A multi-source precipitation data fusion method based on meta-heuristic and machine learning

Abstract

This invention relates to the field of precipitation fusion technology and discloses a multi-source precipitation data fusion method based on metaheuristics and machine learning, comprising the following steps: collecting ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data, and performing data preprocessing; dividing the preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data into training and validation sets according to a certain ratio; setting the initial value range of hyperparameters of a multilayer perceptron model, and inputting them along with the training set into the multilayer perceptron model for training; optimizing the hyperparameters of the multilayer perceptron model using a sparrow search method to obtain a trained multilayer perceptron model; inputting the validation set into the trained multilayer perceptron model to identify the precipitation observation values ​​of the validation set, thereby obtaining the predicted ground precipitation observation data, i.e., the fused precipitation data; this method improves the accuracy of fused precipitation data while avoiding reliance on assumptions that are not valid in reality.

Description

Technical Field

[0001] This invention relates to the field of precipitation fusion technology, specifically to a method for fusing multi-source precipitation data based on metaheuristics and machine learning. Background Technology

[0002] Precipitation is the most active meteorological element in the global hydrological cycle, climate, and ecological balance. Obtaining high-precision precipitation data is crucial for improving the accuracy of hydrological modeling and flood forecasting, as well as accurately revealing the changing patterns of the water cycle. Currently, precipitation information mainly comes from three sources: ground-based rain gauge observations, satellite remote sensing observations and inversion, and radar estimation. Rain gauge observations are the most common, direct, and accurate method for acquiring precipitation information. However, due to limitations imposed by natural conditions and the number and spatial distribution of precipitation stations, precipitation observation data in some sparsely populated areas are not representative. Radar and satellite remote sensing observations can obtain effective precipitation information by detecting atmospheric physical quantities related to rainfall, to some extent compensating for the limitations of ground-based observation data in terms of spatial distribution and temporal continuity. However, affected by retrieval algorithms and sensor errors, radar and satellite remote sensing data are difficult to accurately estimate the corresponding precipitation data that changes over time. Existing research on reducing the uncertainty of precipitation data products largely focuses on fusing individual satellite precipitation products with ground-based precipitation observation data. This approach suffers from limitations and cannot accurately capture the spatial variation of precipitation. Furthermore, common methods for fusing satellite precipitation products, such as Inverse Error Variance Weighted (IEVW), Geographically Weighted Regression (GWR), and Triple Combination (TC), rely on assumptions that are not valid in reality. These methods are constrained by the density of ground observation stations, the quality of the dataset, and the strategies employed by researchers. They often fuse individual satellite precipitation products with precipitation observation data, or fuse multiple satellite precipitation products without considering observed precipitation, resulting in a limited number of fused precipitation products. In addition, machine learning algorithms are widely used in precipitation product fusion because they can effectively capture the complex nonlinear relationships between variables and target outputs, do not rely on strict statistical assumptions, and possess efficient self-learning and data processing capabilities, enabling them to handle high-dimensional and large-scale datasets. However, due to the control of hyperparameters (the setting of model hyperparameters affects the learning process and the reliability of the results) and the influence of the dataset, machine learning algorithms may experience overfitting or fail to achieve optimal fitting results. Hyperparameters are parameters set before the learning process begins, rather than being learned from training data. More accurate model hyperparameters are more conducive to accurate prediction and reducing model error. Determining hyperparameters is a challenging task. In machine learning, there are many types of model hyperparameters, and each hyperparameter has multiple potential values, which increases the search space for parameter values. However, in most studies, hyperparameters are usually manually tuned or use default values, especially on daily or smaller time scales, which greatly affects model accuracy and results in suboptimal model performance. Summary of the Invention

[0003] To address the aforementioned shortcomings in existing technologies, this invention provides a multi-source precipitation data fusion method based on metaheuristics and machine learning. This method solves the problems of existing technologies relying on assumptions that are not valid in reality, being limited by the density of ground data observation stations, the quality of datasets, the strategies adopted by researchers, the small amount of precipitation product data fused, and the low model prediction accuracy caused by inaccurate hyperparameter selection in machine learning.

[0004] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0005] A multi-source precipitation data fusion method based on metaheuristics and machine learning includes the following steps:

[0006] S1. Collect ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data;

[0007] Among them, auxiliary variable data include DEM data, latitude and longitude data, slope and aspect data, ground temperature data, and spatial autocorrelation variable data;

[0008] S2. Perform data preprocessing on ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data to obtain preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data;

[0009] S3. Divide the preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data into training set and validation set according to the proportion;

[0010] S4. Set the initial range of hyperparameter values ​​for the multilayer perceptron model and input them into the multilayer perceptron model along with the training set for training. Use the sparrow search method to optimize the hyperparameters of the multilayer perceptron model to obtain the trained multilayer perceptron model.

[0011] The hyperparameters include the number of hidden layers, the number of neurons in each hidden layer, and the learning rate.

[0012] S5. Input the validation set into the trained multilayer perceptron model to identify the precipitation observations in the validation set, and obtain the predicted surface precipitation observation data, i.e., the fused precipitation data.

[0013] Furthermore, the spatial autocorrelation variable data in step S1 are the precipitation data calculated by rain gauges based on neighboring observations using the inverse distance weighting method, i.e.:

[0014]

[0015] SAC represents the precipitation calculated by the rain gauge based on neighboring observations. w represents the squared distance from the rain gauge station to its i1th adjacent station. i1 OBS represents the weight of the i1th neighboring site. i1 This represents the precipitation at the i1th adjacent station.

[0016] Furthermore, step S2 specifically includes:

[0017] S21. The spatial resolution of the auxiliary variable data and the multi-source satellite precipitation product data is unified by the resampling method to obtain the preprocessed auxiliary variable data;

[0018] S22. Match the multi-source satellite precipitation product data with the ground precipitation observation data in terms of time and location to obtain preprocessed ground precipitation observation data and multi-source satellite precipitation product data.

[0019] Furthermore, step S3 specifically includes:

[0020] Preprocessed multi-source satellite precipitation product data and auxiliary variable data were used as raw data, and preprocessed ground precipitation observation data were used as labeled data. The labeled raw data were divided into training set and validation set according to the proportion.

[0021] Furthermore, the specific process of optimizing the hyperparameters of the multilayer perceptron model using the sparrow search method in step S4 is as follows:

[0022] Step 1: Set the initial range of hyperparameter values ​​for the multilayer perceptron model, and use the initial range of hyperparameter values ​​as the upper and lower bounds of the search space for the sparrow search method. At the same time, set the maximum number of iterations for the sparrow search method.

[0023] Step 2: Based on the upper and lower bounds of the search space, randomly generate an initial sparrow population in the search space and calculate the fitness value of each individual sparrow in the initial sparrow population.

[0024] Among them, the fitness value of each sparrow in the initial sparrow population is the performance index of the multilayer perceptron model trained by the hyperparameter combination on the validation set.

[0025] Step 3: Based on the fitness value of each sparrow in the initial sparrow population, find the best and worst sparrows at present. Use the sparrow with the best current fitness value to retain the sparrow with the best current fitness value, and take the position of the sparrow with the best current fitness value as the global best position of the sparrow population.

[0026] Step 4: Determine if the current iteration count has reached the maximum iteration count. If so, output the global best position of the sparrow population as the optimal solution; otherwise, proceed to Step 5.

[0027] The optimal solution is the optimal combination of hyperparameters.

[0028] Step 5: Based on the alarm value and safety threshold, update the location of the discoverer for the next iteration;

[0029] Step 6: Based on the position of the discoverer, update the position of the joiner in the next iteration;

[0030] Step 7: Update the position of the vigilant for the next iteration;

[0031] Step 8: After executing Steps 5-7, recalculate the fitness value of each sparrow individual, find the best and worst sparrow individuals, and retain the sparrow individual with the best current fitness value. Take the position of the sparrow individual with the best current fitness value as the global best position of the sparrow population, increment the iteration count by 1, and return to Step 4.

[0032] Furthermore, the formula for updating the discoverer's position in the next iteration is:

[0033]

[0034] Where d represents the total number of dimensions, t represents the current iteration number, and P represents the discoverer. This represents the position of the discoverer P in the j-th dimension at the (t+1)-th iteration. Let P represent the position of the discoverer P in the j-th dimension at the t-th iteration, e represent the exponential function, α represent a random number between 0 and 1, and iter max R2 represents the maximum number of iterations, ST represents the safety threshold, Q represents a random number that follows a normal distribution, and L represents a 1×d matrix with all elements equal to 1.

[0035] Furthermore, the formula for updating the position of the joiner in the next iteration is:

[0036]

[0037] Where n represents the total number of sparrows, and J represents the number of participants. This indicates the position of participant J in the j-th dimension at the (t+1)-th iteration. This represents the position of participant J in the j-th dimension at the t-th iteration. This represents the worst-case position in the t-th iteration, A represents a 1×d matrix whose elements are randomly assigned the value 1 or -1, T represents the transpose, and "otherwise" represents the condition. + This represents an intermediate temporary variable with no practical meaning.

[0038] Furthermore, the formula for updating the position of the vigilant in the next iteration is:

[0039]

[0040] In this context, V represents the vigilant. This represents the position of the vigilant V in the j-th dimension at the (t+1)-th iteration. This represents the position of the vigilant V in the j-th dimension at the t-th iteration. Let f represent the globally optimal position at the t-th iteration, β represent the step size control parameter, K represent a random number between -1 and 1, and f V f represents the fitness value of the vigilant V at the t-th iteration. g f represents the globally best fitness value at the t-th iteration. ω ε represents the worst fitness value globally at the t-th iteration, and ε represents the minimum constant to avoid a denominator of zero.

[0041] Further, in step S5, the validation set is input into the trained multilayer perceptron model to identify the precipitation observations in the validation set, thereby obtaining the predicted surface precipitation observation data. The calculation formula for the fused precipitation data is as follows:

[0042] P o =MLP(P1,P2,…,P) m ,X)

[0043] Among them, P o The MLP represents the predicted surface precipitation observation data, and P1, P2, P3 represent the trained multilayer perceptron model. m represents the precipitation product data of the 1st, 2nd and mth satellites in the preprocessing, respectively, and X represents the auxiliary variable data in the preprocessing.

[0044] The present invention has the following beneficial effects:

[0045] 1. The multi-source precipitation data fusion method based on metaheuristics and machine learning proposed in this invention uses the sparrow search method to optimize the hyperparameters of the multilayer perceptron model. It can automatically optimize the hyperparameters on a daily or smaller time scale, rather than by default or manual adjustment, thereby reducing the occurrence of overfitting in the training process of the multilayer perceptron model, so that the performance of the multilayer perceptron model reaches the optimal level and improving the prediction accuracy of the multilayer perceptron model.

[0046] 2. Using multi-source satellite precipitation product data and auxiliary variable data as inputs to the multilayer perceptron model at ground precipitation observation stations, and introducing spatial autocorrelation variable data into the auxiliary variable data, the precipitation observation data predicted by the multilayer perceptron model is better, more naturally distributed, and the prediction accuracy is improved.

[0047] 3. The introduction of the multilayer perceptron model, compared with other machine learning models, can capture more data features through its network structure, better fit the nonlinear relationship between multi-source satellite precipitation product data and auxiliary variables, and avoid relying on assumptions that do not hold true in reality, the density of ground observation stations, the quality of the dataset, and the influence of the strategies adopted by researchers, thus solving the problem of insufficient precipitation data. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating a multi-source precipitation data fusion method based on metaheuristics and machine learning proposed in this invention.

[0049] Figure 2 This is a schematic diagram illustrating the training principle of the multilayer perceptron model in the embodiment. Detailed Implementation

[0050] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0051] like Figure 1 As shown, a multi-source precipitation data fusion method based on metaheuristics and machine learning includes the following steps S1-S5:

[0052] S1. Collect ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data; among which, auxiliary variable data include DEM data, latitude and longitude data, slope and aspect data, ground temperature data, and spatial autocorrelation variable data.

[0053] In this embodiment, the collected ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data are diurnal scale data. The multi-source satellite precipitation product data includes precipitation product data collected by satellites such as IMERG_Early, GSMap_NRT, PDIR-Now, and SM2RAIN-ASCAT. The purpose of collecting multi-source satellite precipitation product data is to address the problem of insufficient precipitation product data in existing technologies. The fusion of multiple satellite precipitation products with ground observation data, compared to fusion with a single satellite precipitation product data, can more accurately capture the spatial variation of precipitation, avoiding the shortcomings of single-satellite precipitation product data. The auxiliary variable data includes DEM data, latitude and longitude data, slope and aspect data, surface temperature data, and spatial autocorrelation variable data. Among them, DEM data is digital elevation model data. The purpose of introducing auxiliary variable data is to supplement and correct the deficiencies of multi-source satellite precipitation product data, enhance the model's ability to understand and describe precipitation phenomena from multiple dimensions, and ultimately improve the accuracy and spatial resolution of precipitation prediction. At the same time, the purpose of collecting surface precipitation observation data is to use the surface precipitation observation data collected by rain gauges as label data or dependent variables, and input them into the model for training, so as to provide the model with a high-precision and reliable reference benchmark and improve the model training accuracy.

[0054] Specifically, in step S1, the spatial autocorrelation variable data are the precipitation data calculated by rain gauges based on neighboring observations using the inverse distance weighting method, i.e.:

[0055]

[0056] SAC represents the precipitation calculated by the rain gauge based on neighboring observations. w represents the squared distance from the rain gauge station to its i1th adjacent station. i1 OBS represents the weight of the i1th neighboring site. i1 This represents the precipitation at the i1th adjacent station.

[0057] In this embodiment, the purpose of collecting spatial autocorrelation variable data is to improve the model's understanding and prediction of the spatial distribution of precipitation by introducing spatial dependence information, thereby making the fused precipitation data predicted by the model more accurate and more evenly distributed.

[0058] S2. Perform data preprocessing on ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data to obtain preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data.

[0059] In this embodiment, the purpose of preprocessing the ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data is to facilitate the extraction of multi-source satellite precipitation product data values ​​corresponding to the locations of the multi-source satellite precipitation product data, auxiliary variable data, and ground precipitation observation data, as well as the data values ​​of each covariate in the auxiliary variable data (i.e., DEM data, latitude and longitude data, slope and aspect data, ground temperature data, and spatial autocorrelation variable data), so that these data can be easily input into the model for training.

[0060] Specifically, step S2 includes S21-S22:

[0061] S21. The spatial resolution of the auxiliary variable data and the multi-source satellite precipitation product data is unified by the resampling method to obtain the preprocessed auxiliary variable data.

[0062] S22. Match the multi-source satellite precipitation product data with the ground precipitation observation data in terms of time and location to obtain preprocessed ground precipitation observation data and multi-source satellite precipitation product data.

[0063] In this embodiment, the purpose of matching the time and location of multi-source satellite precipitation product data and ground precipitation observation data is to ensure that the time and location of ground precipitation observation data and multi-source satellite precipitation product data correspond, avoid errors caused by time or space misalignment, and ensure that multi-source satellite precipitation product data and ground observation data are compared and fused under the same conditions, thereby maximizing the scientificity and accuracy of the fused data.

[0064] S3. The preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data are divided into training set and validation set according to the proportion.

[0065] Specifically, step S3 includes:

[0066] Preprocessed multi-source satellite precipitation product data and auxiliary variable data were used as raw data, and preprocessed ground precipitation observation data were used as labeled data. The labeled raw data were divided into training set and validation set according to the proportion.

[0067] S4. Set the initial range of hyperparameter values ​​for the multilayer perceptron model and input them along with the training set into the multilayer perceptron model for training. Use the sparrow search method to optimize the hyperparameters of the multilayer perceptron model to obtain the trained multilayer perceptron model. The hyperparameters include the number of hidden layers, the number of neurons in each hidden layer, and the learning rate.

[0068] In this embodiment, the hyperparameters of the multilayer perceptron model include the number of hidden layers, the number of neurons in each hidden layer, and the learning rate. Since the setting of hyperparameters affects the reliability of the learning process and results of the multilayer perceptron model, setting the optimal hyperparameter combination can prevent overfitting during training and achieve the best fitting effect. In this embodiment, the sparrow search algorithm has a simple structure and fewer parameters, making the algorithm easy to understand and implement. It also has strong global search capabilities, excellent ability to escape local optima, and fast convergence speed, exhibiting good adaptability and high robustness in handling various optimization problems. Therefore, this invention uses the sparrow search algorithm to optimize the hyperparameters, thereby obtaining the optimal hyperparameter combination. Using the optimal hyperparameter combination and the training set to train the multilayer perceptron model, a trained multilayer perceptron model can be obtained for predicting fused precipitation data.

[0069] Specifically, the process of optimizing the hyperparameters of the multilayer perceptron model using the sparrow search method in step S4 is as follows:

[0070] Step 1: Set the initial range of hyperparameter values ​​for the multilayer perceptron model, and use the initial range of hyperparameter values ​​as the upper and lower bounds of the search space for the sparrow search method. At the same time, set the maximum number of iterations for the sparrow search method.

[0071] In this embodiment, the parameters of the sparrow search method include population size, maximum number of iterations, and upper and lower bounds of the search space (i.e., the range of hyperparameter values), so as to be used for hyperparameter optimization of the multilayer perceptron model.

[0072] Step 2: Based on the upper and lower bounds of the search space, an initial sparrow population is randomly generated in the search space, and the fitness value of each sparrow in the initial sparrow population is calculated; where the fitness value of each sparrow in the initial sparrow population is the performance index of the multilayer perceptron model trained by the hyperparameter combination on the validation set.

[0073] Step 3: Based on the fitness value of each sparrow in the initial sparrow population, find the best and worst sparrow individuals at present. Use the sparrow individual with the best current fitness value to retain the sparrow individual, and take the position of the sparrow individual with the best current fitness value as the global best position of the sparrow population.

[0074] Step 4: Determine if the current iteration count has reached the maximum iteration count. If so, output the global best position of the sparrow population as the optimal solution. Otherwise, proceed to Step 5. The optimal solution is the optimal combination of hyperparameters.

[0075] Step 5: Based on the alarm value and the safety threshold, update the location of the discoverer for the next iteration, i.e.:

[0076]

[0077] Where d represents the total number of dimensions, t represents the current iteration number, and P represents the discoverer. This represents the position of the discoverer P in the j-th dimension at the (t+1)-th iteration. Let P represent the position of the discoverer P in the j-th dimension at the t-th iteration, e represent the exponential function, α represent a random number between 0 and 1, and iter max R2 represents the maximum number of iterations, ST represents the safety threshold, Q represents a random number that follows a normal distribution, and L represents a 1×d matrix with all elements equal to 1.

[0078] Step 6: Based on the discoverer's position, update the position of the joiner in the next iteration, i.e.:

[0079]

[0080] Where n represents the total number of sparrows, and J represents the number of participants. This indicates the position of participant J in the j-th dimension at the (t+1)-th iteration. This represents the position of participant J in the j-th dimension at the t-th iteration. This represents the worst-case position in the t-th iteration, A represents a 1×d matrix whose elements are randomly assigned the value 1 or -1, T represents the transpose, and "otherwise" represents the condition. + This represents a temporary variable with no practical meaning. Step 7: Update the position of the watcher in the next iteration, i.e.:

[0081]

[0082] In this context, V represents the vigilant. This represents the position of the vigilant V in the j-th dimension at the (t+1)-th iteration. This represents the position of the vigilant V in the j-th dimension at the t-th iteration. Let f represent the globally optimal position at the t-th iteration, β represent the step size control parameter, K represent a random number between -1 and 1, and f V f represents the fitness value of the vigilant V at the t-th iteration. g f represents the globally best fitness value at the t-th iteration. ω ε represents the worst fitness value globally at the t-th iteration, and ε represents the minimum constant to avoid a denominator of zero.

[0083] Step 8: After executing Steps 5-7, recalculate the fitness value of each sparrow individual, find the best and worst sparrow individuals, and retain the sparrow individual with the best current fitness value. Take the position of the sparrow individual with the best current fitness value as the global best position of the sparrow population, increment the iteration count by 1, and return to Step 4.

[0084] S5. Input the validation set into the trained multilayer perceptron model to identify the precipitation observations in the validation set, and obtain the predicted surface precipitation observation data, i.e., the fused precipitation data.

[0085] Specifically, in step S5, the validation set is input into the trained multilayer perceptron model to identify the precipitation observations in the validation set, thereby obtaining the predicted surface precipitation observation data. The calculation formula for the fused precipitation data is as follows:

[0086] P o =MLP(P1,P2,…,P) m ,X)

[0087] Among them, P o The MLP represents the predicted surface precipitation observation data, and P1, P2, P3 represent the trained multilayer perceptron model. m represents the precipitation product data of the 1st, 2nd and mth satellites in the preprocessing, respectively, and X represents the auxiliary variable data in the preprocessing.

[0088] In this embodiment, the training set and optimal hyperparameters are input into the multilayer perceptron model for training, and the training process is as follows: Figure 2 As shown, Figure 2 The input data for the multilayer perceptron model includes preprocessed multi-source satellite precipitation product data and auxiliary variable data, ground precipitation observation data, and optimal hyperparameters. After training, the multilayer perceptron model can output fused precipitation products, i.e., predicted ground precipitation observation data. The multilayer perceptron model consists of an input layer, hidden layers, and an output layer, with each layer containing multiple neurons connected by weights. The input layer receives the input data; each input neuron corresponds to a feature of the input data, meaning the number of neurons in the input layer equals the number of features in the input data. The hidden layer captures the complex interactions between the input data (input features). The output layer outputs the predicted data; each neuron in the output layer corresponds to a possible category or predicted value. Figure 2 Both LST-Night and LST-Day in the data are ground temperature data.

[0089] S6. Input the validation set into the trained multilayer perceptron model to identify the precipitation observations in the validation set, and obtain the predicted surface precipitation observation data, i.e., the fused precipitation data.

[0090] Specifically, in step S6, the validation set is input into the trained multilayer perceptron model to identify the precipitation observations in the validation set, thereby obtaining the predicted surface precipitation observation data. The calculation formula for the fused precipitation data is as follows:

[0091] P o =MLP(P1,P2,…,P) m ,X)

[0092] Among them, P o The MLP represents the predicted surface precipitation observation data, and P1, P2, P3 represent the trained multilayer perceptron model. m All represent preprocessed multi-source satellite precipitation product data, and X represents preprocessed auxiliary variable data.

[0093] In summary, the multi-source precipitation data fusion method based on metaheuristics and machine learning proposed in this invention firstly optimizes the hyperparameters of the multilayer perceptron model using a sparrow search method. This automatic optimization, performed on daily or smaller time scales rather than by default or manual adjustment, reduces overfitting during training, optimizes the model's performance, and improves its prediction accuracy. Secondly, multi-source satellite precipitation product data and auxiliary variable data are used as inputs to the multilayer perceptron model at ground precipitation observation stations. The introduction of spatial autocorrelation variables into the auxiliary variable data further enhances the accuracy and naturalness of the final precipitation observation data predicted by the multilayer perceptron model. Furthermore, compared to other machine learning models, the multilayer perceptron model can capture more data features through its network structure, better fit the nonlinear relationship between precipitation and auxiliary variables, and avoid reliance on unfounded assumptions, the density of ground observation stations, dataset quality, and the strategies employed by researchers, thus addressing the problem of insufficient precipitation data.

[0094] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

[0095] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A multi-source precipitation data fusion method based on meta-heuristic and machine learning, characterized in that, Includes the following steps: S1. Collect ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data; Among them, auxiliary variable data include DEM data, latitude and longitude data, slope and aspect data, ground temperature data, and spatial autocorrelation variable data; S2. Perform data preprocessing on ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data to obtain preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data; S3. Divide the preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data into training set and validation set according to the proportion; S4. Set the initial range of hyperparameter values ​​for the multilayer perceptron model and input them into the multilayer perceptron model along with the training set for training. Use the sparrow search method to optimize the hyperparameters of the multilayer perceptron model to obtain the trained multilayer perceptron model. The hyperparameters include the number of hidden layers, the number of neurons in each hidden layer, and the learning rate. The specific process of optimizing the hyperparameters of the multilayer perceptron model using the sparrow search method in step S4 is as follows: Step 1: Set the initial range of hyperparameter values ​​for the multilayer perceptron model, and use the initial range of hyperparameter values ​​as the upper and lower bounds of the search space for the sparrow search method. At the same time, set the maximum number of iterations for the sparrow search method. Step 2: Based on the upper and lower bounds of the search space, randomly generate an initial sparrow population in the search space and calculate the fitness value of each individual sparrow in the initial sparrow population. Among them, the fitness value of each sparrow in the initial sparrow population is the performance index of the multilayer perceptron model trained by the hyperparameter combination on the validation set. Step 3: Based on the fitness value of each sparrow in the initial sparrow population, find the best and worst sparrows at present. Use the sparrow with the best current fitness value to retain the sparrow with the best current fitness value, and take the position of the sparrow with the best current fitness value as the global best position of the sparrow population. Step 4: Determine if the current iteration count has reached the maximum iteration count. If so, output the global best position of the sparrow population as the optimal solution; otherwise, proceed to Step 5. The optimal solution is the optimal combination of hyperparameters. Step 5: Based on the alarm value and safety threshold, update the location of the discoverer for the next iteration; Step 6: Based on the position of the discoverer, update the position of the joiner in the next iteration; Step 7: Update the position of the vigilant for the next iteration; Step 8: After executing Steps 5-7, recalculate the fitness value of each sparrow individual, find the best and worst sparrow individuals, and retain the sparrow individual with the best current fitness value. Take the position of the sparrow individual with the best current fitness value as the global best position of the sparrow population, increment the iteration count by 1, and return to Step 4. S5. Input the validation set into the trained multilayer perceptron model to identify the precipitation observations in the validation set, and obtain the predicted surface precipitation observation data, i.e., the fused precipitation data.

2. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 1, characterized in that, In step S1, the spatial autocorrelation variable data are the precipitation data calculated by rain gauges based on neighboring observations using the inverse distance weighting method, i.e.: wherein, represents the precipitation calculated by the rain gauge from the neighboring observations, represents the square of the distance from the rain gauge to the th neighboring site, represents the weight of the th neighboring site, represents the precipitation at the th neighboring site.

3. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 2, characterized in that, Step S2 specifically includes: S21. The spatial resolution of the auxiliary variable data and the multi-source satellite precipitation product data is unified by the resampling method to obtain the preprocessed auxiliary variable data; S22. Match the multi-source satellite precipitation product data with the ground precipitation observation data in terms of time and location to obtain preprocessed ground precipitation observation data and multi-source satellite precipitation product data.

4. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 3, characterized in that, Step S3 specifically includes: Preprocessed multi-source satellite precipitation product data and auxiliary variable data were used as raw data, and preprocessed ground precipitation observation data were used as labeled data. The labeled raw data were divided into training set and validation set according to the proportion.

5. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 1, characterized in that, The formula for updating the discoverer's position in the next iteration is: in, Indicates the total number of dimensions. Indicates the current iteration number. Indicates the discoverer, Indicates the discoverer In the During the nth iteration, at the... Position in each dimension Indicates the discoverer In the During the nth iteration, at the... Position in each dimension Represents an exponential function. Represents a random number between 0 and 1. Indicates the maximum number of iterations. Indicates the alarm value. Indicates the safety threshold. Represents a random number that follows a normal distribution. Indicates that all elements are 1 matrix.

6. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 5, characterized in that, The formula for updating the position of the joiner in the next iteration is: in, This represents the total number of sparrows. Indicates the participants, Indicates the joiner In the During the nth iteration, at the... Position in each dimension Indicates the joiner In the During the nth iteration, at the... Position in each dimension Indicates the first The worst-case position in the next iteration This indicates that the element is randomly assigned the value 1 or -1. The matrix, Indicates transpose. Indicates the judgment condition. This represents an intermediate temporary variable with no practical meaning.

7. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 6, characterized in that, The formula for updating the position of the watchdog in the next iteration is: in, Those who are warning, Indicating a warning In the During the nth iteration, at the... Position in each dimension Indicating a warning In the During the nth iteration, at the... Position in each dimension Indicates the first The global optimal position at the next iteration. This indicates the step size control parameter. Represents a random number between -1 and 1. Indicates the first Vigilant during the next iteration fitness value, Indicates the first The best fitness value globally in the next iteration. Indicates the first The worst fitness value globally at the next iteration. This represents the smallest constant that avoids a denominator of zero.

8. The multi-source precipitation data fusion method based on meta-heuristic and machine learning according to claim 7, characterized in that, In step S5, the validation set is input into the trained multilayer perceptron model to identify the precipitation observations in the validation set, thereby obtaining the predicted surface precipitation observation data. The calculation formula for the fused precipitation data is as follows: in, This represents the predicted surface precipitation observation data. This represents a trained multilayer perceptron model. , , These represent the first, second, and third preprocessing steps, respectively. Precipitation product data from one satellite, This represents the preprocessed auxiliary variable data.

Images