Power grid net load fluctuation scene generation method, system and device based on ARIMA and Copula combined model and medium

Abstract

The invention belongs to the technical field of power system operation and planning, and discloses a power grid net load fluctuation scene generation method, system and device based on an ARIMA and Copula combined model and a medium, so as to solve the problem of poor scene generation accuracy. The method comprises the following steps: decomposing and reconstructing an original time sequence of the net load of the power grid by using discrete wavelet transform; taking permutation entropy minimization as an optimization target, adopting a variable chromosome length hybridization genetic algorithm to divide time segments for the low-frequency linear subsequences, and respectively establishing ARIMA models to generate linear trend scenes; establishing a joint probability distribution model of the high-frequency fluctuation subsequences at adjacent moments based on a Copula function, and deducing conditional probability distribution in combination with a Bayesian formula to generate a fluctuation scene; and the linear trend scene and the fluctuation scene are superposed to form an initial net load scene set, and a k-means clustering algorithm is adopted to reduce the initial net load scene set to obtain a representative scene set.

Description

Technical Field

[0001] This invention belongs to the field of power system operation and planning technology, specifically relating to a method, system, equipment and medium for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model. Background Technology

[0002] With the continuous integration of renewable energy sources such as wind and solar power into the grid, the intermittency and uncertainty of their output have led to increasingly complex fluctuations in the grid's net load (total load minus renewable energy output), exhibiting dynamic characteristics of high frequency, nonlinearity, and multi-scale coupling. This fluctuation poses severe challenges to the planning (such as energy storage configuration and power source layout) and operational decisions (such as unit combination and real-time dispatch) of new power systems.

[0003] As the core data foundation for system analysis and decision-making, the quality of power grid net load scenarios directly determines the reliability of planning schemes and the flexibility of operational strategies. Specifically, a high-quality scenario set needs to accurately capture the long-term evolution trend (low-frequency component) and short-term fluctuation characteristics (high-frequency component) of the net load, ensuring that key indicators such as offset rate (reflecting the degree of deviation between the scenario and actual data), ramp similarity (measuring the matching degree of power change rate), and time autocorrelation (characterizing the persistence of fluctuation patterns) meet the high reliability requirements of the system. Only in this way can planning schemes effectively balance economy and safety, and operational strategies flexibly respond to various fluctuating conditions.

[0004] Currently, various research methods exist for generating power system scenarios, but their technical principles and application limitations differ. For example, while constructing scenario generation models that consider the spatiotemporal correlation of wind power output using Copula functions or conditional generative adversarial networks can describe spatial and temporal correlations, they are prone to significant deviations in mean drift and long-term trends. Improving the representativeness of renewable energy output scenarios using modified k-means clustering methods often results in insufficient representativeness after reduction due to poor quality of the original scenarios. These scenario generation methods primarily rely on a single model to model the original data sequence, which limits their effectiveness.

[0005] Regarding the optimization of feature extraction methods, although various methods combining genetic algorithms can effectively reduce the complexity of the original dataset, their application in generating power grid net load scenarios is not yet perfect.

[0006] For the analysis and modeling of net load data in power grids, wavelet analysis, Kalman filtering, and time series analysis are commonly used, with the autoregressive integrated moving average (ARIMA) model being widely applied. However, the accuracy of the ARIMA model decreases when modeling predictions for longer time series. Furthermore, existing methods fail to adequately handle both the linear and fluctuating components of the net load data simultaneously, resulting in high scenario bias and significant differences in ramp-up rates compared to the original data, thus failing to accurately reflect the changing trends of the original data.

[0007] In summary, none of the above methods can simultaneously address the issue of balancing low-frequency trend accuracy and high-frequency fluctuation correlation. Consequently, the generated scenarios fail to meet the high reliability requirements of new power system planning and operation in terms of key indicators such as offset rate, ramp similarity, and time autocorrelation. Summary of the Invention

[0008] Based on the aforementioned shortcomings and deficiencies in the existing technology, one of the objectives of this invention is to at least solve one or more of the aforementioned problems in the existing technology. In other words, one of the objectives of this invention is to provide a method, system, device, and medium for generating power grid net load fluctuation scenarios based on the ARIMA and Copula combined model that meets one or more of the aforementioned requirements, thereby achieving comprehensive scenario assessment to provide reliable decision-making basis, improving practicality and adaptability, supporting power system decision-making, and ensuring its safe and economical operation.

[0009] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: In a first aspect, this invention provides a method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model, comprising the following steps: S1, decomposing and reconstructing the original time series of power grid net load using discrete wavelet transform, separating the original time series into a low-frequency linear subsequence and a high-frequency fluctuation subsequence; S2, using the minimization of permutation entropy as the optimization objective, using a variable chromosome length hybridization genetic algorithm to divide the low-frequency linear subsequence into time segments, and establishing an ARIMA model in each time segment to generate a linear trend scenario; S3, establishing a joint probability distribution model of the high-frequency fluctuation subsequence at adjacent times based on the Copula function, deriving the conditional probability distribution using Bayes' theorem, and generating fluctuation scenarios; S4, superimposing the linear trend scenario generated in step S2 with the fluctuation scenario generated in step S3 to form an initial net load scenario set, and using a k-means clustering algorithm to reduce the initial scenario set to obtain a representative scenario set; S5, constructing an evaluation index system to quantitatively evaluate the performance of the representative scenario set.

[0010] As a preferred embodiment, step S1 includes the following steps: acquiring historical net load data of the power grid to form an original time series; pre-setting the wavelet decomposition level; using a low-pass filter and a high-pass filter to decompose the original time series of the power grid net load level by level according to the pre-set decomposition level, reconstructing the original time series to obtain a low-frequency linear subsequence representing the long-term trend and a high-frequency fluctuation subsequence representing the short-term fluctuation characteristics.

[0011] As a preferred approach, step S2, which involves using a variable chromosome length hybridization genetic algorithm to divide the low-frequency linear subsequence into time segments, includes the following steps: minimizing the mean of the permutation entropy within each time segment after division as the objective function; employing a variable chromosome length encoding strategy to dynamically set the chromosome length to the length of the corresponding segment in order to solve the objective function and obtain the optimal number of segment divisions and the corresponding segment division scheme; and introducing a survival factor as a criterion for whether an individual participates in the crossover operation, wherein the survival factor is determined by the individual's fitness, and the fitness is the reciprocal of the permutation entropy.

[0012] As a preferred approach, step S2, which involves establishing an ARIMA model for each time interval to generate a linear trend scenario, includes the following steps: determining the difference order based on the unit root test. d， Determining the autoregressive order based on the Akaike information criterion p and moving average order q The Markov transition matrix is ​​determined using the permutation entropy information of the initial scene. Q The Markov transition matrix Q Used to characterize the probability of transitioning from the current segment to the next segment; based on the Markov transition matrix. Q Determine the segment category to which the current scene belongs, and call the corresponding ARIMA model to generate a linear trend scene.

[0013] As a preferred embodiment, step S3, which involves establishing a joint probability distribution model of the high-frequency fluctuation subsequence at adjacent time points based on the Copula function, includes the following steps: selecting the Copula function type and calculating the Kendall rank correlation coefficient. Spearman rank correlation coefficient To evaluate the goodness of fit of the selected Copula function to the dependent structure of the original wave sequence; define t The net load fluctuation ratio at each time point is determined, and the joint probability density function of the fluctuation ratio at adjacent time points is obtained based on Copula theory. The probability sampling model is obtained by solving the joint probability density function using Bayes' theorem, and the fluctuation scenario is generated accordingly.

[0014] As a preferred approach, in step S4, when the k-means clustering algorithm is used to reduce the initial scene set: using net load power as the sample attribute and Euclidean distance as the measure of distance between samples, the initial scene set is reduced to a representative scene set by iteratively updating the cluster centers.

[0015] As a preferred option, the evaluation index system in step S5 includes at least two of the following indicators: time autocorrelation index, average offset rate index, climbing similarity index, and mean absolute percentage error index.

[0016] Secondly, the present invention provides a power grid net load fluctuation scenario generation system based on a combined ARIMA and Copula model, for implementing the power grid net load fluctuation scenario generation method as described in the first aspect.

[0017] Thirdly, the present invention provides an electronic device, the computer device including a memory, a processor and a computer program, wherein when the computer program is executed by the processor, it implements the grid net load fluctuation scenario generation method as described in the first aspect.

[0018] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the grid net load fluctuation scenario generation method as described in the first aspect.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. Existing technologies often struggle to comprehensively and accurately capture the complex dynamic changes in net load across different time scales when generating power grid net load fluctuation scenarios. This invention, however, decomposes the original time series of power grid net load into low-frequency linear subsequences and high-frequency fluctuating subsequences using discrete wavelet transform, achieving time-frequency domain decoupling of the signal. This decomposition method allows for in-depth analysis and modeling of both the linear and fluctuating components, resulting in a more comprehensive and accurate characterization of the power grid net load fluctuations and providing richer and more accurate foundational information for subsequent scenario generation.

[0020] 2. When processing low-frequency linear subsequences, this invention uses minimizing permutation entropy as the optimization objective and employs a variable chromosome length hybridization genetic algorithm to divide time segments. Compared with the fixed segmentation methods commonly used in existing technologies, this algorithm can identify locally stationary segments in the sequence with relatively consistent statistical characteristics, avoiding modeling errors caused by unreasonable segmentation and further improving the accuracy of scene generation. This precise segmentation helps to more accurately simulate the changing trend of net load in different time periods, providing a more reliable basis for power system planning.

[0021] 3. Existing technologies have limitations in handling the correlation between multiple random variables in power grid net load fluctuations. This invention introduces a Copula function to establish a joint probability distribution model of high-frequency fluctuation subsequences at adjacent time points. The Copula function can decompose the joint distribution of multiple random variables into their respective marginal distributions and a Copula function, thereby independently characterizing the dependencies between variables without being limited by the form of the marginal distribution. This allows us to more accurately describe the correlation between fluctuation characteristics. Compared with existing technologies, the generated fluctuation scenario more closely matches the actual probability distribution, providing a more accurate reference for the operation and control of the power system.

[0022] 4. Based on the joint probability distribution model established using Copula functions, this invention combines Bayes' theorem to derive conditional probability distributions, enabling random sampling and scenario generation of fluctuation characteristics. This approach fully utilizes the probabilistic reasoning capabilities of Bayes' theorem, allowing for more reasonable generation of fluctuation scenarios based on known information, further improving the accuracy of correlation modeling. Compared to existing technologies, the generated scenarios better reflect the changes in net load fluctuations under different conditions, helping power systems formulate more scientific response strategies.

[0023] 5. Existing technologies typically employ a single or a few indicators when evaluating generated power grid net load fluctuation scenarios, making it difficult to comprehensively and objectively assess the quality of the scenarios. This invention introduces multiple evaluation indicators, such as time autocorrelation, average offset rate, ramp similarity, and mean absolute percentage error, to comprehensively evaluate the generated representative scenario set from different perspectives.

[0024] 6. When faced with a large number of initial net load scenarios, existing technologies may not be able to effectively reduce the scenario set, leading to excessive computational load and affecting the efficiency of subsequent power system analysis and decision-making. This invention employs a k-means clustering algorithm to reduce the initial net load scenario set, extracting several representative scenarios from a large number of initial scenarios. Compared with existing technologies, this algorithm reduces computational load while preserving typical scenario characteristics, thus improving the efficiency of subsequent power system analysis and decision-making. This reasonable scenario set reduction method allows us to perform power system analysis and decision-making more efficiently while ensuring scenario quality.

[0025] Further or more detailed beneficial effects will be described in conjunction with specific embodiments in the detailed implementation. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating the method for generating power grid net load fluctuation scenarios as described in Embodiment 1 of the present invention.

[0028] Figure 2 This is a structural diagram of the electronic device provided in the embodiment of the present invention.

[0029] Figure 3 This is a schematic diagram of the ARIMA-Copula combined modeling framework described in Embodiment 5 of the present invention.

[0030] Figure 4 This is a numerical diagram of the MAPE index of each segment of the ARIMA model described in Embodiment 5 of the present invention.

[0031] Figure 5 This is a graph showing the fitting effect of the fluctuation sequence based on the normal Copula function as described in Embodiment 5 of the present invention.

[0032] Figure 6 This is the scene generation result diagram described in Embodiment 5 of the present invention.

[0033] Figure 7 This is a performance index diagram of the different scene generation methods described in Embodiment 5 of the present invention.

[0034] Icon labels: 200. Electronic devices; 201. Processor; 202. Communication bus; 203. User interface; 204. Network interface; 205. Memory. Detailed Implementation

[0035] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0036] In the following description, several embodiments of the present invention are provided. Different embodiments can be substituted or combined. Therefore, the present invention can also be considered to include all possible combinations of the same and / or different embodiments described. Thus, if one embodiment includes features A, B, and C, and another embodiment includes features B and D, then the present invention should also be considered to include embodiments containing one or more other possible combinations of A, B, C, and D, even if such embodiments are not explicitly described in the following text.

[0037] The following description provides examples and does not limit the scope, applicability, or examples set forth in the claims. Changes may be made to the function and arrangement of the described elements without departing from the scope of the invention. Various processes or components may be appropriately omitted, substituted, or added to the various examples. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined into other examples.

[0038] To facilitate a better understanding of the embodiments of the present invention, its application scenarios will be explained before providing a detailed explanation of the specific implementation methods.

[0039] The power grid net load fluctuation scenario generation method described in the embodiments of this specification is applied to the planning, operation, and control processes of power systems. In these scenarios, the application of this method aims to accurately depict the dynamic changes in power grid net load at different time scales, providing a reliable basis for the power system to formulate reasonable generation plans, energy storage configuration strategies, and responses to load fluctuations, thereby improving the stability, reliability, and economy of the power system. By generating representative net load fluctuation scenarios, various load conditions that may occur in actual operation can be simulated, helping power dispatchers to prepare in advance and reduce the risks and losses caused by load fluctuations.

[0040] The following is a brief explanation of the ARIMA, Copula, power grid net load fluctuation scenario generation, discrete wavelet transform, permutation entropy, variable chromosome length hybridization genetic algorithm, Bayesian formula, k-means clustering algorithm, time autocorrelation index, average offset index, ramp similarity index, and mean absolute percentage error index involved in the various embodiments of this specification: ARIMA (Autoregressive Integral Moving Average) is a classic time series forecasting model. It combines the features of autoregressive (AR) models, differencing (I) operations, and moving average (MA) models. The autoregressive component uses past values ​​of the time series for forecasting, the differencing operation transforms non-stationary time series into stationary ones, and the moving average component considers past values ​​of the forecast error. By analyzing time series data, the ARIMA model determines appropriate parameters (autoregressive order p, differencing order d, and moving average order q) to establish a mathematical model for predicting future values, demonstrating good performance in handling time series data with linear characteristics.

[0041] The Copula function is a function used to describe the correlation structure between multiple random variables. It decomposes the joint distribution of multiple random variables into their respective marginal distributions and a Copula function. The Copula function can independently characterize the dependencies between variables, without being limited by the form of the marginal distributions. In the generation of power grid net load fluctuation scenarios, the Copula function can be used to establish a joint probability distribution model of high-frequency fluctuation subsequences at adjacent time points, thereby more accurately describing the correlation between fluctuation characteristics and providing strong support for generating realistic fluctuation scenarios.

[0042] Grid net load fluctuation scenario generation refers to the process of generating a set of scenarios that reflect potential future fluctuations in grid net load based on historical grid net load data and using specific methods and models. These scenarios can simulate the changing trends and fluctuation characteristics of net load at different time scales, helping power system stakeholders gain a deeper understanding of the dynamic characteristics of net load and providing decision-making support for power system planning, operation, and control. For example, they can be used to evaluate the rationality of generation plans, the effectiveness of energy storage system configurations, and the ability to cope with sudden load changes.

[0043] Discrete wavelet transform (DWT) is a time-frequency analysis method that decomposes a signal into different frequency sub-bands while preserving its temporal information. Unlike Fourier transform, DWT can analyze the frequency components of a signal within a local range, better capturing its non-stationary characteristics. In power grid net load data processing, DWT can decompose the original time series of power grid net load into low-frequency linear subsequences and high-frequency fluctuating subsequences, achieving time-frequency domain decoupling and facilitating subsequent modeling of the linear and fluctuating components separately.

[0044] Permutation entropy is an indicator used to measure the complexity and irregularity of a time series. It calculates the entropy value by analyzing the permutation patterns of adjacent data points in the time series. A higher entropy value indicates greater complexity and randomness in the time series; a lower entropy value indicates stronger regularity and determinism. In the generation of power grid net load fluctuation scenarios, minimizing permutation entropy is used as the optimization objective. A variable chromosome length hybridization genetic algorithm is employed to divide the low-frequency linear subsequence into time segments. This can identify locally stationary segments with relatively consistent statistical characteristics within the sequence, improving the modeling accuracy.

[0045] Variable chromosome length hybridization genetic algorithm is an improved genetic algorithm. Traditional genetic algorithms typically use a fixed chromosome length, while variable chromosome length hybridization allows the chromosome length to change during evolution. This algorithm combines hybridization operations, simulating gene crossover and mutation in biological evolution to search for the optimal solution in the solution space. In generating power grid net load fluctuation scenarios, this algorithm uses minimizing permutation entropy as its optimization objective, dividing low-frequency linear subsequences into time segments. This effectively avoids getting trapped in local optima, improves search efficiency and global search capability, and thus more accurately divides time segments.

[0046] Bayes' theorem is a probability-based reasoning method used to update the probability of an event given certain known conditions. It describes how to calculate the posterior probability based on prior and conditional probabilities after acquiring new information. In generating power grid net load fluctuation scenarios, after establishing a joint probability distribution model of high-frequency fluctuation subsequences at adjacent times based on the Copula function, the conditional probability distribution can be derived by combining it with Bayes' theorem. This enables random sampling of fluctuation characteristics and scenario generation, making the generated fluctuation scenario more consistent with the actual probability distribution.

[0047] k-means clustering is a commonly used unsupervised learning algorithm for dividing a dataset into k distinct clusters. This algorithm iteratively adjusts the cluster centroids to minimize the sum of the distances from each data point to the centroid of its cluster. In generating power grid net load fluctuation scenarios, k-means clustering is used to reduce the initial net load scenario set. This allows for the extraction of several representative scenarios from a large set of initial scenarios, preserving typical scenario characteristics while reducing computational load and improving the efficiency of subsequent power system analysis and decision-making.

[0048] The time autocorrelation index is used to measure the correlation between time series at different points in time. It reflects the degree of dependence between the current value and past values ​​of a time series. In the assessment of power grid net load fluctuation scenarios, the time autocorrelation index can evaluate whether the generated representative scenario set can accurately reflect the correlation characteristics of net load at different points in time, that is, whether the fluctuation of net load has a certain continuity and regularity, thereby judging the effectiveness of the scenario generation method.

[0049] The average deviation rate metric measures the average degree of deviation between the generated representative scenario set and the actual net load data. It reflects the magnitude of the deviation by calculating the difference between each scenario in the scenario set and the corresponding time point in the actual data, and then taking the average of these differences. This metric provides a direct assessment of how closely the generated scenario set approximates the actual situation; the smaller the average deviation rate, the more accurately the generated scenario set reflects changes in the actual net load.

[0050] The ramp similarity index is primarily used to assess the similarity between the ramp characteristics of the net load in a generated representative scenario set and the actual net load ramp characteristics. Ramp characteristics reflect the significant changes in net load over a short period of time, which is crucial for power system dispatching and operation. By calculating the ramp similarity index, it is possible to determine whether the generated scenario set can accurately simulate the ramp behavior of the actual net load, providing a reference for the formulation of power system response strategies.

[0051] The Mean Absolute Percentage Error (MAPE) is a commonly used metric for evaluating forecast accuracy. It measures the accuracy of a forecast by calculating the average absolute percentage error between the predicted and actual values. In assessing power grid net load fluctuation scenarios, this metric can evaluate the forecast accuracy of a representative set of scenarios for the actual net load. The smaller the MAPE, the more accurately the generated scenario set can predict changes in the actual net load, and the better the forecasting effect of the method.

[0052] Example 1: like Figure 1 As shown, this embodiment provides a method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model, including the following steps: Step S1: Use discrete wavelet transform to decompose and reconstruct the original time series of net load of the power grid, and separate the original time series into a low-frequency linear subsequence and a high-frequency fluctuation subsequence.

[0053] The specific operation of step S1 is as follows: First, obtain the historical net load data of the power grid to form an original time series, and a preset wavelet decomposition level. Then, use a low-pass filter and a high-pass filter to decompose the original time series of the power grid net load level by level according to the preset decomposition level, and reconstruct the original time series to obtain a low-frequency linear subsequence that represents the long-term trend and a high-frequency fluctuation subsequence that represents the short-term fluctuation characteristics.

[0054] The more specific mathematical expression is: In new power systems, given the strong randomness of new energy output and load fluctuations, net load... It can be represented as: (1) In formula (1): For load power, For new energy power generation; The DWT method was used to decompose the net load time series, extracting its high-frequency and low-frequency components to obtain high-frequency fluctuation subsequences and low-frequency linear subsequences, respectively reflecting the load fluctuation characteristics and long-term trends. Among these, the low-frequency linear subsequence... It can be represented as: (2) In formula (2): For DWT refactoring functions, For low-pass filter wavelet decomposition coefficients, This represents the number of decomposition levels in DWT. For low-pass filter The length of the decomposed sequence; High-frequency fluctuation subsequence The degree of fluctuation in a dataset can be represented as: (3) In formula (3): For high-pass filters wavelet decomposition coefficients, For high-pass filters The length of the decomposed sequence.

[0055] Step S2: With minimizing permutation entropy as the optimization objective, the low-frequency linear subsequence is divided into time segments using a variable chromosome length hybridization genetic algorithm, and an ARIMA model is established in each time segment to generate a linear trend scenario.

[0056] The specific operation of dividing time segments in step S2 is as follows: In order to improve the stationarity and prediction accuracy of subsequent time series modeling, the objective function is to minimize the mean of permutation entropy within each time segment after division, so as to achieve global optimization of the stationarity characteristics of the sequence and make the dynamic characteristics within each sub-segment as consistent as possible, thereby meeting the modeling requirements of ARIMA-type models for the local stationarity of the sequence; a variable chromosome length encoding strategy is adopted to dynamically set the chromosome length to the length of the corresponding segment in order to solve the objective function and obtain the optimal number of segment divisions and the corresponding segment division scheme; a survival factor is introduced as a criterion for whether an individual participates in the crossover operation. The survival factor is determined by the individual fitness, where fitness is the reciprocal of permutation entropy.

[0057] The specific operation for generating the linear trend scenario in step S2 is as follows: determine the difference order based on the unit root test. d， Determining the autoregressive order based on the Akaike information criterion p and moving average order q The Markov transition matrix is ​​determined using the permutation entropy information of the initial scene. Q The Markov transition matrix Q Used to characterize the probability of transitioning from the current segment to the next segment; based on the Markov transition matrix. Q Determine the segment category to which the current scene belongs, and call the corresponding ARIMA model to generate a linear trend scene.

[0058] The more detailed mathematical calculation process is as follows: For low-frequency linear subsequences Reconstructing the phase space using permutation entropy yields the following reconstruction matrix: (4) In equation (4): , For the embedding dimension, For delay time; No. Permutation entropy of segment sequence It can be represented as: (5) In equation (5): For the first The probability of a sort occurring; The objective function, taking the minimization of the mean entropy within each partitioned segment as its objective, can be expressed as: (6) In equation (6): The number of segments to be divided; Given that the objective function (6) is nonlinear, a variable chromosome length genetic algorithm is used to solve it. This algorithm introduces a variable chromosome length encoding mechanism and combines it with corresponding crossover operators to achieve adaptive segmentation and optimal solution approximation in the search space. Specifically: 1. Variable chromosome length coding: Traditional genetic algorithms employ fixed-length chromosome encoding, where the encoding length of each chromosome is typically equal to the total length of the linear sequence being processed. When optimizing the segment lengths of the linear sequence, a fixed chromosome length leads to a verbose encoding structure, significantly increasing computational complexity and impacting the algorithm's solution efficiency and convergence speed. This embodiment introduces a variable chromosome length encoding strategy, dynamically setting the length of each chromosome to match the length of the corresponding segment based on the actual linear segment division, thereby achieving more efficient encoding representation and search performance.

[0059] 2. Hybridization operator: To simulate the phenomenon in biological hybridization where some individuals are unable to produce fertile offspring due to poor adaptability, survival factors are introduced. As a criterion for an individual's eligibility to participate in reproduction. When an individual's survival factor... When the value is less than 1, the individual is considered incapable of reproduction and will not participate in the crossover process during genetic manipulation. This includes the individual survival factor. It can be represented as: (7) In equation (7): For the population number Individual Survival factors For the first Individual Adaptability, For the first The minimum fitness of an individual is determined by taking the reciprocal of the permutation entropy as the fitness value. Then, the fitness of the individual in the population... The fitness is: (8) Determined based on Akaike Information Criteria d and q The value of can be expressed as: (9) (10) (11) In equations (9)-(11): and They are respectively Time and A linear sequence of time steps. It is normal white noise. For backward shift operands, These are the autoregressive coefficients. The moving average coefficient; Markov transition matrix Q The expression is: (12) In equation (12): To divide the transition probabilities between scene sets, and .

[0060] Step S3: Based on the Copula function, establish the joint probability distribution model of the high-frequency fluctuation subsequence at adjacent time points, and derive the conditional probability distribution using Bayes' theorem to generate the fluctuation scenario.

[0061] The specific steps for generating the fluctuation scenario in step S3 are as follows: select the Copula function type and calculate the Kendall rank correlation coefficient. Spearman rank correlation coefficient To evaluate the goodness of fit of the selected Copula function to the dependent structure of the original wave sequence; define t The net load fluctuation ratio at each time point is determined, and the joint probability density function of the fluctuation ratio at adjacent time points is obtained based on Copula theory. The probability sampling model is obtained by solving the joint probability density function using Bayes' theorem, and the fluctuation scenario is generated accordingly.

[0062] The more detailed mathematical calculation process is as follows: Regarding Copula theory, the following definition applies: Hypothetical variables The joint distribution function is The cumulative probability marginal distribution is Then the Copula function satisfies: (13) like Continuous, then The joint probability density function of the random vector can be uniquely determined by taking its partial derivative: (14) Kendall rank correlation coefficient for: (15) Spearman rank correlation coefficient for: (16) In equations (15)-(16): ( V 1 ,V 2) and ( U 1 ,U 2) These are independent random vectors that follow the same distribution. P ( ∙ ) is the probability density function.

[0063] Define the net load fluctuation ratio at time t. for: (17) In equation (16): for t Time-based net load fluctuation sequence for t Linear sequence of net load at any given time.

[0064] Based on Copula theory, the net load fluctuation ratio between adjacent time points can be obtained. t- 1 and t The joint probability density function is: (18) In equation (16): and They are respectively t and t- The marginal distribution function of the net load fluctuation sequence at time 1. for t and t- The joint probability density function of the net load fluctuation sequence at time 1; By using Bayes' theorem to solve for the conditional probability distribution function between adjacent time steps, a probability sampling model for the fluctuations in scene generation can be obtained. for: (19).

[0065] Step S4: Overlay the linear trend scenario generated in step S2 with the fluctuation scenario generated in step S3 to form an initial net load scenario set, and use the k-means clustering algorithm to reduce the initial scenario set to obtain a representative scenario set.

[0066] The specific operation of reducing the initial scene set in step S4 is as follows: In order to reduce the redundancy of the generated scenes and improve the computational efficiency in subsequent optimization scheduling and other applications, while retaining the typical features and key probability information in the original scene set, the net load power is used as the sample attribute, and Euclidean distance is used as the measure of the distance between samples. By iteratively updating the cluster center, the initial scene set is reduced to a representative scene set.

[0067] The more detailed mathematical calculation process is as follows: Pick k Points As cluster centers, the sample attribute is net load power. arrive distance Represented as: (20) In equation (20): This is the distance calculation function. and Samples i and j The k One attribute, q It is an order, and q =2; Cluster centers Iterative updates can be represented as: (twenty one) (twenty two) In equations (20)-(22): m The number of the sample with the smallest distance. For the sample j To the cluster center The distance.

[0068] Step S5: Construct an evaluation index system to quantitatively evaluate the performance of the representative scenario set.

[0069] The evaluation index system described in step S5 includes at least two of the following indicators: time autocorrelation index, average offset rate index, climbing similarity index, and mean absolute percentage error index.

[0070] Example 2: This embodiment provides a power grid net load fluctuation scenario generation system based on a combined ARIMA and Copula model, used to implement the power grid net load fluctuation scenario generation method as described in Embodiment 1.

[0071] Example 3: like Figure 2 As shown, this embodiment provides an electronic device, which may include: at least one processor, at least one network interface, a user interface, a memory, and at least one communication bus.

[0072] The communication bus can be used to enable communication between the various components mentioned above.

[0073] The user interface may include buttons, and optional user interfaces may also include standard wired interfaces and wireless interfaces.

[0074] The network interface may include, but is not limited to, Bluetooth modules, NFC modules, Wi-Fi modules, etc.

[0075] The processor may include one or more processing cores. It connects various parts of the electronic device via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in memory, and accessing data stored in memory to perform various functions and process data. Optionally, the processor can be implemented using at least one hardware form of DSP, FPGA, or PLA. The processor may integrate one or more of the following: CPU, GPU, and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor.

[0076] The memory may include RAM or ROM. Optionally, the memory may include a non-transitory computer-readable medium. The memory may be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor. The memory, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a generated application program. The processor may be used to call the generated application program stored in the memory and execute the steps of the power grid net load fluctuation scenario generation method mentioned in the foregoing embodiments.

[0077] Example 4: This embodiment provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform one or more steps as described in the above embodiments. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0078] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0079] Those skilled in the art will understand that all or part of the processes in the method of Embodiment 1 described above can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and the implementation scheme can be combined arbitrarily.

[0080] Example 5: To verify the effectiveness of the method for generating power grid net load fluctuation scenarios based on the ARIMA and Copula combined model described in this specification, this embodiment selects historical net load data of a power grid in a certain region of my country as a case study for analysis and verification.

[0081] During implementation, the original net load sequence is first preprocessed, and then based on, as follows: Figure 3 The ARIMA-Copula combinatorial modeling framework shown is used for scene generation and evaluation. The specific steps are as follows: Step 1: The Discrete Wavelet Transform (DWT) method is used to perform multi-resolution decomposition on the net load time series, with the number of decomposition levels set to 6. Through step-by-step processing with low-pass and high-pass filter banks, the original sequence is reconstructed into low-frequency trend components (linear part) and high-frequency detail components (nonlinear fluctuation part), thereby achieving time-frequency domain decoupling of the dynamic characteristics of the net load.

[0082] Step 2: For the extracted linear trend components, with the optimization objective of minimizing the mean permutation entropy within each partitioned segment, an adaptive segmentation algorithm using variable chromosome length hybridization is employed. Specific parameter configurations are as follows: chromosome length of population 1... L Population 1 = 10, initial population size is 40; chromosome length of population 2 L The algorithm has a 2^20 subset of 40 individuals, an initial population size of 40, a maximum number of iterations of 300 generations, a crossover probability of 0.43, a deletion probability of 0.03, and a splicing probability of 0.03. After optimization, the optimal number of segment divisions is 15.

[0083] Subsequently, the difference order was determined using the unit root test. d The optimal autoregression order was selected based on the Akaike information criterion. p With moving average order q， ARIMA is constructed in each segment separately. p , d , q The ARIMA model for each segment has the following mean absolute percentage error (MAPE) values: Figure 4 As shown. Analysis Figure 4 It can be seen that when the permutation entropy value of the net load subsequence within each segment is low, the MAPE index value of the corresponding ARIMA model also decreases, indicating that the dynamic complexity of the sequence is negatively correlated with its modeling accuracy. That is, the stronger the regularity and the weaker the randomness of the sequence, the better the fitting effect of the ARIMA model. Further analysis reveals that when traditional ARIMA models perform global modeling on complete long sequences, they are prone to model parameter mismatch due to the difficulty in taking into account the time-varying and non-stationarity of the statistical characteristics (such as mean, variance, and autocorrelation) of the sequence in different time periods, which leads to a decrease in prediction accuracy. In contrast, the adaptive segmented modeling strategy based on permutation entropy minimization proposed in this invention can effectively identify locally stationary segments with relatively consistent statistical characteristics in the sequence, significantly reduce the inherent complexity of each sub-segment, and improve the fitting ability and prediction stability of the ARIMA model in the local range. This strategy overcomes the technical deficiency of traditional single ARIMA models in their insufficient ability to model high-complexity long sequences, enhances the model's adaptability and expressive ability to time-varying load characteristics, and thus significantly improves the accuracy and reliability of the overall scene generation.

[0084] Step 3: For the high-frequency fluctuation component, a multidimensional joint probability distribution model is established using the Copula function. To select the optimal Copula structure, the fitting effect of various functions such as the normal Copula, t-Copula, and Gumbel Copula is evaluated using Kendall's rank correlation coefficient. The results show that the normal Copula function has the highest goodness of fit and can accurately characterize the interdependence structure of the fluctuation sequence between adjacent time points. Therefore, this embodiment selects the normal Copula function to construct the probability density model of the fluctuation subsequence and combines it with Bayesian inference methods to achieve random sampling of fluctuation features and scene generation. The corresponding fitting results are as follows: Figure 5 As shown Based on the above modeling results, the combined model constructed in this invention consists of a linear trend module composed of 15 local ARIMA models and a fluctuation module based on the normal Copula-Bayes distribution. By superimposing the two types of components, a total of 1000 net load time series scenarios are generated, such as... Figure 6 As shown, the statistical characteristics and dynamic fluctuation patterns of the original data are illustrated.

[0085] Step 4: In order to further improve computational efficiency and retain typical scenario features, this embodiment uses the k-means clustering algorithm to reduce the 1,000 original scenarios generated, and finally extracts several representative scenarios for subsequent power system analysis and decision-making.

[0086] To verify the superiority of the method of this invention, this embodiment selects Monte Carlo sampling, the traditional Copula function method, and the Conditional Generative Adversarial Network (CGAN) method as comparative methods. All methods generate 1000 scenes and use the same k-means clustering strategy for scene reduction. To quantitatively evaluate the quality of the generated scenes, a multi-dimensional evaluation system is constructed, including temporal autocorrelation indicators, average offset indicators, and climbing similarity indicators. The calculation results are as follows: Figure 7 As shown. By Figure 7 It is known that the Monte Carlo sampling method generates scenes with weak temporal autocorrelation and significant deviation from the original data, exhibiting low similarity in the climbing process and failing to accurately reflect the dynamic trend of net load changes. The traditional Copula method performs well in maintaining temporal correlation and climbing characteristics, but its average deviation rate remains high. The CGAN method shows improvement in reducing deviation and increasing climbing similarity. In contrast, the method proposed in this invention significantly improves the fidelity of scene generation by decoupling and modeling linear and nonlinear components separately. The resulting average deviation rate and climbing similarity are significantly better than the comparative methods, fully verifying the effectiveness and advancement of this invention in terms of statistical consistency, trend preservation ability, and overall modeling accuracy.

[0087] In summary, this embodiment fully implements a method for generating net load scenarios for power grids based on a combined ARIMA and Copula model, demonstrating good engineering applicability and promotional value. This embodiment verifies the effectiveness of the method proposed in this specification.

[0088] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0089] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0090] The above description is merely an exemplary embodiment of the present invention and should not be construed as limiting the scope of the invention. Any equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of embodiments of the invention upon considering the specification and practicing the disclosure herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of the invention are defined by the claims.

Claims

1. A method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model, characterized in that, Including the following steps: S1. The original time series of net load of the power grid is decomposed and reconstructed using discrete wavelet transform, and the original time series is separated into a low-frequency linear subsequence and a high-frequency fluctuation subsequence. S2. With minimizing permutation entropy as the optimization objective, a variable chromosome length hybridization genetic algorithm is used to divide the low-frequency linear subsequence into time segments, and an ARIMA model is established in each time segment to generate a linear trend scenario. S3. Based on the Copula function, establish a joint probability distribution model of the high-frequency fluctuation subsequence at adjacent time points, and derive the conditional probability distribution using Bayes' theorem to generate the fluctuation scenario. S4. Overlay the linear trend scenario generated in step S2 with the fluctuation scenario generated in step S3 to form an initial net load scenario set, and use the k-means clustering algorithm to reduce the initial scenario set to obtain a representative scenario set. S5. Construct an evaluation index system to quantitatively evaluate the performance of the representative scenario set.

2. The method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model according to claim 1, characterized in that, Step S1 includes the following steps: Historical net load data of the power grid is acquired to form a raw time series; Preset the number of wavelet decomposition levels; The original time series of the net load of the power grid is decomposed step by step using low-pass and high-pass filters according to a preset number of decomposition levels. The original time series is then reconstructed to obtain a low-frequency linear subsequence that represents the long-term trend and a high-frequency fluctuation subsequence that represents the short-term fluctuation characteristics.

3. The method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model according to claim 2, characterized in that, Step S2, which involves using a variable chromosome length hybridization genetic algorithm to divide the low-frequency linear subsequence into time segments, includes the following steps: The objective function is to minimize the mean of the permutation entropy within each time interval after the division. A variable chromosome length encoding strategy is adopted, in which the chromosome length is dynamically set to the length of the corresponding segment in order to solve the objective function and obtain the optimal number of segment divisions and the corresponding segment division scheme. A survival factor is introduced as a criterion for whether an individual participates in the crossover operation. The survival factor is determined by the individual's fitness, where fitness is the reciprocal of the permutation entropy.

4. The method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model according to claim 3, characterized in that, Step S2, which describes establishing an ARIMA model for each time interval to generate a linear trend scenario, includes the following steps: Determining the difference order based on the unit root test d， Determining the autoregressive order based on the Akaike information criterion p and moving average order q ; Determine the Markov transition matrix using the permutation entropy information of the initial scene. Q The Markov transition matrix Q Used to characterize the probability of transitioning from the current segment to the next segment; According to the Markov transition matrix Q Determine the segment category to which the current scene belongs, and call the corresponding ARIMA model to generate a linear trend scene.

5. The method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model according to claim 4, characterized in that, Step S3, which involves establishing a joint probability distribution model of the high-frequency fluctuation subsequence at adjacent time points based on the Copula function, includes the following steps: Select the Copula function type and calculate the Kendall rank correlation coefficient. Spearman rank correlation coefficient To evaluate the goodness of fit of the selected Copula function to the dependent structure of the original wave sequence; definition t The net load fluctuation ratio at time t is determined, and the joint probability density function of the fluctuation ratios at adjacent time t is obtained based on Copula theory. The joint probability density function is solved using Bayes' theorem to obtain the probability sampling model, which is then used to generate the fluctuation scenario.

6. The method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model according to claim 5, characterized in that, In step S4, when the k-means clustering algorithm is used to reduce the initial scene set: Using net load power as a sample attribute and Euclidean distance as a measure of distance between samples, the initial scene set is reduced to a representative scene set by iteratively updating the cluster centers.

7. The method for generating power grid net load fluctuation scenarios based on a combined ARIMA and Copula model according to claim 6, characterized in that, The evaluation index system described in step S5 includes at least two of the following indicators: Time autocorrelation index, average offset index, climbing similarity index, and mean absolute percentage error index.

8. A power grid net load fluctuation scenario generation system based on a combined ARIMA and Copula model, characterized in that, This method is used to implement the grid net load fluctuation scenario generation method as described in any one of claims 1 to 7.

9. A computer device, the computer device comprising a memory, a processor, and a computer program, characterized in that, When the computer program is executed by the processor, it implements the grid net load fluctuation scenario generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the grid net load fluctuation scenario generation method as described in any one of claims 1 to 7.

Images