Method and device for determining peak and valley times of a charging load

Abstract

The embodiment of the application provides a kind of method and device for determining charging load peak and valley time, obtain and determine the charging load of each device at each time;Determine the total charging load and carry out CEEMDAN algorithm decomposition to total charging load, and the IMF component obtained is based on weight again CEEMDAN algorithm decomposition, obtain the denoising of reconstructed total charging load, obtain target total charging load;Target total charging load is operated clustering, and multiple clustering clusters are obtained.In the scheme, the total charging load is reconstructed after multiple decomposition using CEEMDAN algorithm, the noise of total charging load is reduced, the effective information is retained while the signal-to-noise ratio is improved.In addition, the number of clusters is determined by multiple clustering number calculation indexes, which avoids the deviation of clustering result caused by artificial setting of the number of clusters, improves the accuracy and robustness of total charging load classification, improves the accuracy of multiple charging load analysis, can help to analyze the power utilization law of charging equipment user, and guide power grid planning and real-time scheduling.

Description

Technical Field

[0001] This invention belongs to the field of electric vehicle charging technology, specifically relating to a method and apparatus for determining peak and valley charging load times. Background Technology

[0002] The industrial sector is a crucial component of my country's economy, and high-energy-consuming industrial parks are of paramount importance. Carbon reduction in industrial parks requires not only in-depth research on the electricity economics related to industrial restructuring, but also on relevant consumer behavior.

[0003] With the continuous improvement of charging and battery swapping infrastructure in industrial parks, the number of electric vehicles in these parks is experiencing new growth, and more electricity users are transforming from simple consumers into hybrid prosumers. Compared with conventional gasoline vehicles, electric vehicles have significant advantages such as energy saving, low carbon emissions, low operating noise, and high efficiency, offering broad prospects for development and application. However, currently, due to the influence of characteristic parameters such as charging time, charging mode, and charging duration, the uncertainty of electric vehicle charging is quite significant. This uncertain charging behavior can lead to a decline in power grid quality and increased charging costs. Studying the charging behavior of electric vehicles is of great significance for promoting energy conservation and carbon reduction on the consumer side.

[0004] Data-driven electric vehicle charging load modeling methods have become a hot topic in the field of electric vehicle charging model research in recent years due to their high degree of autonomy. However, in actual data acquisition, measurement errors, transmission errors, and calculation errors can lead to noisy data in the charging load data, which may affect the accuracy of charging load data analysis.

[0005] Therefore, improving the accuracy of charging load analysis, enabling precise determination of peak and off-peak charging load times for electric vehicles within specific time periods, and thus enhancing the accuracy of analyzing the electricity consumption patterns of electric vehicle users, is a problem that needs to be solved. Summary of the Invention

[0006] This invention provides a method and apparatus for determining the peak and valley times of charging load, which addresses the problem of how to improve the accuracy of charging load analysis in the prior art.

[0007] The first aspect of this invention provides a method for determining the peak and valley times of charging load, characterized in that the method includes:

[0008] The charging voltage and charging current of multiple devices at multiple times within a target time period are obtained, and the charging load of each device at each time time is determined based on the charging voltage and charging current.

[0009] The total charging load is obtained by summing the sums of all the charging loads at the same time.

[0010] The total charging load is decomposed using the CEEMDAN algorithm, and the obtained IMF components are further decomposed using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components; the weights are determined based on the information entropy of the IMF components.

[0011] The reconstructed total charging load is obtained based on the selected components and the residual components, and the reconstructed total charging load is smoothed and filtered to obtain the noise-reduced target total charging load.

[0012] The target total charging load at each time point is clustered according to the difference between the target total charging loads to obtain multiple clusters;

[0013] From the plurality of clusters, obtain the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load;

[0014] Peak electricity consumption times are determined based on the time corresponding to the total target charging load included in the first cluster, and off-peak electricity consumption times are determined based on the time corresponding to the total target charging load included in the second cluster.

[0015] Optionally, before performing clustering operations on the target total charging load at each time point according to the differences between the target total charging loads to obtain multiple clusters, the following steps are included:

[0016] Calculate the sum of squared errors, average profile coefficient, and Davidson-Baudin index of the target total charging load at each time point within the target time period, and determine the first number of clusters based on the sum of squared errors, the second number of clusters based on the average profile coefficient, and the third number of clusters based on the Davidson-Baudin index.

[0017] The target number of clusters for the target total charging load is determined based on the first cluster number, the second cluster number, and the third cluster number.

[0018] Optionally, determining the first cluster number based on the sum of squared errors, determining the second cluster number based on the average silhouette coefficient, and determining the third cluster number based on the Davidson-Bourdin index specifically includes:

[0019] Calculate the sum of squared errors of the target total charging load at each time point within the target time period to obtain an error sum of squared curve; determine the first cluster number based on the x-coordinate of the point with the largest curvature in the error sum of squared curve;

[0020] Calculate the profile coefficient of the target total charging load at each time point within the target time period; obtain the average profile coefficient curve at each time point within the target time period based on the profile coefficient, and determine the second cluster number based on the x-coordinate of the point with the largest average profile coefficient in the average profile coefficient curve;

[0021] Calculate the Davidson-Baudin index of the target total charging load at each time point within the target time period to obtain the Davidson-Baudin index curve; determine the number of the third cluster based on the x-coordinate of the point with the largest Davidson-Baudin index in the Davidson-Baudin index curve.

[0022] Optionally, determining the target number of clusters for the target total charging load based on the first cluster number, the second cluster number, and the third cluster number specifically includes:

[0023] If the first cluster number, the second cluster number, and the third cluster number are exactly the same, then the first cluster number is determined to be the target cluster number;

[0024] If the number of the first cluster, the number of the second cluster, and the number of the third cluster are not completely the same, then the weighted average of the number of the first cluster, the number of the second cluster, and the number of the third cluster is calculated according to the preset weights of the sum of squared errors, the average profile coefficient, and the Davidson-Bourdin index.

[0025] The target number of clusters is determined based on the weighted average.

[0026] Optionally, determining the target cluster number based on the weighted average includes:

[0027] If the weighted average is an integer, then the weighted average is determined to be the target cluster number;

[0028] If the weighted average is not an integer, the weighted average is rounded down to obtain the first average, and the weighted average is rounded up to obtain the second average.

[0029] The target number of clusters is determined based on the Davidson-Baudin index of the first and second average values.

[0030] Optionally, the step of decomposing the total charging load using the CEEMDAN algorithm specifically includes:

[0031] The total charging load is decomposed using the CEEMDAN algorithm to obtain a first number of first selected IMF components and a first residual signal component;

[0032] The weights of the first selected IMF components are calculated using the entropy weight method;

[0033] The first number of first selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target first selected IMF components with weights greater than or equal to a preset first weight threshold and candidate first selected IMF components with weights less than the preset first weight threshold.

[0034] Optionally, the obtained IMF components are further decomposed using the CEEMDAN algorithm based on their weights to obtain multiple selected components and residual components, including:

[0035] The candidate first selected IMF components are decomposed using the CEEMDAN algorithm to obtain a second number of second selected IMF components and a second residual signal component;

[0036] The weights of the second selected IMF components are calculated using the entropy weight method;

[0037] The second number of second selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target second selected IMF components with weights greater than or equal to the preset first weight threshold, and candidate second selected IMF components with weights less than the preset first weight threshold.

[0038] Optionally, the step of reconstructing the total charging load based on the selected component and the residual component includes:

[0039] The reconstructed total charging load is obtained by adding the first selected IMF component of the plurality of targets, the first residual signal component, the second selected IMF component of the plurality of targets, and the second residual signal component.

[0040] A second aspect of the present invention provides an apparatus for determining the peak and valley times of charging load, characterized in that the apparatus comprises:

[0041] The first acquisition module is used to acquire the charging voltage and charging current of multiple devices at multiple times within a target time period, and to determine the charging load of each device at each time based on the charging voltage and charging current.

[0042] The first determining module is used to determine the sum of all the charging loads at the same time to obtain the total charging load;

[0043] The decomposition module is used to decompose the total charging load using the CEEMDAN algorithm, and to decompose the obtained IMF components again using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components; the weights are determined based on the information entropy of the IMF components.

[0044] The noise reduction module is used to obtain the reconstructed total charging load based on the selected components and the residual components, and to perform smooth filtering on the reconstructed total charging load to obtain the noise-reduced target total charging load.

[0045] The clustering module is used to perform clustering operations on the target total charging load at each time point according to the difference between the target total charging loads, and obtain multiple clusters;

[0046] The second acquisition module is used to acquire, from the plurality of clusters, the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load;

[0047] The second determining module is used to determine the peak electricity consumption time based on the time corresponding to the target total charging load included in the first cluster, and to determine the off-peak electricity consumption time based on the time corresponding to the target total charging load included in the second cluster.

[0048] Optionally, the decomposition module is specifically used for:

[0049] The total charging load is decomposed using the CEEMDAN algorithm to obtain a first number of first selected IMF components and a first residual signal component;

[0050] The weights of the first selected IMF components are calculated using the entropy weight method;

[0051] The first number of first selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target first selected IMF components with weights greater than or equal to a preset first weight threshold and candidate first selected IMF components with weights less than the preset first weight threshold.

[0052] Optionally, the decomposition module is further configured to:

[0053] The candidate first selected IMF components are decomposed using the CEEMDAN algorithm to obtain a second number of second selected IMF components and a second residual signal component;

[0054] The weights of the second selected IMF components are calculated using the entropy weight method;

[0055] The second number of second selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target second selected IMF components with weights greater than or equal to the preset first weight threshold, and candidate second selected IMF components with weights less than the preset first weight threshold.

[0056] A third aspect of the present invention provides an electronic device for determining peak and valley times of charging load. The electronic device includes a processor and a memory. The memory stores at least one instruction, at least one program, a code set, or an instruction set. The at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method for determining peak and valley times of charging load as described in any one of the present invention.

[0057] A fourth aspect of the present invention provides a computer-readable storage medium storing at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the method for determining the peak and valley times of charging load as described in any one of the present invention.

[0058] The embodiments of the present invention have at least the following beneficial effects:

[0059] The process involves acquiring the charging voltage and current of multiple devices at multiple moments within a target time period, and determining the charging load of each device at each moment based on the charging voltage and current. The sum of the charging loads at the same moment is then determined to obtain the total charging load. The total charging load is decomposed using the CEEMDAN algorithm, and the obtained IMF components are further decomposed using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components. The weights are determined based on the information entropy of the IMF components. A reconstructed total charging load is obtained based on the selected components and the residual components, and a smoothing filter is applied to the reconstructed total charging load to obtain a denoised target total charging load. The target total charging load at each moment is clustered according to the differences between the target total charging loads to obtain multiple clusters. From these multiple clusters, a first cluster with the largest target total charging load and a second cluster with the smallest target total charging load are selected. Peak electricity consumption times are determined based on the times corresponding to the target total charging loads included in the first cluster, and off-peak electricity consumption times are determined based on the times corresponding to the target total charging loads included in the second cluster. In this scheme, the total charging load is decomposed and reconstructed multiple times using the CEEMDAN algorithm to reduce noise in the total charging load, preserving effective information while improving the signal-to-noise ratio. Furthermore, the number of clusters is determined comprehensively through multiple clustering indexes, avoiding excessive deviation in clustering results caused by manually setting the clustering number. This improves the accuracy and robustness of the total charging load classification, enhances the accuracy of analysis of multiple charging loads, and helps analyze the electricity consumption patterns of charging equipment users, guiding power grid planning and real-time dispatch. Attached Figure Description

[0060] 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.

[0061] Figure 1 A flowchart illustrating a method for determining peak and valley charging load times according to an embodiment of the present invention;

[0062] Figure 2 A structural block diagram of a device for determining peak and valley times of charging load provided in an embodiment of the present invention;

[0063] Figure 3 A comparison chart of the original value and the charging power data after noise reduction is provided for an embodiment of the present invention;

[0064] Figure 4 An SSE variation curve of charging load is provided for an embodiment of the present invention;

[0065] Figure 5 An average profile coefficient variation curve of charging load is provided for an embodiment of the present invention;

[0066] Figure 6 A DBI variation curve of charging load is provided for an embodiment of the present invention;

[0067] Figure 7 A peak-valley cluster center variation curve of charging load is provided in an embodiment of the present invention;

[0068] Figure 8 This is a diagram showing the distribution relationship between peak and valley charging load and charging start time, provided as an embodiment of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Figure 1 A flowchart of a method for determining peak and valley charging load times provided in an embodiment of the present invention is shown. The method includes the following steps:

[0071] Step 101: Obtain the charging voltage and charging current of multiple devices at multiple times within the target time period, and determine the charging load of each device at each time based on the charging voltage and charging current.

[0072] Specifically, the charging load is calculated by multiplying the charging voltage and charging current. A given period is divided into multiple corresponding time periods, with multiple charging devices charging at different times. Charging power sampling data for each charging device at each time period is collected, including charging time, battery SOC, charging voltage, charging current, and information such as the identification marker gun number and order number. The data collection frequency is 30 seconds per sampling. Effective charging power, or effective charging load, is calculated by multiplying the charging voltage and charging current for each charging device at each time period. In this embodiment of the invention, the charging device is an electric vehicle, the target time period is one day, and the data collection area is a consumer-side enterprise park.

[0073] Step 102: Determine the sum of all the charging loads at the same time to obtain the total charging load.

[0074] Specifically, the total charging load is obtained by superimposing the charging loads of each charging device collected at the same time. Similarly, the total charging load at multiple times within the target time period is obtained.

[0075] Step 103: Decompose the total charging load using the CEEMDAN algorithm, and then decompose the obtained IMF components again using the CEEMDAN algorithm based on the weights to obtain multiple selected components and residual components; the weights are determined based on the information entropy of the IMF components.

[0076] Specifically, the CEEMDAN (Adaptive Noise Complete Ensemble Empirical Mode Decomposition) algorithm is a signal decomposition algorithm that can decompose a signal into multiple components of different frequencies. These signal components of different frequencies are represented by IMF components. After decomposition by the CEEMDAN algorithm, the signal yields residual signal components and multiple IMF components. In this embodiment of the invention, the total charging load is used as the signal input to the CEEMDAN algorithm. The total charging load is decomposed twice using the CEEMDAN algorithm, resulting in an improved CEEMDAN algorithm.

[0077] Entropy weighting is an objective method for determining weights. The principle of entropy weighting is that the smaller the information entropy of an indicator, the greater the degree of variation in its value, the greater the amount of information it provides, and the greater its role in the comprehensive evaluation; therefore, its weight should also be greater. Conversely, the larger the information entropy of an indicator, the smaller the degree of variation in its value, the smaller the amount of information it provides, and the smaller its role in the comprehensive evaluation; therefore, its weight should also be smaller. The principle of entropy weighting is exactly the opposite of the effect of entropy. Therefore, the weight of each indicator can be calculated using information entropy. An improved CEEMDAN algorithm is used to decompose the total charging load. First, the CEEMDAN algorithm is used to perform a first decomposition of the total charging load at each time point, resulting in multiple IMF components and residual signal components. The weights of the multiple IMF components obtained from the first decomposition are calculated using the entropy weighting method. Then, the IMF components with smaller weights are subjected to a second decomposition, again obtaining residual signals and multiple IMF components, and the weights of the IMF components obtained from the second decomposition are calculated for each.

[0078] In this embodiment of the invention, the larger the information entropy of a certain IMF component, the more unstable the component signal is, the greater the fluctuation, and the more noise it contains. Therefore, this component occupies a smaller weight.

[0079] Step 104: Based on the selected components and the residual components, obtain the reconstructed total charging load, and perform smoothing filtering on the reconstructed total charging load to obtain the noise-reduced target total charging load.

[0080] Specifically, charging load is essentially a nonlinear, non-stationary time series data. During actual charging load data acquisition, measurement errors, transmission errors, and calculation errors can lead to noisy data within the charging load data. Under various noise interferences, it is difficult to accurately and effectively extract the charging load signal characteristics of the charging equipment, which is detrimental to the prediction and scheduling of charging behavior. Therefore, it is necessary to perform noise reduction processing on the original signal to improve the signal-to-noise ratio of the charging load signal.

[0081] After the total charging load is decomposed twice using the CEEMDAN algorithm, the signal components obtained are the selected IMF components with a large proportion of effective signal components and the residual signal components. The reconstructed total charging load is the target total charging load.

[0082] The Savitzky-Golay (SG) filtering algorithm is a digital signal processing algorithm used for signal smoothing. In this embodiment of the invention, after obtaining the target total charging load, the various components constituting the target total charging load are fitted to achieve smoothing. The SG filtering algorithm uses the least squares method to fit local data segments, and then uses the fitted function to estimate the value of each data point, thereby achieving smoothing. The advantages of the SG filtering algorithm are that it can simultaneously achieve smoothing and noise reduction, effectively filter out high-frequency noise, and has good adaptability to nonlinear signals. In addition, the algorithm has a fast calculation speed, does not require frequency domain transformation, and is suitable for real-time signal processing.

[0083] In this embodiment of the invention, the improved CEEMDAN algorithm determines the weights of multiple attributes by calculating index entropy weights for denoising and optimization, thus improving the signal-to-noise ratio of the charging load signal compared to the traditional CEEMDAN algorithm. Furthermore, the denoised and optimized signal components are then smoothed using SG filtering to obtain clean signal components and reconstructed, further improving the signal-to-noise ratio of the charging load signal.

[0084] Step 105: Perform clustering operations on the target total charging load at each time point according to the difference between the target total charging loads to obtain multiple clusters.

[0085] Specifically, after obtaining the target total charging load at each time point, the charging load varies in different charging processes due to differences in the type of charging equipment, charging time, etc., resulting in significant differences in the order of magnitude between load data. To improve the efficiency of algorithm analysis, the data after noise reduction using the improved CEEMDAN algorithm is standardized, i.e., normalized. The processing formula is as follows:

[0086]

[0087] In formula (1), p i Let i be the i-th number after standardizing the charging load.

[0088] After normalization, the target total charging load at each time step is calculated using K-means. Based on the set number of clusters, the target total charging load at one time step in each cluster is randomly selected as the initial cluster center p′. i Calculate the minimum difference L(p) between the target total charging load and the initial cluster center at other times. i The formula for calculating the minimum difference is as follows:

[0089]

[0090] p jLet n be the total target charging load excluding the initial cluster center, and n be the number of target total charging loads. Similarly, calculate the minimum difference when other points become the initial cluster centers.

[0091] Calculate the probability that the target total charging load at each time step could become a cluster center. Select the target total charging load at the time with the highest probability as the cluster center, and determine all cluster centers based on the optimal number of clusters. The probability p(p0) of the target total charging load at each time step could become a cluster center is given. i The calculation formula is as follows:

[0092]

[0093] In formula (3), P is the sample set of the target total charging load.

[0094] After determining the cluster centers, the target total charging loads within the target time period are clustered based on the similarity between the various charging loads, minimizing the intra-cluster differences and maximizing the inter-cluster differences. Each cluster is a group of multiple target total charging loads of similar size.

[0095] Step 106: Obtain the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load from the plurality of clusters.

[0096] Specifically, among multiple clusters, the total target charging load within each cluster is similar, and the average value of the total target charging load among the clusters has a certain difference. The first cluster with the largest average value and the second cluster with the smallest average value are determined.

[0097] Step 107: Determine the peak electricity consumption time based on the time corresponding to the target total charging load included in the first cluster, and determine the off-peak electricity consumption time based on the time corresponding to the target total charging load included in the second cluster.

[0098] Specifically, based on the target total charging load included in the first cluster, the corresponding time points are determined and acquired. These time points are then identified as the peak times of the charging load within the target time period, i.e., peak electricity consumption times. Based on the target total charging load included in the second cluster, the corresponding time points are determined and acquired. These time points are then identified as the trough times of the charging load within the target time period, i.e., off-peak electricity consumption times.

[0099] Based on the determined peak and valley times of electricity consumption, it has significant application value in the practical application of electric vehicle charging load models. It can help analyze the electricity consumption patterns of charging equipment users, guide power grid planning and real-time dispatch, and has good development prospects.

[0100] In summary, the present invention provides a method for determining peak and valley charging load times. This method acquires the charging voltage and charging current of multiple devices at multiple times within a target time period, and determines the charging load of each device at each time based on the charging voltage and charging current. It then determines the sum of the charging loads at the same time to obtain the total charging load. The total charging load is decomposed using the CEEMDAN algorithm, and the obtained IMF components are further decomposed using the CEEMDAN algorithm based on weights to obtain multiple refined components and residual components. The weights are determined based on the information entropy of the IMF components. Based on the refined components… The reconstructed total charging load is obtained by selecting components and the residual components, and then smoothed and filtered to obtain the denoised target total charging load. The target total charging load at each time point is clustered according to the differences between the target total charging loads, resulting in multiple clusters. From these multiple clusters, the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load are selected. Peak electricity consumption times are determined based on the time corresponding to the target total charging load included in the first cluster, and off-peak electricity consumption times are determined based on the time corresponding to the target total charging load included in the second cluster. In this scheme, the total charging load is decomposed and reconstructed multiple times using the CEEMDAN algorithm, achieving denoising of the total charging load while preserving effective information and improving the signal-to-noise ratio. Furthermore, the number of clusters is determined comprehensively through multiple clustering index calculations, avoiding excessive deviation in clustering results caused by manually setting the clustering number. This improves the accuracy and robustness of total charging load classification and enhances the accuracy of analysis of multiple charging loads. It can help analyze the electricity consumption patterns of charging equipment users and guide power grid planning and real-time scheduling.

[0101] In one possible implementation, before performing clustering operations on the target total charging load at each time point according to the differences between the target total charging loads to obtain multiple clusters, the process includes:

[0102] Calculate the sum of squared errors, average profile coefficient, and Davidson-Baudin index of the target total charging load at each time point within the target time period, and determine the first number of clusters based on the sum of squared errors, the second number of clusters based on the average profile coefficient, and the third number of clusters based on the Davidson-Baudin index.

[0103] The target number of clusters for the target total charging load is determined based on the first cluster number, the second cluster number, and the third cluster number.

[0104] Specifically, clustering is a method for dividing data samples. In most cases, the categories of data samples are unknown. Clustering is performed based on the ideas of intra-cluster cohesion and low inter-cluster coupling. In terms of algorithms, the clustering effect is evaluated based on clustering effectiveness evaluation metrics. Clustering effectiveness evaluation metrics include Sum of Squared Error (SSE), mean silhouette coefficient, and Davies-Bouldin Index (DBI), but each has its own advantages and disadvantages.

[0105] For the sum of squared errors (SSE), the SSE represents the Euclidean distance from each point within the same cluster to the cluster center of that cluster. The SSE is then summed to obtain the total SSE. A smaller SSE indicates that the points within the same cluster are more densely packed, indicating a higher degree of clustering. A larger SSE indicates a larger distance from each data point to its respective cluster center, resulting in poorer clustering. In this embodiment, the number of clusters determined by the SSE is the first cluster number. Regarding the mean silhouette coefficient (DBI), the DBI ranges from -1 to 1. The DBI is the average of the silhouette coefficients of all data points. A DBI closer to 1 indicates a more reasonable clustering result, while a DBI closer to -1 indicates a worse clustering result. In this embodiment, the number of clusters determined by the DBI is the second cluster number. A smaller DBI indicates a smaller distance between points within the same cluster and a larger distance between different clusters, indicating a more reasonable clustering result. A larger DBI indicates a worse clustering result. In this embodiment, the number of clusters determined by the DBI is the third cluster number.

[0106] In this embodiment of the invention, the number of clusters obtained by combining three indicators is used to determine the target number of clusters, which improves the accuracy of determining the number of clusters.

[0107] In one possible implementation, determining the first cluster number based on the sum of squared errors, determining the second cluster number based on the average silhouette coefficient, and determining the third cluster number based on the Davidson-Bourdin index specifically includes:

[0108] Calculate the sum of squared errors of the target total charging load at each time point within the target time period to obtain the sum of squared error curve; determine the first cluster number based on the x-coordinate of the point with the largest curvature in the sum of squared error curve.

[0109] Specifically, the sum of squared errors (SSE) is calculated as follows:

[0110]

[0111] In equation (4), SSE is the sum of squared clustering errors for all sample points, k represents the number of clusters, and M... i Let p be the sample set of class i, p be the points in class i, and m be the sample set of class i. iLet be the cluster center point in the i-th cluster. Calculate the SSE (Sum of Squared Errors) of the sample set for different numbers of clusters. Obtain the sum of squared errors curve based on the SSE of the samples corresponding to different numbers of clusters. Determine the x-coordinate value corresponding to the point with the largest curvature in the curve as the first cluster number. The data in the sample set represents the target total charging load at each time point within the target time period.

[0112] Calculate the profile coefficient of the target total charging load at each time point within the target time period; obtain the average profile coefficient curve at each time point within the target time period based on the profile coefficient, and determine the second cluster number based on the x-coordinate of the point with the largest average profile coefficient in the average profile coefficient curve.

[0113] Specifically, the contour coefficient is calculated as follows:

[0114]

[0115] In equation (5), s i Meanis is the contour coefficient for a certain point. out Meanis is the average distance between this point and other points in other categories. in This represents the average distance between the point and other points in its category. For different numbers of clusters, the average silhouette coefficient is obtained by summing and averaging the silhouette coefficients of all data points. An average silhouette coefficient curve is then generated based on these average silhouette coefficients for different numbers of clusters. The x-axis value corresponding to the point with the largest average silhouette coefficient is determined as the second cluster number.

[0116] Calculate the Davidson-Baudin index of the target total charging load at each time point within the target time period to obtain the Davidson-Baudin index curve; determine the number of the third cluster based on the x-coordinate of the point with the largest Davidson-Baudin index in the Davidson-Baudin index curve.

[0117] Specifically, the DBI calculation formula is as follows:

[0118]

[0119] In equation (6), d(i,j) is the Euclidean distance between the cluster centers of samples in class i and class j, k represents the number of clusters, and Meanis in,i Meanis is the average distance between samples in class i and the cluster center. in,j denoted as the average distance between the sample in class j and the cluster center.

[0120] Under different numbers of clusters, the average profile coefficients corresponding to different numbers of clusters are obtained. Based on the DBI index corresponding to different numbers of clusters, the DBI curve is obtained. The value of the horizontal axis corresponding to the point with the smallest DBI value in the DBI curve is determined as the third cluster number.

[0121] In one possible implementation, determining the target cluster number of the target total charging load based on the first cluster number, the second cluster number, and the third cluster number specifically includes:

[0122] If the first cluster number, the second cluster number, and the third cluster number are exactly the same, then the first cluster number is determined to be the target cluster number;

[0123] If the number of the first cluster, the number of the second cluster, and the number of the third cluster are not completely the same, then the weighted average of the number of the first cluster, the number of the second cluster, and the number of the third cluster is calculated according to the preset weights of the sum of squared errors, the average profile coefficient, and the Davidson-Bourdin index.

[0124] The target number of clusters is determined based on the weighted average.

[0125] Specifically, weights are assigned to these three indicators: SSE has a weight of 0.3, the average silhouette coefficient has a weight of 0.4, and DBI has a weight of 0.3. After obtaining the first, second, and third cluster numbers based on these three indicators, if all three cluster numbers are identical, this identical number is taken as the target cluster number. If they are not identical—that is, any one of the three cluster numbers is different from the other two, or all three cluster numbers are different—a weighted average of the first, second, and third cluster numbers is calculated based on the weights corresponding to the different indicators, and the target cluster number is determined based on this weighted average.

[0126] In one possible implementation, determining the target cluster number based on the weighted average includes:

[0127] If the weighted average is an integer, then the weighted average is determined to be the target cluster number;

[0128] If the weighted average is not an integer, the weighted average is rounded down to obtain the first average, and the weighted average is rounded up to obtain the second average.

[0129] The target number of clusters is determined based on the Davidson-Baudin index of the first and second average values.

[0130] Specifically, if the first, second, and third cluster numbers obtained from the three indicators are not completely identical, a weighted average is obtained based on the weights of each indicator. If the weighted average is an integer, then this integer is determined as the target cluster number. If the weighted average is not an integer, then the weighted average is rounded down to obtain the first average and rounded up to obtain the second average, with the second average being greater than the first average.

[0131] Calculate the DBI index corresponding to the first average and the second average respectively. The average with the smaller DBI index has the best clustering effect, and this average is determined as the target number of clusters.

[0132] In one possible implementation, the decomposition of the total charging load using the CEEMDAN algorithm specifically includes:

[0133] The total charging load is decomposed using the CEEMDAN algorithm to obtain a first number of first selected IMF components and a first residual signal component;

[0134] The weights of the first selected IMF components are calculated using the entropy weight method;

[0135] The first number of first selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target first selected IMF components with weights greater than or equal to a preset first weight threshold and candidate first selected IMF components with weights less than the preset first weight threshold.

[0136] Specifically, the total charging load at each moment within the target time period is decomposed using the CEEMDAN algorithm.

[0137] The total charging load is used as the signal input for the CEEMDAN algorithm decomposition. After the first decomposition by the CEEMDAN algorithm, a first number of first selected IMF components and a first residual signal component are obtained. In this embodiment of the invention, the first decomposition yields a first selected IMF components. i , and a residual signal component R1(n).

[0138] The entropy weighting method is applied to assign weights to IMF components, specifically involving the calculation of the information entropy and weights of the IMF components. Information entropy describes the uncertainty of the IMF component; the greater the uncertainty of the variable, the greater the information entropy, which also means that the IMF component contains more implicit information. A larger entropy value indicates a greater degree of disorder in the IMF signal compared to the original signal, and therefore a smaller weight. The information entropy is calculated as follows:

[0139]

[0140]

[0141] In the above formula That is, IMF n Information entropy of components, IMF n Let be the nth IMF component, and a be the number of IMF components obtained from the first decomposition.

[0142] After the IMF was determined n After calculating the information entropy of the components, the IMF is then calculated. n The weight η of the componentn for:

[0143]

[0144] Using formula (9), all IMFs from the first decomposition are calculated. i After weighting the components, all IMFs i The components are arranged in descending order of weight, and all IMFs are determined. i Among the components, the IMFs with weights less than a preset first weight threshold i The component is the first selected candidate IMF component, and the IMF with a weight greater than or equal to the preset first weight threshold. i The components are the first selected IMF components. A preset first weight threshold is used to measure the amount of information reflected in the IMF components; IMF components with weights less than the first weight threshold have less information. Among the *a* first selected IMF components obtained from the first decomposition, the IMF components with weights greater than or equal to the preset first weight threshold... i There are 1 components.

[0145] In one possible implementation, the obtained IMF components are further decomposed using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components, including:

[0146] The candidate first selected IMF components are decomposed using the CEEMDAN algorithm to obtain a second number of second selected IMF components and a second residual signal component;

[0147] The weights of the second selected IMF components are calculated using the entropy weight method;

[0148] The second number of second selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target second selected IMF components with weights greater than or equal to the preset first weight threshold, and candidate second selected IMF components with weights less than the preset first weight threshold.

[0149] Specifically, IMFs with weights less than a preset first weight threshold i The candidate first-selected IMF components contain both valid information and noisy data, with a large proportion of noise. Therefore, each candidate first-selected IMF component is decomposed again using the CEEMDAN algorithm to obtain the second residual signal and the second number of second-selected IMF components. The second residual signal component is denoted as R. 2g (n), the second selected component is denoted as IMF. jm The second quantity is g.

[0150] For each candidate first-selected IMF component, the entropy weight method is used to calculate the weight of each second-selected IMF component obtained by decomposing the first-selected IMF component. The second-selected IMF component with a weight greater than or equal to a preset first threshold is the target second-selected IMF component, denoted as IIMF. jm The second selected IMF component with a weight less than the preset first threshold is the candidate second selected IMF component.

[0151] In one possible implementation, the process of reconstructing the total charging load based on the selected component and the residual component includes:

[0152] The reconstructed total charging load is obtained by adding the first selected IMF component of the plurality of targets, the first residual signal component, the second selected IMF component of the plurality of targets, and the second residual signal component.

[0153] Specifically, the signals corresponding to the IMF components with higher information content obtained after decomposing the total charging load using the CEEMDAN algorithm are added together with the residual signals obtained from each decomposition. The signals corresponding to the second selected IMF components with lower weights obtained from the second decomposition are discarded to obtain the reconstructed total charging load.

[0154] The reconstructed total charging load is the total charging load after noise reduction. The reconstructed total charging load P′(n) is expressed as:

[0155]

[0156] In equation (10), a represents the number of the first selected IMF components obtained from the first decomposition, and among the a first selected IMF components, the IMFs with weights greater than or equal to a preset first weight threshold are... i The first l components are considered. g represents the number of second-selected IMF components obtained from the decomposition of each candidate first-selected IMF component. The second-selected IMF components are sorted in descending order of weight, where m is the number of IMF components in the second-selected IMF components with a weight greater than or equal to a preset first threshold. jm The second selected IMF component is the IMF component whose weight is greater than or equal to the preset first threshold.

[0157] Figure 2 This is a structural block diagram of a device for determining peak and valley times of charging load according to an embodiment of the present invention. The device 200 includes:

[0158] The first acquisition module 201 is used to acquire the charging voltage and charging current of multiple devices at multiple times within a target time period, and to determine the charging load of each device at each time based on the charging voltage and charging current.

[0159] The first determining module 202 is used to determine the sum of all the charging loads at the same time to obtain the total charging load.

[0160] The decomposition module 203 is used to decompose the total charging load using the CEEMDAN algorithm, and to decompose the obtained IMF components again using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components; the weights are determined based on the information entropy of the IMF components.

[0161] The noise reduction module 204 is used to obtain the reconstructed total charging load based on the selected components and the residual components, and to perform smooth filtering on the reconstructed total charging load to obtain the noise-reduced target total charging load.

[0162] Clustering module 205 is used to perform clustering operations on the target total charging load at each time point according to the difference between the target total charging loads, and obtain multiple clusters.

[0163] The second acquisition module 206 is used to acquire the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load from the plurality of clusters.

[0164] The second determining module 207 is used to determine the peak electricity consumption time based on the time corresponding to the target total charging load included in the first cluster, and to determine the off-peak electricity consumption time based on the time corresponding to the target total charging load included in the second cluster.

[0165] In another embodiment of the present invention, an electronic device for determining peak and valley charging load times is also provided. The electronic device includes a processor and a memory. The memory stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method for determining peak and valley charging load times as described in any one of the present invention.

[0166] In another embodiment of the present invention, a computer-readable storage medium is also provided, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement the method for determining the peak and valley times of charging load as described in any one of the present invention.

[0167] The following examples further illustrate some of the beneficial effects of the embodiments of the present invention:

[0168] The charging load for each effective charging process is calculated from the charging voltage and charging current. Finally, the charging loads of different electric vehicles at the same time are summed to obtain the total charging load. To illustrate the specific implementation process of the method described in this paper, the charging load of one day will be used for relevant calculations and analysis.

[0169] Charging data from 11:59 AM to 12:02 PM on a certain morning were selected and noise was reduced. The original values ​​were then compared with the noise-reduced data.

[0170] Figure 3 This invention provides a comparison chart of the original and noise-reduced charging power data. (See the image below.) Figure 3 As shown, after the improved CEEMDAN algorithm is used to denoise the original values, the denoised load data is relatively smoother than the original values ​​and can maintain the basic trend of the original curve. It can effectively eliminate noise while ensuring that the signal shape remains unchanged, which is beneficial for subsequent cluster analysis calculations.

[0171] Table 1 shows the comparison data of charging load noise reduction before and after processing with the improved CEEMDAN algorithm:

[0172] Original signal SNR SNR of the noise-reduced signal 8.3743dB 15.8247dB

[0173] Table 1

[0174] To further illustrate the denoising effect of the improved algorithm, the signal-to-noise ratio (SNR) of the original signal and the denoised signal was calculated, as shown in Table 1. The improved CEEMDAN algorithm can significantly improve the signal-to-noise ratio, increasing it from 8.3743 dB to 15.8247 dB.

[0175] The noisy charging load signal was denoised using EEMD (Ensemble Empirical Mode Decomposition), CEEMDAN, and the improved CEEMDAN method proposed in this invention, respectively. The denoised signals were compared and analyzed based on SNR and RMSE (Root Mean Square Error). The denoising performance of each method is shown in Table 2.

[0176] method SNR / dB RMSE EEMD 14.3829 0.1443 CEEMDAN 12.2365 0.1656 Improved CEEMDAN 15.8247 0.0936

[0177] Table 2

[0178] The comparison of noise reduction metrics in Table 2 shows that the signal-to-noise ratios of the EEMD and CEEMDAN methods are low, indicating that these two methods have poor noise reduction performance. Figure 3It can be seen that the data change trend after denoising based on the improved CEEMDAN algorithm is almost consistent with the original signal, and the signal-to-noise ratio is higher than the other two methods, with the lowest root mean square error. The improved CEEMDAN algorithm proposed in this paper not only preserves the effective information in the high-frequency IMF, but also retains the true components of the low-frequency IMF, and its denoising effect is better than other methods.

[0179] Secondly, consider electric vehicle charging load cluster analysis:

[0180] The daily peak and valley values ​​of charging load and their corresponding times from March 30, 2021 to April 15, 2021 were determined and analyzed. Before selecting the initial cluster centers using K-means++, this embodiment of the invention first calculated three indicators for evaluating the clustering effect: SSE, mean silhouette coefficient, and DBI. The number of clusters was automatically selected by combining the three indicators to avoid the influence of manually setting the clustering parameter k. Figure 4 A charging load SSE variation curve is provided for an embodiment of the present invention. Figure 5 This invention provides an average profile coefficient variation curve of charging load. Figure 6 This is a DBI variation curve of charging load provided in an embodiment of the present invention.

[0181] Figure 4 , Figure 5 and Figure 6 The horizontal axis represents the number of cluster centers, i.e., the number of clusters. From... Figure 4 By finding the inflection point of the SSE curve in the SSE index curve graph, we can see that the curvature is maximized when k=3, so the optimal number of clusters can be 3. Figure 5 The optimal cluster number obtained from the average silhouette coefficient curve is 2, which is derived from... Figure 6 The DBI index curve shows that the DBI is minimized when the number of clusters k = 3. Combining the SEE, average profile coefficient, DBI, and the weights of each index, the weighted average number of clusters is calculated to be 2.7. Substituting 2 and 3 into the DBI calculation formula, we find that the DBI is minimized when the number of clusters is 3. Therefore, the optimal number of clusters is chosen as 3. The calculation of the number of clusters for other days is similar.

[0182] When the optimal number of clusters k=3, the improved K-Means++ clustering algorithm, the weighted K-Means++ algorithm, the K-Means++ algorithm, and the DBI index of K-Means are compared, as shown in Table 3.

[0183] algorithm DBI Improved K-Means++ 1.205 Weighted K-Means++ 1.263 K-Means++ 1.308 K-Means 1.537

[0184] Table 3

[0185] As shown in Table 3, the improved K-Means++ clustering algorithm has the lowest DBI index and the best clustering effect.

[0186] After determining the number of clusters, the initial cluster centers were optimized. Finally, the peak and valley cluster centers of the daily charging load from March 30, 2021 to April 15, 2021 were calculated using the K-means method. Figure 7 This is a peak-valley cluster center variation curve of charging load provided in an embodiment of the present invention.

[0187] Depend on Figure 7 It can be seen that the peak load center values ​​were relatively high on the 10th and 14th, but the number of charging times did not increase significantly compared with other days, and was even lower. This is due to the relatively concentrated charging time of electric vehicles; the more concentrated the charging time, the greater the charging power during peak load.

[0188] Figure 8 This is a diagram showing the distribution relationship between peak and valley charging load and charging start time, provided as an embodiment of the present invention.

[0189] like Figure 8 As shown in the figure, the larger the value on the vertical axis, the more frequently it occurs at the corresponding time. Figure 8 The image shows the frequency distribution of peak and off-peak load times at different times during the period from March 30, 2021 to April 15, 2021, as well as the frequency distribution of charging start times for all charging devices at different times. Frequency is also referred to as ratio. Figure 8 As can be seen, the trends of the ratio of peak charging load to charging start time are basically consistent, indicating that peak charging load is greatly affected by the charging start time. In contrast, the changes in off-peak charging load are relatively stable and more dispersed compared to peak charging load. Within a certain period, the peak charging load of charging stations exhibits certain regularity, occurring primarily in three time periods: early morning, afternoon, and evening. Peak charging load is directly proportional to the frequency of charging start time occurrences, while off-peak charging load is not significantly related to the frequency of charging start time occurrences.

[0190] This invention is based on the electric vehicle charging load in consumer-side enterprise parks. It uses entropy weighting and SG smoothing filtering to denoise and reconstruct the charging load signal, and uses comprehensive indicators to determine the number of clusters to perform cluster analysis on the charging load. This improves the accuracy of charging load classification and analysis, which is beneficial for subsequent analysis of electric vehicle charging behavior. It is of great significance for enabling high-carbon emission enterprises to carry out research and increase the proportion of new energy power in end-use energy, and has good development prospects.

[0191] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0192] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for determining the peak and valley times of charging load, characterized in that, The method includes: The charging voltage and charging current of multiple devices at multiple times within a target time period are obtained, and the charging load of each device at each time time is determined based on the charging voltage and charging current. The total charging load is obtained by summing the sums of all the charging loads at the same time. The total charging load is decomposed using the CEEMDAN algorithm, and the obtained IMF components are further decomposed using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components; the weights are determined based on the information entropy of the IMF components. The reconstructed total charging load is obtained based on the selected components and the residual components, and the reconstructed total charging load is smoothed and filtered to obtain the noise-reduced target total charging load. The target total charging load at each time point is clustered according to the difference between the target total charging loads to obtain multiple clusters; From the plurality of clusters, obtain the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load; Peak electricity consumption times are determined based on the time corresponding to the total target charging load included in the first cluster, and off-peak electricity consumption times are determined based on the time corresponding to the total target charging load included in the second cluster. The decomposition of the total charging load using the CEEMDAN algorithm specifically includes: The total charging load is decomposed using the CEEMDAN algorithm to obtain a first number of first selected IMF components and a first residual signal component; The weights of the first selected IMF components are calculated using the entropy weight method; The first number of first selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target first selected IMF components with weights greater than or equal to a preset first weight threshold and candidate first selected IMF components with weights less than the preset first weight threshold. The obtained IMF components are further decomposed using the CEEMDAN algorithm based on their weights to obtain multiple selected components and residual components, including: The candidate first selected IMF components are decomposed using the CEEMDAN algorithm to obtain a second number of second selected IMF components and a second residual signal component; The weights of the second selected IMF components are calculated using the entropy weight method; The second number of second selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target second selected IMF components with weights greater than or equal to the preset first weight threshold, and candidate second selected IMF components with weights less than the preset first weight threshold.

2. The method according to claim 1, characterized in that, Before performing clustering operations on the target total charging load at each time point according to the differences between the target total charging loads to obtain multiple clusters, the following steps are included: Calculate the sum of squared errors, average profile coefficient, and Davidson-Baudin index of the target total charging load at each time point within the target time period, and determine the first number of clusters based on the sum of squared errors, the second number of clusters based on the average profile coefficient, and the third number of clusters based on the Davidson-Baudin index. The target number of clusters for the target total charging load is determined based on the first cluster number, the second cluster number, and the third cluster number.

3. The method according to claim 2, characterized in that, The first cluster number is determined based on the sum of squared errors, the second cluster number is determined based on the average silhouette coefficient, and the Davidson coefficient is used to determine the cluster number. The Burding index determines the number of the third cluster, specifically including: Calculate the sum of squared errors of the target total charging load at each time point within the target time period to obtain an error sum of squared curve; determine the first cluster number based on the x-coordinate of the point with the largest curvature in the error sum of squared curve; Calculate the profile coefficient of the target total charging load at each time point within the target time period; obtain the average profile coefficient curve at each time point within the target time period based on the profile coefficient, and determine the second cluster number based on the x-coordinate of the point with the largest average profile coefficient in the average profile coefficient curve; Calculate the Davidson-Baudin index of the target total charging load at each time point within the target time period to obtain the Davidson-Baudin index curve; determine the number of the third cluster based on the x-coordinate of the point with the largest Davidson-Baudin index in the Davidson-Baudin index curve.

4. The method according to claim 2, characterized in that, The step of determining the target cluster number for the target total charging load based on the first cluster number, the second cluster number, and the third cluster number specifically includes: If the first cluster number, the second cluster number, and the third cluster number are exactly the same, then the first cluster number is determined to be the target cluster number; If the first cluster number, the second cluster number, and the third cluster number are not completely identical, then based on the sum of squared errors, the average profile coefficient, and the Davidson coefficient... The preset weights of the Burding index are used to calculate the weighted average of the number of the first cluster, the number of the second cluster, and the number of the third cluster; The target number of clusters is determined based on the weighted average.

5. The method according to claim 4, characterized in that, Determining the target cluster number based on the weighted average includes: If the weighted average is an integer, then the weighted average is determined to be the target cluster number; If the weighted average is not an integer, the weighted average is rounded down to obtain the first average, and the weighted average is rounded up to obtain the second average. Davidson based on the first average and the second average The Burding index determines the target number of clusters.

6. The method according to claim 1, characterized in that, The process of reconstructing the total charging load based on the selected components and the residual components includes: The reconstructed total charging load is obtained by adding the first selected IMF component of the plurality of targets, the first residual signal component, the second selected IMF component of the plurality of targets, and the second residual signal component.

7. A device for determining the peak and valley times of charging load, characterized in that, The device includes: The first acquisition module is used to acquire the charging voltage and charging current of multiple devices at multiple times within a target time period, and to determine the charging load of each device at each time based on the charging voltage and charging current. The first determining module is used to determine the sum of all the charging loads at the same time to obtain the total charging load; The decomposition module is used to decompose the total charging load using the CEEMDAN algorithm, and to decompose the obtained IMF components again using the CEEMDAN algorithm based on weights to obtain multiple selected components and residual components; the weights are determined based on the information entropy of the IMF components. The noise reduction module is used to obtain the reconstructed total charging load based on the selected components and the residual components, and to perform smooth filtering on the reconstructed total charging load to obtain the noise-reduced target total charging load. The clustering module is used to perform clustering operations on the target total charging load at each time point according to the difference between the target total charging loads, and obtain multiple clusters; The second acquisition module is used to acquire, from the plurality of clusters, the first cluster with the largest target total charging load and the second cluster with the smallest target total charging load; The second determining module is used to determine the peak electricity consumption time based on the time corresponding to the target total charging load included in the first cluster, and to determine the off-peak electricity consumption time based on the time corresponding to the target total charging load included in the second cluster. When performing CEEMDAN algorithm decomposition on the total charging load, the decomposition module is configured to execute: The total charging load is decomposed using the CEEMDAN algorithm to obtain a first number of first selected IMF components and a first residual signal component; The weights of the first selected IMF components are calculated using the entropy weight method; The first number of first selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target first selected IMF components with weights greater than or equal to a preset first weight threshold and candidate first selected IMF components with weights less than the preset first weight threshold. When performing a second CEEMDAN algorithm decomposition on the obtained IMF components based on weights to obtain multiple selected components and residual components, the decomposition module is configured to perform: The candidate first selected IMF components are decomposed using the CEEMDAN algorithm to obtain a second number of second selected IMF components and a second residual signal component; The weights of the second selected IMF components are calculated using the entropy weight method; The second number of second selected IMF components are sorted from largest to smallest according to their weights to obtain multiple target second selected IMF components with weights greater than or equal to the preset first weight threshold, and candidate second selected IMF components with weights less than the preset first weight threshold.

8. An electronic device, characterized in that, The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method for determining the peak and valley times of charging load as described in any one of claims 1-6.

9. A computer-readable storage medium storing at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the method for determining peak and valley times of charging load as described in any one of claims 1-6.

Images