High-proportion new energy power grid energy management method, device, system and equipment based on digital twinning

Abstract

The invention relates to a high-proportion new energy power grid energy management method, device, system and equipment based on digital twinning. The method comprises the steps of obtaining a sampling budget and an initial iteration frequency, and obtaining n pieces of real-time working condition information and m power dispatching decisions of high-proportion new energy power grid operation; performing iteration processing according to the sampling budget, the initial iteration times, the n pieces of real-time working condition information and the m power dispatching decisions to obtain k groups of dispatching decision data; performing clustering analysis and optimization processing on each simulation data and each prediction result of each group of scheduling decision data to obtain an optimal power scheduling decision; and screening out the optimal power dispatching decision with the maximum corresponding posteriori mean value from all the optimal power dispatching decisions as the optimal power dispatching scheme for the operation of the high-proportion new energy power grid. According to the method, under the condition that the accurate simulation difficulty of an extremely-high-proportion new energy power grid is high, good electric power energy dispatching and regulation are achieved with low communication cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power dispatching of power distribution network, and in particular to a high-proportion new energy power grid energy management method, device, system and equipment based on digital twinning. BACKGROUND

[0002] With the increasingly serious predatory exploitation of non-renewable energy, new energy represented by wind energy and solar energy gradually emerges, and new energy power generation is changing from supplementary energy to alternative energy, and the power distribution network also presents the characteristics of power electronics. In the power distribution network of the high-proportion new energy gathering area, the transmission line is long, the support capacity of the conventional power source is weak, the grid strength is low, the voltage changes greatly, and the fluctuation is strong, which brings certain risks to the safe and stable operation of the power distribution network.

[0003] Because of the lack of conventional power sources and weak grid strength of the power distribution network, the voltage and frequency of each node in the power distribution network are easily disturbed. In addition, the active power fluctuation of new energy is large, and the control characteristics of new energy, dynamic reactive power source, energy storage and other devices in the power distribution network have a great influence on the operation risk of large-scale new energy base. Further, the power distribution network model containing a large number of new energy, dynamic reactive power source, energy storage and other devices is difficult to build, and a precise simulation system cannot be established for real-time control strategy simulation. Moreover, in the face of a large number of real-time dispatching requirements, the new energy node covers a wide range, and the communication is difficult, so that the dispatching center cannot accurately obtain the real-time information feedback of all power nodes, and the existing power dispatching system cannot realize high-precision and fast-response power regulation. SUMMARY

[0004] The present application provides a high-proportion new energy power grid energy management method, device, system and equipment based on digital twinning, which is used to solve the technical problem that the existing new energy power system cannot perform power dispatching according to the power node feedback information, resulting in low and unstable precision of power dispatching.

[0005] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0006] On the one hand, a high-proportion new energy power grid energy management method based on digital twinning is provided, comprising the following steps:

[0007] Obtaining a sampling budget and an initial iteration number, and obtaining n real-time working condition information and m power dispatching decisions of the high-proportion new energy power grid operation;

[0008] According to the sampling budget, the initial iteration number, n real-time working condition information and m power dispatching decisions, iterative processing is performed to obtain k groups of dispatching decision data, and each group of dispatching decision data includes a plurality of simulation data and a plurality of prediction results;

[0009] performing clustering analysis and optimization processing on each of the simulation data and each of the prediction results of each of the scheduling decision data, to obtain a posterior mean value of each of the power scheduling decisions and an optimal power scheduling decision corresponding to each of the scheduling decision data;

[0010] selecting, from all of the optimal power scheduling decisions, an optimal power scheduling decision corresponding to a maximum value of the posterior mean values as an optimal power scheduling scheme for the high-proportion new-energy power grid operation.

[0011] Preferably, performing clustering analysis and optimization processing on each of the simulation data and each of the prediction results of each of the scheduling decision data, to obtain a posterior mean value of each of the power scheduling decisions and an optimal power scheduling decision corresponding to each of the scheduling decision data comprises:

[0012] obtaining an ideal sampling mean value;

[0013] performing clustering analysis on each of the power scheduling decisions according to all of the simulation data and all of the prediction results of each of the scheduling decision data by using a likelihood function and a Bayesian information criterion of an expectation maximization algorithm, to obtain a clustering category and a category model parameter matrix corresponding to the clustering category;

[0014] calculating a sample mean value according to all of the simulation data corresponding to each of the power scheduling decisions, and calculating a posterior mean value corresponding to each of the power scheduling decisions according to the category model parameter matrix corresponding to each of the power scheduling decisions, the ideal sampling mean value, the sample mean value, and the prediction result;

[0015] selecting, from all of the posterior mean values, a power scheduling decision corresponding to a maximum value as an optimal power scheduling decision of the scheduling decision data.

[0016] Preferably, the high-proportion new-energy power grid energy management method based on digital twinning further comprises: calculating a posterior mean value corresponding to each of the power scheduling decisions according to the category model parameter matrix corresponding to each of the power scheduling decisions, the ideal sampling mean value, the sample mean value, and the prediction result by using a posterior formula; and the posterior formula is:

[0017]

[0018]

[0019] wherein, is a posterior mean value of a jth power scheduling decision under ith real-time working condition information, is a sample variance of the jth power scheduling decision under the ith real-time working condition information, a posteriori variance of the jth power dispatch decision under the ith real-time working condition information, a matrix of simulation data outputted by the power system simulation in the clustering category c, a matrix of prediction results outputted by the machine learning model in the clustering category c, 、 a matrix of category model parameters of different values in the clustering category c, l j，i simulation data of the jth power dispatch decision under the ith real-time working condition information, a sample mean of the jth power dispatch decision under the ith real-time working condition information, g j，i a prediction result of the jth power dispatch decision under the ith real-time working condition information, T is a matrix transpose.

[0020] Preferably, obtaining a set of the dispatch decision data comprises: performing simulation on each of the real-time working condition information by using a power system simulation according to each of the power dispatch decisions, to obtain a plurality of simulation data; and performing prediction learning on each of the real-time working condition information by using K machine learning models according to each of the power dispatch decisions, to obtain K prediction results.

[0021] Preferably, the iterative processing according to the sampling budget, the initial iteration number, n real-time working condition information and m power dispatch decisions obtains k sets of dispatch decision data, which comprises:

[0022] obtaining an initial simulation data amount and a simulation sample number;

[0023] performing first iteration data acquisition on n real-time working condition information and m power dispatch decisions according to the initial iteration number, to obtain a first set of the dispatch decision data;

[0024] calculating a simulation consumption budget according to the initial simulation data amount and a total number of decisions of the first set of the dispatch decision data; if the simulation consumption budget is greater than the sampling budget, ending the iteration data acquisition, and obtaining one set of the dispatch decision data with k being 1;

[0025] if the simulation consumption budget is not greater than the sampling budget, performing iteration data acquisition again on n real-time working condition information and m power dispatch decisions according to an updated iteration number, to obtain a kth set of the dispatch decision data, until a total simulation consumption budget of the kth set of the dispatch decision data after accumulation is greater than the sampling budget, and k sets of the dispatch decision data are obtained;

[0026] wherein, the total simulation consumption budget L k+1 = L k + ΔL, L k ​A simulation consumption budget of the scheduling decision data is obtained for the kth iteration, and ΔL is the number of simulation samples consumed in the k+1th iteration.

[0027] Preferably, the digital-twin-based energy management method for high-proportion new energy power grids further comprises: obtaining a static voltage stability margin according to the static voltage stability margin index data using a margin formula; , wherein, is the static voltage stability margin index data, and Δ is the static voltage stability margin.

[0028] In another aspect, a digital-twin-based energy management device for high-proportion new energy power grids is provided, comprising a data acquisition module, a data processing module, a clustering optimization module, and a screening and determination module.

[0029] The data acquisition module is configured to acquire a sampling budget and an initial iteration number, and acquire n real-time working condition information and m power scheduling decisions of a high-proportion new energy power grid.

[0030] The data processing module is configured to perform iteration processing according to the sampling budget, the initial iteration number, n real-time working condition information, and m power scheduling decisions, to obtain k groups of scheduling decision data, each group of the scheduling decision data comprising a plurality of simulation data and a plurality of prediction results.

[0031] The clustering optimization module is configured to perform clustering analysis and optimization processing on each simulation data and each prediction result of each group of the scheduling decision data, to obtain a posterior mean value of each power scheduling decision and a best power scheduling decision corresponding to each group of the scheduling decision data.

[0032] The screening and determination module is configured to screen a best power scheduling decision corresponding to a maximum posterior mean value from all the best power scheduling decisions as an optimal power scheduling scheme for the high-proportion new energy power grid.

[0033] Preferably, the clustering optimization module comprises an acquisition submodule, a clustering analysis submodule, a calculation submodule, and a screening submodule.

[0034] The acquisition submodule is configured to acquire an ideal sampling mean value.

[0035] The clustering analysis submodule is configured to perform clustering analysis on each power scheduling decision according to all simulation data and all prediction results of each group of the scheduling decision data using a likelihood function and a Bayesian information criterion of an expectation maximization algorithm, to obtain a clustering category and a category model parameter matrix corresponding to the clustering category.

[0036] The calculating sub-module is configured to calculate, according to all the simulation data corresponding to each of the power dispatching decisions, a sample mean value; and calculate, according to the category model parameter matrix corresponding to each of the power dispatching decisions, the ideal sample mean value, the sample mean value and the prediction result, a posteriori mean value corresponding to each of the power dispatching decisions.

[0037] The screening sub-module is configured to screen, from all the posteriori mean values, a power dispatching decision with the largest value as the optimal power dispatching decision of the group of the dispatching decision data.

[0038] Preferably, the calculating sub-module is further configured to calculate, according to the category model parameter matrix corresponding to each of the power dispatching decisions, the ideal sample mean value, the sample mean value and the prediction result, the posteriori mean value corresponding to each of the power dispatching decisions by using a posteriori formula; and the posteriori formula is as follows:

[0039]

[0040]

[0041] In the formula, μj(i) is the posteriori mean value of the jth power dispatching decision under the ith real-time working condition information, μj(i) is the sample mean value of the jth power dispatching decision under the ith real-time working condition information, σj(i) is the sample variance of the jth power dispatching decision under the ith real-time working condition information, and σj(i) is the posteriori variance of the jth power dispatching decision under the ith real-time working condition information. j，i In the formula, Xj(i) is the simulation data of the jth power dispatching decision under the ith real-time working condition information. j，i

[0042] In another aspect, a high-proportion new energy power grid energy management system based on digital twinning is provided, which comprises a data acquisition unit, a prediction simulation learning unit and a real-time digital twinning power dispatching decision unit.

[0043] The data acquisition unit is configured to acquire a sampling budget and an initial iteration number, and acquire n real-time working condition information and m power dispatching decisions of high-proportion new energy power grid operation. ​​​​​​​​​

[0044] The prediction simulation learning unit is used to simulate the real-time operating information based on the power dispatch decisions using power system simulation to obtain several simulation data; and to perform prediction learning on the real-time operating information using K machine learning models based on the power dispatch decisions to obtain m prediction results.

[0045] The real-time digital twin power dispatch decision unit is used to combine several simulation data and several prediction results of k sets of dispatch decision data with the above-mentioned digital twin-based high-proportion renewable energy grid energy management method to obtain the optimal power dispatch scheme for the operation of the high-proportion renewable energy grid.

[0046] On the other hand, a terminal device is provided, including a processor and a memory;

[0047] The memory is used to store program code and transmit the program code to the processor;

[0048] The processor is used to execute the above-described high-proportion renewable energy grid energy management method based on digital twins according to the instructions in the program code.

[0049] This invention discloses a method, device, system, and equipment for energy management of a high-proportion renewable energy power grid based on digital twins. The method includes acquiring a sampling budget and initial iteration count, as well as acquiring n real-time operating conditions and m power dispatch decisions for the high-proportion renewable energy power grid. Iterative processing is performed based on the sampling budget, initial iteration count, n real-time operating conditions, and m power dispatch decisions to obtain k sets of dispatch decision data. Each set of dispatch decision data includes several simulation data points and several prediction results. Cluster analysis and optimization processing are performed on each simulation data point and prediction result of each set of dispatch decision data to obtain the posterior mean of each power dispatch decision and the optimal power dispatch decision corresponding to each set of dispatch decision data. The optimal power dispatch decision with the largest posterior mean value is selected from all optimal power dispatch decisions as the optimal power dispatch scheme for the operation of the high-proportion renewable energy power grid.

[0050] As can be seen from the above technical solutions, this application has the following advantages: The energy management method for high-proportion new energy power grids based on digital twins obtains the optimal power dispatch decision corresponding to each set of dispatch decision data through cluster analysis using the prediction results and simulation data of each set of dispatch decision data. Then, the optimal power dispatch scheme is selected from the optimal power dispatch decisions corresponding to multiple sets of dispatch decision data. This achieves better power energy dispatch and control with lower communication costs, even when the simulation of extremely high-proportion new energy power grids is difficult. It solves the technical problem that the power dispatch of existing new energy power systems cannot be carried out based on the feedback information of power nodes, resulting in low accuracy and instability of power dispatch.

[0051] This digital twin-based high-proportion renewable energy grid energy management device uses a data acquisition module, a data processing module, a clustering optimization module, and a screening and determination module to obtain the optimal power dispatch decision corresponding to each set of dispatch decision data through cluster analysis of the prediction results and simulation data. Then, it selects the optimal power dispatch scheme from the optimal power dispatch decisions corresponding to multiple sets of dispatch decision data. This enables good power energy dispatch and control with low communication costs, even when accurate simulation of a very high proportion of renewable energy grids is difficult.

[0052] This digital twin-based high-proportion renewable energy grid energy management system trains multiple machine learning models using simulation data from power system simulations in its predictive simulation learning unit. This allows for the generation of multiple predictions based on real-time operating conditions and power dispatch decisions. The real-time digital twin power dispatch decision-making unit performs real-time simulations based on real-time operating conditions and power dispatch decisions. It then combines the simulation data with the machine learning predictions, uses cluster analysis to categorize the results, and finally obtains the optimal power dispatch scheme. Applying this optimal power dispatch scheme to the high-proportion renewable energy grid, the system receives real-time operating information from each power node, inputs it into the power system simulation for data updates, performs simulations based on the updated data, and then performs offline learning on the machine learning models. This enables the optimal power dispatch scheme to generate the optimal power dispatch plan based on the real-time operating information from each power node, achieving better power energy dispatch and control. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating the steps of the high-proportion renewable energy grid energy management method based on digital twins described in this application embodiment;

[0055] Figure 2 This is a schematic diagram of the framework of the high-proportion renewable energy grid energy management method based on digital twins described in the embodiments of this application;

[0056] Figure 3 This is a schematic diagram of the high-proportion renewable energy grid energy management method based on digital twins as described in the embodiments of this application;

[0057] Figure 4 This is a schematic diagram of the framework of the high-proportion renewable energy grid energy management device based on digital twins described in the embodiments of this application;

[0058] Figure 5 This is a schematic diagram of the terminal device described in an embodiment of this application. Detailed Implementation

[0059] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] In the description of the embodiments of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0061] In the embodiments of this application, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this application according to the specific circumstances.

[0062] This application provides a method, device, system, and equipment for high-proportion renewable energy power grid energy management based on digital twins, which solves the technical problem that the power dispatching of existing renewable energy power systems cannot be carried out based on the feedback information of power nodes, resulting in low accuracy and instability of power dispatching.

[0063] Example 1:

[0064] Figure 1 This is a flowchart illustrating the steps of the high-proportion renewable energy grid energy management method based on digital twins described in this application embodiment. Figure 2 This is a schematic diagram illustrating the framework of the high-proportion renewable energy grid energy management method based on digital twins described in this application embodiment. Figure 3 This is a schematic diagram of the energy management method for high-proportion renewable energy power grids based on digital twins, as described in an embodiment of this application.

[0065] like Figures 1 to 3 As shown in the figure, this application provides a method for energy management of a high proportion of new energy power grids based on digital twins, including the following steps:

[0066] S1. Obtain the sampling budget and initial iteration count, as well as n real-time operating condition information and m power dispatch decisions for the high-proportion renewable energy power grid.

[0067] It should be noted that obtaining the sampling budget and initial iteration count in step S1 provides data for obtaining a certain number of sets of scheduling decision data; the n real-time operating condition information and m power dispatch decisions obtained in step S1 for the high-proportion renewable energy power grid operation provide the basic data for subsequently obtaining multiple sets of scheduling decision data. In this embodiment, the real-time operating condition information is denoted as... , , This represents the actual operating condition information of the high-proportion renewable energy power grid in the i-th time slot, and the power dispatch decision is denoted as . , , Let j represent the j-th power dispatch decision. A high-proportion renewable energy grid refers to a power system with large-scale integration of distributed renewable energy sources (such as wind, solar, and hydropower).

[0068] S2. Based on the sampling budget, initial iteration count, n real-time operating condition information and m power dispatch decisions, perform iterative processing to obtain k sets of dispatch decision data. Each set of dispatch decision data includes several simulation data and several prediction results.

[0069] It should be noted that in step S2, based on the sampling budget and initial iteration count obtained in step S1, k sets of scheduling decision data are iteratively processed on n real-time operating conditions and m power dispatch decisions to obtain data for obtaining the optimal power dispatch scheme. In this embodiment, because the power simulation system constructed with a high proportion of new energy power grids has many uncertainties and random variables, even if multiple simulations are performed at each operating condition time, it may not be able to output ideal values. Due to the existence of simulation noise, the power dispatch decision selected using a set of scheduling decision data may not be the optimal power dispatch scheme. Therefore, k sets of scheduling decision data are obtained under the maximized sampling budget L, and the optimal power dispatch scheme is correctly selected from the best power dispatch decision determined by the k sets of scheduling decision data.

[0070] S3. Perform cluster analysis and optimization on each simulation data and prediction result of each group of scheduling decision data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each group of scheduling decision data.

[0071] It should be noted that in step S3, the prediction results of each set of scheduling decision data are used to filter each power scheduling decision, and the results are optimized in combination with simulation data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each set of scheduling decision data. This provides data for the final selection of the optimal power scheduling scheme for high-proportion renewable energy grid operation.

[0072] S4. Select the best power dispatch decision with the largest corresponding posterior mean value from all the best power dispatch decisions as the optimal power dispatch scheme for high-proportion renewable energy power grid operation.

[0073] It should be noted that in step S4, the best power dispatch decision obtained from each group of dispatch decision data in step S3 is selected based on the condition that its posterior mean value is the largest, and the best power dispatch decision with the largest posterior mean value is taken as the optimal power dispatch scheme for the operation of the high proportion of new energy power grid.

[0074] In this embodiment, the energy management method for high-proportion renewable energy power grids based on digital twins uses cluster analysis to obtain the optimal power dispatch decision corresponding to each set of dispatch decision data based on the prediction results and simulation data of each set of dispatch decision data. Then, the optimal power dispatch scheme is selected from the optimal power dispatch decisions corresponding to multiple sets of dispatch decision data. The operation of the high-proportion renewable energy power grid is controlled by the optimal power dispatch scheme. This solves the problems of power dispatch difficulties and heavy reliance on accurate power grid models and real-time feedback signals from power nodes in the prior art for extremely high-proportion renewable energy power grids. It has strong adaptability and can perform power dispatch even without real-time feedback signals from power nodes, thus improving the robustness and reliability of power grid dispatch.

[0075] It should be noted that this digital twin-based energy management method for high-proportion renewable energy power grids can achieve good power energy dispatch and control even when the accurate simulation of extremely high-proportion renewable energy power grids is difficult.

[0076] This application provides a method for energy management of a high-proportion renewable energy power grid based on digital twins. The method includes obtaining a sampling budget and initial iteration count, as well as acquiring n real-time operating conditions and m power dispatch decisions for the high-proportion renewable energy power grid. Iterative processing is performed based on the sampling budget, initial iteration count, n real-time operating conditions, and m power dispatch decisions to obtain k sets of dispatch decision data. Each set of dispatch decision data includes several simulation data points and several prediction results. Cluster analysis and optimization processing are performed on each simulation data point and prediction result of each set of dispatch decision data to obtain the posterior mean of each power dispatch decision and the optimal power dispatch decision corresponding to each set of dispatch decision data. The optimal power dispatch decision with the largest corresponding posterior mean value is selected from all optimal power dispatch decisions as the optimal power dispatch scheme for the high-proportion renewable energy power grid operation. This digital twin-based energy management method for high-proportion renewable energy power grids uses cluster analysis to obtain the optimal power dispatch decision corresponding to each set of dispatch decision data based on the prediction results and simulation data. Then, it selects the optimal power dispatch scheme from the optimal power dispatch decisions corresponding to multiple sets of dispatch decision data. This method achieves better power energy dispatch and control with lower communication costs, even when accurate simulation of extremely high-proportion renewable energy power grids is difficult. It solves the technical problem that existing renewable energy power systems cannot perform power dispatch based on feedback information from power nodes, resulting in low accuracy and instability in power dispatch.

[0077] In one embodiment of this application, obtaining a set of scheduling decision data includes: simulating various real-time operating conditions using power system simulation based on each power scheduling decision to obtain several simulation data sets; and using K machine learning models to predict and learn from each real-time operating condition based on each power scheduling decision to obtain K prediction results. The number of simulation data sets can be m.

[0078] It should be noted that the simulation data output by the power system simulation... It is based on real-time power grid status data consisting of n real-time operating condition information and m power dispatch decisions. Unbiased estimation is performed, and the distribution of the simulation output has a mean of [value missing]. And the variance is The power system simulation output does not follow a normal distribution. When the distribution is not normal, the average of batch data is needed to obtain the real-time power grid status data. Then, according to the central limit theorem, its distribution approximates a normal distribution. Besides power system simulation, there are also... A machine learning model performs predictive learning on various real-time operating condition information. Using... This represents the prediction result of the Kth machine learning model, i.e., the output of the machine learning prediction. .

[0079] In one embodiment of this application, cluster analysis and optimization processing are performed on each simulation data and each prediction result of each set of scheduling decision data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each set of scheduling decision data, including:

[0080] Obtain the ideal sample mean;

[0081] Based on all simulation data and all prediction results of each group of scheduling decision data, cluster analysis is performed on each power scheduling decision using the likelihood function of the expectation-maximization algorithm and the Bayesian information criterion to obtain cluster categories and the category model parameter matrix corresponding to the cluster categories.

[0082] The sample mean is calculated based on all simulation data corresponding to each power dispatch decision; the posterior mean corresponding to each power dispatch decision is calculated based on the category model parameter matrix, ideal sampling mean, sample mean, and prediction results.

[0083] The power dispatch decision with the largest value among all posterior means is selected as the best power dispatch decision for this set of dispatch decision data.

[0084] Specifically, the posterior mean for each power dispatch decision is calculated using a posterior formula based on the category model parameter matrix, ideal sampling mean, sample mean, and prediction results. The posterior formula is as follows:

[0085]

[0086]

[0087] In the formula, Let be the posterior mean of the j-th power dispatch decision under the i-th real-time operating condition information. Let be the sample variance of the j-th power dispatch decision under the ith real-time operating condition information. Let be the posterior variance of the j-th power dispatch decision under the i-th real-time operating condition information. This is a matrix composed of simulation data output from power system simulation in cluster category c. This is the prediction matrix output by the machine learning model in cluster category c. , These are the category model parameter matrices with different values ​​in cluster category c, lj，i The simulation data is for the j-th power dispatch decision under the ith real-time operating condition information. Let g be the sample mean of the j-th power dispatch decision under the i-th real-time operating condition information. j，i Let T be the prediction result of the j-th power dispatch decision under the i-th real-time operating condition information, and T be the matrix transpose.

[0088] It should be noted that a set of scheduling decision data is defined as denoted as , f0(X) j Z j ) represents the real-time status data of the power grid, consisting of n real-time operating condition information and m power dispatch decisions, where g(X) is the real-time status data of the power grid. j Z j () represents the prediction result of the scheduling decision data; assuming It follows a Gaussian mixture model, and the expression for a Gaussian mixture model is:

[0089]

[0090] In the formula, It is the probability density function of a multidimensional, multivariate normal distribution. This represents the number of clusters after clustering using the Gaussian mixture model; according to the expression for the Gaussian mixture model, if... The performance value belongs to the category Then we get the prediction results. and real-time status data of the power grid Follows the mean vector The sum and covariance matrix are The multivariate normal distribution; parameters Indicates category Given weights.

[0091] Machine learning predictions and power grid simulation data can be used to inform power dispatch decisions based on real-time operating conditions in the current time slot. i The data is clustered into C categories. In existing online decision-making systems (i.e., real-time decisions based on real-time feedback signals from the power grid), the amount of simulation data for power system simulation is very limited. When noise is significant, including this simulation data in cluster analysis may not be helpful. Based on this consideration, this digital twin-based high-proportion renewable energy grid energy management method uses a fully observable machine learning prediction function. Cluster analysis was performed on some of the observable data. The clustering structure provides an approximation. Specifically, in the embodiments of this application, cluster analysis is performed on each power dispatching decision using the likelihood function of the expectation-maximization algorithm and the Bayesian information criterion based on all simulation data and all prediction results of each group of dispatching decision data, resulting in cluster categories and category model parameter matrices corresponding to the cluster categories, including:

[0092] Based on all simulation data and all prediction results of each group of scheduling decision data, a data matrix is ​​obtained using a Gaussian mixture model. The data matrix is ​​composed of the category model parameter matrix.

[0093] The estimated parameters are obtained by calculating the likelihood function using the expectation-maximization algorithm based on the data matrix and the prediction results.

[0094] The estimated parameters are adjusted using the Bayesian information criterion to obtain the clustering categories for each power dispatch decision.

[0095] It should be noted that the expression for the data matrix is:

[0096]

[0097] The likelihood function is:

[0098]

[0099] The expression for cluster categories is:

[0100]

[0101] In the formula, and These respectively represent the results based on power system simulation and The matrix formed by the output data of the nth machine learning model is contained in the nth... The average performance of power dispatch decisions in each cluster; based on simulation data from power system simulation, the... The variance of power dispatch decision performance in each cluster is , The covariance matrix represents the correlation between the output simulation data and each machine learning model. It is the first All in each cluster The covariance matrix of the predicted data from a machine learning model. inverse matrix Parameters directly from , , It was calculated. Here and Depends solely on prediction results Data. (Used) This represents the set of parameters for the power system simulation to be estimated. The likelihood function can be maximized using the Expectation-Maximization (EM) algorithm, and the estimated parameters can be expressed as... Due to the total number of clusters. Since it is unknown, it can be considered as a tuning parameter. The Bayesian Information Criterion (BIC) can be used to adjust the given likelihood function to determine the optimal number of classes. ; It is necessary to estimate the number of parameters. It is the number of cluster categories. The preset upper limit. Once these parameters are determined, upon receiving real-time power grid operating information for the current time slot. After that, it will generate The initial online power simulation data. In this embodiment, in the expression of the data matrix, the first... Variance of power dispatch decisions in each cluster It is a number. The correlation between the output simulation data and each machine learning model is represented by a K*1 vector, and the covariance matrix is... It is a K*K matrix. A number, two vectors, and a matrix are concatenated to form a new matrix. For the matrix Finding the inverse matrix yields the inverse matrix. Inverse matrix It can also be divided into four different components according to their corresponding positions. Parameters It refers to the accuracy of the Gaussian mixture model. YE can be understood as the accuracy based on... Calculate the inverse matrix and then split it into four category model parameter matrices.

[0102] In the embodiments of this application, subscripts are used instead of... For example, using to indicate As the amount of simulation data increases, this weighting parameter... Increase. Posterior formula Using the estimated mean of the mixed clustering categories Instead, the expression for the estimated mean of the mixed cluster categories is:

[0103]

[0104] binary variable Represents the j-th power dispatch decision Whether it is grouped into a cluster category The value, if grouped into cluster category , =1; if grouped into cluster category , =0. Therefore, Grouping into the same cluster category Online power simulation estimates of all power dispatch decisions enable the sharing of limited online power simulation information among multiple power dispatch decisions. Therefore, the prediction results are used... Directly predict the posterior mean .

[0105] In one embodiment of this application, iterative processing is performed based on the sampling budget, the initial number of iterations, n real-time operating condition information, and m power dispatch decisions to obtain k sets of dispatch decision data, including:

[0106] Obtain the initial simulation data volume and the number of simulation samples;

[0107] Based on the initial iteration number, the first iteration data is obtained by acquiring n real-time operating condition information and m power dispatch decisions to obtain the first set of dispatch decision data.

[0108] The total number of decisions is calculated based on the initial simulation data volume and the first set of scheduling decision data; if the simulation consumption budget is greater than the sampling budget, the iteration data acquisition ends, and a set of scheduling decision data with k=1 is obtained;

[0109] If the simulation cost budget is not greater than the sampling budget, then... The update iterations are performed again to acquire data from n real-time operating conditions and m power dispatching decisions to obtain the kth set of dispatching decision data. This process continues until the total simulation cost budget exceeds the sampling budget after the kth set of dispatching decision data is accumulated, thus obtaining k sets of dispatching decision data.

[0110] The total simulation cost budget L k+1 =L k +ΔL,L k Let ΔL be the cumulative simulation consumption budget for obtaining scheduling decision data in the k-th iteration, and let ΔL be the number of simulation samples consumed in the (k+1)-th iteration.

[0111] It should be noted that in obtaining the first set of scheduling decision data, power system simulation was used to simulate various real-time operating conditions based on each power scheduling decision, resulting in several simulation data sets. Then, K machine learning models were used to predict and learn from each real-time operating condition based on each power scheduling decision, yielding K prediction results. Subsequently, the simulation cost budget L1 for the first iteration was calculated based on the values ​​of the simulation data and the total number of decisions in the first set of scheduling decision data. Let m be the total number of power dispatch decisions and l0 be the initial simulation data volume. For example, simulating m different power dispatch decisions based on a sample of l0 real-time operating conditions will produce m*l0 simulation results. Assuming that the budget consumed in each simulation is different and follows a Gaussian distribution, when the number of samples and power dispatch decisions is sufficient, the result will eventually be a mean regression, i.e., the budget consumed in each iteration is m*l0*(average cost per simulation). If the overall data is normalized, it becomes m*l0, where the cost per simulation can be omitted as a coefficient.

[0112] In one embodiment of this application, in the optimal power dispatch decision, let and In the formula, To achieve the optimal power dispatching scheme, f is the mean variable. j*，i This is sample data for the optimal power dispatch scheme. (Compared to the optimal power dispatch scheme) The corresponding power simulation data is denoted as The simulation data for optimal power dispatch decision must satisfy the following conditions:

[0113]

[0114]

[0115] In the formula, The mean variable is the optimal power dispatch scheme.

[0116] Example 2:

[0117] Figure 4 This is a schematic diagram of the framework of the high-proportion renewable energy grid energy management device based on digital twins described in the embodiments of this application.

[0118] like Figure 4 As shown, this application embodiment provides a high-proportion new energy grid energy management device based on digital twin, including a data acquisition module 10, a data processing module 20, a clustering optimization module 30, and a screening and determination module 40;

[0119] Data acquisition module 10 is used to acquire the sampling budget and initial iteration number, as well as n real-time operating condition information and m power dispatch decisions for the operation of a high-proportion renewable energy power grid.

[0120] The data processing module 20 is used to iteratively process the sampling budget, initial iteration count, n real-time operating condition information and m power dispatch decisions to obtain k sets of dispatch decision data. Each set of dispatch decision data includes several simulation data and several prediction results.

[0121] Clustering optimization module 30 is used to perform clustering analysis and optimization processing on each simulation data and each prediction result of each group of scheduling decision data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each group of scheduling decision data.

[0122] The selection and determination module 40 is used to select the best power dispatch decision with the largest corresponding posterior mean value from all the best power dispatch decisions as the optimal power dispatch scheme for the operation of the high-proportion renewable energy power grid.

[0123] It should be noted that the content of the modules in the device of Embodiment 2 has already been described in the steps of the method of Embodiment 1, and the content of the module of the high-proportion renewable energy grid energy management device based on digital twin will not be described again in this embodiment. In this embodiment, the high-proportion renewable energy grid energy management device based on digital twin uses a data acquisition module, a data processing module, a clustering optimization module, and a screening and determination module to obtain the optimal power dispatch decision corresponding to each set of dispatch decision data by using cluster analysis on the prediction results and simulation data of each set of dispatch decision data. Then, the optimal power dispatch scheme is selected from the optimal power dispatch decisions corresponding to multiple sets of dispatch decision data. This achieves better power energy dispatch and control with lower communication costs, even when the accurate simulation of a very high proportion of renewable energy grid is difficult.

[0124] In this embodiment, the clustering optimization module 30 includes an acquisition submodule, a clustering analysis submodule, a calculation submodule, and a filtering submodule;

[0125] The `get` submodule is used to obtain the ideal sample mean.

[0126] The clustering analysis submodule is used to perform clustering analysis on each power dispatching decision based on all simulation data and all prediction results of each group of dispatching decision data, using the likelihood function of the expectation-maximization algorithm and the Bayesian information criterion, to obtain cluster categories and the category model parameter matrix corresponding to the cluster categories;

[0127] The calculation submodule is used to calculate the sample mean based on all simulation data corresponding to each power dispatch decision; and to calculate the posterior mean corresponding to each power dispatch decision based on the category model parameter matrix, ideal sampling mean, sample mean, and prediction results.

[0128] The filtering submodule is used to select the power dispatch decision with the largest value from all posterior means as the best power dispatch decision for this set of dispatch decision data.

[0129] In this embodiment, the calculation submodule is further configured to calculate, using a posterior formula, the categorical model parameter matrix corresponding to each power dispatch decision, the ideal sample mean, the sample mean, and the prediction result, to obtain the posterior mean corresponding to each power dispatch decision; the posterior formula is:

[0130]

[0131]

[0132] In the formula, Let be the posterior mean of the j-th power dispatch decision under the i-th real-time operating condition information. Let be the sample variance of the j-th power dispatch decision under the ith real-time operating condition information. Let be the posterior variance of the j-th power dispatch decision under the i-th real-time operating condition information. This is a matrix composed of simulation data output from power system simulation in cluster category c. This is the prediction matrix output by the machine learning model in cluster category c. , These are the category model parameter matrices with different values ​​in cluster category c, l j，i The simulation data is for the j-th power dispatch decision under the ith real-time operating condition information. Let g be the sample mean of the j-th power dispatch decision under the i-th real-time operating condition information. j，i Let T be the prediction result of the j-th power dispatch decision under the i-th real-time operating condition information, and T be the matrix transpose.

[0133] Example 3:

[0134] like Figure 2 As shown, this application provides a high-proportion renewable energy grid energy management system based on digital twins, including a data acquisition unit, a prediction simulation learning unit, and a real-time digital twin power dispatch decision unit;

[0135] The data acquisition unit is used to acquire the sampling budget and initial iteration number, as well as n real-time operating condition information and m power dispatch decisions for the operation of a high-proportion renewable energy power grid.

[0136] The prediction simulation learning unit is used to simulate various real-time operating conditions based on various power dispatch decisions using power system simulation, and obtain several simulation data; and to use K machine learning models to predict and learn various real-time operating conditions based on various power dispatch decisions, and obtain m prediction results.

[0137] The real-time digital twin power dispatch decision unit is used to combine several simulation data and several prediction results of k sets of dispatch decision data and process them according to the above-mentioned energy management method for high-proportion renewable energy power grid based on digital twins to obtain the optimal power dispatch scheme for the operation of high-proportion renewable energy power grid.

[0138] It should be noted that in the predictive simulation learning unit, multiple machine learning models are trained based on simulation data from power system simulation, yielding multiple prediction results based on real-time operating conditions and power dispatch decisions. The real-time digital twin power dispatch decision unit performs real-time simulation based on real-time operating conditions and power dispatch decisions. Then, the simulation data and machine learning prediction results are combined, and cluster analysis is used to categorize the results, ultimately obtaining the optimal power dispatch scheme. Applying this optimal power dispatch scheme to a high-proportion renewable energy grid, the scheme receives real-time operating condition information from each power node, inputs it into the power system simulation for data updates, performs simulations based on the updated data, and then performs offline learning on the machine learning models. This allows the optimal power dispatch scheme to obtain the optimal power dispatch plan based on the real-time operating condition information from each power node, achieving better power energy dispatch and control.

[0139] Example 4:

[0140] Figure 5 This is a schematic diagram of the terminal device described in an embodiment of this application.

[0141] like Figure 5 As shown, this application provides a terminal device, including a processor and a memory;

[0142] Memory is used to store program code and transfer the program code to the processor;

[0143] The processor is used to execute the aforementioned energy management method for high-proportion renewable energy grids based on digital twins, according to the instructions in the program code.

[0144] It should be noted that the processor is used to execute the steps in the above-described embodiment of a high-proportion renewable energy grid energy management method based on digital twins, according to the instructions in the program code. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described system / device embodiments.

[0145] For example, a computer program can be divided into one or more modules / units, one or more of which are stored in memory and executed by a processor to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a terminal device.

[0146] Terminal devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. Terminal devices may include, but are not limited to, processors and memory. Those skilled in the art will understand that this does not constitute a limitation on the terminal device, which may include more or fewer components than illustrated, or combinations of certain components, or different components. For example, a terminal device may also include input / output devices, network access devices, buses, etc.

[0147] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0148] Memory can be an internal storage unit of a terminal device, such as a hard drive or RAM. Memory can also be an external storage device, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, memory can include both internal and external storage units. Memory is used to store computer programs and other programs and data required by the terminal device. Memory can also be used to temporarily store data that has been output or will be output.

[0149] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0150] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0152] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0153] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0154] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for energy management of high-proportion renewable energy power grids based on digital twins, characterized in that, Includes the following steps: Obtain the sampling budget and initial iteration count, as well as n real-time operating condition information and m power dispatch decisions for the operation of a high-proportion renewable energy power grid; Based on the sampling budget, the initial number of iterations, n real-time operating conditions, and m power dispatching decisions, iterative processing is performed to obtain k sets of dispatching decision data. Each set of dispatching decision data includes several simulation data and several prediction results. Cluster analysis and optimization processing are performed on each simulation data and each prediction result of each group of scheduling decision data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each group of scheduling decision data. The optimal power dispatching decision with the largest posterior mean value is selected from all the optimal power dispatching decisions and used as the optimal power dispatching scheme for the operation of the high-proportion renewable energy power grid.

2. The method for high-proportion renewable energy grid energy management based on digital twins according to claim 1, characterized in that, Cluster analysis and optimization are performed on each set of simulation data and each set of prediction results for each group of scheduling decision data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each group of scheduling decision data, including: Obtain the ideal sample mean; Based on all the simulation data and all the prediction results of each group of scheduling decision data, cluster analysis is performed on each of the power scheduling decisions using the likelihood function of the expectation-maximization algorithm and the Bayesian information criterion to obtain cluster categories and the category model parameter matrix corresponding to the cluster categories; The sample mean is calculated based on all the simulation data corresponding to each power dispatch decision; the posterior mean corresponding to each power dispatch decision is calculated based on the category model parameter matrix, the ideal sampling mean, the sample mean, and the prediction result corresponding to each power dispatch decision. The power dispatch decision with the largest value is selected from all the posterior means as the best power dispatch decision for that set of dispatch decision data.

3. The energy management method for high-proportion renewable energy power grids based on digital twins according to claim 2, characterized in that, Also includes: The posterior mean for each power dispatch decision is calculated using a posterior formula based on the category model parameter matrix, the ideal sample mean, the sample mean, and the prediction result corresponding to each power dispatch decision; the posterior formula is: In the formula, Let be the posterior mean of the j-th power dispatch decision under the i-th real-time operating condition information. Let be the sample variance of the j-th power dispatch decision under the ith real-time operating condition information. Let be the posterior variance of the j-th power dispatch decision under the i-th real-time operating condition information. This is a matrix composed of simulation data output from power system simulation in cluster category c. This is the prediction matrix output by the machine learning model in cluster category c. , These are the category model parameter matrices with different values ​​in cluster category c, l j，i The simulation data is for the j-th power dispatch decision under the ith real-time operating condition information. Let g be the sample mean of the j-th power dispatch decision under the ith real-time operating condition information. j，i Let T be the prediction result of the j-th power dispatch decision under the i-th real-time operating condition information, and T be the matrix transpose.

4. The method for high-proportion renewable energy grid energy management based on digital twins according to any one of claims 1-3, characterized in that, Obtaining a set of the aforementioned scheduling decision data includes: simulating the real-time operating information of each of the aforementioned power scheduling decisions using power system simulation to obtain several simulation data; and using K machine learning models to predict and learn the real-time operating information of each of the aforementioned power scheduling decisions to obtain K prediction results.

5. The method for high-proportion renewable energy grid energy management based on digital twins according to any one of claims 1-3, characterized in that, Based on the sampling budget, the initial iteration count, n real-time operating condition information items, and m power dispatch decisions, iterative processing is performed to obtain k sets of dispatch decision data, including: Obtain the initial simulation data volume and the number of simulation samples; Based on the initial iteration number, the first iteration data acquisition is performed on n real-time operating condition information and m power dispatch decisions to obtain the first set of dispatch decision data; The simulation consumption budget is calculated based on the initial simulation data volume and the total number of decisions in the first set of scheduling decision data; if the simulation consumption budget is greater than the sampling budget, the iterative data acquisition ends, and a set of scheduling decision data with k=1 is obtained; If the simulation cost budget is not greater than the sampling budget, then... The update iterations are performed again to acquire data from n real-time operating conditions and m power dispatch decisions to obtain the kth set of dispatch decision data. This process continues until the total simulation consumption budget after accumulating the kth set of dispatch decision data is greater than the sampling budget, thus obtaining k sets of dispatch decision data. The total simulation cost budget L k+1 =L k +ΔL,L k Let ΔL be the simulation cost budget for obtaining scheduling decision data in the k-th iteration, and let ΔL be the number of simulation samples consumed in the (k+1)-th iteration.

6. A high-proportion renewable energy grid energy management device based on digital twins, characterized in that, include: The module includes a data acquisition module, a data processing module, a clustering optimization module, and a filtering and determination module. The data acquisition module is used to acquire the sampling budget and the initial iteration number, as well as to acquire n real-time operating condition information and m power dispatch decisions for the operation of a high-proportion renewable energy power grid. The data processing module is used to perform iterative processing based on the sampling budget, the initial number of iterations, n real-time operating condition information and m power dispatch decisions to obtain k sets of dispatch decision data. Each set of dispatch decision data includes several simulation data and several prediction results. The clustering optimization module is used to perform clustering analysis and optimization processing on each simulation data and each prediction result of each group of scheduling decision data to obtain the posterior mean of each power scheduling decision and the optimal power scheduling decision corresponding to each group of scheduling decision data. The filtering and determination module is used to select the best power dispatching decision with the largest corresponding posterior mean value from all the best power dispatching decisions as the optimal power dispatching scheme for the operation of the high-proportion new energy power grid.

7. The high-proportion renewable energy grid energy management device based on digital twins according to claim 6, characterized in that, The clustering optimization module includes an acquisition submodule, a clustering analysis submodule, a calculation submodule, and a filtering submodule; The acquisition submodule is used to acquire the ideal sample mean; The clustering analysis submodule is used to perform clustering analysis on each of the power dispatching decisions based on all the simulation data and all the prediction results of each group of dispatching decision data, using the likelihood function of the expectation-maximization algorithm and the Bayesian information criterion, to obtain cluster categories and the category model parameter matrix corresponding to the cluster categories; The calculation submodule is used to calculate the sample mean based on all the simulation data corresponding to each power dispatch decision; The posterior mean corresponding to each power dispatch decision is calculated based on the category model parameter matrix, the ideal sampling mean, the sample mean, and the prediction result corresponding to each power dispatch decision. The filtering submodule is used to select the power dispatch decision with the largest value from all the posterior means as the best power dispatch decision for that set of dispatch decision data.

8. The high-proportion renewable energy grid energy management device based on digital twins according to claim 7, characterized in that, The calculation submodule is further configured to calculate, using a posterior formula, the category model parameter matrix corresponding to each power dispatch decision, the ideal sampling mean, the sample mean, and the prediction result, to obtain the posterior mean corresponding to each power dispatch decision; the posterior formula is: In the formula, Let be the posterior mean of the j-th power dispatch decision under the i-th real-time operating condition information. Let be the sample variance of the j-th power dispatch decision under the ith real-time operating condition information. Let be the posterior variance of the j-th power dispatch decision under the i-th real-time operating condition information. This is a matrix composed of simulation data output from power system simulation in cluster category c. This is the prediction matrix output by the machine learning model in cluster category c. , These are the category model parameter matrices with different values ​​in cluster category c, l j，i The simulation data is for the j-th power dispatch decision under the ith real-time operating condition information. Let g be the sample mean of the j-th power dispatch decision under the ith real-time operating condition information. j，i Let T be the prediction result of the j-th power dispatch decision under the i-th real-time operating condition information, and T be the matrix transpose.

9. A high-proportion renewable energy grid energy management system based on digital twins, characterized in that, It includes a data acquisition unit, a predictive simulation learning unit, and a real-time digital twin power dispatching decision unit; The data acquisition unit is used to acquire the sampling budget and the initial iteration number, as well as to acquire n real-time operating condition information and m power dispatch decisions for the operation of a high-proportion renewable energy power grid. The prediction simulation learning unit is used to simulate the real-time operating information based on the power dispatch decisions using power system simulation to obtain several simulation data; and to perform prediction learning on the real-time operating information using K machine learning models based on the power dispatch decisions to obtain m prediction results. The real-time digital twin power dispatch decision unit is used to combine several simulation data and several prediction results of k sets of dispatch decision data and process them according to the energy management method of high-proportion new energy power grid based on digital twin as described in any one of claims 1-5 to obtain the optimal power dispatch scheme for the operation of the high-proportion new energy power grid.

10. A terminal device, characterized in that, Including the processor and memory; The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the high-proportion renewable energy grid energy management method based on digital twins as described in any one of claims 1-5, according to the instructions in the program code.

Images