A method for tracking renewable energy generation uncertainty based on a recursive Gaussian mixture model

Abstract

The application discloses a renewable energy power generation uncertainty tracking method based on a recursive Gaussian mixture model, comprising the following steps: 1, obtaining a deterministic prediction curve of renewable energy power generation, forming a joint observation set containing predicted and actual power generation; 2, setting the cluster number as k, performing a k-means algorithm to obtain k clusters; 3, based on the k clusters, calculating initial parameters of each Gaussian component in the Gaussian mixture model; 4, updating the Gaussian mixture model parameters by using an expectation maximization algorithm to obtain starting optimal parameters; 5, constructing a recursive updating probability model, and based on newly-arrived samples, recursively updating the probability model; 6, using the updated Gaussian mixture model, calculating the conditional probability distribution of the actual power under the condition of the latest predicted power, so as to realize real-time quantification and tracking of the renewable energy power generation uncertainty.

Description

Technical Field

[0001] This invention belongs to the field of power system operation and control technology, specifically a method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model; Background Technology

[0002] With the large-scale integration of renewable power generation such as wind and solar power, the randomness and time-varying uncertainty faced by the power system have significantly increased. Traditional operational decisions based on deterministic predictions are unable to fully reflect this uncertainty, thus affecting the safety and economy of the power system. Therefore, establishing a probabilistic model that can simultaneously represent the joint distribution of predictions and measurements can more accurately describe the uncertainty of renewable energy output, which is of great significance for applications such as short- and medium-term dispatch, opportunity-constrained optimization, and risk assessment. Gaussian mixture models (GMMs), with their strong expressive power for arbitrary non-Gaussian and multimodal distributions, have become a commonly used tool for describing the joint distribution of renewable energy output and are widely used in probabilistic dispatch and risk constraint-based systems.

[0003] However, classic parameter estimation in GMMs typically employs an offline batch training method based on the full sample size using the Expectation-Maximization (EM) algorithm: first, a large number of historical samples are collected, initialized using methods such as k-means, and then the expectation and maximization steps are repeatedly executed until convergence. This offline approach incurs significant computational overhead in scenarios with large sample sizes or continuous sample arrivals, making it unsuitable for real-time tracking of the time-varying characteristics of renewable energy probability distributions during power system operation. Furthermore, simply using the approach of "accumulating new samples and re-batch training" lacks scalability. Against this backdrop, there is an urgent need for an engineered recursive update framework that can both efficiently utilize streaming data to continuously refine the model online and avoid accumulated errors through low-frequency calibration, in order to meet the real-time and accuracy requirements of power system operation. Summary of the Invention

[0004] This invention addresses the shortcomings of existing technologies by proposing a method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model. The method aims to progressively correct the parameters of the Gaussian mixture model under varying renewable energy power conditions, thereby achieving adaptive modeling of the uncertainties in renewable energy generation and improving the accuracy and stability of estimating the probability distribution of renewable energy generation.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: The present invention provides a method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model, characterized by the following steps: Step 1: Obtain Includes A renewable energy power generation The deterministic prediction curve for the first look-ahead point, and based on the first... Includes A renewable energy power generation The deterministic prediction curve for the nth look-ahead point is obtained. indivual dimensional predicted power generation vector and the indivual dimensional actual power generation vector Thus constructing the first A collection of predicted and actual power generation. Joint observation vector of dimension Thus forming a joint observation set ;in, This represents the total number of joint observation vectors; Step 2: Set the number of clusters to [number]. and to Execute the k-means algorithm to obtain Clusters ,in, Indicates the first k One cluster; Step 3, based on Clusters Calculate the first Gaussian mixture model (GMM) in Gaussian Mixture Model (GMM). k Initial parameters of the Gaussian components ,in, , , These represent the first and second halves of the Gaussian mixture model (GMM). k Initial weights, initial expectations, and initial covariance matrices of each Gaussian component; Step 4, based on Using the EM-maximization algorithm to Update the parameters to obtain the initial optimal parameters. ;in, , , These represent the first and second halves of the Gaussian mixture model (GMM). k The initial optimal weights, initial optimal expectations, and initial optimal covariance matrix of each Gaussian component; Step 5: Initialization Define and initialize the total number of joint observation vectors to be deleted. =0; Step Six: Obtain the first b batch A new joint observation vector and used for , , Update the model to obtain the Gaussian Mixture Model (GMM) of the first generation. k The weights of the Gaussian components are learned after new samples. Expectations after learning new samples The covariance matrix after learning new samples ;in, Indicates the first b The first batch A new joint observation vector; Indicates the first b The number of new joint observation vectors in the batch; Will and After piecing them together in order, we get the first... b Complete joint observation set of the batch ;in, express The first in Joint observation vectors; Step 7: Judgment Whether it is true or not, if true, then from Delete the earliest one Joint observation vectors , obtained the b Updated joint observation vector set for batches Proceed to step eight; otherwise, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed to step ten; among which, express The deleted number A joint observation vector, express The first in A joint observation vector, Indicates the sample threshold; Step 8: Joint observation vector based on deletion ,right , , Update the model to obtain the Gaussian Mixture Model (GMM) of the first generation. k The weights of the Gaussian components after discarding old samples Expectations after discarding old samples Covariance matrix after discarding old samples ; Step Nine, Assign to Then, make a judgment Is it true? If yes, proceed to step ten; otherwise, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed to step eleven; among which, For the threshold; Step 10, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed according to step four to obtain... , , and will Set to 0; Step 11, Judgment Is it true? If so, then... , , Each is used as the final optimal weight. Final optimal expectation The final optimal covariance matrix And proceed to step twelve; otherwise, Assign to b Return to step six and execute sequentially; among them, This represents the total number of batches. Step 12: Utilize the final optimal parameters The corresponding Gaussian Mixture Model (GMM) calculates the predicted renewable energy values ​​given a given vector. Actual power vector under certain conditions conditional probability distribution This is to enable real-time quantification and tracking of uncertainties in renewable energy generation.

[0006] The characteristic of the renewable energy generation uncertainty tracking method based on the recursive Gaussian mixture model described in this invention is that, in step three, the equations (1)-(3) are used to calculate the first Gaussian mixture model (GMM). k Initial weights of the Gaussian components Initial expectations Initial covariance matrix : (1) (2) (3) In equations (1)-(3), express The number of joint observation vectors in; express The first in A joint observation vector.

[0007] Furthermore, step four includes: Step 4.1: Calculate the first step using equation (4). i Joint observation vectors In the Gaussian Mixture Model (GMM), the first k The responsibility value of each Gaussian component : (4) In equation (4), Indicates expectation as The covariance matrix is The independent variable of the probability density function of the multivariate Gaussian distribution takes The function value at time, and we have: (5) In equation (5), This represents the transpose of a matrix; Represents a random vector The dimension; Represents the covariance matrix The determinant of; Step 4.2: Using equations (6)-(8), obtain the first Gaussian mixture model (GMM). k The updated weights of the Gaussian components Updated expectations The updated covariance matrix : (6) (7) (8) Step 4.3: Calculate the log-likelihood estimate using equation (9). and the updated log-likelihood estimate and judge and Does the difference between them meet the threshold? If so, then the updated weights will be used. Updated expectations The updated covariance matrix As the initial optimal weights Initial optimal expectation Initial optimal covariance matrix Otherwise, Assign to , Assign to , Assign to Then, return to step 4.1 and execute in sequence; (9).

[0008] Furthermore, step six includes: Step 6.1: Calculate the first step using equation (10). u New joint observation vector In the Gaussian Mixture Model (GMM), the first k New responsibility value for each Gaussian component : (10) Step 6.2: Use equations (11)-(13) to obtain the first Gaussian mixture model (GMM). k The weights of the Gaussian components after learning new samples Expectations after learning new samples The covariance matrix after learning new samples ; (11) (12) (13).

[0009] Furthermore, in step eight, the Gaussian Mixture Model (GMM) is obtained using equations (14)-(16). k The weights of the Gaussian components after discarding old samples Expectations after discarding old samples Covariance matrix after discarding old samples : (14) (15) (16) In equations (14)-(16), Indicates the deletion of the first Joint observation vectors In the Gaussian Mixture Model (GMM), the first k The responsibility value of each Gaussian component.

[0010] Furthermore, step twelve includes: Step 12.1, for Decompose to obtain the first k The mean vector of renewable energy predictions in Gaussian components and the k The mean vector of the actual values ​​of renewable energy in each Gaussian component ; right Decompose, the first k Covariance of renewable energy forecasts in Gaussian components , No. k Covariance of renewable energy predicted values ​​to actual values ​​in each Gaussian component , No. k Covariance of actual to predicted values ​​of renewable energy in each Gaussian component and the k Covariance of the actual value of renewable energy in each Gaussian component ; Step 12.2: Calculate the conditional probability distribution parameters using equations (17)-(19), including: the first... k Weights of Gaussian components ,expect Covariance matrix : (17) (18) (19) In equations (17)-(19), Represents a vector of predicted renewable energy values; Indicates expectation as The covariance matrix is The independent variable of the probability density function of the multivariate Gaussian distribution takes The function value at that time; Step 12.3, calculate using equation (20) Actual power vector under certain conditions conditional probability distribution : (20).

[0011] The present invention provides an electronic device, including a memory and a processor, characterized in that the memory is used to store a program that supports the processor in executing the uncertainty tracking method for renewable energy generation based on a recursive Gaussian mixture model, and the processor is configured to execute the program stored in the memory.

[0012] The present invention discloses a computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, performs the steps of the uncertainty tracking method for renewable energy generation based on a recursive Gaussian mixture model.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention employs a recursive update method, enabling the Gaussian mixture model describing the probability distribution of renewable energy generation to rapidly correct its parameters as new predicted and actual renewable energy power generation samples continuously arrive, without needing to repeatedly train on all historical renewable energy power generation data. This significantly reduces the computational burden of correcting the Gaussian mixture model parameters and allows the renewable energy generation probability model to reflect the latest changes in renewable energy output in a timely manner, thus meeting the efficiency and flexibility requirements for modeling the probability distribution of renewable energy generation in real-time power system operation scenarios.

[0014] 2. This invention effectively avoids the error accumulation problem caused by long-term iteration by introducing a periodic calibration mechanism on the basis of recursive updates. This method ensures real-time updates of the Gaussian mixture model parameters describing the probability distribution of renewable energy generation, while maintaining the accuracy and stability of modeling the probability distribution of renewable energy generation during the long-term operation of the power system. Attached Figure Description

[0015] Figure 1 This is a flowchart of the renewable energy generation uncertainty tracking method of the present invention; Figure 2 for Figure 1 Detailed flowchart of Part A. Detailed Implementation

[0016] In this embodiment, a method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model (GMM) organically combines three functions: incremental learning of GMM parameters, ordered replacement of outdated samples, and periodic calibration. The GMM parameters are updated recursively, ensuring the probabilistic model maintains both high update frequency and long-term accuracy in a streaming data environment. This enables real-time tracking of uncertainties in renewable energy generation. Specifically, as... Figure 1 As shown, the method includes the following steps: Step 1: Obtain Includes A renewable energy power generation The deterministic prediction curve for the first look-ahead point, and based on the first... Includes A renewable energy power generation The deterministic prediction curve for the nth forward point is obtained. indivual dimensional predicted power generation vector and the indivual dimensional actual power generation vector Thus constructing the first A collection of predicted and actual power generation. Joint observation vector of dimension Thus forming a joint observation set ;in, This represents the total number of joint observation vectors; Step 2: Set the number of clusters to [number]. and to Execute the k-means algorithm to obtain Clusters ,in, Indicates the first k One cluster; In the initial stage of model construction, the parameters of the Gaussian Mixture Model (GMM) need to be properly initialized. Direct random initialization may cause the Expectation-Maximization (EM) algorithm to get stuck in local optima, affecting the overall performance of the model. To avoid this problem, this invention first uses k-means clustering to partition the training samples, i.e., given the components of the GMM... ,use Each cluster performs the k-means algorithm on the training sample set.

[0017] Step 3, based on Clusters Calculate the first Gaussian mixture model (GMM) in Gaussian Mixture Model (GMM). k Initial parameters of the Gaussian components ,in, , , These represent the first and second halves of the Gaussian mixture model (GMM). k Initial weights, initial expectations, and initial covariance matrices of each Gaussian component; This Gaussian mixture model (GMM) is used to describe The probability distribution, The probability density function is shown in equation (1): (1) In equation (1), Indicates expectation as The covariance matrix is The independent variable of the probability density function of the multivariate Gaussian distribution takes The function value at time, and we have: (2) (3) Equation (2) is The calculation formula, where This represents the transpose of a matrix; Represents a random vector The dimension; Represents the covariance matrix The determinant of the matrix; Equation (3) indicates that the weights of each Gaussian component are greater than 0 and the sum is 1.

[0018] A Gaussian mixture model (GMM) is a probabilistic model that approximates arbitrarily complex distributions by weighting a sum of several Gaussian distributions. Compared to a single Gaussian model, the GMM can flexibly characterize multimodality, skewness, and nonlinear correlations, making it particularly suitable for describing the uncertainty characteristics of renewable energy power generation. In this invention, the predicted and actual values ​​of renewable energy power generation together form a sample vector. By modeling with a Gaussian mixture model, the joint probability distribution of the two can be obtained. This not only reflects the statistical characteristics of the prediction error, but also allows us to further derive the probability distribution of the actual value given the predicted value through the derivation of the conditional distribution.

[0019] This step is based on the first k Clusters The training samples in the calculation of the first k The parameters of each component are used to calculate the initial weights, initial expectations, and initial covariance matrix of the Gaussian mixture model using equations (1)-(3).

[0020] (4) (5) (6) In equations (4)-(6), express The number of joint observation vectors in; express The first in A joint observation vector.

[0021] Step 4: Estimate using the Expectation-Maximization (EM) algorithm indivual The parameters of a Gaussian mixture model (GMM) training sample set with joint observation vectors of dimension . .

[0022] Step 4.1: This step is the expected E step. The first step is calculated using equation (7). i Joint observation vectors In the Gaussian Mixture Model (GMM), the first kThe responsibility value of each Gaussian component : (7) Step 4.2: This step is to maximize the M-step. Using equations (8)-(10), the M-th step in the Gaussian Mixture Model (GMM) is obtained. k The updated weights of the Gaussian components Updated expectations The updated covariance matrix : (8) (9) (10) Step 4.3: Calculate the log-likelihood estimate using equation (11). and the updated log-likelihood estimate and judge and Does the difference between them meet the threshold? If so, then the updated weights will be used. Updated expectations The updated covariance matrix As the initial optimal weights Initial optimal expectation Initial optimal covariance matrix Otherwise, Assign to , Assign to , Assign to Then, return to step 4.1 and execute in sequence; (11).

[0023] Starting with the initial parameters obtained from clustering, the model gradually approaches a stable solution by repeatedly running the Expectation Maximization (EM) algorithm, thereby obtaining a set of reasonable and stable initial parameters, which provide reliable initial conditions for subsequent recursive updates.

[0024] Step 5: Initialization Define and initialize the total number of joint observation vectors to be deleted. =0; Step Six: Obtain the first b The batch of joint observation vector training sample set, i.e. A new joint observation vector and used for , , Update the model to obtain the Gaussian Mixture Model (GMM) of the first generation.k The weights of the Gaussian components are learned after new samples. Expectations after learning new samples The covariance matrix after learning new samples ;in, Indicates the first b The first batch A new joint observation vector; Indicates the first b The number of new joint observation vectors in the batch; Will and After piecing them together in order, we get the first... b Complete joint observation set of the batch ;in, express The first in Joint observation vectors; Step 6.1: Calculate the first step using equation (12). u New joint observation vector In the Gaussian Mixture Model (GMM), the first k New responsibility value for each Gaussian component : (12) Step 6.2: Use equations (13)-(15) to obtain the first Gaussian mixture model (GMM). k New weights after learning new samples from Gaussian components New expectations after learning from new samples The new covariance matrix after learning the new sample .

[0025] (13) (14) (15).

[0026] In the operation of the power system, the predicted-actual power generation of renewable energy is constantly generated, and new sample data arrives in a stream. If all samples are retrained each time, it will lead to huge computational overhead. To solve this problem, this invention proposes a recursive incremental update method. When a new sample arrives, the responsibility degree of the sample under each Gaussian component is calculated first using the current Gaussian mixture model parameters according to equation (12). The responsibility degree reflects the probability weight of the sample belonging to different components. Subsequently, the responsibility degree is used as a weight to incrementally correct the Gaussian mixture model parameters according to equations (13)-(15). Since the update only depends on the new sample and its responsibility degree, and does not need to re-traverse all historical samples, the computational complexity is significantly reduced. In this way, the Gaussian mixture model can quickly adapt to the distribution changes of new data without increasing the computational burden of updating parameters too much, and maintain a real-time reflection of the probability distribution of renewable energy power generation.

[0027] Step 7: Judgment Whether it is true or not, if true, then from Delete the earliest one Joint observation vectors , obtained the b Updated joint observation vector set for batches and will Assign to Proceed to step eight; otherwise, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed to step ten; among which, express The deleted number A joint observation vector, express The first in A joint observation vector, Indicates the sample threshold; Over time, older historical data may gradually lose its representativeness. If the training set expands indefinitely, it will not only cause storage pressure but may also reduce the model's sensitivity to the latest state. To address this issue, this invention designs an old sample deletion mechanism within the recursive update framework. In step seven, when the number of samples exceeds a preset capacity, the oldest sample is removed from the training set. The number of removed samples equals the number of newly acquired samples, thus achieving the replacement of old samples with new samples.

[0028] Step 8: Joint observation vector based on deletion According to formulas (16)-(18) , , Update the model to obtain the Gaussian Mixture Model (GMM) of the first generation. k The weights of the Gaussian components after discarding old samples Expectations after discarding old samples Covariance matrix after discarding old samples ; (16) (17) (18) In equations (16)-(18), Indicates the deletion of the first Joint observation vectors In the Gaussian Mixture Model (GMM), the first k The responsibility value of each Gaussian component.

[0029] Since the responsibility of a sample is calculated and saved when it enters the model, equations (16)-(18) indicate that when a sample is deleted, its contribution to the statistic can be directly subtracted, achieving rapid updates. In this way, the statistic always remains consistent with the current sample set, and the parameter estimates of the Gaussian mixture model are updated accordingly. Steps seven and eight ensure that the model focuses on the latest data, avoids the adverse effects of outdated information on modeling accuracy, and maintains limited computational and storage resource consumption.

[0030] Step Nine, Assign to Then, make a judgment Is it true? If yes, proceed to step ten; otherwise, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed to step eleven; among which, For the threshold; Step 10, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed according to step four to obtain... , , and will Set to 0; While recursive updates can significantly improve efficiency, repeated iterations over a long period can lead to accumulated biases, causing parameters to gradually deviate from the optimal solution. To address this, this invention proposes a periodic calibration mechanism. Steps nine and ten indicate that when the number of replaced old samples exceeds a certain threshold, an Expectation-Maximization (EM) training is triggered based on all samples. This process clears the gradually accumulated errors from recursive updates, restoring the overall optimality of the model. This design ensures the real-time performance of recursive updates while maintaining the long-term accuracy and stability of the Gaussian mixture model parameters through low-frequency calibration, achieving a balance between efficiency and accuracy.

[0031] Step 11, Judgment Is it true? If so, then... , , Each is used as the final optimal weight. Final optimal expectation The final optimal covariance matrix And proceed to step twelve; otherwise, Assign to b Return to step six and execute sequentially; among them, This represents the total number of batches. Steps five through eleven constitute a complete recursive Gaussian mixture model. The detailed process is as follows: Figure 2 As shown.

[0032] Step 12: Utilize the final optimal parameters The corresponding Gaussian Mixture Model (GMM) calculates the predicted renewable energy values ​​given a given vector. Actual power vector under certain conditions conditional probability distribution This is to enable real-time quantification and tracking of uncertainties in renewable energy generation.

[0033] Step 12.1, for Performing vector decomposition, as shown in equation (19), yields the first... k The mean vector of renewable energy predictions in Gaussian components and the mean vector of actual renewable energy values ; (19) right Perform matrix partitioning as shown in equation (20) to obtain the first... k Covariance of renewable energy forecasts in Gaussian components , No. k Covariance of renewable energy predicted values ​​to actual values ​​in each Gaussian component , No. kCovariance of actual to predicted values ​​of renewable energy in each Gaussian component and the k Covariance of the actual value of renewable energy in each Gaussian component ; (20) Step 12.2: Calculate the conditional probability distribution parameters using equations (21)-(23), including: the first... k Weights of Gaussian components ,expect Covariance matrix : (twenty one) (twenty two) (twenty three) In equations (21)-(23), Represents a vector of predicted renewable energy values; Indicates expectation as The covariance matrix is The independent variable of the probability density function of the multivariate Gaussian distribution takes The function value at that time.

[0034] Step 12.3: Using equation (24), calculate Actual power vector under certain conditions conditional probability distribution This enables real-time quantification and tracking of uncertainties in renewable energy power generation.

[0035] (twenty four) In power system applications, it is often necessary to obtain the probability distribution of the actual value under the condition that the predicted value is known. In the modeling stage, this invention uses a joint Gaussian mixture model to describe the "prediction-actual pair". After obtaining the joint distribution parameters, the expression for the mean and covariance of the conditional distribution can be derived based on the conditional properties of the Gaussian distribution, and the probability distribution function of the actual value under the prediction condition can be formed accordingly (24).

[0036] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0037] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model, characterized in that, Includes the following steps: Step 1: Obtain Includes A renewable energy power generation The deterministic prediction curve for the first look-ahead point, and based on the first... Includes A renewable energy power generation The deterministic prediction curve for the nth look-ahead point is obtained. indivual dimensional predicted power generation vector and the indivual dimensional actual power generation vector Thus constructing the first A collection of predicted and actual power generation. Joint observation vector of dimension Thus forming a joint observation set ;in, This represents the total number of joint observation vectors; Step 2: Set the number of clusters to [number]. and to Execute the k-means algorithm to obtain Clusters ,in, This represents the k-th cluster; Step 3, based on Clusters Calculate the initial parameters of the k-th Gaussian component in a Gaussian mixture model (GMM). ,in, , , Let represent the initial weights, initial expectations, and initial covariance matrix of the k-th Gaussian component in the Gaussian Mixture Model (GMM), respectively. Step 4, based on Using the EM-maximization algorithm to Update the parameters to obtain the initial optimal parameters. ;in, , , Let represent the initial optimal weights, initial optimal expectations, and initial optimal covariance matrix of the k-th Gaussian component in the Gaussian Mixture Model (GMM), respectively. Step 5: Initialization Define and initialize the total number of joint observation vectors to be deleted. =0; Step 6: Obtain the b batch A new joint observation vector and used for , , The update is performed to obtain the weights of the k-th Gaussian component in the Gaussian Mixture Model (GMM) after learning the new samples. Expectations after learning new samples The covariance matrix after learning new samples ;in, This indicates the first batch of batch b. A new joint observation vector; This indicates the number of new joint observation vectors in batch b; Will and After sequential splicing, the complete joint observation set of batch b is obtained. ;in, express The first in Joint observation vectors; Step 7: Judgment Whether it is true or not, if true, then from Delete the earliest one Joint observation vectors The updated joint observation vector set for the bth batch is obtained. Proceed to step eight; otherwise, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed to step ten; among which, express The deleted number A joint observation vector, express The first in A joint observation vector, Indicates the sample threshold; Step 8: Joint observation vector based on deletion Using equations (14)-(16) , , The update is performed to obtain the weight of the k-th Gaussian component in the Gaussian Mixture Model (GMM) after discarding old samples. Expectations after discarding old samples Covariance matrix after discarding old samples : (14) (15) (16) In equations (14)-(16), Indicates the deletion of the first Joint observation vectors The responsibility value for the k-th Gaussian component in a Gaussian mixture model (GMM); Step Nine, Assign to Then, make a judgment Is it true? If yes, proceed to step ten; otherwise, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed to step eleven; among which, For the threshold; Step 10, Assign to ,Will Assign to ,Will Assign to ,Will Assign to Then, proceed according to step four to obtain... , , and will Set to 0; Step 11, Judgment Is it true? If so, then... , , Each is used as the final optimal weight. Final optimal expectation The final optimal covariance matrix And proceed to step twelve; otherwise, Assign the value to b, then return to step six and execute sequentially; where, This represents the total number of batches. Step 12: Utilize the final optimal parameters The corresponding Gaussian Mixture Model (GMM) is used to calculate the given renewable energy prediction vector. Actual power vector under certain conditions conditional probability distribution This is to enable real-time quantification and tracking of uncertainties in renewable energy generation.

2. The method for tracking uncertainty in renewable energy generation based on a recursive Gaussian mixture model according to claim 1, characterized in that, In step three, the initial weights of the k-th Gaussian component in the Gaussian Mixture Model (GMM) are calculated using equations (1) to (3). Initial expectations Initial covariance matrix : (1) (2) (3) In equations (1)-(3), express The number of joint observation vectors in; express The first in A joint observation vector.

3. The method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model according to claim 1, characterized in that, Step four includes: Step 4.1: Calculate the i-th joint observation vector using equation (4). Responsibility value of the k-th Gaussian component in a Gaussian mixture model (GMM) : (4) In equation (4), Indicates expectation as The covariance matrix is The independent variable of the probability density function of the multivariate Gaussian distribution takes The function value at time, and we have: (5) In equation (5), This represents the transpose of a matrix; Represents a random vector The dimension; Represents the covariance matrix The determinant of; Step 4.2: Use equations (6)-(8) to obtain the updated weights of the k-th Gaussian component in the Gaussian Mixture Model (GMM). Updated expectations The updated covariance matrix : (6) (7) (8) Step 4.3: Calculate the log-likelihood estimate using equation (9). and the updated log-likelihood estimate and judge and Does the difference between them meet the threshold? If so, then the updated weights will be used. Updated expectations The updated covariance matrix As the initial optimal weights Initial optimal expectation Initial optimal covariance matrix Otherwise, Assign to , Assign to , Assign to Then, return to step 4.1 and execute in sequence; (9)。 4. The method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model according to claim 3, characterized in that, Step six includes: Step 6.1: Calculate the u-th new joint observation vector using equation (10). New responsibility value for the k-th Gaussian component in a Gaussian mixture model (GMM) : (10) Step 6.2: Use equations (11)-(13) to obtain the weight of the k-th Gaussian component in the Gaussian Mixture Model (GMM) after learning new samples. Expectations after learning new samples The covariance matrix after learning new samples ; (11) (12) (13)。 5. The method for tracking uncertainties in renewable energy generation based on a recursive Gaussian mixture model according to claim 1, characterized in that, Step twelve includes: Step 12.1, for By decomposing the components, we obtain the mean vector of the renewable energy predictions in the k-th Gaussian component. and the mean vector of the actual values ​​of renewable energy in the k-th Gaussian component ; right By decomposing the data, we obtain the covariance of the renewable energy predictions in the k-th Gaussian component. The covariance between the predicted and actual values ​​of renewable energy in the k-th Gaussian component The covariance between the actual and predicted values ​​of renewable energy in the k-th Gaussian component The covariance of the actual values ​​of renewable energy in the k-th Gaussian component ; Step 12.2: Calculate the conditional probability distribution parameters using equations (17)-(19), including the weight of the k-th Gaussian component in the conditional Gaussian mixture model. ,expect Covariance matrix : (17) (18) (19) In equations (17)-(19), Represents a vector of predicted renewable energy values; Indicates expectation as The covariance matrix is The independent variable of the probability density function of the multivariate Gaussian distribution takes The function value at that time; Step 12.3, calculate using equation (20) Actual power vector under certain conditions conditional probability distribution : (20)。 6. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store programs that support the processor in executing the renewable energy generation uncertainty tracking method based on the recursive Gaussian mixture model as described in any of claims 1-5, and the processor is configured to execute the programs stored in the memory.

7. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program, when run by the processor, executes the steps of the renewable energy generation uncertainty tracking method based on the recursive Gaussian mixture model as described in any of claims 1-5.

Images