New energy collaborative scheduling optimization method, system, equipment and medium

Abstract

The invention discloses a new energy collaborative scheduling optimization method, system and device and a medium. The method comprises the following steps: acquiring historical wind and light power generation data, meteorological data and geographic information data; based on the preprocessed data, obtaining different wind and light output scenes through an unsupervised learning technology; based on different wind and light output scenes, constructing a first optimization model through an adaptive adjustment strategy; combining the first optimization model with a system constraint condition, and establishing a first scheduling model by using mixed integer linear programming; and applying a dualization processing and column constraint generation algorithm to solve the first scheduling model, and outputting an optimal scheduling strategy, so that the economical efficiency and robustness of power system scheduling can be improved.

Description

Technical Field

[0001] This invention relates to the field of new energy collaborative scheduling technology, and in particular to a new energy collaborative scheduling optimization method, system, equipment and medium. Background Technology

[0002] As the world vigorously promotes energy structure transformation and actively practices the concept of green and low-carbon development, the power system is undergoing a critical period of profound change. The traditional power supply pattern, dominated by thermal power, is gradually shifting towards a diversified model with a high proportion of renewable energy. However, this transformation is not without its challenges; the characteristics of renewable energy generation pose severe challenges to the operation and dispatch of the power system. With the rapid development of renewable energy, the coordinated dispatch of different types of renewable energy (such as wind power and photovoltaics) and between renewable and traditional energy sources has become increasingly complex. In actual operation, wind farms and photovoltaic power plants are often located in different geographical locations, and their output characteristics are affected by various factors such as local climate and topography, exhibiting strong spatiotemporal differences. For example, multiple wind farms in a region may have different peak and trough periods due to varying wind speed patterns caused by different terrains such as valleys and plains; photovoltaic power plants also show significant differences in their power generation curves due to differences in orientation and shading. This spatiotemporal difference makes it difficult to accurately grasp the output status of each power source during coordinated dispatch, increasing the difficulty of dispatch optimization.

[0003] In past power system dispatch optimization, deterministic optimization methods have been predominantly employed. These methods rely on fixed parameters and assumptions for calculations, failing to adequately consider the uncertainties in renewable energy output and the complex constraints of actual operation. For instance, in computer grouping and power allocation, renewable energy output is treated as a fixed value or a simple predicted average, ignoring its significant fluctuations under different weather and seasonal conditions. When actual renewable energy output deviates considerably from the preset values, it can lead to power deficits or surpluses, impacting the safe and stable operation of the power system. Existing models involving renewable energy dispatch exhibit significant limitations in adaptability to the diversity and dynamic changes of real-world operating scenarios. Model parameters are often statically set based on historical data, failing to track real-time changes in renewable energy output characteristics. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a new energy collaborative scheduling optimization method, system, equipment, and medium to solve the problems of difficult collaborative scheduling of wind and solar power caused by spatiotemporal differences and poor model adaptability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a new energy collaborative scheduling optimization method, comprising:

[0008] Acquire historical wind and solar power generation data, meteorological data, and geographic information data;

[0009] Based on the preprocessed data, different wind and solar power output scenarios are obtained through unsupervised learning techniques;

[0010] Based on different wind and solar power output scenarios, an adaptive adjustment strategy is used to construct the first optimization model;

[0011] By combining the first optimization model with system constraints, and using mixed-integer linear programming, a first scheduling model is established.

[0012] By applying dualization and column constraint generation algorithms, the first scheduling model is solved, and the optimal scheduling strategy is output.

[0013] As a preferred embodiment of the new energy collaborative scheduling optimization method described in this invention, the preprocessing includes:

[0014] The acquired data will be cleaned and standardized.

[0015] After cleaning and standardization, the data is processed by variational mode decomposition to reduce noise and obtain multiple intrinsic mode function components.

[0016] By solving the variational problem, the original constraint problem is transformed into a form that is easier to solve, and the number of each eigenmode function is obtained.

[0017] The number of intrinsic mode functions is dynamically adjusted by using profile coefficients to determine the optimal number of modes and obtain the set of intrinsic mode functions.

[0018] As a preferred embodiment of the new energy collaborative scheduling optimization method described in this invention, the obtaining of different wind and solar power output scenarios includes:

[0019] Based on the intrinsic mode function set, multiple cluster centers are obtained through unsupervised learning techniques;

[0020] Based on multiple cluster centers, the optimal cluster centers and membership assignments are found through objective function optimization.

[0021] The optimal cluster centers and membership assignments are transformed into specific wind and solar power output scenarios. The clustering quality is evaluated by the profile coefficient, resulting in different wind and solar power output scenarios.

[0022] The beneficial effect of this preferred technical solution is that by combining the clustering optimization method of intrinsic mode function, Mahalanobis distance and dynamic covariance matrix, it can generate high-quality wind and solar power output scenarios more accurately and adaptively, thereby improving the power system dispatch model's ability to characterize the uncertainty of new energy sources and its robustness.

[0023] As a preferred embodiment of the new energy collaborative scheduling optimization method described in this invention, the construction of the first optimization model includes:

[0024] Considering the uncertainty of historical data distribution, a fuzzy set is constructed with the empirical distribution as the center and the bulldozer distance as the radius;

[0025] The radius of the fuzzy set is dynamically adjusted based on the standard deviation and sample size of the data.

[0026] The bulldozer distance is approximated using an entropy-regularized transport algorithm.

[0027] Establish a multi-objective robust optimization framework;

[0028] Risk under extreme conditions is assessed by using an improved conditional value-at-risk adaptive weighting function.

[0029] The beneficial effects of this preferred technical solution are that by constructing a distributed robust optimization framework, combining dynamic radius adjustment and entropy regularization transmission algorithms, and introducing kernel density estimation to improve risk assessment, the model's adaptability to the uncertainty of new energy output and the economy and robustness of the scheduling strategy are effectively improved.

[0030] As a preferred embodiment of the new energy collaborative scheduling optimization method described in this invention, the establishment of the first scheduling model includes:

[0031] Define power balance constraints, generator generation upper and lower limit constraints, line power flow constraints, and energy storage charging and discharging constraints;

[0032] Based on the defined constraints, the physical limitations are transformed into mathematical expressions through mixed-integer linear programming and embedded into the objective function to form a complete optimization problem framework;

[0033] An uncertainty modeling mechanism is introduced, and a scheduling strategy is output under the worst-case distribution by utilizing the generated wind and solar power output scenarios and fuzzy sets.

[0034] As a preferred embodiment of the new energy collaborative scheduling optimization method described in this invention, the step of solving the first scheduling model includes:

[0035] By using dualization, the distributed robust optimization problem is transformed into a convex optimization problem;

[0036] The transformed convex optimization problem is solved iteratively using a column constraint generation algorithm.

[0037] Set a convergence criterion, and when convergence reaches the optimal solution, output the optimal scheduling strategy.

[0038] As a preferred embodiment of the new energy collaborative scheduling optimization method described in this invention, the establishment of the multi-objective robust optimization framework is expressed as follows:

[0039] min x sup p∈p (E p [f(x,ξ)]+λ1CVaR 0.9 [g wind (x,ξ)]+λ2CVaR 0.9 [g solar (x,ξ)])

[0040] Where f(x,ξ) is the economic cost function; g wind (x,ξ) is the function of wind curtailment; g solar (x,ξ) represents the light waste function; λ is the weighting coefficient between economic efficiency and environmental friendliness, CVaR 0.9 This is the conditional risk value.

[0041] Secondly, this invention provides a new energy collaborative scheduling optimization system, comprising:

[0042] The acquisition module is used to acquire historical wind and solar power generation data, meteorological data, and geographic information data;

[0043] The scene generation module is used to obtain different wind and solar power output scenes based on preprocessed data through unsupervised learning techniques;

[0044] The optimization model building module is used to build the first optimization model based on different wind and solar power output scenarios by adaptively adjusting strategies;

[0045] The scheduling model establishment module is used to combine the first optimization model with system constraints and establish the first scheduling model using mixed integer linear programming.

[0046] The solution module is used to apply dualization processing and column constraint generation algorithms to solve the first scheduling model and output the optimal scheduling strategy.

[0047] Thirdly, the present invention provides an electronic device, comprising:

[0048] Memory, used to store programs;

[0049] A processor is configured to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the new energy collaborative scheduling optimization method.

[0050] Fourthly, the present invention provides a computer-readable storage medium, comprising: when the program is executed by a processor, the steps of implementing the new energy collaborative scheduling optimization method.

[0051] The beneficial effects of this invention are as follows: This invention utilizes unsupervised learning technology, replacing Euclidean distance with Mahalanobis distance, and combines it with dynamically updated covariance matrices to accurately capture the correlation between features such as wind speed and light intensity, achieving clustering results far exceeding traditional methods. Simultaneously, it dynamically determines the number of modes and clusters based on silhouette coefficients, making scene generation more closely resemble actual operating conditions and providing more representative input for subsequent optimization. The constructed optimization model, through adaptive adjustment of the fuzzy set radius, better addresses the uncertainty of historical data distribution. The innovative multi-objective robust optimization framework and the improved conditional value-at-risk calculation method effectively balance economic costs and wind curtailment volume while controlling extreme wind curtailment risks. During model solving, dualization and column constraint generation algorithms simplify complex problems and solve them efficiently, significantly improving solution efficiency and result accuracy, and comprehensively enhancing the economy and robustness of power system dispatch. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0053] Figure 1 This is a schematic diagram of the basic process of a new energy collaborative scheduling optimization method provided in one embodiment of the present invention. Detailed Implementation

[0054] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0055] Example 1, referring to Figure 1 As an embodiment of the present invention, a new energy collaborative scheduling optimization method is provided, comprising:

[0056] S100: Acquire historical wind and solar power generation data, meteorological data, and geographic information data;

[0057] S200: Based on preprocessed data, different wind and solar power output scenarios are obtained through unsupervised learning techniques;

[0058] S300: Based on different wind and solar power output scenarios, an adaptive adjustment strategy is used to construct the first optimization model;

[0059] S400: Combine the first optimization model with system constraints and use mixed-integer linear programming to establish the first scheduling model;

[0060] S500: Apply dualization processing and column constraint generation algorithm to solve the first scheduling model and output the optimal scheduling strategy.

[0061] It should be noted that existing power system dispatch optimization faces numerous challenges during operation, primarily including the strong uncertainty and volatility of renewable energy output, the time-varying characteristics of load demand, and the complexity of system operating constraints. With the large-scale integration of renewable energy sources such as wind and solar power, their output is significantly affected by meteorological conditions, making it difficult for traditional dispatch methods based on deterministic or static probability distributions to accurately reflect actual operating conditions. This can easily exacerbate wind and solar power curtailment or lead to insufficient system regulation capacity. Therefore, exploring real-time tracking of changes in renewable energy output characteristics is of great significance for improving the adaptability and robustness of dispatch models.

[0062] Therefore, addressing the aforementioned difficulties in coordinated scheduling of wind and solar power due to spatiotemporal differences, as well as the poor model adaptability, the S100-S500 steps, through the introduction of fuzzy sets, a dynamic radius adjustment mechanism, an entropy regularized transmission algorithm, and an improved conditional risk value assessment method, effectively enhance the scheduling model's adaptability and decision robustness in the face of wind and solar power output fluctuations. Simultaneously, by combining system operation constraints and a multi-objective optimization structure, a coordinated optimization of economic efficiency and environmental protection is achieved, providing strong support for the safe, economical, and stable operation of the power system under the background of high-proportion renewable energy integration.

[0063] Example 2, this is an embodiment of the present invention, which provides a new energy collaborative scheduling optimization method based on the previous embodiment, including:

[0064] In this embodiment, step S100 involves acquiring historical wind and solar power generation data, meteorological data, and geographic information data, including historical power output data for wind and solar power generation, recording the actual power output of wind and solar power generation at different time points. Simultaneously, corresponding meteorological data is collected, such as wind speed, solar irradiance, and temperature, as well as geographic information data of the wind and solar power station locations. Power grid operation data, including bus voltage and line power flow, needs to be collected.

[0065] In this embodiment of the application, the preprocessing in step S200 includes:

[0066] The acquired data will be cleaned and standardized.

[0067] After cleaning and standardization, the data is processed by variational mode decomposition (VMD) to reduce noise and obtain multiple intrinsic mode function components.

[0068] By solving the variational problem, the original constraint problem is transformed into a form that is easier to solve, and the number of each eigenmode function is obtained.

[0069] The number of intrinsic mode functions is dynamically adjusted by using profile coefficients to determine the optimal number of modes and obtain the set of intrinsic mode functions.

[0070] In this embodiment of the application, the acquired data undergoes data cleaning and standardization. This includes the possibility that the collected data may contain missing or outlier values ​​due to equipment malfunctions, transmission interference, or other reasons. Missing values ​​can be filled in; outliers can be removed or corrected.

[0071] In this embodiment, the signal denoising process using variational mode decomposition (VMD) addresses the non-stationary characteristics of renewable energy output and load data. This method achieves signal decomposition by solving the following constrained variational problem:

[0072]

[0073] Among them, u k Let ω be the k-th modal component (IMF), representing the decomposed sub-signal; k is the center frequency of the k-th mode; f is the original signal.

[0074] In this embodiment, the variational problem is solved by introducing a quadratic penalty term and an augmented Lagrangian function, transforming the problem into a solvable form using the Alternating Direction Multiplier Method (ADMM). This solves the constrained optimization problem, transforming the constrained variational problem into an unconstrained variational problem. VMD adaptively decomposes the original signal into multiple intrinsic mode functions (IMFs), each with a defined center frequency and finite bandwidth, eliminating high-frequency noise. Specifically, the bandwidth of the modal components is controlled by adjusting the bandwidth constraint parameter α (a larger α value produces a narrower bandwidth, suitable for extracting high-frequency noise), and the number of modes K is dynamically determined based on the profile coefficients, avoiding false modes caused by over-decomposition or information loss caused by under-decomposition. The augmented Lagrangian function is constructed as follows:

[0075]

[0076] Where α is the bandwidth constraint parameter and λ is the Lagrangian multiplier.

[0077] In this embodiment, the dynamic adjustment of the number of intrinsic mode functions using the profile coefficient is expressed as follows:

[0078]

[0079] Where a(i) is the average distance from sample i to other samples in the same cluster; b(i) is the average distance from sample i to the nearest other sample in the cluster. When SC > 0.6, it is considered an effective decomposition to avoid overfitting.

[0080] In this embodiment of the application, different wind and solar power output scenarios are obtained in step S200, including:

[0081] Based on the intrinsic mode function set, multiple cluster centers are obtained through unsupervised learning techniques;

[0082] Based on multiple cluster centers, the optimal cluster centers and membership assignments are found through objective function optimization.

[0083] The optimal cluster centers and membership assignments are transformed into specific wind and solar power output scenarios. The clustering quality is evaluated by the profile coefficient, resulting in different wind and solar power output scenarios.

[0084] In one alternative implementation, the unsupervised learning technique in step S200 can be an improved FCM clustering algorithm, or a method based on self-organizing map (SOM); or a method based on DBSCAN (Density-Based Spatial Clustering of Applications with Noise).

[0085] In one alternative implementation, the improved FCM clustering algorithm includes an improved FCM clustering algorithm based on historical wind and solar power output data, which incorporates Mahalanobis distance and a sliding window mechanism to dynamically capture changes in feature correlation, thereby generating more representative and time-varying typical wind and solar power output scenarios.

[0086] In one alternative implementation, the self-organizing map (SOM) based method includes using historical wind and solar power output data, employing the self-organizing map (SOM) method, and mapping high-dimensional data to a low-dimensional space through competitive learning, dynamically capturing and visualizing typical wind and solar power output patterns, thereby generating wind and solar power output scenarios with topology preservation characteristics.

[0087] In one alternative implementation, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method includes using the DBSCAN clustering method based on historical wind and solar power output data, utilizing density distribution characteristics to automatically identify different wind and solar power output patterns, and effectively capturing abnormal scenes, thereby generating typical wind and solar power output scenes that do not require a preset number of clusters and have strong robustness.

[0088] It should be noted that the improved FCM clustering algorithm used in this invention demonstrates stronger adaptability and practicality in generating wind and solar power output scenarios compared to methods based on Self-Organizing Maps (SOM) and DBSCAN. The improved FCM effectively considers the correlation between wind and solar power output features (such as the dynamic negative correlation between wind speed and light intensity) by introducing Mahalanobis distance instead of traditional Euclidean distance, and improves the ability to characterize time-varying uncertainties by dynamically updating the covariance matrix through a sliding window mechanism. Furthermore, compared to the SOM method's reliance on network structure settings and complex training, and the DBSCAN method's sensitivity to parameters and unsuitability for high-dimensional data, the improved FCM maintains clustering accuracy while exhibiting better stability and interpretability. It is particularly suitable for handling wind and solar power output scenarios with fuzzy boundaries and overlapping characteristics in power systems, thereby generating a more representative and robust set of typical scenarios.

[0089] In this embodiment, the improved FCM clustering algorithm addresses the issue that traditional k-Means uses Euclidean distance to measure similarity, which fails to account for strong correlations between power data features (e.g., wind speed and light intensity may be negatively correlated due to weather changes), thus affecting clustering performance. This invention proposes an improved k-Means algorithm using Mahalanobis distance instead of Euclidean distance:

[0090]

[0091] Where x is the sample vector; c is the cluster center; and S is the covariance matrix, which is dynamically updated every 24 hours using a sliding window mechanism to reflect the dynamic correlation between features in real time.

[0092] In this embodiment, the preprocessed data is input into the improved FCM clustering algorithm, and the algorithm iteratively optimizes the objective function as follows:

[0093]

[0094] Where N is the number of samples, C is the number of clusters, and u ijD represents the membership degree of sample i to class j, and m is the fuzzy index (usually m=2). When the power system data characteristics are complex and the cluster boundaries are fuzzy, the value of m can be appropriately increased to enhance the fuzziness of the clustering and make the clustering results more consistent with the actual situation; if the data characteristics are relatively clear, the value of m can be appropriately decreased to speed up the convergence speed. M (x i ,c j ) is the relationship between sample i and cluster center c j The Mahalanobis distance is used to minimize the objective function by continuously adjusting the membership degree and cluster center, thus completing the clustering.

[0095] In this embodiment, evaluating clustering quality using the silhouette coefficient includes identifying different wind and solar power output scenarios based on the clustering results. Each scenario represents a wind and solar power output pattern with similar characteristics. The silhouette coefficient is used to evaluate clustering quality, with a threshold of 0.6. When the silhouette coefficient is greater than this threshold, the clustering results are considered to effectively distinguish different operating scenarios, providing diverse and representative scenario inputs for subsequent optimization and scheduling.

[0096] In this embodiment of the application, step S300, which involves constructing the first optimization model (Wasserstein-DRO optimization model), includes:

[0097] Considering the uncertainty of historical data distribution, a fuzzy set is constructed with the empirical distribution as the center and the bulldozer distance as the radius;

[0098] The radius of the fuzzy set is dynamically adjusted based on the standard deviation and sample size of the data.

[0099] The bulldozer distance is approximated using an entropy-regularized transport algorithm.

[0100] Establish a multi-objective robust optimization framework;

[0101] Risk under extreme conditions is assessed by using an improved conditional value-at-risk adaptive weighting function.

[0102] In this embodiment of the application, a fuzzy set representation centered on the experience distribution and with the bulldozer distance as the radius is constructed as follows:

[0103]

[0104] Where P is the true probability distribution. Since it is difficult to obtain accurately in practice, a fuzzy set is constructed to cover the possible distribution range. For empirical distribution (historical data); W p ∈ is the Wasserstein distance, which measures the difference between two probability distributions; ∈ is the radius, which controls the size of the fuzzy set. The larger ∈ is, the wider the range of distributions covered by the fuzzy set.

[0105] In this embodiment, the fuzzy set radius is dynamically adjusted based on the standard deviation and sample size of the data, as expressed as:

[0106]

[0107] Where σ is the sample standard deviation, n is the sample size, and k is the adjustment coefficient. When the power system has high stability requirements and it is desired that the fuzzy set can more accurately reflect the data distribution, the value of k can be appropriately reduced; if the data fluctuates drastically and the model's adaptability is high, the value of k can be appropriately increased. Through this adaptive adjustment, the fuzzy set can better adapt to the uncertainty of the data.

[0108] In this embodiment, the Wasserstein distance is approximated using the Sinkhorn algorithm with entropy regularization:

[0109]

[0110] Among them, the newly added nuclear norm regularization term effectively alleviates the ill-conditioned problem of the transfer matrix.

[0111] In this embodiment of the application, the multi-objective robust optimization framework is represented as follows:

[0112] min x sup p∈p (E p [f(x,ξ)]+λ1CVaR 0.9 [g wind (x,ξ)]+λ2CVaR 0.9 [g solar (x,ξ)])

[0113] Where f(x,ξ) is the economic cost function; g wind (x,ξ) is the function of wind curtailment; g solar (x,ξ) represents the curtailment rate function; λ is the weighting coefficient between economic efficiency and environmental friendliness, which should be adjusted according to the actual needs of the power system, λ1+λ2=0.7; CVaR 0.9 Conditional risk value is used to control the risks of extreme wind and solar curtailment.

[0114] In this embodiment, the improved conditional value at risk (CVaR) adaptively adjusted weight function is expressed as follows:

[0115]

[0116] in, It is the quantile function of the random variable X, and the weight function w(u) is adaptively adjusted through kernel density estimation. Kernel density estimation is a nonparametric estimation method used to estimate the probability density function of a random variable. By adaptively adjusting the weight function, the risk under extreme conditions can be estimated more accurately, improving the accuracy of tail risk measurement and enabling the model to achieve better optimization results under different wind and solar power output scenarios.

[0117] In an optional implementation, the first scheduling model in step S400 can be an optimization scheduling model based on distributed robust optimization, or a scheduling model based on stochastic programming (SP); or a rolling scheduling framework based on model predictive control (MPC).

[0118] In one alternative implementation, the distributed robust optimization-based scheduling model includes constructing a robust optimization model based on system constraint analysis, using Wasserstein fuzzy sets to characterize the uncertainty of wind and solar power output, combining constraints such as power balance, unit operation, line power flow and energy storage operation, using a mixed integer linear programming framework for modeling, and solving it through the Lagrange duality and column constraint generation algorithm (C&CG), ultimately outputting an optimal scheduling strategy that balances economy and robustness.

[0119] In one alternative implementation, the scheduling model based on stochastic programming (SP) includes constructing a two-stage stochastic programming model with the goal of minimizing expected costs based on the generated wind and solar power output scenarios and their probability distributions. Under the premise of satisfying constraints such as power balance, unit limits, line power flow and energy storage operation, a mixed integer linear programming framework is used for modeling, and the model is solved directly through scenario decomposition or a solver to output the optimal scheduling strategy that considers the probabilistic characteristics of multiple scenarios.

[0120] In one optional implementation, the rolling scheduling framework based on Model Predictive Control (MPC) includes constructing an MPC scheduling framework with rolling optimization as the goal, based on short-term forecasts of wind and solar power output and system operation constraints. In each scheduling cycle, the optimal power generation and energy storage scheduling plan within a finite time window is solved, and the scheduling strategy is dynamically adjusted in combination with real-time updated forecast information to achieve rapid response to the volatility of new energy sources and continuous optimization of system operation.

[0121] It should be noted that the distributed robust optimization (DRO)-based optimization scheduling model adopted in this invention, compared with the stochastic programming (SP)-based scheduling model and the model predictive control (MPC)-based rolling scheduling framework, has stronger adaptability and theoretical support in dealing with the uncertainty of renewable energy output. Compared with the SP method, DRO does not rely on precise assumptions about the probability distribution of wind and solar power output. Instead, it seeks the optimal scheduling strategy under the worst-case distribution by constructing a fuzzy set that includes empirical distributions, thereby avoiding decision-making biases caused by probability miscalculations and improving the robustness of the scheduling scheme. Compared with MPC, DRO not only considers the fluctuations caused by short-term prediction errors, but also theoretically guarantees the existence of the optimal solution in uncertain but bounded scenarios, enhancing the stability and reliability of the scheduling model under extreme weather or data-scarce scenarios. Therefore, this model, while taking into account economic efficiency, has stronger risk resistance capabilities and is suitable for the power system optimization scheduling needs in complex and uncertain environments with a high proportion of renewable energy integration.

[0122] In this embodiment of the application, the optimized scheduling model based on distributed robust optimization in step S400 includes system constraint analysis, scheduling model construction, and how to consider uncertainty factors.

[0123] In this application embodiment, system constraint analysis includes defining power balance constraints, generator generation upper and lower limit constraints, line power flow constraints, and energy storage charging and discharging constraints;

[0124] In this embodiment of the application, the power balance constraint requires that the sum of the generated power, the load power, and the network loss in the system be equal, that is: ∑ i＝G P i,t =∑ j∈ P j,t +P loss,t Where G represents the set of power generation units, including various power generation equipment such as thermal power, wind power, and photovoltaic power; D represents the set of load nodes, representing various electricity consumption areas in the power system; P i,t P represents the power generation of power generation unit i at time t; j,t P represents the load power of load node j at time t. loss,t Let be the network loss at time t. The network loss can be calculated using methods based on power flow calculations, such as the Newton-Raphson method and the fast decoupling method, etc., without limitation here. The applicable scenarios and differences in accuracy of different methods depend on factors such as the scale and structural complexity of the power system.

[0125] In this embodiment of the application, the upper and lower limits of unit power generation constraints include an upper limit P for the power generation capacity of each power generation unit. i,max and lower limit P i,min That is, P i,min ≤P i,t ≤P i,maxThis is determined by the physical characteristics and operational requirements of power generation equipment. For example, thermal power units have minimum stable combustion load limits during startup and shutdown, while wind power and photovoltaic power are limited by the installed capacity of the equipment.

[0126] In this embodiment of the application, the line power flow constraint is expressed as:

[0127] -P l,max ≤P l,t ≤P l,max

[0128] Among them, P l,t Let P be the power flow of the line at time t at time l. l,max This represents the maximum transmission power of the line. The transmission power of a line is limited by factors such as conductor heating and voltage stability. Exceeding the maximum transmission power may lead to line damage or system voltage collapse.

[0129] In this embodiment of the application, the energy storage charging and discharging constraints include charging power constraints and discharging power constraints, wherein the charging power constraint is expressed as:

[0130] P s,ch,min ≤P s,ch,t ≤P s,ch,max

[0131] The discharge power constraint is expressed as:

[0132] P s,dch,min ≤P s,dch,t ≤P s,dch,max

[0133] Among them, P s,ch,t and P s,dch,t P represents the charging power and discharging power of the energy storage at time t, respectively. s,ch,min P s,ch,max P s,dch,min P s,dch,max This refers to the corresponding power limit value. Additionally, there is an energy storage capacity constraint E. s,min ≤E s,t ≤E s,max E s,t E represents the remaining energy stored at time t. s,min and E s,max These represent the minimum and maximum capacities, reflecting the storage capacity limitations of energy storage devices.

[0134] In this embodiment, the scheduling model construction includes combining the Wasserstein-DRO optimization model and system constraints to construct a distributed robust optimization-based scheduling model. In actual model construction, a mixed-integer linear programming (MILP) framework can be used to handle these constraints and objective functions. Various constraints are transformed into linear inequalities or equations, and the MILP solver is used to find scheduling strategies that satisfy the constraints and optimize the objective function. This model uses the objective function from step two as the optimization objective, and determines the scheduling strategies for each power generation unit (including new energy power generation, thermal power, etc.) and energy storage devices under the premise of satisfying the aforementioned system constraints such as power balance, unit generation upper and lower limits, line power flow, and energy storage charging and discharging. For example, based on different wind and solar power output scenarios, the power generation of thermal power units is rationally arranged, and the charging and discharging state of energy storage is adjusted to achieve economical and reliable operation of the power system.

[0135] In this embodiment, considering uncertainties includes fully incorporating the uncertainty of wind and solar power output into the scheduling model. The uncertainty is integrated into the model's optimization process through the wind and solar power output scenarios generated in step one. A distributed robust optimization method is employed to find the optimal scheduling strategy within a fuzzy set, ensuring the model exhibits good robustness under different wind and solar power output scenarios. For example, in a scenario where wind and solar power output fluctuates significantly due to extreme weather, the model can adjust the power generation of thermal power units and increase the charging and discharging regulation of energy storage based on the optimization results corresponding to different probability distributions within the fuzzy set, thereby maintaining the stable operation of the power system.

[0136] It should be noted that step S400 is a further refinement of step S300, combining specific system constraints (such as power balance, upper and lower limits of power generation, line power flow, energy storage charging and discharging, etc.) to construct a complete distributed robust optimization scheduling model. Step S300 mainly aims to provide an optimization framework that can effectively handle the uncertainty of new energy output, while step S400 applies this framework to specific power system scheduling problems, achieving economical and reliable scheduling of the power system through detailed system constraint analysis and optimization solutions.

[0137] In this embodiment of the application, step S500, solving the first scheduling model, includes:

[0138] By using dualization, the distributed robust optimization problem is transformed into a convex optimization problem;

[0139] The transformed convex optimization problem is solved iteratively using a column constraint generation algorithm.

[0140] Set a convergence criterion, and when convergence reaches the optimal solution, output the optimal scheduling strategy.

[0141] In this embodiment, the dualization process includes transforming the original infinite-dimensional distributed robust optimization problem into a finite-dimensional convex optimization problem using strong duality theory. Consider a distributed robust optimization problem of the following form:

[0142]

[0143] At the same time, the problem satisfies the following constraints:

[0144] g i (x,ξ)≤0,i=1,…,m

[0145] h j (x,ξ)=0,j=1,…,n

[0146] The Lagrange function is constructed as follows:

[0147]

[0148] Where, λ i and μ j Let Lagrange multipliers be the Lagrange multipliers. By introducing Lagrange multipliers to relax the probability measure constraints, the original problem is transformed into a saddle point problem with respect to the Lagrange function. Then, through Fenchel duality, the sup-inf problem is transformed into a min problem, thus simplifying the solution process.

[0149] In this embodiment, the Column Constraint Generation Algorithm (C&CG) includes iteratively solving the transformed convex optimization problem using the C&CG algorithm. When generating feasible solutions in the outer loop, the search range can be determined based on a greedy strategy or a random search strategy. Under the greedy strategy, solutions that significantly improve the objective function value and satisfy some constraints under the current state are prioritized as search directions; the random search strategy randomly generates a series of candidate solutions in the solution space to increase the probability of finding the global optimum. The inner loop processes subproblems using the Alternating Direction Multiplier Method (ADMM). ADMM decomposes the global problem into three subproblems—thermal power, energy storage, and new energy—which are solved in parallel. During parallel computation, subproblems interact and coordinate through shared variables. For example, the power generation calculated in the thermal power subproblem affects the calculation of wind and solar power consumption in the new energy subproblem, while the charging and discharging status of the energy storage subproblem is also fed back to the thermal power and new energy subproblems to ensure the power balance and operational constraints of the entire system. When solving the sub-problems related to thermal power, the main considerations are constraints such as the power generation cost and ramp-up rate of thermal power units; the energy storage sub-problem focuses on charging and discharging power and capacity constraints; and the new energy sub-problem emphasizes the uncertainty and power generation characteristics of wind and solar power output. By parallel computing of these sub-problems, the overall solution speed is accelerated.

[0150] In this embodiment, setting convergence criteria includes defining convergence criteria during the iteration process, such as considering the algorithm convergent when the change in the objective function value is less than a certain threshold. For example, when the difference between the objective function values ​​of two adjacent iterations is less than 10⁻⁶, the algorithm is considered to have converged to a relatively optimal solution. Furthermore, the number of iterations can be used as an auxiliary judgment condition. If the number of iterations reaches a pre-set upper limit but the objective function value still has not converged to a satisfactory level, the algorithm parameters can be appropriately adjusted and the solution re-solved.

[0151] In this embodiment, when the optimal solution is converged, the optimal scheduling strategy is output, including the power generation plan of each power generation unit, such as the power generation setting of thermal power units at different times, and the expected power generation of wind and solar power; the charging and discharging strategy of energy storage devices, i.e., when to charge, how much charging power, when to discharge, and how much discharging power, etc., providing a decision-making basis for the actual operation of the power system. After the output results are obtained, the scheduling strategy can be verified and evaluated using simulation software, such as Power System Analysis and Synthesis Program (PSASP) or DigSILENT PowerFactory. The output power generation plan and energy storage strategy are input into the simulation software to simulate the operation of the power system under different operating conditions, check whether the constraints such as power balance and voltage stability are met, evaluate the economic indicators of the system, such as total power generation cost and wind and solar curtailment, and further optimize the scheduling strategy based on the evaluation results.

[0152] Example 3 is an embodiment of the present invention. This embodiment differs from the first embodiment in that it provides a new energy collaborative scheduling optimization system.

[0153] It should be noted that the technical solution of the new energy collaborative scheduling optimization system and the technical solution of the new energy collaborative scheduling optimization method mentioned above belong to the same concept. For details not described in detail in the technical solution of the new energy collaborative scheduling optimization system in this embodiment, please refer to the description of the technical solution of the new energy collaborative scheduling optimization method mentioned above.

[0154] This embodiment provides a new energy collaborative scheduling and optimization system, comprising:

[0155] The acquisition module is used to acquire historical wind and solar power generation data, meteorological data, and geographic information data;

[0156] The scene generation module is used to obtain different wind and solar power output scenes based on preprocessed data through unsupervised learning techniques;

[0157] The optimization model building module is used to build the first optimization model based on different wind and solar power output scenarios by adaptively adjusting strategies;

[0158] The scheduling model establishment module is used to combine the first optimization model with system constraints and establish the first scheduling model using mixed integer linear programming.

[0159] The solution module is used to apply dualization processing and column constraint generation algorithms to solve the first scheduling model and output the optimal scheduling strategy.

[0160] This embodiment also provides an electronic device applicable to a new energy collaborative scheduling optimization method, including:

[0161] The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a new energy collaborative scheduling optimization method as proposed in the above embodiments.

[0162] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a new energy collaborative scheduling optimization method as proposed in the above embodiments.

[0163] The storage medium proposed in this embodiment and the method for optimizing the coordinated scheduling of new energy sources proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0164] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0165] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A new energy coordinated dispatching optimization method, characterized in that, The application relates to a wind-solar power generation scheduling method and device. The method comprises the following steps: acquiring historical wind-solar power generation data, meteorological data and geographical information data; based on the preprocessed data, different wind-solar power generation scenarios are obtained through unsupervised learning technology; based on the different wind-solar power generation scenarios, a first optimization model is constructed through an adaptive adjustment strategy; the first optimization model is combined with system constraint conditions, and a first scheduling model is established by using mixed integer linear programming; 2.The new energy coordinated dispatching optimization method of claim 1, wherein: the first scheduling model is solved by using a dualization processing and column constraint generation algorithm, and an optimal scheduling strategy is output. The preprocessing comprises the following steps: the acquired data is subjected to data cleaning and standardization; the data subjected to cleaning and standardization is subjected to signal denoising processing through a variational modal decomposition technology, and a plurality of intrinsic modal function components are obtained; through variational problem solving, the original constraint problem is converted into a form easy to solve, and the number of each intrinsic modal function is obtained; 3.The new energy coordinated dispatching optimization method according to claim 1 or 2, characterized in that: the number of intrinsic modal functions is dynamically adjusted through a contour coefficient, the optimal modal number is determined, and an intrinsic modal function set is obtained. The different wind-solar power generation scenarios are obtained through the following steps: based on the intrinsic modal function set, a plurality of clustering centers are obtained through unsupervised learning technology; based on the plurality of clustering centers, optimal clustering centers and membership degree distribution are found through objective function optimization; 4. The new energy coordinated dispatching optimization method of claim 3, wherein: the optimal clustering centers and membership degree distribution are converted into specific wind-solar power generation scenarios, the clustering quality is evaluated through a contour coefficient, and different wind-solar power generation scenarios are obtained. The first optimization model is constructed through the following steps: a fuzzy set with an experience distribution as a center and a bulldozer distance as a radius is constructed by considering the uncertainty of historical data distribution; the fuzzy set radius is dynamically adjusted through the standard deviation and sample size of data; the bulldozer distance is approximately calculated by using an entropy regularization transfer algorithm; a multi-objective robust optimization framework is established; 5. The new energy coordinated dispatching optimization method of claim 4, wherein: the risk in an extreme situation is evaluated by using an improved conditional value at risk adaptive adjustment weight function. The first scheduling model is established through the following steps: power balance constraints, unit power generation upper and lower limit constraints, line power flow constraints and energy storage charging and discharging constraints are defined; based on the defined constraint conditions, physical limitations are converted into mathematical expressions through mixed integer linear programming, and are embedded into an objective function to form a complete optimization problem framework; 6. The new energy coordinated dispatching optimization method of claim 5, wherein: an uncertainty modeling mechanism is introduced, the generated wind-solar power generation scenarios and the fuzzy set are used, and a scheduling strategy is output under the worst case distribution. The first scheduling model is solved through the following steps: a distributed robust optimization problem is converted into a convex optimization problem through dualization processing; the converted convex optimization problem is iteratively solved by using a column constraint generation algorithm; 7. The new energy coordinated dispatching optimization method of claim 6, wherein: a convergence criterion is set, and when the optimal solution is converged, the optimal scheduling strategy is output. min x sup p∈p (E p [f(x,ξ)]+λ1CVaR 0.9 [g wind (x,ξ)]+λ2CVaR 0.9 [g solar (x,ξ)]) Where f(x, ξ) is the economic cost function; g wind (x, ξ) is the wind curtailment function; g solar (x, ξ) is the light curtailment function; λ is the weight coefficient of economy and environmental protection; CVaR 0.9 is the conditional value at risk.

8. A new energy coordinated dispatching optimization system, applying the method of any one of claims 1-7, characterized in that, The multi-objective robust optimization framework is represented as: The application further provides a wind-solar power generation scheduling device. The device comprises the following modules: an acquisition module, which is used for acquiring historical wind-solar power generation data, meteorological data and geographical information data; a scenario generation module, which is used for obtaining different wind-solar power generation scenarios based on preprocessed data through unsupervised learning technology; an optimization model construction module, which is used for constructing a first optimization model based on different wind-solar power generation scenarios through an adaptive adjustment strategy; The scheduling model establishing module is configured to combine the first optimization model with system constraint conditions, and establish a first scheduling model by using mixed integer linear programming. The solving module is configured to apply dualization processing and column constraint generation algorithm to solve the first scheduling model, and output an optimal scheduling strategy.

9. An electronic device, comprising: The computer program product comprises: a memory for storing a program; a processor for loading the program to perform the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium storing a program, characterized in that, The program is executed by the processor to implement the steps of the method according to any one of claims 1-7.

Images