Method for predicting output of new energy station

Abstract

The invention relates to the technical field of data reasoning, and discloses a new energy station output prediction method, which comprises the following steps of: extracting environment characteristics in meteorological data, coding the environment characteristics into environment characteristic vectors, and combining inverter change conditions and component attenuation characteristics of a new energy station into equipment characteristic vectors; fusing the environment feature vector and the equipment feature vector into a dynamic feature vector of the new energy station; predicting an initial output value of the new energy station based on the dynamic feature vector; constructing a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station, and superposing the illumination intensity in the meteorological data into the spatial relationship matrix to obtain the spatial weight of the new energy station; when the new energy station is a hub station, strengthening the spatial weight according to the output fluctuation correlation between the hub station and the peripheral stations, and predicting the final prediction value of the new energy station based on the strengthened spatial weight; according to the invention, the accuracy and speed of new energy station output prediction can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data reasoning technology, and in particular to a method for predicting output of a new energy station. Background Art

[0002] In the field of renewable energy station output forecasting, existing technologies primarily rely on single-dimensional data features for modeling, such as considering only meteorological data or equipment operating data. This makes it difficult to fully capture the dynamic changes in renewable energy output. This inability to fully integrate the synergistic effects of environmental and equipment status characteristics results in insufficient adaptability of forecast models to complex scenarios, leading to low forecast accuracy.

[0003] At the same time, traditional forecasting methods lack effective modeling of spatial correlations. This is particularly true when dealing with the correlation between output fluctuations between hub stations and surrounding stations. This inability to accurately construct spatial weight relationships prevents the forecast process from fully utilizing the topological structure of the regional power grid. Furthermore, the lack of an iterative optimization mechanism prevents the model from adjusting eigenvectors in real time based on forecast deviations, further impacting forecast efficiency and accuracy. This makes it difficult to meet the forecast speed and accuracy requirements for renewable energy grid-connected dispatch. Summary of the Invention

[0004] The present invention provides a method for predicting the output of a new energy station, the main purpose of which is to solve the problem of low accuracy and rate of output prediction of a new energy station.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for predicting the output of a new energy station, comprising: S1. Extract environmental features from meteorological data, encode them into environmental feature vectors, and combine the inverter changes and component attenuation characteristics of the new energy station into a device feature vector; S2. Fusion of the environmental feature vector and the device feature vector into a dynamic feature vector of the new energy station; S3. Predicting the initial output value of the new energy station based on the dynamic feature vector; S4. Constructing a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station, superimposing the light intensity in the meteorological data into the spatial relationship matrix to obtain the spatial weight of the new energy station; S5. When the new energy station is a hub station, the spatial weight is strengthened according to the output fluctuation correlation between the hub station and the surrounding stations, and the final predicted value of the new energy station is predicted based on the strengthened spatial weight.

[0006] In a preferred embodiment, the environmental features extracted from the meteorological data are encoded into an environmental feature vector, and the inverter changes of the new energy station and the component attenuation characteristics are combined into a device feature vector, including: Performing Gaussian filtering on the temperature data, wind speed data, and irradiation data in the meteorological data to obtain an environmental feature vector of the new energy station; Fitting the component power attenuation curve during historical operation to the attenuation rate parameter of the new energy station; Adding the running time to the attenuation rate parameter to obtain a real-time attenuation coefficient of the energy station; The real-time attenuation coefficient and the change of the inverter of the new energy station are combined into the equipment feature vector of the energy station.

[0007] In a preferred embodiment, the step of fusing the environment feature vector and the device feature vector into a dynamic feature vector of the new energy station includes: Encoding the covariance between wind speed, light intensity and temperature in the environmental feature vector into an environmental covariance matrix; Encoding the covariance between the device eigenvector inverter efficiency and the component attenuation coefficient into a device covariance matrix; Determining a weight distribution ratio of the new energy station based on eigenvalue distributions of the environment covariance matrix and the equipment covariance matrix; The environmental feature vector and the device feature are fused based on the weight distribution ratio to obtain a dynamic feature vector of the new energy station.

[0008] In a preferred embodiment, the predicting the initial output value of the new energy station based on the dynamic feature vector includes: Extracting the time-dependent features of the dynamic feature vector; Performing a linear transformation on the time series dependency characteristics to obtain an output probability distribution of the new energy station; The maximum probability in the output probability distribution is used as the initial output value of the new energy station.

[0009] In a preferred embodiment, the constructing of a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station includes: Converting the longitude and latitude coordinates of the new energy station into a position coordinate set in a plane rectangular coordinate system; A spatial relationship matrix is ​​generated according to the spherical distance of the new energy station in the position coordinate set.

[0010] In a preferred embodiment, the step of superimposing the light intensity in the meteorological data into the spatial relationship matrix to obtain the spatial weight of the new energy station includes: Performing inverse scaling processing on the spatial relationship matrix; Sampling the grid number of the light intensity to the position coordinates of the new energy station to obtain a diagonal matrix of the light intensity of the new energy station; Superimposing the illumination intensity diagonal matrix onto the inversely scaled spatial relationship matrix to obtain an illumination enhancement matrix for the new energy station; The main diagonal elements of the illumination enhancement matrix are extracted, and the spatial weights of the new energy stations are generated based on the main diagonal elements.

[0011] In a preferred embodiment, when the new energy station is a hub station, the method includes: When the grid-connected capacity of the new energy station exceeds the regional threshold and is located at a key node of the power grid, it is marked as a hub station; With the hub station as the center, all new energy stations within a preset radius are selected as associated stations.

[0012] In a preferred embodiment, the strengthening of the spatial weight according to the output fluctuation correlation between the hub station and the surrounding stations includes: According to the output order of the hub station and the associated stations, the fluctuation correlation between the hub station and the associated stations is calculated, wherein the calculation formula of the fluctuation correlation is as follows: ; Where, is the fluctuation correlation, is the mathematical expectation, For the The output fluctuation intensity of each hub station, For the The output fluctuation intensity of the associated stations, For the Output time series data of hub stations, For the The output fluctuation intensity of the associated stations, For the The average output level of each hub station, For the The average output level of the associated stations; Arranging the fluctuation correlation into a fluctuation matrix of the new energy station according to the topological relationship between the hub station and the associated stations; Extracting the maximum eigenvalue in the fluctuation matrix as the eigenvector corresponding to the new energy station; The spatial weight is enhanced based on the eigenvector to obtain an enhanced spatial weight.

[0013] In a preferred embodiment, the method of predicting the final predicted value of the new energy station based on the enhanced spatial weight includes: Dividing the initial output value into subsequences according to the time dimension; performing convolution processing on the subsequence based on the enhanced spatial weight; The final predicted value of the new energy station is obtained by connecting the subsequence after residual convolution processing and the initial output value.

[0014] In a preferred embodiment, when the new energy station is a hub station, the spatial weight is strengthened according to the output fluctuation correlation between the hub station and the surrounding stations, and the final predicted value of the new energy station is predicted based on the strengthened spatial weight, which includes: When the deviation between the initial output value and the final predicted value exceeds a preset tolerance, returning to step S2 to update the dynamic feature vector; The final predicted value of the new energy station is re-predicted based on the updated dynamic feature vector.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This method constructs a dynamic feature vector by fusing environmental and device feature vectors. It then generates spatial weights based on the light intensity and spatial relationship matrix from meteorological data. This method comprehensively captures the factors influencing the output of new energy stations, thereby improving the accuracy of output forecasts. Furthermore, the method predicts initial output values ​​based on the dynamic feature vectors and strengthens spatial weights by correlating output fluctuations between hub stations and surrounding stations, enabling precise calculation of the final predicted value and effectively improving forecast accuracy.

[0016] 2. During the prediction process, when the deviation between the initial output value and the final predicted value exceeds the preset tolerance, the present invention can re-update the dynamic feature vector and make another prediction. This iterative optimization mechanism ensures the adaptability and accuracy of the model, further improves the accuracy and speed of the output prediction of the new energy station, and provides more reliable prediction support for the grid-connected scheduling of new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic flow chart of a method for predicting output of a new energy station provided by one embodiment of the present invention; The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] The embodiment of the present application provides a method for predicting the output of a new energy station. The execution subject of the method for predicting the output of a new energy station includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for predicting the output of a new energy station can be executed by software or hardware installed on a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 FIG. 1 is a flow chart of a method for predicting the output of a new energy station provided by an embodiment of the present invention. In this embodiment, the method for predicting the output of a new energy station includes: S1. Extract environmental features from meteorological data, encode them into environmental feature vectors, and combine the inverter changes and component attenuation characteristics of the new energy station into a device feature vector; In an embodiment of the present invention, the environmental features extracted from the meteorological data are encoded into an environmental feature vector, and the inverter change of the new energy station and the component attenuation characteristics are combined into a device feature vector, including: Performing Gaussian filtering on the temperature data, wind speed data, and irradiation data in the meteorological data to obtain an environmental feature vector of the new energy station; Fitting the component power attenuation curve during historical operation to the attenuation rate parameter of the new energy station; Adding the running time to the attenuation rate parameter to obtain a real-time attenuation coefficient of the energy station; The real-time attenuation coefficient and the change of the inverter of the new energy station are combined into the equipment feature vector of the energy station.

[0021] Specifically, temperature data, wind speed data and irradiation data are extracted from the meteorological data, and a Gaussian filtering method is used to process these data.

[0022] Furthermore, Gaussian filtering processing is based on the characteristics of the Gaussian function to construct a smoothing filter. During data processing, weighted averaging calculations are performed according to the values ​​of the neighboring data points around each data point and the weights determined by the Gaussian function, thereby obtaining new, smoother temperature data, wind speed data, and irradiation data. These processed data are combined together to form the environmental characteristic vector of the new energy station.

[0023] Furthermore, data on the changes in component power over time during the historical operation of the new energy station are collected, and the curve fitting method is used to fit these data into a curve. This curve reflects the attenuation trend of the component power. Relevant information that can describe the degree of component power attenuation is extracted from this fitting curve. This information is the attenuation rate parameters of the new energy station.

[0024] Furthermore, the current operating time of the new energy station is obtained, and this operating time information is superimposed with the previously obtained attenuation rate parameter. Specifically, the influencing factors corresponding to the operating time are incorporated into the attenuation rate parameter to obtain a new value that can reflect the power attenuation of the components at the current moment. This value is the real-time attenuation coefficient of the energy station.

[0025] Furthermore, various change information of the new energy station inverter during operation is obtained, and the real-time attenuation coefficient obtained previously is combined with the inverter change information, and they are organized into an ordered set. This set contains key information reflecting the operating status of the energy station equipment, that is, the equipment characteristic vector of the energy station.

[0026] In general, the present invention can comprehensively capture the impact of environmental factors such as temperature, wind speed, and radiation on the output of new energy stations by extracting environmental features from meteorological data and encoding them into environmental feature vectors. At the same time, it uses Gaussian filtering to remove data noise, ensuring the accuracy and stability of environmental features, and providing reliable environmental dimension data support for subsequent predictions.

[0027] In general, when the inverter changes and component attenuation characteristics are combined into a device feature vector, the attenuation rate parameters are generated by fitting the historical component power attenuation curve, and the real-time attenuation coefficient is obtained by superimposing the operating time. This can dynamically reflect the equipment aging status and real-time operating efficiency, so that the device feature vector can accurately characterize the hardware operating status of the new energy station, thereby laying the foundation for constructing a dynamic feature vector that includes both environmental and equipment dimensions, and effectively improving the adaptability and accuracy of the output prediction model to dynamic changes in equipment.

[0028] S2. Fusion of the environmental feature vector and the device feature vector into a dynamic feature vector of the new energy station; In an embodiment of the present invention, the step of fusing the environment feature vector and the device feature vector into a dynamic feature vector of the new energy station includes: Encoding the covariance between wind speed, light intensity and temperature in the environmental feature vector into an environmental covariance matrix; Encoding the covariance between the device eigenvector inverter efficiency and the component attenuation coefficient into a device covariance matrix; Determining a weight distribution ratio of the new energy station based on eigenvalue distributions of the environment covariance matrix and the equipment covariance matrix; The environmental feature vector and the device feature are fused based on the weight distribution ratio to obtain a dynamic feature vector of the new energy station.

[0029] Specifically, the obtained environmental feature vector includes wind speed, light intensity and temperature data.

[0030] Furthermore, covariance is used to measure the mutual change relationship between these data. By calculating the covariance between wind speed and light intensity, wind speed and temperature, and light intensity and temperature, and then arranging these covariance values ​​in a specific row and column order, a matrix is ​​formed. This matrix is ​​the environmental covariance matrix, which reflects the degree of correlation between the data in the environmental feature vector.

[0031] Furthermore, the device feature vector includes inverter efficiency and component attenuation coefficient data.

[0032] Furthermore, the covariance calculation method is also used to measure the mutual change relationship between the inverter efficiency and the component attenuation coefficient. After calculating the covariance between the two, this covariance value is constructed into a matrix form with only one element (or a numerical value) according to the matrix rules, namely the device covariance matrix. This matrix reflects the correlation between the two data in the device eigenvector.

[0033] Furthermore, after obtaining the environment covariance matrix and the equipment covariance matrix, the eigenvalue distribution of the two matrices is analyzed.

[0034] Furthermore, the eigenvalues ​​reflect the importance of the data features represented by the matrix. The larger parts of the eigenvalues ​​of the environmental covariance matrix that can better reflect the important features of the environmental data, as well as the important parts of the eigenvalues ​​of the equipment covariance matrix, are found. According to the size and distribution characteristics of these eigenvalues, the importance of the environmental covariance matrix and the equipment covariance matrix in the whole are determined, and then converted into the weight distribution ratio of the new energy sites, that is, the proportion of environmental-related factors and equipment-related factors in the subsequent analysis is determined.

[0035] Furthermore, after the weight distribution ratio is determined, the environment feature vector and the device feature vector are fused according to their respective corresponding weights.

[0036] Furthermore, each data in the environmental characteristic vector and the equipment characteristic vector is weighted respectively, and then the weighted environmental characteristic vector data and the equipment characteristic vector data are added correspondingly to obtain a new set of data. The vector composed of this new set of data is the dynamic characteristic vector of the new energy station, which comprehensively considers the environmental and equipment factors and can reflect the dynamic operation status of the new energy station.

[0037] In general, when the present invention fuses the environmental feature vector and the device feature vector into a dynamic feature vector, it can accurately capture the dynamic correlation between environmental factors and device status by encoding the environmental covariance matrix and the device covariance matrix.

[0038] In general, the weight distribution ratio is determined based on the eigenvalue distribution of the two types of covariance matrices, and the fusion weight can be adaptively adjusted according to the feature importance, so that the dynamic eigenvector can reflect both the immediate impact of environmental changes on the output and the attenuation effect of long-term conditions such as equipment aging on the output.

[0039] In general, the dynamic feature vector constructed by this fusion mechanism realizes the organic integration of environmental dimension and device dimension data, forming a composite feature representation that includes spatiotemporal dynamic characteristics and device health status.

[0040] In general, compared with single-dimensional feature modeling, this method can more comprehensively characterize the factors affecting the output of new energy stations, provide richer feature inputs for subsequent initial output value predictions, effectively improve the adaptability of the prediction model to complex operating scenarios, and ensure the accuracy and dynamic response capabilities of output predictions from the feature fusion level.

[0041] S3. Predicting the initial output value of the new energy station based on the dynamic feature vector; In an embodiment of the present invention, predicting the initial output value of the new energy station based on the dynamic feature vector includes: Extracting the time-dependent features of the dynamic feature vector; Performing a linear transformation on the time series dependency characteristics to obtain an output probability distribution of the new energy station; The maximum probability in the output probability distribution is used as the initial output value of the new energy station.

[0042] Specifically, the dynamic feature vectors are observed, which contain the environment and equipment-related information of the new energy station at different time points.

[0043] Furthermore, by analyzing the changing patterns and mutual correlations of this information in time series, the temporal dependence characteristics can be found.

[0044] Furthermore, we specifically check the changing trends of data in the dynamic feature vectors of adjacent time points, as well as the influence relationship between data at different time points, and extract these features that reflect the temporal relationship to form a time-dependent feature set that can reflect the temporal changes of new energy station data.

[0045] Furthermore, the extracted temporal dependency features are processed by linear transformation.

[0046] Furthermore, linear transformation operates on each data in the time-dependent feature according to certain rules, so that the relationship between the data changes.

[0047] Furthermore, the specific operation is to assign a specific weight to each time-dependent feature data, multiply the data by the weight, and then add up all the product results to obtain a new set of data.

[0048] Furthermore, through this linear transformation, the time-series dependent characteristics are converted into a form that can reflect the possibility of different output conditions of new energy stations, that is, the output probability distribution of new energy stations is obtained. This distribution shows the probability of new energy stations appearing at different output values ​​in the future.

[0049] Furthermore, in the obtained probability distribution of the output of the new energy station, the probability value corresponding to each output value is checked.

[0050] Furthermore, the one with the largest probability value is found out, and the output value corresponding to the maximum probability is determined as the initial output value of the new energy station.

[0051] Furthermore, this initial output value is the most likely output predicted after a comprehensive analysis of multiple factors such as the new energy station environment and equipment, providing basic data for subsequent further analysis and prediction of the new energy station output.

[0052] In summary, this method, when predicting the initial output value of a new energy station based on a dynamic feature vector, extracts the time-dependent characteristics of the dynamic feature vector, capturing temporal variations in environmental and equipment status, such as daily temperature variations and equipment degradation trends. Linearly transforming the time-dependent characteristics yields an output probability distribution, with the maximum probability value used as the initial output value. This probabilistic prediction method quantifies the likelihood of different output scenarios, avoiding the limitations of single-value predictions and improving the reliability of the prediction results.

[0053] In general, this forecasting process is directly based on dynamic feature vectors that integrate dual-dimensional characteristics of the environment and equipment. It can comprehensively reflect the immediate impact of fluctuating meteorological conditions and equipment operating status on output. Compared with forecasting methods that rely solely on a single data dimension, it can more accurately depict the dynamic changes in the output of renewable energy stations. Furthermore, by determining the initial output value through a probability distribution mechanism, the forecast results are statistically reasonable, providing a more solid initial forecast foundation for subsequent optimization using spatial weights. This ensures the accuracy and scientific nature of the initial output forecast through time series feature mining and probabilistic modeling.

[0054] S4. Constructing a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station, superimposing the light intensity in the meteorological data into the spatial relationship matrix to obtain the spatial weight of the new energy station; In an embodiment of the present invention, the step of constructing a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station includes: Converting the longitude and latitude coordinates of the new energy station into a position coordinate set in a plane rectangular coordinate system; A spatial relationship matrix is ​​generated according to the spherical distance of the new energy station in the position coordinate set.

[0055] The step of superimposing the light intensity in the meteorological data onto the spatial relationship matrix to obtain the spatial weight of the new energy station includes: Performing inverse scaling processing on the spatial relationship matrix; Sampling the grid number of the light intensity to the position coordinates of the new energy station to obtain a diagonal matrix of the light intensity of the new energy station; Superimposing the illumination intensity diagonal matrix onto the inversely scaled spatial relationship matrix to obtain an illumination enhancement matrix for the new energy station; The main diagonal elements of the illumination enhancement matrix are extracted, and the spatial weights of the new energy stations are generated based on the main diagonal elements.

[0056] Specifically, the longitude and latitude coordinates of the new energy stations are obtained and converted into position coordinates in a plane rectangular coordinate system using a coordinate conversion method. Specifically, the longitude and latitude points on the earth's surface are mapped onto a plane through a specific conversion method to obtain the corresponding x and y coordinate values. These plane coordinate values ​​of all new energy stations are collected to form a set, which is the position coordinate set in the plane rectangular coordinate system.

[0057] Furthermore, based on the obtained set of location coordinates, the spherical distances between each renewable energy station are calculated. Spherical distance refers to the shortest distance between two points on the Earth's surface along a sphere. After calculating the spherical distances between all pairs of renewable energy stations, these distance values ​​are arranged into a matrix according to certain rules.

[0058] Furthermore, the rows and columns of the matrix correspond to different new energy stations, and each element in the matrix corresponds to the spherical distance between two new energy stations. The matrix formed in this way is a spatial relationship matrix, which shows the spatial position relationship between each new energy station.

[0059] Specifically, the spatial relationship matrix obtained previously is obtained, and each element in the matrix is ​​inversely scaled.

[0060] Furthermore, for each distance value in the spatial relationship matrix, find a suitable value to divide it, so that the processed value decreases as the original distance value increases, and increases as the original distance value decreases, thereby obtaining the spatial relationship matrix after inverse scaling processing. This processing can highlight the relationship between new energy stations that are close to each other.

[0061] Furthermore, it is known that light intensity is distributed in a grid form, and grid data of light intensity are obtained, and these grid data are sampled according to the position coordinates of the new energy site in a plane rectangular coordinate system.

[0062] Furthermore, find the grid corresponding to the location coordinates of each new energy station, extract the light intensity value at the grid, and then arrange these light intensity values ​​in the order of the new energy stations to form a matrix with only non-zero elements on the main diagonal and zero elements in the rest of the positions. This matrix is ​​the light intensity diagonal matrix of the new energy station, which reflects the light intensity corresponding to each new energy station.

[0063] Furthermore, the light intensity diagonal matrix and the inversely scaled spatial relationship matrix are superimposed. Specifically, the elements at corresponding positions in the two matrices are added together to obtain a new matrix, which is the light enhancement matrix for the new energy station. Through this superposition, light intensity information is integrated into the spatial relationship, highlighting the impact of light factors on the spatial relationship of the new energy station.

[0064] Furthermore, the illumination enhancement matrix is ​​observed and its main diagonal elements are extracted. The main diagonal is the element on the line from the upper left corner to the lower right corner of the matrix.

[0065] Furthermore, these extracted main diagonal elements are processed according to certain rules, such as sorting according to the size of the values, weighting according to the importance, etc., and finally a set of values ​​that can reflect the importance of the spatial location of the new energy station is obtained. This set of values ​​is the spatial weight of the new energy station, which comprehensively considers the impact of spatial distance and light intensity on the new energy station.

[0066] In general, when constructing a spatial relationship matrix based on the longitude and latitude coordinates of renewable energy stations, the present invention converts the longitude and latitude into plane rectangular coordinates and generates a matrix based on spherical distance. This can accurately characterize the spatial position correlation characteristics between renewable energy stations, laying a geometric relationship foundation for the subsequent analysis of the spatial correlation of the output of stations in the regional power grid.

[0067] In general, this matrix construction method based on actual geographic coordinates can effectively reflect the transmission law of meteorological conditions caused by geographical differences between stations, making the spatial relationship matrix reasonable in a physical sense.

[0068] In general, when superimposing light intensity onto the spatial relationship matrix, the weight influence of nearby stations is enhanced through inverse scaling processing. Combining light intensity raster sampling to generate a diagonal matrix and superimposing it can deeply integrate light, a key environmental factor, with spatial position relationships, forming a spatial weight that includes the dual dimensions of geographic distance and light intensity.

[0069] In general, this weight not only reflects the spatial topological relationship between stations, but also incorporates the direct impact of light intensity on output, so that the spatial weight can more comprehensively characterize the spatial correlation characteristics of the output of new energy stations, and provide more accurate weight support for strengthening the spatial correlation calculation in the subsequent hub station output forecast, thereby improving the accuracy of output forecast and regional coordination from the spatial dimension.

[0070] S5. When the new energy station is a hub station, the spatial weight is strengthened according to the output fluctuation correlation between the hub station and the surrounding stations, and the final predicted value of the new energy station is predicted based on the strengthened spatial weight.

[0071] In an embodiment of the present invention, when the new energy station is a hub station, the steps include: When the grid-connected capacity of the new energy station exceeds the regional threshold and is located at a key node of the power grid, it is marked as a hub station; With the hub station as the center, all new energy stations within a preset radius are selected as associated stations.

[0072] The step of strengthening the spatial weight according to the output fluctuation correlation between the hub station and surrounding stations includes: According to the output order of the hub station and the associated stations, the fluctuation correlation between the hub station and the associated stations is calculated, wherein the calculation formula of the fluctuation correlation is as follows: ; Where, is the fluctuation correlation, is the mathematical expectation, For the The output fluctuation intensity of each hub station, For the The output fluctuation intensity of the associated stations, For the Output time series data of hub stations, For the The output fluctuation intensity of the associated stations, For the The average output level of each hub station, For the The average output level of the associated stations; Arranging the fluctuation correlation into a fluctuation matrix of the new energy station according to the topological relationship between the hub station and the associated stations; Extracting the maximum eigenvalue in the fluctuation matrix as the eigenvector corresponding to the new energy station; The spatial weight is enhanced based on the eigenvector to obtain an enhanced spatial weight.

[0073] The final predicted value of the new energy station is predicted based on the enhanced spatial weight, including: Dividing the initial output value into subsequences according to the time dimension; performing convolution processing on the subsequence based on the enhanced spatial weight; The final predicted value of the new energy station is obtained by connecting the subsequence after residual convolution processing and the initial output value.

[0074] When the new energy station is a hub station, the spatial weight is enhanced according to the output fluctuation correlation between the hub station and surrounding stations, and a final predicted value of the new energy station is predicted based on the enhanced spatial weight, including: When the deviation between the initial output value and the final predicted value exceeds a preset tolerance, returning to step S2 to update the dynamic feature vector; The final predicted value of the new energy station is re-predicted based on the updated dynamic feature vector.

[0075] Specifically, the grid-connected capacity data of the new energy station and the node information of the station in the power grid are obtained, and the specific range of the regional threshold and the key nodes of the power grid are determined.

[0076] Furthermore, the grid-connected capacity of each new energy station is compared with the regional threshold. If the grid-connected capacity of a new energy station is greater than the regional threshold, it is checked whether the station is within the range of the predetermined key nodes of the power grid.

[0077] Furthermore, when the grid-connected capacity of a new energy station exceeds the regional threshold and is located at a key node in the power grid, the station will be given a specific mark and marked as a hub station to distinguish the new energy stations that play an important role in the power grid.

[0078] Furthermore, after determining the hub station, the specific value of the preset radius should be clarified.

[0079] Furthermore, with each hub station as the center, a circle is drawn according to a preset radius on the map or power grid layout, and all new energy stations within the circle are screened out. Regardless of the size and type of these stations, as long as they are within this circular area, they will be selected. These selected new energy stations are uniformly regarded as associated stations, thereby determining a set of new energy stations that have a close spatial connection with the hub station.

[0080] Specifically, the output data of hub stations and related stations at different time points are collected, and these output data are sorted in chronological order.

[0081] Furthermore, by analyzing the changes in the output data of the hub station and each associated station over time, the similarity of their output fluctuations can be determined.

[0082] Furthermore, the time points and the amplitude of the increase or decrease in the output of the two are compared, so as to calculate the fluctuation correlation between the hub station and each associated station, and obtain a series of numerical values ​​reflecting the degree of their fluctuation correlation.

[0083] Furthermore, the topological relationship between hub stations and associated stations should be clarified, that is, their connection method and layout structure in the power grid.

[0084] Furthermore, according to this topological relationship, the fluctuation correlation values ​​calculated previously are arranged.

[0085] Furthermore, the fluctuation correlation values ​​related to each station are filled into a matrix according to the connection order and correspondence of the stations in the topological structure, so that the rows and columns of the matrix correspond to different stations respectively. The elements in the matrix correspond to the fluctuation correlation between the two stations. The matrix formed in this way is the fluctuation matrix of the new energy station.

[0086] Furthermore, the fluctuation matrix is ​​analyzed to find the largest eigenvalue.

[0087] Furthermore, eigenvalues ​​reflect the importance of the data features represented by the matrix. In the fluctuation matrix, the eigenvector corresponding to the largest eigenvalue contains the most important information about the output fluctuations of the renewable energy station. Using a specific method, the data set corresponding to the largest eigenvalue is extracted from the fluctuation matrix. This data set is the eigenvector corresponding to the renewable energy station.

[0088] Furthermore, the previously obtained spatial weights and the newly generated eigenvectors are obtained, and the information in the eigenvectors is integrated into the spatial weights.

[0089] Furthermore, according to the important characteristics of the output fluctuation of the new energy station reflected by the eigenvector, each value in the spatial weight is adjusted.

[0090] For example, if the eigenvector shows that the fluctuations of certain stations have a greater impact on the overall situation, the corresponding values ​​of these stations in the spatial weight will be increased accordingly. In this way, the spatial weight is strengthened, and the strengthened spatial weight is finally obtained, which can more accurately reflect the importance of the new energy station under comprehensive factors.

[0091] Specifically, the first The output time series data of each hub station is obtained by collecting the actual output values ​​of the hub station at different time points, and these values ​​arranged in chronological order are used as the output time series data of the hub station; The output fluctuation intensity of each associated station is also recorded by recording the output values ​​of the associated station at each time point, analyzing the changes of these values ​​over time, and obtaining the degree of output fluctuation; The average output level of a hub station is obtained by adding up all the output time series data of the hub station and dividing it by the total number of data; The average output level of each associated station is also calculated by summing up all the output data of the associated station and dividing it by the number of data. The output fluctuation intensity of each hub station is determined by observing the degree to which the output time series data of the station deviates from the average output level. The greater the deviation, the greater the fluctuation intensity. The output fluctuation intensity of an associated station is an indicator to measure the degree of dispersion of the output data of the associated station. It is obtained by calculating the square of the difference between each output data and the average output level, averaging these square values, and then taking the square root.

[0092] Furthermore, the formula is used to calculate the fluctuation correlation between the hub station and the associated stations, and to determine the degree of correlation between their output fluctuations by analyzing the deviation of the output data of the two stations from their respective average output levels, as well as the degree of dispersion of the output data.

[0093] Furthermore, the product of the difference between the output data of each of the two stations and the average output level is first calculated, the mathematical expectation of these products is obtained, and then divided by the measurement value of the degree of dispersion of the output data of each of the two stations, so as to obtain a value reflecting the correlation between the fluctuations of the two.

[0094] Furthermore, when the output fluctuations of the hub station and the associated stations show similar changing trends, that is, the deviations of their output data from the average output level are similar and the degree of dispersion is also similar, the calculated fluctuation correlation value will be larger, indicating that the fluctuation correlation between the two is strong; if the output fluctuations of the two stations have no obvious correlation, when the output of one station increases, the output of the other station decreases, or the degree of dispersion of their output data is very different, then the calculated fluctuation correlation value will be smaller, which means that the fluctuation correlation between the two is weak.

[0095] Specifically, the initial output value data of the new energy station is obtained, and these data are recorded at different time points.

[0096] Furthermore, these initial output value data are divided according to chronological order, and the initial output values ​​within a continuous period of time are grouped together to form a subsequence. In this way, the entire initial output value data is split into multiple subsequences according to the time dimension, and each subsequence contains the output information within a specific period of time.

[0097] Furthermore, the obtained enhanced spatial weights are used to perform convolution processing on the divided subsequences.

[0098] Furthermore, the specific operation is to regard the enhanced spatial weight as a "filter" and use this "filter" to perform a sliding operation on each subsequence.

[0099] Furthermore, during the sliding process, the subsequence data covered by the "filter" is multiplied by the corresponding weight value of the "filter", and then these products are added together to obtain a new value. As the "filter" slides over the subsequence, each slide generates a new value. These new values ​​are combined to form the subsequence after the convolution process. In this way, the information contained in the spatial weight is integrated into the subsequence data.

[0100] Furthermore, a residual connection operation is performed on the subsequence after convolution processing and the original initial output value.

[0101] Furthermore, the data at each time point in the convolved subsequence is added to the data at the corresponding time point in the initial output value to generate a new set of data. This new set of data combines the characteristics of the original data and the convolved data, and can more comprehensively reflect the changes in the output of the new energy station. This new set of data is used as the final predicted value of the new energy station, and is used to predict the future output of the new energy station.

[0102] In general, when the new energy station serves as a hub station, the present invention strengthens the spatial weight by exploring the correlation between its output fluctuations and those of surrounding stations, and can accurately capture the coordinated change pattern of the station output in the regional power grid.

[0103] In general, hub stations are identified based on grid-connected capacity and grid node location, and associated stations are selected with them as the center. The degree of coordination of output fluctuations of each station is quantified through the fluctuation correlation formula. This correlation is converted into the eigenvector of the fluctuation matrix and then the spatial weight is strengthened. The spatial weight not only includes geographical distance and light intensity information, but also incorporates dynamic output correlation characteristics, thereby constructing a spatial correlation model that is more in line with actual operation scenarios and providing more scientific weight support for the final prediction.

[0104] In general, this enhancement mechanism achieves in-depth modeling of the output fluctuation law around the hub station through the dual constraints of topological relationship and fluctuation correlation, so that the prediction process can fully utilize the spatial coupling characteristics of the regional power grid.

[0105] In general, when predicting the final value, by dividing the initial output value into subsequences and performing convolution processing with the enhanced spatial weights, and then fusing the initial prediction with spatial correlation characteristics through residual connection, the prediction model's ability to capture hub station output fluctuations can be effectively improved. In particular, under complex meteorological conditions, it can more accurately reflect the coordinated change trend of regional output, thereby significantly improving the accuracy of new energy station output prediction and the efficiency of regional coordinated prediction.

[0106] In the several embodiments provided by the present invention, it should be understood that the disclosed methods can be implemented in other ways.

[0107] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0108] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method and technology of using digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to achieve optimal results.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting the output of a new energy station, characterized in that: The method comprises: S1. Extract environmental features from meteorological data, encode them into environmental feature vectors, and combine the inverter changes and component attenuation characteristics of the new energy station into a device feature vector; S2. Fusion of the environmental feature vector and the device feature vector into a dynamic feature vector of the new energy station; S3. Predicting the initial output value of the new energy station based on the dynamic feature vector; S4. Constructing a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station, superimposing the light intensity in the meteorological data into the spatial relationship matrix to obtain the spatial weight of the new energy station; S5. When the new energy station is a hub station, the spatial weight is strengthened according to the output fluctuation correlation between the hub station and the surrounding stations, and the final predicted value of the new energy station is predicted based on the strengthened spatial weight.

2. The method for predicting the output of a new energy station according to claim 1, wherein: The environmental features extracted from the meteorological data are encoded into an environmental feature vector, and the inverter changes and component attenuation characteristics of the new energy station are combined into a device feature vector, including: Performing Gaussian filtering on the temperature data, wind speed data, and irradiation data in the meteorological data to obtain an environmental feature vector of the new energy station; Fitting the component power attenuation curve during historical operation to the attenuation rate parameter of the new energy station; Adding the running time to the attenuation rate parameter to obtain a real-time attenuation coefficient of the energy station; The real-time attenuation coefficient and the change of the inverter of the new energy station are combined into the equipment feature vector of the energy station.

3. The method for predicting the output of a new energy station according to claim 1, wherein: The fusing of the environment feature vector and the equipment feature vector into the dynamic feature vector of the new energy station includes: Encoding the covariance between wind speed, light intensity and temperature in the environmental feature vector into an environmental covariance matrix; Encoding the covariance between the device eigenvector inverter efficiency and the component attenuation coefficient into a device covariance matrix; Determining a weight distribution ratio of the new energy station based on eigenvalue distributions of the environment covariance matrix and the equipment covariance matrix; The environmental feature vector and the device feature are fused based on the weight distribution ratio to obtain a dynamic feature vector of the new energy station.

4. The method for predicting the output of a new energy station according to claim 1, wherein: The predicting the initial output value of the new energy station based on the dynamic feature vector includes: Extracting the time-dependent features of the dynamic feature vector; Performing a linear transformation on the time series dependency characteristics to obtain an output probability distribution of the new energy station; The maximum probability in the output probability distribution is used as the initial output value of the new energy station.

5. The method for predicting the output of a new energy station according to claim 1, wherein: The constructing of a spatial relationship matrix based on the latitude and longitude coordinates of the new energy station includes: Converting the longitude and latitude coordinates of the new energy station into a position coordinate set in a plane rectangular coordinate system; A spatial relationship matrix is ​​generated according to the spherical distance of the new energy station in the position coordinate set.

6. The method for predicting the output of a new energy station according to claim 1, wherein: The step of superimposing the light intensity in the meteorological data onto the spatial relationship matrix to obtain the spatial weight of the new energy station includes: Performing inverse scaling processing on the spatial relationship matrix; Sampling the grid number of the light intensity to the position coordinates of the new energy station to obtain a diagonal matrix of the light intensity of the new energy station; Superimposing the illumination intensity diagonal matrix onto the inversely scaled spatial relationship matrix to obtain an illumination enhancement matrix for the new energy station; The main diagonal elements of the illumination enhancement matrix are extracted, and the spatial weights of the new energy stations are generated based on the main diagonal elements.

7. The method for predicting the output of a new energy station according to claim 1, wherein: When the new energy station is a hub station, it includes: When the grid-connected capacity of the new energy station exceeds the regional threshold and is located at a key node of the power grid, it is marked as a hub station; With the hub station as the center, all new energy stations within a preset radius are selected as associated stations.

8. The method for predicting the output of a new energy station according to claim 7, wherein: The step of strengthening the spatial weight according to the output fluctuation correlation between the hub station and surrounding stations includes: According to the output order of the hub station and the associated stations, the fluctuation correlation between the hub station and the associated stations is calculated, wherein the calculation formula of the fluctuation correlation is as follows: ； Where, is the fluctuation correlation, is the mathematical expectation, For the The output fluctuation intensity of each hub station, For the The output fluctuation intensity of the associated stations, For the Output time series data of hub stations, For the The output fluctuation intensity of the associated stations, For the The average output level of each hub station, For the The average output level of the associated stations; Arranging the fluctuation correlation into a fluctuation matrix of the new energy station according to the topological relationship between the hub station and the associated stations; Extracting the maximum eigenvalue in the fluctuation matrix as the eigenvector corresponding to the new energy station; The spatial weight is enhanced based on the eigenvector to obtain an enhanced spatial weight.

9. The method for predicting the output of a new energy station according to claim 8, characterized in that: The final predicted value of the new energy station is predicted based on the enhanced spatial weight, including: Dividing the initial output value into subsequences according to the time dimension; performing convolution processing on the subsequence based on the enhanced spatial weight; The final predicted value of the new energy station is obtained by connecting the subsequence after residual convolution processing and the initial output value.

10. The method for predicting the output of a new energy station according to claim 1, wherein: When the new energy station is a hub station, the spatial weight is enhanced according to the output fluctuation correlation between the hub station and surrounding stations, and a final predicted value of the new energy station is predicted based on the enhanced spatial weight, including: When the deviation between the initial output value and the final predicted value exceeds a preset tolerance, returning to step S2 to update the dynamic feature vector; The final predicted value of the new energy station is re-predicted based on the updated dynamic feature vector.

Images