Drainage basin water-wind-light system power prediction model construction method and device and storage medium

Abstract

The invention relates to the technical field of electrical energy, and discloses a method and a device for constructing a power prediction model of a watershed water-wind-light complementary system, and a storage medium, the method for constructing the power prediction model of the watershed water-wind-light complementary system comprises the following steps: obtaining a data sample, the data sample comprises historical meteorological data, historical operation data and historical output data of the watershed water-wind-light system, and the output data comprises hydropower output data, wind power output data and photovoltaic output data; an initial power prediction sub-model is constructed on each water-wind-light subsystem, and all the initial power prediction sub-models on the water-wind-light system are connected in series based on the drainage basin runoff direction to construct an initial power prediction model; and taking the historical meteorological data and the historical operation data as training samples, taking the historical output data as labels, and training the initial power prediction model by using the data samples to obtain a power prediction model. Large-scale combined prediction can be performed on the drainage basin, and the overall utilization efficiency of the drainage basin is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power energy, and particularly relates to a method, device and storage medium for constructing a power prediction model of a basin water-wind-solar system. Background Art

[0002] With the construction and development of basin-level water-wind-solar multi-energy complementary bases in the lower reaches of the Jinsha River Basin, the Yalong River Basin, the Yellow River Basin, etc., the coordinated dispatching operation and comprehensive consumption and utilization of hydropower, wind power, and photovoltaic power are of great significance for ensuring the safe and stable operation of the power grid and improving the comprehensive efficiency of the basin. In the basin, the hydraulic connection between upstream and downstream reservoirs is close, and the incoming water and water level of the downstream reservoir are restricted by the regulation of the upstream reservoir. For the base of integrated water-wind-solar operation, due to the significant volatility and randomness of wind and solar power output, it is necessary to use hydropower regulation to suppress the fluctuations of wind and solar power, which further exacerbates the impact of upstream reservoir regulation on the downstream reservoir.

[0003] Most of the existing technologies conduct joint prediction for single energy or multi-energy in a regional scope, and have not carried out power joint prediction based on large-scale and multi-energy in the basin, so it is impossible to improve the overall utilization efficiency of various energy sources such as hydropower and new wind-solar energy in the basin and ensure the safe and stable operation of the power grid. Summary of the Invention

[0004] In view of this, the present invention provides a method, device and storage medium for constructing a power prediction model of a basin water-wind-solar system, so as to conduct large-scale joint prediction on the basin and improve the overall utilization efficiency of various energy sources such as hydropower and new wind-solar energy in the basin.

[0005] In a first aspect, the present invention provides a method for constructing a power prediction model of a basin water-wind-solar complementary system. The method for constructing a power prediction model of a basin water-wind-solar complementary system includes: obtaining data samples, where the data samples include historical meteorological data, historical operation data and historical output data of the basin water-wind-solar system, and the output data includes hydropower output data, wind power output data and photovoltaic output data; constructing initial power prediction sub-models on each water-wind-solar subsystem, and connecting all the initial power prediction sub-models on the water-wind-solar system in series based on the basin runoff direction to construct an initial power prediction model; using the historical meteorological data and historical operation data as training samples, using the historical output data as labels, and training the initial power prediction model with the data samples to obtain a power prediction model.

[0006] In this implementation, by separately constructing a power prediction sub-model for each hydro-wind-solar complementary subsystem in the entire basin and serially constructing a power prediction model for the power prediction sub-models within the basin scope, large-scale joint prediction of the basin can be performed. From the perspective of the entire basin and considering the impact of upstream and downstream reservoir regulation, this application can achieve power prediction for multiple hydro-wind-solar energy sources in the basin scope with upstream and downstream coordinated linkage, improve the accuracy of power prediction, and enhance the overall utilization efficiency of multiple energy sources such as hydropower and new wind-solar energy in the basin.

[0007] In an optional implementation, obtaining data samples includes: obtaining initial data for each season, where the initial data includes initial historical meteorological data, initial historical operation data, and historical output data; screening the initial historical meteorological data and initial historical operation data with a correlation degree greater than a preset value with respect to the historical output data according to each season to obtain data samples corresponding to each season; training an initial power prediction model using the data samples to obtain a power prediction model, including: training the initial power prediction model based on the data samples corresponding to each season respectively to obtain a power prediction model for each season.

[0008] In this implementation, considering the impact of seasons on the power of the hydro-wind-solar system and constructing models by season can make full use of the advantages of different models, train corresponding power prediction models for the hydro-wind-solar system in each season, and improve the model prediction accuracy. At the same time, by separately finding sample data with a greater correlation with historical output data for the four seasons, the data samples can be made more refined.

[0009] In an optional implementation, training the initial power prediction model based on the data samples corresponding to each season respectively to obtain a power prediction model for each season includes: for one season, training multiple initial power prediction models based on the data samples to obtain multiple power prediction models; constructing a decision matrix based on the multiple power prediction models; jointly evaluating the decision matrix using a fuzzy decision method, calculating the relative membership degree of each power prediction model, and determining the optimal power prediction model from the multiple power prediction models based on the principle of the maximum membership degree.

[0010] In this implementation, by selecting the best model from multiple power prediction models for subsequent power prediction in combination with the specific seasonal usage scenario, it can provide data support for the preparation of short-term scheduling plans and the like.

[0011] In an optional implementation, constructing a decision matrix based on multiple power prediction models includes: constructing a power prediction accuracy evaluation index matrix based on the certainty coefficient and root mean square error of each power prediction model; constructing a power supply capacity index matrix based on the load tracking risk coefficient and residual load standard deviation of each power prediction model; constructing a decision matrix based on the power prediction accuracy evaluation index matrix and the power supply capacity index matrix.

[0012] In an alternative embodiment, historical meteorological data and historical operation data are used as training samples, and historical output data is used as labels. Training the initial power prediction model with the data samples to obtain the power prediction model includes: when the hydropower, wind and photovoltaic subsystem is the first subsystem in the basin runoff direction, using the historical meteorological data and historical operation data as the training samples of the initial power prediction sub-model, and using the historical output data as the sample labels of the initial power prediction sub-model; when the hydropower, wind and photovoltaic subsystem is not the first subsystem in the basin runoff direction, using the historical meteorological data, historical operation data and the predicted discharge flow of the previous hydropower, wind and photovoltaic subsystem as the training samples of the initial power prediction sub-model, and using the historical output data as the sample labels of the initial power prediction sub-model; training the initial power prediction model based on the training samples and sample labels to obtain the power prediction model.

[0013] In this implementation, from the perspective of the whole basin and considering the influence of upstream and downstream reservoir regulation, different training samples are set for the leading subsystem and other subsystems. Considering the influence of the flow of the upstream hydropower, wind and photovoltaic subsystem on the downstream hydropower, wind and photovoltaic subsystem, adding the predicted discharge flow of the upstream to the training samples of the downstream hydropower, wind and photovoltaic subsystem can realize the power prediction of multiple hydropower, wind and photovoltaic energy sources in the basin range and the coordinated linkage between upstream and downstream, improve the accuracy of power prediction, and enhance the overall utilization efficiency of various energy sources such as hydropower and new wind and photovoltaic energy in the basin.

[0014] In an alternative embodiment, before using the historical meteorological data, historical operation data and the predicted discharge flow of the previous hydropower, wind and photovoltaic system as the training samples of the initial power prediction sub-model when the hydropower, wind and photovoltaic system is not the first subsystem in the basin runoff direction, it further includes: obtaining the basic data of the hydropower station in the hydropower, wind and photovoltaic system, where the basic data of the hydropower station includes the NHQ characteristic curve; calculating the predicted discharge flow of the previous hydropower, wind and photovoltaic system based on the predicted hydropower output data of the previous hydropower, wind and photovoltaic system and the basic data of the hydropower station.

[0015] In an alternative embodiment, training the initial power prediction model based on the training samples and sample labels to obtain the power prediction model includes: inputting the training samples into the initial power prediction model to obtain predicted output data; calculating the determination coefficient, root mean square error, load tracking risk coefficient and residual load standard deviation based on the predicted output data and sample labels; constructing a total loss function based on the determination coefficient, root mean square error, load tracking risk coefficient and residual load standard deviation; and training multiple times based on the total loss function to obtain the power prediction model.

[0016] In an alternative embodiment, constructing the total loss function based on the determination coefficient, root mean square error, load tracking risk coefficient and residual load standard deviation includes: constructing the first loss function based on the determination coefficient as: Wherein, is the first loss function, is the hydropower certainty coefficient of the k-th power prediction model, is the wind power certainty coefficient of the k-th power prediction model, is the photovoltaic power certainty coefficient of the k-th power prediction model; The second loss function is constructed based on the root mean square error as: where, is the second loss function, is the root mean square error of hydropower of the k-th power prediction model, is the root mean square error of wind power of the k-th power prediction model, is the root mean square error of photovoltaic power of the k-th power prediction model; The third loss function is constructed based on the load tracking risk coefficient as: l RL = RL; where, l RL is the third loss function, and RL is the load tracking risk coefficient. The fourth loss function is constructed based on the residual load standard deviation as: l SL = αSL; where, l SL is the fourth loss function, and SL is the residual load standard deviation; The total loss function is constructed as: where, Loss is the total loss function.

[0017] In this implementation manner, four sub-loss functions are constructed by comprehensively considering the certainty coefficient affecting the prediction accuracy, the root mean square error affecting the prediction accuracy, the load tracking risk coefficient affecting the output power tracking level, and the residual load standard deviation affecting the power grid supply and demand level, and the total loss function is obtained by combination, which is used to train the power prediction model. It can comprehensively consider various factors affecting the output power, realize the power prediction of multi-energy sources of water, wind and light and the coordinated linkage between upstream and downstream in the basin range, and improve the accuracy of power prediction.

[0018] In a second aspect, the present invention provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the method for constructing a power prediction model of a basin water-wind-light complementary system according to the first aspect or any corresponding implementation manner thereof.

[0019] In a third aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute the method for constructing a power prediction model of a basin water-wind-light complementary system according to the first aspect or any corresponding implementation manner thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 is a schematic flowchart of a method for constructing a power prediction model of a basin water-wind-solar complementary system according to an embodiment of the present invention;

[0022] Figure 2 is a flowchart of another method for constructing a power prediction model of a basin water-wind-solar complementary system according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of a water-wind-solar complementary system according to an embodiment of the present invention;

[0024] Figure 4 is a structural block diagram of a device for constructing a power prediction model of a basin water-wind-solar complementary system according to an embodiment of the present invention;

[0025] Figure 5 is a schematic hardware structure diagram of a computer device according to an embodiment of the present invention. Specific Embodiments

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0027] Regarding the current power prediction problem, it mainly conducts joint prediction for a single energy source or multiple energy sources within a regional scope. However, when conducting multi-energy power prediction, it is impossible to achieve power prediction for multiple water-wind-solar energy sources within a basin scope with coordinated linkage between upstream and downstream. Therefore, this application proposes a method for constructing a power prediction model of a basin water-wind-solar complementary system, which improves the accuracy of power prediction through joint basin prediction.

[0028] According to an embodiment of the present invention, an embodiment of a method for constructing a power prediction model of a basin water-wind-solar complementary system 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 the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0029] In this embodiment, a method for constructing a power prediction model of a basin water-wind-solar complementary system is provided. Figure 1 It is a flowchart of a method for constructing a power prediction model of a basin water-wind-solar complementary system according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment does not limit Figure 1 to the shown process sequence. As Figure 1 shown, the process includes the following steps:

[0030] Step S101, obtain data samples.

[0031] For a basin system, multiple reservoirs are built from upstream to downstream. Hydroelectric power stations are built in the reservoirs, and wind farms and photovoltaic power station sites are built on both banks of each reservoir. The hydroelectric power stations, wind farms, and photovoltaic power station sites on both banks of each reservoir form a water-wind-solar complementary system. All the water-wind-solar complementary subsystems in the basin form a complete water-wind-solar complementary system of the basin. This application obtains data related to the power generation of the water-wind-solar complementary system.

[0032] Among them, the data samples include the historical meteorological data, historical operation data, and historical output data of the basin water-wind-solar complementary system. The output data includes hydroelectric power output data, wind power output data, and photovoltaic power output data.

[0033] Specifically, collect the historical meteorological data of the control ranges of each hydroelectric power station, the sites of the wind farms and photovoltaic power stations in the basin. The meteorological data includes wind speed, wind direction, air pressure, precipitation, evaporation, irradiance, sunshine duration, air temperature, etc.

[0034] Collect the historical operation data of each hydroelectric power station, wind farm, and photovoltaic power station site. The operation data includes reservoir water level, reservoir inflow, reservoir outflow, power generation flow of hydroelectric generating units, inflow between reservoirs, etc.

[0035] Among them, the historical meteorological data can affect the wind power output data and photovoltaic power output data, and the historical operation data can affect the hydroelectric power output data.

[0036] Furthermore, the data samples also include the static parameter materials of the hydroelectric power stations and wind-solar power stations at the hydropower station level. Specifically, it includes the hydroelectric power station dispatching regulations, the water level-storage capacity relationship curve of the hydroelectric power station, the NHQ characteristic curve, the tail water level-discharge curve, the theoretical output curve of wind turbines, the characteristic curve of photovoltaic panels, the basic parameters of water, wind, and photovoltaic power stations, etc.

[0037] Furthermore, the data samples also include the data of the grid-level external transmission channels. Specifically, it includes the load curve, the capacity of the external transmission channel, the consumption area / province, the utilization hours of the channel, etc.

[0038] Step S102: Construct an initial power prediction sub-model for each hydro-wind-solar complementary subsystem, and based on the basin runoff direction, connect all the initial power prediction sub-models on the hydro-wind-solar complementary system in series to construct an initial power prediction model.

[0039] For each hydro-wind-solar complementary subsystem, establish an initial power prediction sub-model for jointly predicting the output of the hydropower station, wind farm, and photovoltaic power station. And name the hydro-wind-solar complementary subsystems in sequence as hydro-wind-solar complementary subsystem 1, hydro-wind-solar complementary subsystem 2,..., hydro-wind-solar complementary subsystem RN according to the basin runoff direction.

[0040] Among them, the initial power prediction sub-model is used to predict the output of the hydro-wind-solar complementary subsystem based on meteorological data and the operation data of the hydro-wind-solar complementary subsystem.

[0041] For the entire hydro-wind-solar complementary system, connect all the initial power prediction sub-models in series according to the basin runoff direction to obtain an initial power prediction model.

[0042] Among them, the initial power prediction sub-model is used to predict the output of the entire hydro-wind-solar complementary system based on meteorological data and the operation data of the entire hydro-wind-solar complementary system.

[0043] Step S103: Use the historical meteorological data and historical operation data as training samples, and the historical output data as labels, and train the initial power prediction model with the data samples to obtain a power prediction model.

[0044] Divide the data samples into a training set, a validation set, and a test set, and divide the training samples and sample labels from the training set. Specifically, use the historical meteorological data and historical operation data as training samples to input into the initial power prediction model, use the historical output data as labels, and compare with the prediction results of the initial power prediction model to adjust the model parameters of the initial power prediction model to obtain a power prediction model.

[0045] The method for constructing a power prediction model of a basin hydro-wind-solar complementary system provided in this embodiment separately constructs a power prediction sub-model for each hydro-wind-solar complementary subsystem in the entire basin, and connects the power prediction sub-models in series within the basin range to construct a power prediction model for large-scale joint prediction of the basin. This application can realize the power prediction of multiple energy sources of hydro, wind, and solar in the basin range and the coordinated linkage between upstream and downstream, improve the accuracy of power prediction, and enhance the overall utilization efficiency of various energy sources such as basin hydropower and new wind and solar energy from the perspective of the whole basin and considering the influence of upstream and downstream reservoir regulation.

[0046] In this embodiment, a method for constructing a power prediction model of a basin hydro-wind-solar complementary system is provided. Figure 2It is a flowchart of another method for constructing a power prediction model of a basin water-wind-solar complementary system according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not limited to Figure 2 the process sequence shown. As Figure 2 shown, this process includes the following steps:

[0047] Step S201, obtain data samples.

[0048] Specifically, the above step S201 includes:

[0049] Step S2011, obtain the initial data for each season.

[0050] The initial data includes initial historical meteorological data, initial historical operation data, and historical output data.

[0051] Specifically, collect the historical meteorological data within the control ranges of each hydropower station, wind farm, and photovoltaic power station site in the basin. The meteorological data includes wind speed, wind direction, air pressure, precipitation, evaporation, irradiance, sunshine duration, air temperature, etc.

[0052] Collect the historical operation data of each hydropower station, wind farm, and photovoltaic power station site. The operation data includes reservoir water level, reservoir inflow, reservoir outflow, hydropower unit generation flow, reservoir reach inflow, etc.

[0053] Collect the static parameter materials of hydropower stations and wind-solar power stations. Specifically, it includes hydropower station dispatching regulations, hydropower station water level-storage capacity relationship curves, NHQ characteristic curves, tail water level-discharge curves, theoretical output curves of wind turbines, photovoltaic panel characteristic curves, basic parameters of water, wind, and solar power stations, etc.

[0054] Collect the data of the grid-level external transmission channels. Specifically, it includes load curves, external transmission channel capacities, consumption regions / provinces, channel utilization hours, etc.

[0055] Furthermore, there may be missing data in the collected historical meteorological data, initial historical operation data, and historical output data.

[0056] In one implementation, the data with more than 8 consecutive missing periods is regarded as missing data and deleted. The outliers are identified using a box plot and processed using the K-nearest neighbor complement method. The specific processing method is:

[0057]

[0058] Among them, is the value of the i-th period after processing using the K-nearest neighbor complement method, x i-k and x i+k are the outlier x iSample values in the k-th time period before and after.

[0059] In one implementation, if there is missing wind farm output data, the missing wind farm output data is deduced using the theoretical processing curve of wind turbines and interpolation is performed.

[0060] In one implementation, if there is missing photovoltaic power station output data, the missing photovoltaic power station output data is deduced using the characteristic curve of photovoltaic panels and interpolation is performed.

[0061] Step S2012: According to each season, screen the initial historical meteorological data and initial historical operation data with a correlation degree greater than a preset value with the historical output data to obtain data samples corresponding to each season.

[0062] Specifically, the solar calendar method is used to divide seasons, that is, March to May is spring, June to August is summer, September to November is autumn, and December to February is winter. The initial historical meteorological data, initial historical operation data, and historical output data are divided into four groups of data sets according to spring, summer, autumn, and winter.

[0063] For the data set of each season, calculate the correlation degree between the initial historical meteorological data, initial historical operation data, and historical output data, and screen out the initial historical meteorological data and initial historical operation data with a correlation degree greater than the preset value with the historical output data to obtain data samples corresponding to each season. The methods for calculating the correlation degree include clustering techniques such as Pearson correlation coefficient, random forest method, KMeans, and principal hierarchical analysis. This step can screen out the main factors affecting historical output data and improve the accuracy of subsequent output prediction.

[0064] In a specific implementation, the Pearson correlation coefficient is used to calculate the correlation coefficient between the initial historical meteorological data, initial historical operation data, and historical output data. The calculation formula is:

[0065]

[0066] where x i is the value of a certain input factor in the i-th time period, which is any element in the initial historical meteorological data and initial historical operation data, such as wind speed, wind direction, air pressure, precipitation, evaporation, irradiance, sunshine duration, temperature, runoff, reservoir water level, reservoir discharge flow, hydropower unit generation flow, etc.; y i is the historical output data in the i-th time period, is the average value of each input factor, is the average value of historical output data, and n is the total number of time periods in the data set.

[0067] It can be understood that when x i is the initial historical meteorological data in the i-th time period, yi is the wind power output data or photovoltaic power output data for the i-th time period; when x i is the initial historical operation data for the i-th time period, y i is the hydropower output data for the i-th time period.

[0068] Furthermore, the initial historical meteorological data and initial historical operation data with a correlation coefficient r xy greater than a preset value are used as data samples. Exemplarily, the preset value is 0.6.

[0069] Using the above method to consider the influence of seasons on the power of the water-wind-solar complementary system and constructing models by season can make full use of the advantages of different models, train corresponding power prediction models for the water-wind-solar complementary system in each season, and improve the prediction accuracy of the models. At the same time, sample data with a greater correlation with historical output data is found for each of the four seasons respectively. This method can make the data samples more refined and improve the accuracy of subsequent model training and prediction.

[0070] In a further solution, the present application conducts a complementary analysis on the water-wind-solar complementary system.

[0071] In one implementation, the complementary correlation coefficients among the hydropower output data, wind power output data, and photovoltaic power output data are calculated.

[0072] Specifically, the Pearson correlation coefficient calculation method is used to calculate the Pearson correlation coefficient between the hydropower output data and the wind power output data, the Pearson correlation coefficient between the hydropower output data and the photovoltaic power output data, and the Pearson correlation coefficient between the wind power output data and the photovoltaic power output data respectively.

[0073] Then, the complementary correlation coefficient is calculated based on the three Pearson correlation coefficients as:

[0074] c = ω 1 cc wp + ω 2 cc wh + ω 3 cc hp .

[0075] Among them, c is the complementary correlation coefficient among the hydropower output data, wind power output data, and photovoltaic power output data, and cc wp , cc wh , cc hp are the Pearson correlation coefficient between the wind power output data and the photovoltaic power output data, the Pearson correlation coefficient between the hydropower output data and the wind power output data, and the Pearson correlation coefficient between the hydropower output data and the photovoltaic power output data respectively.

[0076] Among them, the value range of the complementary correlation coefficient is [0, 1]. The closer the value is to -1, the stronger the complementarity between the hydropower output data, wind power output data, and photovoltaic output data.

[0077] In one implementation, calculate the complementarity rate between the hydropower output data, wind power output data, and photovoltaic output data.

[0078] Respectively use the following formulas to calculate the standard deviations of the hydropower output data, wind power output data, and photovoltaic output data:

[0079]

[0080] Specifically, the complementarity rate is:

[0081]

[0082] Among them, is the complementarity rate between the hydropower output data, wind power output data, and photovoltaic output data, and σ hydro , σ wind and σ pv are the standard deviations of the hydropower output data, wind power output data, and photovoltaic output data respectively, and σ all is the standard deviation of the sum of the hydropower output data, wind power output data, and photovoltaic output data.

[0083] Among them, the value range of the complementarity rate is [0, 1]. The smaller the value, the better the complementarity between the hydropower output data, wind power output data, and photovoltaic output data.

[0084] In one implementation, calculate the complementary coefficient between the hydropower output data, wind power output data, and photovoltaic output data.

[0085] Specifically, the complementary coefficient is:

[0086]

[0087] Among them, χ is the complementary coefficient between the hydropower output data, wind power output data, and photovoltaic output data, are the per-unit values of the power change amounts of the hydropower output data, wind power output data, and photovoltaic output data in the i-th time period respectively.

[0088] Step S202, construct an initial power prediction sub-model on each hydropower-wind-solar complementary subsystem, and based on the basin runoff direction, connect all the initial power prediction sub-models on the hydropower-wind-solar complementary system in series to construct an initial power prediction model.

[0089] Specifically, please refer to Figure 3 , Figure 3 is a schematic diagram of the hydropower-wind-solar complementary system according to the embodiment of the present invention. As Figure 3As shown in the figure, multiple water-wind-solar complementary subsystems are set in the basin of the entire water-wind-solar complementary system. According to the basin runoff direction of the water-wind-solar complementary system, the water-wind-solar complementary subsystems are named as water-wind-solar complementary subsystem 1, water-wind-solar complementary subsystem 2, ……, water-wind-solar complementary subsystem RN. Each water-wind-solar complementary subsystem includes a hydropower station, a wind farm, and a photovoltaic power station. There is an electrical connection between the hydropower station and the wind farm, and between the hydropower station and the photovoltaic power station. An initial power prediction sub-model is constructed in each water-wind-solar complementary subsystem, and all the initial power prediction sub-models are connected in series according to the basin runoff direction to obtain an initial power prediction model.

[0090] Among them, each initial power prediction sub-model includes three models, which respectively correspond to predicting the output of the hydropower station, the wind farm, and the photovoltaic power station in the water-wind-solar complementary subsystem.

[0091] Step S203: Use the historical meteorological data and historical operation data as training samples, and the historical output data as labels, and train the initial power prediction model with the data samples to obtain a power prediction model.

[0092] The data set of the data samples is divided into a training set, a validation set, and a test set. Exemplarily, 60% of the data set is used as the training set, 20% of the data set is used as the validation set, and 20% of the data set is used as the test set. Further, the training samples and sample labels are divided from the training set. The historical meteorological data and historical operation data are used as training samples, and the historical output data is used as labels to train the initial power prediction model.

[0093] Among them, the initial power prediction model includes one of the models such as LSTM, CNN, GRU, decision regression tree, RNN, DeepTCN, PTST, etc.

[0094] In one implementation, the initial power prediction model is trained for each of the four seasons of spring, summer, autumn, and winter.

[0095] Specifically, the above step S203 includes:

[0096] Step S2031: For one season, train multiple initial power prediction models based on the data samples to obtain multiple power prediction models.

[0097] Specifically, the data samples are divided into data sets corresponding to each season, and further the training set, the validation set, and the test set are divided for each season. Train the initial power prediction model from the training set corresponding to each season to obtain the power prediction model for each season.

[0098] Among them, each hydropower station, the wind farms on both sides of it, and the photovoltaic power station fields form a complementary water-wind-solar power subsystem, and the entire complementary water-wind-solar power system formed by the series connection of the basin runoff directions from top to bottom is a complementary water-wind-solar power system. Therefore, the operation of the hydropower station in the upstream complementary water-wind-solar power subsystem of the basin will affect the operation of the hydropower stations in the downstream complementary water-wind-solar power subsystems.

[0099] When the complementary water-wind-solar power subsystem is the first system in the basin runoff direction, that is, when the hydropower station in the complementary water-wind-solar power subsystem is the leading reservoir, there is no upstream complementary water-wind-solar power subsystem that affects it. Directly use the historical meteorological data and historical operation data as the training samples of the initial power prediction sub-model of this complementary water-wind-solar power subsystem, and use the historical output data as the sample labels of the initial power prediction sub-model. Input the training samples into the initial power prediction model to obtain the predicted output data, and calculate the determination coefficient and root mean square error of this complementary water-wind-solar power subsystem based on the predicted output data and sample labels. Specifically, calculate the determination coefficient and root mean square error for the hydropower station, wind farm, and photovoltaic power station field respectively.

[0100] When the complementary water-wind-solar power system is not the first system in the basin runoff direction, that is, when the hydropower station in the complementary water-wind-solar power subsystem is not the leading reservoir, there is an upstream complementary water-wind-solar power subsystem that affects it. Obtain the predicted outflow of the previous complementary water-wind-solar power subsystem based on the static parameters of the hydropower station-level wind power station and the predicted output data of the previous complementary water-wind-solar power subsystem. Specifically, calculate the predicted outflow of the previous complementary water-wind-solar power subsystem by combining the actual values of the upstream and downstream water levels and the NHQ characteristic curve.

[0101] Use the historical meteorological data, historical operation data, and the predicted outflow of the previous complementary water-wind-solar power subsystem as the training samples of the initial power prediction sub-model, and use the historical output data as the sample labels of the initial power prediction sub-model. Input the training samples into the initial power prediction model to obtain the predicted output data, and calculate the determination coefficient and root mean square error of this complementary water-wind-solar power subsystem based on the predicted output data and sample labels. Specifically, calculate the determination coefficient and root mean square error for the hydropower station, wind farm, and photovoltaic power station field respectively.

[0102] Repeat the above method to calculate the determination coefficient and root mean square error of each complementary water-wind-solar power subsystem respectively.

[0103] Among them, the determination coefficient is:

[0104]

[0105] The root mean square error is:

[0106]

[0107] Among them, is the coefficient of determination when training the initial power prediction sub-model m for season s. The larger its value, the higher the prediction accuracy. is the root mean square error when training the initial power prediction sub-model m for season s. The smaller its value, the higher the prediction accuracy. NT is the number of samples or the total number of time periods in the training set for season s, and y i,s is the historical output data of the corresponding training set for the i-th time period in season s, is the predicted output data of the corresponding training set for the i-th time period in season s using the initial power prediction sub-model m, is the average value of the actual power for all time periods in the corresponding training set for season s.

[0108] Furthermore, calculate the load tracking risk coefficient and the standard deviation of the remaining load for the entire water-wind-solar-hydro system.

[0109] Among them, the load tracking risk coefficient is:

[0110]

[0111] Among them, is the load tracking risk coefficient of the entire water-wind-solar-hydro system when training the initial power prediction sub-model m for season s, are respectively the predicted hydropower output data, predicted wind power output data, and predicted photovoltaic output data of the k-th water-wind-solar complementary sub-system for the i-th time period in season s using the initial power prediction sub-model m, is the load accommodation demand of the k-th water-wind-solar complementary sub-system for the i-th time period in season s, and RN is the total number of water-wind-solar complementary sub-systems in the basin. Among them, the load accommodation demand is calculated based on the basic parameters of the water, wind, and photovoltaic power stations.

[0112] Among them, the smaller the value of the load tracking risk coefficient, the stronger the load tracking ability.

[0113] The standard deviation of the remaining load is:

[0114]

[0115] Among them,

[0116] Among them,

[0117] Among them, is the standard deviation of the remaining load of the entire water-wind-solar-hydro system when training the initial power prediction sub-model m for season s, is the remaining load value calculated after prediction for the kth hydropower-wind-solar complementary subsystem in the ith time period of season s using the initial power prediction sub-model m. is the average value of the remaining loads for all time periods of the kth hydropower-wind-solar complementary subsystem in season s using the initial power prediction sub-model m.

[0118] Furthermore, a total loss function is constructed based on the determination coefficient, root mean square error, load tracking risk coefficient, and standard deviation of the remaining load.

[0119] Specifically, the first loss function constructed based on the determination coefficient is:

[0120]

[0121] where is the first loss function, is the hydropower determination coefficient of the kth power prediction model, is the wind power determination coefficient of the kth power prediction model, is the photovoltaic determination coefficient of the kth power prediction model;

[0122] The second loss function constructed based on the root mean square error is:

[0123]

[0124] where is the second loss function, is the root mean square error of hydropower of the kth power prediction model, is the root mean square error of wind power of the kth power prediction model, is the root mean square error of photovoltaic of the kth power prediction model;

[0125] The third loss function constructed based on the load tracking risk coefficient is:

[0126] l RL = RL.

[0127] where l RL is the third loss function, and RL is the load tracking risk coefficient.

[0128] The fourth loss function constructed based on the standard deviation of the remaining load is:

[0129] l SL = αSL.

[0130] where l SL is the fourth loss function, SL is the standard deviation of the remaining load, and α is a normalization coefficient such that the value range of l SL is [0, 1].

[0131] The total loss function is constructed as follows:

[0132]

[0133] where Loss is the total loss function, and λ 1 , λ 2 , λ 3 , λ 4 are weight coefficients, which satisfy 0.6 ≤ λ 1 + λ 2 ≤ 0.8, and λ 1 + λ 2 + λ 3 + λ 4 = 1.

[0134] Based on the total loss function, train multiple times, and calculate and select the power prediction model with the minimum loss function for subsequent verification and testing. Specifically, use the validation set and the test set to replace the training set to verify and test the power prediction model. Among them, in the model verification and result testing stage, use the above-mentioned coefficient of determination, root mean square error, load tracking risk coefficient, residual load standard deviation, and loss function as evaluation indicators to evaluate the power prediction model.

[0135] Based on the above method, multiple power prediction models are trained. During the actual application process, use the following method to select one of them for use.

[0136] Step S2032: Construct a decision matrix based on multiple power prediction models.

[0137] First, collect the forecast information. Specifically, collect the forecast data of meteorology and hydrology at a 15-minute scale for the next day, including but not limited to wind speed, wind direction, air pressure, precipitation, evaporation, irradiance, sunshine duration, temperature, and reservoir inflow; as well as the scheduling control plan for the reservoir water level for the next day and the load forecast curve.

[0138] Input the collected forecast information into multiple power prediction models to obtain the predicted hydropower output data, predicted wind power output data, and predicted photovoltaic output data of each power prediction model.

[0139] Furthermore, based on the predicted hydropower output data, predicted wind power output data, and predicted photovoltaic output data of each power prediction model, calculate the coefficient of determination, root mean square error, load tracking risk coefficient, and residual load standard deviation of each power prediction model. Among them, the specific calculation method is the same as the above steps and will not be elaborated here.

[0140] Specifically, construct a power prediction accuracy evaluation index matrix based on the coefficient of determination and root mean square error of each power prediction model:

[0141]

[0142] Construct a power supply capacity index matrix based on the load tracking risk coefficient and the standard deviation of the remaining load of each power prediction model:

[0143]

[0144] Construct a decision matrix based on the power prediction accuracy evaluation index matrix and the power supply capacity index matrix:

[0145] EC MN×(6×RN+2) ={EC 1 , …, EC k , …, EC RN , EC sys}.

[0146] Step S2033: Use the fuzzy decision-making method to jointly evaluate the decision matrix, calculate the relative membership degree of each power prediction model, and determine the power prediction model for each season based on the principle of maximum membership degree.

[0147] According to the above evaluation indicators, by using the fuzzy optimal selection decision-making method, the best power prediction model for power prediction under the current season, current prediction day, and current forecast conditions can be determined. The evaluation results can provide data support for the preparation of short-term scheduling plans and other needs.

[0148] Specifically, first calculate the information entropy as:

[0149]

[0150] Calculate the index weight according to the information entropy as:

[0151]

[0152] where H j is the information entropy value corresponding to the j-th column index in the decision matrix, and p m,j is the probability of predicting the j-th index using the m-th power prediction model, which satisfies where ec m,j is the index value of the m-th row and j-th column of the decision matrix.

[0153] If, then define lim(p m,j ×lnp m,j ) = 0.

[0154]

[0155] where u m is the relative membership degree when predicting using the m-th power prediction model.

[0156] Set the principle of maximum membership degree as: BM = arg max1≤m≤Mn u m 。

[0157] Among them, BM is the best power prediction model that should be used for power prediction under the current season, current prediction day, and current forecast conditions.

[0158] This application comprehensively considers the certainty coefficient affecting the prediction accuracy, the root mean square error affecting the prediction accuracy, the load tracking risk coefficient affecting the output power tracking level, and the standard deviation of the remaining load affecting the power grid supply and demand level to construct four sub-loss functions, and combines them to obtain the total loss function for training the power prediction model. It can comprehensively consider various factors affecting the output power, realize the power prediction of multiple water, wind, and light energy sources in the basin range and the coordinated linkage between upstream and downstream, and improve the accuracy of power prediction.

[0159] The method for constructing a power prediction model of a basin water, wind, and light complementary system provided in this embodiment constructs a power prediction model within the entire basin range, separately constructs a power prediction sub-model for each water, wind, and light complementary subsystem, and conducts large-scale joint prediction of the basin. This application, from the perspective of the entire basin and considering the impact of upstream and downstream reservoir regulation, sets different training samples for the leading subsystem and other subsystems, considers the impact of the flow of the upstream water, wind, and light complementary subsystem on the downstream water, wind, and light complementary subsystem, adds the predicted discharge flow of the upstream to the training samples of the downstream water, wind, and light complementary subsystem, and comprehensively considers the certainty coefficient affecting the prediction accuracy, the root mean square error affecting the prediction accuracy, the load tracking risk coefficient affecting the output power tracking level, and the standard deviation of the remaining load affecting the power grid supply and demand level to perform model training and model selection. It can realize the power prediction of multiple water, wind, and light energy sources in the basin range and the coordinated linkage between upstream and downstream, improve the accuracy of power prediction, enhance the overall utilization efficiency of various energy sources such as basin hydropower and new wind and light energy, and ensure the safe and stable operation of the power grid.

[0160] In this embodiment, a device for constructing a power prediction model of a basin water, wind, and light complementary system is also provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0161] This embodiment provides a device for constructing a power prediction model of a basin water, wind, and light complementary system, Figure 4 which is a structural block diagram of the device for constructing a power prediction model of a basin water, wind, and light complementary system according to an embodiment of the present invention. As Figure 4 shown, it includes:

[0162] An acquisition module 401, configured to acquire data samples, where the data samples include historical meteorological data, historical operation data, and historical output data of a basin water-wind-solar complementary system, and the output data includes hydropower output data, wind power output data, and photovoltaic output data.

[0163] A construction module 402, configured to construct an initial power prediction sub-model on each water-wind-solar subsystem, and cascade all the initial power prediction sub-models on the water-wind-solar system based on the basin runoff direction to construct an initial power prediction model.

[0164] A training module 403, configured to use the historical meteorological data and historical operation data as training samples, use the historical output data as labels, and train the initial power prediction model with the data samples to obtain a power prediction model.

[0165] In some optional embodiments, the acquisition module 401 includes:

[0166] An acquisition unit, configured to acquire initial data for each season, where the initial data includes initial historical meteorological data, initial historical operation data, and historical output data.

[0167] A screening unit, configured to screen the initial historical meteorological data and initial historical operation data with a correlation degree greater than a preset value with the historical output data according to each season to obtain data samples corresponding to each season.

[0168] In some optional embodiments, the training module 403 includes:

[0169] A seasonal training sub-module, configured to train the initial power prediction model based on the data samples corresponding to each season respectively to obtain a power prediction model for each season.

[0170] In some optional embodiments, the seasonal training sub-module includes:

[0171] An initial training unit, configured to train, for one season, multiple initial power prediction models based on the data samples to obtain multiple power prediction models.

[0172] A construction unit, configured to construct a decision matrix based on the multiple power prediction models.

[0173] A determination unit, configured to jointly evaluate the decision matrix by using a fuzzy decision method, calculate the relative membership degree of each power prediction model, and determine the optimal power prediction model from the multiple power prediction models based on the principle of maximum membership degree.

[0174] In some optional embodiments, the construction unit includes:

[0175] The first construction subunit is used to construct a power prediction accuracy evaluation index matrix based on the determination coefficient and root mean square error of each power prediction model.

[0176] The second construction subunit is used to construct a power supply capacity index matrix based on the load tracking risk coefficient and residual load standard deviation of each power prediction model.

[0177] The third construction subunit is used to construct a decision matrix based on the power prediction accuracy evaluation index matrix and the power supply capacity index matrix.

[0178] In some alternative embodiments, the training module 403 includes:

[0179] The data partitioning unit is used to, when the water-wind-solar subsystem is the first system in the basin runoff direction, use the historical meteorological data and historical operation data as the training samples of the initial power prediction sub-model, and use the historical output data as the sample labels of the initial power prediction sub-model; when the water-wind-solar subsystem is not the first system in the basin runoff direction, use the historical meteorological data, historical operation data, and the predicted out-flow of the previous water-wind-solar subsystem as the training samples of the initial power prediction sub-model, and use the historical output data as the sample labels of the initial power prediction sub-model.

[0180] The training unit is used to train the initial power prediction model based on the training samples and sample labels to obtain the power prediction model.

[0181] In some alternative embodiments, the device for constructing the power prediction model of the basin water-wind-solar complementary system further includes:

[0182] The second acquisition module is used to acquire the basic data of the hydropower station in the water-wind-solar system. The basic data of the hydropower station includes the NHQ characteristic curve, and calculate the predicted out-flow of the previous water-wind-solar system based on the predicted hydropower output data of the previous water-wind-solar system and the basic data of the hydropower station.

[0183] In some alternative embodiments, the second acquisition module includes:

[0184] The prediction unit is used to input the training samples into the initial power prediction model to obtain the predicted output data.

[0185] The calculation unit is used to calculate the determination coefficient, root mean square error, load tracking risk coefficient, and residual load standard deviation based on the predicted output data and sample labels.

[0186] The construction unit is used to construct the total loss function based on the determination coefficient, root mean square error, load tracking risk coefficient, and residual load standard deviation.

[0187] The training unit is used to train the power prediction model multiple times based on the total loss function.

[0188] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be repeated here.

[0189] The device for constructing the power prediction model of the basin water-wind-solar complementary system in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0190] The embodiment of the present invention also provides a computer device having the above Figure 4 shown device for constructing the power prediction model of the basin water-wind-solar complementary system.

[0191] Please refer to Figure 5 , Figure 5 which is a schematic structural diagram of a computer device provided by an optional embodiment of the present invention. As Figure 5 shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display 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. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (such as an array of servers, a set of blade servers, or a multi-processor system). Figure 5 In

[0192] FIG. 1, one processor 10 is taken as an example.

[0193] Among them, the memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0194] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories may be connected to the computer device through 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 combinations thereof.

[0195] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, a hard disk, or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0196] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 may be connected through a bus or other means. Figure 5 Taking connection through a bus as an example.

[0197] The input device 30 can receive input digital or character information and generate key signal inputs related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (such as an LED), and a tactile feedback device (such as a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device may be a touch screen.

[0198] Embodiments of the present invention also provide a computer-readable storage medium. The methods according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code that is originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the methods described herein can be stored as such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated 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 drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods shown in the above embodiments are implemented.

[0199] A part of the present invention can be applied as a computer program product, for example, computer program instructions, which, when executed by a computer, can call or provide the methods and / or technical solutions according to the present invention through the operation of the computer. Those skilled in the art should be able to understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are 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. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible by the computer.

[0200] Although the embodiments of the present 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 present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for constructing a power prediction model for a watershed water, wind and solar system, characterized in that: The method comprises: Acquire data samples, wherein the data samples include historical meteorological data, historical operation data, and historical output data of the water, wind, and solar systems in the basin, and the output data include hydropower output data, wind power output data, and photovoltaic output data; Constructing an initial power prediction sub-model on each water-wind-solar sub-system, and based on the runoff direction of the watershed, connecting all the initial power prediction sub-models on the water-wind-solar system in series to construct an initial power prediction model; The historical meteorological data and the historical operating data are used as training samples, the historical output data are used as labels, and the data samples are used to train the initial power prediction model to obtain a power prediction model.

2. The method for constructing a watershed water-wind-solar system power prediction model according to claim 1, characterized in that: The obtaining of data samples comprises: Acquire initial data for each season, the initial data including initial historical meteorological data, initial historical operation data and the historical output data; According to each season, the initial historical meteorological data and the initial historical operation data whose correlation degree with the historical output data is greater than a preset value are screened to obtain the data sample corresponding to each season; The using the data sample to train the initial power prediction model to obtain the power prediction model comprises: The initial power prediction model is trained respectively based on the data samples corresponding to each season to obtain a power prediction model for each season.

3. The method for constructing a watershed water-wind-solar system power prediction model according to claim 2, characterized in that: The initial power prediction model is trained based on the data samples corresponding to each season to obtain the power prediction model for each season, including: For one season, training a plurality of the initial power prediction models based on the data samples to obtain a plurality of the power prediction models; Constructing a decision matrix based on a plurality of said power prediction models; The decision matrix is ​​jointly evaluated by using a fuzzy decision method, the relative superiority of each power prediction model is calculated, and the optimal power prediction model is determined from the power prediction models based on the maximum superiority principle.

4. The method for constructing a watershed water, wind and solar power system power prediction model according to claim 3 is characterized in that: The constructing a decision matrix based on the multiple power prediction models comprises: Constructing a power prediction accuracy evaluation index matrix based on the certainty coefficient and root mean square error of each power prediction model; Constructing a power supply capability index matrix based on the load tracking risk coefficient and the residual load standard deviation of each power prediction model; The decision matrix is ​​constructed based on the power prediction accuracy evaluation index matrix and the power supply capability index matrix.

5. The method for constructing a watershed water, wind and solar power system power prediction model according to claim 3 is characterized in that: The method of using the historical meteorological data and the historical operation data as training samples, using the historical output data as labels, and using the data samples to train the initial power prediction model to obtain the power prediction model includes: When the water-wind-photovoltaic subsystem is the first system in the runoff direction of the watershed, the historical meteorological data and the historical operating data are used as training samples of the initial power prediction submodel, and the historical output data are used as sample labels of the initial power prediction submodel; when the water-wind-photovoltaic subsystem is the first system in the runoff direction of a non-watershed, the historical meteorological data, the historical operating data and the predicted outflow flow of the previous water-wind-photovoltaic subsystem are used as training samples of the initial power prediction submodel, and the historical output data are used as sample labels of the initial power prediction submodel; The initial power prediction model is trained based on the training samples and the sample labels to obtain the power prediction model.

6. The method for constructing a watershed water, wind and solar power system power prediction model according to claim 5, characterized in that: When the water-wind-solar system is the first system in the non-basin runoff direction, the historical meteorological data, the historical operation data and the predicted outflow flow of the previous water-wind-solar system are used as training samples of the initial power prediction sub-model, and the following further includes: Acquire basic data of a hydropower station in a water-wind-solar system, wherein the basic data of the hydropower station includes an NHQ characteristic curve; The predicted outflow flow of the previous water, wind and solar system is calculated based on the predicted hydropower output data of the previous water, wind and solar system and the basic data of the hydropower station.

7. The method for constructing a watershed water, wind and solar power system power prediction model according to claim 6, characterized in that: The training of the initial power prediction model based on the training samples and the sample labels to obtain the power prediction model comprises: Inputting the training samples into the initial power prediction model to obtain predicted output data; Calculating a certainty coefficient, a root mean square error, a load tracking risk coefficient, and a residual load standard deviation based on the predicted output data and the sample labels; constructing a total loss function based on the certainty coefficient, the root mean square error, the load tracking risk coefficient and the residual load standard deviation; The power prediction model is obtained by multiple training based on the total loss function.

8. The method for constructing a watershed water, wind and solar power system power prediction model according to claim 7, characterized in that: The constructing of a total loss function based on the certainty coefficient, the root mean square error, the load tracking risk coefficient and the residual load standard deviation comprises: The first loss function is constructed based on the certainty coefficient: Among them, the is the first loss function, is the hydropower deterministic coefficient of the kth power prediction model, is the wind power certainty coefficient of the kth power prediction model, is the photoelectric deterministic coefficient of the kth power prediction model; The second loss function is constructed based on the root mean square error: Among them, the is the second loss function, is the hydropower root mean square error of the kth power prediction model, is the wind power root mean square error of the kth power prediction model, is the photoelectric root mean square error of the kth power prediction model; The third loss function constructed based on the load tracking risk coefficient is: l RL =RL; Among them, the l RL is a third loss function, and RL is the load following risk coefficient; The fourth loss function constructed based on the residual load standard deviation is: l SL =αSL; Among them, the l SL is a fourth loss function, and SL is the residual load standard deviation; The total loss function is constructed as: Among them, the Loss is the total loss function.

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 constructing a watershed water-wind-solar system power prediction model according to any one of claims 1 to 8 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for constructing a watershed water-wind-solar system power prediction model according to any one of claims 1 to 8.

Images