types of land, the thickness of the water layer is different, so the underground runoff index is calculated for each type of land respectively and summed up to reflect the overall underground runoff intensity of the target area. Since evaporation is the main way of groundwater loss, , , .

[0102] Step 3: Assigning the impervious area proportion weight to different types of land in the target area based on the water permeability, generating the regional comprehensive impervious area proportion based on the area proportion of different types of land and the impervious area proportion weight, and generating the artificial disturbance parameter based on the human activity data and the regional comprehensive impervious area proportion;

[0103] In this embodiment, the principle for generating the regional comprehensive impervious area proportion is as follows:

[0104] The land use types in the target area are classified into forest land, grassland, water area, agricultural land, and urban land. A weighted average of the impervious area proportions is assigned to each land type: forest land 0.1, grassland 0.1, water area 0.01, agricultural land 0.2, and urban land 0.8. The overall impervious area proportion for the region is then calculated by weighting the proportions of each land type with their respective impervious area proportions. The formula used is as follows:

[0105] ;

[0106] in, This indicates the proportion of the overall impermeable area in the region. Indicates the first Weighting of the proportion of impermeable area for each land type.

[0107] The regional comprehensive impervious area ratio reflects the degree of decline in surface permeability caused by changes in land use due to human activities, i.e., it reflects the impact of different types of land on surface runoff and groundwater recharge capacity. A higher impervious area ratio indicates that the surface is less susceptible to water infiltration, resulting in less groundwater recharge. The weighting of the impervious area ratio is based on the land's permeability. This is used to quantify the ability of different land types to impede precipitation infiltration. The value is between 0 and 1, and The smaller the value, the weaker the surface's ability to prevent rainwater infiltration, the more rainwater seeps into the ground, and the more it replenishes the groundwater. The larger the area, the stronger the surface's ability to impede rainwater infiltration, making it more difficult for rainwater to penetrate the ground. In areas with high human activity, such as urban land, the surface is often covered by materials with extremely poor permeability, such as cement and asphalt, making it difficult for rainwater to infiltrate. Therefore, the proportion of impermeable area in such land has a higher weighting. In natural environments such as woodlands and grasslands, the surface is mostly soil and vegetation, offering almost no obstruction to rainwater replenishment; therefore, the proportion of impermeable area has a lower weighting. The overall proportion of impermeable area in the target area is calculated by combining all land types within the target area. , The larger the value, the stronger the ability of the target area to prevent rainwater from seeping into the ground.

[0108] The mathematical expression upon which the artificial perturbation parameters are based is:

[0109] ;

[0110] in, Indicates the parameters of artificial perturbation. This indicates the amount of groundwater extracted. This term represents the synergistic relationship between groundwater extraction volume and the proportion of the overall impermeable area in a region. These represent the weighting coefficients of groundwater extraction volume, the proportion of the overall impermeable area in the region, and the synergistic term, respectively. ,and .

[0111] Artificial disturbance parameters reflect the direct impact of human activities on the groundwater system, mainly including groundwater extraction and changes in land type caused by human activities, which in turn affect precipitation infiltration. This reflects the negative impact of direct mining on groundwater levels. This indicates the amount of groundwater extracted. The larger the value, the more groundwater is extracted and the more severe the impact of human activities. This reflects the indirect impact of land use type on the groundwater system. The higher the proportion of impermeable area, the weaker the rainwater infiltration capacity, the less natural recharge, which indirectly leads to a drop in the groundwater level. This reflects the synergistic effect of groundwater extraction and impermeable area. Higher elevations generally indicate higher levels of urbanization. In these areas, groundwater extraction makes groundwater regeneration more difficult, thus amplifying the impact of human activities. While direct extraction directly affects groundwater, the overall proportion of impermeable surface area in a region indirectly impacts groundwater levels. ,Pick , , .

[0112] Step 4: Correct the natural recharge parameters based on the average values ​​of meteorological and hydrological data, and correct the artificial disturbance parameters based on the cumulative amount of artificial extraction.

[0113] In this embodiment, the principle underlying the correction of the natural replenishment parameter is as follows:

[0114] For each time period, the average precipitation intensity and average evaporation intensity of the past 7 days were statistically analyzed and standardized using the Z-score. The soil moisture index was calculated based on the average precipitation intensity and average evaporation intensity using the following formula:

[0115] ;

[0116] in, Indicates soil moisture index, This represents the standardized average precipitation intensity. This represents the standardized average evaporation intensity;

[0117] The soil moisture index is calculated based on the average precipitation intensity and the average evaporation intensity in the historical data according to the current dry-wet state of the soil, and the higher the precipitation intensity, the higher the soil moisture, and the lower the amount of water that can be reabsorbed by the soil, and more precipitation is used as recharge of groundwater without being absorbed by the soil; the lower the precipitation intensity, the lower the soil moisture, and the higher the amount of water that can be reabsorbed by the soil, and a larger part of the precipitation is absorbed by the soil without flowing into the underground, and the soil moisture is calculated based on the average precipitation intensity and the average evaporation intensity, and the higher the average precipitation intensity, the greater the soil moisture index, and the more humid the soil is;

[0118] The formula for correcting the natural recharge parameter is:

[0119] ;

[0120] wherein, represents the corrected natural recharge parameter, represents the natural recharge parameter;

[0121] represents the correction coefficient of the natural recharge parameter, which is corrected based on the soil moisture index, and the higher the soil moisture index, the higher the soil moisture index under the influence of precipitation and evaporation in the past 7 days, and the more water of the natural recharge enters the groundwater rather than the dry soil; the higher the calculated , the closer to the true value, and the correction is relatively small; the lower the soil moisture index, the drier the soil, and the water of the natural recharge is preferentially absorbed by the dry soil before entering the groundwater, so the calculated is reduced more to consider the water absorption of the soil.

[0122] The principle for correcting the artificial disturbance parameter is:

[0123] For each time period, the cumulative effect index is generated by calculating the net groundwater extraction amount in the past 30 days, and the formula is:

[0124] ;

[0125] wherein, represents the cumulative effect index, represents the sensitivity coefficient, and , represents the net groundwater extraction amount in the past 30 days, represents the extraction amount threshold, is a hyperbolic tangent function, which maps the value of to ;

[0126] The corrected artificial disturbance parameter is:

[0127] ;

[0128] wherein, represents the revised artificial disturbance parameter, represents the artificial disturbance parameter.

[0129] The purpose of revising the artificial disturbance parameter is to reflect the cumulative effect of groundwater exploitation behavior, that is, the continuous influence of long-term exploitation on the groundwater system, represents the cumulative intensity of human exploitation activities in the past 30 days, which is the total exploitation amount minus the recharge amount, is a preset threshold value determined based on the sustainable exploitation amount of groundwater, The purpose is to limit between , so that the cumulative effect coefficient is , when , it means that the net exploitation amount of groundwater is positive, the cumulative effect of exploitation is enhanced, and the artificial disturbance parameter is amplified, when , , it means that there is no cumulative effect in the past 30 days, and no revision is made, when , , it means that there is still recharge,

[0130] Step 5: Construct a deep learning network, use the natural loss parameter in the same time period in the historical data and the revised natural recharge parameter and artificial disturbance parameter as input, and the groundwater level in the same time period as label, train the water level prediction model;

[0131] In this embodiment, the deep learning network structure constructed is as follows:

[0132] Input layer: contains 3 neurons, used for inputting natural loss parameter and revised natural recharge parameter and artificial disturbance parameter;

[0133] First hidden layer: contains 64 neurons, activated using ReLU activation function;

[0134] Second hidden layer: contains 32 neurons, activated using ReLU activation function;

[0135] Third hidden layer: contains 16 neurons, activated using ReLU activation function;

[0136] Output layer: contains 1 neuron, used for outputting groundwater level;

[0137] The following natural loss parameters, corrected natural recharge parameters and corrected artificial disturbance parameters in the same time period in the historical data are used as input features, and the actual observation value of the underground water level at the corresponding time is used as a label, the daily data of the past several years (such as 3-5 years) is used as a training set to ensure covering different seasons and extreme weather conditions, the mean square error function is used as a loss function, and the loss function is minimized by back propagation, the data of the last year in the historical data is divided out as a validation set, the loss is calculated through the validation set, and the training is stopped when the validation set loss no longer decreases, and the current model is saved.

[0138] Step 6: input the natural loss parameters, the corrected natural recharge parameters and the artificial disturbance parameters of the current time period into the water level prediction model to obtain the underground water level prediction value of the current time period.

[0139] Please refer to Fig. 3 The application also provides a groundwater level prediction system based on multi-source data, which is used to implement the above-mentioned groundwater level prediction method based on multi-source data, and specifically comprises:

[0140] A data acquisition module is configured to acquire water level data, meteorological data, hydrological data, geological data, remote sensing data and human activity data of a target region in a past time period.

[0141] A natural activity calculation module is configured to calculate the area proportion of different types of land in the target region, generate a natural recharge parameter of the target region based on the meteorological data, the hydrological data and the area proportion of different types of land, calculate a vegetation water consumption evaluation index based on the remote sensing data, and then calculate an evaporation loss index of the groundwater, and combine the evaporation loss index with the geological data to generate a natural loss parameter.

[0142] A human activity calculation module is configured to assign an impervious area proportion weight to different types of land in the target region based on the water permeability, generate a regional comprehensive impervious area proportion based on the area proportion of different types of land and the impervious area proportion weight, and generate an artificial disturbance parameter based on the human activity data and the regional comprehensive impervious area proportion.

[0143] A parameter correction module is configured to correct the natural recharge parameter based on the average values of the meteorological data and the hydrological data, and correct the artificial disturbance parameter based on the cumulative amount of artificial exploitation.

[0144] A model training module is configured to construct a deep learning network, use the natural loss parameters and the corrected natural recharge parameters and artificial disturbance parameters in the same time period in the historical data as input, and use the underground water level at the corresponding time as a label to train a water level prediction model.

[0145] The comprehensive output module is used for calculating the natural loss parameter, the corrected natural recharge parameter and the artificial disturbance parameter in real time based on the steps of the natural activity calculation module, the human activity calculation module and the parameter correction module, inputting the three real-time parameters into the water level prediction model, and outputting the real-time underground water level prediction value.

[0146] The above formulas are all dimensionless values calculated, the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formulas are set by the person skilled in the art according to the actual situation.

[0147] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software methods depends on the specific application and design constraints of the technical solutions.

[0148] The units described as separate components can or can not be physically separated, the components shown as units can or can not be physical units, and can be located in one place or distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0149] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for predicting groundwater level based on multi-source data, characterized in that, The specific steps include: Step 1: Collecting water level data, meteorological data, hydrological data, geological data, remote sensing data and human activity data of the target area in the past time period; Step 2: Calculating the area proportion of different types of land in the target area, generating the natural recharge parameter of the target area based on meteorological data, hydrological data and the area proportion of different types of land, calculating the vegetation water consumption evaluation index based on remote sensing data, and then calculating the evaporation loss index of groundwater, combining the evaporation loss index with the geological data to generate the natural loss parameter; Step 3: Assigning different types of land in the target area based on the water permeability to generate the impervious area proportion weight, generating the regional comprehensive impervious area proportion based on the area proportion of different types of land and the impervious area proportion weight, and generating the artificial disturbance parameter based on the human activity data and the regional comprehensive impervious area proportion; Step 4: Correcting the natural recharge parameter based on the average value of meteorological data and hydrological data, and correcting the artificial disturbance parameter based on the cumulative amount of artificial exploitation; Step 5: Constructing a deep learning network, taking the natural loss parameter and the corrected natural recharge parameter and artificial disturbance parameter in the same time period in the historical data as input, and taking the groundwater level in the same time period as label, to train the water level prediction model; Step 6: Inputting the natural loss parameter, the corrected natural recharge parameter and the artificial disturbance parameter of the current time period into the water level prediction model to obtain the predicted value of the groundwater level in the current time period; The principle for generating the natural recharge parameter is: Classify the land use types of the target area, calculate the area proportion of different types of land, and the formula is: wherein D i represents the proportion of the area of the i-th land type to the total area of the target region, i represents an index of the land type, and i∈[1,I], I represents the number of land types, A i represents the area of the i-th land type; The formula for calculating the natural recharge parameter is: wherein R t represents a natural recharge parameter, P t (norm) represents a precipitation intensity, k1 represents a weight coefficient of the precipitation intensity, and k1 ∈ [0.5, 2], k1 × P t (norm) represents a precipitation recharge index, W t (norm) represents a river level, k2 represents a weight coefficient of the river level, and k2 ∈ (0, 1.5], K(i) represents a soil permeability coefficient of the ith type of land, and k2 × lg[1 + W t (norm)] × K(i) represents a river recharge index of the ith type of land. The principle for generating the natural loss parameter is: Calculate the vegetation water consumption evaluation index based on the normalized difference vegetation index, and the formula is: wherein N t represents the vegetation water consumption evaluation index, NDVI represents the normalized difference vegetation index, T t represents the environmental temperature, T0 represents the environmental temperature corresponding to the strongest vegetation transpiration of the target region; Generate the evaporation loss index based on the vegetation water consumption evaluation index and the evaporation intensity, and the formula is: M t = N t × E t (norm) wherein M t represents the evaporation loss index, E t represents the evaporation intensity; Adjust the seasonal correction coefficient of evaporation intensity based on the season, adjust the underground runoff based on the soil moisture index, and generate the natural loss parameter, and the formula is: where S t represents the natural loss parameter, M t represents the evaporation loss index, C season represents the seasonal correction coefficient, and C season = 1.2 in summer, C season = 0.8 in winter, C season = 1.0 in spring and autumn, H i represents the soil water layer thickness of the i-th type of land, and u1 and u2 represent the weight coefficients of the evaporation loss index and the underground runoff index, respectively, with u1 + u2 = 1 and u1 > u2. 2.The groundwater level prediction method based on multi-source data according to claim 1, characterized in that: In step 1, the time period is one day, and the water level data, meteorological data, hydrological data, geological data and remote sensing data are all taken as the average value of multiple collections in one day, the meteorological data includes precipitation intensity, evaporation intensity and environmental temperature; the hydrological data includes river water level and underground runoff; the geological data includes soil permeability coefficient and soil water layer thickness; the remote sensing data represents the normalized difference vegetation index of the target area; the human activity data includes groundwater exploitation amount and land use type, the land use type includes forest land, grassland, water area, agricultural land and urban land, and the water level data, meteorological data and hydrological data are standardized by Z-score, and the above data used in subsequent calculation are all the data after Z-score standardization. 3.The groundwater level prediction method based on multi-source data according to claim 1, characterized in that: The principle for generating the regional comprehensive impervious area proportion in step 3 is: The land use types of the target area are classified, specifically including forest land, grassland, water area, agricultural land and urban land, and a non-permeable area proportion weight is assigned to each type of land, and the specific weight assignment is 0.1 for forest land, 0.1 for grassland, 0.01 for water area, 0.2 for agricultural land and 0.8 for urban land; the area proportions of various types of land are weighted and averaged with the non-permeable area proportion weights to generate a regional comprehensive non-permeable area proportion, and the formula is: where B represents the proportion of the impervious area of the regional synthesis, W i represents the impervious area proportion weight of the i-th land type.

4. The method of claim 3, wherein: The mathematical expression for generating the artificial disturbance parameter in step 3 is: H t = v1 x Q t + v2 x B + v3 x B x Q where H t represents the artificial disturbance parameter, Q t represents the groundwater exploitation amount, BxQ represents a synergy term of the groundwater exploitation amount and the proportion of the regional comprehensive impervious area, v1, v2, and v3 respectively represent weight coefficients of the groundwater exploitation amount, the proportion of the regional comprehensive impervious area, and the synergy term, v1+v2+v3=1, and v1>v2>v3.

5. The method of claim 2, wherein: The principle for correcting the natural recharge parameter in step 4 is: For each time period, the average precipitation intensity and average evaporation intensity of the past 7 days are calculated and Z-score standardized, and the soil moisture index is calculated based on the average precipitation intensity and average evaporation intensity, and the formula is: where SMI represents the soil moisture index, P acc represents the normalized average precipitation intensity, E acc (norm) represents the normalized average evaporation intensity; The formula for correcting the natural recharge parameter is: R' t = R t x SMI t 2 wherein R' t represents the corrected natural recharge parameter, a represents an adjustment coefficient, and a = 2, R t represents the natural recharge parameter; The principle for correcting the artificial disturbance parameter is: For each time period, the net groundwater exploitation amount of the past 30 days is calculated to generate an accumulated effect index, and the formula is: where β represents an accumulative effect index, γ represents a sensitivity coefficient, and γ ∈ [0.1, 0.3], Q net represents the net exploitation of groundwater in the past 30 days, Q0 represents an exploitation threshold, is a hyperbolic tangent function that maps the value of to between (-1, 1); The corrected artificial disturbance parameter is: H' t = β x H t where H' = H + AH t represents the modified artificial disturbance parameter, H t represents the artificial disturbance parameter.

6. A multi-source data based groundwater level prediction system, characterized in that: The system is used to implement the groundwater level prediction method based on multi-source data according to any one of claims 1-5, and specifically includes: A data acquisition module is configured to acquire water level data, meteorological data, hydrological data, geological data, remote sensing data and human activity data of a target area in a past time period; A natural activity calculation module is configured to calculate the area proportions of different types of land in the target area, generate a natural recharge parameter of the target area based on meteorological data, hydrological data and the area proportions of different types of land, calculate a vegetation water consumption evaluation index based on remote sensing data, and further calculate an evaporation loss index of groundwater, and combine the evaporation loss index with geological data to generate a natural loss parameter; A human activity calculation module is configured to assign a non-permeable area proportion weight to different types of land in the target area based on permeability, generate a regional comprehensive non-permeable area proportion based on the area proportions of different types of land and the non-permeable area proportion weight, and generate an artificial disturbance parameter based on human activity data and the regional comprehensive non-permeable area proportion; A parameter correction module is configured to correct the natural recharge parameter based on the average values of meteorological data and hydrological data, and correct the artificial disturbance parameter based on the cumulative amount of artificial exploitation; A model training module is configured to construct a deep learning network, use the natural loss parameter of the same time period and the corrected natural recharge parameter and artificial disturbance parameter in the historical data as input, and use the groundwater level of the same time period as label to train a water level prediction model; A comprehensive output module is configured to input the natural loss parameter, the corrected natural recharge parameter and the artificial disturbance parameter of the current time period into the water level prediction model to obtain the groundwater level prediction value of the current time period.

Images