Distributed new energy regulation and control method and system based on double-layer clustering

Abstract

The invention discloses a distributed new energy regulation and control method and system based on double-layer clustering, and the method comprises the steps: carrying out the primary clustering according to a first aggregation index of distributed new energy, and obtaining a partitioning result of the distributed new energy; carrying out primary polymerization and secondary polymerization on the distributed new energy to obtain a new energy polymerization group; generating random power generation scenes according to the power generation capacity prediction error limit of the new energy aggregation group, and constructing a random optimization scheduling model by taking minimization of the power generation cost, the new energy abandoning penalty cost and the carbon emission cost in each scene as an objective function; based on new energy aggregation constraint, equipment model constraint and power balance constraint, a solver is adopted to solve the model, a final generator set output arrangement result is obtained, the result can adapt to nonlinear output fluctuation change of a time sequence, the output prediction precision and controllability are improved, and the method is suitable for large-scale popularization and application. The problems of low power dispatching efficiency and poor effect caused by incapability of high-precision partition classification and random scene optimization dispatching in the prior art are solved.

Description

Technical Field

[0001] This application relates to the technical field of renewable energy management and power grid dispatching, and specifically provides a distributed new energy regulation method and system based on double-layer clustering. Background Art

[0002] The proportion of distributed new energy in the power grid is increasing day by day. The characteristics of its power generation, such as dispersion, intermittency, and volatility, make the regulation problem of its management objects numerous and complex, posing a great challenge to the information collection and task allocation of the power dispatching center. If a large number of distributed new energy individuals are managed and regulated one by one, it will not only lead to low management efficiency, but also cause large communication delays, seriously affecting the rapid response ability of the power grid.

[0003] In the prior art, the partition aggregation technology is used to manage the distributed new energy by partition classification. However, the existing partition aggregation technology does not fully consider the characteristic differences of distributed new energy power generation at different times and in different spaces, and does not combine the power generation characteristics of distributed new energy with indicators such as equipment performance for classification aggregation, resulting in low accuracy of the partition classification results, thus leading to low output prediction accuracy. As a result, the deterministic optimization will cause scheduling decision deviation due to fixed assumptions, and it is difficult for power dispatching to accurately grasp the power generation situation of distributed new energy. At the same time, it will also lead to the uncontrollability of distributed new energy in actual operation, increasing the risks and uncertainties of power grid operation.

[0004] Therefore, there is an urgent need for a distributed new energy regulation method to solve the problems existing in the prior art. Summary of the Invention

[0005] The purpose of this application is to address the problems that the existing partition aggregation technology cannot achieve high-precision partition classification and stochastic scenario optimal scheduling, resulting in low power dispatching efficiency and poor effect. A distributed new energy regulation method based on double-layer clustering is proposed. After the first aggregation and the second aggregation of the distributed new energy, a new energy aggregation group is obtained, which can reduce the data processing and real-time control burden of the power dispatching center and improve the system response speed. According to the power generation prediction error limit of the new energy aggregation group, a stochastic power generation scenario is generated, and a stochastic optimal scheduling model is constructed with the minimization of the power generation cost, new energy abandonment penalty cost, and carbon emission cost under each scenario as the objective function. Based on the new energy aggregation constraint, equipment model constraint, and power balance constraint, a solver is used to solve the model to obtain the final output arrangement result of the generator set. This result can adapt to the non-linear output fluctuation changes of the time series, improve the output prediction accuracy and controllability, and overcome the problems that the existing technology cannot achieve high-precision partition classification and stochastic scenario optimal scheduling, resulting in low power dispatching efficiency and poor effect.

[0006] To solve the above technical problems, according to the first aspect of the embodiments of the present application, a distributed new energy regulation method and system based on double-layer clustering are provided, including the following steps: Perform primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy; perform secondary clustering on the partitioning result according to the second aggregation index of distributed new energy to obtain new energy aggregation groups; generate random power generation scenarios for the new energy aggregation groups according to the power generation prediction error limits of the new energy aggregation groups; construct a stochastic optimization scheduling model with the objective function of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each random power generation scenario; establish new energy aggregation constraints according to the random power generation scenarios, and combine the equipment model constraints and power balance constraints to solve the stochastic optimization scheduling model using a solver to obtain the optimized scheduling result of distributed new energy.

[0007] In this solution, through primary clustering and partitioning, new energy systems with close geographical locations can be integrated into virtual power plants, significantly reducing the number of management objects; through secondary clustering, new energy aggregation groups with similar output characteristics can be obtained, further classifying and partitioning the primary clustering results, reducing the data processing and real-time control burden of the power dispatching center, improving the system response speed. The new energy aggregation groups can participate in power grid dispatching as virtual power generation units, reducing the uncertainty caused by individual output fluctuations, thereby enhancing the stability of the power system. At the same time, after secondary clustering, not only the operating characteristics of the new energy aggregation groups can be obtained, but also the output curves and fluctuation characteristics of the new energy aggregation groups can be obtained, so as to extract the power generation prediction error limits of the new energy cluster, providing basic data for generating random power generation scenarios; a stochastic optimization scheduling model is constructed according to the random power generation scenarios combined with the minimization objective function, fully considering the random volatility of new energy during power grid dispatching. Further, the optimization model is solved by integrating new energy aggregation constraints, equipment model constraints, and power balance constraints to obtain the optimal output arrangement of the generator sets, taking into account both robustness while improving the operating economy of the power system, enhancing the stability of the power grid while reducing the system operating cost.

[0008] Preferably, the performing primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy includes: the first aggregation index at least includes a spatial geographical location index and an electrical distance index; calculate the spatial geographical location index values and electrical distance index values of each node based on the topological structure of the distribution network, normalize each index value, and establish the first data set of distributed new energy; based on the first data set, use the K-means++ algorithm to perform primary clustering to obtain several initial clustering clusters of distributed new energy, and use the initial clustering clusters as the partitioning result.

[0009] Preferably, based on the first data set, perform primary clustering using the K-means++ algorithm to obtain a number of initial clustering clusters, including: determining an initial clustering center based on the first data set, and calculating the Euclidean distance between each new energy individual and the first initial clustering center; determining the probability of the corresponding new energy individual as the target clustering center according to the Euclidean distance, and taking the new energy individuals with probabilities exceeding the set threshold as the target clustering centers; when the number of the target clustering centers reaches the target value, allocating the remaining new energy individuals in the first data set to the clustering clusters corresponding to the target clustering center with the smallest Euclidean distance to them to obtain the initial clustering clusters.

[0010] Preferably, perform secondary clustering on the partitioning result according to the second aggregation index of distributed new energy to obtain new energy aggregation groups, including: determining the second aggregation index according to the spatio-temporal complementary characteristics of distributed new energy and the new energy power generation regulation performance, at least including the installed capacity index, the regulation capacity index, and the historical power generation curve index; obtaining the installed capacity index data, the regulation capacity index data, and the historical power generation curve index data of each new energy unit, normalizing each index data, and establishing a second data set of distributed new energy; based on the second data set, perform secondary clustering on the partitioning result using the K-medoids algorithm to obtain a number of new energy aggregation groups including the geographical characteristics and spatio-temporal characteristics of distributed new energy.

[0011] Preferably, generate a random power generation scenario for the new energy aggregation groups according to the power generation prediction error limit of the new energy aggregation groups, including: taking the maximum value of the distances between the data points and the clustering center points in each type of new energy aggregation group as the power generation prediction error limit of the current new energy aggregation group; obtaining the predicted power generation power of the new energy aggregation group according to the clustering center points of each type of new energy aggregation group, and calculating the random output by combining the power generation prediction error limit to generate the random power generation scenario.

[0012] Preferably, obtaining the predicted power generation power of the new energy aggregation group according to the clustering center points of each type of new energy aggregation group, and calculating the random output by combining the power generation prediction error limit to generate the random power generation scenario, including: taking the power generation curve corresponding to the clustering center points of each type of new energy aggregation group as the basic output curve of the new energy aggregation group; obtaining the predicted power generation power of the new energy aggregation group according to the basic output curve, and calculating the random output of the new energy aggregation group by integrating the power generation prediction error limit and the probability fluctuation error term; constructing a multi-dimensional scenario set according to the scenario probabilities corresponding to each new energy aggregation group when reaching the random output to obtain the random power generation scenario.

[0013] Preferably, a stochastic optimization scheduling model is constructed with the objective function of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario, including: establishing the objective function based on the minimum value of the sum of the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario, so as to construct a stochastic optimization scheduling model.

[0014] Preferably, the construction of the stochastic optimization scheduling model with the objective function of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario further includes: calculating the power generation cost of the system under each stochastic power generation scenario according to the probability of occurrence of the stochastic power generation scenario, combined with the new energy power generation cost and the power generation power of the new energy under the stochastic power generation scenario; calculating the new energy abandonment penalty cost of the system under each stochastic power generation scenario according to the unit penalty cost of the aggregated new energy abandonment and the power generation amount of the aggregated new energy under the stochastic power generation scenario; calculating the carbon emission cost of the system under each stochastic power generation scenario according to the carbon emission cost of the new energy and the carbon emission factor.

[0015] Preferably, the new energy aggregation constraint includes the output constraint of the new energy aggregation group under the stochastic power generation scenario, specifically including: the available power generation power constraint of the new energy aggregation group under the stochastic power generation scenario and the ramp-up and ramp-down ability constraint of the distributed new energy monomers in the new energy aggregation group under the stochastic power generation scenario.

[0016] In a second aspect, an embodiment of the present application provides a distributed new energy regulation system based on double-layer clustering. The system includes: a primary clustering module for performing primary clustering according to the first aggregation index of the distributed new energy to obtain the partitioning result of the distributed new energy; a secondary clustering module for performing secondary clustering on the partitioning result according to the second aggregation index of the distributed new energy to obtain a new energy aggregation group; a stochastic power generation module for generating stochastic power generation scenarios of the new energy aggregation group according to the power generation prediction error limit of the new energy aggregation group; a model construction module for constructing a stochastic optimization scheduling model with the objective function of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario; and an optimal scheduling module for establishing new energy aggregation constraints according to the stochastic power generation scenarios, and solving the stochastic optimization scheduling model by using a solver in combination with equipment model constraints and power balance constraints to obtain the optimal scheduling result of the distributed new energy.

[0017] Advantages of the present application: 1. Perform primary clustering through the K-means++ algorithm and secondary clustering through the K-medoids algorithm to obtain a more accurate partitioning result, thereby more effectively reducing the number of managed objects, decreasing the data processing and real-time control burden on the power dispatching center, and enhancing the system response speed. Among them, when performing secondary clustering, the historical power generation curve is used as one of the classification indicators, which can more accurately classify new energy according to the power generation characteristics of new energy, and further ensure that the aggregation based on partition classification can characterize the characteristic differences of distributed new energy power generation at different times and in different spaces, realizing the aggregation of distributed new energy with complementary regulation performance, flexibly adapting to the non-linear changes of time series, significantly improving the accuracy and controllability of power output prediction, and providing robust support for the dispatching of distributed new energy. 2. Obtain the typical power output curve and fluctuation characteristics of the new energy cluster through the aggregation process, and then generate corresponding stochastic power generation scenarios to characterize the uncertainty of new energy power output. By considering the impact of new energy uncertainty in the stochastic optimization model, comprehensively considering economy and volatility, and through the coordinated dispatching of the new energy cluster and thermal power units, the operating cost is reduced while the power grid stability is enhanced. Compared with traditional decentralized management, this method makes full use of spatio-temporal correlation, optimizes resource allocation, and significantly improves the efficient management of distributed new energy in the distribution network. 3. Generate stochastic scenarios based on the prediction error limit, which can accurately simulate the randomness and volatility of new energy, and then realize the incorporation of multi-time scale and multi-dimensional random variables into the optimization model, avoiding the decision-making deviation caused by fixed assumptions in deterministic optimization, and ensuring that the optimization result is closer to the actual operating conditions. Description of the Drawings

[0018] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, purposes, and advantages of this application will become more obvious. The drawings are only for the purpose of showing the preferred embodiments and are not considered as a limitation of this application. Moreover, throughout the drawings, the same reference numerals are used to represent the same components.

[0019] Figure 1 It is a flowchart of a distributed new energy regulation method based on double-layer clustering provided by an embodiment of this application.

[0020] Figure 2 It is a flowchart of double-layer clustering provided by an embodiment of this application.

[0021] Figure 3 It is a schematic diagram of the modules of a distributed new energy regulation system based on double-layer clustering provided by an embodiment of this application. Detailed Embodiments

[0022] To make the objectives, technical solutions and advantages of this application more clear and understandable, the following further elaborates on this application in combination with the accompanying drawings and embodiments. It should be understood that the specific implementation manners described herein are only one of the best embodiments of this application, which is only used to explain this application and does not limit the protection scope of this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of this application.

[0023] Embodiment 1: As Figure 1 shown, a distributed new energy regulation method based on double-layer clustering includes the following steps: S101. Perform primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy.

[0024] In this embodiment, by performing primary aggregation partitioning on distributed new energy, new energy systems with close geographical locations can be integrated into virtual power plants, significantly reducing the number of management objects, facilitating real-time data upload and dispatching command issuance, reducing communication latency, and enhancing the grid's rapid response ability, thereby supporting the efficient information collection and task allocation of the power dispatching center.

[0025] Furthermore, the power generation characteristics of distributed new energy are highly correlated with its geographical location. New energy systems in the same region are affected by similar light intensity and weather conditions, and their power output waveforms are similar. After aggregation, the output characteristics of the regional new energy group are more regular, facilitating power system prediction and optimization management. Compared with decentralized management, aggregation partitioning can make full use of spatial correlation and enhance the unified optimization ability.

[0026] Specifically, S101 includes: The first aggregation index includes at least a spatial geographical location index and an electrical distance index; Based on the topological structure of the distribution network, calculate the spatial geographical location index values and electrical distance index values of each node, normalize each index value, and establish the first data set of distributed new energy; Based on the first data set, use the K-means++ algorithm to perform primary clustering to obtain several initial clustering clusters of distributed new energy, and use the initial clustering clusters as the partitioning result.

[0027] Specifically, based on the first data set, using the K-means++ algorithm to perform primary clustering to obtain several initial clustering clusters includes: Determine the initial clustering centers based on the first data set, and calculate the Euclidean distance between each new energy individual and the first initial clustering center; Determine the probability of the corresponding new energy individual as the target clustering center according to the Euclidean distance, and take the new energy individuals with probabilities exceeding the set threshold as the target clustering centers; When the number of the target clustering centers reaches the target value, allocate the remaining new energy individuals in the first dataset to the clustering clusters corresponding to the target clustering center with the smallest Euclidean distance from them, to obtain the initial clustering clusters.

[0028] In some embodiments, as Figure 2 shown, during the first clustering, partition aggregation is mainly performed according to the spatial characteristics of distributed new energy. Its spatial features include but are not limited to spatial geographical location and electrical distance. In this embodiment, the spatial geographical location and electrical distance are used as the first aggregation indicators for the first clustering, and the sensitivity of the voltage change amplitude ΔUi of the grid topological node i relative to the load injection change amount at node j is measured according to the electrical distance indicator. The strength of the electrical connection between nodes can be effectively characterized by the electrical distance indicator value, so as to reflect the mutual correlation between nodes in terms of energy transmission and communication levels, and ensure the accuracy and reliability of the first clustering partition.

[0029] In some examples, the specific steps for partitioning and clustering distributed new energy based on the k-means++ algorithm are as follows: Step 1: Collect the topological parameters of the distribution network, calculate the physical distances (i.e., obtain the spatial geographical location indicator values) and electrical distance indicator values of each node, and establish the first dataset after normalization processing , where each feature vector in the dataset is composed of the spatial geographical location indicator value and the electrical distance indicator value, and is expressed as ; Step 2: Preset the number K1 of target clustering centers according to the grid construction requirements or using the elbow method; Step 3: Randomly select an individual from the first dataset as the initial clustering center ; Step 4: For each new energy individual , calculate its Euclidean distance from the initial clustering center , and determine the probability of being selected as the target clustering center according to the Euclidean distance value. The probability is proportional to the distance, so as to ensure that the distribution of new center points is more dispersed; repeat this step to obtain K1 target clustering centers; Step 5: After the clustering centers are selected, for the remaining objects, according to the Euclidean distance D, allocate each distributed new energy individual to the clustering cluster corresponding to the nearest center point, and iteratively update the center point until the clustering result converges or reaches the maximum number of iterations, to obtain several initial clustering clusters, that is, the partition result of the first clustering of distributed new energy.

[0030] Further, normalizing the spatial geographical location index value and the electrical distance index value can ensure the fairness and consistency of calculations when there are significant differences in the dimensions and numerical ranges of different index data, standardizing each index to a unified range, such as [0, 1]. The normalization calculation method is as follows: (1) Where: respectively represent the values of the normalized index, is the original index value, and respectively represent the maximum and minimum values in the original index value.

[0031] In this embodiment, using the k-means++ algorithm with distance metric for one clustering can obtain center points with more uniform distribution and greater spacing, so as to improve the clustering effect and convergence stability. This improvement effectively reduces the local optimum problem that may be caused by the random selection of the initial center point, thereby improving the accuracy and robustness of distributed new energy resource clustering.

[0032] S102. Perform secondary clustering on the partition result according to the second aggregation index of distributed new energy to obtain a new energy aggregation group.

[0033] In this embodiment, the secondary clustering is a two-layer aggregation of the partition result of the primary clustering, which can identify new energy clusters with similar power characteristics, thereby improving the predictability and controllability of new energy power generation.

[0034] It can be understood that the K-medoids algorithm, as an optimization algorithm based on time series clustering and Dynamic Time Warping (DTW) distance, uses the DTW distance to measure the similarity between time series data of different new energy power plants. Compared with the traditional Euclidean distance, DTW can more effectively adapt to time shifts and non-linear deformations in time series, thereby improving the accuracy of clustering analysis and realizing the efficient management and real-time scheduling of distributed new energy systems.

[0035] Specifically, the S102 includes: Determine the second aggregation index according to the spatio-temporal complementary characteristics of distributed new energy and the new energy power generation regulation performance, including at least the installed capacity index, the regulation capacity index, and the historical power generation curve index; Obtain the installed capacity index data, the regulation capacity index data, and the historical power generation curve index data of each new energy unit, normalize each index data, and establish a second data set of distributed new energy; Based on the second dataset, the K-medoids algorithm is used to perform secondary clustering on the partitioning results to obtain several new energy aggregation groups that include the geographical characteristics and spatio-temporal characteristics of distributed new energy.

[0036] In some embodiments, as Figure 2 shown, the installed capacity, regulation capacity, and historical power generation curve are used as the second aggregation indicators. After collecting the installed capacity, regulation capacity, and historical power generation curve data of each new energy unit and normalizing each data, a second dataset is established. The normalization method for the second aggregation indicators is the same as that for the first aggregation indicators in the above primary clustering. Then, the K-medoids algorithm is used for secondary clustering, and the specific steps are as follows: A1. Randomly select several distinct data points from the second dataset as the initial aggregation points to obtain an initial aggregation point set; A2. Assign the data points to the nearest initial aggregation point. For the remaining data points in the second dataset, calculate the DTW distance between them and all the initial aggregation points. For example , where represents the remaining data points in the second dataset, represents the initial aggregation points in the initial aggregation point set, and is assigned to the cluster corresponding to the nearest ; A3. Recalculate the total distance between the points within each cluster and other points, and select the point that minimizes the total distance as the new target initial aggregation point to optimize the central representativeness of the cluster; A4. Repeat steps A2 and A3 until the initial aggregation point set no longer changes for two consecutive iterations or reaches a predetermined number of iterations; A5. Based on the DTW distance, assign each data point to the cluster with the minimum distance to obtain several new energy aggregation clusters, and use each new energy aggregation cluster as a new energy aggregation group.

[0037] Further, in step A2, the distance between each data point and the initial aggregation point is the DTW distance, and the DTW distance calculation method is as follows: For any two time series and , initialize a matrix , represents the DTW distance between the first y elements of and the first l elements of ; Use dynamic programming to fill the matrix : , , That is and the DTW distance between them, is the initial aggregation point of the data point and the DTW distance of .

[0038] Among them, the installed capacity can directly reflect the power generation potential and scale effect, and is crucial for the assessment of the total output in the spatio-temporal complementary analysis. The regulation ability reflects the flexibility of new energy individuals in operation and the response ability to grid dispatching, and is the core parameter for measuring the regulation performance, which is particularly critical in dealing with load fluctuations and new energy consumption. The historical power generation curve integrates the output characteristics in the time dimension, contains information such as the seasonality, daily variation and weather dependence of new energy power generation, and can reflect the spatio-temporal distribution law of distributed new energy.

[0039] In this embodiment, through the selection of the initial aggregation point and the calculation based on the DTW distance, it can be ensured that the time series data of each distributed new energy is accurately assigned to the cluster that is most similar to its behavior pattern, thus laying a foundation for subsequent optimization and management, and improving the operation efficiency and dispatching performance of the new energy system.

[0040] In some other embodiments, based on the first-layer clustering (i.e., the first clustering) and the second-layer clustering (i.e., the second clustering), the distributed new energy resources can be divided into multiple categories according to their geographical characteristics and spatio-temporal characteristics, and then a new energy aggregation model can be constructed according to the new energy aggregation groups of different categories. This model includes key parameters such as the total capacity and regulation ability of the new energy cluster after new energy aggregation. Furthermore, through the aggregation process of distributed new energy, the number of dispatching decision variables can be effectively reduced, thereby enhancing the feasibility and practicality of high-proportion distributed new energy participating in the dispatching optimization of the power system.

[0041] Furthermore, the constraints on the total capacity and regulation ability of the new energy cluster after new energy aggregation in the new energy aggregation model are represented as follows: (2) (3) Among them, is the output of the m-th new energy aggregation group, is the maximum output of the m-th new energy aggregation group, is the maximum output of the distributed new energy monomer d belonging to the m-th new energy, is the ramp rate (i.e., the maximum regulation capacity) of the new energy monomer d.

[0042] S103. Generate the random power generation scenarios of the new energy aggregation group according to the prediction error limit of the power generation of the new energy aggregation group.

[0043] Specifically, the S103 includes: Taking the maximum distance between the data points and the clustering center point in each type of new - energy aggregation group as the power generation prediction error limit of the current new - energy aggregation group; Obtaining the predicted power generation of the new - energy aggregation group according to the clustering center points of each type of new - energy aggregation group, calculating the random output by combining the power generation prediction error limit, so as to generate the random power generation scenario.

[0044] Specifically, the step of obtaining the predicted power generation of the new - energy aggregation group according to the clustering center points of each type of new - energy aggregation group, calculating the random output by combining the power generation prediction error limit, so as to generate the random power generation scenario includes: Taking the power generation curve corresponding to the clustering center point of each type of new - energy aggregation group as the basic output curve of the new - energy aggregation group; Obtaining the predicted power generation of the new - energy aggregation group according to the basic output curve, and calculating the random output of the new - energy aggregation group by fusing the power generation prediction error limit and the probability fluctuation error term; Constructing a multi - dimensional scenario set according to the scenario probabilities corresponding to the random output of each new - energy aggregation group, so as to obtain the random power generation scenario.

[0045] In some embodiments, as Figure 2 shown, by secondary clustering, the clustering center point of each new - energy aggregation group is extracted, and the power generation curve corresponding to this clustering center point can be regarded as the typical daily power generation curve of this new - energy cluster, so as to obtain the new - energy output prediction curve; in this type of new - energy aggregation group, the distance between the data point with the farthest distance and the clustering center point, for example , can be used as the power generation prediction error limit of this aggregation group, and then a random power generation scenario can be generated based on the new - energy output prediction curve and the power generation prediction error limit to simulate the uncertainty of new - energy output.

[0046] In some embodiments, taking the new - energy output prediction curve as the basic output curve, extracting the predicted power generation of the new - energy aggregation group according to the basic output curve, and the calculation formula is as follows: (4) Wherein, represents the set of predicted values of new - energy power generation, represents the predicted power generation of the new - energy aggregation group b at time t.

[0047] Furthermore, a probability fluctuation error term is introduced into the predicted power generation at each moment based on the power generation prediction error limit, and the specific formula is as follows: (5) Among them, represents the output of the randomly generated new energy aggregation group at time t, represents an error term that follows a normal distribution, and its value range is between 0 and 1. represents the power generation prediction error limit of this category of new energy aggregation group.

[0048] Further, calculate the probability of the random power generation scenario according to the introduced probability fluctuation error term: (6) Among them, represents the error term corresponding occurrence probability, the probability of scenario s occurring.

[0049] Further, by repeatedly executing the above steps of extracting the predicted power generation power of the new energy aggregation group, the probability fluctuation error term, and calculating the corresponding probability of the random power generation scenario, a number of random power generation scenario samples and their corresponding probabilities are obtained, and then the probabilities are normalized. The processing formula is as follows: (7) Among them, represents the normalized value of the probability of scenario s occurring.

[0050] It can be understood that in the process of power system dispatching control, generally, the predicted output curve of new energy is used as the upper limit of new energy power generation capacity, and then dispatching optimization is carried out. However, due to the volatility of new energy, there are often certain errors in the predicted output curve. Traditional methods often rely on the "predicted situation" to formulate plans, ignoring the risks that may be brought by extreme situations, so it is difficult to ensure the reliability and robustness of the dispatching plan. Therefore, in this embodiment, a random optimization method is used to generate random power generation scenarios by combining the power generation prediction error characteristics obtained by the above classification, so as to more accurately simulate the randomness and volatility of new energy. Further, the uncertainty is described by the new energy predicted output scenario set, and multi-time scale and multi-dimensional random variables are incorporated into the optimization model to avoid decision-making biases caused by fixed assumptions in deterministic optimization, making the optimization scheme closer to the actual operation situation.

[0051] S104. With the goal of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each of the random power generation scenarios, construct a random optimization scheduling model.

[0052] Specifically, the S104 includes: Establish the objective function based on the minimum value of the sum of the power generation cost, the new energy abandonment penalty cost, and the carbon emission cost under each of the random power generation scenarios, so as to construct a random optimization scheduling model.

[0053] Specifically, the S104 further includes: Calculating the power generation cost of the system in each of the random power generation scenarios according to the probability of occurrence of the random power generation scenario, combined with the new energy power generation cost and the power generation power of the new energy in the random power generation scenario; Calculating the new energy abandonment penalty cost of the system in each of the random power generation scenarios according to the abandonment unit penalty cost of the aggregated new energy and the power generation amount of the aggregated new energy in the random power generation scenario; Calculating the carbon emission cost of the system in each of the random power generation scenarios according to the carbon emission cost of the new energy and the carbon emission factor.

[0054] As an alternative implementation manner, the calculation formula of the objective function can be expressed as: (8) In the formula, is the objective function, is the power generation cost, is the new energy abandonment penalty cost, is the carbon emission cost.

[0055] In this embodiment, the power generation cost includes, but is not limited to, the power generation cost of thermal power units. A stochastic optimal scheduling model is constructed by combining the power generation cost of the new energy aggregation group, the new energy abandonment penalty cost, and the carbon emission cost, which fully considers the operation economy of the power system and the energy utilization efficiency. While improving the operation economy of the power system, it takes into account robustness, enhances the grid stability while reducing the system operation cost, realizes the optimization of renewable energy output, the improvement of operation efficiency, and the improvement of energy supply services, so as to adapt to the increase of renewable energy penetration rate and the dynamic change of energy demand.

[0056] Furthermore, the calculation formulas of the power generation cost, the new energy abandonment penalty cost, and the carbon emission cost can be expressed as: (9) In the formula, is the number of the random power generation scenarios, is the time, is the number of grid nodes, is the number of new energy categories, is the probability of occurrence of the random power generation scenario s, is the unit power generation cost of the thermal power unit, is the active power generation of the thermal power unit at node i at time t in the random power generation scenario s, is the start-up cost of the thermal power unit, is the shutdown cost of the thermal power unit, is the penalty cost per unit of abandoned power of aggregated new energy of type m at node i at time t. is the actual power generation of aggregated new energy of type m at node i at time t under the random power generation scenario s. is the maximum power generation of aggregated new energy of type m at node i at time t under the random power generation scenario s. is the carbon emission cost of distributed new energy at node i at time t. is the carbon emission factor of the generator at node i.

[0057] In this embodiment, the costs of different types of new energy power generation are compared and analyzed according to the power generation cost, and the power generation methods and combinations with lower costs are preferentially selected to ensure that the entire power generation system reduces the power generation cost as much as possible while meeting the power demand during the power grid dispatching process; at the same time, considering the penalty cost of new energy abandonment, the characteristics of intermittent and unstable new energy power generation are fully considered, further improving the economy of new energy power generation, reasonably allocating power generation resources, and improving energy utilization efficiency; then the carbon emission cost is incorporated into the objective function to reduce the proportion of traditional energy power generation with high carbon emissions. By comprehensively considering these three aspects, an objective function is established and a stochastic optimization scheduling model considering the random power generation scenario of the new energy aggregation group is constructed to obtain the optimal processing arrangement of each unit within the aggregation group, significantly improving the new energy scheduling efficiency and reliability.

[0058] S105. Establish new energy aggregation constraints according to the random power generation scenario, and use a solver to solve the stochastic optimization scheduling model in combination with equipment model constraints and power balance constraints to obtain the optimal scheduling result of distributed new energy.

[0059] Specifically, the new energy aggregation constraints include the output constraints of the new energy aggregation group under the random power generation scenario, specifically including: The available power generation constraint of the new energy aggregation group under the random power generation scenario and the ramp rate constraint of the distributed new energy monomers in the new energy aggregation group under the random power generation scenario.

[0060] As an optional implementation manner, the available power generation constraint and the ramp rate constraint are expressed as: (10) (11) In the formula, is the available power generation of the m - type new energy aggregation group at time t under the random power generation scenario s. is the available power generation of the m - type new energy aggregation group at time t + 1 under the random power generation scenario s. is the random output of the m - type new energy aggregation group at time t under the random power generation scenario s. The ramping ability of the distributed new energy monomer d of the m - type new energy aggregation group under the random power generation scenario s.

[0061] In some embodiments, the new energy aggregation constraint is specifically the constraint on the overall capacity and regulation ability of the new energy cluster after aggregation of new energy under the corresponding random power generation scenario generated above. Among them, Equation (10) represents the range of power generation that the new energy aggregation group can achieve, and Equation (11) represents the power regulation ability of the new energy aggregation group.

[0062] In this embodiment, through the constraints on the power generation range and power regulation, the intermittency and volatility during the new energy power generation process are overcome, so that the output change of the new energy aggregation group remains within a reasonable range, improving the operation efficiency and safety stability of the equipment. At the same time, it helps to flexibly adjust the output of new energy according to the constraint range to quickly respond to system requirements, enhancing the power system's ability to cope with the uncertainty of the new energy aggregation group and further improving the overall dispatching efficiency and flexibility of the system.

[0063] Specifically, the equipment model constraint includes the thermal power unit power generation model constraint, which is as follows: (12) (13) (14) (15) (16) (17) In the formula, is the upper limit of the active power generation of the thermal power unit at node i, is the lower limit of the active power generation of the thermal power unit at node i; is the reactive power generation of the thermal power unit at node i, is the upper limit of the reactive power generation of the thermal power unit at node i, is the lower limit of the reactive power generation of the thermal power unit at node i; is the state variable, taking values of 0 and 1, representing the working state of the thermal power unit. Among them, 1 represents that the thermal power unit is in the working state, and 0 represents that the thermal power unit is in the shutdown state; is the upper limit of the ramping ability of the thermal power unit at node i, is the lower limit of the ramping ability of the thermal power unit at node i; is the minimum continuous operation time of the thermal power unit at node i, is the minimum continuous shutdown time of the thermal power unit at node i; is the working state of the thermal power unit at any moment under the minimum continuous shutdown time, It is the working state of the thermal power unit at any moment under the minimum continuous operation time.

[0064] In this embodiment, the power generation power range of the new energy aggregation group is restricted by equations (12) and (13) to ensure the balance between power supply and demand in the power system; the regulation range of the power generation power is restricted by equations (14) and (15), so that the thermal power unit has a certain rapid regulation ability under the power fluctuation in the new energy random power generation scenario to make up for the fluctuation of new energy power generation, ensuring that the power system always operates safely and stably during the scheduling period. At the same time, the thermal power unit is regulated within the economic operation range to reduce the power generation cost; the start-stop operation of the thermal power unit is restricted by equations (16) and (17). Since it takes a certain time for the thermal power unit to reach full load operation from startup, and there are also time limits for restarting after shutdown, therefore, the start-stop of the thermal power unit can be preset in advance according to this constraint condition, combined with the prediction of new energy power generation, to optimize the overall operation mode of the system and improve the operation efficiency and reliability of the power system.

[0065] Specifically, the power balance constraint includes the following expressions: (18) (19) (20) (21) (22) (23) (24) Wherein, and respectively represent the active power and reactive power transmitted in the branch from node i to node j under the random power generation scenario s, and respectively represent the active power and reactive power injected at node j under the random power generation scenario s, and respectively represent the resistance and reactance of line ij, and respectively represent the current flowing from node i to node j and the square of the current value under the random power generation scenario s, and respectively represent the voltage amplitude and the square of the voltage amplitude at node i at time t under the random power generation scenario s, represents the voltage amplitude at node j at time t under the random power generation scenario s, and respectively represent the upper limit and the lower limit of the voltage amplitude, represents the upper limit of the current amplitude of line ij, represents the set of branches with node j as the head end, represents the active power transmitted in each branch in represents the reactive power transmitted in each branch in.

[0066] In this embodiment, the above power balance constraint restricts the adjustment relationship among the power flow, current, and terminal voltages transmitted in the line, ensuring that the power injection and outflow at each node are balanced, ensuring the power transmission security of each new energy aggregation group under the stochastic power generation scenario, and guaranteeing the economic and stable operation of the power system.

[0067] In this embodiment, by comprehensively considering the above constraint conditions to solve the stochastic optimization scheduling model, the optimal output curves and start-stop arrangements of the new energy cluster and thermal power units under the stochastic scenario of new energy can be obtained. For the internal control of the new energy cluster, only by proportionally allocating the output plan according to the capacity of each unit, the optimal output arrangement of each unit within the cluster can be obtained, which not only satisfies the efficient utilization of energy of the new energy aggregation group but also ensures the safe and stable operation of the power system.

[0068] Based on the same inventive concept, an embodiment of the present application also provides a distributed new energy regulation system based on double-layer clustering corresponding to the distributed new energy regulation method based on double-layer clustering, as Figure 3 shown. The system includes: A primary clustering module 301, configured to perform primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy; A secondary clustering module 302, configured to perform secondary clustering on the partitioning result according to the second aggregation index of distributed new energy to obtain a new energy aggregation group; A stochastic power generation module 303, configured to generate a stochastic power generation scenario of the new energy aggregation group according to the power generation prediction error limit of the new energy aggregation group; A model construction module 304, configured to construct a stochastic optimization scheduling model with the objective of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario; An optimal scheduling module 305, configured to establish new energy aggregation constraints according to the stochastic power generation scenario, and solve the stochastic optimization scheduling model by using a solver in combination with equipment model constraints and power balance constraints to obtain the optimal scheduling result of distributed new energy.

[0069] As an alternative implementation, the primary clustering module 301 is specifically configured to: the first aggregation metric at least includes a spatial geographical location metric and an electrical distance metric; calculate the spatial geographical location metric value and the electrical distance metric value of each node based on the topological structure of the distribution network, normalize each metric value, and establish a first data set of distributed new energy sources; based on the first data set, perform primary clustering using the K-means++ algorithm to obtain several initial clustering clusters of the distributed new energy sources, and use the initial clustering clusters as the partitioning result.

[0070] Furthermore, the primary clustering module 301 is specifically further configured to: determine an initial clustering center based on the first data set, and calculate the Euclidean distance between each new energy individual and the first initial clustering center; determine the probability of the corresponding new energy individual as the target clustering center according to the Euclidean distance, and use the new energy individuals with probabilities exceeding the set threshold as the target clustering centers; when the number of the target clustering centers reaches the target value, allocate the remaining new energy individuals in the first data set to the clustering clusters corresponding to the target clustering center with the smallest Euclidean distance to them to obtain the initial clustering clusters.

[0071] As an alternative implementation, the secondary clustering module 302 is specifically configured to: determine the second aggregation metric according to the spatio-temporal complementary characteristics of the distributed new energy sources and the new energy power generation regulation performance, at least including an installed capacity metric, a regulation capacity metric, and a historical power generation curve metric; obtain the installed capacity metric data, the regulation capacity metric data, and the historical power generation curve metric data of each new energy unit, normalize each metric data, and establish a second data set of the distributed new energy sources; based on the second data set, perform secondary clustering on the partitioning result using the K-medoids algorithm to obtain several new energy aggregation groups including the geographical characteristics and spatio-temporal characteristics of the distributed new energy sources.

[0072] As an alternative implementation, the random power generation module 303 is specifically configured to: use the maximum value of the distance between the data points and the clustering center points in each type of new energy aggregation group as the power generation prediction error limit of the current new energy aggregation group; obtain the predicted power generation of the new energy aggregation group according to the clustering center points of each type of new energy aggregation group, and calculate the random output in combination with the power generation prediction error limit to generate the random power generation scenario.

[0073] Furthermore, the random power generation module 303 is specifically further configured to: use the power generation curve corresponding to the clustering center points of each type of new energy aggregation group as the basic output curve of the new energy aggregation group; obtain the predicted power generation of the new energy aggregation group according to the basic output curve, and calculate the random output of the new energy aggregation group by fusing the power generation prediction error limit and the probability fluctuation error term; Construct a multi-dimensional scenario set based on the scenario probabilities corresponding to each new energy aggregation group reaching the stochastic output, so as to obtain the stochastic power generation scenario.

[0074] As an optional implementation manner, the model construction module 304 is specifically configured to: establish the objective function with the minimum value of the sum of the power generation cost, the new energy abandonment penalty cost, and the carbon emission cost in each of the stochastic power generation scenarios as the target, so as to construct a stochastic optimal scheduling model.

[0075] Further, the model construction module 304 is specifically further configured to: calculate the power generation cost of the system in each of the stochastic power generation scenarios according to the probability of occurrence of the stochastic power generation scenario, combined with the new energy power generation cost and the power generation power of the new energy in the stochastic power generation scenario; Calculate the new energy abandonment penalty cost of the system in each of the stochastic power generation scenarios according to the abandonment unit penalty cost of the aggregated new energy and the power generation amount of the aggregated new energy in the stochastic power generation scenario; Calculate the carbon emission cost of the system in each of the stochastic power generation scenarios according to the carbon emission cost of the new energy and the carbon emission factor.

[0076] The above specific implementation manners are the preferred implementation manners of the present application. The specific implementation scope of the present application is not limited thereby. The scope of the present application includes but is not limited to this specific implementation manner. All equivalent changes made according to the shape, structure, and method of the present application are within the protection scope of the present application.

Claims

1. A distributed new energy regulation method based on double-layer clustering, characterized in that: It includes the following steps: Perform primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy; Perform secondary clustering on the partitioning result according to the second aggregation index of distributed new energy to obtain new energy aggregation groups; Generate random power generation scenarios for the new energy aggregation groups according to the power generation prediction error limits of the new energy aggregation groups; Construct a stochastic optimization scheduling model with the objective function of minimizing the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each of the random power generation scenarios; Establish new energy aggregation constraints according to the random power generation scenarios, and use a solver to solve the stochastic optimization scheduling model in combination with equipment model constraints and power balance constraints to obtain the optimized scheduling result of distributed new energy.

2. The distributed new energy regulation method based on double-layer clustering according to claim 1, wherein: The step of performing primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy includes: The first aggregation index at least includes a spatial geographical location index and an electrical distance index; Calculate the spatial geographical location index values and electrical distance index values of each node based on the topological structure of the distribution network, normalize each index value, and establish a first data set of distributed new energy; Based on the first data set, perform primary clustering using the K-means++ algorithm to obtain several initial clustering clusters of distributed new energy, and use the initial clustering clusters as the partitioning result.

3. The distributed new energy regulation method based on double-layer clustering according to claim 2, characterized in that: The step of performing primary clustering using the K-means++ algorithm based on the first data set to obtain several initial clustering clusters includes: Determine the initial clustering centers based on the first data set, and calculate the Euclidean distance between each new energy individual and the first initial clustering center; Determine the probability of the corresponding new energy individual as the target clustering center according to the Euclidean distance, and use the new energy individuals with probabilities exceeding the set threshold as the target clustering centers; When the number of the target clustering centers reaches the target value, allocate the remaining new energy individuals in the first data set to the clustering clusters corresponding to the target clustering centers with the smallest Euclidean distance to them to obtain the initial clustering clusters.

4. The distributed new energy regulation method based on double-layer clustering according to claim 1, characterized in that: The step of performing secondary clustering on the partitioning result according to the second aggregation index of distributed new energy to obtain new energy aggregation groups includes: Determine the second aggregation index according to the spatio-temporal complementarity characteristics of distributed new energy and the new energy power generation regulation performance, which at least includes an installed capacity index, a regulation ability index, and a historical power generation curve index; Obtain the installed capacity index data, regulation ability index data, and historical power generation curve index data of each new energy unit, normalize each index data, and establish a second data set of distributed new energy; Based on the second data set, use the K-medoids algorithm to perform secondary clustering on the partitioning result to obtain several new energy aggregation groups including the geographical characteristics and spatio-temporal characteristics of distributed new energy.

5. The distributed new energy regulation method based on double-layer clustering according to claim 4, characterized in that: The step of generating random power generation scenarios for the new energy aggregation groups according to the power generation prediction error limits of the new energy aggregation groups includes: Take the maximum value of the distances between the data points and the clustering center points in each type of new energy aggregation group as the power generation prediction error limit of the current new energy aggregation group; Obtain the predicted power generation of the new energy aggregation group according to the clustering center points of various new energy aggregation groups, and calculate the stochastic output by combining the power generation prediction error limit to generate the stochastic power generation scenario.

6. The distributed new energy regulation method based on double-layer clustering according to claim 5, characterized in that: The obtaining the predicted power generation of the new energy aggregation group according to the clustering center points of various new energy aggregation groups, and calculating the stochastic output by combining the power generation prediction error limit to generate the stochastic power generation scenario includes: Use the power generation curves corresponding to the clustering center points of various new energy aggregation groups as the basic output curves of the new energy aggregation groups; Obtain the predicted power generation of the new energy aggregation group according to the basic output curve, and calculate the stochastic output of the new energy aggregation group by fusing the power generation prediction error limit and the probability fluctuation error term; Construct a multi-dimensional scenario set according to the scenario probabilities corresponding to each new energy aggregation group when reaching the stochastic output to obtain the stochastic power generation scenario.

7. The distributed new energy regulation method based on double-layer clustering according to claim 1, wherein: Taking the minimization of the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario as the objective function to construct a stochastic optimal scheduling model includes: Establish the objective function based on the minimum value of the sum of the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario to construct a stochastic optimal scheduling model.

8. The distributed new energy regulation method based on double-layer clustering according to claim 7, characterized in that: Taking the minimization of the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario as the objective function to construct a stochastic optimal scheduling model further includes: Calculate the power generation cost of the system under each stochastic power generation scenario according to the probability of occurrence of the stochastic power generation scenario, combined with the new energy power generation cost and the power generation power of new energy under the stochastic power generation scenario; Calculate the new energy abandonment penalty cost of the system under each stochastic power generation scenario according to the unit penalty cost of abandoned aggregated new energy and the power generation amount of aggregated new energy under the stochastic power generation scenario; Calculate the carbon emission cost of the system under each stochastic power generation scenario according to the carbon emission cost of new energy and the carbon emission factor.

9. The distributed new energy regulation method based on double-layer clustering according to claim 5, characterized in that: The new energy aggregation constraint includes the output constraint of the new energy aggregation group under the stochastic power generation scenario, specifically including: The available power generation power constraint of the new energy aggregation group under the stochastic power generation scenario and the ramp-up capacity constraint of the distributed new energy monomers of the new energy aggregation group under the stochastic power generation scenario.

10. A distributed new energy regulation system based on double-layer clustering, characterized in that: Applicable to the distributed new energy regulation method based on double-layer clustering as described in any one of claims 1-9 above, including: A primary clustering module for performing primary clustering according to the first aggregation index of distributed new energy to obtain the partitioning result of distributed new energy; A secondary clustering module for performing secondary clustering on the partitioning result according to the second aggregation index of distributed new energy to obtain new energy aggregation groups; A stochastic power generation module for generating the stochastic power generation scenario of the new energy aggregation group according to the power generation prediction error limit of the new energy aggregation group; A model construction module for constructing a stochastic optimal scheduling model with the minimization of the power generation cost, new energy abandonment penalty cost, and carbon emission cost of the system under each stochastic power generation scenario as the objective function; An optimization scheduling module is used to establish new energy aggregation constraints according to the stochastic power generation scenario, and solve the stochastic optimization scheduling model by using a solver in combination with equipment model constraints and power balance constraints to obtain the optimization scheduling results of distributed new energy.

Images