A method for predicting long-term new energy power in a region

Abstract

The application relates to the technical field of data reasoning, and discloses a regional medium and long-term new energy power prediction method, which comprises the following steps: acquiring historical new energy output time series data and corresponding meteorological element data of a target region to generate a regional basic data set; analyzing the space-time coupling characteristics in the regional basic data set and extracting a cross-regional new energy output collaborative fluctuation mode in the space-time coupling characteristics; constructing an output characteristic transfer matrix of the target region based on the cross-regional new energy output collaborative fluctuation mode; fusing the output characteristic transfer matrix and climate mode prediction data to obtain a medium and long-term meteorological-output mapping relationship of the target region; and outputting a new energy power prediction result of the target region according to the medium and long-term meteorological-output mapping relationship; and the application can improve the accuracy of medium and long-term new energy power prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a regional medium and long-term new energy power prediction method. BACKGROUND

[0002] In the field of regional medium and long-term new energy power prediction, existing technologies are often limited to modeling based on historical data of a single region, failing to fully explore the internal correlation of new energy output among different regions. Due to the neglect of cross-regional collaborative fluctuation characteristics, the coupling law of new energy output in the time and space dimensions is not well grasped, making it difficult to accurately reflect the mutual influence of output changes among regions at medium and long-term time scales, resulting in large deviations in the prediction results.

[0003] At the same time, existing methods often stop at the simple association level when fusing meteorological elements and new energy output characteristics, failing to effectively combine the persistent climate characteristics and dynamic output mode. This leads to a lack of depth adaptation to the medium and long-term climate evolution law in the constructed meteorological-output mapping relationship, and the lagging and cumulative effects of meteorological element changes on new energy output cannot be accurately captured, further reducing the reliability of medium and long-term prediction. SUMMARY

[0004] The present application provides a regional medium and long-term new energy power prediction method, which mainly aims to solve the problem of low accuracy of medium and long-term new energy power prediction.

[0005] To achieve the above purpose, the present application provides a regional medium and long-term new energy power prediction method, which comprises:

[0006] S1, obtaining historical new energy output time series data and corresponding meteorological element data of a target region, and generating a regional basic data set;

[0007] S2, analyzing the spatio-temporal coupling characteristics in the regional basic data set, and extracting the cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling characteristics;

[0008] S3, based on the cross-regional new energy output collaborative fluctuation mode, constructing an output characteristic transfer matrix of the target region;

[0009] S4, fusing the output characteristic transfer matrix and climate mode prediction data to obtain a medium and long-term meteorological-output mapping relationship of the target region;

[0010] S5, according to the medium and long-term meteorological-output mapping relationship, outputting the new energy power prediction result of the target region.

[0011] In a preferred embodiment, the historical new energy output time series data and corresponding meteorological element data of the target area are acquired to generate a regional basic data set, which includes:

[0012] The wind farm power output sequence and photovoltaic power station power output sequence of the target area and associated areas are collected;

[0013] The wind farm power output sequence and photovoltaic power station power output sequence are time-aligned to generate a standardized output time series of the target area;

[0014] The standardized output time series and the meteorological element monitoring data of the corresponding site are associated to form a regional basic data set of the target area.

[0015] In a preferred embodiment, the meteorological element monitoring data includes:

[0016] Atmospheric boundary layer height data, surface radiation flux data, and near-surface wind vector data monitored by meteorological stations are extracted;

[0017] The atmospheric boundary layer height data, surface radiation flux data, and near-surface wind vector data are spatially interpolated to generate a meteorological element grid field covering the target area, and the meteorological element grid field is used as the meteorological element monitoring data of the meteorological station.

[0018] In a preferred embodiment, the analysis of the spatio-temporal coupling characteristics in the regional basic data set and the extraction of the cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling characteristics include:

[0019] A new energy output spatio-temporal tensor model of the target area is constructed based on the regional basic data set;

[0020] High-order singular value decomposition is performed on the spatio-temporal tensor model to obtain a spatio-temporal coupling characteristic matrix of the target area;

[0021] From the spatio-temporal coupling characteristic matrix, a feature vector with a variance contribution rate exceeding a preset threshold is selected as the cross-regional new energy output collaborative fluctuation mode.

[0022] In a preferred embodiment, the selection of the feature vector with a variance contribution rate exceeding a preset threshold from the spatio-temporal coupling characteristic matrix as the cross-regional new energy output collaborative fluctuation mode includes:

[0023] A modal energy distribution function of the target area is constructed according to the spatio-temporal coherence coefficient of the feature vector in the spatio-temporal coupling characteristic matrix;

[0024] The selection boundary of the feature vector is determined by the gradient change of the modal energy distribution function.

[0025] The feature vector within the screening boundary is taken as the cross-regional new energy output coordinated fluctuation mode.

[0026] In a preferred embodiment, after analyzing the spatio-temporal coupling features in the regional basic data set and extracting the cross-regional new energy output coordinated fluctuation mode in the spatio-temporal coupling features, the method further comprises:

[0027] Mapping the feature vector to an atmospheric circulation pattern field;

[0028] Verifying the phase correlation relationship between the feature vector and an atmospheric activity center;

[0029] Splicing the feature vector with a strong correlation verification structure into a spatio-temporal correlation matrix.

[0030] In a preferred embodiment, the construction of the output characteristic transfer matrix of the target region based on the cross-regional new energy output coordinated fluctuation mode comprises:

[0031] Identifying a persistent climate feature in the meteorological element monitoring data;

[0032] Dividing historical data into typical climate scenarios according to the persistent climate feature, and generating a climate feature weight of the target region;

[0033] Based on the climate feature weight and the spatio-temporal correlation matrix, constructing an output characteristic transfer matrix of the target region, wherein the expression of the output characteristic transfer matrix is as follows:

[0034]

[0035] In the formula, is the output characteristic transfer matrix, is a representation of the change of the output characteristic transfer matrix over time, is the spatio-temporal correlation matrix, is the climate feature weight, is a Hadamard product operator symbol.

[0036] In a preferred embodiment, the fusion of the output characteristic transfer matrix and the climate mode prediction data to obtain the medium and long-term meteorological-output mapping relationship of the target region comprises:

[0037] Decomposing the output characteristic transfer matrix into a spatial weight matrix and a time-varying mode matrix;

[0038] Performing tensor convolution operation on the climate mode prediction data and the spatial weight matrix to generate a climate feature enhancement field of the target region; ​

[0039] time-scale transform the climate feature enhancement field through the time-varying modal matrix to generate a spatiotemporal modulation feature of the target region;

[0040] perform feature selection on the spatiotemporal modulation feature to extract a key meteorological-power correlation feature of the target region;

[0041] construct a medium-and-long-term meteorological-power mapping relationship of the target region based on the key meteorological-power correlation feature.

[0042] In a preferred embodiment, the step of constructing a medium-and-long-term meteorological-power mapping relationship of the target region based on the key meteorological-power correlation feature comprises:

[0043] dividing the key meteorological-power correlation feature into multi-category features according to meteorological element types and power response characteristics;

[0044] evolving the multi-category features into a causal rule chain according to a physical evolution path of the cross-regional new energy power collaborative fluctuation mode;

[0045] pooling the causal rule chain as the medium-and-long-term meteorological-power mapping relationship of the target region.

[0046] In a preferred embodiment, the step of outputting a new energy power prediction result of the target region according to the medium-and-long-term meteorological-power mapping relationship comprises:

[0047] matching a power relationship in the medium-and-long-term meteorological-power mapping relationship according to a key climate driving factor of the target region;

[0048] outputting the power relationship and a causal rule chain of the key climate driving factor as a visualized new energy power prediction result.

[0049] Compared with the prior art, the present application has the following beneficial effects:

[0050] 1. The present application effectively captures the spatiotemporal coupling rules of new energy power between different regions by extracting a cross-regional new energy power collaborative fluctuation mode and constructing a power characteristic migration matrix. This deep mining of cross-regional collaborative fluctuation characteristics enables the model to more accurately reflect the dynamic change trend of new energy power at a medium-and-long-term time scale, thereby improving the degree of coincidence between the prediction result and the actual power output.

[0051] 2.The method for predicting regional medium and long term new energy power provided by the embodiment of the present application can build a weather-output mapping relationship that adapts to the medium and long term climate evolution rule by fusing the output characteristic migration matrix with the climate model prediction data. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 FIG. 1 is a flowchart of a method for predicting regional medium and long term new energy power provided by an embodiment of the present application.

[0053] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are merely intended to explain the present application and are not intended to limit the present application.

[0055] The embodiment of the present application provides a method for predicting regional medium and long term new energy power. The execution subject of the method for predicting regional medium and long term new energy power includes but is not limited to at least one of electronic devices such as a server, a terminal and the like which can be configured to execute the method provided by the embodiment of the present application. In other words, the method for predicting regional medium and long term new energy power can be executed by software or hardware installed in 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 and the like. The server can be a stand-alone server or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

[0056] Referring to Figure 1 FIG. 1 is a flowchart of a method for predicting regional medium and long term new energy power provided by an embodiment of the present application. In the embodiment, the method for predicting regional medium and long term new energy power includes:

[0057] S1, obtaining historical new energy output time series data and corresponding meteorological element data of a target region to generate a regional basic data set;

[0058] In the embodiment of the present application, the obtaining historical new energy output time series data and corresponding meteorological element data of a target region to generate a regional basic data set includes:

[0059] Collect wind farm power output sequences and photovoltaic power station power output sequences in the target area and associated areas;

[0060] Performing time alignment processing on the wind farm power output sequence and the photovoltaic power station power output sequence to generate a standardized output time sequence for the target area;

[0061] The standardized output time series is associated with meteorological element monitoring data of corresponding sites to form a regional basic data set of the target area.

[0062] The meteorological element monitoring data includes:

[0063] Extract atmospheric boundary layer height data, surface radiation flux data and near-surface wind vector data monitored by meteorological stations;

[0064] The atmospheric boundary layer height data, surface radiation flux data and near-surface wind vector data are spatially interpolated to generate a meteorological element grid field covering the target area, and the meteorological element grid field is used as the meteorological element monitoring data of the meteorological station.

[0065] Specifically, determine the specific geographical scope of the target area and related areas, contact the operating and management parties of all wind farms and photovoltaic power stations within the scope, and obtain the original power output records of each power station. These records must contain continuous time information (accurate to the minute) and the corresponding power output value. Arrange all records of the wind farm in chronological order as a wind farm power output sequence, and arrange all records of the photovoltaic power station in chronological order as a photovoltaic power station power output sequence. At the same time, verify the time continuity of the sequence. If there is a record interruption due to equipment failure, it is necessary to obtain supplementary measurement data from the operator or extract the corresponding time period data through the power station backup monitoring system to ensure that both sequences are complete continuous time series.

[0066] Furthermore, the time granularity of the wind farm power output sequence and the photovoltaic power station power output sequence is unified. With a time interval of 15 minutes, all power output values ​​in the two sequences are assigned to corresponding 15-minute time intervals. The average power output value in each interval is taken as the representative value of the interval. If there is a missing value in a certain interval in one of the sequences, the arithmetic mean of the power output values ​​of the two adjacent intervals before and after the interval in the sequence is used to fill the missing value. After completion, the two sequences will have corresponding power output values ​​at each 15-minute time point. These aligned wind farm and photovoltaic power station power output values ​​are arranged in chronological order to form a standardized output time series for the target area.

[0067] Further, collect the meteorological element monitoring data of each wind farm and photovoltaic power station in the target area and the associated area, including wind speed, wind direction, light intensity, temperature, humidity, etc., and the monitoring time granularity is consistent with the standardized output time sequence (i.e. 15 minutes per data point), determine the correspondence between each power station and the meteorological monitoring site, ensure that the power output data of each power station can be matched to the unique meteorological monitoring site data, then match the power output value of each 15-minute time point in the standardized output time sequence with all meteorological element monitoring data of the corresponding site at that time point, so that each time point contains standardized output data and corresponding meteorological element data, integrate all matched time point data in chronological order to form a complete data set containing time, wind farm power output, photovoltaic power station power output and corresponding meteorological elements, i.e. the regional basic data set of the target area.

[0068] Specifically, contact the operators of all meteorological stations in the target area and the associated area, obtain the atmospheric boundary layer height data, surface radiation flux data and near-surface wind vector data monitored by each station through formal data sharing agreement or interface, the data should include the monitoring time (accurate to 15 minutes) and specific geographic location information (latitude and longitude coordinates) of the meteorological station corresponding to each data, and the data format should be unified as a structured table containing time, latitude and longitude, and element value. Arrange the atmospheric boundary layer height data of each station in chronological order into a continuous sequence, and similarly arrange the surface radiation flux data sequence and near-surface wind vector data sequence of each station, check the integrity of all sequences, if there is short-time data missing due to instrument calibration (for example, the surface radiation flux data of a certain 15-minute point of a certain station is missing), find the values of the nearest two valid monitoring time points before and after the missing point, calculate the time difference ratio, and fill in the missing value using linear interpolation method (if the missing point is located between the two valid points, the interpolation value is the arithmetic mean of the previous and next values), to ensure that the three kinds of data have corresponding values at all monitoring time points, and finally form an original data set containing the location, time and corresponding three kinds of data of each station.

[0069] Further, the target area is divided into uniform square grids with a size of 1 km x 1 km, and the minimum latitude and longitude of the target area are taken as the starting point to generate grids covering the entire area, and each grid is assigned a unique coordinate identifier (such as row number and column number). The inverse distance weighting method is used to spatially interpolate the atmospheric boundary layer height data. The specific steps are as follows: for the center point of each grid, find all the atmospheric boundary layer height data of the meteorological stations within a 50 km range around the center point at the same time point, calculate the straight-line distance between each station and the center point of the grid, and set the weight as the inverse square of the distance (i.e. the closer the distance, the greater the weight), multiply the atmospheric boundary layer height data of each station by the corresponding weight, and then sum them up, and divide by the sum of all weights to get the atmospheric boundary layer height value of the center point of the grid. In this way, the atmospheric boundary layer height values of all grids are calculated to form the atmospheric boundary layer height grid field.

[0070] Further, the same grid division method and inverse distance weighting method are used to interpolate and calculate the surface radiation flux data and near-surface wind vector data. For the near-surface wind vector data, the wind direction and wind speed are processed separately: the wind speed is directly interpolated using the inverse distance weighting method, and the wind direction is first converted into east and north components (for example, the wind direction θ is converted into u = cosθ x wind speed, v = sinθ x wind speed), and then the u and v components are interpolated using the inverse distance weighting method, and finally the wind direction and wind speed are combined to obtain the near-surface wind vector grid field. The atmospheric boundary layer height grid field, the surface radiation flux grid field and the near-surface wind vector grid field are integrated into a meteorological element grid field covering the target area, and the meteorological element grid field is determined as the meteorological element monitoring data of the meteorological station, ensuring that each grid contains three meteorological element values at each 15-minute time point.

[0071] In summary, in the process of obtaining the historical new energy output time series data and corresponding meteorological element data of the target area and generating the regional basic data set, by collecting the new energy output sequence of the target area and the associated area and performing time alignment processing, a standardized output time series is formed, ensuring the consistency and comparability of the data in the time dimension, laying a high-quality data foundation for subsequent analysis of cross-regional collaborative fluctuation characteristics, and helping to improve the effectiveness of data application.

[0072] In summary, by extracting key meteorological element data and performing spatial interpolation to generate a meteorological element grid field covering the target area, and then correlating it with the standardized output time series, the spatial dimension of meteorological data and output data is accurately matched, making the regional basic data set fully reflect the spatio-temporal correlation characteristics of new energy output and meteorological elements, providing comprehensive and accurate data support for subsequent mining of the deep mapping relationship between the two, and thus ensuring the input quality of the prediction model.

[0073] S2, analyze the spatio-temporal coupling features in the regional basic data set, and extract a cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling features;

[0074] In the embodiment of the application, the analysis of the spatio-temporal coupling features in the regional basic data set and the extraction of the cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling features include:

[0075] Based on the regional basic data set, a new energy output spatio-temporal tensor model of the target region is constructed;

[0076] The spatio-temporal tensor model is subjected to high-order singular value decomposition to obtain a spatio-temporal coupling feature matrix of the target region;

[0077] From the spatio-temporal coupling feature matrix, a feature vector with a variance contribution rate exceeding a preset threshold is screened as the cross-regional new energy output collaborative fluctuation mode.

[0078] The screening of the feature vector with the variance contribution rate exceeding the preset threshold from the spatio-temporal coupling feature matrix as the cross-regional new energy output collaborative fluctuation mode includes:

[0079] According to the spatio-temporal coherence coefficient of the feature vector in the spatio-temporal coupling feature matrix, a mode energy distribution function of the target region is constructed;

[0080] The screening boundary of the feature vector is determined through the gradient change of the mode energy distribution function;

[0081] The feature vector within the screening boundary is taken as the cross-regional new energy output collaborative fluctuation mode.

[0082] After the analysis of the spatio-temporal coupling features in the regional basic data set and the extraction of the cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling features, the method further includes:

[0083] The feature vector is mapped to an atmospheric circulation pattern field;

[0084] The phase correlation relationship between the feature vector and an atmospheric activity center is verified;

[0085] The feature vector with a strong correlation is spliced into a spatio-temporal correlation matrix.

[0086] Specifically, all time points, location information of each new energy site (wind farm and photovoltaic power station), and corresponding power output data are extracted from the regional basic data set. The time dimension is divided into continuous time point sequences at 15-minute intervals, the spatial dimension is divided into different spatial sites according to the latitude and longitude coordinates of each site, and the new energy type dimension is divided into wind power and photovoltaic power. A three-dimensional new energy output space-time tensor model is constructed, in which each element value of the tensor corresponds to the power output value of a specific time point, a specific spatial site, and a specific new energy type. If there is data missing for an element (such as the photovoltaic power output of a site at a certain time point not being recorded), the arithmetic mean of the power output values of the same type of new energy at the adjacent two time points of the site is used for filling to ensure that all elements in the tensor model have a determined value, and finally a complete new energy output space-time tensor model of the target region is formed.

[0087] Further, the new energy output space-time tensor model is subjected to high-order singular value decomposition. The tensor is first expanded into a matrix along the time dimension, and singular value decomposition is performed on the matrix to obtain a time factor matrix. Then, the tensor is expanded into a matrix along the spatial site dimension, and singular value decomposition is performed to obtain a spatial factor matrix. Next, the tensor is expanded into a matrix along the new energy type dimension, and singular value decomposition is performed to obtain a type factor matrix. Then, the core tensor is calculated according to the three factor matrices. The elements in the core tensor reflect the coupling strength between different dimensions. The core tensor is multiplied and combined with the time factor matrix, the spatial factor matrix, and the type factor matrix in a specific order to obtain a space-time coupling characteristic matrix of the target region that can reflect the interaction relationship between time, space, and new energy type.

[0088] Further, the variance contribution rate corresponding to each eigenvector in the space-time coupling characteristic matrix is calculated. The specific method is as follows: first, the amount of variance explained by each eigenvector is calculated, and then each variance is divided by the total variance of the characteristic matrix to obtain the respective variance contribution rate. All eigenvectors are arranged in order of variance contribution rate from large to small, and the preset threshold is set to 85%. From the sorted eigenvectors, they are selected in turn, and their variance contribution rates are accumulated. When the cumulative value first exceeds 85%, the selection is stopped. At this time, all selected eigenvectors collectively constitute a set that can explain most of the new energy output fluctuation rules. These eigenvectors are determined as the cross-regional new energy output collaborative fluctuation mode of the target region.

[0089] Specifically, all feature vectors are extracted from the spatio-temporal coupling feature matrix, for each feature vector, in the time dimension, the correlation degree of the values corresponding to any two different time points is calculated, and the average value of all time point combinations is taken as the time correlation of the feature vector; in the spatial dimension, the correlation degree of the values corresponding to any two different stations is calculated, and the average value of all station combinations is taken as the spatial correlation of the feature vector, the time correlation and the spatial correlation are added and divided by 2 to obtain the spatio-temporal coherence coefficient of each feature vector, which reflects the close degree of the feature vector in time and space, the greater the coefficient value, the stronger the consistency of the feature vector in space-time, then based on the spatio-temporal coherence coefficient of the feature vector, the variance contribution rate corresponding to each feature vector is combined, and the two are weighted and summed in a 1:1 ratio (i.e. the spatio-temporal coherence coefficient and the variance contribution rate are directly added), to obtain the comprehensive energy value of each feature vector, and then these comprehensive energy values are arranged in order according to the arrangement order of the feature vectors in the spatio-temporal coupling feature matrix to form a modal energy distribution function of the target area which can reflect the energy size and distribution of different feature vectors.

[0090] Further, the modal energy distribution function is analyzed point by point, starting from the comprehensive energy value corresponding to the first feature vector, the comprehensive energy value of each feature vector is recorded in turn, and the trend of these values is observed, that is, whether it is rising, stable or falling, the gradient value corresponding to each point of the function is calculated, which is specifically the comprehensive energy value of the point minus the comprehensive energy value of the previous point, the result is positive indicating that the comprehensive energy value is rising, and the result is negative indicating that it is falling, the greater the absolute value of the gradient value, the faster the change rate of the comprehensive energy value at the point, when the gradient value changes from positive to negative, the comprehensive energy value corresponding to the point is determined as the screening threshold, and the comprehensive energy value greater than or equal to the threshold is within the screening boundary, and the comprehensive energy value less than the threshold is outside the screening boundary, so as to determine the screening boundary of the feature vector.

[0091] Further, all feature vectors in the spatio-temporal coupling feature matrix are traversed, and the comprehensive energy value of each feature vector is calculated one by one (the calculation method is consistent with that of constructing the modal energy distribution function), the calculated comprehensive energy value is compared with the screening threshold, and the feature vectors with a comprehensive energy value greater than or equal to the screening threshold are selected out, these feature vectors are within the screening boundary, they have strong spatio-temporal coherence and high energy, and can more accurately reflect the collaborative fluctuation rule of new energy output between different regions in the target area, and these selected feature vectors are determined as the cross-regional new energy output collaborative fluctuation mode of the target area.

[0092] Overall, in the process of analyzing the spatio-temporal coupling characteristics in the regional basic data set and extracting the cross-regional new energy output collaborative fluctuation mode, by constructing the new energy output spatio-temporal tensor model and performing high-order singular value decomposition, the spatio-temporal correlation law hidden in the data can be deeply mined, and the collaborative change mode of new energy output between different regions can be accurately captured. The effective extraction of the cross-regional collaborative fluctuation characteristics enables the model to break through the limitations of single-region data, fully utilizes the correlation information between regions, provides key feature support for subsequent construction of a more practical prediction model, and helps to improve the prediction accuracy.

[0093] Overall, by screening the feature vectors with variance contribution rates meeting the threshold and verifying the phase correlation between the extracted cross-regional new energy output collaborative fluctuation mode and the atmospheric activity center in combination with the atmospheric circulation pattern field, the extracted cross-regional new energy output collaborative fluctuation mode has clear physical meaning and strong correlation.

[0094] Overall, this process not only eliminates redundant information and focuses on key features, but also enhances the adaptability of features to meteorological system evolution, providing high-quality basic features for constructing the output characteristic transfer matrix and the medium and long-term meteorological-output mapping relationship, and further ensuring the scientificity and effectiveness of the prediction method.

[0095] S3, based on the cross-regional new energy output collaborative fluctuation mode, constructing the output characteristic transfer matrix of the target region;

[0096] In the embodiments of the present application, the construction of the output characteristic transfer matrix of the target region based on the cross-regional new energy output collaborative fluctuation mode comprises:

[0097] Identifying the persistent climate features in the meteorological element monitoring data;

[0098] Dividing the historical data into typical climate scenarios according to the persistent climate features, and generating the climate feature weights of the target region;

[0099] Based on the climate feature weights and the spatio-temporal correlation matrix, constructing the output characteristic transfer matrix of the target region, wherein the expression of the output characteristic transfer matrix is as follows:

[0100] ;

[0101] In the formula, is the output characteristic transfer matrix, is a symbol representing that the output characteristic transfer matrix changes with time, is the spatio-temporal correlation matrix, is the climate feature weight, is the Hadamard product operator.

[0102] Specifically, the continuous records of key meteorological elements such as wind speed, light intensity, temperature, etc. are extracted from meteorological element monitoring data, time intervals are divided according to seasons, daily average values of each meteorological element in each season are calculated, daily average values of the same season in the past three years are compared, and the number of days in which each meteorological element remains stable fluctuation (fluctuation range does not exceed 5%) within three days before and after the same date is counted. The meteorological element characteristics that show stable fluctuation mode in the same season in consecutive years are determined as persistent climate characteristics, for example, the light intensity in the afternoon in a certain region in summer remains stable within a similar range for three consecutive years, which is one of the persistent climate characteristics of the region.

[0103] Further, according to the identified persistent climate characteristics, the historical meteorological data is divided into four typical climate scenarios of spring, summer, autumn and winter according to seasons, the frequency of occurrence of each persistent climate characteristic in each scenario is calculated, for example, the proportion of the number of days of "afternoon strong light" characteristics in summer scenario to the total number of days in summer, the frequency value of each characteristic is divided by the total frequency of all characteristics in the scenario, to obtain the weight value of each persistent climate characteristic in the corresponding scenario, and then the weight values of all persistent climate characteristics in the same seasonal scenario are integrated to form the climate characteristic weight of the target region, ensuring that the sum of the climate characteristic weights in each scenario is 1.

[0104] Further, based on the climate characteristic weight, combined with the correlation coefficients between stations in the space-time correlation matrix, the climate characteristic weight corresponding to each station is multiplied by the space-time correlation coefficient between the station and all other stations to obtain the migration influence value of the station to other stations, and all migration influence values are arranged in matrix form in order of stations, wherein the rows of the matrix represent the source stations, the columns represent the target stations, and the element values in the matrix are the output characteristic migration coefficients of the source stations to the target stations. In this way, the output characteristic migration matrix of the target region reflecting the mutual migration relationship of the output characteristics of different stations under the influence of climate characteristics is constructed.

[0105] Specifically, all feature vectors are extracted from the spatio-temporal coupling feature matrix, for each feature vector, in the time dimension, the correlation degree of the values corresponding to any two different time points is calculated, and the average value of all time point combinations is taken as the time correlation of the feature vector; in the spatial dimension, the correlation degree of the values corresponding to any two different stations is calculated, and the average value of all station combinations is taken as the spatial correlation of the feature vector, the time correlation and the spatial correlation are added and divided by 2 to obtain the spatio-temporal coherence coefficient of each feature vector, which reflects the close degree of the feature vector in time and space, the greater the coefficient value, the stronger the consistency of the feature vector in space-time, then based on the spatio-temporal coherence coefficient of the feature vector, the variance contribution rate corresponding to each feature vector is combined, and the two are weighted and summed in a 1:1 ratio (i.e. the spatio-temporal coherence coefficient and the variance contribution rate are directly added), to obtain the comprehensive energy value of each feature vector, and then these comprehensive energy values are arranged in order according to the arrangement order of the feature vectors in the spatio-temporal coupling feature matrix to form a modal energy distribution function of the target area which can reflect the energy size and distribution of different feature vectors.

[0106] Further, the modal energy distribution function is analyzed point by point, starting from the comprehensive energy value corresponding to the first feature vector, the comprehensive energy value of each feature vector is recorded in turn, and the trend of these values is observed, that is, whether it is rising, stable or falling, the gradient value corresponding to each point of the function is calculated, which is specifically the comprehensive energy value of the point minus the comprehensive energy value of the previous point, the result is positive indicating that the comprehensive energy value is rising, and the result is negative indicating that it is falling, the greater the absolute value of the gradient value, the faster the change rate of the comprehensive energy value at the point, when the gradient value changes from positive to negative, the comprehensive energy value corresponding to the point is determined as the screening threshold, and the comprehensive energy value greater than or equal to the threshold is within the screening boundary, and the comprehensive energy value less than the threshold is outside the screening boundary, so as to determine the screening boundary of the feature vector.

[0107] Further, all feature vectors in the spatio-temporal coupling feature matrix are traversed, and the comprehensive energy value of each feature vector is calculated one by one (the calculation method is consistent with that of constructing the modal energy distribution function), the calculated comprehensive energy value is compared with the screening threshold, and the feature vectors with a comprehensive energy value greater than or equal to the screening threshold are selected out, these feature vectors are within the screening boundary, they have strong spatio-temporal coherence and high energy, and can more accurately reflect the collaborative fluctuation rule of new energy output between different regions in the target area, and these selected feature vectors are determined as the cross-regional new energy output collaborative fluctuation mode of the target area.

[0108] Overall, by identifying the persistent climate characteristics in the meteorological element monitoring data, dividing the typical climate scenarios and generating the climate feature weights, and then combining the spatio-temporal correlation matrix for Hadamard product operation, the matrix can dynamically reflect the influence of cross-regional coordinated fluctuation characteristics on the output characteristics of the target region under different climate scenarios, effectively capturing the migration law of the output characteristics over time and climate conditions, and providing a precise feature carrier for subsequent fusion of climate model prediction data.

[0109] Overall, the output characteristic migration matrix integrates the key features of cross-regional coordinated fluctuation and climate feature weights, not only retaining the correlation information of regional new energy output, but also highlighting the influence weight of climate factors, so that the matrix has spatio-temporal dynamic adaptability.

[0110] Overall, this structured matrix construction method lays a reliable foundation for subsequent generation of climate feature enhancement field and spatio-temporal modulation characteristics, which helps to improve the accuracy of medium and long-term meteorological-output mapping relationship, and further enhances the stability of new energy power prediction results.

[0111] S4, fusing the output characteristic migration matrix and the climate model prediction data to obtain the medium and long-term meteorological-output mapping relationship of the target region;

[0112] In the embodiments of the present application, the fusion of the output characteristic migration matrix and the climate model prediction data to obtain the medium and long-term meteorological-output mapping relationship of the target region comprises:

[0113] decomposing the output characteristic migration matrix into a spatial weight matrix and a time-varying modal matrix;

[0114] performing tensor convolution operation on the climate model prediction data and the spatial weight matrix to generate a climate feature enhancement field of the target region;

[0115] performing time scale transformation on the climate feature enhancement field through the time-varying modal matrix to generate a spatio-temporal modulation characteristic of the target region;

[0116] performing feature selection on the spatio-temporal modulation characteristic to extract key meteorological-output correlation characteristics of the target region;

[0117] constructing a medium and long-term meteorological-output mapping relationship of the target region based on the key meteorological-output correlation characteristics.

[0118] The construction of the medium and long-term meteorological-output mapping relationship of the target region based on the key meteorological-output correlation characteristics comprises:

[0119] dividing the key meteorological-output correlation characteristics into multi-class features according to meteorological element types and output response characteristics;

[0120] The physical evolution path of the cross-regional new energy output collaborative fluctuation mode evolves the multiple-category features into a causal rule chain;

[0121] The causal rule chain is aggregated into a medium and long-term meteorological-output mapping relationship for the target region.

[0122] Specifically, the part related to the spatial position and the part changing over time in the output characteristic migration matrix are separated, specifically, the elements in the matrix that do not change over time and reflect the spatial correlation strength of each site are extracted, arranged in the row and column order of the original matrix, and a spatial weight matrix is formed; the elements in the matrix that change over time and embody the modal characteristics at different times are extracted, and arranged in the same time order and site order, and a time-varying modal matrix is formed, ensuring that the original numerical relationship of the output characteristic migration matrix can be restored after the multiplication of the two matrices.

[0123] Further, the climate model prediction data of the target region is obtained, which needs to include the predicted values of each meteorological element in a future period of time, and the time granularity matches the spatial resolution of the spatial weight matrix. The climate model prediction data is converted into a tensor form consistent with the dimension of the spatial weight matrix, where each element corresponds to the predicted value of a specific spatial position and time point, and then a tensor convolution operation is performed, that is, each element in the climate model prediction data tensor is multiplied by the element at the corresponding position in the spatial weight matrix, and the product results of the surrounding adjacent elements are accumulated to obtain a new tensor element value. Through this point-by-point operation, a climate feature enhancement field covering the entire target region is generated, which not only retains the trend of the prediction data but also integrates spatial correlation information.

[0124] Further, the conversion ratio of the time scale in the time-varying modal matrix is determined, for example, the time interval of the climate feature enhancement field is converted from daily to weekly. Specifically, the elements corresponding to the same time period in the time-varying modal matrix are averaged to obtain a representative value for that time period, and then all element values corresponding to the time period in the climate feature enhancement field are multiplied by the representative value to obtain the converted data, which is rearranged according to the new time scale to generate the spatiotemporal modulation characteristics of the target region, ensuring that the time interval of the new characteristics is uniform and can reflect the climate and output correlation trend over a longer period of time.

[0125] Further, the importance of all elements in the spatiotemporal modulation feature is evaluated, specifically, the correlation degree between each element and the actual value of new energy output is calculated, and elements with high correlation degree are considered more important. A fixed importance threshold is set, and elements with correlation degree higher than the threshold are selected. These elements are the key parts that significantly affect new energy output. These elements are sorted according to their position in the spatiotemporal modulation feature and time sequence to form the key meteorological-output correlation feature of the target area, ensuring that the extracted feature can effectively reflect the core relationship between meteorological conditions and new energy output.

[0126] Further, based on the key meteorological-output correlation feature, the meteorological element data in the feature is taken as input, and the corresponding new energy output data is taken as output to establish a corresponding relationship between them. Specifically, each key meteorological feature is paired with the corresponding output value in time sequence to form a series of data pairs. By analyzing the change rule of these data pairs, the corresponding change mode of new energy output when the meteorological element presents a certain feature is determined. These modes are systematically arranged to form a rule set that can describe how meteorological conditions affect new energy output in the future for a long time, i.e., the medium and long-term meteorological-output mapping relationship of the target area, ensuring that this relationship can accurately reflect the stable corresponding trend between key meteorological features and output.

[0127] Specifically, all meteorological elements involved in the key meteorological-output correlation feature are extracted, and wind speed, light intensity, temperature, humidity, and air pressure are clearly distinguished. The output response characteristics are also identified, including the rising / falling trend of wind power, the peak / valley change of photovoltaic power, etc. Clear classification criteria are established for each meteorological element type and output response characteristic, such as wind speed class features requiring wind speed value range and change rate, and wind power response class features requiring power change amplitude and duration. The classification framework is formed by horizontally dividing by meteorological element type and vertically dividing by output response characteristic. Each key meteorological-output correlation feature is matched with the categories in the framework, such as a feature containing wind speed increasing by 2 meters per hour and wind power rising by 10%, which is classified into the cross category of wind speed class-wind power rising trend. The classification results of all features are confirmed one by one to ensure no repetition or omission, and finally a multi-class feature with clear structure is obtained.

[0128] Further, the order of appearance and influence relationship of various categories of features in the cross-regional new energy output collaborative fluctuation mode are combed. The direction of the causal relationship is determined by analyzing the time sequence and physical correlation of the features in the historical data. For example, when the wind speed in a certain region increases by 5 m / s in the next 3 15-minute periods, and then the wind power of the wind farm in the region and the adjacent region increases by 20% in the next 3 periods, it is determined that “continuous increase of wind speed” is the cause of “collaborative increase of wind power”. According to this logic, the multi-category features in the same physical evolution path are arranged in causal order, and the result of the previous feature is used as the cause of the next feature, forming an initial rule chain, such as “continuous increase of wind speed → increase of wind power in region A → increase of wind power in region B”. Different initial rule chains are compared, and if there are shared features or continuous causal relationships, they are merged, for example, “continuous increase of wind speed → increase of wind power in region A” and “increase of wind power in region A → increase of wind power in region C” are merged into a longer chain. Finally, a causal rule chain covering the whole process of cross-regional collaborative fluctuation is formed, which fully reflects the physical evolution path from meteorological element change to cross-regional output response.

[0129] Further, all verified causal rule chains are collected and grouped according to the core meteorological element types involved in the rule chains, such as wind speed dominant type, light intensity dominant type, temperature dominant type, etc. Each group is sorted by the length and complexity of the rule chain, from simple two-link chain to multi-link chain. The logical consistency of each group of rule chains is checked to verify whether the causal relationship of different chains under the same meteorological element type is contradictory. For example, check all wind speed dominant type chains to see if the direction of wind speed change and wind power change is consistent. If there is a contradiction, recheck the original data and evolution path analysis process, and correct the wrong chain. The verified rule chains in each group are integrated according to the weight of meteorological element influence, and the weight is determined according to the frequency of the rule chain appearing in the historical data. The higher the frequency, the higher the priority in integration. Finally, a complete system containing all core causal relationships, logically consistent and covering all types of meteorological-output correlation scenarios is formed, i.e. the medium and long-term meteorological-output mapping relationship of the target region. This relationship can clearly show how different meteorological elements affect cross-regional new energy output through multi-link causal action, ensuring accurate prediction of output response based on meteorological changes on a medium and long-term time scale.

[0130] In summary, by fusing the output characteristic transfer matrix and the climate model prediction data to obtain the medium and long-term meteorological-output mapping relationship, the output characteristic transfer matrix is decomposed into a spatial weight matrix and a time-varying mode matrix, and the climate feature enhancement field is generated through tensor convolution operation, which can effectively enhance the adaptability of climate model prediction data to the spatial characteristics of the target region, making the meteorological data more consistent with the actual situation of the region.

[0131] Overall, time-scale transformation is performed by means of time-varying modal matrix to generate spatiotemporal modulation features, and key meteorological-output correlation features are extracted to construct a mapping relationship, which not only accurately captures the dynamic correlation between meteorological elements and new energy output in the time dimension, but also focuses on core influencing factors, so that the constructed mapping relationship can deeply adapt to the long-term climate evolution law, providing a reliable correlation basis for new energy power prediction, and helping to improve the accuracy and applicability of prediction.

[0132] S5、According to the long-term meteorological-output mapping relationship, output the new energy power prediction result of the target area.

[0133] In the embodiments of the present application, according to the long-term meteorological-output mapping relationship, the new energy power prediction result of the target area is output, which comprises:

[0134] According to the key climate driving factor of the target area, match the output relationship in the long-term meteorological-output mapping relationship;

[0135] The output relationship and the causal rule chain of the key climate driving factor are output as a visual new energy power prediction result.

[0136] Specifically, the continuous data of the key climate driving factor is obtained from the real-time meteorological monitoring system of the target area, including the specific values of wind speed, light intensity, temperature, etc. recorded every hour, as well as the change rate (such as the increase or decrease of wind speed per hour) and duration (such as the number of hours a certain light intensity lasts) of each factor. These data are arranged into a structured table, which contains information such as timestamp, meteorological element type, specific value, change rate and duration. All preset meteorological element conditions in the long-term meteorological-output mapping relationship are retrieved, which also includes clear definitions of value range, change rate and duration, for example, "wind speed in 3-5 meters / second, hourly increase amplitude not more than 1 meter / second, duration ≥4 hours" corresponds to a specific wind power output relationship. Compare the real-time key climate driving factor data with the meteorological element conditions in the mapping relationship one by one, match the meteorological element type first, then check whether the value is within the preset range, whether the change rate meets the definition, and whether the duration meets the standard. If the conditions are fully met, the corresponding output relationship is directly extracted; if there is a slight deviation (such as wind speed value exceeding the range within 0.3 meters / second), calculate the deviation rate, and if the deviation rate ≤5%, it is considered as approximate matching, the corresponding output relationship is extracted and the deviation item is marked, to ensure that the correlation between the output relationship obtained and the real-time key climate driving factor has clear basis.

[0137] Further, the extracted output relationship is disassembled into specific new energy power values and change trends, such as "the wind power will rise from 100 MW to 150 MW in the next 12 hours, with an increase of 5 MW per hour", and the corresponding key climate driving factor causal rule chain is sorted out, which includes complete links such as "wind speed increases from 3 m / s to 5 m / s → wind power of wind farm A rises → wind power of wind farm B rises synchronously". A visualization prediction result is generated using a chart tool, the horizontal axis is set as time (in hours, covering the next 72 hours), the vertical axis is set as power (in megawatts), the blue line is used to represent the change of wind power, the red line is used to represent the change of photovoltaic power, and each point on the line is marked with a specific power value and a corresponding time. A causal rule chain explanation area is added below the chart, the key climate driving factors and output changes are connected by horizontal arrows, the influence path (such as "wind speed ↑ → wind power ↑") is marked beside the arrow, the influence intensity is distinguished by the thickness of the arrow (thick arrow represents strong correlation), and the correlation basis (such as "based on 4 hours of monitoring data") is marked above the arrow. The prediction start time is marked in the upper left corner of the chart, and the data source (real-time monitoring system and medium and long-term mapping relationship) is marked in the upper right corner, so that the visualization result not only shows the power prediction values, but also clearly presents the causal logic between the climate driving factors and the output changes, and the user can intuitively read the power trend in the future period and the meteorological driving mechanism behind it.

[0138] In summary, when the present application outputs the new energy power prediction result of the target area according to the medium and long-term meteorological-output mapping relationship, the output prediction result can accurately respond to the long-term influence of meteorological element change on new energy output by matching the key climate driving factor with the output relationship in the mapping relationship, which improves the pertinence and accuracy of the prediction result.

[0139] In summary, the causal rule chain of the output relationship and the key climate driving factor is output in a visual form, which not only intuitively presents the generation logic of the prediction result, but also enhances the explainability of the prediction process, facilitates the user to understand the correlation mechanism between meteorological factors and new energy output, provides a clear and reliable reference basis for new energy dispatching, planning and other decisions, and further improves the practical value of the prediction result.

[0140] In the several embodiments of the present application, it should be understood that the disclosed method can be implemented in other ways.

[0141] Obviously, the present application is not limited to the details of the above exemplary embodiments, but can be implemented in other concrete forms without departing from the spirit or essential characteristics of the present application.

[0142] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. The artificial intelligence is to use a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for predicting long-term new energy power in a region, characterized in that, The method comprises: S1, obtaining historical new energy output time series data and corresponding meteorological element data of a target area, and generating a regional basic data set; S2, analyzing the spatio-temporal coupling characteristics in the regional basic data set, and extracting the cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling characteristics, comprising: constructing a new energy output spatio-temporal tensor model of the target area based on the regional basic data set; performing high-order singular value decomposition on the spatio-temporal tensor model to obtain a spatio-temporal coupling characteristic matrix of the target area; screening feature vectors with variance contribution rates exceeding a preset threshold from the spatio-temporal coupling characteristic matrix as the cross-regional new energy output collaborative fluctuation mode, comprising: constructing a mode energy distribution function of the target area according to the spatio-temporal coherence coefficients of the feature vectors in the spatio-temporal coupling characteristic matrix; determining the screening boundary of the feature vectors through the gradient change of the mode energy distribution function; taking the feature vectors within the screening boundary as the cross-regional new energy output collaborative fluctuation mode; S3, constructing an output characteristic transfer matrix of the target area based on the cross-regional new energy output collaborative fluctuation mode, comprising: identifying persistent climate characteristics in the meteorological element monitoring data; dividing the historical data into typical climate scenarios according to the persistent climate characteristics, and generating climate characteristic weights of the target area; constructing an output characteristic transfer matrix of the target area based on the climate characteristic weights and the spatio-temporal correlation matrix, wherein the expression of the output characteristic transfer matrix is as follows: ； wherein is the output characteristic transfer matrix, is the output characteristic transfer matrix, is the spatio-temporal correlation matrix, is the climate feature weight, is the Hadamard product operator. S4, fusing the output characteristic transfer matrix and climate mode prediction data to obtain a medium and long term meteorological-output mapping relationship of the target area; S5, outputting a new energy power prediction result of the target area according to the medium and long term meteorological-output mapping relationship.

2. The method of claim 1, wherein the step of predicting the long-term new energy power in the region comprises: The method comprises: collecting wind farm power output sequences and photovoltaic power station power output sequences of the target area and associated areas; performing time alignment processing on the wind farm power output sequences and the photovoltaic power station power output sequences to generate standardized output time series of the target area; associating the standardized output time series with meteorological element monitoring data of corresponding sites to form a regional basic data set of the target area.

3. The method of claim 2, wherein the step of predicting the long-term new energy power in the region comprises: The meteorological element monitoring data comprises: extracting atmospheric boundary layer height data, surface radiation flux data, and near-surface wind vector data monitored by meteorological stations; performing spatial interpolation processing on the atmospheric boundary layer height data, surface radiation flux data, and near-surface wind vector data to generate a meteorological element grid field covering the target area, and taking the meteorological element grid field as the meteorological element monitoring data of the meteorological station. 4.The method of claim 1, wherein, After analyzing the spatio-temporal coupling characteristics in the regional basic data set and extracting the cross-regional new energy output collaborative fluctuation mode in the spatio-temporal coupling characteristics, the method further comprises: mapping the feature vectors to an atmospheric circulation pattern field; verifying the phase correlation relationship between the feature vectors and the atmospheric activity center; The verification structure is spliced into a space-time correlation matrix.

5. The method for predicting regional medium- and long-term new energy power according to claim 1, wherein: The output characteristic transfer matrix is fused with the climate model prediction data to obtain a medium and long term meteorological-output mapping relationship of the target area, including: The output characteristic transfer matrix is decomposed into a spatial weight matrix and a time-varying modal matrix; The climate model prediction data and the spatial weight matrix are subjected to tensor convolution operation to generate a climate feature enhancement field of the target area; The climate feature enhancement field is subjected to time scale transformation through the time-varying modal matrix to generate a space-time modulation feature of the target area; The space-time modulation feature is subjected to feature selection to extract key meteorological-output correlation features of the target area; The key meteorological-output correlation features are used to construct a medium and long term meteorological-output mapping relationship of the target area.

6. The method for predicting regional medium- and long-term new energy power according to claim 5, characterized in that: The key meteorological-output correlation features are used to construct a medium and long term meteorological-output mapping relationship of the target area, including: The key meteorological-output correlation features are divided into multi-category features according to meteorological element types and output response characteristics; The multi-category features are evolved into a causal rule chain through the physical evolution path of the cross-regional new energy output collaborative fluctuation mode; The causal rule chain is collected as a medium and long term meteorological-output mapping relationship of the target area.

7. The method for predicting regional medium- and long-term new energy power according to claim 6, characterized in that: The new energy power prediction result of the target area is output according to the medium and long term meteorological-output mapping relationship, including: The output relationship is matched with the causal rule chain of the key climate driving factor to output a visual new energy power prediction result. The output relationship is matched with the causal rule chain of the key climate driving factor to output a visual new energy power prediction result.

Images