A Data-Driven Photovoltaic Power Generation Estimation Method and System Based on Betty Curves

Abstract

This invention discloses a data-driven photovoltaic power generation estimation method and system based on Betty curves, comprising: collecting data on a first object, performing a first reconstruction calculation to obtain a first feature of the first object; performing statistical calculations based on the first object data to obtain a second feature of the first object; performing a first transformation task based on specific information of the first object data to obtain the result of the first transformation task; concatenating the first feature, the second feature, and the result of the first transformation task to form a feature space, inputting it into a first prediction model, and outputting the prediction result of the first object. This invention captures the complex dynamic patterns of time series, achieves diversification of the feature space, and improves the generalization ability of the prediction model. By combining multi-dimensional features and employing efficient machine learning algorithms, it achieves high-precision estimation of photovoltaic power generation while improving the model's adaptability to nonlinear and multi-scale characteristics.

Description

Technical Field

[0001] This invention relates to the field of new energy technology, specifically to a data-driven photovoltaic power generation estimation method and system based on the Betty curve. Background Technology

[0002] In recent years, with the continuous growth of global demand for clean energy, photovoltaic (PV) power generation, as a green and efficient energy form, has been widely used. However, due to its dependence on weather conditions and environmental factors, PV power output exhibits significant randomness and volatility. Traditional PV power generation estimation methods mainly rely on physical models and empirical formulas. While these methods can provide preliminary estimation results under certain conditions, they often suffer from large estimation errors and insufficient real-time performance when faced with complex and variable weather conditions and long-term data fluctuations. With the development of big data and machine learning technologies, data-driven PV power generation estimation methods have gradually become a research hotspot. Among these methods, how to utilize deep feature extraction and modeling of complex time series data has become crucial for improving estimation accuracy.

[0003] Existing data-driven photovoltaic (PV) power generation estimation methods typically employ a combination of time series analysis and machine learning models. These methods rely primarily on simple statistical features (such as mean and standard deviation) for feature extraction, often failing to fully exploit the topological information within the time series. Furthermore, most methods neglect higher-order relationships and complex dynamic characteristics of the time series, using only shallow models for regression prediction, making them ill-suited to handle the nonlinear and multi-scale characteristics of PV power generation. The accuracy and robustness of traditional methods are further reduced, especially in the face of extreme weather and environmental changes. Therefore, there is an urgent need for an innovative estimation method that can effectively capture the deep topological features of time series data and combine them with multidimensional statistical features. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that traditional methods only utilize basic statistical features, ignoring the topological structure information in time series. This invention, by introducing Betty curves, can effectively extract deep-level topological features of time series. Existing methods perform poorly when dealing with multi-scale time series and nonlinear relationships. This invention combines multi-dimensional feature input and the XGBoost model to achieve efficient modeling of complex dynamic characteristics. Traditional methods typically do not consider the periodicity and categorical properties of the time dimension. This invention enhances the ability to capture time-dimensional features through one-hot encoding of specific information.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a data-driven photovoltaic power generation estimation method based on the Betty curve, comprising: collecting data of a first object, performing a first reconstruction calculation, and obtaining a first feature of the first object.

[0007] Statistical calculations are performed on the data of the first object to obtain the second feature of the first object.

[0008] A first transformation task is performed on the specific information of the first object data to obtain the result of the first transformation task.

[0009] The first feature, the second feature, and the result of the first transformation task are combined to form a feature space, which is then input into the first prediction model to output the prediction result of the first object.

[0010] As a preferred embodiment of the data-driven photovoltaic power generation estimation method based on the Betty curve described in this invention, the step of collecting first object data and performing the first analysis includes collecting time-related data of the first object, converting the time series into point cloud data, and filtering and calculating to obtain the first feature.

[0011] As a preferred embodiment of the data-driven photovoltaic power generation estimation method based on the Betty curve described in this invention, the step of performing statistical calculations based on the first object data to obtain the second feature of the first object includes calculating and extracting multiple statistical features based on the first object data.

[0012] As a preferred embodiment of the data-driven photovoltaic power generation estimation method based on the Betty curve described in this invention, the first transformation task for specific information of the first object data includes encoding the specific information of the first object data and converting the specific information into the result of the first transformation task.

[0013] As a preferred embodiment of the data-driven photovoltaic power generation estimation method based on the Betty curve described in this invention, the first object includes, but is not limited to, photovoltaic power generation equipment, and the data of the first object includes historical photovoltaic power generation time series and historical load time series.

[0014] The first reconstruction computation includes, but is not limited to, phase space reconstruction using Takens embedding theory.

[0015] The first feature is the topological feature, and the second feature is the statistical feature.

[0016] The first transformation task is to convert date features into category features, and the result of the first transformation task is category features.

[0017] The first predictive model includes, but is not limited to, machine learning regressors.

[0018] As a preferred embodiment of the data-driven photovoltaic power generation estimation method based on Betty curves described in this invention, the step of performing a first reconstruction calculation to obtain the first feature of the first object includes using Takens embedding theory to perform phase space reconstruction and converting the time series into point cloud data. X = {x0,…,x} n-1 Given a one-dimensional time series}, and a delay parameter τ and a dimension parameter d, the Takens phase space reconstruction point at time t is defined as:

[0019] z t :=(x t ,x t+τ ,…,x t+(d-1)τ )∈R d

[0020] Vietoris-Rips filtering is applied to the point cloud data to generate a filtered complex and calculate the continuous graph. The 0-dimensional and 1-dimensional Betty curves are extracted from the graph as the topological features of the time series and used as the first feature.

[0021] As a preferred embodiment of the data-driven photovoltaic power generation estimation method based on the Betty curve described in this invention, the step of performing statistical calculations based on the data of the first object to obtain the second feature of the first object includes calculating the absolute energy feature, the average value feature, the standard deviation feature, the skewness feature, the kurtosis feature, the absolute average change feature, the average second difference center feature, and the median feature.

[0022] The first transformation task is performed, and the results of the first transformation task include converting date information into categorical features through one-hot encoding. A date has 12 possible month attributes, 31 possible day attributes, and 7 possible week attributes. One-hot encoding is vectorized to obtain a 50-dimensional categorical feature vector.

[0023] A data-driven photovoltaic power generation estimation system based on Betty curves, characterized in that it includes:

[0024] The feature calculation module collects data on the first object, performs a first reconstruction calculation, and obtains the first feature of the first object. Based on the first object data, it performs statistical calculations to obtain the second feature of the first object.

[0025] The conversion module performs a first conversion task based on specific information of the first object data and obtains the result of the first conversion task.

[0026] The prediction module combines the first feature, the second feature, and the result of the first transformation task to form a feature space, which is then input into the first prediction model, and outputs the prediction result for the first object.

[0027] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0028] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0029] The beneficial effects of this invention are as follows: Time series data is transformed into point cloud data using Takens embedding theory, and a persistence graph is calculated through Vietoris-Rips filtering, from which topological features of the Betty curve are extracted. This effectively captures the complex dynamic patterns of the time series, providing a unique high-dimensional feature description for subsequent predictions. Compared to traditional simple statistical features, this method can describe the nonlinear dynamic structure of the time series, improving the ability to characterize photovoltaic power changes under complex weather conditions. It extracts various statistical features, including mean, standard deviation, and skewness, reflecting the distribution characteristics of the time series. It supplements the low-dimensional statistical information not fully covered by topological features, achieving diversification of the feature space and improving the generalization ability of the prediction model.

[0030] One-hot encoding transforms date features into category features, further supplementing the temporal dimension of the time series. This enhances the feature space's ability to interpret temporal information, enabling the model to better capture periodic patterns. Topological, statistical, and category features are concatenated into a feature space, which is then input into the XGBoost regression model for prediction. By combining multidimensional features and employing efficient machine learning algorithms, high-precision estimation of photovoltaic power generation is achieved, while simultaneously improving the model's adaptability to nonlinear and multi-scale characteristics. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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. Wherein:

[0032] Figure 1 The first embodiment of the present invention provides an overall flowchart of a data-driven photovoltaic power generation estimation method and system based on the Betty curve.

[0033] Figure 2 This is a periodic time series diagram of a data-driven photovoltaic power generation estimation method and system based on Betty curves, provided as the first embodiment of the present invention.

[0034] Figure 3This is a chaotic time series diagram of a data-driven photovoltaic power generation estimation method and system based on Betty curves, provided as the first embodiment of the present invention.

[0035] Figure 4 The first embodiment of the present invention provides a 0-dimensional Betty curve plot of a data-driven photovoltaic power generation estimation method and system based on Betty curves.

[0036] Figure 5 The first embodiment of the present invention provides a 1D Betty curve diagram of a data-driven photovoltaic power generation estimation method and system based on Betty curves.

[0037] Figure 6 The first embodiment of the present invention provides a 2D Betty curve diagram of a data-driven photovoltaic power generation estimation method and system based on Betty curves.

[0038] Figure 7 Elia photovoltaic data graph, which is provided as a second embodiment of the present invention, is a data-driven photovoltaic power generation estimation method and system based on the Betty curve.

[0039] Figure 8 The prediction results of the present invention, the NbeatsX model and the persistence model, based on the data-driven photovoltaic power generation estimation method and system based on the Beatty curve, provided in the second embodiment of the present invention, over a continuous 10 days.

[0040] Figure 9 The second embodiment of the present invention provides a data-driven photovoltaic power generation estimation method and system based on the Betty curve. The method and the prediction results over 10 consecutive days without using topological features. Detailed Implementation

[0041] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0042] Example 1, referring to Figures 1-6 As an embodiment of the present invention, a data-driven photovoltaic power generation estimation method based on the Betty curve is provided, comprising:

[0043] S1: Collect the data of the first object, perform the first reconstruction calculation, and obtain the first feature of the first object.

[0044] In this invention, the first object is a photovoltaic power generation device, the first object data includes historical photovoltaic power generation time series and historical load time series, the first reconstruction calculation is phase space reconstruction using Takens embedding theory, and the first feature is a topological feature.

[0045] To analyze the topological properties of historical photovoltaic and historical load time series, it is necessary to first convert the time series into a point cloud representation. The method used in this invention is phase space reconstruction, and its theoretical basis comes from the following theorem:

[0046] Takens' embedding theorem, if... For an n-dimensional compact manifold, φ: R×M→M and f: M→R are almost everywhere smooth mappings. Given a delay parameter τ, for a point x in M, define x as... i =φ(i·τ,x), then,g:M→R 2n+1 ,x→(f(x),f(x1),…·,f(x 2n )) is an embedding mapping between manifolds.

[0047] Specifically, in the scenario studied in this invention, if X = {x0, ..., x...} n-1 If} is a one-dimensional time series, then given the delay parameter τ and the dimension parameter d, the Takens phase space reconstruction point at time t is defined as:

[0048] z t :=(x t ,x t+τ ,…,x t+(d-1)τ )∈R d

[0049] Given parameters (d, τ), phase space reconstruction transforms the time series into point cloud data.

[0050] Figure 2 This is a schematic diagram of a periodic time series. Figure 3 This is a schematic diagram of a chaotic time series.

[0051] Based on phase space reconstruction, Vietoris-Rips filtering is used to filter the point cloud data to obtain a filtered complex, and a persistence graph is obtained based on this. Using the persistence graph, the Betti curve of the time series can be calculated.

[0052] Using a persistence graph, the Betti curves of the time series are calculated. For a given dimension d, the d-dimensional persistent homology Betti curve is the curve formed by the number of free generators of homology groups in each filtering layer. By definition, a 0-dimensional Betti curve is clearly a curve that monotonically decreases along the filtering time axis, while higher-dimensional Betti curves reflect the fine-grained topological properties of the time series phase space. Figure 3 Taking time series as an example, its 0-dimensional Betty curve is as follows: Figure 4 1D Betty curves Figure 5 2D Betty curves Figure 6 .

[0053] It should be noted that the first object includes, but is not limited to, photovoltaic power generation equipment, and may also include wind power generation equipment and power load monitoring equipment. The first reconstruction calculation includes, but is not limited to, phase space reconstruction using Takens embedding theory, and may also include the sliding window method based on delayed embedding and the method based on RecurrencePlot (RP).

[0054] In an optional embodiment of the present invention, the first object is a wind power generation device, and the data of the first object is data collected based on historical wind speed time series and wind power generation time series.

[0055] A phase space reconstruction calculation based on Takens embedding theory was performed. For a wind speed time series, given a delay parameter τ = 10 minutes and an embedding dimension m = 3, the time series was mapped to a 3D point cloud using the formula: xt = {xt, xt + τ, xt + 2τ}. A complex was constructed using Vietoris-Rips filtering, and a persistence graph was calculated. 0-dimensional Betty curves (number of connected components) and 1-dimensional Betty curves (number of cycles) were extracted as topological features of the wind speed time series.

[0056] In an optional embodiment of the present invention, the first object is a power load monitoring device, and the data for the first object is data collected based on historical power load time series and voltage fluctuation time series. Takens embedding theory phase space reconstruction calculations are performed. For the power load time series, given a delay parameter τ = 15 minutes and an embedding dimension m = 4, the formula xt = {xt, xt+τ, xt+2τ, xt+3τ} is used to map it into a four-dimensional point cloud. An Alpha filtering algorithm is applied to the generated point cloud data to construct a topological complex and calculate 0-dimensional and 1-dimensional Betty curves to describe the topological characteristics of the load time series.

[0057] In an optional embodiment of the present invention, the first reconstruction calculation method is a sliding window method based on delayed embedding. For photovoltaic power generation equipment, based on the historical photovoltaic power generation time series, the sliding window width is set to W = 60 minutes, and the step size is S = 15 minutes. Each time series segment within the sliding window is represented by delayed embedding:

[0058] x t ={x t ,x t+S ,x t+2S ,…,x t+(W-1)S}

[0059] Where, x t A collection representing time series.

[0060] Topological feature extraction: The Persistent Cohomology method is applied to the embedded point cloud of each sliding window to extract the persistence graph and calculate the Betty curve to describe the topological characteristics of photovoltaic power generation.

[0061] In an optional embodiment of the present invention, the first reconstruction calculation method is a Recurrence Plot (RP) based method. For the time series data of photovoltaic power generation equipment, an RP matrix is ​​constructed:

[0062] R(i,j)=Θ(∈-|x i -x j |)

[0063] Where Θ is the step function, ∈ is the distance threshold, and x i and x j It is the point of embedding.

[0064] The Cliques Topology algorithm is applied to the similarity network in the RP matrix to extract 0-dimensional and 1-dimensional Betty curves, which describe the cyclic and connectivity properties of the time series.

[0065] S2: Perform statistical calculations based on the data of the first object to obtain the second feature of the first object.

[0066] Calculate the statistical characteristics of historical weather, historical photovoltaic, and historical load time series data. The specific characteristics and formulas involved are as follows:

[0067] Absolute energy, which is the sum of the squares of the components of the time series:

[0068] E=∑x i 2

[0069] The average value, i.e., the mean of a time series:

[0070]

[0071] Standard deviation:

[0072]

[0073] Skewness, i.e., the third normalized moment of a time series:

[0074]

[0075] It is a measure that describes the direction and degree of skewness in the distribution of statistical data.

[0076] Kurtosis, also known as the fourth-order normalized moment of a time series:

[0077]

[0078] It is a characteristic quantity that describes the peak value of the probability density distribution curve at the mean.

[0079] The absolute average change is the mean of the absolute values ​​of continuous changes in a time series.

[0080]

[0081] Mean second-order difference center: that is, the mean of the second-order change of the time series.

[0082]

[0083] The median is the median of each value in a time series when arranged from smallest to largest.

[0084] S3: Perform a first transformation task based on specific information of the first object data to obtain the result of the first transformation task.

[0085] In this invention, the specific information of the first object data is date information, the first transformation task is to convert the date feature into a category feature, and the result of the first transformation task is a category feature.

[0086] Furthermore, one-hot encoding, that is, for a variable with n states, using e i Express its i-th state (here e) i (A column vector of n dimensions, where the i-th component is 1 and the rest are 0) Specifically, a date has 12 possible month attributes, 31 possible day attributes, and 7 possible week attributes. One-hot encoding uses a 50-dimensional feature vector to represent the attribute information of the date through vectorization.

[0087] S4: Combine the first feature, the second feature, and the result of the first transformation task to form a feature space, input it into the first prediction model, and output the prediction result of the first object.

[0088] All features are concatenated to form a feature space. Finally, the feature space is input into a machine learning regressor for training to predict photovoltaics. The first prediction model of this invention is the XGBoost regressor. The XGBoost regressor is a powerful, flexible, and efficient machine learning tool that provides excellent performance and extensive functionality through a gradient boosting framework, suitable for solving various complex regressor problems. Therefore, XGBoost was chosen as the machine learning regressor.

[0089] It should be noted that the first prediction model includes, but is not limited to, the XGboos regressor, and can also be a random forest regressor and a support vector machine regressor.

[0090] In an optional embodiment of the present invention, the first prediction model is a random forest regressor, which is an ensemble learning method based on decision trees. It improves prediction accuracy by weighted averaging of the results of multiple decision trees and has strong robustness and generalization ability.

[0091] The number of decision trees is set to N=100, the maximum depth D=10, and the minimum number of split samples is 5. The feature space is formed by concatenating the first feature, the second feature, and the results of the first transformation task. A random forest model is constructed using training data of historical photovoltaic power generation time series. The feature space of the test data is input into the random forest regressor, which outputs the predicted value of photovoltaic power generation.

[0092] Random forests can effectively handle the relationships between nonlinear features, are less prone to overfitting, and are suitable for complex multidimensional regression problems such as photovoltaic power generation prediction.

[0093] In an optional embodiment of the present invention, the first prediction model is a support vector machine regressor. The support vector machine regressor makes predictions by finding the optimal hyperplane between data points, and has strong global optimization capabilities, making it particularly suitable for small sample and high-dimensional data problems.

[0094] A radial basis function (RBF) kernel with kernel parameter γ = 0.1 and regularization parameter C = 1 is used. The concatenated feature space is input. A support vector machine (SVR) regressor is trained by minimizing the loss function within the error range (∈). The trained SVR model is then used to predict the test data, outputting an estimated photovoltaic power generation value.

[0095] SVR has strong processing capabilities for high-dimensional spatial data, enabling it to accurately fit nonlinear relationships in photovoltaic power generation prediction while avoiding over-reliance on sample size.

[0096] The computer device can be a server. This computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores data cluster data from a power monitoring system. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a data-driven photovoltaic power generation estimation method based on Betty curves.

[0097] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0098] Example 2, refer to Figures 7-9 As an embodiment of the present invention, a data-driven photovoltaic power generation estimation method and system based on the Betty curve is provided. To verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0099] This invention selects the publicly available Elia dataset as experimental data. Elia is a major Belgian power company responsible for the construction and maintenance of the country's power transmission and distribution infrastructure. Elia focuses on the construction and maintenance of electrical infrastructure, including high-voltage transmission lines and distribution networks. The Elia dataset is sampled at 15-minute intervals, i.e., four data points per hour. The visualization of the data used is as follows: Figure 7 As shown in Table 1.

[0100] Table 1 Descriptive Statistics

[0101] average value 339.5 Standard deviation 535.1 Minimum value 0 25％ 0 50％ 1.68 75％ 515.9 Maximum value 2383.2

[0102] In the experiments of this invention, the dataset was divided into training and test sets. The last 20% of the dataset (approximately three months, from September 2016 to December 2016) was used as the test set. Prediction performance was evaluated using mean absolute error (MAE), standardized mean absolute error (nMAE), and standardized root mean square error (nRMSE). These are defined as follows (where y represents the true value of PV, ...). (This represents the predicted value of PV):

[0103]

[0104] The final experimental results are as follows Figure 8 As shown in Table 2.

[0105] Table 2. Prediction results of the proposed method, NbeatsX model, and persistence model.

[0106] This article's method NbeatsX Continuous model MAE 53.7 77.6 84.2 nMAE 0.028 0.038 0.041 nRMSE 0.051 0.088 0.100

[0107] from Figure 8 As shown in Table 2, the method presented in this paper can effectively reduce the error of photovoltaic prediction: compared with other methods, the method presented in this paper reduces MAE by at least 21.1%, nMAE by at least 1%, and nRMSE by at least 3.7%.

[0108] Table 3. Prediction results of the proposed method and the method with removed topological features.

[0109] This article's method Remove topological features MAE 53.7 66.7 nMAE 0.028 0.032 nRMSE 0.051 0.068

[0110] from Figure 9 As shown in Table 3, topological features effectively reduce photovoltaic prediction errors: compared to not using topological features, using topological features reduces MAE by 13%, nMAE by 0.4%, and nRMSE by 2.2%. Furthermore, from... Figure 9 It is easy to see that for some atypical clear-sky days (such as the third, fourth, seventh, and eighth days), using only the statistical characteristics of photovoltaics and loads and historical weather data will result in deviations in the time scale (peak lag) or in the amplitude scale (peak overestimation). In this case, topological features can effectively improve the accuracy of prediction.

[0111] Example 3, an embodiment of the present invention, includes a data-driven photovoltaic power generation estimation system based on the Betty curve, specifically:

[0112] The feature calculation module collects data on the first object, performs a first reconstruction calculation, and obtains the first feature of the first object. Based on the first object data, it performs statistical calculations to obtain the second feature of the first object.

[0113] The conversion module performs a first conversion task based on specific information of the first object data and obtains the result of the first conversion task.

[0114] The prediction module combines the first feature, the second feature, and the result of the first transformation task to form a feature space, which is then input into the first prediction model, and outputs the prediction result for the first object.

[0115] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data-driven photovoltaic power generation estimation method based on the Betty curve, characterized in that, include: Collect data of the first object, perform the first reconstruction calculation, and obtain the first feature of the first object; Statistical calculations are performed on the data of the first object to obtain the second feature of the first object; Based on specific information of the first object data, perform a first transformation task to obtain the result of the first transformation task; The first feature, the second feature, and the result of the first transformation task are combined to form a feature space, which is then input into the first prediction model to output the prediction result of the first object. The first object includes photovoltaic power generation equipment, and the data of the first object includes historical photovoltaic power generation time series and historical load time series; The first reconstruction computation includes phase space reconstruction using Takens embedding theory; The first feature is a topological feature, and the second feature is a statistical feature; The first transformation task is to convert date features into category features, and the result of the first transformation task is category features. The first predictive model includes a machine learning regressor; The first reconstruction calculation to obtain the first feature of the first object includes using Takens embedding theory to perform phase space reconstruction and converting the time series into point cloud data; It is a one-dimensional time series, given a delay parameter. and dimension parameters The Takens phase space reconstruction point at time t is defined as: ； Vietoris-Rips filtering is applied to point cloud data to generate filtered complexes and calculate a continuous graph. The 0-dimensional and 1-dimensional Betty curves are extracted from the graph as the topological features of the time series and used as the first feature. The step of performing statistical calculations based on the data of the first object to obtain the second characteristic of the first object includes calculating the absolute energy characteristic, the mean characteristic, the standard deviation characteristic, the skewness characteristic, the kurtosis characteristic, the absolute mean change characteristic, the mean second difference center characteristic, and the median characteristic. The first transformation task is performed, and the results of the first transformation task include converting date information into categorical features through one-hot encoding. A date has 12 possible month attributes, 31 possible day attributes, and 7 possible week attributes. One-hot encoding is vectorized to obtain a 50-dimensional categorical feature vector.

2. The data-driven photovoltaic power generation estimation method based on the Betty curve as described in claim 1, characterized in that: The step of collecting data of the first object and performing the first reconstruction calculation includes collecting time-related data of the first object, converting the time series into point cloud data, and filtering and calculating to obtain the first feature.

3. The data-driven photovoltaic power generation estimation method based on the Betty curve as described in claim 2, characterized in that: The step of performing statistical calculations based on the first object data to obtain the second feature of the first object includes calculating and extracting multiple statistical features based on the first object data.

4. The data-driven photovoltaic power generation estimation method based on the Betty curve as described in claim 3, characterized in that: The first transformation task for specific information of the first object data includes encoding the specific information of the first object data and converting the specific information into the result of the first transformation task.

5. A dynamic load capacity management system for an oil-immersed power transformer employing the method described in any one of claims 1-4, characterized in that: The feature calculation module collects data of the first object, performs the first reconstruction calculation, and obtains the first feature of the first object. Statistical calculations are performed on the data of the first object to obtain the second feature of the first object; The conversion module performs a first conversion task based on specific information of the first object data and obtains the result of the first conversion task. The prediction module combines the first feature, the second feature, and the result of the first transformation task to form a feature space, which is then input into the first prediction model, and outputs the prediction result for the first object.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images