Determination method for determining optimal operation parameters of model based on historical operation data

Abstract

The invention relates to the technical field of water conservancy professional model application, in particular to a method for determining optimal model operation parameters based on historical operation data, and the method comprises the steps: obtaining historical operation data generated in the operation process of a water conservancy professional model; extracting the historical operation data to obtain key features; establishing a nonlinear mapping relation between parameter configuration of the water conservancy professional model and a model operation result based on a mathematical correlation model established by a machine learning mechanism; determining an optimization mechanism for optimizing parameter configuration based on the historical operation data, the key features, a water conservancy professional model and a nonlinear mapping relationship; performing global search on the parameter configuration by using an optimization mechanism, and performing iteration to obtain the optimal parameter configuration of the water conservancy professional model; and according to the optimal parameter configuration, obtaining an optimized water conservancy professional model for deeper application. Different optimization mechanisms are allocated for different water conservancy professional models, the generalization ability of parameter optimization is effectively improved, and different requirements are met.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy professional model application, and in particular to a method for determining optimal model operation parameters based on historical operation data. Background Art

[0002] In the water conservancy industry, the importance of models is increasing in many aspects such as reservoir optimization scheduling, safety assessment, and water quality management.

[0003] The setting of model parameters directly impacts operational performance and result accuracy. Currently, most model parameter selection relies on empirical settings or simple tentative adjustments, lacking systematic analysis and optimization tools. This can make it difficult to test all possible scenarios based solely on staff experience, and it can be impossible to identify the parameter combination that optimizes model performance. Furthermore, parameter settings by individuals with different experience levels can vary significantly, leading to significant fluctuations in model performance, reducing the reliability of model results and increasing decision-making risks for users. Summary of the Invention

[0004] (1) Technical issues to be resolved

[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for determining the optimal parameters of model operation based on historical operation data, which solves the technical problem in the existing reservoir event management that model parameters are often set based on staff, resulting in significant fluctuations in model performance and reduced reliability.

[0006] (2) Technical solution

[0007] In order to achieve the above objectives, the main technical solutions adopted by the present invention include:

[0008] In the first aspect, the data processing of emergency dispatch events proposed in an embodiment of the present invention includes: applied to the field of application of water conservancy professional models, the method includes: obtaining historical operation data generated by the water conservancy professional model during operation; extracting the historical operation data to obtain key features; establishing a nonlinear mapping relationship between the parameter configuration of the water conservancy professional model and the model operation results based on a mathematical association model established by a machine learning mechanism; determining an optimization mechanism for optimizing the parameter configuration based on the historical operation data, the key features, the water conservancy professional model and the nonlinear mapping relationship; using the optimization mechanism to perform a global search on the parameter configuration, and iteratively obtain the optimal parameter configuration of the water conservancy professional model; according to the optimal parameter configuration, obtaining the optimized water conservancy professional model.

[0009] Optionally, when the optimization mechanism is a particle swarm optimization mechanism, the optimization mechanism is used to perform a global search on the parameter configuration, and the optimal parameter configuration of the water conservancy professional model is iteratively obtained, including: in any iteration, obtaining the current position and update speed of the current parameter configuration in the parameter space; based on the current position and the update speed, updating the parameter configuration; using the objective function to evaluate the fitness of the updated parameter configuration; wherein, the fitness is used to characterize the quality of the updated parameter configuration; the objective function includes at least one of the following: mean square error, mean absolute error; when the updated parameter configuration meets the preset conditions, the iteration is stopped to obtain the optimal parameter configuration; wherein, the preset conditions include at least one of the following: reaching the maximum number of iterations, and the fitness meets the first preset threshold.

[0010] Optionally, the method also includes: in any iteration, obtaining a current water conservancy professional model based on the current parameter configuration; using the current water conservancy professional model to process the training data to obtain a training result; when the error between the training result and the actual data is greater than a second preset threshold, adjusting the search direction of the optimization mechanism.

[0011] Optionally, the method further includes: utilizing the Pareto mechanism to obtain the optimal compromise value of each parameter in the water conservancy professional model; utilizing a multi-objective optimization mechanism to iterate the optimal compromise value of each parameter to obtain the optimal parameter configuration.

[0012] Optionally, the method also includes: using a machine learning mechanism to determine the key parameters in the water conservancy professional model; generating a parameter performance evaluation matrix based on the key parameters and the historical operation data; wherein the parameter performance evaluation matrix is ​​used to characterize the degree of influence of each key parameter on the water conservancy professional model; using the parameter performance evaluation matrix to evaluate the optimized water conservancy professional model to obtain an evaluation result.

[0013] Optionally, extracting the historical operation data to obtain key features includes: reducing the dimension of the historical operation data using a principal component analysis mechanism or a recursive feature elimination mechanism to obtain processed data; and extracting the processed data to obtain the key features.

[0014] Optionally, the optimization mechanism includes at least one of the following: a genetic mechanism, a particle swarm optimization mechanism, and a simulated annealing mechanism; the machine learning mechanism includes at least one of the following: a support vector machine mechanism and a neural network mechanism.

[0015] Optionally, the historical operation data includes at least one of the following: input, output, running time, resource consumption, and exception records.

[0016] Optionally, the method further includes: visually outputting the index information of the water conservancy professional model using a visualization tool; wherein the parameter information includes at least one of the following: a parameter optimization trend graph, a performance difference comparison graph, and an adaptability evaluation graph.

[0017] Optionally, the method is applied to a distributed processing system.

[0018] (3) Beneficial effects

[0019] The beneficial effects of the present invention are as follows: the present invention provides a method for determining the optimal parameters for model operation based on historical operation data, the present invention obtains the historical operation data generated by the water conservancy professional model during operation; extracts the historical operation data to obtain key features; establishes a nonlinear mapping relationship between the parameter settings of the water conservancy professional model and the model operation results based on a mathematical association model established by a machine learning mechanism; determines an optimization mechanism for optimizing the parameter configuration based on the historical operation data, the key features, the mathematical association model and the nonlinear mapping relationship; utilizes the optimization mechanism to perform a global search on the parameter configuration, and iteratively obtains the optimal parameter configuration of the water conservancy professional model; obtains the optimized water conservancy professional model based on the optimal parameter configuration. In this way, different optimization mechanisms can be assigned to different application scenarios in a targeted manner. Due to different needs, the required models are also different. Such a setting can effectively improve the generalization ability of parameter optimization and adapt to different needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a method for determining optimal model operation parameters based on historical operation data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] At present, the importance of models is increasing in many aspects such as reservoir optimization scheduling, safety assessment, and water quality management.

[0022] The setting of model parameters directly impacts operational performance and result accuracy. Currently, most model parameter selection relies on empirical settings or simple tentative adjustments, lacking systematic analysis and optimization tools. This can make it difficult to test all possible scenarios based solely on staff experience, and it can be impossible to identify the parameter combination that optimizes model performance. Furthermore, parameter settings by individuals with different experience levels can vary significantly, leading to significant fluctuations in model performance, reducing the reliability of model results and increasing decision-making risks for users.

[0023] In order to solve the above problems, the present invention provides a method for determining optimal parameters of model operation based on historical operation data.

[0024] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0025] The present invention obtains the historical operation data generated by the water conservancy professional model during operation; extracts the historical operation data to obtain key features; establishes a nonlinear mapping relationship between the parameter settings of the water conservancy professional model and the model operation results based on a mathematical association model established by a machine learning mechanism; determines an optimization mechanism for optimizing the parameter configuration based on the historical operation data, the key features, the mathematical association model and the nonlinear mapping relationship; uses the optimization mechanism to perform a global search on the parameter configuration and iteratively obtain the optimal parameter configuration of the water conservancy professional model; and obtains the optimized water conservancy professional model based on the optimal parameter configuration. In this way, different optimization mechanisms can be assigned to different application scenarios in a targeted manner. Due to different needs, the required models are also different. Such a setting can effectively improve the generalization ability of parameter optimization and adapt to different needs.

[0026] For easier understanding, please refer to Figure 1 The present invention provides a method for determining optimal model operation parameters based on historical operation data, which is applied to the field of reservoir operation management. Specifically, the present invention can be applied to a reservoir data management platform or system. The method may include:

[0027] S101, obtaining historical operation data generated during the operation of the water conservancy professional model.

[0028] The historical operation data may include but is not limited to at least one of the following: input, output, operation time, resource consumption, and abnormal records.

[0029] Specifically, based on the input data in the historical operation data, relevant information about the reservoir can be obtained, including but not limited to: water flow, water demand, evaporation, precipitation, etc. By analyzing the key characteristics of the input data, the importance of different information can be obtained, so that adjustment directions and adjustment focuses can be provided for subsequent parameter optimization based on the importance.

[0030] The output of the model can directly reflect the operating effect of the water conservancy professional model. According to the output and parameters, the impact on the water conservancy professional model under the current parameter configuration can be obtained.

[0031] The operation time data can record the operation status of the water conservancy professional model in different time periods. By analyzing the key characteristics of the operation time, including but not limited to response time, model operation cycle and other data, the impact of different operations on the reservoir status can be obtained, thereby obtaining subsequent optimization solutions.

[0032] It should be noted that the present invention does not specifically limit the type of water conservancy professional models, including but not limited to at least one of the following: water balance model, water quality model, hydraulic model, and ecological model.

[0033] S102: Extract the historical operation data to obtain key features.

[0034] The key features involved in this step can be used to reflect the operating conditions and patterns of the reservoir. Key features play an important role in the parameter configuration of water conservancy models. For example, in a water balance model constructed using a neural network model, key features can be introduced during the initial parameter configuration phase. This makes the model's parameters highly sensitive to the key features from the outset, thereby accelerating the model's convergence.

[0035] Key features can also be used to guide subsequent optimization mechanisms in refining model parameters. By analyzing the relationship between key features and model parameters, we can understand which parameters have the greatest impact on model performance. For example, in a linear regression model, the mean squared error of the model's results is a key feature. If this error is large, it means that the model's weight parameters are improperly set and need to be adjusted.

[0036] S103, establishing a nonlinear mapping relationship between the parameter configuration of the water conservancy professional model and the model operation results based on the mathematical association model established by the machine learning mechanism.

[0037] Specifically, by establishing a nonlinear mapping relationship between the model operation results and the parameter configuration, the connection and rules between the model parameters and results can be accurately characterized. In the subsequent parameter adjustment process, the changes in the operation results can be displayed as the parameters change, thereby more realistically reflecting the actual operation of the water conservancy professional model and improving the parameter optimization efficiency.

[0038] It should be understood that the present invention is not limited to the specific type of mathematical association model, including but not limited to: artificial neural network model, decision tree model, support vector machine model, random forest model. The machine learning mechanism includes at least one of the following: support vector machine mechanism, neural network mechanism.

[0039] S104: Determine an optimization mechanism for optimizing the parameter configuration based on the historical operation data, the key characteristics, the water conservancy professional model, and the nonlinear mapping relationship.

[0040] Specifically, the complexity of the water conservancy model can be obtained from the type of historical operating data and / or water conservancy model. For example, if the input of the model in the historical operating data includes multiple operating conditions (such as normal operation, design flood, freezing, and emptying) and conditions (such as data of reservoirs in different seasons and water levels), and the input data presents complex change patterns and interrelationships, then a relatively complex model is needed to describe and fit the data. The water conservancy model itself can also show the complexity of the model. For example, if the model is mainly used to analyze the key elements and interrelationships of the reservoir and the model does not involve precise numerical calculations, the model can be considered to be relatively simple. When the model is more complex, it can be considered that the parameter space of the model is more complex. Genetic mechanisms can be used to find a better parameter combination in the complex parameter space to make the water conservancy model more in line with the actual operation of the reservoir.

[0041] The parameter configuration includes multiple model parameters, which can be understood as the structure and complexity of the model. It should be understood that the types of parameters vary depending on the type of water conservancy model. Taking the water balance model constructed by the neural network model as an example, the parameters at this time may include but are not limited to: the number of layers, the number of neurons in each layer, the connection method, and the feature weight. Taking the water balance model constructed by the decision tree model as an example, the parameter configuration may include but is not limited to: depth and number of branches.

[0042] As can be seen from the above, different models have different characteristics and requirements. Therefore, the present invention uses different optimization mechanisms to optimize model parameters for different model complexities. That is to say, for models with lower complexity, a simpler optimization mechanism is used, and for models with higher complexity, a more complex optimization mechanism is used. For example, for a hydraulic model constructed with a linear regression model, the model parameters can be optimized using the least squares method. For a water balance model constructed with a neural network model, a particle swarm optimization mechanism can be used for optimization. In this way, a simple optimization mechanism can be provided for models of different complexities, which can not only improve the performance of the model, but also reduce the consumption of computing resources and improve the operating efficiency of the model. Among them, the optimization mechanism includes at least one of the following: genetic mechanism, particle swarm optimization mechanism, simulated annealing mechanism, and least squares mechanism.

[0043] S105 , performing a global search on the parameter configuration using the optimization mechanism, and iteratively obtaining the optimal parameter configuration of the water conservancy professional model.

[0044] After obtaining the optimization mechanism, the parameter configuration can be globally searched in the parameter space.

[0045] For ease of understanding, the following explanation will be provided with specific examples. Consider the optimization of a water balance model based on a neural network model using a particle swarm optimization mechanism. Optionally, an optimization objective can be defined first. Specifically, the model parameters can be optimized by reducing the error between the output of the neural network model and actual observations. The actual observations here can be understood as actual data from reservoir operation or the results obtained using a well-performing model. The specific choice can be made based on actual circumstances. The error can be, for example, mean squared error (MSE), mean absolute error (MAE), or other error types. When the MSE error type is used, the error can be applied to values ​​where the results obtained by the water conservancy model under the current parameter configuration deviate significantly from the actual observations. This will encourage the model to pay more attention to these large errors and allow for more targeted parameter adjustments to reduce their impact on the model's overall performance, thereby improving model accuracy. For complex and changing operational environments, such as those caused by climate change and human factors, the MEA can better address diverse operating conditions and data variations, providing reliable prediction and decision support. Users can choose the MSE based on their actual circumstances, and this is not a limitation of the present invention.

[0046] Next, parameters can be initialized based on the type of reservoir treatment model and the reservoir's historical data. Specifically, in particle swarm optimization, the particle's velocity and position are initialized to search for the optimal solution in the parameter space of the water conservancy model. Each particle here represents a complete set of parameter combinations in the reservoir treatment model, and the particle's initial position is the position of the initial parameter configuration in the parameter space. For example, a particle swarm includes three particles, A1, A2, and A3, whose positions in the parameter space are (1,1), (0,2), and (2,0), respectively. These are the initial positions of the three particles.

[0047] Next, the particles are moved and iterated in the parameter space to find the optimal parameter set that makes the objective function (such as minimizing the error between the model output and the actual observation) reach the optimal value. In each iteration, each particle will update its speed according to its current position and historical optimal position, and update its position in the search space. After each position update, its fitness is evaluated by the objective function, and the particles select the optimal path by comparing the fitness. For example, the initial velocities of the above three particles are V1 = (0.2, 0.2), V2 = (0.1, 0.1), V3 = (0.3, 0.3), and the actual observation points are (1, 3), (2, 5), and (3, 7). If the objective function is the mean square error, the outputs of A1, A2, and A3 are 2, 3, and 4 respectively, then according to the mean square error formula Where y1 is the output of the model obtained according to the parameter configuration, y2 is the actual observation value, and N is an integer greater than 1. Thus, particle A1 is obtained, and the errors for the three observation points are 1, 4, and 9 respectively. Its fitness is 4.67. The fitness of A2 and A3 are 1 and 11.67 respectively. Thus, A2 is the historical optimal position. Next, the speed can be updated. Optionally, the speed can be updated using a random number. Taking random numbers of 0.6 and 0.8 as an example, the update speed of A1 is (-1.04, 1.36), and the update position of A1 is (-0.04, 2.36). The fitness of the updated position is calculated, and so on, the updated positions of A2 and A3 are obtained, thereby obtaining the optimal position. Finally, when the updated parameter configuration meets the preset conditions, the iteration is stopped to obtain the optimal parameter configuration; wherein the preset conditions include at least one of the following: reaching the maximum number of iterations, and the fitness meeting the first preset threshold.

[0048] In addition, the present invention also provides an example of optimizing a hydraulic model built based on a convolutional neural network using a genetic mechanism. For example, the convolution layer of the convolutional neural network: uses 4 3x3 convolution kernels. Pooling layer: 2x2 maximum pooling. Fully connected layer: flatten the output of the pooling layer and connect it to a fully connected layer containing 10 neurons. Output layer: use the Softmax activation function for classification. Randomly generate 10 individuals, each individual contains the weights of 4 3x3 convolution kernels of the convolution layer (a total of 4x3x3=36 parameters) and the weights of the fully connected layer (assuming that the output of the pooling layer has 144 elements after flattening and is connected to 10 neurons, then the weights of the fully connected layer have 144x10=1440 parameters), so each individual has a total of 36+1440=1476 parameters.

[0049] For each individual (i.e., a set of CNN model parameters), apply it to the CNN model, perform forward propagation on the training set, and calculate the model's accuracy as the fitness value. For example, if the first individual has an accuracy of 0.85 on the training set, its fitness value is 0.85. Use the roulette wheel selection method to select the parent individual based on the individual's fitness value. Assume that the fitness values ​​of individuals in the population are [0.85, 0.82, 0.78, 0.90, 0.75, 0.88, 0.80, 0.83, 0.77, 0.86], calculate the selection probability of each individual, and then randomly select based on the selection probability.

[0050] Next, a crossover operation is performed on the selected parent individuals to generate offspring individuals. Assuming the crossover point is randomly selected at the 500th parameter, the parameters of the two parent individuals before and after the crossover point are swapped to generate two offspring individuals. A mutation operation is performed on the offspring individuals to increase the diversity of the population. Some parameters are randomly selected with a mutation probability of 0.1 and subjected to small random perturbations. For example, the 200th parameter is randomly selected and its value is increased or decreased by a small random number. Finally, some of the parent individuals are replaced with offspring individuals to form a new population. This iteration is repeated until the maximum number of iterations is reached, resulting in the offspring with the highest fitness, that is, the optimal parameter configuration.

[0051] S106, obtaining the optimized water conservancy professional model according to the optimal parameter configuration to apply the water conservancy professional model.

[0052] In this way, the water conservancy model, derived from the optimally configured parameters, can produce more accurate predictions, improving accuracy and efficiency. Furthermore, by optimizing parameter configuration, reservoir operation and management can be made more scientific and rational, reducing unnecessary operations and losses. For example, by rationally scheduling equipment operating hours and maintenance plans, equipment wear and repair costs can be reduced; optimizing water resource scheduling can reduce water loss and waste caused by improper scheduling, thereby reducing reservoir operating costs.

[0053] Optionally, before extracting the historical operation data (i.e., step S102), it can be preprocessed, and the preprocessing includes but is not limited to: removing outliers that exceed a reasonable range, and deleting missing or duplicate records in the data. Specifically, the historical operation data can be standardized and normalized first so that all data are on the same scale for easy comparison and analysis. It should be noted that the present invention does not specifically limit the standardization method, including but not limited to: z-score standardization, min-max normalization. Optionally, missing values ​​can also be filled by interpolation, mean filling or other suitable methods. Avoid incomplete feature extraction due to data omissions.

[0054] Optionally, in any iteration of step S105, a current water conservancy model is obtained based on the current parameter configuration; the training data is processed using the current water conservancy model to obtain a training result; when the error between the training result and the actual data is greater than a second preset threshold, it indicates that the prediction result of the current water conservancy model deviates significantly from the actual situation, and the accuracy of the model needs to be improved. At this time, it is necessary to adjust the search direction of the optimization mechanism. Adjusting the search direction of the optimization mechanism allows the model to be optimized in a more accurate direction, thereby preventing the algorithm from falling into a local optimal solution.

[0055] In obtaining the optimal parameter configuration, optionally, the Pareto mechanism can be used to obtain the optimal compromise value of each parameter in the water conservancy professional model, which includes multiple compromise points; each compromise point represents a parameter that can no longer be further optimized among multiple objectives. The optimal compromise value of each parameter is iterated using a multi-objective optimization mechanism (such as NSGA-II, MOEA / D), that is, through continuous iteration, the individuals in the population gradually approach the optimal parameter configuration, so that the model can achieve better performance in multiple objectives, thereby obtaining the optimal parameter configuration. In this way, through continuous iteration, the individuals in the population gradually approach the optimal parameter configuration, so that the water conservancy professional model can achieve better performance in each objective. The objectives here include but are not limited to at least one of the following: minimizing prediction error, maximizing computational efficiency, and minimizing resource consumption.

[0056] Optionally, cross-validation can be used to validate data outside the training set to ensure that the optimal parameters have good generalization capabilities. Cross-validation divides the data into multiple subsets, training and validating the water conservancy model on different training and validation sets. This allows the performance of the water conservancy model to be validated using existing data, reducing the difficulty of verification.

[0057] Optionally, the impact of each parameter on different performances can be scored to improve the efficiency of parameter optimization. Specifically, the key parameters in the water conservancy professional model are determined using a machine learning mechanism; that is, the parameters that play a role in the performance of the water conservancy professional model can be determined based on the optimization goal, which are the key parameters. For example, when the goal is to maximize computational efficiency, the number of layers of the convolutional network may affect computational efficiency. In this case, the number of layers of the convolutional network is the key parameter. Next, based on the key parameters and the historical operating data, a parameter performance evaluation matrix is ​​generated; wherein the parameter performance evaluation matrix is ​​used to characterize the degree of influence of each key parameter on the water conservancy professional model. Finally, the optimized water conservancy professional model is evaluated using the parameter performance evaluation matrix to obtain an evaluation result. In this way, the degree of influence of each parameter on different performances can be obtained more intuitively. According to the matrix, the adjustment direction of each parameter can be obtained, thereby improving the efficiency of parameter optimization and reducing the time and resources used in the parameter optimization process.

[0058] Optionally, the principal component analysis mechanism or the recursive feature elimination mechanism can be used to reduce the dimension of the historical operation data to obtain processed data; through the principal component analysis mechanism, several comprehensive variables (principal components) that play a major role in the data can be found. These principal components can remove redundant information while retaining most of the information of the original data, thereby improving computing efficiency and avoiding problems such as data sparsity and inaccurate calculations caused by excessively high dimensions.

[0059] After obtaining the optimized water conservancy model, an online learning mechanism based on real-time data can be used to adapt to changes in the environment or data distribution. This online learning mechanism ensures the model's adaptability in dynamic environments by updating model parameters in real time. The model can also be run in a simulation environment to obtain application feedback, allowing for further fine-tuning of model parameters and evaluating the effectiveness and robustness of the optimized parameters. The simulation environment constructs a virtual environment similar to real-world scenarios to test the performance of the water conservancy model under various conditions. Actual application feedback includes performance metrics collected from actual operations, including model runtime, resource consumption, prediction accuracy, and exception handling capabilities.

[0060] The above describes a method for determining optimal model operation parameters based on historical operational data. This method can be applied to distributed management systems. Each module implements an automated process for data preprocessing, feature extraction, association modeling, parameter optimization, and result verification, ensuring the independence and scalability of each functional module.

[0061] Optionally, the system has a user interface that allows users to set optimization goals, constraints, and view the optimization progress. The user interface provides a graphical operation interface that supports user-defined optimization tasks and real-time monitoring of optimization progress.

[0062] After obtaining the optimal parameter configuration, visualization tools can be used to visualize the water conservancy model's indicator information; this parameter information includes at least one of the following: a parameter optimization trend chart, a performance difference comparison chart, and an adaptability assessment chart. Optimization results can be dynamically adjusted based on real-time operational data. Through interfaces with business systems, the model's operational status is monitored in real time, and parameter optimization plans are updated based on new data, ensuring the model's continued efficiency and adaptability in actual operation.

[0063] After the optimization is complete, all optimization processes and result data can be stored in a historical database, forming a reusable optimization case library. By reusing cases, the time cost of subsequent similar optimization tasks is reduced, while promoting the accumulation and sharing of knowledge.

[0064] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0065] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0066] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; and internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0067] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0068] In the description of this specification, the terms "one embodiment", "some embodiments", "embodiments", "examples", "specific examples" or "some examples" refer to the specific features, structures, materials or characteristics described in conjunction with the embodiment or example and included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and features of different embodiments or examples, unless they are mutually inconsistent.

[0069] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for determining optimal parameters of model operation based on historical operation data, characterized in that: Applied to the field of water conservancy professional model application, the method includes: Obtain historical operation data generated by water conservancy professional models during operation; Extracting the historical operation data to obtain key features; A mathematical correlation model established based on a machine learning mechanism is used to establish a nonlinear mapping relationship between the parameter configuration of the water conservancy professional model and the model operation results; Determining an optimization mechanism for optimizing the parameter configuration based on the historical operation data, the key characteristics, the water conservancy professional model, and the nonlinear mapping relationship; Using the optimization mechanism to perform a global search on the parameter configuration, iteratively obtaining the optimal parameter configuration of the water conservancy professional model; According to the optimal parameter configuration, the optimized water conservancy professional model is obtained.

2. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: When the optimization mechanism is a particle swarm optimization mechanism, the optimization mechanism is used to perform a global search on the parameter configuration to iteratively obtain the optimal parameter configuration of the water conservancy professional model, including: In any iteration, obtain the current position and update speed of the current parameter configuration in the parameter space; updating the parameter configuration based on the current position and the update speed; The fitness of the updated parameter configuration is obtained by evaluating the objective function; wherein the fitness is used to characterize the quality of the updated parameter configuration; the objective function includes at least one of the following: mean square error, mean absolute error; When the updated parameter configuration meets the preset conditions, the iteration is stopped to obtain the optimal parameter configuration; wherein the preset conditions include at least one of the following: reaching the maximum number of iterations, and the fitness meeting the first preset threshold.

3. The method for determining optimal parameters of model operation based on historical operation data according to claim 2, characterized in that: The method further comprises: In any iteration, the current water conservancy professional model is obtained based on the current parameter configuration; Use the current water conservancy professional model to process the training data and obtain the training results; When the error between the training result and the actual data is greater than a second preset threshold, the search direction of the optimization mechanism is adjusted.

4. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The method further comprises: Using the Pareto mechanism, the optimal compromise value of each parameter in the water conservancy professional model is obtained; The optimal compromise value of each parameter is iterated using a multi-objective optimization mechanism to obtain the optimal parameter configuration.

5. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The method further comprises: Using machine learning mechanisms to determine key parameters in the water conservancy professional model; Based on the key parameters and the historical operation data, a parameter effectiveness evaluation matrix is ​​generated; wherein the parameter effectiveness evaluation matrix is ​​used to characterize the degree of influence of each key parameter on the water conservancy professional model; The parameter performance evaluation matrix is ​​used to evaluate the optimized water conservancy professional model to obtain an evaluation result.

6. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The historical operation data is extracted to obtain key features, including: Using a principal component analysis mechanism or a recursive feature elimination mechanism to reduce the dimension of the historical operation data to obtain processed data; The processed data is extracted to obtain the key features.

7. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The optimization mechanism includes at least one of the following: genetic mechanism, particle swarm optimization mechanism, simulated annealing mechanism; The machine learning mechanism includes at least one of the following: a support vector machine mechanism and a neural network mechanism.

8. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The historical operation data includes at least one of the following: input, output, operation time, resource consumption, and abnormal records.

9. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The method further comprises: Visualization tools are used to output the index information of the water conservancy professional model; wherein the parameter information includes at least one of the following: a parameter optimization trend diagram, a performance difference comparison diagram, and an adaptability evaluation diagram.

10. The method for determining optimal parameters of model operation based on historical operation data according to claim 1, characterized in that: The method is applied to a distributed processing system.

Images