Power grid and park collaborative probabilistic scheduling method based on gaussian mixture model and conditional value at risk

Abstract

The application discloses a power grid park collaborative probability scheduling method and system based on a Gaussian mixture model and a conditional risk value, and belongs to the technical field of power dispatching. The application collects and normalizes multi-source data of power grid thermal power units, park hydrogen energy, multi-node loads, grid-park tie lines and bilateral wind and light prediction error; adopts a Gaussian algorithm to construct a Gaussian mixture model to depict wind and light multi-peak error distribution, generates discrete output scenarios through Monte Carlo sampling; builds a deterministic constraint containing thermal power, hydrogen energy and power interaction, and introduces a conditional risk value linear constraint to control park load shedding, power grid up / down reserve shortage tail risk, and wind and light abandonment as a scene expected penalty; and constructs a mixed integer linear programming model to solve the minimum system comprehensive operation cost as a target. The application accurately represents new energy random fluctuation, quantifies extreme scene risk, and takes into account grid-park collaboration and hydrogen energy storage regulation, while guaranteeing power supply reliability and reducing wind and light abandonment and load shedding loss, and is suitable for high new energy penetration rate park grid collaborative scheduling scene.

Description

Technical Field

[0001] This invention relates to the field of power system operation and dispatching, specifically to a probabilistic dispatching method for power grid and industrial park collaboration based on Gaussian mixture model and conditional value risk. Background Technology

[0002] Due to the significant randomness, volatility, and intermittency of wind and solar power output, the increasing penetration rate of these technologies in the power grid significantly enhances the uncertainty of system operation. This can easily lead to problems such as wind and solar curtailment, power imbalance, and insufficient reserves, seriously affecting the safe and stable operation of the power grid. Existing dispatching methods mostly employ deterministic or simple probabilistic models, which are insufficient to accurately characterize the multi-peak distribution characteristics of renewable energy output and do not adequately consider operational risks under extreme scenarios. Furthermore, while traditional energy storage methods can mitigate volatility to some extent, they suffer from high costs and limited adjustment mechanisms, making it difficult to balance economy and reliability in large-scale applications. Therefore, it is necessary to propose an optimized dispatching method that can accurately describe the uncertainty of renewable energy, balance risk control, and improve grid-park coordination capabilities to enhance renewable energy absorption and ensure safe system operation. Summary of the Invention

[0003] To address the aforementioned issues, this invention proposes a grid-industrial park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk. By introducing a hydrogen energy system and risk constraint mechanism, it achieves effective buffering of new energy fluctuations and multi-entity collaborative optimization, significantly reducing wind and solar curtailment rates and the risk of load shedding in industrial parks.

[0004] The technical solution adopted in this invention is as follows: a probabilistic scheduling method for power grid and industrial park coordination based on Gaussian mixture model and conditional value risk, comprising the following steps:

[0005] Step 1: Collect and initialize multi-source basic data of the system in real time, and complete the calibration and modeling of full-dimensional operating parameters on both the grid side and the park side. The operating parameters include: the output range, ramping constraints and cost coefficients of thermal power units on the grid side, the prediction benchmark values ​​of multi-node loads on the park side and the power transmission limits of the grid-park interconnection line, the rated power, conversion efficiency and hydrogen storage tank capacity limits of electrolyzers and fuel cells in the hydrogen energy system; and simultaneously collect historical prediction error data of wind power and photovoltaic power on both the grid and park sides, and obtain the uncertainty modeling input sample set after normalization preprocessing.

[0006] Step 2: Generate a source-load scenario based on a Gaussian mixture model: The historical prediction error data is fitted using the expectation-maximization algorithm to construct a Gaussian mixture model reflecting the multi-peak distribution characteristics of new energy power output. Parameters such as the weights, mean, and covariance of each Gaussian component in the Gaussian mixture model are identified, and a Monte Carlo sampling method is used to generate... A representative output scenario is used to transform the continuous probability distribution into a discrete set of optimized scenarios, providing probabilistic input support for collaborative scheduling; and capacity boundary correction is performed on the output of each site scenario to make the output between 0 and the corresponding installed capacity.

[0007] Step 3: Construct a risk management and collaborative scheduling model that takes into account conditional risk value. Based on the aforementioned optimization scenario set, establish a joint constraint system between the power grid and the industrial park. First, construct physical constraints covering the power grid's thermal power output, dynamic hydrogen energy storage, power grid-industrial park interaction, and industrial park balance. Second, based on conditional risk value theory, quantify the systemic risks under extreme and severe scenarios. By introducing risk auxiliary variables and linearization, establish probabilistic constraints on insufficient reserve adjustments and industrial park load shedding risks to ensure that, at the set confidence level, the system can cope with the fluctuations in new energy sources by reserving sufficient flexible resources. Among these constraints, only industrial park load shedding and insufficient reserve adjustments by the power grid are included in the conditional risk value probabilistic constraints, while wind and solar curtailment losses are only included as economic penalties in the optimization objectives.

[0008] Step 4: Comprehensive Cost Target Optimization and Scheduling Decision Solving: With the goal of minimizing the total expected operating cost of the system, the objective function integrates the costs of thermal power generation, hydrogen energy operation, grid-park interaction transactions, reserve capacity reservation, and expected penalties for wind and solar curtailment and load shedding under multiple scenarios. A mixed-integer linear programming algorithm is used to solve the model globally, outputting the start-up and shutdown plans and output plans of thermal power units within the scheduling cycle, the charging and discharging paths of the hydrogen energy system, and the interaction power schemes of the grid-park interconnection lines, achieving the optimal balance between economy and reliability.

[0009] Furthermore, in step one, the parameter calibration and data preprocessing steps are as follows:

[0010] 1) Grid-side thermal power unit operating parameter calibration: For the first grid-side... For a thermal power unit, its upper limit of active power output, lower limit of active power output, uphill rate, downhill rate, start-stop status, and power generation cost coefficient are collected and initialized; in the subsequent objective function construction, the grid-side... Taiwan thermal power units Power generation cost function at time satisfy:

[0011]

[0012] In the formula, For the power grid side Number of output segments for a thermal power unit; For the first The linear cost slope of each output segment; For the power grid side Taiwan thermal power units The moment fell into the first The output variables of each segment; For unit start-up costs; These are the unit startup status variables;

[0013] 2) Grid-Park Link Transmission Parameter Calibration: Collect and initialize the power transmission limit of the grid-park link between the power grid and the park. Interaction power at time satisfy:

[0014]

[0015] In the formula, for The power exchange between the grid and the park at any given time; a positive value indicates that the park purchases electricity from the grid, and a negative value indicates that the park sells electricity to the grid. This refers to the maximum permissible transmission power of the network-to-campus link; superscript This indicates the power interaction between the power grid and the industrial park;

[0016] 3) Calibration of equipment parameters and conversion efficiency of the hydrogen energy system on the park side: For the hydrogen energy system on the park side, collect and initialize the maximum input power of the electrolyzer, the maximum output power of the fuel cell, the upper limit of the hydrogen storage tank capacity, the lower limit of the hydrogen storage tank capacity, the electrolysis efficiency of the electrolyzer, and the power generation efficiency of the fuel cell; wherein, the parameters and conversion efficiency of the hydrogen energy system on the park side are as follows: Each electrolytic cell is in The energy transformation relationship at any given moment satisfies:

[0017]

[0018] The park side One fuel cell The energy transformation relationship at any given moment satisfies:

[0019]

[0020] In the formula, For the side of the park Each electrolytic cell is in The electrical power consumed at all times; For the side of the park Each electrolytic cell is in Constantly input the hydrogen energy power into the hydrogen storage system; For the side of the park Electrolysis efficiency of each electrolytic cell; For the side of the park One fuel cell The hydrogen energy consumed at all times; For the side of the park One fuel cell The electrical power output at any given time; For the side of the park The power generation efficiency of a fuel cell;

[0021] 4) Construct a multi-node load parameter and total load model for the park side: This model is designed for the first... Each load node is used to collect and initialize its baseline load power, load node weighting coefficient, load time distribution characteristic coefficient, and park-side load fluctuation term; the park-side load is... Total load demand power at any given time satisfy:

[0022]

[0023] In the formula, For the park side Total load demand power at any given time; This represents the total number of load nodes on the park side. Number the load nodes on the park side; For the side of the park The weighting coefficient of each load node; For the side of the park Each load node The baseline load power at any given time; For the side of the park Each load node The load time distribution characteristic coefficient at any given time; For the park side Load fluctuation term at any given time; superscript Indicates the park side;

[0024] 5) Collect historical predicted output data and error data of wind and solar power on the grid side and in the park side and perform normalization preprocessing: Collect historical actual output of wind power, historical predicted output of wind power, historical actual output of photovoltaic power, historical predicted output of photovoltaic power, historical actual output of wind power, historical predicted output of wind power, historical actual output of photovoltaic power, and historical predicted output of photovoltaic power on the park side, and construct the prediction error of wind power, photovoltaic power, wind power, and photovoltaic power on the grid side respectively.

[0025] Among them, the first power grid side A wind farm station Wind power prediction error at any time satisfy:

[0026]

[0027] The first power grid side A photovoltaic power station Photovoltaic prediction error at time satisfy:

[0028]

[0029] The park side A wind farm station Wind power prediction error at any time satisfy:

[0030]

[0031] The park side A photovoltaic power station Photovoltaic prediction error at time satisfy:

[0032]

[0033] In the formula, For the power grid side A wind farm station Actual output at any given moment; For the power grid side A wind farm station Predicting output at any given moment; For the power grid side A photovoltaic power station Actual output at any given moment; For the power grid side A photovoltaic power station Predicting output at any given moment; For the side of the park A wind farm station Actual output at any given moment; For the side of the park A wind farm station Predicting output at any given moment; For the side of the park A photovoltaic power station Actual output at any given moment; For the side of the park A photovoltaic power station Predicted output at any given moment; superscript Indicates wind power, superscript Indicates photovoltaic;

[0034] The prediction errors of wind power on the grid side, photovoltaic power on the grid side, wind power on the park side, and photovoltaic power on the park side are normalized respectively to obtain the sample sets of wind and solar prediction errors on the grid side and the sample sets of wind and solar prediction errors on the park side for fitting the Gaussian mixture model.

[0035] Furthermore, in step two, the steps for constructing typical power output scenarios based on Gaussian mixture models are as follows:

[0036] 1) Constructing a Gaussian mixture model based on preprocessed wind and solar prediction error samples: Based on the grid-side wind power prediction error samples, grid-side solar prediction error samples, park-side wind power prediction error samples, and park-side solar prediction error samples obtained in step one, construct wind and solar prediction error sample vectors. satisfy:

[0037]

[0038] In the formula, For the first A vector of historical prediction error samples; For the power grid side A wind farm station Wind power prediction error at any given time; For the power grid side A photovoltaic power station Photovoltaic prediction error at any given time; For the side of the park A wind farm station Wind power prediction error at any given time; For the side of the park A photovoltaic power station Photovoltaic prediction error at any given time;

[0039] Based on the wind power and photovoltaic prediction error sample vector , construct by The probability density function of a Gaussian mixture model composed of Gaussian components is... satisfy:

[0040]

[0041]

[0042] In the formula, The number of Gaussian components; Number the Gaussian components; No. The weighting coefficients of each Gaussian component; For the first The multidimensional normal distribution corresponding to each Gaussian component; For the first The mean vector of Gaussian components; For the first The covariance matrix of Gaussian components;

[0043] 2) Solving for Gaussian mixture model parameters using the expectation-maximization algorithm: This involves calculating the weight coefficients in the Gaussian mixture model. Mean vector and covariance matrix Perform iterative solution; in the desired step, calculate the th... The historical prediction error sample belongs to the first The posterior probability of each Gaussian component It satisfies:

[0044]

[0045] In the formula, For the first The historical prediction error sample belongs to the first The posterior probability of each Gaussian component;

[0046] In the maximum step, according to the posterior probability Update the Gaussian mixture model parameters. The updated parameters satisfy the following:

[0047]

[0048]

[0049]

[0050]

[0051] In the formula, This represents the total number of historical prediction error samples. To belong to the The equivalent number of Gaussian components; by alternately executing the expectation step and the maximization step until the log-likelihood function converges, a Gaussian mixture model for sampling prediction errors of wind power and photovoltaic power is obtained.

[0052] 3) Constructing the day-ahead forecast baseline output curves for wind and solar power within the dispatch period: Assume the dispatch period includes For each scheduling period, the day-ahead forecast baseline output curves for grid-side wind power, grid-side photovoltaic power, park-side wind power, and park-side photovoltaic power are constructed respectively;

[0053] Among them, the first on the power grid side or the park side A wind farm station Day-ahead forecast baseline output satisfy:

[0054]

[0055] The first power grid side or the first park side A photovoltaic power station Day-ahead forecast baseline output satisfy:

[0056]

[0057] In the formula, As the identifier of the station's ownership, when Time indicates the grid side, when The time indicates the side of the park; For the subject of ownership Next The installed capacity of each wind farm station; For the subject of ownership Next The installed capacity of each photovoltaic power station; This serves as the benchmark coefficient for wind power output. This refers to the wind power output fluctuation coefficient. For wind power output phase correction parameters; This serves as the baseline coefficient for photovoltaic power output. This corresponds to the time period of sunrise; This corresponds to the time period of sunset;

[0058] 4) Generating discrete wind power and photovoltaic output scenarios using Monte Carlo sampling: Monte Carlo sampling is performed based on the converged Gaussian mixture model to generate... There are three independent wind power and solar power prediction error scenarios; among them, the first... The scene is Prediction error sample vector at time step satisfy:

[0059]

[0060] In the formula, Number the sampling scene; For the first The scene is The wind power and solar power prediction error sample vectors at each moment;

[0061] Based on the prediction error sample vector obtained from Monte Carlo sampling, the following data are generated for grid-side wind power, grid-side photovoltaic power, park-side wind power, and park-side photovoltaic power in the [missing data]. A scene, The actual output at any given moment, of which, the power grid side's first The scene of each wind farm station contributes to the effort satisfy:

[0062]

[0063] The first power grid side The scene output of a photovoltaic power station satisfy:

[0064]

[0065] Park side The scene of each wind farm station contributes to the effort satisfy:

[0066]

[0067] Park side The scene output of a photovoltaic power station satisfy:

[0068]

[0069] The generated grid-side wind power scenario output, grid-side photovoltaic scenario output, park-side wind power scenario output, and park-side photovoltaic scenario output are each subject to capacity boundary correction to ensure that the scenario output of each power station is not less than zero and does not exceed the corresponding installed capacity; after capacity boundary correction, the following is obtained: Each scenario contains [number] scenarios. A set of typical power output scenarios for wind and solar power on the grid side and in the industrial park side during a scheduling period, wherein the probability values ​​of each scenario satisfy:

[0070]

[0071] In the formula, For the first The probability of typical wind power and solar power output scenarios.

[0072] Furthermore, in step three, a collaborative scheduling model that takes into account conditional value risk control is constructed, comprising two parts:

[0073] The first part establishes deterministic physical topology and energy operation constraints for both the power grid and the industrial park, specifically including:

[0074] 1) Grid-side all-time power balance constraint: During any dispatch period, the output of grid-side thermal power units, the grid-side wind power day-ahead forecast reference output, and the grid-side photovoltaic day-ahead forecast reference output together meet the power demand of the grid-park interconnection line. The grid-side all-time power balance constraint satisfies:

[0075]

[0076] 2) All-time power balance constraint on the park side: During any scheduling period, the day-ahead forecasted baseline output of wind power on the park side, the day-ahead forecasted baseline output of photovoltaic power on the park side, the power exchanged between the grid and the park, and the net output of the hydrogen energy system on the park side together meet the total load demand on the park side. The all-time power balance constraint on the park side satisfies:

[0077]

[0078] 3) Time-series dynamic evolution equation of hydrogen storage capacity in the hydrogen energy system: The change in hydrogen storage capacity of the park-side hydrogen energy system between adjacent scheduling periods is jointly determined by the hydrogen power input to the hydrogen storage system from the electrolyzer and the hydrogen power consumed by the fuel cell. The time-series dynamic evolution equation of hydrogen storage capacity in the hydrogen energy system satisfies:

[0079]

[0080] In the formula, For the side of the park A hydrogen storage device in Hydrogen storage capacity at any given time; For the side of the park The amount of hydrogen stored in each hydrogen storage device during the previous scheduling period; The time interval between two adjacent scheduling periods;

[0081] 4) Limits on the power output of hydrogen energy equipment and the capacity of hydrogen storage tanks: The input power of the electrolyzer, the output power of the fuel cell, and the capacity of the hydrogen storage tank on the park side must meet the following requirements:

[0082]

[0083]

[0084]

[0085] In the formula, For the side of the park The maximum input power of each electrolytic cell; For the side of the park The maximum output power of each fuel cell; and The first on each side of the park The lower and upper limits of the capacity of each hydrogen storage device;

[0086] The second part constructs a multi-scenario probabilistic risk linear constraint system based on conditional value at risk theory. This system is designed to address risks such as load shedding in industrial parks, insufficient reserve requirements due to grid-side increases, and insufficient reserve requirements due to grid-side decreases. Specifically, it includes:

[0087] 1) Basic definition of conditional value at risk and method for linearizing nonlinear risk terms: Let the first... A scene, The risk loss function corresponding to time is Among them, risk types This includes risks such as load shedding in industrial parks, insufficient reserve requirements on the grid side, and insufficient reserve requirements on the grid side; a conditional value of risk auxiliary variable is introduced. and excess risk loss variables The positive part of the function term where the risk loss exceeds the threshold of the conditional value of risk is transformed into a linear constraint, satisfying:

[0088]

[0089]

[0090] The type of risk exist Conditional Value at Time satisfy:

[0091]

[0092] In the formula, This is the tail risk probability parameter;

[0093] 2) Definition of loss function for the park side-shear load scenario: For the first In one scenario, when the actual output of wind power, photovoltaic power, grid-park interconnection power, and net output of the hydrogen energy system on the park side cannot meet the total load demand on the park side, a load shedding variable on the park side is introduced. The loss function for the park side-cut load scenario satisfy:

[0094]

[0095]

[0096]

[0097] In the formula, For the side of the park The load node at the ... A scene, The load shedding power at any given moment; For the first A scene, Loss function for the park side-cut load scenario at any given time;

[0098] 3) Linear constraint on conditional value of risk corresponding to tail risk of side-shear load in the park: The loss function for the side-shear load scenario in the park... Substituting the conditional value-at-risk linearization constraint, we obtain the conditional value-at-risk linearization constraint corresponding to the tail risk of the park's side-shear load, which satisfies:

[0099]

[0100]

[0101]

[0102] In the formula, As an auxiliary variable for the risk value of side-shear load in the park; For the excess loss due to side-shear load risk in the industrial park; The threshold for controlling the risk of side-shear load in the park;

[0103] 4) Risk constraints on the risk value of grid-side adjustments to reserve shortage loss function and associated conditions: [Regarding the first...] In one scenario, when the actual output of wind and solar power on both the grid and industrial park sides is lower than the day-ahead forecast baseline output, the system generates an upward adjustment of reserve demand. This upward adjustment of reserve demand variable... satisfy:

[0104]

[0105]

[0106] Upgrade the reserve shortage loss function on the grid side satisfy:

[0107]

[0108]

[0109] The conditional value-at-risk linear constraint corresponding to the insufficient reserve on the grid side is satisfied as follows:

[0110]

[0111]

[0112]

[0113] Regarding the first In one scenario, when the actual output of wind and solar power on both the grid and industrial park sides exceeds the day-ahead forecast baseline output, the system generates a downward adjustment of reserve demand. This downward adjustment of reserve demand variable... satisfy:

[0114]

[0115]

[0116] The power grid side lowers the reserve shortage loss function satisfy:

[0117]

[0118]

[0119] The conditional value-at-risk linear constraint corresponding to the grid-side reduction in reserve requirements satisfies:

[0120]

[0121]

[0122]

[0123] In the formula, For the first A scene, Adjusting the reserve requirement variable at any time; For the first A scene, Adjusting the reserve requirement variable at any time; For the first A scene, The grid side adjusts the reserve shortage loss function at any given time. For the first A scene, The grid side adjusts the reserve shortage loss function at a given time. For the power grid side Taiwan thermal power units Reserved standby capacity at all times; For the power grid side Taiwan thermal power units Reserved standby capacity at all times; and These are auxiliary variables for adjusting the value-at-risk (VAT) of insufficient reserves on the grid side and adjusting the VAT of insufficient reserves on the grid side, respectively. and These represent the upward adjustment of the excess variable for risk loss due to insufficient reserves on the grid side and the downward adjustment of the excess variable for risk loss due to insufficient reserves on the grid side. and These represent raising the reserve shortage risk control threshold on the grid side and lowering the reserve shortage risk control threshold on the grid side, respectively.

[0124] A multi-scenario probabilistic risk linear constraint system is formed by the linear constraints of conditional risk value corresponding to the tail risk of load shedding in the park, the linear constraints of conditional risk value corresponding to insufficient reserve on the grid side, and the linear constraints of conditional risk value corresponding to insufficient reserve on the grid side. Among them, the power of wind and solar curtailment is not used as a conditional risk value risk loss variable, but is included in the system's comprehensive economic objective function as a multi-scenario expected penalty cost item in step four.

[0125] Furthermore, in step four, a comprehensive cost objective function is constructed and optimized, including constructing a power grid-side specific cost term, a park-side specific cost term, a full-scenario expected penalty cost term, and the output of the model solution and scheduling scheme, as detailed below:

[0126] 1) Construct grid-side dedicated cost items: The grid-side dedicated cost items include the fuel cost of grid-side thermal power units, the start-up and shutdown costs of the units, and the reserved costs for increasing and decreasing reserve capacity; wherein, the grid-side dedicated cost items satisfy:

[0127]

[0128] In the formula, This is a cost item specific to the power grid side; For the power grid side The unit of Taiwan's thermal power plant increased its reserve capacity to reserve costs; For the power grid side The unit of Taiwan's thermal power plant reduced its reserve capacity to reserve costs;

[0129] 2) Constructing a dedicated cost item for the park side: The dedicated cost item for the park side includes the operation and maintenance loss cost of the hydrogen energy equipment on the park side and the power exchange transaction cost of the grid-park interconnection line; to maintain the linearity of the objective function, the power exchange of the grid-park interconnection line is decomposed into power purchased and power sold, satisfying:

[0130]

[0131]

[0132] The park-side specific cost items satisfy:

[0133]

[0134] In the formula, This is a cost item specific to the park. For the park side Power purchased from the power grid at all times via the grid-park interconnection line; For the park side The power sold to the grid at all times through the grid-park interconnection line; For the side of the park Unit operation and maintenance cost coefficient of each electrolytic cell; For the side of the park Unit operation and maintenance cost coefficient of a fuel cell; for The electricity purchase price at the industrial park; for Electricity price at the park's side;

[0135] 3) Constructing the full-scenario expected penalty cost term: The full-scenario expected penalty cost term calculates the expected penalty costs for all sampled scenarios generated in step two, including grid-side wind power curtailment, grid-side photovoltaic power curtailment, park-side wind power curtailment, park-side photovoltaic power curtailment, and park-side multi-node load shedding compensation; wherein, the grid-side and park-side wind and solar power curtailment power respectively satisfy:

[0136]

[0137]

[0138]

[0139]

[0140]

[0141] The expected penalty cost item in the whole scenario satisfy:

[0142]

[0143] In the formula, This represents the expected penalty cost across all scenarios. For the power grid side The first wind farm station A scene, Wind curtailment power at any given time; For the power grid side The first photovoltaic power station A scene, The power of light discarded at any given moment; For the side of the park The first wind farm station A scene, Wind curtailment power at any given time; For the side of the park The first photovoltaic power station A scene, The power of light discarded at any given moment; The unit cost coefficient for wind and solar power curtailment penalties; The unit load shedding compensation cost coefficient;

[0144] Among them, the curtailed wind and solar power includes four parts: grid-side wind power curtailment, grid-side solar power curtailment, park-side wind power curtailment, and park-side solar power curtailment, which respectively correspond to the over-limit curtailment losses of the actual output of wind and solar power on the grid side and the park side relative to the day-ahead predicted baseline output; the curtailed wind and solar power is included in the system's comprehensive economic objective function as the expected penalty cost item for the whole scenario, and is not included as the conditional risk value risk loss variable in step three.

[0145] 4) Model solution and scheduling scheme output: Output the grid-side specific cost item Park-side dedicated cost items And the expected penalty cost item for the entire scenario Integrated into the system's overall cost objective function It satisfies:

[0146]

[0147] The optimization objective is to minimize the overall system cost objective function, satisfying:

[0148]

[0149] The system comprehensive cost objective function The model is integrated with the deterministic physical constraints, multi-scenario conditional risk value probability risk linear constraints, grid-park interconnection power purchase and sale power split constraints, and wind and solar power curtailment non-negativity constraints constructed in step three to form a mixed-integer linear programming model. The mixed-integer linear programming solver is called to perform global optimization and output a grid-park collaborative probabilistic scheduling scheme that meets the requirements of economy and reliability.

[0150] The proposed power grid and park collaborative probabilistic dispatch scheme includes upper-level power grid dispatch results, park hydrogen energy dispatch results, grid-park interactive dispatch results, and scenario risk handling results.

[0151] The upper-level power grid dispatch results include the start-up and shutdown status of thermal power units on the power grid side, the segmented output of thermal power units, and the increase or decrease of reserve capacity.

[0152] The hydrogen energy dispatch results of the park include the input power of the electrolyzer on the park side, the output power of the fuel cell, and the hydrogen storage capacity of the hydrogen storage equipment.

[0153] The grid-park interactive scheduling results include the grid-park interconnection line interactive power, the park-side power purchase power, and the park-side power sales power.

[0154] The scenario risk management results include grid-side wind power curtailment, grid-side photovoltaic power curtailment, park-side wind power curtailment, park-side photovoltaic power curtailment, park-side load shedding power, and the conditional risk value indicators corresponding to park-side load shedding risk, grid-side upward reserve shortage risk, and grid-side downward reserve shortage risk.

[0155] This invention also provides a power grid and park collaborative probabilistic scheduling system based on Gaussian mixture model and conditional value risk, used to implement the power grid and park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described above, including a data acquisition and initialization module, a source-load scenario generation module, a collaborative scheduling modeling module, and an optimization solution decision module;

[0156] The data acquisition and initialization module is used to collect and initialize multi-source basic data of the system, operating parameters of the grid side and the park side in real time, and synchronously collect historical prediction error data of wind power and photovoltaic power on both sides of the grid and the park and complete normalization preprocessing to provide input sources for uncertainty modeling. The operating parameters include the output range, ramping constraints and cost coefficients of thermal power units on the grid side, the benchmark value of multi-node load prediction on the park side, the power transmission limit of grid-park interconnection line, and the rated power, conversion efficiency and hydrogen storage tank capacity limit of the hydrogen energy system electrolyzer and fuel cell.

[0157] The source load scene generation module is connected to the data acquisition initialization module. It is used to construct a Gaussian mixture model based on the expectation-maximization algorithm to fit the historical prediction error data of wind and solar power, complete the identification of model weights, mean and covariance parameters, and use Monte Carlo sampling to generate multiple sets of discrete representative wind and solar power output scene sets, transforming the continuous probability distribution into discrete scenes that can be used for optimization calculation.

[0158] The collaborative scheduling modeling module, connected to the source-load scenario generation module, is used to build a grid-park joint constraint system based on a discrete wind and solar power output scenario set. On the one hand, it constructs deterministic physical operation constraints that include thermal power output, dynamic hydrogen energy storage, grid-park power interaction, and park load balance. On the other hand, it quantifies the system risk of extreme scenarios based on conditional risk value theory, linearizes the risk loss function by introducing risk auxiliary variables, and establishes probabilistic constraints for three types of tail risks: park load shedding, grid up-and-down adjustment of insufficient reserves, etc. Under a pre-set confidence level, it ensures that the system's flexible resources can mitigate the impact of new energy fluctuations.

[0159] The optimization solution decision module, connected to the collaborative scheduling modeling module, is used to construct a comprehensive cost function with the objective of minimizing the total expected operating cost of the system. The objective function integrates the cost of thermal power generation, hydrogen energy operation, grid-park interaction transaction costs, reserve costs, and expected penalties for wind and solar curtailment and load shedding in multiple scenarios. The mixed-integer linear programming solver is invoked to solve the overall model globally, outputting the start-up and shutdown and output plans of thermal power units within the scheduling cycle, the hydrogen charging and discharging paths of the hydrogen energy system, and the collaborative scheduling scheme for interactive power of grid-park interconnection lines, so as to achieve the optimal balance between system operating economy and power supply reliability.

[0160] The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a power grid and park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described above.

[0161] The present invention also provides a computer program product, which, when executed by a processor, implements a power grid and park collaborative probabilistic scheduling method based on a Gaussian mixture model and conditional value risk as described above.

[0162] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a power grid and park collaborative probabilistic scheduling method based on a Gaussian mixture model and conditional value risk as described above.

[0163] Advantages and benefits of this invention: This invention utilizes a Gaussian mixture model (GMM) to accurately characterize the multi-peak and asymmetric error features of renewable energy output. Compared to the traditional single normal distribution model, it can more realistically restore the stochastic attributes of wind and solar resources. Combined with the Conditional Value at Risk (CVaR) theory, it quantifies potential losses under extreme scenarios in scheduling decisions, effectively balancing the system's operational economy and power supply reliability, and significantly reducing wind and solar curtailment rates and the risk of load shedding in industrial parks. This invention not only enhances the self-balancing ability of uncertainties within industrial parks and reduces reliance on the expensive reserve capacity of the main power grid, but also achieves long-term regulation through cross-period energy time-shifting. This invention fully explores the mutual support potential of distributed resources, providing a scientific and highly economical solution for the operation of low-carbon microgrids with a high proportion of renewable energy access, while also offering new ideas for the safe and economical operation of high-proportion renewable energy power systems. Attached Figure Description

[0164] Figure 1 This is a flowchart of the present invention;

[0165] Figure 2 The image shows the wind power and photovoltaic output scenarios fitted using the method of this invention.

[0166] Figure 3 A diagram illustrating the grid-park renewable energy consumption situation using the method of this invention. Detailed Implementation

[0167] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0168] Example 1

[0169] like Figure 1 As shown, a probabilistic scheduling method for power grid and industrial park coordination based on Gaussian mixture model and conditional value risk includes the following steps:

[0170] Step 1: Real-time acquisition and initialization of multi-source basic data for the system, completing the calibration and modeling of full-dimensional operating parameters on both the grid side and the industrial park side. These operating parameters include: the output range, ramp-up constraints, and cost coefficients of thermal power units on the grid side; the predicted baseline values ​​of multi-node loads on the industrial park side and the power transmission limits of the grid-industrial park interconnection lines; the rated power, conversion efficiency, and hydrogen storage tank capacity limits of the electrolyzers and fuel cells in the hydrogen energy system; and the simultaneous acquisition of historical prediction error data for wind and solar power on both the grid and industrial park sides, followed by normalization preprocessing to obtain uncertainty modeling input samples. The parameter calibration and data preprocessing steps are as follows:

[0171] 1) Grid-side thermal power unit operating parameter calibration: For the first grid-side... For a thermal power unit, its upper limit of active power output, lower limit of active power output, uphill rate, downhill rate, start-stop status, and power generation cost coefficient are collected and initialized; in the subsequent objective function construction, the grid-side... Taiwan thermal power units Power generation cost function at time satisfy:

[0172]

[0173] In the formula, For the power grid side Number of output segments for a thermal power unit; For the first The linear cost slope of each output segment; For the power grid side Taiwan thermal power units The moment fell into the first The output variables of each segment; For unit start-up costs; These are the unit startup status variables;

[0174] 2) Grid-Park Link Transmission Parameter Calibration: Collect and initialize the power transmission limit of the grid-park link between the power grid and the park. Interaction power at time satisfy:

[0175]

[0176] In the formula, for The power exchange between the grid and the park at any given time; a positive value indicates that the park purchases electricity from the grid, and a negative value indicates that the park sells electricity to the grid. This refers to the maximum permissible transmission power of the network-to-campus link; superscript This indicates the power interaction between the power grid and the industrial park;

[0177] 3) Calibration of equipment parameters and conversion efficiency of the hydrogen energy system on the park side: For the hydrogen energy system on the park side, collect and initialize the maximum input power of the electrolyzer, the maximum output power of the fuel cell, the upper limit of the hydrogen storage tank capacity, the lower limit of the hydrogen storage tank capacity, the electrolysis efficiency of the electrolyzer, and the power generation efficiency of the fuel cell; wherein, the parameters and conversion efficiency of the hydrogen energy system on the park side are as follows: Each electrolytic cell is in The energy transformation relationship at any given moment satisfies:

[0178]

[0179] The park side One fuel cell The energy transformation relationship at any given moment satisfies:

[0180]

[0181] In the formula, For the side of the park Each electrolytic cell is in The electrical power consumed at all times; For the side of the park Each electrolytic cell is in Constantly input the hydrogen energy power into the hydrogen storage system; For the side of the park Electrolysis efficiency of each electrolytic cell; For the side of the park One fuel cell The hydrogen energy consumed at all times; For the side of the park One fuel cell The electrical power output at any given time; For the side of the park The power generation efficiency of a fuel cell;

[0182] 4) Construct a multi-node load parameter and total load model for the park side: This model is designed for the first... Each load node is used to collect and initialize its baseline load power, load node weighting coefficient, load time distribution characteristic coefficient, and park-side load fluctuation term; the park-side load is... Total load demand power at any given time satisfy:

[0183]

[0184] In the formula, For the park side Total load power demand at any given time; This represents the total number of load nodes on the park side. Number the load nodes on the park side; For the side of the park The weighting coefficient of each load node; For the side of the park Each load node The baseline load power at any given time; For the side of the park Each load node The load time distribution characteristic coefficient at any given time; For the park side Load fluctuation term at any given time; superscript Indicates the park side;

[0185] 5) Collect historical predicted output data and error data of wind and solar power on the grid side and in the park side and perform normalization preprocessing: Collect historical actual output of wind power, historical predicted output of wind power, historical actual output of photovoltaic power, historical predicted output of photovoltaic power, historical actual output of wind power, historical predicted output of wind power, historical actual output of photovoltaic power, and historical predicted output of photovoltaic power on the park side, and construct the prediction error of wind power, photovoltaic power, wind power, and photovoltaic power on the grid side respectively.

[0186] Among them, the first power grid side A wind farm station Wind power prediction error at any time satisfy:

[0187]

[0188] The first power grid side A photovoltaic power station Photovoltaic prediction error at time satisfy:

[0189]

[0190] The park side A wind farm station Wind power prediction error at any time satisfy:

[0191]

[0192] The park side A photovoltaic power station Photovoltaic prediction error at time satisfy:

[0193]

[0194] In the formula, For the power grid side A wind farm station Actual output at any given moment; For the power grid side A wind farm station Predicting output at any given moment; For the power grid side A photovoltaic power station Actual output at any given moment; For the power grid side A photovoltaic power station Predicting output at any given moment; For the side of the park A wind farm station Actual output at any given moment; For the side of the park A wind farm station Predicting output at any given moment; For the side of the park A photovoltaic power station Actual output at any given moment; For the side of the park A photovoltaic power station Predicted output at any given moment; superscript Indicates wind power, superscript Indicates photovoltaic;

[0195] The prediction errors of wind power on the grid side, photovoltaic power on the grid side, wind power on the park side, and photovoltaic power on the park side are normalized to obtain the sample sets of wind and solar prediction errors on the grid side and the sample sets of wind and solar prediction errors on the park side for fitting the Gaussian mixture model.

[0196] Step 2: Generate a source-load scenario based on a Gaussian mixture model: The historical prediction error data is fitted using the expectation-maximization algorithm to construct a Gaussian mixture model reflecting the multi-peak distribution characteristics of new energy power output. Parameters such as the weights, mean, and covariance of each Gaussian component in the Gaussian mixture model are identified, and a Monte Carlo sampling method is used to generate... A representative output scenario is used to transform the continuous probability distribution into a discrete set of optimization scenarios, providing probabilistic input support for collaborative scheduling. The steps for constructing typical source-load output scenarios based on a Gaussian mixture model are as follows:

[0197] 1) Constructing a Gaussian mixture model based on preprocessed wind and solar prediction error samples: Based on the grid-side wind power prediction error samples, grid-side solar prediction error samples, park-side wind power prediction error samples, and park-side solar prediction error samples obtained in step one, construct wind and solar prediction error sample vectors. satisfy:

[0198]

[0199] In the formula, For the first A vector of historical prediction error samples; For the power grid side A wind farm station Wind power forecast error at any given time; For the power grid side A photovoltaic power station Photovoltaic prediction error at any given time; For the side of the park A wind farm station Wind power forecast error at any given time; For the side of the park A photovoltaic power station Photovoltaic prediction error at any given time;

[0200] Based on the wind power and photovoltaic prediction error sample vector , construct by The probability density function of a Gaussian mixture model composed of Gaussian components is... satisfy:

[0201]

[0202]

[0203] In the formula, The number of Gaussian components; Number the Gaussian components; No. The weighting coefficients of each Gaussian component; For the first The multidimensional normal distribution corresponding to each Gaussian component; For the first The mean vector of Gaussian components; For the first The covariance matrix of Gaussian components;

[0204] 2) Solving for Gaussian mixture model parameters using the expectation-maximization algorithm: This involves calculating the weight coefficients in the Gaussian mixture model. Mean vector and covariance matrix Perform iterative solution; in the desired step, calculate the th... The historical prediction error sample belongs to the first... The posterior probability of each Gaussian component It satisfies:

[0205]

[0206] In the formula, For the first The historical prediction error sample belongs to the first... The posterior probability of each Gaussian component;

[0207] In the maximum step, according to the posterior probability Update the Gaussian mixture model parameters. The updated parameters satisfy the following:

[0208]

[0209]

[0210]

[0211]

[0212] In the formula, This represents the total number of historical prediction error samples. To belong to the The equivalent number of Gaussian components; by alternately executing the expectation step and the maximization step until the log-likelihood function converges, a Gaussian mixture model for sampling prediction errors of wind power and photovoltaic power is obtained.

[0213] 3) Constructing the day-ahead forecast baseline output curves for wind and solar power within the dispatch period: Assume the dispatch period includes For each scheduling period, the day-ahead forecast baseline output curves for grid-side wind power, grid-side photovoltaic power, park-side wind power, and park-side photovoltaic power are constructed respectively;

[0214] Among them, the first on the power grid side or the park side A wind farm station Day-ahead forecast baseline output satisfy:

[0215]

[0216] The first power grid side or the first park side A photovoltaic power station Day-ahead forecast baseline output satisfy:

[0217]

[0218] In the formula, As the identifier of the station's ownership, when Time indicates the grid side, when The time indicates the side of the park; For the subject of ownership Next The installed capacity of each wind farm station; For the subject of ownership Next The installed capacity of each photovoltaic power station; This serves as the benchmark coefficient for wind power output. This refers to the wind power output fluctuation coefficient. For wind power output phase correction parameters; This serves as the baseline coefficient for photovoltaic power output. This corresponds to the time period of sunrise; This corresponds to the time period of sunset;

[0219] 4) Generating discrete wind power and photovoltaic output scenarios using Monte Carlo sampling: Monte Carlo sampling is performed based on the converged Gaussian mixture model to generate... There are three independent wind power and solar power prediction error scenarios; among them, the first... The scene is Prediction error sample vector at time step satisfy:

[0220]

[0221] In the formula, Number the sampling scene; For the first The scene is The wind power and solar power prediction error sample vectors at each moment;

[0222] Based on the prediction error sample vector obtained from Monte Carlo sampling, the following data are generated for grid-side wind power, grid-side photovoltaic power, park-side wind power, and park-side photovoltaic power in the [missing data]. A scene, The actual output at any given moment, of which, the power grid side's first The scene of each wind farm station contributes to the effort satisfy:

[0223]

[0224] The first power grid side The scene output of a photovoltaic power station satisfy:

[0225]

[0226] Park side The scene of each wind farm station contributes to the effort satisfy:

[0227]

[0228] Park side The scene output of a photovoltaic power station satisfy:

[0229]

[0230] The generated grid-side wind power scenario output, grid-side photovoltaic scenario output, park-side wind power scenario output, and park-side photovoltaic scenario output are each subject to capacity boundary correction to ensure that the scenario output of each power station is not less than zero and does not exceed the corresponding installed capacity; after capacity boundary correction, the following is obtained: Each scenario contains [number] scenarios. A set of typical power output scenarios for wind and solar power on the grid side and in the industrial park side during a scheduling period, wherein the probability values ​​of each scenario satisfy:

[0231]

[0232] In the formula, For the first The probability of typical wind power and solar power output scenarios.

[0233] Step 3: Constructing a risk management and collaborative scheduling model considering conditional value of risk: Based on the aforementioned optimization scenario set, a joint constraint system for the power grid and industrial parks is established. First, physical constraints covering power grid thermal power output, dynamic hydrogen energy storage, power grid-park power interaction, and park balance are constructed. Second, based on conditional value of risk theory, systemic risks under extreme and severe scenarios are quantified. By introducing risk auxiliary variables and linearization processing, probabilistic constraints on insufficient reserve adjustments and park load shedding risks are established to ensure that, at the set confidence level, the system can cope with the fluctuations in new energy sources by reserving sufficient flexible resources. The collaborative scheduling model considering conditional value of risk control consists of two parts:

[0234] The first part establishes deterministic physical topology and energy operation constraints for both the power grid and the industrial park, specifically including:

[0235] 1) Grid-side all-time power balance constraint: During any dispatch period, the output of grid-side thermal power units, the grid-side wind power day-ahead forecast reference output, and the grid-side photovoltaic day-ahead forecast reference output together meet the power demand of the grid-park interconnection line. The grid-side all-time power balance constraint satisfies:

[0236]

[0237] 2) All-time power balance constraint on the park side: During any scheduling period, the day-ahead forecasted baseline output of wind power on the park side, the day-ahead forecasted baseline output of photovoltaic power on the park side, the power exchanged between the grid and the park, and the net output of the hydrogen energy system on the park side together meet the total load demand on the park side. The all-time power balance constraint on the park side satisfies:

[0238]

[0239] 3) Time-series dynamic evolution equation of hydrogen storage capacity in the hydrogen energy system: The change in hydrogen storage capacity of the park-side hydrogen energy system between adjacent scheduling periods is jointly determined by the hydrogen power input to the hydrogen storage system from the electrolyzer and the hydrogen power consumed by the fuel cell. The time-series dynamic evolution equation of hydrogen storage capacity in the hydrogen energy system satisfies:

[0240]

[0241] In the formula, For the side of the park A hydrogen storage device in Hydrogen storage capacity at any given time; For the side of the park The amount of hydrogen stored in each hydrogen storage device during the previous scheduling period; The time interval between two adjacent scheduling periods;

[0242] 4) Limits on the power output of hydrogen energy equipment and the capacity of hydrogen storage tanks: The input power of the electrolyzer, the output power of the fuel cell, and the capacity of the hydrogen storage tank on the park side must meet the following requirements:

[0243]

[0244]

[0245]

[0246] In the formula, For the side of the park The maximum input power of each electrolytic cell; For the side of the park The maximum output power of each fuel cell; and The first on each side of the park The lower and upper limits of the capacity of each hydrogen storage device;

[0247] The second part constructs a multi-scenario probabilistic risk linear constraint system based on conditional value at risk theory. This system is designed to address risks such as load shedding in industrial parks, insufficient reserve requirements due to grid-side increases, and insufficient reserve requirements due to grid-side decreases. Specifically, it includes:

[0248] 1) Basic definition of conditional value at risk and method for linearizing nonlinear risk terms: Let the first... A scene, The risk loss function corresponding to time is Among them, risk types This includes risks such as load shedding in industrial parks, insufficient reserve requirements on the grid side, and insufficient reserve requirements on the grid side; a conditional value of risk auxiliary variable is introduced. and excess risk loss variables The positive part of the function term where the risk loss exceeds the threshold of the conditional value of risk is transformed into a linear constraint, satisfying:

[0249]

[0250]

[0251] The type of risk exist Conditional Value at Time satisfy:

[0252]

[0253] In the formula, This is the tail risk probability parameter;

[0254] 2) Definition of loss function for the park side-shear load scenario: For the first In one scenario, when the actual output of wind power, photovoltaic power, grid-park interconnection power, and net output of the hydrogen energy system on the park side cannot meet the total load demand on the park side, a load shedding variable on the park side is introduced. The loss function for the park side-cut load scenario satisfy:

[0255]

[0256]

[0257]

[0258] In the formula, For the side of the park The load node at the ... A scene, The load shedding power at any given moment; For the first A scene, Loss function for the park side-cut load scenario at any given time;

[0259] 3) Linear constraint on conditional value of risk corresponding to tail risk of side-shear load in the park: The loss function for the side-shear load scenario in the park... Substituting the conditional value-at-risk linearization constraint, we obtain the conditional value-at-risk linearization constraint corresponding to the tail risk of the park's side-shear load, which satisfies:

[0260]

[0261]

[0262]

[0263] In the formula, As an auxiliary variable for the risk value of side-shear load in the park; For the excess loss due to side-shear load risk in the industrial park; The threshold for controlling the risk of side-shear load in the park;

[0264] 4) Risk constraints on the risk value of grid-side adjustments to reserve shortage loss function and associated conditions: [Regarding the first...] In one scenario, when the actual output of wind and solar power on both the grid and industrial park sides is lower than the day-ahead forecast baseline output, the system generates an upward adjustment of reserve demand. This upward adjustment of reserve demand variable... satisfy:

[0265]

[0266]

[0267] Upgrade the reserve shortage loss function on the grid side satisfy:

[0268]

[0269]

[0270] The conditional value-at-risk linear constraint corresponding to the insufficient reserve on the grid side is satisfied as follows:

[0271]

[0272]

[0273]

[0274] Regarding the first In one scenario, when the actual output of wind and solar power on both the grid and industrial park sides exceeds the day-ahead forecast baseline output, the system generates a downward adjustment of reserve demand. This downward adjustment of reserve demand variable... satisfy:

[0275]

[0276]

[0277] The power grid side lowers the reserve shortage loss function satisfy:

[0278]

[0279]

[0280] The conditional value-at-risk linear constraint corresponding to the grid-side reduction in reserve requirements satisfies:

[0281]

[0282]

[0283]

[0284] In the formula, For the first A scene, Adjusting the reserve requirement variable at any time; For the first A scene, Adjusting the reserve requirement variable at any time; For the first A scene, The grid side adjusts the reserve shortage loss function at any given time. For the first A scene, The grid side adjusts the reserve shortage loss function at a given time. For the power grid side Taiwan thermal power units Reserved standby capacity at all times; For the power grid side Taiwan thermal power units Reserved standby capacity at all times; and These are auxiliary variables for adjusting the value-at-risk (VAT) of insufficient reserves on the grid side and adjusting the VAT of insufficient reserves on the grid side, respectively. and These represent the upward adjustment of the excess variable for risk loss due to insufficient reserves on the grid side and the downward adjustment of the excess variable for risk loss due to insufficient reserves on the grid side. and These represent raising the reserve shortage risk control threshold on the grid side and lowering the reserve shortage risk control threshold on the grid side, respectively.

[0285] A multi-scenario probabilistic risk linear constraint system is formed by using the conditional risk value linear constraints corresponding to the tail risk of load shedding on the industrial park side, the conditional risk value linear constraints corresponding to insufficient reserve adjustments on the grid side, and the conditional risk value linear constraints corresponding to insufficient reserve adjustments on the grid side. In this system, wind and solar power curtailment is not considered as a conditional risk value loss variable, but rather as a multi-scenario expected penalty cost item included in the system's comprehensive economic objective function in step four. This embodiment incorporates three extreme losses—load shedding on the industrial park side, insufficient reserve adjustments on the grid side, and insufficient reserve adjustments on the grid side—into conditional value risk constraints for tail risk management. Wind and solar power curtailment losses are only included as multi-scenario expected penalties in the economic objective function and are not included as conditional value risk loss variables in the construction of linear risk constraints.

[0286] Step 4: Comprehensive Cost Target Optimization and Scheduling Decision Solving: The optimization objective is to minimize the total expected operating cost of the system. The objective function integrates the costs of thermal power generation, hydrogen energy operation, grid-park interaction transactions, reserve capacity, and expected penalties for wind and solar curtailment and load shedding in multiple scenarios. A mixed-integer linear programming algorithm is used to globally solve the model, outputting the start-up and shutdown plans and output schedules of thermal power units within the scheduling cycle, the charging and discharging paths of the hydrogen energy system, and the interaction power schemes of the grid-park interconnection lines, achieving an optimal balance between economy and reliability. Specifically, the comprehensive cost objective function is constructed and optimized, including constructing grid-side specific cost items, park-side specific cost items, expected penalty cost items for all scenarios, and the output of the model solution and scheduling scheme, as detailed below:

[0287] 1) Construct grid-side dedicated cost items: The grid-side dedicated cost items include the fuel cost of grid-side thermal power units, the start-up and shutdown costs of the units, and the reserved costs for increasing and decreasing reserve capacity; wherein, the grid-side dedicated cost items satisfy:

[0288]

[0289] In the formula, This is a cost item specific to the power grid side; For the power grid side The unit of Taiwan's thermal power plant increased its reserve capacity to reserve costs; For the power grid side The unit of Taiwan's thermal power plant reduced its reserve capacity to reserve costs;

[0290] 2) Constructing a dedicated cost item for the park side: The dedicated cost item for the park side includes the operation and maintenance loss cost of the hydrogen energy equipment on the park side and the power exchange transaction cost of the grid-park interconnection line; to maintain the linearity of the objective function, the power exchange of the grid-park interconnection line is decomposed into power purchased and power sold, satisfying:

[0291]

[0292]

[0293] The park-side specific cost items satisfy:

[0294]

[0295] In the formula, This is a cost item specific to the park. For the park side Power purchased from the power grid at all times via the grid-park interconnection line; For the park side The power sold to the grid at all times through the grid-park interconnection line; For the side of the park Unit operation and maintenance cost coefficient of each electrolytic cell; For the side of the park Unit operation and maintenance cost coefficient of a fuel cell; for The electricity purchase price at the industrial park; for Electricity price at the park's side;

[0296] 3) Constructing the full-scenario expected penalty cost term: The full-scenario expected penalty cost term calculates the expected penalty costs for all sampled scenarios generated in step two, including grid-side wind power curtailment, grid-side photovoltaic power curtailment, park-side wind power curtailment, park-side photovoltaic power curtailment, and park-side multi-node load shedding compensation; wherein, the wind and solar power curtailment power on the grid side and the park side respectively satisfy:

[0297]

[0298]

[0299]

[0300]

[0301]

[0302] The expected penalty cost item in the whole scenario satisfy:

[0303]

[0304] In the formula, This represents the expected penalty cost across all scenarios. For the power grid side The first wind farm station A scene, Wind curtailment power at any given time; For the power grid side The first photovoltaic power station A scene, The power of light discarded at any given moment; For the side of the park The first wind farm station A scene, Wind curtailment power at any given time; For the side of the park The first photovoltaic power station A scene, The power of light discarded at any given moment; The unit cost coefficient for wind and solar power curtailment penalties; The unit load shedding compensation cost coefficient;

[0305] Among them, the curtailed wind and solar power includes four parts: grid-side wind power curtailment, grid-side solar power curtailment, park-side wind power curtailment, and park-side solar power curtailment, which respectively correspond to the over-limit curtailment losses of the actual output of wind and solar power on the grid side and the park side relative to the day-ahead predicted baseline output; the curtailed wind and solar power is included in the system's comprehensive economic objective function as the expected penalty cost item for the whole scenario, and is not included as the conditional risk value risk loss variable in step three.

[0306] 4) Model solution and scheduling scheme output: Output the grid-side specific cost item Park-side dedicated cost items And the expected penalty cost item for the entire scenario Integrated into the system's overall cost objective function It satisfies:

[0307]

[0308] The optimization objective is to minimize the overall system cost objective function, satisfying:

[0309]

[0310] The system comprehensive cost objective function The model is integrated with the deterministic physical constraints, multi-scenario conditional risk value probability risk linear constraints, grid-park interconnection power purchase and sale power split constraints, and wind and solar power curtailment non-negativity constraints constructed in step three to form a mixed-integer linear programming model. The mixed-integer linear programming solver is called to perform global optimization and output a grid-park collaborative probabilistic scheduling scheme that meets the requirements of economy and reliability.

[0311] The proposed power grid and park collaborative probabilistic dispatch scheme includes upper-level power grid dispatch results, park hydrogen energy dispatch results, grid-park interactive dispatch results, and scenario risk handling results.

[0312] The upper-level power grid dispatch results include the start-up and shutdown status of thermal power units on the power grid side, the segmented output of thermal power units, and the increase or decrease of reserve capacity.

[0313] The hydrogen energy dispatch results of the park include the input power of the electrolyzer on the park side, the output power of the fuel cell, and the hydrogen storage capacity of the hydrogen storage equipment.

[0314] The grid-park interactive scheduling results include the grid-park interconnection line interactive power, the park-side power purchase power, and the park-side power sales power.

[0315] The scenario risk management results include grid-side wind power curtailment, grid-side photovoltaic power curtailment, park-side wind power curtailment, park-side photovoltaic power curtailment, park-side load shedding power, and the conditional risk value indicators corresponding to park-side load shedding risk, grid-side upward reserve shortage risk, and grid-side downward reserve shortage risk.

[0316] Example 2

[0317] like Figure 2 As shown in the figure, this diagram illustrates typical wind and solar power output scenarios generated based on the method of this invention. Wind power output fluctuates significantly throughout the day, exhibiting randomness and uncertainty; solar power output, on the other hand, is concentrated during the day, peaking at midday. The distribution of multiple scenario curves around the baseline curve demonstrates that this method can effectively characterize the prediction error of new energy power output and the changing characteristics of multiple scenarios.

[0318] like Figure 3 As shown in the figure, this graph illustrates the overall absorption of new energy sources over 24 hours. The results indicate that, under the scheduling method of this invention, the power grid and the park's new energy output are effectively utilized, with high-level absorption of wind and solar power, and extremely low levels of wind and solar curtailment. During the midday peak solar power output period, the system can fully absorb new energy through grid-park coordination, flexible load response, and hydrogen energy regulation, demonstrating that this method can improve the overall absorption level of the grid and park, and enhance its adaptability to wind and solar power fluctuations.

Claims

1. A probabilistic scheduling method for power grid and industrial park coordination based on Gaussian mixture model and conditional value risk, characterized in that, Includes the following steps: Step 1: Collect and initialize multi-source basic data of the system in real time, and complete the calibration and modeling of full-dimensional operating parameters on both the grid side and the park side. The operating parameters include: the output range, ramping constraints and cost coefficients of thermal power units on the grid side, the prediction benchmark values ​​of multi-node loads on the park side and the power transmission limits of the grid-park interconnection line, the rated power, conversion efficiency and hydrogen storage tank capacity limits of electrolyzers and fuel cells in the hydrogen energy system; and simultaneously collect historical prediction error data of wind power and photovoltaic power on both the grid and park sides, and obtain the uncertainty modeling input sample set after normalization preprocessing. Step 2: Generate a source-load scenario based on a Gaussian mixture model: The historical prediction error data is fitted using the expectation-maximization algorithm to construct a Gaussian mixture model reflecting the multi-peak distribution characteristics of new energy power output. Parameters such as the weights, mean, and covariance of each Gaussian component in the Gaussian mixture model are identified, and a Monte Carlo sampling method is used to generate... A representative output scenario is used to transform the continuous probability distribution into a discrete set of optimized scenarios, and the capacity boundary of each site scenario is corrected so that the output is between 0 and the corresponding installed capacity. Step 3: Construct a risk management and collaborative scheduling model that takes into account conditional risk value. Based on the aforementioned optimization scenario set, establish a joint constraint system between the power grid and the industrial park. First, construct physical constraints covering the power grid's thermal power output, dynamic hydrogen energy storage, power grid-industrial park power interaction, and industrial park balance. Second, based on conditional risk value theory, quantify the systemic risks under extreme and severe scenarios. By introducing risk auxiliary variables and linearization processing, establish probabilistic constraints regarding insufficient reserve adjustments and industrial park load shedding risks. Among these, only industrial park load shedding and insufficient reserve adjustments by the power grid are included in the conditional risk value probabilistic constraints, while wind and solar curtailment losses are only included as economic penalties in the optimization objectives. Step 4: Comprehensive Cost Target Optimization and Scheduling Decision Solving: With the goal of minimizing the total expected operating cost of the system, the objective function integrates the costs of thermal power generation, hydrogen energy operation, grid-park interaction transactions, reserve capacity reservation, and expected penalties for wind and solar curtailment and load shedding under multiple scenarios. A mixed-integer linear programming algorithm is used to solve the model globally, outputting the start-up and shutdown plans and output plans of thermal power units within the scheduling cycle, the charging and discharging paths of the hydrogen energy system, and the interaction power schemes of the grid-park interconnection lines, achieving the optimal balance between economy and reliability.

2. The power grid and industrial park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in claim 1, characterized in that: In step one, the parameter calibration and data preprocessing steps are as follows: 1) Grid-side thermal power unit operating parameter calibration: For the first grid-side... For a thermal power unit, its upper limit of active power output, lower limit of active power output, uphill rate, downhill rate, start-stop status, and power generation cost coefficient are collected and initialized; in the subsequent objective function construction, the grid-side... Taiwan thermal power units Power generation cost function at time satisfy: In the formula, For the power grid side Number of output segments for a thermal power unit; For the first The linear cost slope of each output segment; For the power grid side Taiwan thermal power units The moment fell into the first The output variables of each segment; For unit start-up costs; This refers to the unit startup status variables; 2) Grid-Park Link Transmission Parameter Calibration: Collect and initialize the power transmission limit of the grid-park link between the power grid and the park. Interaction power at time satisfy: In the formula, for The power exchange between the grid and the park at any given time; a positive value indicates that the park purchases electricity from the grid, and a negative value indicates that the park sells electricity to the grid. This refers to the maximum permissible transmission power of the network-to-campus link; superscript This indicates the power interaction between the power grid and the industrial park; 3) Calibration of equipment parameters and conversion efficiency of the hydrogen energy system on the park side: For the hydrogen energy system on the park side, collect and initialize the maximum input power of the electrolyzer, the maximum output power of the fuel cell, the upper limit of the hydrogen storage tank capacity, the lower limit of the hydrogen storage tank capacity, the electrolysis efficiency of the electrolyzer, and the power generation efficiency of the fuel cell; wherein, the parameters and conversion efficiency of the hydrogen energy system on the park side are as follows: Each electrolytic cell is in The energy transformation relationship at any given moment satisfies: The park side One fuel cell The energy transformation relationship at any given moment satisfies: In the formula, For the side of the park Each electrolytic cell is in The electrical power consumed at all times; For the side of the park Each electrolytic cell is in Constantly input the hydrogen energy power into the hydrogen storage system; For the side of the park Electrolysis efficiency of each electrolytic cell; For the side of the park One fuel cell The hydrogen energy consumed at all times; For the side of the park One fuel cell The electrical power output at any given time; For the side of the park The power generation efficiency of a fuel cell; 4) Construct a multi-node load parameter and total load model for the park side: This model is designed for the first... Each load node is used to collect and initialize its baseline load power, load node weighting coefficient, load time distribution characteristic coefficient, and park-side load fluctuation term; the park-side load is... Total load demand power at any given time satisfy: In the formula, For the park side Total load power demand at any given time; This represents the total number of load nodes on the park side. Number the load nodes on the park side; For the side of the park The weighting coefficient of each load node; For the side of the park Each load node The baseline load power at any given time; For the side of the park Each load node The load time distribution characteristic coefficient at any given time; For the park side Load fluctuation term at any given time; superscript Indicates the park side; 5) Collect historical predicted output data and error data of wind and solar power on the grid side and in the park side and perform normalization preprocessing: Collect historical actual output of wind power, historical predicted output of wind power, historical actual output of photovoltaic power, historical predicted output of photovoltaic power, historical actual output of wind power, historical predicted output of wind power, historical actual output of photovoltaic power, and historical predicted output of photovoltaic power on the park side, and construct the prediction error of wind power, photovoltaic power, wind power, and photovoltaic power on the grid side respectively. Among them, the first power grid side A wind farm station Wind power prediction error at any time satisfy: The first power grid side A photovoltaic power station Photovoltaic prediction error at time satisfy: The park side A wind farm station Wind power prediction error at any time satisfy: The park side A photovoltaic power station Photovoltaic prediction error at time satisfy: In the formula, For the power grid side A wind farm station Actual output at any given moment; For the power grid side A wind farm station Predicting output at any given moment; For the power grid side A photovoltaic power station Actual output at any given moment; For the power grid side A photovoltaic power station Predicting output at any given moment; For the side of the park A wind farm station Actual output at any given moment; For the side of the park A wind farm station Predicting output at any given moment; For the side of the park A photovoltaic power station Actual output at any given moment; For the side of the park A photovoltaic power station Predicted output at any given moment; superscript Indicates wind power, superscript Indicates photovoltaic; The prediction errors of wind power on the grid side, photovoltaic power on the grid side, wind power on the park side, and photovoltaic power on the park side are normalized to obtain the sample sets of wind and solar prediction errors on the grid side and the sample sets of wind and solar prediction errors on the park side for fitting the Gaussian mixture model.

3. The power grid and industrial park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in claim 2, characterized in that: In step two, the steps for constructing typical power output scenarios based on Gaussian mixture models are as follows: 1) Constructing a Gaussian mixture model based on preprocessed wind and solar prediction error samples: Based on the grid-side wind power prediction error samples, grid-side solar prediction error samples, park-side wind power prediction error samples, and park-side solar prediction error samples obtained in step one, construct wind and solar prediction error sample vectors. satisfy: In the formula, For the first A vector of historical prediction error samples; For the power grid side A wind farm station Wind power prediction error at any given time; For the power grid side A photovoltaic power station Photovoltaic prediction error at any given time; For the side of the park A wind farm station Wind power prediction error at any given time; For the side of the park A photovoltaic power station Photovoltaic prediction error at any given time; Based on the wind power and photovoltaic prediction error sample vector , construct by The probability density function of a Gaussian mixture model composed of Gaussian components is... satisfy: In the formula, The number of Gaussian components; Number the Gaussian components; No. The weighting coefficients of each Gaussian component; For the first The multidimensional normal distribution corresponding to each Gaussian component; For the first The mean vector of Gaussian components; For the first The covariance matrix of Gaussian components; 2) Solving for Gaussian mixture model parameters using the expectation-maximization algorithm: This involves calculating the weight coefficients in the Gaussian mixture model. Mean vector and covariance matrix Perform iterative solution; in the desired step, calculate the th... The historical prediction error sample belongs to the first... The posterior probability of each Gaussian component It satisfies: In the formula, For the first The historical prediction error sample belongs to the first... The posterior probability of each Gaussian component; In the maximum step, according to the posterior probability Update the Gaussian mixture model parameters. The updated parameters satisfy the following: In the formula, This represents the total number of historical prediction error samples. To belong to the The equivalent number of Gaussian components; by alternately executing the expectation step and the maximization step until the log-likelihood function converges, a Gaussian mixture model for sampling prediction errors of wind power and photovoltaic power is obtained. 3) Constructing the day-ahead forecast baseline output curves for wind and solar power within the dispatch period: Assume the dispatch period includes For each scheduling period, the day-ahead forecast baseline output curves for grid-side wind power, grid-side photovoltaic power, park-side wind power, and park-side photovoltaic power are constructed respectively; Among them, the first on the power grid side or the park side A wind farm station Day-ahead forecast baseline output satisfy: The first power grid side or the first park side A photovoltaic power station Day-ahead forecast baseline output satisfy: In the formula, As the identifier of the station's ownership, when Time indicates the grid side, when The time indicates the side of the park; For the subject of ownership Next The installed capacity of each wind farm station; For the subject of ownership Next The installed capacity of each photovoltaic power station; This serves as the benchmark coefficient for wind power output. This refers to the wind power output fluctuation coefficient. For wind power output phase correction parameters; This serves as the baseline coefficient for photovoltaic power output. This corresponds to the time period of sunrise; This corresponds to the time period of sunset; 4) Generating discrete wind power and photovoltaic output scenarios using Monte Carlo sampling: Monte Carlo sampling is performed based on the converged Gaussian mixture model to generate... There are three independent wind power and solar power prediction error scenarios; among them, the first... The scene is Prediction error sample vector at time step satisfy: In the formula, Number the sampling scene; For the first The scene is The wind power and solar power prediction error sample vectors at each moment; Based on the prediction error sample vector obtained from Monte Carlo sampling, the following data are generated for grid-side wind power, grid-side photovoltaic power, park-side wind power, and park-side photovoltaic power in the [missing data]. A scene, The actual output at any given moment, of which, the power grid side's first The scene of each wind farm station contributes to the effort satisfy: The first power grid side The scene output of a photovoltaic power station satisfy: Park side The scene of each wind farm station contributes to the effort satisfy: Park side The scene output of a photovoltaic power station satisfy: The generated grid-side wind power scenario output, grid-side photovoltaic scenario output, park-side wind power scenario output, and park-side photovoltaic scenario output are each subject to capacity boundary correction to ensure that the scenario output of each power station is not less than zero and does not exceed the corresponding installed capacity; after capacity boundary correction, the following is obtained: Each scenario contains A set of typical power output scenarios for wind and solar power on the grid side and in the industrial park side during a scheduling period, wherein the probability values ​​of each scenario satisfy: In the formula, For the first The probability of typical wind power and solar power output scenarios.

4. The power grid and industrial park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in claim 3, characterized in that: Step three involves constructing a collaborative scheduling model that takes into account conditional value risk control, comprising two parts: The first part establishes deterministic physical topology and energy operation constraints for both the power grid and the industrial park, specifically including: 1) Grid-side all-time power balance constraint: During any dispatch period, the output of grid-side thermal power units, the grid-side wind power day-ahead forecast reference output, and the grid-side photovoltaic day-ahead forecast reference output together meet the power demand of the grid-park interconnection line. The grid-side all-time power balance constraint satisfies: 2) All-time power balance constraint on the park side: During any scheduling period, the day-ahead forecasted baseline output of wind power on the park side, the day-ahead forecasted baseline output of photovoltaic power on the park side, the power exchanged between the grid and the park, and the net output of the hydrogen energy system on the park side together meet the total load demand on the park side. The all-time power balance constraint on the park side satisfies: 3) Time-series dynamic evolution equation of hydrogen storage capacity in the hydrogen energy system: The change in hydrogen storage capacity of the park-side hydrogen energy system between adjacent scheduling periods is jointly determined by the hydrogen power input to the hydrogen storage system from the electrolyzer and the hydrogen power consumed by the fuel cell. The time-series dynamic evolution equation of hydrogen storage capacity in the hydrogen energy system satisfies: In the formula, For the side of the park A hydrogen storage device in Hydrogen storage capacity at any given time; For the side of the park The amount of hydrogen stored in each hydrogen storage device during the previous scheduling period; The time interval between two adjacent scheduling periods; 4) Limits on the power output of hydrogen energy equipment and the capacity of hydrogen storage tanks: The input power of the electrolyzer, the output power of the fuel cell, and the capacity of the hydrogen storage tank on the park side must meet the following requirements: In the formula, For the side of the park The maximum input power of each electrolytic cell; For the side of the park The maximum output power of each fuel cell; and The first on each side of the park The lower and upper limits of the capacity of each hydrogen storage device; The second part constructs a multi-scenario probabilistic risk linear constraint system based on conditional value at risk theory. This system is designed to address risks such as load shedding in industrial parks, insufficient reserve requirements due to grid-side increases, and insufficient reserve requirements due to grid-side decreases. Specifically, it includes: 1) Basic definition of conditional value at risk and method for linearizing nonlinear risk terms: Let the first... A scene, The risk loss function corresponding to time is Among them, risk types This includes risks such as load shedding in industrial parks, insufficient reserve requirements on the grid side, and insufficient reserve requirements on the grid side; a conditional value of risk auxiliary variable is introduced. and excess risk loss variables The positive part of the function term where the risk loss exceeds the threshold of the conditional value of risk is transformed into a linear constraint, satisfying: The type of risk exist Conditional Value at Time satisfy: In the formula, This is the tail risk probability parameter; 2) Definition of loss function for the park side-shear load scenario: For the first In one scenario, when the actual output of wind power, photovoltaic power, grid-park interconnection power, and net output of the hydrogen energy system on the park side cannot meet the total load demand on the park side, a load shedding variable on the park side is introduced. The loss function for the park side-cut load scenario satisfy: In the formula, For the side of the park The load node at the ... A scene, The load shedding power at any given moment; For the first A scene, Loss function for the park side-cut load scenario at any given time; 3) Linear constraint on conditional value of risk corresponding to tail risk of side-shear load in the park: The loss function for the side-shear load scenario in the park... Substituting the conditional value-at-risk linearization constraint, we obtain the conditional value-at-risk linearization constraint corresponding to the tail risk of the park's side-shear load, which satisfies: In the formula, As an auxiliary variable for the risk value of side-shear load in the park; For the excess loss due to side-shear load risk in the industrial park; The threshold for controlling the side-shear load risk in the park; 4) Risk constraints on the risk value of grid-side adjustments to reserve shortage loss function and associated conditions: [Regarding the first...] In one scenario, when the actual output of wind and solar power on both the grid and industrial park sides is lower than the day-ahead forecast baseline output, the system generates an upward adjustment of reserve demand. This upward adjustment of reserve demand variable... satisfy: Upgrade the reserve shortage loss function on the grid side satisfy: The conditional value-at-risk linear constraint corresponding to the insufficient reserve on the grid side is satisfied as follows: Regarding the first In one scenario, when the actual output of wind and solar power on both the grid and industrial park sides exceeds the day-ahead forecast baseline output, the system generates a downward adjustment of reserve demand. This downward adjustment of reserve demand variable... satisfy: The power grid side lowers the reserve shortage loss function satisfy: The conditional value-at-risk linear constraint corresponding to the grid-side reduction in reserve requirements satisfies: In the formula, For the first A scene, Adjusting the reserve requirement variable at any time; For the first A scene, Adjusting the reserve requirement variable at any time; For the first A scene, The grid side adjusts the reserve shortage loss function at any given time. For the first A scene, The grid side adjusts the reserve shortage loss function at a given time. For the power grid side Taiwan thermal power units Reserved standby capacity at all times; For the power grid side Taiwan thermal power units Reserved standby capacity at all times; and These are auxiliary variables for adjusting the value-at-risk (VAT) of insufficient reserves on the grid side and adjusting the VAT of insufficient reserves on the grid side, respectively. and These represent the upward adjustment of the excess variable for risk loss due to insufficient reserves on the grid side and the downward adjustment of the excess variable for risk loss due to insufficient reserves on the grid side. and These represent raising the reserve shortage risk control threshold on the grid side and lowering the reserve shortage risk control threshold on the grid side, respectively. By using the conditional risk value linear constraints corresponding to the tail risk of load shedding in the industrial park, the conditional risk value linear constraints corresponding to insufficient reserve adjustment on the grid side, and the conditional risk value linear constraints corresponding to insufficient reserve adjustment on the grid side, a multi-scenario probabilistic risk linear constraint system is formed. Among them, the losses from load shedding in the industrial park and insufficient reserve adjustment on the grid side are used to construct the CVaR risk constraints. The wind and solar curtailment losses caused by wind and solar curtailment are not included in the conditional value risk variables, but are only used as expected penalty terms in the objective function.

5. The power grid and industrial park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in claim 4, characterized in that: Step four involves constructing the comprehensive cost objective function and completing the optimization solution. This includes constructing the power grid-side specific cost term, the park-side specific cost term, the expected penalty cost term for the entire scenario, and the output of the model solution and scheduling scheme, as detailed below: 1) Construct grid-side dedicated cost items: The grid-side dedicated cost items include the fuel cost of grid-side thermal power units, the start-up and shutdown costs of the units, and the reserved costs for increasing or decreasing reserve capacity; wherein, the grid-side dedicated cost items satisfy: In the formula, This is a cost item specific to the power grid side; For the power grid side The unit of Taiwan's thermal power plant increased its reserve capacity to reserve costs; For the power grid side The unit of Taiwan's thermal power plant reduced its reserve capacity to reserve costs; 2) Constructing a dedicated cost item for the park side: The dedicated cost item for the park side includes the operation and maintenance loss cost of the hydrogen energy equipment on the park side and the power exchange transaction cost of the grid-park interconnection line; to maintain the linearity of the objective function, the power exchange of the grid-park interconnection line is decomposed into power purchased and power sold, satisfying: The park-side specific cost items satisfy: In the formula, This is a cost item specific to the park. For the park side Power purchased from the power grid at all times via the grid-park interconnection line; For the park side The power sold to the grid at all times through the grid-park interconnection line; For the side of the park Unit operation and maintenance cost coefficient of each electrolytic cell; For the side of the park Unit operation and maintenance cost coefficient of a fuel cell; for The electricity purchase price at the industrial park; for Electricity price at the park's side; 3) Constructing the full-scenario expected penalty cost term: The full-scenario expected penalty cost term calculates the expected penalty costs for all sampled scenarios generated in step two, including grid-side wind power curtailment, grid-side photovoltaic power curtailment, park-side wind power curtailment, park-side photovoltaic power curtailment, and park-side multi-node load shedding compensation; wherein, the grid-side and park-side wind and solar power curtailment power respectively satisfy: The expected penalty cost item in the whole scenario satisfy: In the formula, This represents the expected penalty cost across all scenarios. For the power grid side The first wind farm station A scene, Wind curtailment power at any given time; For the power grid side The first photovoltaic power station A scene, The power of light discarded at any given moment; For the side of the park The first wind farm station A scene, Wind curtailment power at any given time; For the side of the park The first photovoltaic power station A scene, The power of light discarded at any given moment; The unit cost coefficient for wind and solar power curtailment penalties; The unit load shedding compensation cost coefficient; Among them, the curtailed wind and solar power includes four parts: grid-side wind power curtailment, grid-side solar power curtailment, park-side wind power curtailment, and park-side solar power curtailment, which respectively correspond to the over-limit curtailment losses of the actual output of wind and solar power on the grid side and the park side relative to the day-ahead predicted baseline output; the curtailed wind and solar power is included in the system's comprehensive economic objective function as the expected penalty cost item for the whole scenario, and is not included as the conditional risk value risk loss variable. 4) Model solution and scheduling scheme output: Output the grid-side specific cost item Park-side dedicated cost items And the expected penalty cost item for the entire scenario Integrated into the system's overall cost objective function It satisfies: The optimization objective is to minimize the overall system cost objective function, satisfying: The system comprehensive cost objective function The model is integrated with the deterministic physical constraints, multi-scenario conditional risk value probability risk linear constraints, grid-park interconnection power purchase and sale power split constraints, and wind and solar power curtailment non-negativity constraints constructed in step three to form a mixed-integer linear programming model. The mixed-integer linear programming solver is called to perform global optimization and output a grid-park collaborative probabilistic scheduling scheme that meets the requirements of economy and reliability. The proposed power grid and park collaborative probabilistic dispatch scheme includes upper-level power grid dispatch results, park hydrogen energy dispatch results, grid-park interactive dispatch results, and scenario risk handling results. The upper-level power grid dispatch results include the start-up and shutdown status of thermal power units on the power grid side, the segmented output of thermal power units, and the increase or decrease of reserve capacity. The hydrogen energy dispatch results of the park include the input power of the electrolyzer on the park side, the output power of the fuel cell, and the hydrogen storage capacity of the hydrogen storage equipment. The grid-park interactive scheduling results include the grid-park interconnection line interactive power, the park-side power purchase power, and the park-side power sales power. The scenario risk management results include grid-side wind power curtailment, grid-side photovoltaic power curtailment, park-side wind power curtailment, park-side photovoltaic power curtailment, park-side load shedding power, and the conditional risk value indicators corresponding to park-side load shedding risk, grid-side upward reserve shortage risk, and grid-side downward reserve shortage risk.

6. A power grid and industrial park collaborative probabilistic dispatching system based on Gaussian mixture model and conditional value risk, used to implement the power grid and industrial park collaborative probabilistic dispatching method based on Gaussian mixture model and conditional value risk as described in any one of claims 1-5, characterized in that, It includes a data acquisition and initialization module, a source-load scenario generation module, a collaborative scheduling modeling module, and an optimization solution decision module; The data acquisition and initialization module is used to collect and initialize multi-source basic data of the system, operating parameters of the grid side and the park side in real time, and synchronously collect historical prediction error data of wind power and photovoltaic power on both sides of the grid and the park and complete normalization preprocessing to provide input sources for uncertainty modeling. The operating parameters include the output range, ramping constraints and cost coefficients of thermal power units on the grid side, the benchmark value of multi-node load prediction on the park side, the power transmission limit of grid-park interconnection line, and the rated power, conversion efficiency and hydrogen storage tank capacity limit of the hydrogen energy system electrolyzer and fuel cell. The source load scene generation module is connected to the data acquisition initialization module. It is used to construct a Gaussian mixture model based on the expectation-maximization algorithm to fit the historical prediction error data of wind and solar power, complete the identification of model weights, mean and covariance parameters, and use Monte Carlo sampling to generate multiple sets of discrete representative wind and solar power output scene sets, transforming the continuous probability distribution into discrete scenes that can be used for optimization calculation. The collaborative scheduling modeling module, connected to the source-load scenario generation module, is used to build a grid-park joint constraint system based on a discrete wind and solar power output scenario set. On the one hand, it constructs deterministic physical operation constraints that include thermal power output, dynamic hydrogen energy storage, grid-park power interaction, and park load balance. On the other hand, it quantifies the system risk of extreme scenarios based on conditional risk value theory, linearizes the risk loss function by introducing risk auxiliary variables, and establishes probabilistic constraints for three types of tail risks: park load shedding, grid up-and-down adjustment of insufficient reserves, etc. Under a pre-set confidence level, it ensures that the system's flexible resources can mitigate the impact of new energy fluctuations. The optimization solution decision module, connected to the collaborative scheduling modeling module, is used to construct a comprehensive cost function with the objective of minimizing the total expected operating cost of the system. The objective function integrates the cost of thermal power generation, hydrogen energy operation, grid-park interaction transaction costs, reserve costs, and expected penalties for wind and solar curtailment and load shedding in multiple scenarios. The mixed-integer linear programming solver is invoked to solve the overall model globally, outputting the start-up and shutdown and output plans of thermal power units within the scheduling cycle, the hydrogen charging and discharging paths of the hydrogen energy system, and the collaborative scheduling scheme for interactive power of grid-park interconnection lines, so as to achieve the optimal balance between system operating economy and power supply reliability.

7. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a power grid and park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in any one of claims 1-5.

8. A computer program product, characterized in that, When the computer program is executed by the processor, it implements a power grid and park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor to implement a power grid and park collaborative probabilistic scheduling method based on Gaussian mixture model and conditional value risk as described in any one of claims 1-5.

Images