A method, device and equipment for extracting a joint regulation rule of a river basin cascade reservoir group

Abstract

The application discloses a kind of river basin cascade reservoir group joint scheduling rule extraction method, device and equipment, wherein the method comprises: from the preset cascade reservoir scheduling process, identify influencing factor and decision variable, and the preset cascade reservoir scheduling process is at least used to optimize one of flood control, power generation, ecology, navigation, water supply;With influencing factor as input, decision variable as output, support vector machine model is generated by training, and the hyperparameters of support vector machine model are optimized in training process based on improved henry gas solubility optimization algorithm, and the individual updating process is improved based on multiple intelligent optimization algorithms in improved henry gas solubility optimization algorithm;The hyperplane corresponding to support vector machine model is extracted to use hyperplane to represent river basin cascade reservoir group joint scheduling rule.The technical scheme provided by the application improves the accuracy of representing the complex nonlinear relationship between scheduling factor and decision variable.

Description

Technical Field

[0001] This invention relates to the field of reservoir scheduling research, specifically to a method, apparatus, and equipment for extracting joint scheduling rules for a cascade reservoir group in a river basin. Background Technology

[0002] Hydropower is an important clean and renewable energy source, playing a crucial role in addressing global energy security, environmental protection, and climate change. However, in the actual operation and scheduling of cascade reservoir groups, inflow is highly random, and the lead time and accuracy of medium- and long-term runoff forecasts are insufficient to meet the requirements for developing joint scheduling plans for cascade reservoir groups. To improve the joint operation and scheduling capabilities of cascade reservoir groups under conditions of limited lead time and low accuracy in runoff forecasts, this paper explores extracting the hidden nonlinear functional relationships between various scheduling factors of the reservoirs from deterministic optimization scheduling results. This leads to the development of joint scheduling rules for cascade reservoir groups, enabling the rational arrangement of water storage and release in cascade reservoirs. This has become an important research area in reservoir scheduling.

[0003] Currently, commonly used reservoir scheduling rules take two forms: scheduling diagrams and scheduling functions. The former is a relatively conservative scheduling strategy based on historical operating patterns, which has the advantages of simplicity and intuitiveness, but is limited in practical applications because it cannot consider water and rainfall forecast information. The latter uses scheduling factors as inputs and decision variables as outputs, formulating flexible scheduling rules to adapt to actual scheduling needs, and can make scheduling decisions under conditions of uncertainty in inflow.

[0004] There are two main methods for generating scheduling functions: regression analysis and machine learning. Regression analysis yields expressions based on prediction, testing, and correction, a complex process heavily influenced by subjective human factors. Machine learning, on the other hand, uses data mapping to characterize the nonlinear mapping relationship between scheduling influencing factors and decision variables, thus overcoming some of the shortcomings of regression analysis.

[0005] However, the hyperparameter optimization process of existing machine learning methods based on grid search is time-consuming and not suitable for the real-time scheduling needs of cascade reservoirs. Under the conditions of low inflow forecast accuracy and short forecast period, it is still difficult to characterize the complex nonlinear relationship between scheduling factors and decision variables in the joint scheduling of cascade reservoir groups. Summary of the Invention

[0006] In view of this, the present invention provides a method, apparatus and equipment for extracting joint scheduling rules of a cascade reservoir group in a watershed, which improves the accuracy of characterizing the complex nonlinear relationship between scheduling factors and decision variables.

[0007] According to a first aspect, embodiments of the present invention provide a method for extracting joint scheduling rules for a cascade reservoir group in a river basin. The method includes: identifying influencing factors and decision variables from a preset cascade reservoir scheduling process, wherein the preset cascade reservoir scheduling process is used to optimize at least one of flood control, power generation, ecology, navigation, and water supply; training and generating a support vector machine model using the influencing factors as input and the decision variables as output, and optimizing the hyperparameters of the support vector machine model during the training process based on an improved Henry's Law gas solubility optimization algorithm, wherein the improved Henry's Law gas solubility optimization algorithm is an improvement on the individual update process based on multiple intelligent optimization algorithms; and extracting the hyperplane corresponding to the support vector machine model to represent the joint scheduling rules of the cascade reservoir group in the river basin using the hyperplane.

[0008] Optionally, the optimization of the hyperparameters of the support vector machine model during training based on the improved Henry's Law gas solubility optimization algorithm includes: initializing multiple sets of hyperparameters using the Henry's Law gas solubility optimization algorithm, and evaluating and updating each set of initialized hyperparameters; updating each set of initialized hyperparameters using multiple intelligent optimization algorithms; calculating the fitness of each set of updated hyperparameters; for the same initialized hyperparameters, selecting a set of better updated hyperparameters based on the fitness from the hyperparameters updated by the Henry's Law gas solubility optimization algorithm and the hyperparameters updated by multiple intelligent optimization algorithms; continuing to iteratively evaluate and update each set of better updated hyperparameters using the Henry's Law gas solubility optimization algorithm until the iteration termination condition is met, and using the set of hyperparameters with the best evaluation result at termination as the hyperparameters of the support vector machine model.

[0009] Optionally, the intelligent optimization algorithm includes a multiverse optimization algorithm and a quadratic interpolation algorithm.

[0010] Optionally, identifying influencing factors and decision variables from the preset cascade reservoir scheduling process includes: establishing a joint power generation optimization scheduling model for the cascade reservoir group with the objective of maximizing power generation during the scheduling period of the cascade reservoir group, and with water balance constraints and boundary ranges as constraints; solving the joint scheduling model to obtain the preset cascade reservoir scheduling process, and identifying the influencing factors and decision variables from the preset cascade reservoir scheduling process; wherein the preset cascade reservoir scheduling process includes: time-period operating water level process, outflow process, and total plant output process.

[0011] Optionally, the constraints for establishing the joint power generation optimization scheduling model of the cascade reservoir group may also include: water level constraints, flow constraints, flow balance constraints, output constraints, and variable non-negativity constraints.

[0012] Optionally, before training to generate the support vector machine model, the method further includes: performing performance testing on the improved Henry's gas solubility optimization algorithm based on a benchmark function, and adjusting the selection of the intelligent optimization algorithm based on the test results.

[0013] Optionally, the benchmark function includes: a single-peak test function, a multi-peak test function, and a multi-peak test function with a fixed dimension, wherein the single-peak test function is used to test the convergence speed and solution accuracy of the algorithm; the multi-peak test function is used to test the global optimization ability of the algorithm when there are multiple local optima in the function; and the multi-peak test function with a fixed dimension is used to test the algorithm's ability to avoid local optima and its ability to balance global and local optimization.

[0014] According to a second aspect, embodiments of the present invention provide a device for extracting joint scheduling rules for a cascade reservoir group in a river basin. The device includes: an identification module, used to identify influencing factors and decision variables from a preset cascade reservoir scheduling process, wherein the preset cascade reservoir scheduling process is used to optimize at least one of flood control, power generation, ecology, navigation, and water supply; a model training module, used to train and generate a support vector machine model with the influencing factors as input and the decision variables as output, and to optimize the hyperparameters of the support vector machine model during the training process based on an improved Henry's Law gas solubility optimization algorithm, wherein the improved Henry's Law gas solubility optimization algorithm is an improvement based on multiple intelligent optimization algorithms; and a rule extraction module, used to extract the hyperplane corresponding to the support vector machine model, so as to use the hyperplane to characterize the joint scheduling rules of the cascade reservoir group in the river basin.

[0015] According to a third aspect, embodiments of the present invention provide a device for extracting joint scheduling rules for a cascade reservoir group in a river basin, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the method described in the first aspect, or any optional embodiment of the first aspect.

[0016] According to a fourth aspect, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing the computer to perform the method described in the first aspect, or any alternative embodiment of the first aspect.

[0017] The technical solution provided in this application has the following advantages:

[0018] The technical solution provided in this application obtains the influencing factors and decision variables that need to be determined for joint scheduling rules. Then, using the influencing factors as input and the decision variables as output, a support vector machine model is trained. During the training process, the hyperparameters of the support vector machine are optimized in real time using an improved Henry's Law gas solubility optimization algorithm. After training, a more accurate support vector machine model is obtained. Finally, the hyperplane of the support vector machine model is used as the joint scheduling rule of the basin cascade reservoir group to represent the nonlinear mapping relationship between the influencing factors and decision variables. Thus, the accuracy of the joint scheduling rule of the basin cascade reservoir group is improved under the conditions of low inflow forecast accuracy and short forecast period.

[0019] This invention specifically utilizes the Henry's Law gas solubility optimization algorithm to initialize several sets of support vector machine hyperparameters, followed by an evaluation and update process. After the Henry's Law gas solubility optimization algorithm updates each set of hyperparameters, two intelligent optimization algorithms—the multiverse optimization algorithm and the quadratic interpolation algorithm—are used to update the hyperparameters. This results in three different update results for each set of hyperparameters. Then, based on fitness, the optimal updated hyperparameters are selected from the three updated sets for each set of hyperparameters. These optimal updated hyperparameters are then used to return to the evaluation step of the Henry's Law gas solubility optimization algorithm. This process is iterated multiple times until the termination condition is met, thereby significantly improving the accuracy of the support vector machine model hyperparameters. This invention employs the multiverse optimization operator and the quadratic interpolation method, improving the solution accuracy, convergence speed, and ability to avoid getting trapped in local optima of the existing Henry's Law gas solubility algorithm, thus enhancing its ability to optimize support vector machine hyperparameters. This invention integrates a support vector machine with an improved Henry's Law gas solubility algorithm to establish a joint scheduling rule extraction model for cascade reservoir groups. This solves the problem in existing technologies of accurately characterizing the complex nonlinear relationship between scheduling influencing factors and decision variables in the joint scheduling of cascade reservoir groups when the inflow forecast accuracy is low and the forecast period is short. Attached Figure Description

[0020] The features and advantages of the invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the drawings:

[0021] Figure 1 This diagram illustrates the steps of a method for extracting joint scheduling rules for a cascade reservoir group in a watershed according to one embodiment of the present invention.

[0022] Figure 2 A flowchart illustrating a method for extracting joint scheduling rules for a cascade reservoir group in a watershed according to one embodiment of the present invention is shown.

[0023] Figure 3This diagram illustrates the structure of a joint scheduling rule extraction device for a cascade reservoir group in a watershed, according to one embodiment of the present invention.

[0024] Figure 4 The diagram shows a structural schematic of a watershed cascade reservoir group joint scheduling rule extraction device according to one embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 In one embodiment, a method for extracting joint scheduling rules for a cascade reservoir group in a river basin specifically includes the following steps:

[0027] Step S101: Identify influencing factors and decision variables from the preset cascade reservoir scheduling process. The preset cascade reservoir scheduling process is used to optimize at least one of flood control, power generation, ecology, navigation, and water supply.

[0028] Step S102: Using the influencing factor as input and the decision variable as output, train and generate a support vector machine model, and optimize the hyperparameters of the support vector machine model during the training process based on the improved Henry gas solubility optimization algorithm. The improved Henry gas solubility optimization algorithm is an improvement on the individual update process based on a variety of intelligent optimization algorithms.

[0029] Step S103: Extract the hyperplane corresponding to the support vector machine model, so as to use the hyperplane to represent the joint scheduling rules of the cascade reservoir group in the basin.

[0030] Specifically, the joint operation of cascade reservoirs refers to the work of determining the relationship between decision variables and influencing factors in the operation of reservoirs of various hydropower stations on the same river. Due to the uncertainty of future inflow into the reservoirs of hydropower stations, the basic requirement of reservoir operation is to coordinate and properly handle the contradiction between the reliability and economy of hydropower station operation. Different joint operation rules apply to different operation objectives. Since the hydropower stations in a cascade are hydraulically connected, reservoir operation is more complex, requiring consideration of the mutual influence of head and discharge between upstream and downstream hydropower stations. To fully utilize head and reduce water wastage, the order of reservoir storage and discharge should be rationally determined, optimizing at least one of the processes: flood control, power generation, ecology, navigation, and water supply.

[0031] Therefore, to achieve the aforementioned joint scheduling, it is first necessary to determine which parameter changes can maximally affect the processes of flood control, power generation, ecology, navigation, and water supply, and which decision variables can directly achieve the effects of flood control, power generation, ecology, navigation, and water supply. This requires identifying influencing factors and decision variables from a pre-set cascade reservoir scheduling process, which is at least used to optimize one of the four processes: flood control, power generation, ecology, navigation, and water supply. In this embodiment of the invention, the cascade reservoir scheduling process is obtained with the goal of achieving the optimal economic benefits of joint reservoir power generation.

[0032] Specifically, the optimization focuses on the power generation process. With the objective of maximizing power generation during the scheduling period of the cascade reservoir group in the basin, and using water balance constraints and boundary ranges as constraints, a joint power generation optimization scheduling model for the cascade reservoir group is established. Then, the joint scheduling model is solved to obtain the preset cascade reservoir scheduling process used in this embodiment of the invention, and influencing factors and decision variables are identified from the preset cascade reservoir scheduling process. The solved preset cascade reservoir scheduling process includes: time-period operating water level process, outflow process, and total plant output process.

[0033] The objective function for maximizing power generation in this embodiment of the invention is:

[0034]

[0035] In the formula, E represents the total power generation of the reservoir group during the scheduling period; P i,t k i,t H i,t and Q i,t Let N be the generator unit power, output coefficient, head (the upstream water level of the hydropower station is the reservoir water level, the downstream water level is the turbine tailwater level, and the head is the difference between the reservoir water level and the turbine tailwater level) and power generation flow of reservoir i during time period t; N is the number of cascade reservoirs; T is the length of the scheduling period; and Δt is the time period step.

[0036] Through solving and identifying the above model, the influencing factors are determined to be the initial water level of the reservoir during the specified time period, the inflow rate, and the initial water level of adjacent cascade reservoirs during the same time period; the decision variable is the outflow rate of the reservoir during the specified time period. Therefore, a support vector machine model is accurately trained based on the identified influencing factors and decision variables.

[0037] Subsequently, using influencing factors as input samples and decision variables as output labels, a support vector machine (SVM) model is trained to represent the nonlinear relationship between influencing factors and decision variables using the nonlinear hyperplane of the SVM. Simultaneously, during training, the hyperparameters of the SVM model are optimized based on an improved Henry's Law gas solubility optimization algorithm, enhancing the SVM's ability to characterize the nonlinear relationship between influencing factors and decision variables. In this embodiment, the improved Henry's Law gas solubility optimization algorithm is an improvement based on multiple intelligent optimization algorithms, including but not limited to particle swarm optimization, genetic algorithms, simulated annealing, tabu search, and ant colony optimization. Multiple intelligent optimization algorithms are used to improve the update process of individuals in the Henry's Law gas solubility optimization algorithm. Specifically, multiple intelligent optimization algorithms are used simultaneously to update individual hyperparameters, and then the optimal hyperparameters are selected from the hyperparameters updated by different algorithms. This improves the solution accuracy, convergence speed, and ability to avoid getting trapped in local optima of the existing Henry's Law gas solubility algorithm, enhancing its ability to optimize SVM hyperparameters. This makes the training of the SVM model for influencing factors and decision variables more accurate, improving the accuracy of the extracted joint scheduling rules for the cascade reservoir group in the watershed.

[0038] Specifically, such as Figure 2 As shown, in one embodiment, step S102 specifically includes the following steps:

[0039] Step 1: Initialize multiple sets of hyperparameters using the Henry's gas solubility optimization algorithm, and evaluate and update the initial hyperparameters for each set.

[0040] Step 2: Update the initial hyperparameters of each group using a variety of intelligent optimization algorithms.

[0041] Step 3: Calculate the fitness of each set of updated hyperparameters.

[0042] Step 4: For the same initialized hyperparameters, select a set of better updated hyperparameters based on fitness from the hyperparameters updated by Henry's gas solubility optimization algorithm and the hyperparameters updated by various intelligent optimization algorithms;

[0043] Step 5: Use the Henry's gas solubility optimization algorithm to iteratively evaluate and update each set of updated hyperparameters until the iteration termination condition is met. Use the set of hyperparameters with the best evaluation results at the time of termination as the hyperparameters of the support vector machine model.

[0044] Specifically, Henry's Law is a famous gas law in the field of physical chemistry, formulated by the renowned 19th-century chemist Henry the Great. In short, this law states that at a given temperature, when a gas and liquid reach equilibrium, the amount of gas dissolved in the liquid is directly proportional to its partial pressure above the liquid. A key concept in understanding Henry's Law is solubility, which changes with temperature and pressure. Specifically, increasing temperature leads to a decrease in gas solubility, while increasing pressure leads to an increase in gas solubility. By controlling both temperature and pressure, the optimal equilibrium state of the gas can be achieved according to Henry's Law. Inspired by this phenomenon, the Henry gas solubility optimization algorithm (HGSO) is a novel, population-based, physics-inspired metaheuristic optimization algorithm developed in recent years. Empirical studies show that HGSO has advantages in equilibrium exploration and development.

[0045] The optimization process of HGSO is divided into four stages: initialization, evaluation, update, and termination. This embodiment of the invention optimizes the hyperparameters of a support vector machine based on HGSO. The hyperparameters optimized in this embodiment include the penalty factor C* used by the support vector machine to adjust the slack variables, and the control parameter gamma of the radial basis kernel function. A set of hyperparameters (C*, gamma) represents a gas individual in the HGSO algorithm. The selection of hyperparameters varies depending on the choice of kernel function; this embodiment is only an example and not a limitation. In the initialization stage, HGSO uses the number of sets N, the upper boundary vector lb, and the lower boundary vector ub as inputs to randomly generate multiple sets of hyperparameters, which can be expressed as:

[0046] g i =lb+r l,i (ub-lb)i=1,2,...,N

[0047] Where: g i Represents the i-th group of hyperparameters; r l,i Let represent a random number uniformly distributed in the interval [0,1] corresponding to hyperparameter i. Multiple sets of hyperparameters are further divided into several sets, the number of which is denoted by l. Each set of hyperparameters in each set has the same Henry's constant.

[0048] During the evaluation phase, each set is evaluated based on the fitness function (which is a function obtained by transforming the objective function of the support vector machine, such as transforming the objective function according to the actual situation, such as requiring the maximum power or the highest water level) to determine the best hyperparameters that obtain the highest equilibrium state from other hyperparameters of the same type. Then, the hyperparameters are sorted to obtain the global optimal hyperparameters.

[0049] The Henry's constant and solubility are updated according to the following formula, and the initial Henry's constant H, partial pressure P, and the ratio of enthalpy of solution to gas constant C are randomly generated.

[0050]

[0051] S i,l (it)=KH l (it+1)P i,l (it)

[0052] In the formula: it represents the current iteration number; H l (it) = r2c1 is the Henry's constant of the l-th group of hyperparameters; P i,l (it) = r3c2 is the partial voltage of hyperparameter i in the l-th group; C l =r4c3 is the ratio of the enthalpy of dissolution to the gas constant of the l-th hyperparameter group; r2, r3, and r4 are random numbers uniformly distributed in the interval [0,1], and c1, c2, and c3 are constants, set to 0.05, 100, and 0.01 respectively; S represents the temperature of the current iteration, where MaxIt represents the maximum number of iterations; i,l (it) represents the solubility of hyperparameter i in the l-th set during the it-th iteration; K is a constant with a value of 1. Based on these parameters in the current iteration, each set of hyperparameters is updated during the update phase according to the following formula:

[0053] g i,l (it+1)=g i,l (it)+r5γF(G l (it)-g i,l (it))+r6αF(S i,l (it)G l (it)-g i,l (it))

[0054]

[0055] Where: g i,l Let be the hyperparameter i in the l-th set; r5 and r6 are random numbers uniformly distributed in the interval [0,1]; γ is the interaction ability of gases in the same group; α, β, and ε are optimization parameters, respectively 1, 1, and 0.5; F is a flag parameter controlling the hyperparameter search direction, which is a random number of 1 or -1; G l (it), F l (it) represents the optimal hyperparameter in the l-th set and the fitness value of the i-th hyperparameter therein; F(it) represents the fitness value of the globally optimal hyperparameter obtained so far.

[0056] To enhance HGSO's global optimization capability, the method for updating poorly performing hyperparameters is the same as the method for initializing the population, as follows:

[0057] N′=N(c4+r7(c5-c4))

[0058] g′ i,l =lb + r8(ub - lb)

[0059] In the formula: N' represents the number of inferior hyperparameters; r7 and r8 are random numbers uniformly distributed in the interval [0,1]; c4 and c5 are constants, set to 0.1 and 0.2 respectively; g' i,l The update hyperparameters are for the inferior hyperparameter i in the l-th set.

[0060] For each set of hyperparameters, after updating using HGSO, various intelligent optimization algorithms are then used to update the hyperparameters. In this embodiment of the invention, the multiverse optimization algorithm and the quadratic interpolation algorithm are used.

[0061] The multiverse theory in physics inspires the multiverse algorithm, explaining the existence of universes other than our own. Objects can be transferred and exchanged between multiple universes, and the individual update mechanism in the multiverse algorithm simulates this exchange. Each universe consists of three basic elements: white holes, black holes, and wormholes. Therefore, there are two paths for object transfer between different universes: white hole / black hole tunnels and wormhole mechanisms. In this embodiment, a "universe" refers to an individual in the population, a universe represents a set of hyperparameters, an "object" can be understood as a specific parameter within that set of hyperparameters, and the "optimal universe" is understood as the set of hyperparameters with the best fitness value among all hyperparameters.

[0062] The material transfer mechanism in white hole / black hole tunnels is as follows:

[0063]

[0064] In the formula, u i Representing the universe i, u i,j u k,j Let represent objects j in universe i and objects j in universe k, respectively; normr(-) represents normalizing the expansion rate of the universe; r1 represents a random number uniformly distributed on [0,1].

[0065] The material transfer mechanism of wormholes is as follows:

[0066]

[0067] In the formula, U jLet represent object j in the optimal universe; r2, r3, and r4 all represent random numbers uniformly distributed on [0,1]; TDR and WEP represent the travel distance rate and the probability of wormhole existence, respectively; ub j lb j Let $\mathbf$ and $\mathbf$ represent the upper and lower limits of object $j$, respectively.

[0068] Then, a quadratic interpolation algorithm is used to update each set of hyperparameters, as follows:

[0069] For a set of hyperparameters S a During the update, two sets of hyperparameters S are randomly selected from all hyperparameters. b S c The function exhibits characteristics of "low in the middle and high at both ends," and then a quadratic polynomial of the interpolation function is fitted. Based on the relationship between the interpolation function and the original function in S... a S b S c Based on the property of equal function values, we solve the interpolation function, calculate its extrema, and obtain the updated hyperparameters. The specific formula is as follows:

[0070]

[0071] In the formula, s a =(s a,1 ,s a,2 ,...,s a,j ,...,s a,D ), s b =(s b,1 ,s b,2 ,...,s b,j ,...,s b,D ), s c =(s c,1 ,s c,2 ,...,s c,j ,...,s c,D ) is the candidate individual selected to form a quadratic curve, s p =(s p,1 ,s p,2 ,...s p,j ,...,s p,D ) is the extreme point of the quadratic curve, and also a promising possible solution based on 3 random individuals in the population. f(·) represents the fitness value of the individual.

[0072] Subsequently, for each set of hyperparameters, a set of hyperparameters updated by HGSO, a set of hyperparameters updated by the multiverse optimization algorithm, and a set of hyperparameters updated by the quadratic interpolation algorithm are obtained. The three sets of hyperparameters are substituted into the preset fitness function, and the best-updated hyperparameters are selected from the three sets of hyperparameters based on the result of the fitness function. This is a set of superior hyperparameters.

[0073] Then, using the updated hyperparameters from each set, the evaluation process of HGSO is repeated. This process continues iterating multiple times until the iteration termination conditions are met (including but not limited to reaching a preset number of iterations, or the hyperparameter results stabilizing). The iteration stops, yielding a set of optimal hyperparameters for the support vector machine model. By employing a multiverse optimization algorithm, the global exploration capability of the existing HGSO is enhanced, preventing inferior individuals from getting trapped in local optima. The quadratic interpolation method improves the local exploitation capability of HGSO, thereby increasing the solution accuracy and convergence speed of the existing HGSO. This comprehensively improves the solution accuracy, convergence speed, and ability to avoid local optima of the existing Henry's Law gas solubility algorithm, enhancing its ability to optimize support vector machine hyperparameters.

[0074] Specifically, in one embodiment, the constraints for establishing the optimized scheduling model for joint power generation of a cascade reservoir group according to the present invention include:

[0075] Water level constraints:

[0076] Z min (j,t)≤Z(j,t)≤Z max (j,t)

[0077] Flow constraints:

[0078]

[0079] Flow balancing constraints:

[0080] Q i (j,t)=Q o (j-1,t)+q(j,t)

[0081] Water balance constraints:

[0082] V(j,t+1)=V(j,t)+(Q i (j,t)-Q o (j,t))ΔT

[0083] Output constraints:

[0084] N min (j,t)≤N(j,t)≤N max (j,t)

[0085] Non-negativity constraint for variables.

[0086] In the formula, Z(j,t) represents the water level of the reservoir corresponding to the j-th hydropower station during time period t; Z min (j,t), Z max (j,t) represent the minimum and maximum water levels of the reservoir corresponding to the j-th hydropower station during time period t, respectively; Qo (j,t) represents the discharge flow of the j-th hydropower station during time period t; Q o min (j,t), Q o max (j,t) represent the minimum and maximum discharge flows of the j-th hydropower station during time period t, respectively; Q i (j,t) represents the inflow rate into the reservoir corresponding to the j-th hydropower station during time period t; Q o (j-1,t) represents the discharge flow of the (j-1)th hydropower station during time period t; q(j,t) represents the inflow of the downstream section of the river during time period t of the jth hydropower station; V(j,t) represents the water storage of the reservoir corresponding to the jth hydropower station during time period t; N(j,t) represents the power output of the jth hydropower station during time period t (the active power of the entire hydropower station); N min (j,t), N max (j,t) represent the minimum and maximum output of the j-th hydropower station during time period t, respectively.

[0087] The above constraints further improve the accuracy of the joint power generation optimization scheduling model of cascade reservoir groups, thereby improving the accuracy of identifying influencing factors and decision variables.

[0088] Specifically, in one embodiment, before step S102 described above, the following step is also included:

[0089] Step 6: Perform performance testing on the improved Henry's gas solubility optimization algorithm based on the benchmark function, and adjust the selection of the intelligent optimization algorithm based on the test results.

[0090] Specifically, this embodiment of the invention also uses a classic benchmark function testing platform, including 23 test functions widely used in the field of intelligent algorithms, to perform performance testing on the improved Henry's Law gas solubility optimization algorithm. The benchmark functions selected in this embodiment include: single-peak test functions, multi-peak test functions, and fixed-dimensional multi-peak test functions. The single-peak test functions are used to test the algorithm's convergence speed and development accuracy; the multi-peak test functions are used to test the algorithm's ability to explore when multiple local optima exist; and the fixed-dimensional multi-peak test functions are used to test the algorithm's ability to avoid local optima and its ability to balance exploration and development.

[0091] By testing the benchmark function, the effectiveness of the improved Henry's Gas Solubility Optimization Algorithm (IHGSO) is ensured before training. If the test results are unsatisfactory, the selection of the intelligent optimization algorithm is adjusted in a timely manner to ensure the reliability of IHGSO and to ensure the accuracy of the support vector machine in characterizing the nonlinear mapping relationship between influencing factors and decision variables.

[0092] In a specific scenario embodiment, a cascade reservoir group in the upper reaches of the Yangtze River is taken as the research object. With the goal of maximizing power generation, a joint power generation optimization scheduling model for the cascade reservoir group is established. Historical runoff data from 2014 to 2020 is used as the input to the scheduling model, and a hybrid algorithm combining dynamic programming, stepwise approximation, and stepwise optimization is employed to solve the model. The scheduling results from 2014 to 2018 serve as the training set, while the results from other years serve as the validation set. Influencing factors identified during the scheduling process include the initial water level of the time period, the inflow during the time period, and the initial water level of adjacent cascade reservoirs at the same time. These are used as the input to the scheduling rule extraction model, while the outflow during the time period is the decision variable and is used as the output of the scheduling rule extraction model. The practical application effect of the joint scheduling rule extraction method for a cascade reservoir group provided by this invention is then verified.

[0093] To analyze the performance of the cascade reservoir group joint scheduling rule extraction model (Improved Henry's Law Gas Solubility Optimization Algorithm-Support Vector Machine, IHGSO-SVM) proposed in this embodiment of the invention, this study also compared other hybrid scheduling rule extraction models, including Henry's Law Gas Solubility Optimization Algorithm-Support Vector Machine (HGSO-SVM), Particle Swarm Optimization Algorithm-Support Vector Machine (PSO-SVM), Sine and Cosine Algorithm-Support Vector Machine (SCA-SVM), and Grid Algorithm-Support Vector Machine (Grid-SVM). The statistical indicators compared included the correlation coefficient R between measured and predicted water levels. 2 The root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) were calculated, and the results are shown in Table 1. As can be seen from Table 1, for the same input data, the R-squared value of the IHGSO-SVM model provided by this invention is significantly higher than that of the standard model. 2 The highest values ​​for RMSE, MAE, and MAPE, and the lowest values ​​for MAPE, indicate that the proposed IHGSO algorithm can find a better combination of hyperparameters to improve the ability of SVM in extracting joint scheduling rules for cascade reservoir groups. The scheduling rules for cascade reservoir groups extracted by the IHGSO-SVM model are closer to the optimized scheduling process.

[0094] Table 2 presents a comparison between the measured cascade power generation and the predicted cascade power generation based on scheduling rules on the validation set. It can be seen that under the joint scheduling rules of the cascade reservoir group extracted by IHGSO-SVM, the power generation of Reservoir 1, Reservoir 2, and the cascade itself are 131.301 billion kWh, 64.778 billion kWh, and 196.079 billion kWh, respectively. These figures are all higher than the measured 129.823 billion kWh, 64.031 billion kWh, and 193.854 billion kWh, and are also closer to the results of other scheduling rule extraction hybrid models, which are closer to the optimized scheduling model's results of 131.41 billion kWh, 65 billion kWh, and 196.41 billion kWh. This also indicates that the cascade reservoir group scheduling operation rules derived by the IHGSO-SVM rule extraction model provided in this invention are superior to existing scheduling rule extraction models, and can obtain cascade power generation closer to the optimal operation and scheduling mode.

[0095] Table 1. Validation index values ​​of different scheduling rule extraction models on the validation set.

[0096]

[0097]

[0098] Table 2 Comparison of measured and predicted cascade power generation based on scheduling rules on the validation set (100 million kWh)

[0099]

[0100] Through the above steps, the technical solution provided in this application obtains the influencing factors and decision variables that need to be determined for the joint scheduling rules. Then, using the influencing factors as input and the decision variables as output, a support vector machine model is trained. During the training process, the hyperparameters of the support vector machine are optimized in real time using an improved Henry's Law gas solubility optimization algorithm. After training, a more accurate support vector machine model is obtained. Finally, the hyperplane of the support vector machine model is used as the joint scheduling rule of the basin cascade reservoir group to represent the nonlinear mapping relationship between the influencing factors and decision variables. Thus, under the conditions of low inflow forecast accuracy and short forecast period, the accuracy of the joint scheduling rule of the basin cascade reservoir group is improved.

[0101] This invention specifically utilizes the Henry's Law gas solubility optimization algorithm to initialize several sets of support vector machine hyperparameters, followed by an evaluation and update process. After the Henry's Law gas solubility optimization algorithm updates each set of hyperparameters, two intelligent optimization algorithms—the multiverse optimization algorithm and the quadratic interpolation algorithm—are used to update the hyperparameters. This results in three different update results for each set of hyperparameters. Then, based on fitness, the optimal updated hyperparameters are selected from the three updated sets for each set of hyperparameters. These optimal updated hyperparameters are then used to return to the evaluation step of the Henry's Law gas solubility optimization algorithm. This process is iterated multiple times until the termination condition is met, thereby significantly improving the accuracy of the support vector machine model hyperparameters. This invention employs the multiverse optimization operator and the quadratic interpolation method, improving the solution accuracy, convergence speed, and ability to avoid getting trapped in local optima of the existing Henry's Law gas solubility algorithm, thus enhancing its ability to optimize support vector machine hyperparameters. This invention integrates a support vector machine with an improved Henry's Law gas solubility algorithm to establish a joint scheduling rule extraction model for cascade reservoir groups. This solves the problem in existing technologies of accurately characterizing the complex nonlinear relationship between scheduling influencing factors and decision variables in the joint scheduling of cascade reservoir groups when the inflow forecast accuracy is low and the forecast period is short.

[0102] like Figure 3 As shown in the figure, this embodiment also provides a device for extracting joint scheduling rules for a cascade reservoir group in a river basin. The device includes:

[0103] The identification module 101 is used to identify influencing factors and decision variables from a preset cascade reservoir scheduling process. The preset cascade reservoir scheduling process is used to optimize at least one of flood control, power generation, ecology, navigation, and water supply. For details, please refer to the relevant description of step S101 in the above method embodiment, which will not be repeated here.

[0104] The model training module 102 is used to train and generate a support vector machine model with influencing factors as input and decision variables as output. During training, it optimizes the hyperparameters of the support vector machine model based on an improved Henry's Law gas solubility optimization algorithm, which is an improvement upon various intelligent optimization algorithms. For details, please refer to the relevant description of step S102 in the above method embodiments, which will not be repeated here.

[0105] The rule extraction module 103 is used to extract the hyperplane corresponding to the support vector machine model, so as to use the hyperplane to represent the joint scheduling rules of the cascade reservoir group in the basin. For details, please refer to the relevant description of step S103 in the above method embodiment, which will not be repeated here.

[0106] The basin-wide cascade reservoir group joint scheduling rule extraction device provided in this embodiment of the invention is used to execute the basin-wide cascade reservoir group joint scheduling rule extraction method provided in the above embodiment. Its implementation method and principle are the same. For details, please refer to the relevant description of the above method embodiment, which will not be repeated here.

[0107] Through the collaborative efforts of the aforementioned components, the technical solution provided in this application obtains the influencing factors and decision variables that need to be determined for joint scheduling rules. Then, using the influencing factors as input and the decision variables as output, a support vector machine model is trained. During the training process, the hyperparameters of the support vector machine are optimized in real time using an improved Henry's Law gas solubility optimization algorithm. After training, a more accurate support vector machine model is obtained. Finally, the hyperplane of the support vector machine model is used as the joint scheduling rule of the basin-wide cascade reservoir group, representing the nonlinear mapping relationship between the influencing factors and decision variables. Thus, under the conditions of low inflow forecast accuracy and short forecast period, the accuracy of the joint scheduling rule of the basin-wide cascade reservoir group is improved.

[0108] This invention specifically utilizes the Henry's Law gas solubility optimization algorithm to initialize several sets of support vector machine hyperparameters, followed by an evaluation and update process. After the Henry's Law gas solubility optimization algorithm updates each set of hyperparameters, two intelligent optimization algorithms—the multiverse optimization algorithm and the quadratic interpolation algorithm—are used to update the hyperparameters. This results in three different update results for each set of hyperparameters. Then, based on fitness, the optimal updated hyperparameters are selected from the three updated sets for each set of hyperparameters. These optimal updated hyperparameters are then used to return to the evaluation step of the Henry's Law gas solubility optimization algorithm. This process is iterated multiple times until the termination condition is met, thereby significantly improving the accuracy of the support vector machine model hyperparameters. This invention employs the multiverse optimization operator and the quadratic interpolation method, improving the solution accuracy, convergence speed, and ability to avoid getting trapped in local optima of the existing Henry's Law gas solubility algorithm, thus enhancing its ability to optimize support vector machine hyperparameters. This invention integrates a support vector machine with an improved Henry's Law gas solubility algorithm to establish a joint scheduling rule extraction model for cascade reservoir groups. This solves the problem in existing technologies of accurately characterizing the complex nonlinear relationship between scheduling influencing factors and decision variables in the joint scheduling of cascade reservoir groups when the inflow forecast accuracy is low and the forecast period is short.

[0109] Figure 4 This invention illustrates a device for extracting joint scheduling rules for a cascade reservoir group in a river basin, comprising a processor 901 and a memory 902, which can be connected via a bus or other means. Figure 4 Taking the example of a connection between China and Israel via a bus.

[0110] Processor 901 can be a Central Processing Unit (CPU). Processor 901 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0111] The memory 902, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above method embodiments. The processor 901 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory 902, thereby implementing the methods in the above method embodiments.

[0112] The memory 902 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor 901, etc. Furthermore, the memory 902 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 902 may optionally include memory remotely located relative to the processor 901, and these remote memories may be connected to the processor 901 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0113] One or more modules are stored in memory 902, and when executed by processor 901, they perform the methods described in the above method embodiments.

[0114] The specific details of the equipment for extracting the joint scheduling rules of the above-mentioned cascade reservoir group can be understood by referring to the relevant descriptions and effects in the above method embodiments, and will not be repeated here.

[0115] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The implemented program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0116] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for extracting joint regulation rules of a cascade reservoir group in a river basin, characterized in that, The method comprises: from a preset cascade reservoir scheduling process, identifying an influencing factor and a decision variable, the preset cascade reservoir scheduling process being used for at least optimizing one of flood control, power generation, ecology, navigation and water supply; training a support vector machine model with the influencing factor as input and the decision variable as output, and optimizing hyperparameters of the support vector machine model in the training process based on an improved Henry gas solubility optimization algorithm, the improved Henry gas solubility optimization algorithm being improved based on individual updating processes of multiple intelligent optimization algorithms; extracting a hyperplane corresponding to the support vector machine model to represent the joint scheduling rule of the cascade reservoir group in the basin by using the hyperplane; the step of optimizing the hyperparameters of the support vector machine model in the training process based on the improved Henry gas solubility optimization algorithm comprises: initializing multiple groups of hyperparameters by using the Henry gas solubility optimization algorithm, and evaluating and updating each group of initialized hyperparameters; updating each group of initialized hyperparameters by using multiple intelligent optimization algorithms; calculating the fitness of each group of updated hyperparameters; for the same initialized hyperparameters, selecting an updated group of optimal hyperparameters from the hyperparameters updated by the Henry gas solubility optimization algorithm and the hyperparameters updated by the multiple intelligent optimization algorithms based on the fitness; continuing to iteratively evaluate and update each group of updated optimal hyperparameters by using the Henry gas solubility optimization algorithm until an iteration termination condition is reached, and taking a group of hyperparameters with the optimal evaluation result at the termination as the hyperparameters of the support vector machine model; the intelligent optimization algorithm comprises a multiverse optimization algorithm and a quadratic interpolation algorithm.

2. The method of claim 1, wherein, The step of identifying the influencing factor and the decision variable from the preset cascade reservoir scheduling process comprises: establishing a joint power generation optimal scheduling model of the cascade reservoir group with the maximum power generation in the scheduling period of the cascade reservoir group in the basin as the target and water balance constraints and boundary ranges as constraint conditions; solving the joint scheduling model to obtain the preset cascade reservoir scheduling process, and identifying the influencing factor and the decision variable from the preset cascade reservoir scheduling process; wherein the preset cascade reservoir scheduling process comprises a per-period operation water level process, a reservoir outflow process and a total plant output process.

3. The method of claim 2, wherein, The constraint conditions for establishing the joint power generation optimal scheduling model of the cascade reservoir group further comprise water level constraints, flow constraints, flow balance constraints, output constraints and variable non-negative constraints.

4. The method of claim 1, wherein, Before the support vector machine model is trained, the method further comprises: testing the performance of the improved Henry gas solubility optimization algorithm based on a benchmark function, and adjusting the selection of the intelligent optimization algorithm based on the test result.

5. The method of claim 4, wherein, The benchmark function comprises unimodal test functions, multimodal test functions and multimodal test functions with fixed dimensions, wherein the unimodal test functions are used to test the convergence speed and solution accuracy of the algorithm; the multimodal test functions are used to test the global optimization ability of the algorithm when there are multiple local optimal solutions in the function; and the multimodal test functions with fixed dimensions are used to test the ability of the algorithm to avoid local optimization and the balance ability between global and local optimization of the algorithm.

6. A device for extracting joint regulation rules of a cascade reservoir group of a river basin, characterized in that, The device comprises: An identification module is configured to identify an influence factor and a decision variable from a preset cascade reservoir scheduling process, the preset cascade reservoir scheduling process being used for optimizing at least one of flood control, power generation, ecology, navigation, and water supply; A model training module is configured to train a support vector machine model with the influence factor as input and the decision variable as output, and to optimize hyperparameters of the support vector machine model in a training process based on an improved Henry gas solubility optimization algorithm, the improved Henry gas solubility optimization algorithm being improved based on multiple intelligent optimization algorithms; the optimization of the hyperparameters of the support vector machine model in the training process based on the improved Henry gas solubility optimization algorithm includes: initializing multiple groups of hyperparameters by using the Henry gas solubility optimization algorithm, and evaluating and updating each group of initialized hyperparameters; updating each group of initialized hyperparameters by using multiple intelligent optimization algorithms; calculating the fitness of each group of updated hyperparameters; for the same initialized hyperparameters, selecting an updated group of superior hyperparameters from the hyperparameters updated by the Henry gas solubility optimization algorithm and the hyperparameters updated by the multiple intelligent optimization algorithms based on the fitness; continuing to iteratively evaluate and update each group of updated superior hyperparameters by using the Henry gas solubility optimization algorithm until an iteration termination condition is reached, and taking a group of hyperparameters with the optimal evaluation result at the termination as the hyperparameters of the support vector machine model; the intelligent optimization algorithms include a multiverse optimization algorithm and a quadratic interpolation algorithm; A rule extraction module is configured to extract a hyperplane corresponding to the support vector machine model, so as to represent a joint scheduling rule of the cascade reservoirs in the river basin by using the hyperplane.

7. A device for extracting a joint dispatching rule of a cascade reservoir group of a river basin, characterized by, The method comprises the following steps: The computer readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the method according to any one of claims 1-5.

Images