A model-fitting-based water resource assessment method
By constructing a water resource spatial and temporal scale assessment model based on BP neural network and LSTM neural network, and using the GBDT algorithm for model fusion, the problems of computational complexity of traditional methods and low accuracy of remote sensing methods are solved, thus achieving efficient and accurate water resource assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-01-30
- Publication Date
- 2026-04-03
AI Technical Summary
In existing technologies, traditional methods rely on a large amount of data and have complex calculation processes, resulting in poor timeliness of water resource assessments, while remote sensing methods have low assessment accuracy.
By constructing a water resource spatial and temporal scale assessment model based on BP neural network and LSTM neural network, and using the GBDT algorithm for model fusion, combined with environmental data for multi-scale assessment, a water level and water volume mapping relationship is constructed to achieve multi-scale assessment of water resources.
It improves the accuracy and timeliness of water resource assessment, reduces the complexity of the calculation process, and realizes automated and real-time water resource quantity measurement.
Smart Images
Figure CN116011713B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of water resources assessment technology, and specifically to a water resources assessment method based on model fitting. Background Technology
[0002] Lakes are among the most widely distributed water bodies on land and an important component of the terrestrial hydrosphere, closely related to the natural water cycle. As a natural resource closely related to human survival and development, lakes play a vital role in regulating regional climate, urban and rural water supply, improving the ecological environment, protecting biodiversity, agricultural irrigation, tourism, shipping, power generation, and aquaculture. Water resources are the core of lake resources, serving as the material basis for maintaining the stability and healthy development of lake ecosystems. Therefore, the rational assessment and prediction of lake water resources are of great significance.
[0003] Lake water volume changes dynamically with variations in its inflow and outflow. Lake water inflow typically includes surface precipitation, inflow from external rivers, and recharge from surrounding aquifers; outflow typically includes surface evaporation, lakebed seepage, downstream river discharge, and artificial water intake. Existing methods for assessing lake water resources mainly include traditional methods and remote sensing methods. Traditional methods primarily obtain values for various inflow and outflow parameters through hydrological station observation data and field experiments, then calculate the lake water volume. This method not only requires high-quality data but also consumes significant human and material resources, and it is difficult to effectively assess the water volume of multiple lakes in a large area, thus having significant limitations. In recent years, with the continuous development of remote sensing technology, more and more scholars have used remote sensing data to assess lake water volume. While this method is more convenient to operate, it generally estimates lake water volume based on lake level and surface area, and the accuracy of the assessment results needs improvement. Therefore, both traditional and remote sensing methods have certain drawbacks. Traditional methods rely on a large amount of data and have a complex calculation process, resulting in poor timeliness of water resource assessment. Remote sensing methods have a single estimation method, resulting in poor accuracy of water resource assessment. Summary of the Invention
[0004] The purpose of this invention is to provide a water resource assessment method based on model fitting, in order to solve the technical problems in the prior art where traditional methods rely on a large amount of data and have a complex calculation process, resulting in poor timeliness of water resource assessment, and remote sensing methods have a single estimation method, resulting in poor accuracy of water resource assessment.
[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:
[0006] A model-fit-based water resource assessment method includes the following steps:
[0007] Step S1: Obtain historical environmental data and historical water level data, and extract the historical time series corresponding to the historical environmental data and historical water level data;
[0008] Step S2: Construct a spatial scale assessment model for water resources using historical environmental data and historical water level data to achieve spatial scale assessment of water resources;
[0009] Step S3: Construct a water resources time-scale assessment model using historical water level data and historical time series to achieve water resources assessment on a time scale;
[0010] Step S4: Use the GBDT algorithm to fuse the spatial scale assessment model and the temporal scale assessment model of water resources to obtain a multi-scale fusion assessment model of water resources, so as to realize multi-scale assessment of water resources and improve the accuracy of water resources assessment.
[0011] Step S5: Using the current water level data calculated by the multi-scale fusion assessment model of water resources, the current water level data is used to obtain the current water resource storage through the preset water level and water volume mapping relationship.
[0012] As a preferred embodiment of the present invention, the step of constructing a spatial-scale assessment model for water resources using historical environmental data and historical water level data includes:
[0013] Historical environmental data is used as the input to a BP neural network, and historical water level data is used as the output. The BP neural network is trained based on the input and output of the BP neural network to obtain the spatial scale assessment model of water resources.
[0014] The model expression for the spatial scale assessment model of water resources is as follows:
[0015] H t =BP(en_data) t );
[0016] In the formula, H t For the water level data at time t, en_data t Let t be the environmental data at the t-th time step, BP be a BP neural network, and t be the time step counting variable.
[0017] As a preferred embodiment of the present invention, the step of constructing a water resources time-scale assessment model using historical water level data and historical time series includes:
[0018] The water level data from the previous historical time series, as well as the previous and subsequent historical time series, are used as input terms of the LSTM neural network. The water level data from the subsequent historical time series are used as output terms of the LSTM neural network. The LSTM neural network is trained using the input terms and output terms of the LSTM neural network to obtain the water resource time scale assessment model.
[0019] The model expression for the water resources time-scale assessment model is as follows:
[0020] H t =LSTM(H t-1 ,T t ,T t-1 );
[0021] In the formula, H t For the water level data at time t, H t-1 For the water level data at time series t-1, T t T t-1 , , are the time series values of the t-th time series and the (t-1)-th time series, respectively. LSTM is an LSTM neural network, and t is the time series count variable.
[0022] As a preferred embodiment of the present invention, the step of using the GBDT algorithm to fuse the spatial scale assessment model and the temporal scale assessment model of water resources to obtain a multi-scale fused assessment model of water resources includes:
[0023] The outputs of both the spatial scale assessment model and the temporal scale assessment model of water resources at the same historical time series are used as inputs to the GBDT algorithm, and historical water level data at the same historical time series are used as outputs to the GBDT algorithm. The GBDT algorithm is trained based on the inputs and outputs to obtain the multi-scale fusion assessment model of water resources.
[0024] The model expression for the multi-scale integrated assessment model of water resources is as follows:
[0025] H t =GBDT(BP[H t ],LSTM[H t ]);
[0026] In the formula, H t For the water level data at time t, BP[H t ] represents the output term of the spatial scale assessment model for water resources at time t, LSTM[H t ] represents the output of the water resources time-scale assessment model at the t-th time series, where t is the time series count variable.
[0027] As a preferred embodiment of the present invention, the environmental data includes at least one data component selected from temperature and humidity data, sunshine data, precipitation data, plant and animal data, and solar term data, and each data component in the environmental data is normalized.
[0028] As a preferred embodiment of the present invention, the output terms of the spatial scale assessment model of water resources and the output terms of the temporal scale assessment model of water resources at the same historical time series are both normalized before fusion training.
[0029] As a preferred embodiment of the present invention, the current water level data calculated using the multi-scale fusion assessment model of water resources includes:
[0030] Acquire current environmental data and water level data at the current preceding time series. Input the current environmental data into the water resources spatial scale assessment model to output the current water level data at the spatial scale. Input the water level data at the current preceding time series, the current preceding time series, and the current time series into the water resources temporal scale assessment model to output the current water level data at the temporal scale.
[0031] The current water level data at both the spatial and temporal scales are used to calculate the multi-scale fused current water level data using a water resources multi-scale fusion assessment model.
[0032] As a preferred embodiment of the present invention, the current water resource reserves are obtained from the current water level data fused from multiple scales according to the water level and water volume mapping relationship.
[0033] As a preferred embodiment of the present invention, the construction of the water level and water volume mapping relationship includes:
[0034] Water level data and water resource storage value are measured at the same time series. Water level data is used as the independent variable and water resource storage value is used as the dependent variable to construct a two-dimensional curve representing the mapping relationship between water level and water volume.
[0035] In this two-dimensional curve, the horizontal axis represents water level data, and the vertical axis represents water resource reserves.
[0036] As a preferred embodiment of the present invention, the construction of the water level and water volume mapping relationship includes:
[0037] Water level data and water resource storage value are measured at the same time series. Water level data and water resource storage value are used as input terms of neural network to construct a water quantity measurement network that represents the mapping relationship between water level and water quantity.
[0038] The expression for the water quantity measurement network is as follows:
[0039] V t =network(Ht );
[0040] Among them, H t Let V be the water level data at the t-th time series. t Let be the water resource storage at the t-th time series, network be a neural network, and t be a time series counting variable.
[0041] Compared with the prior art, the present invention has the following advantages:
[0042] This invention utilizes historical environmental data and historical water level data to construct a spatial-scale assessment model for water resources, enabling assessment of water resources at a spatial scale. It also utilizes historical water level data and historical time series to construct a temporal-scale assessment model for water resources, enabling assessment at a temporal scale. The GBDT algorithm is then used to fuse the spatial-scale and temporal-scale assessment models to obtain a multi-scale fusion assessment model for water resources. This multi-scale assessment improves the accuracy of water resource assessment. The use of model-based assessment reduces the complexity of the computational process, resulting in more timely water resource assessments. Attached Figure Description
[0043] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.
[0044] Figure 1 A flowchart of a water resource assessment method provided in an embodiment of the present invention. Detailed Implementation
[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0046] like Figure 1 As shown, this invention provides a water resource assessment method based on model fitting, comprising the following steps:
[0047] Step S1: Obtain historical environmental data and historical water level data, and extract the historical time series corresponding to the historical environmental data and historical water level data;
[0048] Step S2: Construct a spatial scale assessment model for water resources using historical environmental data and historical water level data to achieve spatial scale assessment of water resources;
[0049] A spatial-scale assessment model for water resources is constructed using historical environmental data and historical water level data, including:
[0050] Historical environmental data is used as the input to a BP neural network, and historical water level data is used as the output. The BP neural network is trained based on the input and output of the BP neural network to obtain a spatial scale assessment model for water resources.
[0051] The model expression for the spatial scale assessment model of water resources is:
[0052] H t =BP(en_data) t );
[0053] In the formula, H t For the water level data at time t, en_data t Let t be the environmental data at the t-th time step, BP be a BP neural network, and t be the time step counting variable.
[0054] A spatial scale assessment model for water resources is constructed to calculate water level data by using the correlation between spatial environmental data and water level data in a model-based manner. This enables the model-based assessment of water level data at a spatial scale. Compared with the manual calculation process using various data, the model-based assessment reduces the manual involvement, thus reducing the complexity of the calculation process and achieving a high degree of automation.
[0055] Step S3: Construct a water resources time-scale assessment model using historical water level data and historical time series to achieve water resources assessment on a time scale;
[0056] A time-scale water resources assessment model is constructed using historical water level data and historical time series data, including:
[0057] The water level data from the previous historical time series, as well as the previous and subsequent historical time series, are used as input terms of the LSTM neural network. The water level data from the subsequent historical time series are used as output terms of the LSTM neural network. The LSTM neural network is trained using the input terms and output terms of the LSTM neural network to obtain a water resources time-scale assessment model.
[0058] The model expression for the water resources time-scale assessment model is as follows:
[0059] H t =LSTM(H t-1 ,Tt ,T t-1 );
[0060] In the formula, H t For the water level data at time t, H t-1 For the water level data at time series t-1, T t T t-1 , , are the time series values of the t-th time series and the (t-1)-th time series, respectively. LSTM is an LSTM neural network, and t is the time series count variable.
[0061] A spatial scale assessment model for water resources is constructed to calculate water level data in a model-based manner by utilizing the temporal correlation of water level data at the time level. This enables model-based assessment of water level data at the time scale. Compared with the manual calculation process using various data, model-based assessment reduces the manual involvement process, that is, reduces the complexity of the calculation process and has a high degree of automation.
[0062] Step S4: Use the GBDT algorithm to fuse the spatial scale assessment model and the temporal scale assessment model of water resources to obtain a multi-scale fusion assessment model of water resources, so as to realize multi-scale assessment of water resources and improve the accuracy of water resources assessment.
[0063] The GBDT algorithm is used to fuse spatial-scale and temporal-scale water resource assessment models to obtain a multi-scale fused water resource assessment model, including:
[0064] The outputs of both the spatial scale assessment model and the temporal scale assessment model of water resources at the same historical time series are used as inputs to the GBDT algorithm, and historical water level data at the same historical time series are used as outputs to the GBDT algorithm. The GBDT algorithm is trained using the inputs and outputs to obtain a multi-scale fusion assessment model of water resources.
[0065] The model expression for the multi-scale integrated assessment model of water resources is as follows:
[0066] H t =GBDT(BP[H t ],LSTM[H t ]);
[0067] In the formula, H t For the water level data at time t, BP[H t ] represents the output term of the spatial scale assessment model for water resources at time t, LSTM[H t ] represents the output of the water resources time-scale assessment model at the t-th time series, where t is the time series count variable.
[0068] The GBDT algorithm integrates spatial and temporal scale assessment models for water resources. During the modeling process, it selects more accurate measurement results at both spatial and temporal scales, enabling multi-scale assessment of water resources to improve the accuracy of water resource assessment. Compared to single-scale water resource assessment, the integrated assessment model with both spatial and temporal scales fully leverages the strengths of each scale assessment model, ensuring the real-time accuracy of the model-based assessment.
[0069] Step S5: Using the current water level data calculated by the multi-scale fusion assessment model of water resources, the current water level data is used to obtain the current water resource storage through the preset water level and water volume mapping relationship.
[0070] Environmental data includes at least one of the following data components: temperature and humidity data, sunshine data, precipitation data, flora and fauna data, and solar term data. Each data component in the environmental data is then normalized.
[0071] Both the outputs of the spatial scale assessment model and the temporal scale assessment model of water resources at the same historical time point are normalized before fusion training.
[0072] Current water level data calculated using a multi-scale fusion assessment model for water resources includes:
[0073] Acquire current environmental data and water level data at the current preceding time series. Input the current environmental data into the water resources spatial scale assessment model to output the current water level data at the spatial scale. Input the water level data at the current preceding time series, the current preceding time series, and the current time series into the water resources temporal scale assessment model to output the current water level data at the temporal scale.
[0074] The current water level data at both the spatial and temporal scales are used to calculate the multi-scale fused current water level data using a water resources multi-scale fusion assessment model.
[0075] The current water resource reserves are obtained from the multi-scale fusion of current water level data based on the mapping relationship between water level and water volume.
[0076] The construction of the water level-water volume mapping relationship includes:
[0077] Water level data and water resource storage value are measured at the same time series. Water level data is used as the independent variable and water resource storage value is used as the dependent variable to construct a two-dimensional curve representing the mapping relationship between water level and water volume.
[0078] In this two-dimensional curve, the horizontal axis represents water level data, and the vertical axis represents water resource reserves.
[0079] The construction of the water level-water volume mapping relationship includes:
[0080] Water level data and water resource storage value are measured at the same time series. Water level data and water resource storage value are used as input terms of neural network to construct a water quantity measurement network that represents the mapping relationship between water level and water quantity.
[0081] The expression for the water quantity measurement network is as follows:
[0082] V t =network(H t );
[0083] Among them, H t Let V be the water level data at the t-th time series. t Let be the water resource storage at the t-th time series, network be a neural network, and t be a time series counting variable.
[0084] This invention utilizes historical environmental data and historical water level data to construct a spatial-scale assessment model for water resources, enabling assessment of water resources at a spatial scale. It also utilizes historical water level data and historical time series to construct a temporal-scale assessment model for water resources, enabling assessment at a temporal scale. The GBDT algorithm is then used to fuse the spatial-scale and temporal-scale assessment models to obtain a multi-scale fusion assessment model for water resources. This multi-scale assessment improves the accuracy of water resource assessment. The use of model-based assessment reduces the complexity of the computational process, resulting in more timely water resource assessments.
[0085] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.
Claims
1. A water resource assessment method based on model fitting, characterized in that: Includes the following steps: Step S1: Obtain historical environmental data and historical water level data, and extract the historical time series corresponding to the historical environmental data and historical water level data; Step S2: Construct a spatial scale assessment model for water resources using historical environmental data and historical water level data to achieve spatial scale assessment of water resources. Historical environmental data is used as the input of a BP neural network, and historical water level data is used as the output of a BP neural network. The BP neural network is trained based on the input and output of the BP neural network to obtain the spatial scale assessment model for water resources. Step S3: Construct a water resources time-scale assessment model using historical water level data and historical time series to achieve water resources assessment on a time scale. Specifically, the water level data from the previous historical time series, as well as the previous and subsequent historical time series, are used as input terms of an LSTM neural network. The water level data from the subsequent historical time series are used as output terms of the LSTM neural network. The LSTM neural network is trained based on the input terms and output terms to obtain the water resources time-scale assessment model. Step S4: The spatial scale assessment model and the temporal scale assessment model of water resources are fused using the GBDT algorithm to obtain a multi-scale fusion assessment model of water resources, thereby improving the accuracy of water resources assessment. Specifically, the outputs of the spatial scale assessment model and the temporal scale assessment model of water resources at the same historical time series are used as inputs to the GBDT algorithm, and the historical water level data at the same historical time series are used as outputs to the GBDT algorithm. The multi-scale fusion assessment model of water resources is obtained by training the GBDT algorithm based on the inputs and outputs of the GBDT algorithm. Step S5: Using the current water level data calculated by the multi-scale fusion assessment model of water resources, the current water level data is used to obtain the current water resource storage through the preset water level-water volume mapping relationship. Specifically, the water level data and water resource storage value at the same time series are calculated, and the water level data and water resource storage value are used as input terms of the neural network to construct a water volume measurement network that represents the water level-water volume mapping relationship.
2. The water resource assessment method based on model fitting according to claim 1, characterized in that: The model expression for the spatial scale assessment model of water resources is as follows: H t =BP(a_data t ); In the formula, H t For the water level data at time t, en_data t Let t be the environmental data at the t-th time step, BP be a BP neural network, and t be the time step counting variable.
3. The water resource assessment method based on model fitting according to claim 2, characterized in that: The model expression for the water resources time-scale assessment model is as follows: H t =LSTM(H t-1 ,T t , T t-1 ); In the formula, H t For the water level data at time t, H t-1 For the water level data at time series t-1, T t T t-1 , , are the time series values of the t-th time series and the (t-1)-th time series, respectively. LSTM is an LSTM neural network, and t is the time series count variable.
4. The water resource assessment method based on model fitting according to claim 3, characterized in that: The model expression for the multi-scale integrated assessment model of water resources is as follows: H t =GBDT(BP[H t ], LSTM[H t ]); In the formula, H t For the water level data at time t, BP[H t ] represents the output term of the spatial scale assessment model for water resources at time t, LSTM[H t ] represents the output of the water resources time-scale assessment model at the t-th time series, where t is the time series count variable.
5. The water resource assessment method based on model fitting according to claim 4, characterized in that: Environmental data includes at least one of the following data components: temperature and humidity data, sunshine data, precipitation data, flora and fauna data, and solar term data. Each data component in the environmental data is then normalized.
6. The water resource assessment method based on model fitting according to claim 5, characterized in that: The outputs of both the spatial scale assessment model and the temporal scale assessment model of water resources at the same historical time point are normalized before fusion training.
7. The water resource assessment method based on model fitting according to claim 6, characterized in that, The current water level data calculated using the multi-scale fusion assessment model for water resources includes: Acquire current environmental data and water level data at the current preceding time series. Input the current environmental data into the water resources spatial scale assessment model to output the current water level data at the spatial scale. Input the water level data at the current preceding time series, the current preceding time series, and the current time series into the water resources temporal scale assessment model to output the current water level data at the temporal scale. The current water level data at both the spatial and temporal scales are used to calculate the multi-scale fused current water level data using a water resources multi-scale fusion assessment model.
8. The water resource assessment method based on model fitting according to claim 7, characterized in that, The current water resource reserves are obtained from the multi-scale fusion of current water level data based on the water level-volume mapping relationship.
9. A water resource assessment method based on model fitting according to claim 8, characterized in that, The construction of the water level and water volume mapping relationship includes: Water level data and water resource storage value are measured at the same time series. Water level data is used as the independent variable and water resource storage value is used as the dependent variable to construct a two-dimensional curve representing the mapping relationship between water level and water volume. In this two-dimensional curve, the horizontal axis represents water level data, and the vertical axis represents water resource reserves.
10. A water resource assessment method based on model fitting according to claim 8, characterized in that, The expression for the water quantity measurement network is: V t =network(H t ); Among them, H t Let V be the water level data at the t-th time series. t Let be the water resource storage at the t-th time series, network be a neural network, and t be a time series counting variable.
Citation Information
Patent Citations
Reservoir water level prediction method based on neural network
CN113887774A