A method for generating a cascade hydropower dispatch scenario set considering water inflow-price-power generation space joint uncertainty

Abstract

The application discloses a kind of considering water-electricity price-generating space joint uncertainty cascade hydropower scheduling scene set generation method, belong to power generation scheduling and electric power market technical field.The application first to water flow, node electricity price and generating space constructs single-variable uncertainty quantification model, utilizes three-dimensional Copula function to construct variable joint probability distribution and depict interdependence, generates initial multidimensional uncertainty scene set according to distribution using Monte Carlo method.Then according to scene probability, classification recommendation is carried out, forms high / low probability scene set and inputs cascade hydropower scheduling model, generates scheduling scheme for different scene.The application breaks through the limitation of single or two-dimensional uncertainty analysis, accurately depicts the high-dimensional interdependence structure of water, electricity price and generating space, realizes the efficient coupling of probability scene and scheduling decision, significantly improves the scheduling robustness and engineering practicability of cascade hydropower under extreme working conditions.

Description

Technical Field

[0001] This invention relates to a method for generating a cascade hydropower dispatch scenario set that considers the joint uncertainty of water inflow, electricity price, and power generation space, and belongs to the field of power generation dispatch and electricity market technology. Background Technology

[0002] With the deepening of market-oriented reforms in the power system, the dispatching model of cascade hydropower is being reshaped, shifting from the traditional planned power generation model of "power generation determined by water availability" and "water generation determined by electricity availability" to a model based on market clearing and electricity price determination under the electricity spot market, aiming to maximize power generation revenue. In the electricity spot market environment, cascade hydropower must consider multiple uncertainties, including electricity prices, congestion, and market competition, in order to achieve stable dispatching.

[0003] However, existing literature has three shortcomings in dealing with the aforementioned uncertainties:

[0004] (1) Limited sources of uncertainty: Most studies at present only consider the uncertainty of electricity price and water inflow, ignoring the key source of uncertainty of hydropower station’s “power generation space” affected by channel, maintenance, cross-sectional constraints, etc.;

[0005] (2) Ignoring the dependency relationship of variables in scenario generation: Traditional scheduling scenario generation usually assumes that variables such as water flow and node electricity price are independent of each other and samples them to generate scenarios separately. However, in actual operation, there is a certain dependency relationship between dry season and high electricity price, and between limited power generation space and rising electricity price. Assuming that each variable is independent will lead to the distortion of the scenario set;

[0006] (3) Mismatch between scenario set and scheduling decision: Most existing technologies calculate the probability of scenario set based on expected value, without classifying scenarios and matching them with different scheduling objectives, resulting in limited engineering application value of scenario set.

[0007] There is an urgent need for a method to systematically quantify the inherent interdependence of uncertainties among water flow, nodal electricity price, and power generation space, to generate a set of scenarios that directly serve different scheduling optimization objectives, and to improve the revenue of cascade hydropower scheduling and operation under the electricity spot market. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for generating a set of cascade hydropower scheduling scenarios that considers the joint uncertainty of inflow water, electricity price, and generation space. This method generates a set of scenarios that directly serve different scheduling optimization objectives, thereby improving the revenue of cascade hydropower scheduling and operation in the electricity spot market.

[0009] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0010] The purpose of this invention is to overcome the aforementioned shortcomings in the prior art and to provide a method for generating a cascade hydropower dispatch scenario set that considers the joint uncertainty of inflow water, electricity price, and power generation space, in order to solve the following technical problems:

[0011] (1) Overcoming the limitations of single or two-dimensional uncertainty: For the first time, “power generation space” is included as an independent uncertainty variable in the uncertainty analysis framework of cascade hydropower dispatch.

[0012] (2) Characterize the complex interdependencies between variables: The three-dimensional Copula function is used to capture the interdependency structure between water flow, nodal electricity price and power generation space, so as to realize the modeling of the joint uncertainty of water flow, nodal electricity price and power generation space.

[0013] (3) Coupling of scenario set and scheduling decision: Through the scenario classification mechanism, the scenario set is divided into "high probability typical scenario" and "low probability extreme scenario", and the constraint logic of "the upper limit of the final output of cascade hydropower is the smaller value of the potential output determined by the inflow and the power generation space" is explicitly introduced into the scheduling model, so that the scenario set can directly drive different scheduling objectives (baseline plan scheduling and robust plan scheduling) and improve the engineering practicality of the scheduling scheme.

[0014] The technical solution adopted by the present invention to solve the above problems is as follows.

[0015] In a first aspect, the present invention provides a method for generating a set of cascade hydropower dispatch scenarios that considers the joint uncertainty of inflow water, electricity price, and power generation space, comprising the following steps:

[0016] Step 1: Construct univariate uncertainty quantification models for water inflow, nodal electricity price, and power generation space respectively, and output the marginal probability distributions of water inflow, nodal electricity price, and power generation space;

[0017] Step 2: Based on the marginal probability distributions of the inflow, nodal electricity price, and power generation space, a three-dimensional Copula function is used to construct the joint probability distribution of the inflow, nodal electricity price, and power generation space to characterize the interdependence among the three.

[0018] Step 3: Based on the joint probability distribution, generate an initial multidimensional uncertainty scenario set using the Monte Carlo method;

[0019] Step 4: Reduce the initial multidimensional uncertainty scene set into scenes, and classify and recommend the reduced scenes according to scene probability and features to obtain high-probability typical scenes and low-probability extreme scenes.

[0020] The power generation space refers to the upper limit of power that a hydropower station can safely generate under the physical boundary conditions of given grid topology, channel capacity, unit maintenance status and vibration zone constraints.

[0021] Furthermore, in step 1, a univariate uncertainty quantification model is constructed for the inflow water flow, including:

[0022] Deterministic forecasts of inflow for cascade hydropower projects based on distributed hydrological models and the actual inflow rate monitored by the station A "forecast-measured" mapping model was constructed using quantile regression technology:

[0023] ;

[0024] in, The simulated flow rate generated by the distributed hydrological model at time t; It is the flow rate at the quantile τ to be predicted; The quantiles covering the uncertainty interval. . The intercept term represents the systematic bias of the physical model; The slope term represents the factor by which the physical model amplifies or reduces runoff fluctuations.

[0025] To obtain robust runoff estimates under extreme drought and flood conditions, the model fits the model parameters by minimizing the checked ball loss function. and The formula is as follows:

[0026]

[0027]

[0028] in, This is the error value, i.e. ; This is an indicator function; its value is 1 if the error is less than 0, and 0 otherwise. =0.001, 0.005, 0.01, 0.02, ..., 0.5, ..., 0.98, 0.99, 0.995, 0.999, for and Each was fitted separately to obtain a set The scatter plot is used to fit the extreme flood portion using the generalized Pareto distribution (GPD), and the remaining portion is fitted using cubic splines, thus obtaining the marginal distribution of the inflow.

[0029] Furthermore, in step 1, a univariate uncertainty quantification model is constructed for nodal electricity prices, including:

[0030] Mapping model of nodal electricity prices and supply and demand data:

[0031]

[0032]

[0033] in, for The total load forecast value of the sending or receiving system at any given time; for The predicted output values ​​of other power sources within the system at any given time; for The net load at any given time represents the remaining demand that must be met by this type of power supply; for Historical marginal electricity prices of the transmission channel nodes at all times; Net load and electricity price Functional mapping relationship between them; , , , , The model parameters to be determined are obtained by fitting using the least squares method; The term is an exponential term, simulating the surge in electricity prices when the net load approaches the system's installed capacity limit; The residual term represents random fluctuations that the model cannot explain, and it is usually assumed that it follows a normal distribution.

[0034] According to the above formula, the main source of electricity price uncertainty is load forecasting error. Other energy forecast errors :

[0035]

[0036]

[0037] Assuming the errors of both are independent and identically distributed, then the joint standard deviation of the net load... for:

[0038]

[0039] Within the standard deviation of net load, massive electricity price results were generated based on Monte Carlo simulation:

[0040]

[0041] in, It is the first The scenario of Monte Carlo sampling; It is the first Sample values ​​of net load scenario obtained from Monte Carlo sampling; It is a deterministic electricity price forecast provided by an external model; It is the first The random disturbance term from the second sampling. Then, the net load and electricity price are mapped, resulting in multiple sets of generated data. Substitute the above-fitted net load-electricity price curve ~ In addition, residual random terms are superimposed to obtain multiple possible electricity price scenarios. .

[0042] The above electricity pricing scenarios are transformed into probabilistic electricity prices, where the expected electricity price is... Represented as:

[0043]

[0044] according to Sort by size from smallest to largest, and take and ( Using the confidence interval of electricity price as a reference, the resulting sample set represents the marginal distribution of electricity price.

[0045] Furthermore, in step 1, a univariate uncertainty quantification model is constructed for the power generation space, including:

[0046] Basic characteristic data preparation: baseline scale of power generation space Based on the predicted total power output of the cascade hydropower stations in the province and the predicted output share of cascade hydropower stations Decide:

[0047] ;

[0048] ;

[0049] in, This represents the daily power generation of the target hydropower station during a historical period. Because the combined power generation capacity of a cascade hydropower station group is far greater than that of small hydropower stations, therefore... It generally falls within a stable range. .because It may be affected by unconventional factors such as policies, channel restrictions, and maintenance events, resulting in a very small probability of significant fluctuations. Therefore, to remove the influence of extreme values, a range of 5% to 95% is selected. Calculate the average value .

[0050] The physical and topological constraint feature data are hard constraints, including maintenance information (start and end times of unit maintenance shutdown, and transmission and transformation maintenance information for the same day). DC channel output power and cross-sectional constraints Research on the output constraints of the target power plant units :

[0051]

[0052] Market and dispatch characteristic data include inter-provincial power transmission agreements, historical and current-day electricity prices and transaction results for each hydropower station node, and historical daily power transmission volume of each hydropower station's DC transmission channels.

[0053] Gaussian process regression modeling: for each 15-minute time node Construct feature vectors:

[0054]

[0055] in, This is the initial power generation share allocated based on the total projected power output; The cross-sectional margin of the passage corresponding to the target hydropower station; Day-ahead / real-time node price; Under maintenance; This is a typical output shape factor for similar days in history.

[0056] Next, a kernel function needs to be selected. Since the power generation space exhibits significant intra-diurnal periodicity and smooth evolution characteristics, a hybrid kernel function is chosen. :

[0057]

[0058] in, It is a radial basis function kernel, which is a commonly used kernel function to capture nonlinear mappings between features; Using a periodic kernel, the cyclical characteristics of a 15-minute sequence within a day are captured; It is a white noise core that absorbs measurement errors and non-systematic fluctuations.

[0059] Finally, model training and parameter optimization are performed. Using the aforementioned historical measured data as the training set, the model is trained by maximizing the log-marginal likelihood function. To determine the hyperparameters of the kernel function, thereby estimating the power generation space:

[0060]

[0061] in, It is the number of samples in the training set; It is the output vector of the training set; that is, the spatial value of power generation measured in history. For kernel function The matrix formed represents the covariance structure among the training samples; For noise variance; The identity matrix; after the GPR model is trained, for the features at a certain time before the current date to be predicted... GPR will provide predictions and probability distributions of the power generation space, including mean predictions. and variance predict:

[0062]

[0063]

[0064]

[0065] in, This is the covariance variable between the newly input independent variable and the training set. The marginal probability distribution of the power generation space at this point is:

[0066]

[0067] Physical constraint verification: Since GPR outputs statistical results, but hydropower dispatch must adhere to hard physical constraints, constraint truncation and uncertainty correction are necessary. Predicted power generation space. The following formula must be satisfied:

[0068]

[0069] in, The interval is defined as uncertain. The marginal distribution of the power generation space follows a right-truncated normal distribution. If transmission line maintenance or unit maintenance occurs at a certain moment, the mean at that moment... The forced correction is set to 0, and the variance is... It should also approach 0. Based on the right-truncated normal distribution, the discrete part follows a 0-value degenerate distribution.

[0070] Furthermore, in step 2, based on the marginal probability distributions of the inflow water flow, nodal electricity price, and power generation space, a joint probability distribution of the inflow water flow, nodal electricity price, and power generation space is constructed using a three-dimensional Copula function to characterize the interdependence among the three, including:

[0071] Using the marginal probability distributions of water flow, nodal electricity price, and power generation space obtained in step 1, a joint probability distribution of water flow, nodal electricity price, and power generation space is constructed using a three-dimensional Gaussian Copula function;

[0072] Inflow rate Node electricity price Power generation space Construct a system state vector :

[0073]

[0074] Fit their marginal cumulative distribution functions respectively ; variables in actual physical space , , Transform it into a probability space variable on the interval [0,1] through its cumulative distribution function, so as to express:

[0075]

[0076] in, Let i represent the i-th probability space variable. These represent the probability space variables of inflow water, nodal electricity price, and power generation space, respectively.

[0077] Let the probability space vector be... Therefore, the true joint distribution of the three variables can be expressed as:

[0078]

[0079] in, The Pearson correlation coefficient matrix is ​​3x3; when using the three-dimensional Gaussian Copula function C, its mathematical expression is:

[0080]

[0081] in, It is the inverse function of the standard normal distribution. It is a vector with a mean of 0 and a correlation coefficient matrix of... The three-dimensional standard normal cumulative function; define the auxiliary vector. ,in:

[0082] ;

[0083] in, Let i represent the i-th auxiliary variable. These represent auxiliary variables: inflow rate, nodal electricity price, and power generation space, respectively.

[0084] Furthermore, in step 3, based on the joint probability distribution, an initial multidimensional uncertainty scenario set is generated using the Monte Carlo method, including:

[0085] N independent three-dimensional random vectors are generated using the Monte Carlo method. :

[0086] ;

[0087] Where the superscript (k) indicates the k-th element generated by the Monte Carlo method, and there are a total of N elements;

[0088] For the correlation coefficient matrix Perform Cholesky decomposition:

[0089] ;

[0090] ;

[0091] at this time, Having established the physical interdependence between water supply, power generation space, and electricity price, the physical space is restored through two inverse transformations:

[0092] ;

[0093] ;

[0094] in, This represents the inverse function of the marginal cumulative distribution function of the i-th variable; This represents the k-th initial scene;

[0095] At this point, N initial scene sets can be obtained. :

[0096] ;

[0097] Because it is Monte Carlo random sampling with equal probability, the initial probability of each scene occurring is... All are:

[0098] .

[0099] Furthermore, in step 4, the initial multidimensional uncertainty scene set is reduced in scope, and the reduced scenes are classified and recommended based on scene probability and features to obtain high-probability typical scenes and low-probability extreme scenes, including:

[0100] For the initial scene set Each variable is standardized using Z-score to obtain the k-th standardized scenario. ;

[0101] Define any two scenarios and Three-dimensional geometric distance between :

[0102]

[0103] in,

[0104] Set the target retention scene set as Initially an empty set, the set of scenes to be reduced is: Initially, there are N scenes, which are selected using the synchronous back-substitution elimination method. Points:

[0105] ① Find the absolute center point: Find the point whose sum of distances to all other points is the smallest, and place it there. From Remove from the middle;

[0106] ② Cyclic selection: For Each of the remaining candidate points Calculate if we take Add In the context of the entire system, the goal is to find the point that maximizes the decrease in the probability distance from the unselected points to the set of selected points. Put it in .

[0107] ③ Termination decision: Repeat the selection until a set is reached. The middle just contains A scenario.

[0108] The reduced scenario will increase its own probability. The final retained scene is the one closest to it. Final recommendation probability for:

[0109]

[0110] in, This represents the set of all eliminated scenarios that are closest to the representative scenario r.

[0111] Select the top M scenarios with the highest probability, ensuring that the cumulative probability of these scenarios exceeds 50%, to obtain a set of high-probability typical scenarios, which represent high-probability typical scenarios and serve the baseline plan scheduling.

[0112] From the remaining scenarios, select scenarios where the physical values ​​of power generation space, water flow, and nodal electricity price are individually or jointly at the boundary to obtain a set of low-probability extreme scenarios, which represent low-probability extreme scenarios and serve robust scheduling.

[0113] Furthermore, the method also includes:

[0114] Step 5: Embed the high-probability typical scenarios and / or the low-probability extreme scenarios into the cascade hydropower optimization scheduling model to generate scheduling schemes for different scheduling objectives.

[0115] Furthermore, in step 5, the high-probability typical scenarios and / or the low-probability extreme scenarios are embedded into the cascade hydropower optimal scheduling model to generate scheduling schemes for different scheduling objectives, including:

[0116] The generated water inflow, node electricity price, and power generation space scenario set are embedded into the cascade hydropower optimization scheduling model. The water inflow determines the total supply and process of hydropower resources; the electricity price guides the optimization model to allocate power output to maximize revenue; and the power generation space serves as an upper limit constraint on power output, reflecting the limitations imposed on power output by external physical boundaries such as power grid channels, maintenance, and cross-sections.

[0117] To maximize power generation efficiency, a day-ahead optimal scheduling model for cascade hydropower is constructed. The objective function is:

[0118] ;

[0119] in, For time period sequence number; This represents the total number of time periods; For hydropower station serial numbers; This represents the total number of hydropower stations in the entire cascade. For power station During the period The power generation capacity; For the scene Next period Predicted clearing price at the node; The time interval is the step size.

[0120] To maximize power generation efficiency, a day-ahead optimal scheduling model for cascade hydropower is constructed; the objective function is:

[0121] ;

[0122] in, The time period number; This represents the total number of time periods; For hydropower station serial numbers; This refers to the total number of hydropower stations in the entire cascade. For power station During the period The power generation capacity; For the scene Next period Predicted clearing price at the node; The time interval step;

[0123] Under the baseline scheduling model, the inflow rate, nodal electricity price, and generation capacity under high-probability typical scenarios are simultaneously input into the day-ahead optimization scheduling model of cascade hydropower to solve for the optimal output process. As the benchmark generation plan for the day-ahead market;

[0124] Under the robust scheduling model, the inflow rate, nodal electricity price, and generation capacity under low-probability extreme scenarios are substituted into the day-ahead optimization scheduling model of cascade hydropower. If the limits are exceeded, the strategy is adjusted, and the optimal output process is obtained by solving the problem. As a robust scheduling scheme;

[0125] The constraints include:

[0126] Water balance constraints:

[0127]

[0128] in and For power station During the period and time period Storage capacity, and Power plants During the period Inbound and outbound flow.

[0129] Power output-generating flow-head relationship:

[0130]

[0131] in This is the NHQ curve function. For power station During the period The water purifier head, For power station During the period The power generation flow.

[0132] Space constraints for power generation:

[0133]

[0134] in Contribute to the medium-term plan, This is a configurable parameter for the allowable positive deviation. This is a configurable parameter for allowing negative deviation. For the scene Next period The maximum power output of the power plant.

[0135] Short-term power generation constraints:

[0136]

[0137] in The daily power generation target is given for medium-term dispatch.

[0138] Reservoir water level constraints:

[0139]

[0140] in , and Power plants The water level, dead water level, and normal water level at time t.

[0141] Power generation flow constraints:

[0142]

[0143] in , and Power plants The power generation flow rate, minimum power generation flow rate, and maximum power generation flow rate at time t.

[0144] Outbound flow constraints:

[0145]

[0146] in , and Power plants The outbound flow rate, minimum outbound flow rate, and maximum outbound flow rate at time t.

[0147] Output climbing constraints:

[0148]

[0149] in For power station The maximum permissible range of output variation between adjacent time periods.

[0150] Vibration zone avoidance constraints:

[0151]

[0152] in The vibration zone number is... The total number of vibration zones, and These are the lower and upper limits of the vibration zone, respectively.

[0153] Water flow stagnation time:

[0154]

[0155] in For power station and power station If the inflow interval between m and m is the head power station, then It's the power station's natural inflow. For water flow from the upstream power station The dam site flows to the downstream power station Time required for dam site construction.

[0156] In a second aspect, the present invention provides a cascade hydropower dispatch scenario generation device that considers the joint uncertainty of water inflow, electricity price and power generation space, including a processor and a storage medium;

[0157] The storage medium is used to store instructions;

[0158] The processor is configured to operate according to the instructions to perform the steps of the method according to the first aspect.

[0159] Compared with the prior art, the present invention has the following advantages and effects:

[0160] (1) A three-dimensional uncertainty joint modeling framework of "water inflow-electricity price-generation space" is proposed for the first time: This solves the problem that existing technologies ignore the physical constraints of power grid channels and unit maintenance on hydropower output. By independently incorporating the generation space into the three-dimensional Copula joint modeling of "water inflow-electricity price-generation space", the generated scenario set can simultaneously reflect the triple uncertainties of resources, market and operation, making the scheduling plan closer to the actual executable capacity.

[0161] (2) By relaxing the minimum constraint of "medium-term plan - power generation space", the precise connection between scenario and scheduling nesting is achieved: After traditional scenario generation, it is either directly used for the expected value calculation of stochastic planning or only used for display purposes, without forming an organic connection with the long-medium-short-term nested process in scheduling business. This invention relaxes the minimum constraint of "medium-term plan - power generation space", so that the power generation space scenario value directly participates in the short-term output upper limit decision, which preserves the economic optimization space while ensuring physical feasibility. This significantly improves the feasibility of the plan.

[0162] (3) Establish a “typical scenario-extreme scenario” classification-driven mechanism to achieve synergistic optimization of benefits and risks: This invention changes the existing method’s mode of seeking the expected value of all inputs in multiple scenarios. It uses typical scenarios to drive benchmark scheduling to pursue high probability benefits, and extreme scenarios are used separately for robustness verification to identify low probability risks. The combination of the two can effectively reduce the risk of water abandonment / power shortage without sacrificing expected benefits, and balance the benefits and risk robustness of scheduling. Attached Figure Description

[0163] To more clearly illustrate the technical solutions in the embodiments of the present invention and / or the prior art, the drawings used in the description of the embodiments and / or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0164] Figure 1 This is a technical roadmap of an embodiment of the present invention. Detailed Implementation

[0165] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following embodiments.

[0166] This invention provides a method for generating a set of cascade hydropower dispatch scenarios that considers the joint uncertainty of inflow water, electricity price, and generation space, in order to solve the following technical problems:

[0167] (1) Overcoming the limitations of single or two-dimensional uncertainty: For the first time, “power generation space” is included as an independent uncertainty variable in the uncertainty analysis framework of cascade hydropower dispatch.

[0168] (2) Characterize the complex interdependencies between variables: The three-dimensional Copula function is used to capture the interdependency structure between water flow, nodal electricity price and power generation space, so as to realize the modeling of the joint uncertainty of water flow, nodal electricity price and power generation space.

[0169] (3) Coupling of scenario set and scheduling decision: Through the scenario classification mechanism, the scenario set is divided into "high probability typical scenario" and "low probability extreme scenario", and the constraint logic of "the upper limit of the final output of cascade hydropower is the smaller value of the potential output determined by the inflow and the power generation space" is explicitly introduced into the scheduling model, so that the scenario set can directly drive different scheduling objectives (baseline plan scheduling and robust plan scheduling) and improve the engineering practicality of the scheduling scheme.

[0170] The implementation of this invention does not depend on specific river basin or power station parameters. The generation of dispatch scenario sets for any cascade hydropower station group in the electricity spot market environment can be achieved by referring to the following steps.

[0171] Example 1:

[0172] like Figure 1 As shown, this embodiment provides a method for generating a cascade hydropower dispatch scenario set considering the joint uncertainty of inflow water, electricity price, and power generation space, including the following steps:

[0173] Step 1: Construct a univariate uncertainty quantification model

[0174] This step constructs probabilistic prediction models for inflow, nodal electricity price, and power generation space, and outputs their respective predicted values ​​and uncertainty intervals.

[0175] (1) Water inflow forecasting and uncertainty quantification

[0176] Deterministic forecasting of inflow for cascade hydropower projects based on distributed hydrological models (such as the Community Water Model, CWatM model). and the actual inflow rate monitored by the station A "forecast-measured" mapping model was constructed using quantile regression technology:

[0177] ;

[0178] in, The simulated flow rate generated by the distributed hydrological model at time t; It is the flow rate at the quantile τ to be predicted; The quantiles covering the uncertainty interval. .in Representing the median prediction, in this embodiment, it can be taken as... Represents the upper limit of runoff prediction (only) The probability is higher than this value, 0 < <0.5), This represents the lower limit of runoff prediction. The intercept term represents the systematic bias of the physical model; This is the slope term, representing the factor by which the physical model amplifies or reduces runoff fluctuations. For different... , and They are different.

[0179] To obtain robust runoff estimates under extreme drought and flood conditions, the model fits the model parameters by minimizing the checked ball loss function. and The formula is as follows:

[0180]

[0181]

[0182] in, This is the error value, i.e. ; This is an indicator function; if the error is less than 0 (the forecast was too high), the value is 1; otherwise, it is 0. Its core logic is: if you want to infer... The formula will underestimate ( High-weight penalties are applied to cases where the value is overestimated, while penalties are applied to cases where the value is overestimated. In cases where only low weight is applied, the same applies to cases where high weight is applied.

[0183] Pick =0.001, 0.005, 0.01, 0.02, ..., 0.5, ..., 0.98, 0.99, 0.995, 0.999, for and Each was fitted separately to obtain a set The scatter plot is used to fit the extreme flood portion using the generalized Pareto distribution (GPD), and the remaining portion is fitted using cubic splines, thus obtaining the marginal distribution of the inflow.

[0184] Therefore, in the forecasting phase, the lower limit of the forecasted flow is... , median and upper limit They are respectively:

[0185]

[0186]

[0187]

[0188] in, For future runoff predicted by a distributed hydrological model.

[0189] (2) Electricity price forecasting and uncertainty quantification

[0190] Electricity price fluctuations are primarily influenced by market supply and demand. Therefore, this invention constructs a mapping relationship between total load, other power output, and nodal prices based on supply and demand curves to perform probabilistic electricity price prediction. First, the predicted values ​​of driving factors such as other hydropower generation capacity and new energy output, intra-provincial demand, and inter-provincial transmission are used as input features. The nonlinear mapping relationship between these features and historical clearing prices is learned, and the predicted values ​​are output at 15-minute intervals. The mapping model between nodal prices and supply and demand data is first performed:

[0191]

[0192]

[0193] in, for The total load forecast value of the sending or receiving system at any given time; for The predicted output values ​​of other power sources within the system at any given time; Net load, for The net load at any given time represents the remaining demand that must be met by this type of power supply; for Historical marginal electricity prices of the transmission channel nodes at all times; , , , , The model parameters to be determined are obtained by fitting using the least squares method; The term is an exponential term, simulating the surge in electricity prices when the net load approaches the system's installed capacity limit; The residual term represents random fluctuations that the model cannot explain, and is usually assumed to follow a normal distribution. .

[0194] According to the above formula, the main source of electricity price uncertainty is load forecasting error. Other energy forecast errors (Assuming the errors of both follow a normal distribution):

[0195]

[0196]

[0197] Assuming the errors of both are independent and identically distributed, the joint standard deviation of the net load is:

[0198]

[0199] Within the standard deviation of net load, massive electricity price results were generated based on Monte Carlo simulation:

[0200]

[0201] in, This is the net load value sequence number. Then, the net load and electricity price are mapped, generating multiple sets... Substituting the fitted curves into (3.6-8) and superimposing the residual random terms, we obtain several possible electricity price scenarios. .

[0202] The above electricity pricing scenarios are transformed into probabilistic electricity prices, where the expected electricity price is... Represented as:

[0203]

[0204] according to Sort by size from smallest to largest, and take and As a confidence interval for electricity prices, i.e. .

[0205] (3) Spatial forecasting of power generation and quantification of uncertainty

[0206] The "power generation space" involved in this invention is defined as: the upper limit of the actual safe power output of a hydropower station under given physical boundary conditions such as grid topology, channel capacity, unit maintenance status, and vibration zone constraints, reflecting the actual power generation capacity at the grid and equipment levels. A Gaussian process regression model is first used to predict the reference power generation space, and then post-processing based on physical constraints ensures that the predicted value conforms to reality.

[0207] Basic characteristic data preparation: baseline scale of power generation space Based on the predicted total power output of the cascade hydropower stations in the province and the predicted output share of cascade hydropower stations Decide:

[0208]

[0209]

[0210] in, This represents the daily power generation of the target hydropower station during a historical period. Because the combined power generation capacity of a cascade hydropower station group is far greater than that of small hydropower stations, therefore... It generally falls within a stable range. .because It may be affected by unconventional factors such as policies, channel restrictions, and maintenance events, resulting in a very small probability of significant fluctuations. Therefore, to remove the influence of extreme values, a range of 5% to 95% is selected. Calculate the average value .

[0211] The physical and topological constraint feature data are hard constraints, including maintenance information (start and end times of unit maintenance shutdown, and transmission and transformation maintenance information for the same day). DC channel output power and cross-sectional constraints Research on the output constraints of the target power plant units :

[0212]

[0213] Market and dispatch characteristic data include inter-provincial power transmission agreements (lower limit of inter-provincial power transmission for West-to-East power transmission, total daily power transmission from Yunnan Province at 96 points in history), historical and current-day electricity prices and transaction results of each hydropower station node, and historical daily power transmission from DC channels involved in each hydropower station.

[0214] Gaussian process regression modeling: for each 15-minute time node Construct feature vectors:

[0215]

[0216] in, This is the initial power generation share allocated based on the total projected power output; The cross-sectional margin of the passage corresponding to the target hydropower station; Day-ahead / real-time node price; Under maintenance; Typical output shape factor for similar historical days (obtained based on market and scheduling characteristics).

[0217] Next, a kernel function needs to be selected. Since the power generation space exhibits significant intra-diurnal periodicity and smooth evolution characteristics, a hybrid kernel function is chosen. :

[0218]

[0219] in, It is a radial basis function kernel, which is a commonly used kernel function to capture nonlinear mappings between features; Using a periodic kernel, the cyclical characteristics of a 15-minute sequence within a day are captured; It is a white noise core that absorbs measurement errors and non-systematic fluctuations.

[0220] Finally, model training and parameter optimization are performed. Using the aforementioned historical measured data as the training set, the hyperparameters (length scale) of the kernel function are determined by maximizing the log-marginal likelihood function. and variance ), thereby estimating the power generation space:

[0221]

[0222] After the GPR model is trained, the features to be predicted at a certain time before the current date are... GPR will provide predictions and probability distributions of power generation space, including mean prediction (expected value of power generation space). and variance Forecast (Measurement of Spatial Uncertainty in Power Generation):

[0223]

[0224]

[0225]

[0226] in, It is the covariance variable between the newly input independent variable and the training set.

[0227] Physical constraint verification: Since GPR outputs statistical results, but hydropower scheduling must adhere to hard physical constraints, constraint truncation and uncertainty correction are necessary. The final prediction space must satisfy the following formula;

[0228]

[0229] in, This represents the uncertainty interval. If a transmission line maintenance or generator maintenance occurs at a certain time, the mean value at that time... The forced correction is set to 0, and the variance is... It should also approach 0.

[0230] Step 2: Construct a three-dimensional joint distribution model based on Copula.

[0231] Using the three marginal distributions obtained in step 1, a joint probability distribution of inflow, nodal electricity price, and generation space is constructed using a three-dimensional Gaussian Copula function. Inflow Node electricity price Power generation space Construct a system state vector :

[0232]

[0233] Fit their marginal cumulative distribution functions respectively Variables in actual physical space , , Transform it into a probability space variable on the interval [0,1] using its cumulative distribution function:

[0234]

[0235]

[0236] Therefore, the true joint distribution of the three variables can be expressed as:

[0237]

[0238] in, This is a 3x3 Pearson correlation coefficient matrix. When using the three-dimensional Gaussian Copula function C, its specific mathematical expression is:

[0239]

[0240] in, It is the inverse function of the standard normal distribution. It is a vector with a mean of 0 and a covariance matrix of The cumulative function of the three-dimensional standard normal distribution. Define an auxiliary vector. ,in:

[0241]

[0242] Step 3: Generate an initial set of multidimensional uncertainty scenarios

[0243] First, the Monte Carlo method is used to generate N independent three-dimensional random vectors. :

[0244]

[0245] in Let be three independent, standard normally distributed random variables. The correlation coefficient matrix estimated in the above standard normally space is... Perform Cholesky decomposition:

[0246]

[0247]

[0248] at this time, The physical interdependence between water supply, power generation space, and electricity price has already been established. Next, two inverse transformations will be used to restore the physical space:

[0249]

[0250]

[0251] At this point, we can obtain N initial scene sets:

[0252]

[0253] Because it is Monte Carlo random sampling with equal probability, the initial probability of each scene occurring is... All are:

[0254]

[0255] Step 4: Scene Reduction and Category Recommendation

[0256] Due to the large number of initial scenarios, this step reduces the number of scenarios and categorizes them according to their application value to reduce the computational burden of optimization and scheduling. First, the K most representative scenarios are selected from the N scenarios, and their weight probabilities are reassigned based on minimizing the Wasserstein distance. Since the numerical magnitudes of water flow, power generation space, and electricity price are different, the scenarios for each variable must first be Z-score standardized (mean 0, variance 1) to obtain the standardized scenarios. Define any two scenarios. and Three-dimensional geometric distance between :

[0257]

[0258] Set the target retention scene set as (Initially an empty set), the set of scenes to be reduced is: (Initially N scenes), selected using the synchronous back-substitution elimination method. Points:

[0259] ① Find the absolute center point: Find the point whose sum of distances to all other points is the smallest, and place it there. From Remove from the middle;

[0260] ② Cyclic selection: Assuming a selection has already been made The point is at In China, for Each of the remaining candidate points Calculate if we take Add In the system, how much does the probabilistic distance between the unselected points and the set of selected points decrease? Find the point that causes the largest decrease in this distance. Put it in .

[0261] ③ Termination decision: Repeat the selection until a set is reached. The middle just contains A scenario.

[0262] The reduced scenario will increase its own probability. The final retained scene is the one closest to it. Final recommendation probability for:

[0263]

[0264] in, This represents the set of all eliminated scenarios that are closest to the representative scenario r.

[0265] Select the top M scenarios with the highest probability, ensuring that the cumulative probability of these scenarios exceeds 50%, representing typical high-probability scenarios, i.e., the baseline plan scheduling.

[0266] From the remaining scenarios, select scenarios where the physical values ​​of power generation space, water flow, and nodal electricity price are individually or jointly at the boundary to represent low-probability extreme scenarios and serve robust scheduling.

[0267] Step 5: Construct a scenario-driven cascade hydropower optimization scheduling model

[0268] The inflow rate, nodal electricity price, and power generation space scenario set generated in steps 1-4 are embedded into the cascade hydropower optimal scheduling model. Inflow rate determines the total supply and process of hydropower resources; electricity price guides the optimization model to allocate power output to maximize revenue; power generation space serves as an upper limit constraint on output, reflecting the limitations imposed on output by external physical boundaries such as power grid channels, maintenance, and cross-sections. This invention aims to maximize power generation efficiency by constructing a day-ahead (short-term) optimal scheduling model for cascade hydropower. The objective function is:

[0269]

[0270] in, The time period number; The total number of time periods (in the case of a 15-minute step size). (96) For hydropower station serial numbers; This represents the total number of hydropower stations in the entire cascade. For power station During the period The power generation capacity; For the scene Next period Predicted clearing price at the node; The time interval is the step size.

[0271] The cascade hydropower optimal scheduling model takes water flow, nodal electricity price, and power generation space as inputs and outputs a scheduling scheme. It can be solved using existing genetic algorithms or other existing optimal scheduling models. This model only needs to use existing technologies to achieve short-term optimal scheduling functions, which will not be elaborated here.

[0272] The constraints include:

[0273] (1) Water balance constraint:

[0274]

[0275] in and For power station During the period and time period Storage capacity, and Power plants During the period Inbound and outbound flow.

[0276] (2) Output-Power Generation Flow-Head Relationship:

[0277]

[0278] in This is the NHQ curve function. For power station During the period The water purifier head, For power station During the period The power generation flow.

[0279] (3) Space constraints for power generation (core):

[0280]

[0281] in Contribute to the medium-term plan, This is a configurable parameter for the allowable positive deviation. This is a configurable parameter for allowing negative deviation. For the scene Next period The maximum power output of the power plant.

[0282] (4) Short-term power generation constraints:

[0283]

[0284] in The daily power generation target is given for medium-term dispatch.

[0285] (5) Reservoir water level constraints:

[0286]

[0287] in , and Power plants The water level, dead water level, and normal water level at time t.

[0288] (6) Power generation flow constraints:

[0289]

[0290] in , and Power plants The power generation flow rate, minimum power generation flow rate, and maximum power generation flow rate at time t.

[0291] (7) Outbound flow constraints:

[0292]

[0293] in , and Power plants The outbound flow rate, minimum outbound flow rate, and maximum outbound flow rate at time t.

[0294] (8) Output ramping constraint:

[0295]

[0296] in For power station The maximum permissible range of output variation between adjacent time periods.

[0297] (9) Vibration zone avoidance constraints:

[0298]

[0299] in The vibration zone number is... The total number of vibration zones, and These are the lower and upper limits of the vibration zone, respectively.

[0300] (10) When the water flow is stagnant:

[0301]

[0302] in For power station and power station (If m is the leading power station, then) This refers to the inflow within the interval between the natural inflow of the power station. For water flow from the upstream power station The dam site flows to the downstream power station Time required for dam site construction.

[0303] Under the baseline scheduling model, the inflow rate, nodal electricity price, and generation capacity under high-probability typical scenarios are simultaneously input into the above model to solve for the optimal power output process. As the benchmark generation plan for the day-ahead market;

[0304] In the robust scheduling mode, the inflow of water, nodal electricity price, and power generation space under low-probability extreme scenarios are substituted. If the limits are exceeded, the strategy is adjusted, such as optimizing the output curve and reserving reserve capacity, to generate a robust scheduling scheme.

[0305] Example 2:

[0306] like Figure 1 As shown, this embodiment provides a method for generating a cascade hydropower dispatch scenario set considering the joint uncertainty of inflow water, electricity price, and power generation space, including the following steps:

[0307] S1 Constructing a Quantitative Model of Spatial Uncertainty in Water Inflow, Electricity Price, and Power Generation

[0308] For each hydropower station in the target cascade hydropower station group, a marginal probability distribution model is established for inflow, nodal electricity price, and power generation space, specifically including:

[0309] S1.1 Quantification of Inflow Uncertainty. Collect historical inflow measurement data and distributed hydrological model prediction data for the same time period from different stages of the cascade hydropower stations. The time period should be no less than one year, ideally including data from the wet season, dry season, and normal water season. Set a set of variables. , representing the lower limit, median, and upper limit of uncertainty, respectively. A quantile regression method is employed, minimizing the check sphere loss function to solve for the regression parameter values ​​at different quantiles. In actual prediction, the flow predicted by the distributed hydrological model is input into the trained quantile regression model to obtain the predicted inflow values ​​at different quantiles, and then the cumulative distribution function of the inflow to different power stations within the cascade is fitted. .

[0310] S1.2 Quantification of Electricity Price Uncertainty. Historical load demand data, renewable energy (wind and solar) output data, thermal power output data, non-market-based mandatory power output data, and day-ahead clearing price data are collected for the target electricity market. A net load sequence is constructed based on "total load - other power output". Using the above formula The supply-demand mapping relationship between the net load sequence and the electricity price sequence is fitted using the least squares method. Simultaneously, the total load forecast error and other power output forecast errors are statistically analyzed to obtain the standard deviation of the net load forecast. By generating a large number of net load samples through Monte Carlo simulations and substituting them into the supply-demand mapping curve to obtain an electricity price sample set, the cumulative distribution function of electricity prices can be fitted. .

[0311] S1.3 Quantification of Generation Spatial Uncertainty. Historical power output ratios of target hydropower stations (groups) relative to the province's total power output are collected. Historical electricity prices, maintenance events, and channel margins at each node are used as input features, and a Gaussian process regression model is trained. The output ratio output by the Gaussian process regression model is multiplied by the predicted total power output for the second day in the province to obtain the predicted day-ahead generation space of the cascade hydropower stations. The Gaussian process regression model outputs the mean and uncertainty interval (posterior variance), thus yielding the cumulative distribution function of the generation space. Finally, the results need to be physically verified using channel blockage and maintenance events. During maintenance periods, the power generation space of the power station / unit should be forcibly set to 0.

[0312] S2 Constructs a three-dimensional joint distribution model based on Copula.

[0313] By selecting historical data from multiple years for the same season as the forecast, the cumulative probability values ​​for water inflow, nodal electricity price, and power generation potential are calculated for each day within that period. Then, it is transformed into a standard normal spatial variable using the inverse function of the standard normal distribution. Calculate the correlation coefficient matrix of the three Z-sequences. The three-dimensional Gaussian Copula function is determined using this matrix. Thus, the joint probability distribution of the three variables is obtained. .

[0314] S3 generates an initial set of multidimensional uncertainty scenarios.

[0315] Monte Carlo sampling method is used to obtain data from the joint probability distribution. Generate a large number of initial scenarios: For the k-th sampling, generate a vector of three independent random numbers that follow a standard normal distribution. Regarding the correlation coefficient matrix Cholesky decomposition yields the lower triangular matrix. According to the formula The result is a sequence with a mean of 0 and a covariance matrix of... The three-dimensional normal distribution. Then... Convert to uniform distribution variable Then, using the inverse function of the cumulative distribution function of each variable obtained in S1, the uniformly distributed variable is converted into its actual value in physical space. Thus, the k-th scenario This is a complete scenario that includes inflow water volume, nodal electricity prices, power generation spatial information, and the physical dependencies between them. Repeating the above process N times (N is determined based on computing resources, typically 1000-2000) yields the initial scenario set. .

[0316] S4 Scene Reduction and Classification Recommendation Based on Probabilistic Distance

[0317] S4.1 Scene Reduction. Variables in each scene are Z-score standardized to eliminate the influence of dimensions, and then the 3D geometric distance between any two scenes is calculated. A simultaneous back-substitution elimination method is used, aiming to minimize the Wasserstein distance. Initial Scene Xue Jianwei Representative scenarios ( And, and reassign weights to the retained scenes.

[0318] S4.2 Scene Category Recommendation. (The following will be retained...) The scenarios are sorted from highest to lowest probability. The top M scenarios with a cumulative probability exceeding 50% are selected as high-probability typical scenarios, representing the most likely "water inflow-electricity price-generation space" state. Among the remaining scenarios, one or more scenarios where the three physical values ​​of water inflow, nodal electricity price, and generation space are at their boundary positions are selected as low-probability extreme scenarios for risk verification of the scheduling scheme.

[0319] S5 Scenario-Driven Cascade Hydropower Optimization Scheduling

[0320] S5.1 Baseline Scheduled Dispatch. The inflow sequence, electricity price sequence, and generation spatial sequence from high-probability typical scenarios are simultaneously input into the short-term day-ahead dispatch model. Its objective function is to maximize generation efficiency under the electricity spot market.

[0321]

[0322] Among these constraints, the key constraint on the utilization of power generation space is the upper limit constraint on power output:

[0323]

[0324] This constraint ensures that the short-term scheduling plan strictly conforms to the physical feasibility of the power grid and generating equipment without deviating excessively from the medium-term optimal scheduling objective. A mixed-integer linear programming solver is used to solve the above model, yielding the output plan and expected revenue for the second day at 96 points.

[0325] S5.2 Robust Scheduler. The solution method for robust schedulers is the same as above. If the solution cannot avoid exceeding the limit ( If the plan is deemed to have a risk of water wastage / power shortage under extreme water inflow or power grid physical limitations, the dispatcher will be advised to adjust the output plan or apply for exemption from assessment, and the adjusted strategy will be reported to the market.

[0326] Those skilled in the art can, based on the above steps and in conjunction with specific cascade hydropower station attribute data, electricity market data, water inflow forecast data, and grid operation constraints, realize the proposed method for generating cascade hydropower dispatch scenarios that considers the joint uncertainty of water inflow, electricity price, and power generation space in the electricity spot market.

[0327] Example 3:

[0328] This embodiment provides a device for generating a cascade hydropower dispatch scenario set considering the joint uncertainty of water inflow, electricity price, and power generation space, including a processor and a storage medium;

[0329] The storage medium is used to store instructions;

[0330] The processor is configured to operate according to the instructions to perform the steps of the method according to Embodiment 1 or Embodiment 2.

[0331] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0332] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0333] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0334] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0335] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for generating a set of cascade hydropower dispatch scenarios considering the joint uncertainty of inflow water, electricity price, and generation space, characterized in that, Includes the following steps: Step 1: Construct univariate uncertainty quantification models for water inflow, nodal electricity price, and power generation space respectively, and output the marginal probability distributions of water inflow, nodal electricity price, and power generation space; Step 2: Based on the marginal probability distributions of the inflow, nodal electricity price, and power generation space, a three-dimensional Copula function is used to construct the joint probability distribution of the inflow, nodal electricity price, and power generation space to characterize the interdependence among the three. Step 3: Based on the joint probability distribution, generate an initial multidimensional uncertainty scenario set using the Monte Carlo method; Step 4: Reduce the initial multidimensional uncertainty scene set into scenes, and classify and recommend the reduced scenes according to scene probability and features to obtain high-probability typical scenes and low-probability extreme scenes. The power generation space refers to the upper limit of power that a hydropower station can safely generate under the physical boundary conditions of given grid topology, channel capacity, unit maintenance status and vibration zone constraints.

2. The method for generating a cascade hydropower dispatch scenario set according to claim 1, characterized in that, In step 1, a univariate uncertainty quantification model is constructed for the inflow water flow, including: Deterministic forecasts of inflow for cascade hydropower projects based on distributed hydrological models and the actual inflow rate monitored by the station A "forecast-measured" mapping model was constructed using quantile regression technology: ； in, The simulated flow rate generated by the distributed hydrological model at time t; It is the flow rate at the quantile τ to be predicted; The quantiles covering the uncertainty interval. ; The intercept term represents the systematic bias of the physical model; The slope term represents the factor by which the physical model amplifies or reduces runoff fluctuations; To obtain robust runoff estimates under extreme drought and flood conditions, the model fits the model parameters by minimizing the checked ball loss function. and The formula is as follows: ； ； in, This is the error value, i.e. ; This is an indicator function; if the error is less than 0, the value is 1; otherwise, it is 0. =0.001, 0.005, 0.01, 0.02, ..., 0.5, ..., 0.98, 0.99, 0.995, 0.999, for and Each was fitted separately to obtain a set The scatter plot is used to fit the extreme flood portion using a generalized Pareto distribution, while the remaining portion is fitted using cubic splines, thus obtaining the marginal distribution of the inflow.

3. The method for generating a cascade hydropower dispatch scenario set according to claim 1, characterized in that, In step 1, a univariate uncertainty quantification model is constructed for nodal electricity prices, including: Mapping model of nodal electricity prices and supply and demand data: ； ； in, for The total load forecast value of the sending or receiving system at any given time; for The predicted output values ​​of other power sources within the system at any given time; for The net load at any given time represents the remaining demand that must be met by this type of power supply; for Historical marginal electricity prices of the transmission channel nodes at all times; Net load and electricity price Functional mapping relationship between them; , , , , The model parameters to be determined are obtained by fitting using the least squares method; The term is an exponential term, simulating the surge in electricity prices when the net load approaches the system's installed capacity limit; The residual term represents random fluctuations that the model cannot explain, and it is usually assumed that it follows a normal distribution. According to the above formula, the main source of electricity price uncertainty is load forecasting error. Other energy forecast errors : ； ； Assuming the errors of both are independent and identically distributed, then the joint standard deviation of the net load... for: ； Within the standard deviation of net load, massive electricity price results were generated based on Monte Carlo simulation: ； in, It is the first The scenario of Monte Carlo sampling; It is the first Sample values ​​of net load scenario obtained from Monte Carlo sampling; It is a deterministic electricity price forecast provided by an external model; It is the first The random disturbance term of the second sampling; then the net load and electricity price are mapped, resulting in multiple sets of generated data. Substitute the above-fitted net load-electricity price curve ~ In addition, residual random terms are superimposed to obtain multiple possible electricity price scenarios. ; The above electricity pricing scenarios are transformed into probabilistic electricity prices, where the expected electricity price is... Represented as: ； according to Sort by size from smallest to largest, and take and ( Using the confidence interval of electricity price as a reference, the resulting sample set represents the marginal distribution of electricity price.

4. The method for generating a cascade hydropower dispatching scenario set according to claim 1, characterized in that, In step 1, a univariate uncertainty quantification model is constructed for the power generation space, including: Basic characteristic data preparation: baseline scale of power generation space Based on the predicted total power output of the cascade hydropower stations in the province and the predicted output share of cascade hydropower stations Decide: ； ； in, This represents the daily power generation of the target hydropower station during a historical period; since the combined power generation capacity of a cascade hydropower station group is far greater than that of small hydropower stations, therefore... It generally falls within a stable range. ;because It may be affected by unconventional factors such as policies, channel restrictions, and maintenance events, resulting in a very small probability of significant fluctuations. Therefore, to remove the influence of extreme values, a range of 5% to 95% is selected. Calculate the average value ; Physical and topological constraint feature data are hard constraints, including maintenance information. DC channel output power and cross-sectional constraints Research on the output constraints of the target power plant units : ； Market and dispatch characteristic data include inter-provincial power transmission agreements, historical and day-ahead electricity prices and transaction results of each hydropower station node, and historical daily power transmission volume of each hydropower station's DC transmission channels; Gaussian process regression modeling: for each 15-minute time node Construct feature vectors: ； in, This is the initial power generation share allocated based on the total projected power output; The cross-sectional margin of the passage corresponding to the target hydropower station; Day-ahead / real-time node price; Under maintenance; This is a typical output shape factor for similar days in history; Next, a kernel function needs to be selected; since the power generation space exhibits significant intra-diurnal periodicity and smooth evolution characteristics, a hybrid kernel function is chosen. : ； in, It is a radial basis function kernel, which is a commonly used kernel function to capture nonlinear mappings between features; Using a periodic kernel, the cyclical characteristics of a 15-minute sequence within a day are captured; It acts as a white noise core, absorbing measurement errors and non-systematic fluctuations; Finally, model training and parameter optimization were performed. Using the aforementioned historical measured data as the training set, the model was trained by maximizing the log-marginal likelihood function. To determine the hyperparameters of the kernel function, thereby estimating the power generation space: ； in, It is the number of samples in the training set; It is the output vector of the training set; that is, the spatial value of power generation measured in history. For kernel function The matrix formed represents the covariance structure among the training samples; For noise variance; The identity matrix is ​​used; after the GPR model is trained, for the features at a certain time before the current date to be predicted... GPR will provide predictions and probability distributions of the power generation space, including mean predictions. and variance predict: ； ； ； in, It is the covariance variable between the newly input independent variable and the training set; at this time, the marginal probability distribution of the power generation space is: ； Physical constraint verification: Since GPR outputs statistical results, but hydropower dispatch must adhere to hard physical constraints, constraint truncation and uncertainty correction are necessary; predicted power generation space. The following formula must be satisfied: ； in, The uncertainty interval is defined; the marginal distribution of the power generation space follows a right-truncated normal distribution; if transmission line maintenance or unit maintenance occurs at a certain moment, the mean value at that moment is... The forced correction is set to 0, and the variance is... It should also approach 0. Based on the right-truncated normal distribution, the discrete part follows a 0-value degenerate distribution.

5. The method for generating a cascade hydropower dispatch scenario set according to claim 4, characterized in that, In step 2, based on the marginal probability distributions of the inflow water flow, nodal electricity price, and power generation space, a three-dimensional Copula function is used to construct a joint probability distribution of the inflow water flow, nodal electricity price, and power generation space to characterize the interdependence among the three, including: Using the marginal probability distributions of water flow, nodal electricity price, and power generation space obtained in step 1, a joint probability distribution of water flow, nodal electricity price, and power generation space is constructed using a three-dimensional Gaussian Copula function; Inflow rate Node electricity price Power generation space Construct a system state vector : ； Fit their marginal cumulative distribution functions respectively ; variables in actual physical space , , Transform it into a probability space variable on the interval [0,1] through its cumulative distribution function, so as to express: ； in, Let i represent the i-th probability space variable. These represent the probability space variables of inflow water, nodal electricity price, and power generation space, respectively. Let the probability space vector be... Therefore, the true joint distribution of the three variables can be expressed as: ； in, The Pearson correlation coefficient matrix is ​​3x3; when using the three-dimensional Gaussian Copula function C, its mathematical expression is: ； in, It is the inverse function of the standard normal distribution. The mean is a vector of 0, and the correlation coefficient matrix is... The three-dimensional standard normal cumulative function; define the auxiliary vector. ,in: ; in, Let i represent the i-th auxiliary variable. These represent auxiliary variables: inflow rate, nodal electricity price, and power generation space, respectively.

6. The method for generating a cascade hydropower dispatching scenario set according to claim 5, characterized in that, In step 3, based on the joint probability distribution, an initial set of multidimensional uncertainty scenarios is generated using the Monte Carlo method, including: N independent three-dimensional random vectors are generated using the Monte Carlo method. : ； Where the superscript (k) indicates the k-th element generated by the Monte Carlo method, and there are a total of N elements; For the correlation coefficient matrix Perform Cholesky decomposition: ； ； at this time, Having established the physical interdependence between water supply, power generation space, and electricity price, the physical space is restored through two inverse transformations: ； ； in, This represents the inverse function of the marginal cumulative distribution function of the i-th variable; This represents the k-th initial scene; At this point, N initial scene sets can be obtained. : ； Because it is Monte Carlo random sampling with equal probability, the initial probability of each scene occurring is... All are: 。 7. The method for generating a cascade hydropower dispatching scenario set according to claim 6, characterized in that, In step 4, the initial multidimensional uncertainty scene set is reduced in size, and the reduced scenes are classified and recommended based on scene probability and features to obtain high-probability typical scenes and low-probability extreme scenes, including: For the initial scene set Each variable is standardized using Z-score to obtain the k-th standardized scenario. ; Define any two scenarios and Three-dimensional geometric distance between : ； in, and Scenes and Coordinates in three-dimensional space; Representing the scene respectively Standardized values ​​for inflow water volume, nodal electricity price, and power generation space; Representing the scene respectively Standardized values ​​for inflow water volume, nodal electricity price, and power generation space; Set the target retention scene set as Initially an empty set, the set of scenes to be reduced is: Initially, there are N scenes, which are selected using the synchronous back-substitution elimination method. Points: ① Find the absolute center point: Find the point whose sum of distances to all other points is the smallest, and place it there. From Remove from the middle; ② Cyclic selection: For Each of the remaining candidate points Calculate if we take Add In the context of the entire system, the goal is to find the point that maximizes the decrease in the probability distance from the unselected points to the set of selected points. Put it in ; ③ Termination decision: Repeat the selection until a set is reached. The middle just contains One scenario; The reduced scenario will increase its own probability. The final retained scene is the one closest to it. Final recommendation probability for: ； in, This represents the set of all eliminated scenarios that are closest to the representative scenario r. Take the top M scenarios with the highest probability, and ensure that the cumulative probability of the above scenarios exceeds 50% to obtain a set of high-probability typical scenarios, which represent high-probability typical scenarios and serve the baseline plan scheduling. From the remaining scenarios, select scenarios where the physical values ​​of power generation space, water flow, and node electricity price are individually or jointly located at the boundary to obtain a set of low-probability extreme scenarios, which represent low-probability extreme scenarios and serve robust scheduling.

8. The method for generating a cascade hydropower dispatch scenario set according to claim 1, characterized in that, The method further includes: Step 5: Embed the high-probability typical scenarios and / or the low-probability extreme scenarios into the cascade hydropower optimization scheduling model to generate scheduling schemes for different scheduling objectives.

9. The method for generating a cascade hydropower dispatch scenario set according to claim 8, characterized in that, In step 5, the high-probability typical scenarios and / or the low-probability extreme scenarios are embedded into the cascade hydropower optimal scheduling model to generate scheduling schemes for different scheduling objectives, including: To maximize power generation efficiency, a day-ahead optimal scheduling model for cascade hydropower is constructed; the objective function is: ； in, For time period sequence number; This represents the total number of time periods; For hydropower station serial numbers; This represents the total number of hydropower stations in the entire cascade. For power station During the period The power generation capacity; For the scene Next period Predicted clearing price at the node; The time interval step; Under the baseline scheduling model, the inflow rate, nodal electricity price, and generation capacity under high-probability typical scenarios are simultaneously input into the day-ahead optimization scheduling model of cascade hydropower to solve for the optimal output process. As the benchmark generation plan for the day-ahead market; Under the robust scheduling model, the inflow rate, nodal electricity price, and generation capacity under low-probability extreme scenarios are substituted into the day-ahead optimization scheduling model of cascade hydropower. If the limits are exceeded, the strategy is adjusted, and the optimal output process is obtained by solving the problem. As a robust scheduling scheme; The constraints of the day-ahead optimization scheduling model for cascade hydropower include: Water balance constraints: ； in and For power station During the period and time period Storage capacity, and Power plants During the period Inbound and outbound flow; Power output - power generation flow - head relationship: ； in This is the NHQ curve function. For power station During the period The water purifier head, For power station During the period The power generation flow rate; Space constraints for power generation: ； in Contribute to the medium-term plan, This is a configurable parameter for the allowable positive deviation. This is a configurable parameter for allowing negative deviation. For the scene Next period The maximum power output of a power plant; Short-term power generation constraints: ； in The daily power generation target given for medium-term dispatch; Reservoir water level constraints: ； in , and Power plants Water level, dead water level, and normal water level at time t; Power generation flow constraints: ； in , and Power plants The power generation flow rate, minimum power generation flow rate, and maximum power generation flow rate at time t; Outbound flow constraints: ； in , and Power plants The outbound flow rate, minimum outbound flow rate, and maximum outbound flow rate at time t; Output climbing constraints: ； in For power station The maximum permissible range of output variation between adjacent time periods; Vibration zone avoidance constraints: ； in The vibration zone number is... The total number of vibration zones, and These are the lower and upper limits of the vibration zone, respectively. Water flow stagnation time: ； in For power station and power station If the inflow interval between m and m is the head power station, then It's the power station's natural inflow. For water flow from the upstream power station The dam site flows to the downstream power station Time required for dam site construction.

10. A device for generating a set of cascade hydropower dispatch scenarios considering the joint uncertainty of inflow water, electricity price, and generation space, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1 to 9.

Images