Reservoir scheduling method and system based on rainfall forecast and remote sensing monitoring, and storage medium

Abstract

The invention provides a reservoir scheduling method based on rainfall forecast and remote sensing monitoring. The reservoir scheduling method comprises the following steps: step 1, collecting basic data of a research area and constructing a plurality of basic databases; 2, the basic data of the basic database are preprocessed; 3, constructing a weather forecast model taking the reservoir catchment range as a core, and forecasting weather data in a forecast period; step 4, constructing a distributed hydrological model considering future climate and land utilization or cover change, and calculating a reservoir incoming flow process in a prediction period; 5, calculating irrigation area water deficit state data based on the irrigation area crop remote sensing image; 6, the incoming flow process and irrigation area water deficit state data of the reservoir in the future period in the prediction period serve as input conditions, boundary conditions and constraint conditions are coupled, and a reservoir adaptability scheduling model is constructed; and 7, solving the reservoir adaptability scheduling model. According to the invention, through comprehensive analysis of multi-source information, a multi-level scheduling model is established, and accurate scheduling of the reservoir is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of reservoir scheduling, and in particular relates to a reservoir scheduling method, system and storage medium based on rainfall forecast and remote sensing monitoring. Background Art

[0002] Reservoir operation plays a crucial role in watershed management, but faces complexities and uncertainties arising from future climate change, land use / cover change, and other human activities. For reservoirs with irrigation functions, traditional operation methods fail to fully integrate the precise data provided by real-time weather forecasts and remote sensing technology. In recent years, the application of remote sensing technology in crop growth monitoring has provided new insights for improving irrigation efficiency in irrigated areas. Simultaneously, rainfall prediction models based on numerical weather forecasts provide forward-looking information for forecasting water inflow conditions. However, the coordinated application of these two technologies in reservoir operation remains limited.

[0003] The present invention aims to integrate remote sensing monitoring and rainfall forecasting technologies, construct a scheduling model based on multi-source data coupling for reservoirs with irrigation tasks, and provide a more efficient and accurate water resources management solution to achieve precise crop irrigation and optimal use of water resources. Summary of the Invention

[0004] This invention provides a reservoir scheduling method, system, and storage medium based on rainfall forecast and remote sensing monitoring. This method establishes a multi-layered scheduling model by comprehensively analyzing multiple sources of information, including rainfall forecasts, remote sensing data, soil moisture, land use / cover data, and meteorological data, to achieve precise reservoir scheduling.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A reservoir scheduling method based on rainfall forecast and remote sensing monitoring includes the following steps:

[0007] Step 1. Collect basic data of the study area and build multiple basic databases;

[0008] Step 2. Preprocess the basic data of multiple basic databases to obtain preprocessed data;

[0009] Step 3. Taking the upstream catchment area of the reservoir with irrigation tasks as the study area, construct a meteorological forecast model with the reservoir catchment area as the core, determine the forecast period that meets the forecast accuracy requirements, and forecast the meteorological data for the forecast period;

[0010] Step 4. Construct a distributed hydrological model that takes into account future climate and land use or cover changes, and calculate the reservoir inflow process during the forecast period in step 3;

[0011] Step 5. Calculate the water shortage status data of the irrigation area based on the remote sensing image of the crops in the irrigation area;

[0012] Step 6. Taking the reservoir inflow process in the future period of the forecast period and the water shortage status data of the irrigation area as input conditions, coupling the boundary conditions and constraints, and constructing the reservoir adaptive operation model;

[0013] Step 7. Use the multi-objective particle swarm optimization algorithm to solve the reservoir adaptive operation model and obtain a non-inferior solution set.

[0014] Furthermore, the database in step 1 includes: an irrigation area characteristic database, a crop growth database, a meteorological and hydrological database, a watershed physical geography database, a watershed land use database, a reservoir characteristic and operation database, a remote sensing database, and a reanalysis database.

[0015] Furthermore, step 2 includes the following sub-steps:

[0016] Step 21, setting the basic data threshold range based on the historical distribution pattern of basic data in the study area;

[0017] Step 22: Based on the threshold range of the basic data, the Z-score method is used to detect outliers in the basic data and process them;

[0018] Step 23: Use ArcGIS software to perform projection conversion and unify the geographic information system coordinate system of each basic database in the study area;

[0019] Step 24, converting basic data with different time resolutions into the same time scale;

[0020] Step 25: normalize the basic data of different scales.

[0021] Furthermore, step 3 includes the following sub-steps:

[0022] Step 31: Based on the DEM data in the watershed physical geographic database, the watershed boundary of the reservoir and the distribution data of the tributaries in the upper reaches of the watershed are extracted, and the spatial resolution is selected based on the size of the watershed area and the size of the irrigation area in the upper reaches of the reservoir.

[0023] Step 32: Use the wrapper feature selection algorithm to extract the preprocessed data in step 2 and input it into the LSTM model, and set the hyperparameters of the LSTM model;

[0024] In step 33, the objective functions of the LTSM model are set as follows: (1) maximizing the Nash efficiency coefficient between the simulated rainfall value and the measured rainfall value, and (2) maximizing the Nash efficiency coefficient between the simulated evaporation value and the measured evaporation value.

[0025] Step 34: randomly divide the preprocessed data into a training set and a test set according to a certain ratio, use the training set to train the LSTM model, and tune the hyperparameters of the LSTM model;

[0026] Step 35, using the test set to evaluate the prediction effect of the model, determining a prediction period that meets the prediction accuracy requirements based on the evaluation results, and evaluating the rainfall and evaporation prediction results for the prediction period;

[0027] Step 38: Select the forecast data of the intersection period of the forecast period that meets the rainfall forecast accuracy requirement and the forecast period that meets the evaporation forecast accuracy requirement as the input for the scheduling model construction.

[0028] Furthermore, the above step 4 includes the following sub-steps:

[0029] Step 41, based on the historical rainfall and evaporation observed in the meteorological and hydrological database and the measured flow process data at the reservoir location, the data are divided into a rate period and a test period according to a certain ratio;

[0030] Step 42, based on the output of the global climate model, a bias correction method is used to obtain precipitation, temperature and evaporation data for the basin prediction period;

[0031] Step 43: Based on the historical land use or land cover scenarios in the watershed land use database and the land use policy information of the local land space planning, simulate the land use / land cover change scenarios in the watershed during the forecast period.

[0032] Step 44: construct a distributed hydrological model that takes into account future climate and land use or land cover changes, and calibrate model parameters based on the regular data set to obtain an optimized distributed hydrological model;

[0033] Step 45: inputting rainfall, temperature, and evaporation data during the test period, predicting the reservoir inflow process during the test period, and verifying the optimized distributed hydrological model based on the reservoir inflow process during the test period to obtain effective model parameters;

[0034] Step 46 , inputting the effective model parameters into the optimized distributed hydrological model, combining the precipitation, evaporation, and LUC scenario data of the forecast period, and calculating the reservoir inflow process during the forecast period.

[0035] Furthermore, the step 5 includes the following sub-steps:

[0036] Step 51, calculating the NDVI value based on the reflectance of crops in different bands in the remote sensing database;

[0037] Step 52, arranging the crop growth stages and growth parameters of each crop in the irrigation area and the NDVI values of the corresponding stages based on the crop growth database;

[0038] Step 53: Construct a multivariate regression model of the target output and other independent variables for each crop at different growth stages. The formula is as follows:

[0039] Y i,t =β i,t,0 +β i,t,1 X i,t,1 +β i,t,2 X i,t,2 +...+β i,t,n X i,t,n +∈ i,t

[0040] Where Y i,t Characterizes the target output of the i-th crop at the t-th growth stage, X i,t,1 ,X i,t,2 ,...,X i,t,n Represents different independent variables, i and t represent different crop types and growth stages respectively, β i,t,1 ,,β i,t,2 ,...,β i,t,n Characterizes the regression coefficient of each independent variable of the i-th crop at the t-th growth stage; β i,t,0 The intercept term of the i-th crop at the t-th growth stage; i,t Characterizes the random error term of the i-th crop at the t-th growth stage;

[0041] Step 54: Use the least squares method to solve the parameters of the multiple regression model and draw a residual graph to check whether the residuals meet the normality, uniform distribution and independence assumptions. If not, readjust the multiple regression model. The objective function of the multiple regression model is to minimize the loss function:

[0042]

[0043] Step 55, calculating the SMI based on the remote sensing image data in the remote sensing database;

[0044] Step 56: construct a linear regression model using the measured soil moisture data for a certain historical period and the SMI value calculated for the image element at the location during the same period. After calculating the regression parameters, calculate the actual soil moisture content of all pixel locations during all historical observation periods.

[0045] Step 57, calculate the crop water demand based on the crop type, NDVI value and evapotranspiration, and calculate the crop water deficit based on the actual available water of the crop. Based on the crop water deficit, calculate the total water deficit of the irrigation area and the water supply required by the water source in turn as the irrigation area water deficit status data.

[0046] Furthermore, the step 6 includes the following sub-steps:

[0047] Step 61, establishing a reservoir irrigation scheduling sub-objective function, including: irrigation water supply guarantee rate target f1, irrigation area water demand damage degree target f2, reservoir storage target f3, and ecological environment protection target f4;

[0048] Step 62, based on the reservoir water inflow conditions, reservoir size, forecast period length, and irrigation area water shortage conditions, determine the constraints for reservoir scheduling, including: water quantity constraints, reservoir water storage size constraints, irrigation water supply capacity constraints for each period, water quality protection constraints, and non-negative constraints on decision variables.

[0049] Furthermore, the reservoir adaptive operation model constructed in step 6 is:

[0050]

[0051] Where, Represents the weight coefficients of each sub-objective function f1, f2, f3, f4, respectively, total It represents the total objective function of the reservoir adaptive operation model after considering the weight coefficient.

[0052] In another aspect, the present invention provides a reservoir scheduling system based on rainfall forecast and remote sensing monitoring, comprising:

[0053] Data collection module: It is used to collect basic data of the study area and build multiple basic databases;

[0054] Data preprocessing module: It is used to preprocess the basic data of multiple basic databases to obtain preprocessed data;

[0055] Weather forecast module: This module is used to construct a weather forecast model with the catchment area of the upstream reservoir with irrigation tasks as the research area, determine the forecast period that meets the forecast accuracy requirements, and forecast the weather data within the forecast period;

[0056] Reservoir inflow flow process forecast module: It is used to build a distributed hydrological model that takes into account future climate and land use or cover changes, and calculate the reservoir inflow flow process during the forecast period;

[0057] Irrigation area water shortage status simulation module: It is used to calculate the water shortage status data of the irrigation area based on remote sensing images of crops in the irrigation area;

[0058] Reservoir adaptive operation model construction module: It is used to take the reservoir inflow process in the future period within the forecast period and the water shortage status data of the irrigation area as input conditions, couple the boundary conditions and constraints, and build the reservoir adaptive operation model;

[0059] Multi-objective non-inferior solution set solving module: It is used to solve the reservoir adaptive scheduling model using the multi-objective particle swarm optimization algorithm to obtain a non-inferior solution set.

[0060] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the reservoir scheduling method based on rainfall forecast and remote sensing monitoring.

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

[0062] 1. This invention constructs a multi-level scheduling model by comprehensively analyzing multiple sources of information, including rainfall forecasts, remote sensing data, soil moisture, land use / cover data, and meteorological data. This multi-source information fusion approach is more systematic and comprehensive in information processing. Compared with traditional single-source scheduling methods, multi-source information fusion can effectively improve the accuracy and reliability of scheduling decisions, especially under complex hydrological and meteorological conditions, and can better cope with uncertainty.

[0063] 2. This paper constructs a distributed hydrological model that accounts for future climate change and land use / cover changes, used to calculate reservoir inflow processes in the future. Distributed hydrological models can more precisely simulate hydrological processes within a watershed, accounting for spatial variations in factors such as topography, soil, and vegetation. Compared to traditional centralized models, distributed models offer significant advantages in simulation accuracy and adaptability, enabling more accurate predictions of reservoir inflow.

[0064] Furthermore, the present invention employs a multi-objective particle swarm optimization algorithm (MOPSO) to solve the reservoir adaptive scheduling model and obtain a set of non-inferior solutions. Compared to traditional optimization algorithms, MOPSO boasts strong global search capabilities, rapid convergence, and high computational efficiency. Furthermore, MOPSO can simultaneously consider multiple objectives (such as flood control, irrigation, and power generation), excelling in multi-objective optimization problems. The application of this algorithm makes reservoir scheduling plans more scientific and rational, enabling optimal resource allocation while meeting diverse needs.

[0065] 1. Application of remote sensing technology in water shortage assessment in irrigation areas

[0066] This technology calculates water shortage status data in irrigation areas based on remote sensing imagery of crops. This technology enables more accurate and timely assessments of water demand in irrigation areas. Compared to traditional manual surveys or calculations based on single meteorological data, remote sensing technology provides large-scale, high-precision, real-time data, enabling more precise irrigation scheduling plans.

[0067] 2. Construction of Adaptive Scheduling Model

[0068] The adaptive reservoir operation model constructed by this invention can dynamically adjust the operation plan by coupling boundary conditions and constraints. This model demonstrates greater adaptability in addressing uncertainties such as climate change and land use / cover change. Compared with traditional fixed-rule operation models, the adaptive operation model can dynamically adjust based on real-time data and forecast information, better meeting the diverse needs of reservoir operation.

[0069] In summary, the present invention has significant innovations and technical advantages in many aspects, including multi-source information fusion, application of distributed hydrological models, introduction of optimization algorithms, application of remote sensing technology, and construction of adaptive scheduling models. It can effectively improve the scientificity, accuracy, and adaptability of reservoir scheduling, and provide strong technical support for the efficient operation of reservoirs and the rational use of water resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0071] Figure 1 A flow chart of a reservoir scheduling method based on rainfall forecast and remote sensing monitoring provided by an exemplary embodiment of the present disclosure is shown;

[0072] Figure 2 A flowchart for predicting reservoir inflows under the influence of future climate change, future land use / cover change and other human activities;

[0073] Figure 3 This is a schematic diagram of the reservoir adaptive operation model;

[0074] Figure 4 This is a flowchart of the multi-objective particle swarm optimization algorithm;

[0075] Figure 5 A schematic structural diagram of a reservoir scheduling system based on rainfall forecast and remote sensing monitoring provided by an exemplary embodiment of the present disclosure is shown;

[0076] Figure 6 It is a structural diagram of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0077] To make the above-mentioned objects, features, and advantages of the present application more clearly understood, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings. The following description sets forth many specific details to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar improvements without violating the scope of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0078] Example 1

[0079] like Figure 1 As shown in FIG, a reservoir operation method based on rainfall forecast and remote sensing monitoring includes the following steps:

[0080] Step 1. Collect basic data of the study area and build multiple basic databases; including:

[0081] Step 11: Through field surveys and visits to relevant departments within the study area, such as agriculture, water conservancy, meteorology, hydrology, and remote sensing, obtain product results from domestic and international open-source databases (such as MODIS (Moderate Resolution Imaging Spectroradiometer) and GPM (Global Precipitation Measurement) satellite observation products). Obtain historical meteorological data (such as precipitation, temperature, humidity, wind speed, and evaporation); high-resolution remote sensing data (such as vegetation, soil moisture, land use, and NDVI); reanalysis data (such as ERA5 and European Centre for Meteorological Forecasting reanalysis data); digital elevation models (DEMs) of watersheds and river system distribution; planting structure and distribution in irrigated areas; irrigation methods and water shortage status; water source conditions and canal system orientation; historical hydrological data; reservoir characteristic parameters, and other underlying data for the study area.

[0082] Step 12: Data collation, building irrigation area characteristic database, crop growth database, meteorological and hydrological database, watershed physical geography database, watershed land use database, reservoir characteristic and operation database, remote sensing database, reanalysis database, etc., to provide basic data support for subsequent analysis and modeling.

[0083] Step 2. Preprocess the basic data of multiple basic databases to obtain preprocessed data; including:

[0084] Step 21: Define reasonable threshold ranges for the data based on the historical distribution patterns of the collected data, such as the reasonable range for daily rainfall: 0 to 500 mm; the reasonable range for temperature: -30 to 50°C.

[0085] Step 22: Use the Z-score method to detect outliers in the data.

[0086] (1) Calculate the Z score for each data point using the following formula:

[0087]

[0088] Where X is the original measurement, μ is the mean of the dataset, and σ is the standard deviation of the dataset. The Z-score reveals how far away a data point is from the mean of the dataset and can be positive or negative.

[0089] (2) Setting the Z-score threshold. In the Z-score method, data points whose absolute value of the Z-score is greater than a certain threshold (e.g., 3) are usually considered outliers. This patent does not limit the value of the Z-score threshold.

[0090] (3) According to the set threshold, all data points whose absolute value of Z score is greater than the threshold are screened out. These points are the outliers in the data set.

[0091] (4) Handling of outliers: After identifying outliers, it is necessary to combine the previous and subsequent data and professional knowledge to handle the outliers. Methods for handling outliers include but are not limited to: a) Deletion: If the outliers are caused by data entry errors or measurement errors, they can be deleted. b) Retention: In some cases, outliers may represent an important feature or extreme case in the data set. In this case, you can choose to retain the outliers. c) Replacement: You can use a statistical method (such as mean, median or mode) to replace outliers to reduce their impact on the analysis results.

[0092] Step 23: Use ArcGIS software to perform projection conversion to unify the geographic information system coordinate systems of various databases in the study area, such as WGS84 or CGCS2000 coordinate systems.

[0093] In step 24, data with different temporal resolutions (e.g., hourly meteorological data and daily crop growth status data) are uniformly converted to the same time scale (e.g., daily). Higher-frequency monitoring data (e.g., hourly meteorological data) are aggregated into daily data using daily averages, while lower-frequency data (e.g., remote sensing NDVI) are interpolated to a daily frequency.

[0094] In step 25, the variable data of different scales are normalized in order to convert the data of different dimensions to the same scale, so that the features of different dimensions can be compared and analyzed on the same scale.

[0095] Step 3. Take the upstream catchment area of the reservoir with irrigation tasks as the study area, build a meteorological forecast model with the reservoir catchment area as the core, determine the forecast period that meets the forecast accuracy requirements, and forecast meteorological data within the forecast period; including:

[0096] Step 31: Use the DEM data to extract the catchment boundary of the reservoir and the distribution data of the tributaries in the upper reaches of the basin through the Hydrology tool of ArcGIS software.

[0097] Step 32 , according to the requirements of reservoir operation, in terms of spatial resolution, an appropriate spatial resolution (such as 0.1°×0.1° or 1 km×1 km) is selected based on the size of the upstream catchment area of the reservoir and the size of the irrigation area.

[0098] In step 33, the wrapper feature selection algorithm is used to extract data, automatically selecting the input data processed and verified in step 2 and inputting it into a multi-layer long short-term memory (LSTM) model.

[0099] Step 34, set the hyperparameters of the multi-layer long short-term memory network model, mainly including: dropout rate, number of LSTM stacking layers, batch processing settings, input feature dimension, hidden layer bias, dimension of hidden layer state, learning rate, and Adam optimization algorithm.

[0100] In step 35, the objective functions are respectively (1) maximizing the Nash efficiency coefficient between the model-simulated rainfall and the measured rainfall (NSE1), and (2) maximizing the Nash efficiency coefficient between the model-simulated evaporation and the measured evaporation (NSE2).

[0101] Step 36: Model Training and Optimization: First, the preprocessed dataset is randomly divided into training and test sets in a specific ratio (e.g., 6:4). When partitioning the data, special attention should be paid to ensuring sample balance to avoid the negative impact of class imbalance on model performance. Next, the LSTM model is trained using the training set, aiming to optimize the model's predictive capabilities by learning the relationship between the input sequence and the target output. During training, a Bayesian optimization algorithm is used to tune the LSTM model's hyperparameters to improve model performance. During training, the model is continuously iterated based on the hyperparameters, meaning the NSE value is calculated for the LSTM model after each iteration until the simulated performance of the LSTM model exceeds a predetermined threshold or the maximum number of iterations is reached.

[0102] Step 37: After model training is complete, the model's predictions are validated using the test set. The model's prediction accuracy is evaluated by calculating the NSE (Nearly Error Separation) between the model-predicted rainfall (or model-predicted evaporation) and the actual rainfall (or actual evaporation) data. Based on a preset evaluation threshold, a determination is made as to whether the model has achieved the required accuracy level. Furthermore, based on the evaluation results, a prediction period length is determined that meets the required accuracy requirements, and the rainfall and evaporation predictions for that period are evaluated.

[0103] Step 38, model application: Select the forecast data of the intersection period of the forecast period that meets the rainfall forecast accuracy requirement and the forecast period that meets the evaporation forecast accuracy requirement (that is, the forecast period that can simultaneously meet the rainfall and evaporation forecast requirements) as the input for scheduling model construction.

[0104] Step 4. Figure 3 As shown in Figure 2, a distributed hydrological model considering future climate and land use or cover changes is constructed to calculate the reservoir inflow process during the prediction period in step 3; this includes:

[0105] In step 41, the historically observed rainfall, evaporation and measured flow process data at the reservoir location are divided into a parameter rate period and a test period according to a certain ratio (eg, 7:3 or 6:4).

[0106] Step 42, based on the output results of Global Climate Models (GCMs) (daily precipitation, temperature, etc. series), a bias correction method is used to obtain the changes in precipitation, temperature, etc. in the basin in the future.

[0107] Step 43, based on the historical land use / cover (LUC) scenario and the land use policy information of the local land space planning, the CA-Markov model is used to predict the future land use / cover (LUC) change scenario of the basin.

[0108] In step 44, hydrological models are effective tools for calculating the flow process at a designated outlet by considering factors such as precipitation and potential evapotranspiration. Based on the characteristics of the reservoir's basin and the completeness of the data, an appropriate distributed hydrological model (such as the SWAT model or the VIC model) is selected for model parameter calibration. For each subbasin, a reasonable range is initially set for the model parameters, and a random value within this range is selected as the initial value for the parameter.

[0109] In step 45, for the data set of the calibration period, maximizing the NSE between the measured runoff value and the simulated runoff value is used as the objective function of parameter calibration, the SCEM-UA (Shuffled Complex Evolution-University of Arizona) algorithm is used to solve the objective function, and the Gelman-Rubin convergence value is less than the threshold of 1.2 to optimize the selection of the values of each parameter.

[0110] In step 46, the model parameters obtained from the periodic optimization are input into the distributed hydrological model, along with rainfall, temperature, and evaporation data from the test period, to predict the reservoir inflow during the test period. The NSE metric is also used to assess the error accuracy between the simulated and measured runoff during the test period. Specifically, if the NSE value between the predicted and measured runoff during the test period exceeds a certain threshold, the corresponding model parameters are output as the validated model parameters after verification.

[0111] In step 47, the calibrated effective model parameters are input into the hydrological model. Combined with the precipitation, temperature, evaporation, and LUC scenario data for the forecast period, the distributed hydrological model is used to calculate and output the flow process for the forecast period, which is the reservoir inflow process. The inflow processes for the observation and forecast periods are combined to form the complete reservoir inflow process.

[0112] Step 5. Calculate the water deficit status data for the irrigation area based on remote sensing images of crops in the irrigation area; including: Step 51, NDVI calculation. The growth status of crops is closely related to their water requirements, and NDVI is commonly used as an indicator of crop growth. NDVI can reflect the greenness of vegetation and thus infer crop growth. NDVI data is calculated by calculating the reflectance of different bands in remote sensing images. The formula is as follows:

[0113]

[0114] Among them, NIR x,y,m Represents the reflectivity of the near-infrared band of the image element with coordinates (x, y) in the remote sensing image at the mth time period, RED x,y,m Represents the reflectance of the red light band at the image element with coordinates (x, y) in the remote sensing image at time period m. NDVI values range from -1 to +1. Values greater than 0.2 typically indicate the presence of vegetation, and higher values indicate denser vegetation.

[0115] Step 52, based on the crop types and distribution data in the irrigation area, using existing literature, local observations and surveys, etc., collect and organize the crop growth stages (including germination, vegetative growth, flowering, filling and maturity, etc.) and growth parameters (such as yield, plant height, leaf area index, etc.) of each crop and the NDVI values of the corresponding stages.

[0116] Step 53: construct a multivariate regression model for each crop at different growth stages, including the target output (e.g., yield) and other independent variables. The formula is as follows:

[0117] Y i,t =β i,t,0 +β i,t,1 X i,t,1 +β i,t,2 X i,t,2 +…+βi,t,n X i,t,n +∈ i,t (3)

[0118] Where Y i,t represents the target output (e.g., yield) of the i-th crop at the t-th growth stage, X i,t,1 ,X i,t,2 ,…,X i,t,n Represents different independent variables (such as the yield of the crop, plant height, leaf area, NDVI time series, cumulative NDVI, temperature, precipitation, etc.), i and t represent different crop types and growth stages respectively. i,t,1 ,,β i,t,2 ,…,β i,t,n Characterizes the regression coefficient of each independent variable of the i-th crop at the t-th growth stage; β i,t,0 The intercept term of the i-th crop at the t-th growth stage; ∈ i,t Characterizes the random error term of the i-th crop at the t-th growth stage.

[0119] Step 54, model parameter initialization: given parameter β i,t,0 and β i,t,1 ,,β i,t,2 ,...,β i,t,n The initial value of , the least squares method is used to solve the model parameters, and the model calculation objective function is to minimize the following loss function:

[0120]

[0121] Where Y i,t,m and Y i,t,m Represents the measured and model-predicted output values of the i-th crop in the m-th period of the t-th growth stage.

[0122] Step 55: Select the corresponding parameter combination scheme with the minimum loss function, and set the parameter β of the scheme to i,t,0 and β i,t,1 ,,β i,t,2 ,…,β i,t,n The value is used as the regression parameter of the i-th crop at the t-th growth stage.

[0123] Step 56, draw the residual graph and check the residual ∈ i,t Whether the normality, uniform distribution and independence assumptions are met. If not, the model can be readjusted.

[0124] Step 57, SMI calculation. SMI is an indicator of soil moisture status based on remote sensing data, which reflects the relative degree of soil moisture content relative to the dry-wet boundary. SMI values closer to 1 indicate wetness, and closer to 0 indicate dryness. The calculation formula is as follows:

[0125]

[0126] Where, LST x,y,max , LST x,y,min They represent the maximum and minimum surface temperature of the image element with coordinates (x, y) under the same vegetation coverage or regional conditions; LST x,y,m Indicates the surface temperature in the current remote sensing image element at the mth period, LST x,y,max , LST x,y,min and LST x,y,m The values can be obtained from remote sensing image data.

[0127] Step 58: A regression model is constructed using the ground-based soil moisture data for a certain historical period and the SMI values calculated for the pixel at that location during the same period. After the regression parameters are calculated, the actual soil moisture content of all pixel locations during all historical observation periods is calculated.

[0128] SM x,y,m =α×SMI x,y,m +β (6)

[0129] Where SM x,y,m and SMI x,y,m Respectively represent the actual soil moisture content (e.g., volumetric moisture content, unit: %) and sudden moisture index at the image element with coordinates (x, y) during the mth time period. Actual soil moisture content is typically measured using sensors such as time domain reflectometry. α and β represent the regression coefficients. Furthermore, if no measured soil moisture value is available at the location of the image element, data from a nearby area or an image element with similar soil characteristics from the same time period is used as a surrogate.

[0130] Step 59: Calculate the water shortage in the irrigation area.

[0131] (1) Calculation of crop water deficit. The water deficit (CWD) of crops in irrigated areas is mainly determined by subtracting the crop water demand (CWR) from the actual available water for the crop. The calculation formula is as follows:

[0132]

[0133] Where, ET 0,x,y,m The crop evapotranspiration of the image element with coordinates (x, y) in time period m can be estimated by combining regional meteorological observation data with the FAO Penman-Monteith formula; K c,i,m Characterizes the growth coefficient of the i-th crop in the m-period. It is a coefficient related to the crop type and NDVI value. and θ represent the regression coefficients in the relationship between K and NDVI, which can be found in existing literature.

[0134] (2) Calculation of total water shortage in the irrigation area.

[0135] There are multiple crops in the irrigation area, and the total water shortage of the irrigation area in period m is calculated using weighted average.

[0136]

[0137] Where TCWD m Represents the total water shortage of the entire irrigation area in period m; CWD x,y,m,i A represents the water shortage of crop i in the image element (x, y) during the m period; i Represents the planting area of crop i in the image element (x, y).

[0138] (3) Calculation of water supply volume from water source.

[0139] Since there are various losses in the process of water resource transportation and distribution from the water source to the crops in the field (such as channel leakage, evaporation loss, etc.), the total amount of water that the reservoir should supply to the irrigation area can be calculated by the following formula:

[0140]

[0141] Where Q xu,m It is the amount of water that the reservoir water source should supply to the irrigation area in period m. It is the loss coefficient, which indicates the ratio of total loss to the water supply of the reservoir. Its value ranges from 0 to 1. The specific value needs to be obtained based on historical data or field measurements.

[0142] Step 6. Take the reservoir's future inflow process and irrigation area water shortage status data within the forecast period as input conditions, couple the boundary conditions and constraints, and build a reservoir adaptive operation model; including:

[0143] Step 61: Establishing a reservoir irrigation scheduling objective function, including: irrigation water supply guarantee rate objective f1, irrigation area water demand damage degree objective f2, reservoir storage objective f3, and ecological environment protection objective f4.

[0144] (1) Irrigation water supply guarantee rate target. Taking the maximization of the total water supply guarantee rate of the irrigation area during the observation and forecast periods as the objective function, the water supply capacity of the existing irrigation reservoirs should be brought into play to maximize the irrigation water supply during the observation and forecast periods, meet the crop water demand during the observation and forecast periods, and improve the irrigation water supply guarantee rate.

[0145]

[0146] Where M is the total number of time periods for calculation (including observation period and prediction period), m = 1~M; S mRepresents whether the irrigation demand of the entire irrigation area is met in time period m. If the irrigation demand is met in time period m, then S m =1; otherwise S m =0.

[0147] (2) Irrigation area water demand damage target. The water demand of crops in the irrigation area is one of the most critical targets in reservoir operation. This target is measured by minimizing the difference between the actual water supply of the reservoir and the water demand of crops in the irrigation area at each time point during the observation and forecast periods.

[0148]

[0149] Where Q g,m Represents the actual water supply from the reservoir to the irrigation area in the mth period.

[0150] (3) Reservoir storage targets. Reservoir water storage is an important goal to ensure future irrigation needs and the long-term operation of the reservoir. Excessive water storage may lead to flood risks, while insufficient water storage may lead to water resource shortages. The goal is to optimize reservoir water storage capacity.

[0151]

[0152] Where V m Represents the water storage capacity of the reservoir in the mth period.

[0153] (4) Ecological and environmental protection goals. In order to protect the water ecological environment downstream of the reservoir, the supply of ecological water should be considered. Ecological water use is usually determined by the minimum ecological flow requirement of the reservoir. If the ecological flow actually supplied by the reservoir to the downstream is greater than or equal to the minimum ecological flow, it is considered to meet the ecological and environmental requirements; otherwise, it is considered to not meet the ecological needs.

[0154]

[0155] Where C st,m represents whether the ecological flow demand of the downstream reservoir in period m is met. If the ecological flow demand of the downstream reservoir is met in period m, then C st,m =1; otherwise C st,m =0.

[0156] Investigate the stakeholders involved in the irrigation district and reservoir, and determine the weight coefficients of each objective function in step 61 through expert consultation.

[0157]

[0158] Where, Represents the weight coefficient of each target, f total Represents the total objective function after considering the weight exponent.

[0159] Step 62, determining the constraints for reservoir operation based on the reservoir water inflow conditions, reservoir size, forecast period length, irrigation area water shortage conditions, and other constraints.

[0160] (1) Water quantity constraints

[0161] The inflow and outflow of the reservoir must meet the water balance:

[0162] V m =V m-1 +(I m -Q g,m -E m -Q st,m )Δt (14)

[0163] Where V m and V m-1 Represents the reservoir water storage at the end of the mth and m-1th period respectively; I m represents the inflow of the reservoir in period m; E m , Q st,m They represent the evaporation and leakage losses and ecological flow of the reservoir in time period m respectively; Δt represents the length of the unit calculation period.

[0164] (2) Reservoir water storage capacity constraints

[0165] Reservoir storage capacity must be within the minimum and maximum storage capacity range, between the minimum and maximum allowable storage capacity. For example, for an irrigation reservoir that also serves as a flood control function, the maximum storage capacity during the flood season cannot exceed the reservoir capacity corresponding to the flood limit water level. During the non-flood season, the maximum storage capacity can be up to the beneficial storage capacity.

[0166] V m,min ≤V m ≤V m,max (15)

[0167] Where V m,min and V m,max They represent the minimum and maximum water storage capacity that the reservoir should ensure in the mth period.

[0168] (3) Constraints on irrigation water supply capacity during the period

[0169] Irrigation water supply must meet irrigation water demand while not exceeding the maximum water supply capacity:

[0170] Q g,m ≤Q g,max (16)

[0171] Where Q g,max It represents the maximum water supply capacity of the reservoir for irrigation. This value is constrained by the reservoir's overflow facilities, the scale of the irrigation canal, etc.

[0172] (4) Water quality protection constraints

[0173] Reservoir water is used for irrigation, and its water quality should meet the standard specifications (such as: "Agricultural Irrigation Water Quality Standard" (GB 5084-2021)).

[0174] C g,m ≤C g,limit (17)

[0175] Where C g,m represents the concentration of pollutants in the outflow of the reservoir at the mth period; C g,limit The maximum permissible concentration of pollutants.

[0176] (5) Non-negative constraints on decision variables

[0177] All decision variables must be non-negative:

[0178] Q g,m ,V m ,I m ,E m ,Q st,m ≥0 (18)

[0179] Step 7. Figure 4 As shown in the figure, a multi-objective particle swarm optimization algorithm is used to solve the reservoir adaptive operation model and obtain a non-inferior solution set. It includes:

[0180] Step 71 couples the reservoir operation model with a multi-objective particle swarm optimization algorithm. This algorithm is a swarm-search-based optimization algorithm that simulates the foraging behavior of a flock of birds. Each particle represents a solution, and the search space gradually approaches the optimal solution by updating the particle's position and velocity.

[0181] Step 72: Set the relevant parameters of the multi-objective particle swarm optimization algorithm: set the population size n, the maximum number of iterations T max , inertia weight ω, individual learning factor c1, group learning factor c2 and other parameters.

[0182] In step 73, each particle represents a scheduling solution, and its position represents the decision variables in the reservoir scheduling solution (such as water supply and storage capacity). Within the constraints, each particle's initial position and velocity are initialized, and the current fitness of each particle is evaluated based on the objective function. The calculated initial position is used as the initial individual optimal solution and the global optimal solution.

[0183] In step 74, the particle's fitness value is calculated based on the scheduling plan corresponding to its current position. Each particle compares its current fitness value with its own individual optimal solution. If the current fitness value is better, the individual optimal solution is updated. The solution with the best fitness value among all particles is recorded as the global optimal solution. If a particle's fitness value is better than the global optimal solution, the global optimal solution is updated. Each particle adjusts its speed based on its individual optimal solution and the global optimal solution, and updates its position accordingly. The particle's position update indicates a change in the reservoir scheduling plan, pushing the search towards the potential optimal solution area.

[0184] Step 75, boundary processing: Check whether the new position of the particle is out of bounds. If a dimension of the particle exceeds the acceptable range (for example, the water supply cannot be negative or exceeds the maximum water supply capacity), its position is restricted to within the boundary or a reasonable position is regenerated.

[0185] Step 76, repeat steps 72-75 until the number of iterations reaches the maximum number of iterations, or the fitness value change of the global optimal solution is less than the set threshold, and output the global optimal solution Gbest, that is, the optimal solution of the reservoir operation model.

[0186] Example 2

[0187] like Figure 5 As shown, this embodiment provides a reservoir scheduling system based on rainfall forecast and remote sensing monitoring, including:

[0188] Data collection module: It is used to collect basic data of the study area and build multiple basic databases;

[0189] Data preprocessing module: It is used to preprocess the basic data of multiple basic databases to obtain preprocessed data;

[0190] Weather forecast module: This module is used to construct a weather forecast model with the catchment area of the upstream reservoir with irrigation tasks as the research area, determine the forecast period that meets the forecast accuracy requirements, and forecast the weather data within the forecast period;

[0191] Reservoir inflow flow process forecast module: It is used to build a distributed hydrological model that takes into account future climate and land use or cover changes, and calculate the reservoir inflow flow process during the forecast period;

[0192] Irrigation area water shortage status simulation module: It is used to calculate the water shortage status data of the irrigation area based on remote sensing images of crops in the irrigation area;

[0193] Reservoir adaptive operation model construction module: It is used to take the reservoir inflow process in the future period within the forecast period and the water shortage status data of the irrigation area as input conditions, couple the boundary conditions and constraints, and build the reservoir adaptive operation model;

[0194] Multi-objective non-inferior solution set solving module: It is used to solve the reservoir adaptive scheduling model using the multi-objective particle swarm optimization algorithm to obtain a non-inferior solution set.

[0195] Example 3

[0196] like Figure 6 As shown, this embodiment provides a computer-readable storage medium storing a computer program, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the reservoir scheduling method based on meteorological forecasting and remote sensing monitoring.

[0197] The above is only a preferred specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any changes or replacements that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed in this application should be covered by the scope of protection of the present application.

[0198] It should be understood that parts not elaborated in detail in this specification belong to the prior art.

[0199] It should be understood that the above description of the preferred embodiment is relatively detailed and cannot be regarded as limiting the scope of protection of the patent of the present invention. Under the guidance of the present invention, ordinary technicians in this field can also make substitutions or modifications without departing from the scope of protection of the claims of the present invention, which all fall within the scope of protection of the present invention. The scope of protection requested by the present invention shall be based on the attached claims.

Claims

1. A reservoir scheduling method based on rainfall forecast and remote sensing monitoring, characterized in that: The steps include: Step 1. Collect basic data of the study area and build multiple basic databases; Step 2. Preprocess the basic data of multiple basic databases to obtain preprocessed data; Step 3. Taking the upstream catchment area of the reservoir with irrigation tasks as the study area, construct a meteorological forecast model with the reservoir catchment area as the core, determine the forecast period that meets the forecast accuracy requirements, and forecast the meteorological data for the forecast period; Step 4. Construct a distributed hydrological model that takes into account future climate and land use or cover changes, and calculate the reservoir inflow process during the forecast period in step 3; Step 5. Calculate the water shortage status data of the irrigation area based on the remote sensing image of the crops in the irrigation area; Step 6. Taking the reservoir inflow process in the future period of the forecast period and the water shortage status data of the irrigation area as input conditions, coupling the boundary conditions and constraints, and constructing the reservoir adaptive operation model; Step 7. Use the multi-objective particle swarm optimization algorithm to solve the reservoir adaptive operation model and obtain a non-inferior solution set.

2. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 1 is characterized in that: The databases in step 1 include: irrigation area characteristics database, crop growth database, meteorological and hydrological database, watershed physical geography database, watershed land use database, reservoir characteristics and operation database, remote sensing database, and reanalysis database.

3. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 2 is characterized in that: The step 2 includes the following sub-steps: Step 21, setting the basic data threshold range based on the historical distribution pattern of basic data in the study area; Step 22: Based on the threshold range of the basic data, the Z-score method is used to detect outliers in the basic data and process them; Step 23: Use ArcGIS software to perform projection conversion and unify the geographic information system coordinate system of each basic database in the study area; Step 24, converting basic data with different time resolutions into the same time scale; Step 25: normalize the basic data of different scales.

4. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 2 is characterized in that: The step 3 includes the following sub-steps: Step 31: Based on the DEM data in the watershed physical geographic database, the watershed boundary of the reservoir and the distribution data of the tributaries upstream of the watershed are extracted, and the spatial resolution is selected based on the size of the watershed area upstream of the reservoir and the size of the irrigation area; Step 32: Use the wrapper feature selection algorithm to extract the preprocessed data in step 2 and input it into the LSTM model, and set the hyperparameters of the LSTM model; In step 33, the objective functions of the LTSM model are set as follows: (1) maximizing the Nash efficiency coefficient between the simulated rainfall value and the measured rainfall value, (2) maximizing the Nash efficiency coefficient between the simulated evaporation value and the measured evaporation value; Step 34: randomly divide the preprocessed data into a training set and a test set according to a certain ratio, use the training set to train the LSTM model, and tune the hyperparameters of the LSTM model; Step 35, using the test set to evaluate the prediction effect of the model, determining a prediction period that meets the prediction accuracy requirements based on the evaluation results, and evaluating the rainfall and evaporation prediction results for the prediction period; Step 38: Select the forecast data of the intersection period of the forecast period that meets the rainfall forecast accuracy requirement and the forecast period that meets the evaporation forecast accuracy requirement as the input for the scheduling model construction.

5. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 2 is characterized in that: The 4 steps include the following sub-steps: Step 41, based on the historical rainfall and evaporation observed in the meteorological and hydrological database and the measured flow process data at the reservoir location, the data are divided into a rate period and a test period according to a certain ratio; Step 42, based on the output of the global climate model, a bias correction method is used to obtain precipitation, temperature and evaporation data for the basin prediction period; Step 43: Based on the historical land use or land cover scenarios in the watershed land use database and the land use policy information of the local land space planning, simulate the land use / land cover change scenarios in the watershed during the forecast period. Step 44: construct a distributed hydrological model that takes into account future climate and land use or land cover changes, and calibrate model parameters based on the regular data set to obtain an optimized distributed hydrological model; Step 45: inputting rainfall, temperature, and evaporation data during the test period, predicting the reservoir inflow process during the test period, and verifying the optimized distributed hydrological model based on the reservoir inflow process during the test period to obtain effective model parameters; Step 46 , inputting the effective model parameters into the optimized distributed hydrological model, combining the precipitation, evaporation, and LUC scenario data of the forecast period, and calculating the reservoir inflow process during the forecast period.

6. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 2 is characterized in that: The step 5 includes the following sub-steps: Step 51, calculating the NDVI value based on the reflectance of crops in different bands in the remote sensing database; Step 52, arranging the crop growth stages and growth parameters of each crop in the irrigation area and the NDVI values of the corresponding stages based on the crop growth database; Step 53: Construct a multivariate regression model of the target output and other independent variables for each crop at different growth stages. The formula is as follows: Y i,t =b i,t,0 +b i,t,1 X i,t,1 +b i,t,2 X i,t,2 +…+b i,t,n X i,t,n +∈ i,t Where Y i,t Characterizes the target output of the i-th crop at the t-th growth stage, X i,t,1 ,X i,t,2 ,...,X i,t,n Represents different independent variables, i and t represent different crop types and growth stages respectively, β i,t,1 ,,β i,t,2 ,...,β i,t,n Characterizes the regression coefficient of each independent variable of the i-th crop at the t-th growth stage; β i,t,0 The intercept term of the i-th crop at the t-th growth stage; ∈ i,t Characterizes the random error term of the i-th crop at the t-th growth stage; Step 54: Use the least squares method to solve the parameters of the multiple regression model and draw a residual graph to check whether the residuals meet the normality, uniform distribution and independence assumptions. If not, readjust the multiple regression model. The objective function of the multiple regression model is to minimize the loss function: Step 55, calculating the SMI based on the remote sensing image data in the remote sensing database; Step 56: construct a linear regression model using the measured soil moisture data for a certain historical period and the SMI value calculated for the image element at the location during the same period. After calculating the regression parameters, calculate the actual soil moisture content of all pixel locations during all historical observation periods. Step 57, calculate the crop water demand based on the crop type, NDVI value and evapotranspiration, and calculate the crop water deficit based on the actual available water of the crop. Based on the crop water deficit, calculate the total water deficit of the irrigation area and the water supply required by the water source in turn as the irrigation area water deficit status data.

7. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 2 is characterized in that: The step 6 includes the following sub-steps: Step 61, establishing a reservoir irrigation scheduling sub-objective function, including: irrigation water supply guarantee rate target f1, irrigation area water demand damage degree target f2, reservoir storage target f3, and ecological environment protection target f4; Step 62, based on the reservoir water inflow conditions, reservoir size, forecast period length, and irrigation area water shortage conditions, determine the constraints for reservoir scheduling, including: water quantity constraints, reservoir water storage size constraints, irrigation water supply capacity constraints for each period, water quality protection constraints, and non-negative constraints on decision variables.

8. The reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to claim 7 is characterized in that: The reservoir adaptive operation model constructed in 6 is: Where, Represents the weight coefficients of each sub-objective function f1, f2, f3, f4, respectively, total It represents the total objective function of the reservoir adaptive operation model after considering the weight coefficient.

9. A reservoir dispatching system based on rainfall forecast and remote sensing monitoring, characterized in that: include: Data collection module: It is used to collect basic data of the study area and build multiple basic databases; Data preprocessing module: It is used to preprocess the basic data of multiple basic databases to obtain preprocessed data; Weather forecast module: This module is used to construct a weather forecast model with the catchment area of the upstream reservoir with irrigation tasks as the research area, determine the forecast period that meets the forecast accuracy requirements, and forecast the weather data within the forecast period; Reservoir inflow flow process forecast module: It is used to build a distributed hydrological model that takes into account future climate and land use or cover changes, and calculate the reservoir inflow flow process during the forecast period; Irrigation area water shortage status simulation module: It is used to calculate the water shortage status data of the irrigation area based on remote sensing images of crops in the irrigation area; Reservoir adaptive operation model construction module: It is used to take the reservoir inflow process in the future period within the forecast period and the water shortage status data of the irrigation area as input conditions, couple the boundary conditions and constraints, and build the reservoir adaptive operation model; Multi-objective non-inferior solution set solving module: It is used to solve the reservoir adaptive scheduling model using the multi-objective particle swarm optimization algorithm to obtain a non-inferior solution set; The reservoir scheduling based on rainfall forecast and remote sensing monitoring is used to execute the steps in the reservoir scheduling method based on rainfall forecast and remote sensing monitoring according to any one of claims 1 to 8.

10. The present invention further provides a computer-readable storage medium storing a computer program, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method of claims 1 to 8. Any of the steps in the reservoir scheduling method based on rainfall forecast and remote sensing monitoring.

Images