Method and device for calculating hydraulic energy parameters of pumped storage power station based on machine learning

Abstract

The invention relates to the technical field of pumped storage, in particular to a method and device for calculating hydraulic energy parameters of a pumped storage power station based on machine learning. According to the method, the preset machine learning model is trained by determining the reservoir capacity curve coefficient and the corresponding water level data as the input and output samples, so that the water level calculation model obtained through training can quickly perform water energy parameter calculation on the pumped storage power station, and the condition that various optimization models are adopted to perform iterative calculation one by one in related technologies is avoided; therefore, the problems of low calculation efficiency and long consumed time are solved. Meanwhile, when the storage capacity curve coefficient is determined, the storage capacity curve constraint is considered, the calculation amount can be reduced, and the problems of inaccuracy and the like possibly caused by random determination of the coefficient are avoided; and penalty constraints determined by the water level data are added in a loss function adopted during model training, so that the model can quickly correct the prediction direction in the training process, and finally water level parameters meeting actual engineering requirements are output.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pumped storage power station, and particularly relates to a pumped storage power station water energy parameter calculation method and device based on machine learning. BACKGROUND

[0002] Pumped storage power station water energy parameter calculation is a relatively complex technical work, which needs to comprehensively consider the reservoir capacity conditions of the upper and lower reservoirs of the power station, the parameter requirements for stable operation of the unit, etc., to find a set of comprehensive optimal parameters as the basis for the planning and design of the power station.

[0003] The existing pumped storage water energy parameter optimization calculation method seeks the optimal solution through repeated iteration and trial calculation, which has defects: in the early planning and site selection stage, a large number of pumped storage power stations need to calculate the water energy parameters, and the traditional method can only calculate iteratively one by one, the overall calculation efficiency is low, and a long time is needed, which often affects the efficiency of the planning and site selection work. SUMMARY

[0004] Therefore, the present application provides a pumped storage power station water energy parameter calculation method and device based on machine learning to solve the problem of long time consumption in the prior art pumped storage power station water energy parameter calculation.

[0005] In a first aspect, the present application provides a pumped storage power station water energy parameter calculation method based on machine learning, which comprises: determining the coefficients of the reservoir capacity curve simulation model based on the reservoir capacity curve constraint and the reservoir bottom elevation distribution interval by using the Monte Carlo method; inputting a plurality of reservoir capacity curves determined by the reservoir capacity curve simulation model into a preset optimization model to obtain the water energy parameters corresponding to each reservoir capacity curve; adding a penalty constraint to the water level data in the water energy parameters in a preset loss function to obtain the loss function of the preset machine learning model; training the preset machine learning model with the coefficients and the reservoir bottom elevation as inputs and the water level data in the water energy parameters as outputs to obtain a water level calculation model; determining the water level data of the to-be-established pumped storage power station according to the reservoir capacity curve coefficients of the to-be-established pumped storage power station and the water level calculation model, and calculating the water energy parameters.

[0006] The method for calculating water energy parameters of a pumped storage power station based on machine learning provided in the embodiments of the present application trains a preset machine learning model by taking the reservoir storage curve coefficient and corresponding water level data as input and output samples, so that the water level calculation model trained in this way can quickly calculate the water energy parameters of the pumped storage power station, thereby avoiding the problem of low calculation efficiency and long time consumption caused by iterative calculation of various optimization models one by one in the related art. Meanwhile, when determining the reservoir storage curve coefficient, the reservoir storage curve constraint is considered, which can reduce the calculation amount and avoid the problem of inaccuracy possibly caused by random determination of the coefficient; and the penalty constraint determined by the water level data is added to the loss function used in model training, so that the model can quickly correct the prediction direction in the training process, and finally output the water level parameters meeting the actual engineering requirements.

[0007] In an optional implementation, based on the reservoir storage curve constraint and the reservoir bottom elevation distribution interval, the coefficient of the reservoir storage curve simulation model is determined by using the Monte Carlo method, including: determining the reservoir bottom elevation distribution interval according to the target area digital elevation model; generating the reservoir bottom elevation randomly in the reservoir bottom elevation distribution interval, and generating the coefficient randomly in a preset interval; substituting the reservoir bottom elevation and the coefficient into the reservoir storage curve model, and judging whether the reservoir storage curve simulation model meets the reservoir storage curve constraint; when the reservoir storage curve constraint is met, the corresponding coefficient is taken as the coefficient of the reservoir storage curve simulation model.

[0008] In the present application, the randomly generated coefficient is verified based on the reservoir bottom elevation. Thus, the coefficient is accurately adapted to the terrain and engineering requirements, the reservoir storage curve simulation model is more in line with the actual situation, and a solid foundation is laid for subsequent water energy parameter calculation, thereby improving the accuracy and reliability of the parameter calculation of the pumped storage power station.

[0009] In an optional implementation, the reservoir storage curve constraint includes a reservoir storage curve monotonicity constraint, a reservoir storage curve convexity constraint and a reservoir bottom volume constraint.

[0010] In the present application, by setting the reservoir storage curve monotonicity constraint, the reservoir storage curve convexity constraint and the reservoir bottom volume constraint, the determined model can meet the terrain rules and physical rules.

[0011] In an optional implementation, the reservoir bottom elevation distribution interval is determined according to the target area digital elevation model, including: taking a preset area range of a pumped storage power station to be established as a target area, and constructing a digital elevation model of the target area; determining the terrain maximum elevation and the terrain minimum elevation according to the digital elevation model; taking the terrain maximum elevation and the terrain minimum elevation as the distribution interval of the reservoir bottom elevation, and the distribution interval is subject to uniform probability distribution.

[0012] In the present application, the target area is generated according to the geographical range, so that the finally trained model can be more in line with the characteristics of the target area.

[0013] In an optional embodiment, the reservoir bottom elevation distribution interval includes a lower reservoir bottom elevation distribution interval and an upper reservoir bottom elevation distribution interval, and the reservoir bottom elevation is randomly generated in the reservoir bottom elevation distribution interval, including: generating a lower reservoir bottom elevation in the lower reservoir bottom elevation distribution interval with a uniform probability, and generating an upper reservoir bottom elevation in the upper reservoir bottom elevation distribution interval with a uniform probability; judging the relationship between the generated lower reservoir bottom elevation and the upper reservoir bottom elevation; when the lower reservoir bottom elevation is greater than the upper reservoir bottom elevation, the lower reservoir bottom elevation and the upper reservoir bottom elevation are exchanged; judging whether the difference between the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation satisfies the applicable water head of the to-be-established pumped storage power station; when it is satisfied, the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation are taken as the randomly generated reservoir bottom elevation.

[0014] In the present application, by judging and verifying the generated reservoir bottom elevation, the reservoir bottom elevation generation is more in line with the actual terrain and engineering requirements, ensuring that the upper reservoir bottom elevation is higher than the lower reservoir bottom elevation and satisfying the applicable water head constraint, providing reasonable and engineering-logic-compliant basic data for subsequent reservoir capacity curve construction, water energy parameter calculation, etc., and improving the scientificity and reliability of pumped storage power station parameter calculation.

[0015] In an optional embodiment, the water level data includes a normal storage level and a dead storage level, and a penalty constraint is added in a preset loss function according to the water level data in the water energy parameter, to obtain a loss function of the preset machine learning model, including: determining a first penalty constraint according to the relationship between the normal storage level and the dead storage level; determining a second penalty constraint according to the relationship between the dead storage level and the reservoir bottom elevation, the reservoir bottom elevation being determined according to the to-be-established pumped storage power station; and determining the loss function of the preset machine learning model according to the preset loss function and the first penalty constraint and the second penalty constraint.

[0016] In an optional embodiment, the first penalty constraint is expressed by the following formula:

[0017] In the formula, represents the normal storage level of the upper reservoir, represents the dead storage level of the upper reservoir, represents the normal storage level of the lower reservoir, represents the dead storage level of the lower reservoir. The second penalty constraint is expressed by the following formula:

[0018] In the formula, Z u represents the reservoir bottom elevation of the upper reservoir, represents the dead storage level of the upper reservoir, represents the reservoir bottom elevation of the lower reservoir, represents the dead storage level of the lower reservoir.

[0019] In the present invention, the penalty constraint is determined according to the relationship between the water level data and the elevation data in the formula, which can ensure the rationality of the water level operation and the matching of the water level and the terrain.

[0020] In the second aspect, the present invention provides a device for calculating water energy parameters of a pumped-storage power station based on machine learning, the device comprising: a coefficient determination module, for determining the coefficients of a reservoir capacity curve simulation model using the Monte Carlo method based on the reservoir capacity curve constraints and the reservoir bottom elevation distribution interval; a water energy parameter optimization module, for inputting multiple groups of reservoir capacity curves determined by the reservoir capacity curve simulation model into a preset optimization model to obtain water energy parameters corresponding to each group of reservoir capacity curves; a loss function construction module, for adding penalty constraints to a preset loss function according to the water level data in the water energy parameters to obtain the loss function of the preset machine learning model; a training module, for training the preset machine learning model with the coefficients and reservoir bottom elevation as input and the water level data in the water energy parameters as output to obtain a water level calculation model; a water energy parameter calculation module, for determining the water level data of the pumped-storage power station to be established according to the reservoir capacity curve coefficients of the pumped-storage power station to be established and the water level calculation model, and calculating the water energy parameters.

[0021] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the method for calculating water energy parameters of a pumped-storage power station based on machine learning according to the above-mentioned first aspect or any corresponding embodiment thereof.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for calculating the water energy parameters of a pumped-storage power station based on machine learning according to the above-mentioned first aspect or any corresponding embodiment thereof.

[0023] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute the method for calculating water energy parameters of a pumped-storage power station based on machine learning according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 is a flowchart of a machine learning-based pumped storage power station water energy parameter calculation method according to an embodiment of the present application; Figure 2 is a structural block diagram of a machine learning-based pumped storage power station water energy parameter calculation device according to an embodiment of the present application; Figure 3 is a hardware structure schematic diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] To make the objects, technical solutions and advantages of embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0027] According to an embodiment of the present application, a machine learning-based pumped storage power station water energy parameter calculation method embodiment is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0028] In the present embodiment, a machine learning-based pumped storage power station water energy parameter calculation method is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 1 is a flowchart of a machine learning-based pumped storage power station water energy parameter calculation method according to an embodiment of the present application, as shown in Figure 1 the flowchart includes the following steps: Step S101, based on the reservoir capacity curve constraint and the reservoir bottom elevation distribution interval, the coefficients of the reservoir capacity curve simulation model are determined by using the Monte Carlo method; wherein the reservoir capacity curve simulation model is a model representing the relationship between the water level and the reservoir capacity of the reservoir, or a model describing the reservoir capacity curve, which can be represented by the following formula:

[0029] In the formula, , , and represent the coefficients of the model, Z represents the water level, and V represents the reservoir capacity. It should be noted that the formula is only an example of the model, and in other embodiments, other formulas can be used as the reservoir capacity curve simulation model according to actual conditions.

[0030] Specifically, in the embodiment, the coefficients of the reservoir capacity curve simulation model are determined, so that a plurality of reservoir capacity curves can be simulated by the model. In determining the coefficients, in order to reduce the amount of calculation and avoid the problems such as inaccuracy caused by random determination of the coefficients, the reservoir capacity curve constraint and the reservoir bottom elevation distribution interval are set. The reservoir capacity curve constraint can be understood as a physical constraint condition that needs to be met in the reservoir capacity curve simulation model, that is, in combination with the reservoir bottom elevation in the reservoir bottom elevation distribution interval, the model needs to meet the reservoir capacity curve constraint when the Monte Carlo method is used to determine the model coefficients.

[0031] It should be noted that the Monte Carlo method is a numerical statistical method of random sampling and statistical analysis. In the embodiment, the reservoir bottom elevation distribution interval is used as the sampling boundary of the Monte Carlo method, and the reservoir capacity curve constraint is used as the screening standard of the Monte Carlo method, so that the determination of the model coefficients that meet the reservoir capacity curve constraint in the reservoir bottom elevation distribution interval is realized.

[0032] In addition, in the pumped storage power station, the upper reservoir and the lower reservoir are included, and the water head difference between the upper reservoir and the lower reservoir is used to generate water power. Therefore, the reservoir capacity curve simulation model includes the upper reservoir capacity curve simulation model and the lower reservoir capacity curve simulation model, and the coefficients of the two models need to be determined in the manner of the step.

[0033] In step S102, a plurality of groups of reservoir capacity curves determined by the reservoir capacity curve simulation model are input into a preset optimization model to obtain water power parameters corresponding to each group of reservoir capacity curves. Specifically, after the coefficients of the upper reservoir capacity curve simulation model and the lower reservoir capacity curve simulation model are determined, the two models can be used to determine the upper reservoir capacity curve and the lower reservoir capacity curve, respectively. The coefficients of the two models determined by the Monte Carlo method according to the above step can include a plurality of coefficients, respectively, so that a plurality of groups of reservoir capacity curves can be obtained. Each group of reservoir capacity curves includes an upper reservoir capacity curve (corresponding to an upper reservoir capacity curve simulation model) and a lower reservoir capacity curve (corresponding to a lower reservoir capacity curve simulation model).

[0034] Specifically, the preset optimization model is used to determine the more optimal water-energy parameters corresponding to each group of reservoir capacity curves, thereby providing samples for subsequent machine learning training. The water-energy parameters include normal water levels of upper and lower reservoirs of the power station, regulation reservoir capacity, dead water level, installed capacity of the power station, maximum water head, minimum water head, rated water head, maximum lift, minimum lift, etc. For the preset optimization model, a water-energy parameter calculation model in related technologies can be used, such as a pumped storage power station water-energy parameter optimization mathematical model containing a maximum scale criterion objective function and constraint conditions used in pumped storage power station water-energy parameter calculation based on the maximum scale criterion, or a pumped storage power station water-energy parameter optimization mathematical model considering standby reservoir capacity used in pumped storage power station water-energy parameter optimization considering standby reservoir capacity, and the like. Thus, the calculation of water-energy parameters can be implemented with reference to related technologies, and the specific calculation process of water-energy parameters is not limited in this embodiment.

[0035] In step S103, a penalty constraint is added in the preset loss function according to the water level data in the water-energy parameters, to obtain a loss function of the preset machine learning model. Specifically, in the machine learning model mentioned in related technologies, the loss function mostly uses absolute error loss, mean square error loss, or cross-entropy, etc., to realize model optimization by measuring the relationship between the true value and the predicted value. In this embodiment, on the basis of the preset loss function (which can be the loss function used in related technologies), the physical constraints of the pumped storage power station are considered, thereby constructing a loss function including the preset loss function and the penalty constraint.

[0036] Meanwhile, when training the preset machine learning model, the water level data in the water-energy parameters is used as the output. Based on this, the penalty constraint is determined based on the water level data in this embodiment, so that the model can quickly correct the prediction direction in the training process, and finally output the water level parameters that meet the actual engineering requirements.

[0037] In step S104, the preset machine learning model is trained with the coefficient and the reservoir bottom elevation as input and the water level data in the water-energy parameters as output, to obtain a water level calculation model. The preset machine learning model can be a deep learning model such as a neural network model, and the type of machine learning model used in this embodiment is not limited. Since the calculation of the water-energy parameters of the pumped storage power station is based on the water level data, the water level data is extracted from the calculated water-energy parameters as the output sample for model training in this embodiment. It should be noted that when calculating the water-energy parameters based on the multiple groups of reservoir capacity curves in the foregoing step, only the water level data can be calculated to determine the output sample for model training.

[0038] Specifically, according to the above analysis, a pumped-storage power station includes an upper reservoir and a lower reservoir. Therefore, the coefficients here include the coefficients of the upper reservoir capacity curve simulation model, the coefficients of the lower reservoir capacity curve simulation model, the upper reservoir bottom elevation, and the lower reservoir bottom elevation. The upper and lower reservoir bottom elevations are determined from the reservoir bottom elevation distribution range when determining the model coefficients. Water level data includes the normal water level of the upper reservoir, the dead water level of the upper reservoir, the normal water level of the lower reservoir, and the dead water level of the lower reservoir. When training the model, one can refer to the model training process in related art, input samples into the model to obtain model predictions, substitute the model predictions and output samples (i.e., true values) into the model's loss function, and adjust the model parameters based on the calculated loss function. By repeatedly repeating this process, i.e., inputting different samples, one ultimately obtains a trained model, i.e., a water level calculation model.

[0039] Step S105 determines the water level data for the pumped-storage power station to be established based on the reservoir capacity curve coefficient and the water level calculation model, and calculates the hydropower parameters. Specifically, for the pumped-storage power station to be established, the corresponding reservoir capacity curve can be determined based on terrain information, thereby fitting the reservoir capacity curve coefficient. This terrain information refers to the topography of the pumped-storage power station's construction location and can be obtained from a topographic map. Based on this acquired terrain information (including the reservoir bottom elevation), the water surface area corresponding to different water levels is determined. Then, a volumetric algorithm, such as the pyramid volume method, is used to calculate the volume corresponding to different water levels to obtain the reservoir capacity. This generates multiple water level-reservoir capacity data (from which the reservoir capacity curve can be further derived). Finally, based on this data, a polynomial fit is performed using the reservoir capacity curve simulation model to obtain the reservoir capacity curve coefficient corresponding to the pumped-storage power station.

[0040] The determined reservoir capacity curve coefficient and reservoir bottom elevation are then input into the trained water level calculation model to obtain the water level data corresponding to the pumped-storage power station. Furthermore, water level parameters are determined based on this water level data. For example, based on the normal water storage level and dead water level of the upper and lower reservoirs, combined with the reservoir capacity curve, the reservoir capacity corresponding to each water level can be obtained. Based on the water level data (normal water storage level and dead water level of the upper and lower reservoirs), the reservoir capacity corresponding to these two water levels can be found on the reservoir capacity curve, and the adjusted reservoir capacity can be obtained by subtracting them. Based on the water level and reservoir capacity, the hydraulic head and storage energy are calculated, and then the installed capacity is calculated. It should be noted that the process of calculating hydropower parameters using water level data can be implemented by referring to related technologies and will not be elaborated here.

[0041] The method for calculating water energy parameters of a pumped storage power station based on machine learning provided in the embodiments of the present application can quickly calculate water energy parameters of a pumped storage power station by determining the reservoir storage curve coefficient and corresponding water level data as input and output samples to train a preset machine learning model, thereby avoiding the problem of low calculation efficiency and long time consumption caused by iterative calculation of various optimization models one by one in the related art. Meanwhile, the reservoir storage curve constraint is considered when the reservoir storage curve coefficient is determined, which can reduce the calculation amount and avoid the problem of inaccuracy possibly caused by random determination of the coefficient; and the penalty constraint determined by the water level data is added to the loss function used in the model training, which can enable the model to quickly correct the prediction direction in the training process, and finally output the water level parameters meeting the actual engineering requirements.

[0042] In the embodiments, a method for calculating water energy parameters of a pumped storage power station based on machine learning is provided, and the flow includes the following steps. In step S201, the coefficient of the reservoir storage curve simulation model is determined by using the Monte Carlo method based on the reservoir storage curve constraint and the reservoir bottom elevation distribution interval.

[0043] Specifically, step S201 includes the following steps. In step S2011, the reservoir bottom elevation distribution interval is determined according to the digital elevation model of the target area. Specifically, the digital elevation model refers to a digital model in which the topographic elevation information of the target area is stored in the form of a regular grid or an irregular triangular net by digital means. Therefore, the elevation information can be determined from the model to obtain the reservoir bottom elevation distribution interval.

[0044] In an optional embodiment, step S2011 includes the following steps. In step a1, the digital elevation model of the target area is constructed by taking a preset area range in which the pumped storage power station is to be established as the target area. Specifically, the target area can be a relatively large area, for example, a province or an area with similar topography, so that the machine learning model trained subsequently can be more in line with the characteristics of the target area. In the embodiments, the area within a certain range of the area in which the pumped storage power station is to be established, for example, a range of several hundred kilometers or one thousand kilometers, is taken as the target area. When the pumped storage power station is established in the target area, the water level calculation model trained in the embodiments can be used to determine. However, if the pumped storage power station is established outside the target area, the data of the model needs to be reacquired and trained by referring to the method of the embodiments to ensure the accuracy of the calculated water energy parameters.

[0045] After the target region is determined, topographic data of the region can be acquired to construct a digital elevation model. The topographic data can be acquired based on a public geographic data platform, aerial remote sensing measurement, or ground measurement, and the present embodiment does not make a specific limitation thereto.

[0046] In step a2, the maximum terrain elevation and the minimum terrain elevation are determined according to the digital elevation model. Specifically, the digital elevation model stores the terrain elevation information of the target region, and thus the maximum terrain elevation Zmax and the minimum terrain elevation Zmin can be extracted therefrom.

[0047] In step a3, the maximum terrain elevation and the minimum terrain elevation are taken as the distribution interval of the reservoir bottom elevation, and within the distribution interval, the reservoir bottom elevation is subject to a uniform probability distribution. Specifically, the pumped storage power station includes an upper reservoir and a lower reservoir, and thus the distribution interval herein includes the distribution interval of the upper reservoir bottom elevation and the distribution interval of the lower reservoir bottom elevation, and within the corresponding distribution interval, the reservoir bottom elevation is uniformly distributed. That is: Lower reservoir bottom elevation Z d : subject to a uniform probability distribution: Z d ~ U(Zmin, Zmax); Upper reservoir bottom elevation Z u : subject to a uniform probability distribution: Z u ~ U(Zmin, Zmax).

[0048] In step S2012, the reservoir bottom elevation is randomly generated within the distribution interval of the reservoir bottom elevation, and the coefficient is randomly generated within a preset interval. Specifically, since the reservoir bottom elevation is subject to a uniform probability distribution within the distribution interval, the reservoir bottom elevation is randomly generated by taking values from the distribution interval with the same probability. As for the coefficient of the reservoir capacity curve simulation model, in order to speed up the calculation, the value range (i.e., the preset interval) of the coefficient can be determined in advance, and then the coefficient is generated within the value range. For example, A3 can take values between -0.01 and 0.01, A2 can take values between -1 and 1, A1 can take values between -1000 and 1000, and A0 can be randomly taken.

[0049] In an alternative embodiment, the above step S222 includes: randomly generating the lower reservoir bottom elevation within the distribution interval of the lower reservoir bottom elevation with a uniform probability, and randomly generating the upper reservoir bottom elevation within the distribution interval of the upper reservoir bottom elevation with a uniform probability; judging the relationship between the generated lower reservoir bottom elevation and the upper reservoir bottom elevation; when the lower reservoir bottom elevation is greater than the upper reservoir bottom elevation, the lower reservoir bottom elevation and the upper reservoir bottom elevation are exchanged; judging whether the difference between the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation satisfies the applicable water head of the pumped storage power station to be established; when satisfied, the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation are taken as the randomly generated reservoir bottom elevation.

[0050] Specifically, for the reservoir bottom elevation randomly generated in the distribution interval, it is also necessary to determine whether it meets the relevant requirements first. First, because the upper reservoir bottom elevation and the lower reservoir bottom elevation are randomly generated in the same interval, the lower reservoir bottom elevation may be greater than the upper reservoir bottom elevation. At this time, the upper reservoir bottom elevation and the lower reservoir bottom elevation can be exchanged, that is, the randomly generated lower reservoir bottom elevation is taken as the upper reservoir bottom elevation, and the randomly generated upper reservoir bottom elevation is taken as the lower reservoir bottom elevation. When the upper reservoir bottom elevation is greater than the lower reservoir bottom elevation, it is necessary to further determine whether the difference (i.e. the height difference) between the upper reservoir bottom elevation and the lower reservoir bottom elevation meets the applicable water head. Among them, any pumped storage power station is designed with the minimum applicable water head Zhmin and the maximum applicable water head Zhmax when manufactured, both of which are greater than 0. For example, Zhmin=50m and Zhmax=800m in a certain area. After the height difference is calculated, it is necessary to determine whether it is between the minimum applicable water head and the maximum applicable water head. When it exceeds the range, the upper reservoir elevation and the lower reservoir elevation need to be regenerated until the above requirements are met, and the upper reservoir elevation and the lower reservoir elevation meeting the above requirements are taken as a set of reservoir bottom elevations.

[0051] Step S2013, substituting the reservoir bottom elevation and the coefficient into the reservoir capacity curve model, and determining whether the reservoir capacity curve simulation model meets the reservoir capacity curve constraint.

[0052] Step S2014, when the reservoir capacity curve constraint is met, the corresponding coefficient is taken as the coefficient of the reservoir capacity curve simulation model.

[0053] Among them, the reservoir capacity curve constraint includes the monotonicity constraint of the reservoir capacity curve, the convexity constraint of the reservoir capacity curve and the reservoir bottom volume constraint. Specifically, the monotonicity constraint of the reservoir capacity curve is expressed as dV / dZ>0, which can be understood as the reservoir capacity increases with the increase of Z, that is, the reservoir capacity curve simulation model formed by the constraint meets the physical law; the convexity constraint of the reservoir capacity curve is expressed as , which can be understood as the increment of the reservoir capacity is also increasing with the increase of Z, that is, the reservoir capacity curve simulation model formed by the constraint meets the terrain law; the reservoir bottom volume constraint is expressed as V(Z0)=0, which can be understood as for each reservoir, the reservoir bottom elevation must be 0.

[0054] Specifically, in the embodiment, the rejection sampling method is used in determining the model coefficients, that is, if the generated coefficients make the model meet the reservoir capacity curve constraint, the coefficients are accepted; if not, the coefficients are rejected. In determining the model coefficients, the coefficients of the upper reservoir capacity curve simulation model and the coefficients of the lower reservoir capacity curve simulation model need to be determined respectively. In determining the coefficients of the upper reservoir capacity curve simulation model, the randomly generated upper reservoir bottom elevation (as the water level) and the coefficients are substituted into the model to calculate the reservoir capacity. Then, it is judged whether the relationship between the reservoir capacity and the water level meets the above reservoir capacity curve constraint, if yes, the coefficients are accepted, if not, the coefficients are rejected. Then, the next upper reservoir bottom elevation and the coefficients are substituted into the model to continue to judge, and finally a plurality of sets of coefficients of the upper reservoir capacity curve simulation model meeting the reservoir capacity curve constraint are obtained, and a set of coefficients can obtain an upper reservoir capacity curve, so a plurality of sets of coefficients can obtain a plurality of upper reservoir capacity curves. Similarly, the lower reservoir bottom elevation and the coefficients are substituted into the model, and a plurality of sets of coefficients of the lower reservoir capacity curve simulation model can be determined by the rejection sampling method, and further a plurality of lower reservoir capacity curves are obtained.

[0055] It should be noted that in the above random generation of the reservoir bottom elevation, a plurality of sets of reservoir bottom elevations are randomly generated, and in this step, the upper reservoir bottom elevation and the lower reservoir bottom elevation in a set of reservoir bottom elevations are respectively substituted into the corresponding model to obtain the corresponding coefficients and the reservoir capacity curve. Therefore, each set of reservoir bottom elevations can correspond to determine a set of reservoir capacity curves, and finally a plurality of sets of reservoir capacity curves can be determined.

[0056] In step S202, a plurality of sets of reservoir capacity curves determined by the reservoir capacity curve simulation model are input into a preset optimization model to obtain water-energy parameters corresponding to each set of reservoir capacity curves; for details, please refer to Figure 1 The step S102 of the embodiment shown in the figure will not be described here again.

[0057] In step S203, a penalty constraint is added in a preset loss function according to the water level data in the water-energy parameters, and a loss function of a preset machine learning model is obtained.

[0058] Specifically, the above step S203 includes: In step S2031, a first penalty constraint is determined according to the relationship between the normal storage level and the dead storage level. Specifically, the first penalty constraint is expressed by the following formula:

[0059] In the formula, the upper reservoir normal storage level is denoted by h n u p p e r, the upper reservoir dead storage level is denoted by h d u p p e r, the lower reservoir normal storage level is denoted by h n l o w e r, the lower reservoir dead storage level is denoted by h d l o w e r.

[0060] In step S2032, the second penalty constraint is determined according to the relationship between the dead water level and the reservoir bottom elevation, and the reservoir bottom elevation is determined according to the to-be-established pumped storage power station. Specifically, the second penalty constraint is expressed by the following formula:

[0061] In the formula, h represents the upper reservoir bottom elevation, h represents the lower reservoir bottom elevation, h represents the upper reservoir dead water level, and h represents the lower reservoir dead water level. Z u h represents the upper reservoir bottom elevation, h represents the upper reservoir dead water level, h represents the lower reservoir bottom elevation, h represents the lower reservoir dead water level. It should be noted that when the penalty constraint is used in the training process of the preset machine learning model, the upper reservoir bottom elevation and the lower reservoir bottom elevation can be substituted into the upper reservoir bottom elevation and the lower reservoir bottom elevation determined according to the topographic information corresponding to the to-be-established pumped storage power station.

[0062] In step S2033, the loss function of the preset machine learning model is determined according to the preset loss function and the first penalty constraint and the second penalty constraint. Specifically, when the preset loss function adopts the mean squared error (MSE), the loss function of the preset machine learning model is expressed by the following formula:

[0063]

[0064] In the formula, λ represents a penalty coefficient. By adding the first penalty constraint and the second penalty constraint to the preset loss function, the water level logical inversion can be avoided, and the rationality of the reservoir operation can be ensured.

[0065] In step S204, the preset machine learning model is trained by taking the coefficient and the reservoir bottom elevation as input and taking the water level data in the water-energy parameter as output to obtain a water level calculation model. For details, refer to step S104 of the embodiment shown in Figure 1 and will not be described here again.

[0066] In step S205, the water level data of the to-be-established pumped storage power station is determined according to the reservoir capacity curve coefficient of the to-be-established pumped storage power station and the water level calculation model, and the water-energy parameter is calculated. For details, refer to step S105 of the embodiment shown in Figure 1 and will not be described here again.

[0067] As a specific application embodiment of the embodiment of the present application, the pumped storage power station water-energy parameter calculation method based on machine learning is implemented by the following process: S1, determine a target region, and obtain DEM (Digital Elevation Model) data of the target region. The target region should be determined in combination with the terrain features. For example, mountainous topography and hilly and plain topography should be taken as different target regions to train and apply the machine learning model respectively, so that the machine learning model formed is more in line with the characteristics of the target region. In this way, a general model is generated in the entire target region, that is, the machine learning model trained in this embodiment can be used in the target region.

[0068] S2, a model for generating upper and lower reservoir capacity curves considering physical constraints is constructed, and a large number of upper and lower reservoir capacity curve samples are generated by using the Monte Carlo method. Each set of upper and lower reservoir capacity curves includes a relationship curve between the water level and the capacity of the upper reservoir and a relationship curve between the water level and the capacity of the lower reservoir, and each set of curves is generated based on the following key parameters, i.e., the upper reservoir bottom elevation and the lower reservoir bottom elevation.

[0069] Specifically, the above step S2 includes the following steps: S21, the distribution range of the reservoir bottom elevation is determined considering the terrain features and the characteristics of the pumped storage unit.

[0070] First, the terrain feature parameters are extracted based on the digital elevation model (DEM) of the target region, mainly the maximum elevation Zmax and the minimum elevation Zmin of the target region.

[0071] Then, the distribution range of the reservoir bottom elevation is determined: Lower reservoir bottom elevation Z d obeys a uniform probability distribution: Z d ~ U(Zmin, Zmax). Upper reservoir bottom elevation Z u obeys a uniform probability distribution: Z u ~ U(Zmin, Zmax).

[0072] S22, a reservoir capacity curve simulation model is established.

[0073] The model is expressed by the following formula:

[0074] The model meets the following constraints: 1) Monotonicity constraint of the capacity curve: dV / dZ>0, that is, the capacity curve is increasing with the increase of Z.

[0075] 2) Convexity constraint of the capacity curve: that is, the increment of the capacity is also increasing with the increase of Z.

[0076] 3) Reservoir bottom volume constraint: V(Z0) = 0. For each reservoir, the reservoir bottom elevation must be 0.

[0077] S23, generating N sets of upper and lower reservoir capacity curve data, where N is a large number, for example, N=100000. Specifically, the upper reservoir capacity curve and the lower reservoir capacity curve are randomly simulated to obtain the power station capacity curve. For each simulation, the specific steps include: S231, randomly generate Z within the distribution interval d , Z u If Z d >Z u , then swap Z d and Z u ; Calculate Z h = Z u - Z d , if Z h If the requirements are not met, regenerate until they are met. h Should meet the following requirements: Zhmin≤Z h ≤Zhmax, where Zhmin and Zhmax are the minimum applicable water head and the maximum applicable water head considering the design and manufacturing level of the pumped storage unit, both of which are greater than 0. For example, in a certain area, Zhmin=50m and Zhmax=800m.

[0078] S232, Z-based d and Z u , use the rejection sampling method to randomly generate the four coefficients A3, A2, A1, and A0 (to ensure that the generated coefficients can meet the various constraints of the storage capacity curve model), and then use the model to generate the storage capacity curve of the lower reservoir and the upper reservoir respectively. When simulating the storage capacity curve, the four coefficients A3, A2, A1, and A0 can be positive or negative. In actual work, these four numbers can also be given a range to speed up the calculation. For example, according to experience, A3 can be between -0.01 and 0.01, A2 between -1 and 1, A1 between -1000 and 1000, and A0 can be randomly selected.

[0079] S3, for each pumped storage sample in S2, the optimization model is used to calculate its hydraulic parameters independently according to the storage capacity curves of its upper and lower reservoirs.

[0080] S4, obtain input and output sample data of N pumped storage power stations, use the input and output data samples, consider physical constraints to improve the artificial neural network, and then train it to obtain a machine learning model for calculating water energy parameters in the target area.

[0081] Specifically, the above step S4 includes: S41, obtain all input and output data. Among them, the input sample includes multiple groups of samples, each group of samples includes the coefficient of the upper reservoir storage curve simulation model, the coefficient of the lower reservoir storage curve simulation model, the upper reservoir bottom elevation and the lower reservoir bottom elevation. For the output sample, the normal storage level of the upper reservoir, the dead water level of each power station are obtained from the water energy parameter calculation results as the output sample.

[0082] S42, use the input and output data samples to improve the artificial neural network considering the physical constraints, retrain, and obtain the machine learning model for water energy parameter calculation. Among them, when improving, considering the physical constraints of the pumped storage power station, the loss function is constructed as shown in the following formula:

[0083]

[0084] In the formula, is a penalty coefficient, which can be valued according to the actual situation, for example, it can be taken as 1000.

[0085] S5, when using, first obtain the coefficients of the upper and lower reservoir storage curve polynomials of the to-be-calculated power station, input the artificial neural network, and obtain the normal storage level of the upper reservoir, the dead water level of the power station, the normal storage level of the lower reservoir, and the dead water level.

[0086] Specifically, in actual design, the reservoir curve of pumped storage is often calculated according to the topographic map, so first calculate the reservoir curve of the upper and lower reservoirs, that is, the data of water level and reservoir capacity, and then fit the polynomials respectively to obtain several parameters A 3， A 2， A 1， A 0， B 3， B 2， B 1， B0, and Z d , Z u , and then input into the machine learning model.

[0087] S6, calculate the water energy parameters (installed capacity, water head, etc.) of the power station according to the above characteristic water level.

[0088] In this embodiment, a pumped storage power station water energy parameter calculation device based on machine learning is also provided, which is used to implement the above embodiments and preferred embodiments, and has been described and will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware is also possible and is contemplated.

[0089] The embodiment provides a machine learning-based water energy parameter calculation device for pumped storage power stations. Figure 2 As shown in the figure, the device comprises: A coefficient determination module 21 configured to determine the coefficient of the reservoir capacity curve simulation model by using the Monte Carlo method based on the reservoir capacity curve constraint and the reservoir bottom elevation distribution interval. A water energy parameter optimization module 22 configured to input the plurality of groups of reservoir capacity curves determined by the reservoir capacity curve simulation model into a preset optimization model to obtain the water energy parameters corresponding to each group of reservoir capacity curves. A loss function construction module 23 configured to add a penalty constraint in a preset loss function according to the water level data in the water energy parameters to obtain the loss function of the preset machine learning model. A training module 24 configured to train the preset machine learning model by taking the coefficient and the reservoir bottom elevation as input and the water level data in the water energy parameters as output to obtain a water level calculation model. A water energy parameter calculation module 25 configured to determine the water level data of the to-be-established pumped storage power station according to the reservoir capacity curve coefficient of the to-be-established pumped storage power station and the water level calculation model and calculate the water energy parameters.

[0090] In an optional implementation, the coefficient determination module comprises: an interval determination module configured to determine the reservoir bottom elevation distribution interval according to a target area digital elevation model; a random generation module configured to randomly generate the reservoir bottom elevation in the reservoir bottom elevation distribution interval and randomly generate the coefficient in a preset interval; a judgment module configured to substitute the reservoir bottom elevation and the coefficient into the reservoir capacity curve model and judge whether the reservoir capacity curve simulation model meets the reservoir capacity curve constraint; and a coefficient determination submodule configured to take the corresponding coefficient as the coefficient of the reservoir capacity curve simulation model when the reservoir capacity curve constraint is met.

[0091] In an optional implementation, the reservoir capacity curve constraint comprises a reservoir capacity curve monotonicity constraint, a reservoir capacity curve convexity constraint and a reservoir bottom volume constraint.

[0092] In an optional implementation, the interval determination module is specifically configured to take the preset area range of the to-be-established pumped storage power station as the target area, construct the digital elevation model of the target area, determine the terrain maximum elevation and the terrain minimum elevation according to the digital elevation model, and take the terrain maximum elevation and the terrain minimum elevation as the distribution interval of the reservoir bottom elevation, and the distribution interval is subject to a uniform probability distribution.

[0093] In an optional implementation, the reservoir bottom elevation distribution interval includes a lower reservoir bottom elevation distribution interval and an upper reservoir bottom elevation distribution interval, and the random generation module is specifically configured to randomly generate a lower reservoir bottom elevation in the lower reservoir bottom elevation distribution interval with a uniform probability and randomly generate an upper reservoir bottom elevation in the upper reservoir bottom elevation distribution interval with a uniform probability; determine the relationship between the generated lower reservoir bottom elevation and the upper reservoir bottom elevation; when the lower reservoir bottom elevation is greater than the upper reservoir bottom elevation, exchange the lower reservoir bottom elevation and the upper reservoir bottom elevation; determine whether the difference between the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation satisfies the applicable water head of the to-be-established pumped storage power station; when satisfied, the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation are taken as the randomly generated reservoir bottom elevation.

[0094] In an optional implementation, the water level data include a normal storage level and a dead storage level, and the loss function construction module is specifically configured to: determine a first penalty constraint according to the relationship between the normal storage level and the dead storage level; determine a second penalty constraint according to the relationship between the dead storage level and the reservoir bottom elevation, the reservoir bottom elevation being determined according to the to-be-established pumped storage power station; and determine the loss function of the preset machine learning model according to the preset loss function and the first penalty constraint and the second penalty constraint.

[0095] In an optional implementation, the first penalty constraint is expressed by the following formula:

[0096] In the formula, represents the normal storage level of the upper reservoir, represents the dead storage level of the upper reservoir, represents the normal storage level of the lower reservoir, represents the dead storage level of the lower reservoir. The second penalty constraint is expressed by the following formula:

[0097] In the formula, Z u represents the reservoir bottom elevation of the upper reservoir, represents the dead storage level of the upper reservoir, represents the reservoir bottom elevation of the lower reservoir, represents the dead storage level of the lower reservoir.

[0098] Further function descriptions of the above modules are the same as those of the corresponding embodiments, and will not be described here.

[0099] The embodiment of the application further provides a computer device having the above Figure 2 The machine learning-based pumped storage power station water energy parameter calculation device shown in the figure.

[0100] Please refer to Figure 3 , Figure 3is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 3 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 3 A processor 10 is taken as an example.

[0101] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0102] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0103] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created based on the use of a computer device for displaying a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0104] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0105] The computer device also comprises a communication interface 30 for communication of the computer device with other devices or communication networks.

[0106] The embodiments of the present application further provide a computer readable storage medium, and the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or stored in a remote storage medium or a non-transitory machine readable storage medium and downloaded from a network and stored in a local storage medium by computer code, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor, or programmable or special purpose hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk or a solid state disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that the computer, processor, microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code, when the software or computer code is accessed and executed by the computer, processor or hardware, the method shown in the above embodiments is implemented.

[0107] Part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, through the operation of the computer, the method and / or technical solutions according to the present application can be called or provided. Those skilled in the art should understand that the form of computer program instructions in computer readable medium includes but is not limited to source file, executable file, installation package file, etc., and accordingly, the way of computer program instructions executed by computer includes but is not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Here, the computer readable medium can be any available computer readable storage medium or communication medium accessible to the computer.

[0108] Although the embodiments of the present application are described in conjunction with the accompanying drawings, various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and changes fall within the scope defined by the appended claims.

Claims

1. A method for calculating water energy parameters of a pumped storage power station based on machine learning, characterized in that: The method comprises: Based on the storage capacity curve constraints and the distribution range of reservoir bottom elevation, the coefficients of the storage capacity curve simulation model are determined using the Monte Carlo method. Input multiple groups of storage capacity curves determined by the storage capacity curve simulation model into the preset optimization model to obtain the water energy parameters corresponding to each group of storage capacity curves; Adding a penalty constraint to a preset loss function according to the water level data in the water energy parameters to obtain a loss function of a preset machine learning model; The preset machine learning model is trained using the coefficient and the reservoir bottom elevation as input and the water level data in the water energy parameters as output to obtain a water level calculation model; The water level data of the pumped storage power station to be established is determined according to the storage capacity curve coefficient of the pumped storage power station to be established and the water level calculation model, and the water energy parameters are calculated.

2. The method according to claim 1, characterized in that Based on the storage capacity curve constraints and the distribution range of reservoir bottom elevation, the Monte Carlo method is used to determine the coefficients of the storage capacity curve simulation model, including: Determine the reservoir bottom elevation distribution range based on the digital elevation model of the target area; Randomly generate the reservoir bottom elevation within the reservoir bottom elevation distribution range, and randomly generate the coefficient within the preset range; Substituting the reservoir bottom elevation and coefficient into the reservoir capacity curve model, and determining whether the reservoir capacity curve simulation model satisfies the reservoir capacity curve constraint; When the storage capacity curve constraint is met, the corresponding coefficient is used as the coefficient of the storage capacity curve simulation model.

3. The method according to claim 1, characterized in that The storage capacity curve constraints include storage capacity curve monotonicity constraints, storage capacity curve convexity constraints and storage bottom volume constraints.

4. The method according to claim 2, characterized in that Determine the reservoir bottom elevation distribution range based on the target area digital elevation model, including: Taking the preset area where the pumped storage power station is to be built as the target area, a digital elevation model of the target area is constructed; Determining a maximum terrain elevation and a minimum terrain elevation based on the digital elevation model; The maximum elevation and the minimum elevation of the terrain are used as the distribution range of the reservoir bottom elevation, and within the distribution range, it obeys a uniform probability distribution.

5. The method according to claim 2, characterized in that The reservoir bottom elevation distribution interval includes a lower reservoir bottom elevation distribution interval and an upper reservoir bottom elevation distribution interval, and randomly generating the reservoir bottom elevation in the reservoir bottom elevation distribution interval includes: Randomly generating the lower reservoir bottom elevation with uniform probability within the lower reservoir bottom elevation distribution interval, and randomly generating the upper reservoir bottom elevation with uniform probability within the upper reservoir bottom elevation distribution interval; Determining the relationship between the generated lower reservoir bottom elevation and the upper reservoir bottom elevation; When the elevation of the lower reservoir bottom is greater than the elevation of the upper reservoir bottom, the elevation of the lower reservoir bottom is exchanged with the elevation of the upper reservoir bottom; determining whether the difference between the upper reservoir bottom elevation and the lower reservoir bottom elevation after the exchange satisfies the applicable water head of the pumped storage power station to be established; When the conditions are met, the exchanged upper reservoir bottom elevation and the lower reservoir bottom elevation are used as the randomly generated reservoir bottom elevation.

6. The method according to claim 1, characterized in that The water level data includes the normal water level and the dead water level. According to the water level data in the hydraulic parameters, a penalty constraint is added to the preset loss function to obtain the loss function of the preset machine learning model, including: determining a first penalty constraint according to the relationship between the normal water level and the dead water level; determining a second penalty constraint based on a relationship between the dead water level and the reservoir bottom elevation, wherein the reservoir bottom elevation is determined based on the pumped storage power station to be established; The loss function of the preset machine learning model is determined based on the preset loss function and the first penalty constraint and the second penalty constraint.

7. The method according to claim 6, characterized in that The first penalty constraint is expressed by the following formula: Where, Indicates the normal water level of the upper reservoir. Indicates the dead water level of the upper reservoir, Indicates the normal water level of the lower reservoir. Indicates the dead water level of the lower reservoir; The second penalty constraint is expressed by the following formula: Where, Z u Indicates the elevation of the upper reservoir bottom. Indicates the dead water level of the upper reservoir, Indicates the elevation of the lower reservoir bottom. Indicates the dead water level of the lower reservoir.

8. A device for calculating water energy parameters of a pumped storage power station based on machine learning, characterized in that: The device comprises: The coefficient determination module is used to determine the coefficients of the reservoir capacity curve simulation model using the Monte Carlo method based on the reservoir capacity curve constraints and the reservoir bottom elevation distribution interval; A water energy parameter optimization module is used to input multiple groups of storage capacity curves determined by the storage capacity curve simulation model into a preset optimization model to obtain the water energy parameters corresponding to each group of storage capacity curves; A loss function construction module, configured to add a penalty constraint to a preset loss function based on the water level data in the hydropower parameters to obtain a loss function of a preset machine learning model; A training module is used to train the preset machine learning model using the coefficient and reservoir bottom elevation as input and the water level data in the hydraulic parameters as output to obtain a water level calculation model; The water energy parameter calculation module is used to determine the water level data of the pumped storage power station to be established according to the storage capacity curve coefficient of the pumped storage power station to be established and the water level calculation model, and calculate the water energy parameters.

9. A computer device, characterized in that: include: 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 method for calculating water energy parameters of a pumped storage power station based on machine learning as described in any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, which are used to enable a computer to execute the method for calculating water energy parameters of a pumped-storage power station based on machine learning according to any one of claims 1 to 7.

Images