Multi-element periodic grid modeling method and system based on power prediction

Abstract

The invention discloses a multivariate periodic grid modeling method and system based on power prediction, and belongs to the technical field of new energy power generation, and the method mainly comprises four steps: meteorological and power field tuple formation, data normalization processing, and multivariate grid formation. According to the method, distributed photovoltaic predicted power correction in a grid is carried out based on meteorological information and generated power information of a space grid where the distributed photovoltaic is located. The method at least has the beneficial technical effects that data required by power prediction can be fully utilized, the application of the method can be combined with a regression model, and influences of various factors in the photovoltaic power generation process are fully considered, so that the prediction accuracy and reliability are improved, and better support is provided for operation and management of a photovoltaic power generation system.

Description

Technical Field

[0001] The present invention belongs to the technical field of renewable energy power generation, and more specifically, relates to a multivariate periodic grid modeling method and system based on power prediction. Background Art

[0002] As renewable energy sources such as solar and wind power continue to gain popularity, power fluctuations caused by weather and climate have become a complex issue in power system management. In order to maintain the balance between supply and demand in the power system and ensure real-time matching of power, power forecasting provides an estimate of future power generation, which helps dispatchers plan and adjust power generation plans more accurately to meet power demand.

[0003] The relationship between meteorological data and power is usually complex and affected by many factors, such as temperature, humidity, wind speed, etc. These factors may show periodic changes in different time and space. At the same time, meteorological data and power data are usually multivariate, that is, they are not only related to time, but also to space. Therefore, using only a single model may not be able to fully capture the complex relationship between the data.

[0004] Prior art document 1 (CN117669960A) discloses a new energy power prediction method based on multivariate meteorological factors, which determines the linear correlation between new energy power generation and meteorological factors, and establishes a data set of power generation and meteorological factors; establishes a multivariate linear regression model based on the data set of power generation and meteorological factors; obtains meteorological factor forecast data, and performs deterministic forecast through the regression model; performs uncertainty forecast and probability forecast based on the meteorological factor forecast error; constructs a convolutional neural network model, and establishes a nonlinear implicit mapping relationship between the meteorological field and new energy power generation; establishes a new energy power generation prediction model combining linear regression and convolutional neural network; adjusts the model according to the newly added data, performs error check on the prediction results, and optimizes the model; performs power dispatch analysis based on the new energy power generation prediction results, and formulates a power dispatch optimization plan; adjusts the power dispatch optimization plan based on extreme weather scenario data.

[0005] It is worth noting that the power forecasting method represented by the existing technology has at least three aspects to be improved: First, although the plan mentions the need to determine the linear correlation between renewable energy power generation and meteorological factors, it does not take into account the possible nonlinear relationship or interaction between different meteorological factors. Relying solely on the linear regression model may not fully capture the nonlinear impact of complex meteorological factors on power generation; Second, although the CNN model is introduced to establish a nonlinear implicit mapping relationship between the meteorological field and renewable energy power generation, it does not clearly explain how to process meteorological time series data. CNN is usually used for image processing, and is not efficient when processing time series data; Third, the plan mentions using new data to adjust the model, but does not explain in detail how to make adaptive adjustments and how to improve the dynamic adaptability of the model through real-time or frequent feedback, that is, no practical implementation means are given. Summary of the invention

[0006] The purpose of the present invention is to solve the shortcomings of the prior art and propose a multivariate periodic grid modeling method and system based on power prediction, which provides an important reference basis for correcting the prediction and can more comprehensively consider the impact of various periodic factors on solar power generation, thereby improving the accuracy and reliability of the prediction.

[0007] A first aspect of the present invention provides a multivariate periodic grid modeling method based on power prediction, comprising the following steps:

[0008] Step 1: Collect meteorological data, measured and predicted power data related to power generation, and pair them to form data tuples. Each data tuple contains meteorological information and corresponding power values ​​at a set time and space point.

[0009] Step 2: Use normalization to process each feature in the above data tuple to form a normalized multivariate data set;

[0010] Step 3: Use Fourier transform to periodically partition the normalized multivariate data set to identify the periodic components in the data;

[0011] Step 4: After determining the periodic components, the normalized multivariate data set is meshed to obtain periodic models under different characteristics and realize multivariate periodic mesh modeling.

[0012] Preferably, in step 1, the meteorological data related to power generation include:

[0013] Temperature data: The unit is degrees Celsius (℃), recorded as Temp;

[0014] Humidity data: The unit is relative humidity (RH), recorded as H;

[0015] Irradiance data: The unit is watt per square meter (W / m 2 ), denoted as Ir;

[0016] Power data: The unit is watt (W), denoted as P;

[0017] Time data: the unit is minute (min), recorded as Time;

[0018] Spatial data: The unit is longitude (°), denoted as S;

[0019] x i : The data tuple at the i×15th minute.

[0020] Preferably, in step 1, the collected meteorological data and power data are smoothed, and the data smoothing process adopts a moving average method, which is expressed by the following formula:

[0021]

[0022] in:

[0023] k represents the size of the window;

[0024] x i represents the value of the i-th data point;

[0025] t represents the current time;

[0026] MA t represents the moving average at time t;

[0027] The collected meteorological data and power data are corrected for missing data. The missing data correction adopts the mean filling method, which is expressed by the following formula:

[0028]

[0029] in:

[0030] x i represents the value of the i-th data point;

[0031] Represents the mean of the entire data set;

[0032] n represents the total number of data points in the dataset.

[0033] Preferably, in step 2, each feature in the above data tuple is processed using L1 norm normalization to form a normalized multivariate data set;

[0034] L1 norm normalization is expressed as follows:

[0035]

[0036] in:

[0037] x ij represents the jth eigenvalue in the i-th data tuple;

[0038] ||x i ||1 represents the L1 norm of the i-th data tuple;

[0039] x′ ij Represents the normalized eigenvalue of the jth eigenvalue in the i-th data tuple.

[0040] Furthermore, the L1 norm of the i-th data tuple is expressed as follows:

[0041] ||x i ||1=|x i1 |+|x i2 |+...+|x in | (4)

[0042] in:

[0043] x i1 ,x i2 ,···,x in : The eigenvalues ​​in the i-th data tuple.

[0044] Preferably, in step 3, the normalized multivariate data set is periodically divided using a two-dimensional discrete Fourier transform, which is expressed as the following formula:

[0045]

[0046] in:

[0047] F(u,v) represents the transformation result in the frequency domain;

[0048] T(x,y) represents the raw data in the time domain; it can be an element in any two-dimensional data matrix, such as temperature, humidity, power, etc.

[0049] x and y are the row and column indices in the data matrix, indicating the spatial location of the data;

[0050] M and N are the number of rows and columns of data respectively

[0051] u and v are frequencies in the frequency domain.

[0052] Preferably, step 3 comprises:

[0053] Step 3.1: Perform a two-dimensional discrete Fourier transform on the normalized multivariate data set;

[0054] Step 3.2: Calculate the amplitude of each frequency point in the spectrum to characterize the importance of the frequency in the data;

[0055] Step 3.3, calculate the average amplitude of the spectrum as a threshold; according to the threshold, mark the frequency points in the spectrum whose amplitude is greater than the threshold as periodic components, otherwise mark them as non-periodic components;

[0056] Step 3.4: Based on the labeling results, the original data in the multivariate data set is periodically divided to separate the periodic components from the non-periodic components.

[0057] Preferably, in step 4, a multivariate normalized feature is used to form a multivariate grid, a grid model is constructed in time and space dimensions, time is divided into continuous time points, and space is divided into discrete position points;

[0058] Each grid point corresponds to a time and space position, as well as the meteorological data and power values ​​related to the position; the normalized meteorological data and actual power values ​​are filled into the corresponding grid points of the multivariate grid to form a complete multivariate periodic grid data set.

[0059] Preferably, the normalized feature data is clustered using a K-means clustering algorithm, and the data points and their cluster labels are used as nodes of a multivariate periodic grid, thereby dividing the data into different spatial regions or clusters to form a multivariate periodic grid structure.

[0060] Preferably, the normalized multivariate data set is divided into a training set and a test set, the training set is used to train a model including a prediction power model and a multivariate periodic grid model, the parameters of the model are adjusted by cross-validation and other methods, and the trained model is used to predict the test set to obtain a predicted value.

[0061] A second aspect of the present invention provides a multivariate periodic grid modeling system based on power prediction, and the multivariate periodic grid modeling method based on power prediction is executed, comprising:

[0062] Data entry module: used to enter the collected meteorological data related to power generation, measured and predicted power data, and pair them to form data tuples;

[0063] Data preprocessing module: used to preprocess the input data to form a normalized multivariate data set;

[0064] Periodic component identification module: used to periodically divide the normalized multivariate data set using Fourier transform to identify periodic components in the data;

[0065] Model generation module: used to grid the normalized multivariate data set to obtain periodic models with different characteristics and realize multivariate periodic grid modeling.

[0066] Compared with the prior art, the beneficial technical effects of the present invention at least include: the present invention proposes a multivariate periodic grid modeling method based on power prediction, which corrects the predicted power of distributed photovoltaic and wind turbines in the grid based on the meteorological information and power generation information of the spatial grid where the distributed photovoltaic and wind turbines are located. First, it is necessary to collect meteorological data related to power generation, such as temperature, humidity, light, etc. The collected meteorological data and power data are combined into tuples, each of which contains meteorological information and corresponding power values ​​at a specific time and space point. These tuples constitute the basic data unit of the multivariate periodic grid. Then the data is normalized by L1 norm to maintain the proportional consistency between different features. Then, the normalized temperature, humidity, power and light data are Fourier transformed, and the periodic components in the data are found by spectrum analysis and other methods, and then periodic division is performed. After the periodic components are determined, the data is grid-divided to form a multivariate periodic grid, and multivariate periodic grid modeling is realized, so that it can be used for power prediction, and finally the power prediction error is corrected by model training and forming a dynamic rolling of the multivariate periodic grid, so as to improve the accuracy of power prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 Flowchart for modeling multivariate periodic grids;

[0068] Figure 2 This is the result diagram of multivariate periodic grid;

[0069] Figure 3 This is the error analysis diagram between predicted power and actual power. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The described embodiments are only embodiments of a part of the present invention, not all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0071] The multivariate periodic grid modeling method can make full use of the data, more accurately capture the periodic changes of the data in time and space, and form the multivariate periodic grid required for power prediction by periodically dividing and clustering the data. The invention can be combined with the regression model in subsequent applications, fully considering the influence of various factors in the photovoltaic power generation process, thereby improving the accuracy and reliability of the prediction and providing better support for the operation and management of the photovoltaic power generation system.

[0072] like Figure 1 As shown, embodiment 1 of the present invention provides a multivariate periodic grid modeling method based on power prediction, which mainly includes: forming meteorological and power domain tuples, processing multiple tuples using L1 norm normalization, periodically dividing data, and forming a multivariate grid using multivariate normalization features. Based on the meteorological information and power generation information of the spatial grid where the distributed photovoltaic is located, the distributed photovoltaic predicted power in the grid is corrected.

[0073] The multivariate periodic grid modeling method based on power prediction specifically comprises the following steps:

[0074] Step 1: It is necessary to collect meteorological data related to power generation, such as temperature, humidity, light, etc., and form them into meteorological tuples. At the same time, the measured and predicted power data are formed into power tuples, each of which contains meteorological information and corresponding power values ​​at a specific time and space point. These tuples constitute the basic data unit of the multivariate periodic grid.

[0075] In a preferred but non-limiting embodiment of the present invention, the required meteorological data and power data are determined, the required data are collected according to the timestamp, the collected meteorological data and power data are smoothed and the missing data are corrected, and the processed meteorological data and power data are paired to form a data tuple. It is expressed in the form of a feature vector, denoted as x i .

[0076] Furthermore, the required data include:

[0077] Temperature data: The unit is degrees Celsius (℃), recorded as Temp;

[0078] Humidity data: The unit is relative humidity (RH), recorded as H;

[0079] Irradiance data: The unit is watt per square meter (W / m 2 ), denoted as Ir;

[0080] Power data: The unit is watt (W), denoted as P;

[0081] Time data: the unit is minute (min), recorded as Time;

[0082] Spatial data: The unit is longitude (°), denoted as S;

[0083] Additionally, timestamps are recorded every 15 minutes;

[0084] x i : The data tuple at the i×15th minute.

[0085] It is worth noting that the meteorological numbers related to power generation given above are a preferred but non-restrictive implementation method of the present invention. In the prior art, the correlation between meteorology and power generation may be different, which may be directly or indirectly related. After research by the inventors, the meteorological numbers related to power generation given above are a preferred way to implement the present invention, but those skilled in the art can choose more or fewer types of meteorological data, which all fall within the scope of the present invention.

[0086] In a preferred but non-limiting embodiment of the present invention, the collected meteorological data and power data are subjected to data smoothing, and the data smoothing adopts a moving average method, which is expressed by the following formula:

[0087]

[0088] in:

[0089] k represents the size of the window;

[0090] x i represents the value of the i-th data point;

[0091] t represents the current time;

[0092] MA t represents the moving average at time t.

[0093] In a preferred but non-limiting embodiment of the present invention, the collected meteorological data and power data are corrected for missing data, and the missing data correction adopts the mean filling method, which is expressed by the following formula:

[0094]

[0095] in:

[0096] x i represents the value of the i-th data point;

[0097] Represents the mean of the entire data set;

[0098] n represents the total number of data points in the dataset.

[0099] Step 2: After obtaining the data tuple processed by step 1, each feature in the data tuple is processed by normalization to form a normalized multivariate data set. In a preferred but non-limiting embodiment of the present invention, the data is normalized by L1 norm to maintain the proportional consistency between different features.

[0100] L1 norm normalization is expressed as follows:

[0101]

[0102] in:

[0103] x ij represents the jth eigenvalue in the i-th data tuple;

[0104] ||x i ||1 represents the L1 norm of the i-th data tuple;

[0105] x′ ij Represents the normalized eigenvalue of the jth eigenvalue in the i-th data tuple.

[0106] Furthermore, the L1 norm of the i-th data tuple is expressed as follows:

[0107] ||x i ||1=|x i1 |+|x i2 |+...+|x in | (4)

[0108] in:

[0109] x i1 ,x i2 ,···,x in : The eigenvalues ​​in the i-th data tuple.

[0110] Step 3: Perform Fourier transform on the normalized data, use spectrum analysis and other methods to find the periodic components in the data, and then perform periodic division. To perform periodic division on the data, you can treat it as a two-dimensional signal and apply the discrete Fourier transform formula to calculate its two-dimensional Fourier transform F(u,v). This can convert the data from the spatial domain to the frequency domain, so that the periodic components in the data can be observed.

[0111] In the results of the 2D Fourier transform, larger amplitude values ​​usually indicate a periodic structure in the original data, expressed as follows:

[0112]

[0113] in:

[0114] F(u,v) represents the transformation result in the frequency domain;

[0115] T(x,y) represents the raw data in the time domain; it can be an element in any two-dimensional data matrix, such as temperature, humidity, power, etc.

[0116] x and y are the row and column indices in the data matrix, indicating the spatial location of the data;

[0117] M and N are the number of rows and columns of data respectively

[0118] u and v are frequencies in the frequency domain.

[0119] In a preferred but non-limiting embodiment of the present invention, the data periodic partitioning process includes:

[0120] Step 3.1: Perform a two-dimensional discrete Fourier transform (DFT) on the normalized data.

[0121] Step 3.2: Calculate the amplitude of each frequency point in the spectrum to indicate the importance of the frequency in the data.

[0122] Step 3.3, calculate the average amplitude of the spectrum as the threshold. According to the threshold, mark the frequency points in the spectrum with amplitude greater than the threshold as periodic components, otherwise mark them as non-periodic components.

[0123] Step 3.4: According to the labeling results, the original data is periodically divided to separate the periodic components from the non-periodic components.

[0124] Step 4: After determining the periodic components, the data is gridded to obtain periodic models under different features and realize multivariate periodic grid modeling. Use multivariate normalization features to form a multivariate grid, build a grid model in the time and space dimensions, divide time into continuous time points, and divide space into discrete location points. Each grid point corresponds to a time and space location, as well as the meteorological data and power values ​​related to the location. Fill the normalized meteorological data and actual power values ​​into the corresponding grid points of the multivariate grid to form a complete multivariate periodic grid data set.

[0125] In a preferred but non-limiting embodiment of the present invention, the multivariate grid establishment involves clustering the normalized feature data using a K-means clustering algorithm, thereby dividing the data into different spatial regions or clusters to form a multivariate grid. K-means clustering is a commonly used unsupervised learning algorithm, which aims to divide a data set into K different groups or clusters, so that each data point belongs to the center point of the cluster closest to it. For example, but not limited to, there is a data set X containing n samples i ={x1,x2,...,x n}, each sample X i There are m features in each. The goal of K-means clustering is to divide the data set into K clusters so that each sample belongs to the center of the cluster closest to it.

[0126] In a further preferred but non-limiting embodiment of the present invention, the specific process of forming a multivariate grid using multivariate normalization features includes:

[0127] Step 4.1: Use the normalized feature data as input to the K-means clustering algorithm.

[0128] Step 4.2: Initialize the center points of K clusters. You can randomly select K data points in the data set as the initial center points.

[0129] The K-means clustering algorithm is used to cluster the data and obtain the cluster label to which each data point belongs.

[0130] Step 4.3: Use the data points and their cluster labels as nodes of the multivariate grid to form a multivariate grid structure.

[0131] It is worth noting that, as a prominent substantive feature of the present invention and one of the significant advances achieved, after the above steps to realize multivariate periodic grid modeling, the data required for power prediction can be fully utilized, and the subsequent application of the invention can be combined with the regression model to fully consider the influence of various factors in the photovoltaic power generation process, thereby improving the accuracy and reliability of the prediction and providing better support for the operation and management of the photovoltaic power generation system.

[0132] In order to more clearly introduce the outstanding substantive features of the multivariate periodic grid modeling method based on power prediction of the present invention and the significant progress it brings to the prior art, the following is a verification example of the present invention to further verify the feasibility of the above multivariate periodic grid modeling process. This verification example uses existing photovoltaic prediction and measured data to perform multivariate periodic grid modeling simulation in MATLAB according to the process. The simulation results are as follows: Figure 2 shown.

[0133] The simulation distinguishes the distribution of different characteristic data in the multivariate periodic grid. It helps to intuitively understand the distribution law of data in time and space, as well as the correlation and periodic changes between different characteristics, thus providing a basis for the subsequent correction of power prediction errors. The modeling effect of the multivariate periodic grid is judged by the separation between different clusters and the similarity between data points in different clusters. Generally, the closer the data points are to the centroid, the finer the grid. Therefore, the degree of grid refinement can be judged by the Euclidean distance between the data points and the centroid under different K values. The following table shows the Euclidean distance under different K values.

[0134] From the final results, we can see that when K = 4, the effect is optimal, and the Euclidean distance is very small, and the grid refinement is high.

[0135] K value 2 3 4 5 Euclidean distance 1.2716 0.91126 0.66292 0.68539

[0136] At the same time, the data can be divided into training sets and test sets to verify the model performance.

[0137] The entire data set is divided into a training set and a test set. The commonly used division ratio is 70% for training and 30% for testing. The ratio can be adjusted according to the actual situation. Now select 2 prediction data for comparative analysis.

[0138] The training set data is used to train the model. The models here include the power prediction model and the multivariate periodic grid model.

[0139] During the training process, the model parameters are adjusted through methods such as cross-validation so that the model performs best on the training set.

[0140] Use the trained model to predict the test set and get the predicted value.

[0141] Calculate the error between the predicted value and the measured value. Commonly used error indicators include root mean square error (RMSE), mean absolute error (MAE), etc.

[0142] The trained model and data are put into the multivariate periodic grid, and the predicted power value is obtained according to the model output.

[0143] Compare the power curves of the training set and the test set to see whether the power curve after grid refinement is close to the actual power curve. The finer the grid, the smaller the error between the predicted curve and the measured curve. From the final result, we can see that the predicted power 1 is closest to the measured power, such as Figure 3 shown.

[0144] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0145] Embodiment 2 of the present invention provides a multivariate periodic grid modeling system based on power prediction, which runs a multivariate periodic grid modeling method based on power prediction according to embodiment 1, including:

[0146] Data entry module: used to enter the collected meteorological data related to power generation, measured and predicted power data, and pair them to form data tuples;

[0147] Data preprocessing module: used to preprocess the input data to form a normalized multivariate data set;

[0148] Periodic component identification module: used to periodically divide the normalized multivariate data set using Fourier transform to identify periodic components in the data;

[0149] Model generation module: used to grid the normalized multivariate data set to obtain periodic models with different characteristics and realize multivariate periodic grid modeling.

[0150] Compared with the prior art, the beneficial technical effects of the present invention at least include: the present invention proposes a multivariate periodic grid modeling method based on power prediction, which corrects the predicted power of distributed photovoltaic and wind turbines in the grid based on the meteorological information and power generation information of the spatial grid where the distributed photovoltaic and wind turbines are located. First, it is necessary to collect meteorological data related to power generation, such as temperature, humidity, light, etc. The collected meteorological data and power data are combined into tuples, each of which contains meteorological information and corresponding power values ​​at a specific time and space point. These tuples constitute the basic data unit of the multivariate periodic grid. Then the data is normalized by L1 norm to maintain the proportional consistency between different features. Then, the normalized temperature, humidity, power and light data are Fourier transformed, and the periodic components in the data are found by spectrum analysis and other methods, and then periodic division is performed. After the periodic components are determined, the data is grid-divided to form a multivariate periodic grid, and multivariate periodic grid modeling is realized, so that it can be used for power prediction, and finally the power prediction error is corrected by model training and forming a dynamic rolling of the multivariate periodic grid, so as to improve the accuracy of power prediction.

[0151] Furthermore, compared with the prediction method represented by prior art document 1, the present invention has significantly improved its three shortcomings, and the multivariate periodic grid modeling provided by the present invention constructs a periodic grid to more carefully process the periodicity, seasonality and long-term trends of time series data. This method is particularly suitable for the periodic fluctuations (such as day and night cycles, seasonal cycles, etc.) that are prevalent in new energy power. It can better capture the long-term and short-term change patterns in time series data, and handle abnormal situations more finely, and it has its own periodic grid structure when modeling, accurately captures the dependencies between multiple variables and their fluctuations, and can effectively reflect the predicted fluctuations caused by uncertain factors, thereby providing a more accurate probability forecast. Finally, a model is established by periodic grids and multivariate time series data, which not only takes into account historical data, but also can adaptively respond to the periodic fluctuation characteristics of different time periods. The optimized model has stronger dynamic adaptability and stability.

[0152] Compared with the traditional multivariate meteorological factor regression and CNN methods, the method based on multivariate periodic grid modeling in the present invention can more effectively handle the periodic fluctuations and uncertainties in renewable energy power. Through a more complex grid structure, capturing periodic and nonlinear relationships, the accuracy and stability of power forecasting can be significantly improved, and stronger adaptability can be demonstrated under extreme weather conditions.

[0153] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A multivariate periodic grid modeling method based on power prediction, characterized by: The following steps are involved: Step 1: Collect meteorological data, measured and predicted power data related to power generation, and pair them to form data tuples. Each data tuple contains meteorological information and corresponding power values ​​at a set time and space point. Step 2: Use normalization to process each feature in the above data tuple to form a normalized multivariate data set; Step 3: Use Fourier transform to periodically partition the normalized multivariate data set to identify the periodic components in the data; Step 4: After determining the periodic components, the normalized multivariate data set is meshed to obtain periodic models under different characteristics and realize multivariate periodic mesh modeling.

2. The multivariate periodic grid modeling method based on power prediction according to claim 1 is characterized by: In step 1, meteorological parameters related to power generation include: Temperature data: The unit is degrees Celsius (℃), recorded as Temp; Humidity data: The unit is relative humidity (RH), recorded as H; Irradiance data: The unit is watt per square meter (W / m 2 ), denoted as Ir; Power data: The unit is watt (W), denoted as P; Time data: the unit is minute (min), recorded as Time; Spatial data: The unit is longitude (°), denoted as S; x i : The data tuple at the i×15th minute.

3. A multivariate periodic grid modeling method based on power prediction according to claim 1 or 2, characterized in that: In step 1, the collected meteorological data and power data are smoothed. The data smoothing process adopts the moving average method, which is expressed by the following formula: in: k represents the size of the window; x i represents the value of the i-th data point; t represents the current time; MA t represents the moving average at time t; The collected meteorological data and power data are corrected for missing data. The missing data correction adopts the mean filling method, which is expressed by the following formula: in: x i represents the value of the i-th data point; Represents the mean of the entire data set; n represents the total number of data points in the dataset.

4. A multivariate periodic grid modeling method based on power prediction according to claim 1 or 2, characterized in that: In step 2, each feature in the above data tuple is processed using L1 norm normalization to form a normalized multivariate data set; L1 norm normalization is expressed as follows: in: x ij represents the jth eigenvalue in the i-th data tuple; ||x i ||1 represents the L1 norm of the i-th data tuple; x i ' j Represents the normalized eigenvalue of the jth eigenvalue in the i-th data tuple. Furthermore, the L1 norm of the i-th data tuple is expressed as follows: ||x i ||1=|x i1 |+|x i2 |+...+|x in | (4) in: x i1 ,x i2 ,···,x in : Each eigenvalue in the i-th data tuple.

5. A multivariate periodic grid modeling method based on power prediction according to claim 1 or 2, characterized in that: In step 3, the normalized multivariate data set is periodically divided using a two-dimensional discrete Fourier transform, which is expressed as follows: in: F(u,v) represents the transformation result in the frequency domain; T(x,y) represents the raw data in the time domain; it can be an element in any two-dimensional data matrix, such as temperature, humidity, power, etc. x and y are the row and column indices in the data matrix, indicating the spatial location of the data; M and N are the number of rows and columns of data respectively u and v are frequencies in the frequency domain.

6. A multivariate periodic grid modeling method based on power prediction according to claim 1 or 2, characterized in that: Step 3 includes: Step 3.1: Perform a two-dimensional discrete Fourier transform on the normalized multivariate data set; Step 3.2: Calculate the amplitude of each frequency point in the spectrum to characterize the importance of the frequency in the data; Step 3.3, calculate the average amplitude of the spectrum as a threshold; according to the threshold, mark the frequency points in the spectrum whose amplitude is greater than the threshold as periodic components, otherwise mark them as non-periodic components; Step 3.4: Based on the labeling results, the original data in the multivariate data set is periodically divided to separate the periodic components from the non-periodic components.

7. A multivariate periodic grid modeling method based on power prediction according to claim 1 or 2, characterized in that: In step 4, a multivariate grid is formed using multivariate normalized features, and a grid model is constructed in the time and space dimensions, dividing time into continuous time points and space into discrete location points; Each grid point corresponds to a time and space position, as well as the meteorological data and power values ​​related to the position; the normalized meteorological data and actual power values ​​are filled into the corresponding grid points of the multivariate grid to form a complete multivariate periodic grid data set.

8. The multivariate periodic grid modeling method based on power prediction according to claim 7 is characterized by: The normalized feature data are clustered using the K-means clustering algorithm, and the data points and their cluster labels are used as nodes of the multivariate periodic grid, thereby dividing the data into different spatial regions or clusters to form a multivariate periodic grid structure.

9. The multivariate periodic grid modeling method based on power prediction according to claim 7 is characterized by: The normalized multivariate data set is divided into a training set and a test set. The training set is used to train a model including a prediction power model and a multivariate periodic grid model. The parameters of the model are adjusted through methods such as cross-validation. The trained model is used to predict the test set to obtain the predicted value.

10. A multivariate periodic grid modeling system based on power prediction, running a multivariate periodic grid modeling method based on power prediction according to any one of claims 1 to 9, characterized in that: include: Data entry module: used to enter the collected meteorological data related to power generation, measured and predicted power data, and pair them to form data tuples; Data preprocessing module: used to preprocess the input data to form a normalized multivariate data set; Periodic component identification module: used to periodically divide the normalized multivariate data set using Fourier transform to identify periodic components in the data; Model generation module: used to grid the normalized multivariate data set to obtain periodic models with different characteristics and realize multivariate periodic grid modeling.

Images