Intelligent new energy power generation power prediction curve decomposition analysis method and system

Abstract

The invention discloses an intelligent new energy power generation power prediction curve decomposition analysis method and system, and belongs to the technical field of new energy prediction analysis, and the method comprises the steps of data collection, power curve decomposition, error distribution analysis, intelligent factor identification and correction suggestion generation. According to the method, a staged integration method combining power curve decomposition, error distribution analysis and intelligent factor identification is adopted, the cause of the prediction error is systematically analyzed, and targeted correction suggestions are provided, so that the precision and reliability of a prediction model are improved; a multi-scale space-time coupling power curve decomposition method is adopted to perform power curve decomposition, so that a reconstructed prediction curve with more physical significance and stability is generated, and the adaptability of a prediction model to complex space-time characteristics is remarkably improved; intelligent factor identification is carried out by adopting an error dominant factor identification method combining modal decoupling and causal response analysis, and accurate identification and quantitative analysis of error influence factors are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy prediction and analysis, and specifically refers to an intelligent new energy power generation prediction curve decomposition and analysis method and system. Background Art

[0002] Intelligent decomposition and analysis of renewable energy power generation forecast curves combines industrial data analysis with artificial intelligence technology to perform refined decomposition, abnormal fluctuation identification, and error attribution analysis on renewable energy power generation forecast curves, enabling intelligent evaluation of forecast curves. This aims to improve the accuracy and stability of renewable energy power generation forecasts and provide reliable data support for grid dispatching, energy storage dispatching, and operation and maintenance.

[0003] However, in the decomposition and analysis process of the new energy power generation prediction curve, existing methods often only focus on the accuracy of the results and lack in-depth analysis of the source of errors. Once the prediction curve deviates, it is difficult to determine whether it is due to insufficient adaptability of the model itself or fluctuations or missing input data, resulting in a large degree of blindness in the model tuning process. In the existing power curve decomposition process, it is difficult to take into account multi-scale changes and time dynamics at the same time, resulting in unstable decomposition results and lack of physical interpretation ability. Single-scale or fixed mode decomposition methods are prone to ignore short-term disturbances or long-term trend changes, which in turn affects the learning effect of subsequent models. In the existing error factor analysis method, the identification of error source factors usually remains at the level of statistical correlation analysis, lacking systematic modeling of causal relationships. In actual applications, errors are often caused by the joint action of multiple factors with complex interactive relationships. Simply relying on significance ranking or correlation coefficient cannot accurately identify the key factors that truly dominate the error. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the existing technology, the present invention provides an intelligent new energy power generation prediction curve decomposition and analysis method and system. In view of the fact that in the process of decomposition and analysis of the new energy power generation prediction curve, the existing methods often only focus on the accuracy of the results, lack of in-depth analysis of the source of errors, and once the prediction curve deviates, it is difficult to judge whether it is due to insufficient adaptability of the model itself or fluctuations or missing input data, resulting in a large degree of blindness in the model tuning process. This solution creatively adopts a phased integration method that combines power curve decomposition, error distribution analysis and intelligent factor identification, and realizes a systematic analysis of the causes of prediction errors, and provides targeted correction suggestions, thereby improving the accuracy and reliability of the prediction model; in the process of existing power curve decomposition, it is difficult to take into account multi-scale changes and time dynamics at the same time, resulting in unstable decomposition results and lack of physical interpretation ability. Single-scale or fixed-mode decomposition methods are easy to In order to solve the technical problem of ignoring short-term disturbances or long-term trend changes, which in turn affects the learning effect of subsequent models, this solution creatively adopts a multi-scale spatiotemporal coupling power curve decomposition method to decompose the power curve, thereby generating a reconstructed prediction curve with more physical significance and stability, and significantly improving the adaptability of the prediction model to complex spatiotemporal characteristics; in the existing error factor analysis method, the identification of error source factors usually remains at the level of statistical correlation analysis, and lacks systematic modeling of causal relationships. In actual applications, errors are often caused by the joint action of multiple factors with complex interactive relationships. Simply relying on significance ranking or correlation coefficients cannot accurately identify the key factors that truly dominate the errors. This solution creatively adopts an error dominant factor identification method that combines modal decoupling and causal response analysis to perform intelligent factor identification, achieving accurate identification and quantitative analysis of error influencing factors, and providing strong support for the optimization of prediction models.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an intelligent new energy power generation prediction curve decomposition and analysis method, which includes the following steps:

[0006] Step S1: data collection;

[0007] Step S2: power curve decomposition;

[0008] Step S3: error distribution analysis;

[0009] Step S4: intelligent factor identification;

[0010] Step S5: Generate correction suggestions.

[0011] Furthermore, in step S1, the data collection is specifically to obtain a standardized time-series power generation prediction data set through multi-source power generation data collection and data optimization processing;

[0012] The multi-source power generation data collection specifically includes collecting new energy power generation power curves, meteorological data, operating condition data and geographic time and space information;

[0013] The new energy power generation curve includes a power generation prediction curve and an actual power generation curve;

[0014] The data optimization processing includes data time series alignment, outlier removal, data standardization and error pattern annotation;

[0015] The error pattern marking is to pre-mark the predicted error type by pre-defining the error type, including excessive error, too low error, violent fluctuation error, system offset error, oscillation error and mutation error;

[0016] The standardized time-series power generation prediction data set includes power generation prediction input samples, new energy power generation curves, error annotation data, space annotation data, and time annotation data.

[0017] Furthermore, in step S2, the power curve decomposition is specifically performed by using a multi-scale spatiotemporal coupling power curve decomposition method to decompose the power curve to obtain a reconstructed power prediction curve, including the following steps:

[0018] Step S21: Multiscale variational modal decomposition, specifically, by introducing a multiscale coupling regularization term on the basis of standard variational modal decomposition, constructing an improved optimization objective function, and performing multiscale decomposition on the renewable energy power generation curve in the standardized time-series power generation prediction dataset to obtain a curve modal component set;

[0019] Step S22: spatial feature extraction, specifically, constructing each curve modal component into a three-dimensional space-time tensor according to time, spatial position and modal dimension, and then performing low-rank decomposition on the three-dimensional space-time tensor using the tensor singular value decomposition method to extract the main modal components of space-time coupling and generate a low-rank reconstruction tensor;

[0020] Step S23: Modal adaptive update, specifically combining the sliding window mechanism and the recursive weighting strategy to construct a modal component dynamic update algorithm, update the spatiotemporal coupled main modal components, and obtain dynamic modal components;

[0021] Step S24: Power weighted fusion is used to fuse the error and structural stability information to reconstruct the prediction curve. Specifically, the local error is obtained by comparing the difference between the power prediction value and the actual power of the dynamic modal component at the time point t, and the modal stability index is extracted from the low-rank reconstruction tensor; the local error and the modal stability index are combined to construct the normalized weight; finally, the dynamic modal components are weighted and summed according to the normalized weight to generate the reconstructed power prediction curve.

[0022] Furthermore, in step S3, the error distribution analysis is used to clarify the distribution type and cause of the prediction error. Specifically, the error distribution analysis is performed using an error statistical calculation method combined with cluster modeling to obtain error pattern classification reference data, including the following steps:

[0023] Step S31: basic error calculation, specifically, based on the reconstructed power prediction curve, by comparing the difference between the reconstructed power prediction value and the actual power at each time point t to obtain the basic error;

[0024] Step S32: Error cluster identification, specifically using a density peak clustering algorithm to cluster the basic errors, extract the latent class structure of the errors, and generate error types;

[0025] The error types include excessively high errors, excessively low errors, violently fluctuating errors, systematic offset errors, oscillatory errors, and sudden errors;

[0026] Step S33: Error probability distribution calculation, specifically, using the probability density function to model the probability density distribution of the error type to obtain an error probability density function model, and by using the error probability density function model, according to the basic error and error type, perform error probability distribution calculation to obtain error pattern classification reference data.

[0027] Furthermore, in step S4, the intelligent factor identification is used to analyze the core factors affecting the prediction error and quantify the effect strength. Specifically, the intelligent factor identification is performed using an error dominant factor identification method combining modal decoupling and causal response analysis to obtain error effect factor identification data, including the following steps:

[0028] Step S41: constructing a disturbance response tensor, specifically, based on the error pattern classification reference data and the reconstructed power prediction curve, introducing a unit derivative disturbance into the reconstructed power prediction curve to construct the disturbance response tensor and obtain the original disturbance response tensor;

[0029] Step S42: local response tensor calculation, specifically, calculating the average time response of each error factor based on the original disturbance response tensor by using a time window sliding method to obtain a local sensitive response tensor;

[0030] Step S43: Calculating the heterogeneity of error influencing factors, specifically, calculating the heterogeneity of error influencing factors based on the error type and the improved response dispersion calculation method combined with the error type, to obtain the error response heterogeneity characteristics;

[0031] Step S44: Screening of dominant factors affecting errors, specifically, calculating the average response intensity of each error factor under the error type based on the heterogeneity characteristics of the error response, and screening the dominant factors affecting errors by constructing a dominant scoring index function to obtain dominant screening index characteristics, and based on the dominant screening index characteristics, screening the dominant factors affecting errors from the original tensor of the disturbance response to obtain error factor identification data.

[0032] Furthermore, in step S5, the correction suggestion is generated to provide prediction correction optimization measures, specifically, prediction correction suggestions are generated based on the error effect factor identification data and the reconstructed power prediction curve to obtain structured prediction correction reference data.

[0033] The present invention provides an intelligent new energy power generation power forecast curve decomposition and analysis system, comprising: a data collection module, a power curve decomposition module, an error distribution analysis module, an intelligent factor identification module, and a correction suggestion generation module;

[0034] The data collection module is used to collect data, obtain a standardized time-series power generation prediction data set through data collection, and send the standardized time-series power generation prediction data set to the power curve decomposition module, the error distribution analysis module and the intelligent factor identification module;

[0035] The power curve decomposition module is used to decompose the power curve, obtain a reconstructed power prediction curve through the power curve decomposition, and send the reconstructed power prediction curve to the error distribution analysis module, the intelligent factor identification module and the correction suggestion generation module;

[0036] The error distribution analysis module is used for error distribution analysis, obtains error pattern classification reference data through error distribution analysis, and sends the error pattern classification reference data to the intelligent factor identification module;

[0037] The intelligent factor identification module is used for intelligent factor identification, obtains error effect factor identification data through intelligent factor identification, and sends the error effect factor identification data to the correction suggestion generation module;

[0038] The correction suggestion generation module is used to generate correction suggestions, and obtains structured prediction correction reference data through correction suggestion generation.

[0039] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0040] (1) In the process of decomposition and analysis of the forecast curve of renewable energy power generation, existing methods often only focus on the accuracy of the results, but lack in-depth analysis of the source of errors. Once the forecast curve deviates, it is difficult to determine whether it is due to insufficient adaptability of the model itself or fluctuations or missing input data, resulting in a large degree of blindness in the model tuning process. This solution creatively adopts a phased integrated method that combines power curve decomposition, error distribution analysis, and intelligent factor identification to systematically analyze the causes of forecast errors and provide targeted correction suggestions, thereby improving the accuracy and reliability of the forecast model.

[0041] (2) In order to address the technical issues that exist in the existing power curve decomposition process, it is difficult to take into account both multi-scale changes and temporal dynamics, resulting in unstable decomposition results and a lack of physical interpretation capabilities. Single-scale or fixed-mode decomposition methods tend to ignore short-term disturbances or long-term trend changes, thereby affecting the learning effect of subsequent models, this scheme creatively adopts a multi-scale spatiotemporal coupled power curve decomposition method to perform power curve decomposition, thereby generating a more physically meaningful and stable reconstructed prediction curve, and significantly improving the adaptability of the prediction model to complex spatiotemporal characteristics.

[0042] (3) In the existing error factor analysis methods, the identification of error source factors usually remains at the level of statistical correlation analysis, lacking systematic modeling of causal relationships. In actual applications, errors are often caused by the combined action of multiple factors with complex interactive relationships. Simply relying on significance ranking or correlation coefficients cannot accurately identify the key factors that truly dominate the errors. This solution creatively adopts an error dominant factor identification method that combines modal decoupling and causal response analysis to perform intelligent factor identification, achieving accurate identification and quantitative analysis of error influencing factors, and providing strong support for the optimization of prediction models. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic flow chart of an intelligent new energy power generation prediction curve decomposition and analysis method provided by the present invention;

[0044] Figure 2 This is a schematic diagram of an intelligent new energy power generation prediction curve decomposition and analysis system provided by the present invention;

[0045] Figure 3 Schematic diagram of the process of step S2;

[0046] Figure 4 Schematic diagram of the process of step S3;

[0047] Figure 5 Schematic diagram of the process of step S4.

[0048] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0050] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0051] Example 1, see Figure 1 The present invention provides an intelligent new energy power generation prediction curve decomposition and analysis method, which includes the following steps:

[0052] Step S1: data collection;

[0053] Step S2: power curve decomposition;

[0054] Step S3: error distribution analysis;

[0055] Step S4: intelligent factor identification;

[0056] Step S5: generating correction suggestions;

[0057] By performing the above operations, in the process of decomposition and analysis of the new energy power generation power prediction curve, the existing methods often only focus on the accuracy of the results, lack of in-depth analysis of the source of errors. Once the prediction curve deviates, it is difficult to judge whether it is due to insufficient adaptability of the model itself or fluctuations or missing input data, resulting in a large degree of blindness in the model tuning process. The technical problem is solved. This solution creatively adopts a phased integration method that combines power curve decomposition, error distribution analysis and intelligent factor identification, realizes the systematic analysis of the causes of prediction errors, and provides targeted correction suggestions, thereby improving the accuracy and reliability of the prediction model.

[0058] Example 2, see Figure 1,This embodiment is based on the above embodiment. In step S1, the data collection is specifically to obtain a standardized time-series power generation prediction data set through multi-source power generation data collection and data optimization processing;

[0059] The multi-source power generation data collection specifically includes collecting new energy power generation power curves, meteorological data, operating condition data and geographic time and space information;

[0060] The new energy power generation curve includes a power generation prediction curve and an actual power generation curve;

[0061] The data optimization processing includes data time series alignment, outlier removal, data standardization and error pattern annotation;

[0062] The error pattern marking is to pre-mark the predicted error type by pre-defining the error type, including excessive error, too low error, violent fluctuation error, system offset error, oscillation error and mutation error;

[0063] The standardized time-series power generation prediction data set includes power generation prediction input samples, new energy power generation curves, error annotation data, space annotation data, and time annotation data.

[0064] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S2, the power curve decomposition is specifically performed by using a multi-scale spatiotemporal coupling power curve decomposition method to decompose the power curve to obtain a reconstructed power prediction curve, including the following steps:

[0065] Step S21: Multiscale variational modal decomposition, specifically, by introducing a multiscale coupling regularization term on the basis of standard variational modal decomposition, constructing an improved optimization objective function, and performing multiscale decomposition on the renewable energy power generation curve in the standardized time-series power generation prediction dataset to obtain a curve modal component set;

[0066] The calculation formula of the optimization objective function is:

[0067] ;

[0068] ;

[0069] Where, is the optimization objective function, is the objective function, u k is the kth modal component of the curve, k is the first index of the modal component, K is the total number of modal components, is the center frequency of the kth modal component of the curve, is the partial derivative symbol, ||·||2 is the L2 normal form symbol, t is the time point index, is the Dirac function, j is the imaginary unit, π is the circumference of a circle, and u k (t) is the value of the kth modal component of the curve at the tth time point, e is the base of the exponential function, is the multiscale coupling regularization term, is the regularization strength coefficient with a value greater than 0, u k+1 (t) is the value of the k+1th curve modal component at the tth time point, is a scale attenuation factor with a value range of (0,1);

[0070] Step S22: spatial feature extraction, specifically, constructing each curve modal component into a three-dimensional space-time tensor according to time, spatial position and modal dimension, and then performing low-rank decomposition on the three-dimensional space-time tensor using the tensor singular value decomposition method to extract the main modal components of space-time coupling and generate a low-rank reconstruction tensor;

[0071] Step S23: Modal adaptive update, specifically combining the sliding window mechanism and the recursive weighting strategy to construct a modal component dynamic update algorithm, update the spatiotemporal coupled main modal components, and obtain the dynamic modal components. The calculation formula is:

[0072] ;

[0073] Where, is the kth dynamic modal component in the cth time window, is the current window component weight with a value range of (0,1), which is used to control the fusion ratio of new and old information. is the spatiotemporal coupling main modal component of the current window, specifically the kth spatiotemporal coupling main modal component in the cth time window, where c is the time window index. is the spatiotemporal coupling main modal component of the previous window, specifically the kth spatiotemporal coupling main modal component in the c-1th time window, It is the historical difference influence coefficient with a value greater than 0, which is used to control the decay rate of the historical difference influence;

[0074] Step S24: Power weighted fusion is used to fuse the error and structural stability information to reconstruct the prediction curve. Specifically, the local error is obtained by comparing the difference between the power prediction value and the actual power of the dynamic modal component at time point t, and the modal stability index is extracted from the low-rank reconstruction tensor; the local error and the modal stability index are combined to construct the normalized weight; finally, the dynamic modal components are weighted and summed according to the normalized weight to generate the reconstructed power prediction curve;

[0075] The calculation formula for constructing the normalized weight by combining the local error and the modal stability index is:

[0076] ;

[0077] Where w k (t) is the normalized weight of the kth dynamic modal component at time point t, exp(·) is the exponential function, is the weight attenuation factor, which is used to adjust the sensitivity of the error to the weight distribution. k (t) is the local error of the kth dynamic modal component at time t, s k (t) is the modal stability index of the kth dynamic modal component at time point t, h is the second index of the modal component, err h (t) is the local error of the hth dynamic modal component at time t, s h (t) is the modal stability index of the hth dynamic modal component at time point t;

[0078] The calculation formula for weighted summation of dynamic modal components is:

[0079] ;

[0080] Where, is the predicted value of the reconstruction power at time point t, is the value of the kth dynamic modal component in the cth time window at the tth time point;

[0081] By performing the above operations, we can address the technical problems in the existing power curve decomposition process, such as the difficulty in taking into account both multi-scale changes and temporal dynamics at the same time, which leads to unstable decomposition results and lack of physical interpretation ability. Single-scale or fixed-mode decomposition methods easily ignore short-term disturbances or long-term trend changes, thereby affecting the learning effect of subsequent models. This solution creatively adopts a multi-scale spatiotemporal coupling power curve decomposition method to perform power curve decomposition, thereby generating a reconstructed prediction curve with greater physical significance and stability, and significantly improving the adaptability of the prediction model to complex spatiotemporal characteristics.

[0082] Example 4, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S3, the error distribution analysis is used to clarify the distribution type and cause of the prediction error. Specifically, the error statistical calculation method combined with cluster modeling is used to perform error distribution analysis to obtain error pattern classification reference data, including the following steps:

[0083] Step S31: basic error calculation, specifically, based on the reconstructed power prediction curve, by comparing the difference between the reconstructed power prediction value and the actual power at each time point t to obtain the basic error;

[0084] Step S32: Error cluster identification, specifically using a density peak clustering algorithm to cluster the basic errors, extract the latent class structure of the errors, and generate error types;

[0085] The error types include excessively high errors, excessively low errors, violently fluctuating errors, systematic offset errors, oscillatory errors, and sudden errors;

[0086] Step S33: Error probability distribution calculation, specifically, using the probability density function to model the probability density distribution of the error type to obtain an error probability density function model, and by using the error probability density function model, according to the basic error and error type, perform error probability distribution calculation to obtain error pattern classification reference data.

[0087] Example 5, see Figure 1 and Figure 5 This embodiment is based on the above embodiment. In step S4, the intelligent factor identification is used to analyze the core factors affecting the prediction error and quantify the effect strength. Specifically, the error dominant factor identification method combining modal decoupling and causal response analysis is used to perform intelligent factor identification to obtain error effect factor identification data, including the following steps:

[0088] Step S41: constructing a disturbance response tensor, specifically, based on the error pattern classification reference data and the reconstructed power prediction curve, introducing a unit derivative disturbance into the reconstructed power prediction curve to construct a disturbance response tensor, and obtaining the original disturbance response tensor. The calculation formula is:

[0089] ;

[0090] Where X is the original tensor of the disturbance response, which is used to represent the set of error factors, R is the dimension space identifier, T is the time window length, F is the total number of input factors of the reconstructed power prediction curve, and M is the total number of output modes of the power generation prediction curve.

[0091] Step S42: Calculate the local response tensor, specifically by using the time window sliding method, according to the original disturbance response tensor, calculate the average time response of each error factor, and obtain the local sensitive response tensor. The calculation formula is:

[0092] ;

[0093] Where, is the local sensitive response tensor, i is the power generation prediction input sample index, f is the error factor index, is the time step index, W is the width of the sliding time window, is the i-th power generation prediction input sample in the The predicted power response output for the time step, is the i-th power generation prediction input sample in the The f-th error contribution factor at time step, It is The gradient response value of the error factor at time step f is used to indicate the sensitivity to disturbance;

[0094] Step S43: Calculate the heterogeneity of the error influencing factors. Specifically, based on the error type, the error influencing factors are calculated by combining the improved response dispersion calculation method with the error type to obtain the error response heterogeneity characteristics. The calculation formula is:

[0095] ;

[0096] Where, The fth error factor corresponds to the The response time heterogeneity index of each error type is used to represent the error response heterogeneity characteristics. is the error type index, It is The total number of power generation prediction input samples of each error type, It is The power generation prediction input sample set of error types, Var t (·) is the time variance calculation function, which is used to represent the volatility of the time series response;

[0097] Step S44: Screening the dominant factors affecting the error, specifically, calculating the average response strength of each error factor under the error type based on the heterogeneity characteristics of the error response, and screening the dominant factors affecting the error by constructing a dominant scoring index function to obtain dominant screening index characteristics, and based on the dominant screening index characteristics, screening the dominant factors affecting the error from the original tensor of the disturbance response to obtain error factor identification data;

[0098] The dominant scoring index function is constructed to screen the dominant factors affecting the error, and the calculation formula for the dominant screening index characteristics is obtained as follows:

[0099] ;

[0100] Where, The fth error factor corresponds to the The dominant screening indicator characteristics of each error type are used for sorting and screening. The fth error factor corresponds to the The average response strength of each error type is calculated by calculating the local sensitive response tensor The average value of is the time heterogeneity weight;

[0101] By performing the above operations, in order to address the technical problem that in existing error factor analysis methods, the identification of error source factors usually remains at the level of statistical correlation analysis, and lacks systematic modeling of causal relationships, while in actual applications, errors are often caused by the joint action of multiple factors with complex interactive relationships, and simply relying on significance ranking or correlation coefficients cannot accurately identify the key factors that truly dominate the errors. This solution creatively adopts an error dominant factor identification method that combines modal decoupling and causal response analysis to perform intelligent factor identification, thereby achieving accurate identification and quantitative analysis of error influencing factors, and providing strong support for the optimization of prediction models.

[0102] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the correction suggestion is generated to provide prediction correction optimization measures, specifically, prediction correction suggestions are generated based on the error factor identification data and the reconstructed power prediction curve to obtain structured prediction correction reference data.

[0103] Example 7, see Figure 2 This embodiment is based on the above embodiment. The present invention provides an intelligent new energy power generation power forecast curve decomposition and analysis system, including: a data collection module, a power curve decomposition module, an error distribution analysis module, an intelligent factor identification module and a correction suggestion generation module;

[0104] The data collection module is used to collect data, obtain a standardized time-series power generation prediction data set through data collection, and send the standardized time-series power generation prediction data set to the power curve decomposition module, the error distribution analysis module and the intelligent factor identification module;

[0105] The power curve decomposition module is used to decompose the power curve, obtain a reconstructed power prediction curve through the power curve decomposition, and send the reconstructed power prediction curve to the error distribution analysis module, the intelligent factor identification module and the correction suggestion generation module;

[0106] The error distribution analysis module is used for error distribution analysis, obtains error pattern classification reference data through error distribution analysis, and sends the error pattern classification reference data to the intelligent factor identification module;

[0107] The intelligent factor identification module is used for intelligent factor identification, obtains error effect factor identification data through intelligent factor identification, and sends the error effect factor identification data to the correction suggestion generation module;

[0108] The correction suggestion generation module is used to generate correction suggestions, and obtains structured prediction correction reference data through correction suggestion generation.

[0109] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0110] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0111] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An intelligent new energy power generation prediction curve decomposition and analysis method, characterized by: The method comprises the following steps: Step S1: Data collection, specifically obtaining a standardized time-series power generation prediction data set through multi-source power generation data collection and data optimization processing; Step S2: power curve decomposition, specifically using a multi-scale spatiotemporal coupled power curve decomposition method to decompose the power curve and obtain a reconstructed power prediction curve, including the following steps: step S21: multi-scale variational mode decomposition; step S22: spatial feature extraction; step S23: mode adaptive update; step S24: power weighted fusion; Step S3: Error distribution analysis, used to clarify the distribution type and cause of the prediction error. Specifically, an error statistical calculation method combined with cluster modeling is used to perform error distribution analysis to obtain error pattern classification reference data. Step S4: Intelligent factor identification, which is used to analyze the core factors affecting the prediction error and quantify the effect strength. Specifically, the error dominant factor identification method combining modal decoupling and causal response analysis is used to perform intelligent factor identification and obtain error effect factor identification data, including the following steps: Step S41: Constructing the disturbance response tensor; Step S42: Calculating the local response tensor; Step S43: Calculating the heterogeneity of the error influencing factors; Step S44: Screening the error influencing dominant factors; Step S5: Generate correction suggestions and obtain structured prediction correction reference data.

2. The intelligent new energy power generation prediction curve decomposition and analysis method according to claim 1 is characterized by: In step S21, the multi-scale variational modal decomposition is specifically performed by introducing a multi-scale coupling regularization term on the basis of standard variational modal decomposition to construct an improved optimization objective function, and performing multi-scale decomposition on the new energy power generation curve in the standardized time-series power generation prediction data set to obtain a curve modal component set; In step S22, the spatial feature extraction is specifically performed by constructing each curve modal component into a three-dimensional space-time tensor according to time, spatial position and modal dimension, and then performing low-rank decomposition on the three-dimensional space-time tensor using the tensor singular value decomposition method to extract the main modal components of space-time coupling and generate a low-rank reconstruction tensor; In step S23, the modal adaptive update is specifically to construct a modal component dynamic update algorithm by combining a sliding window mechanism and a recursive weighting strategy, and to update the spatiotemporal coupled main modal components to obtain dynamic modal components; In step S24, the power weighted fusion is used to fuse the error and structural stability information to reconstruct the prediction curve. Specifically, the local error is obtained by comparing the difference between the power prediction value and the actual power of the dynamic modal component at the time point t, and the modal stability index is extracted from the low-rank reconstruction tensor; the local error and the modal stability index are combined to construct the normalized weight; finally, the dynamic modal components are weighted and summed according to the normalized weight to generate the reconstructed power prediction curve.

3. The intelligent new energy power generation prediction curve decomposition and analysis method according to claim 2 is characterized by: In step S41, the disturbance response tensor is constructed by introducing a unit derivative disturbance into the reconstructed power prediction curve based on the error pattern classification reference data and the reconstructed power prediction curve to construct the disturbance response tensor and obtain the disturbance response original tensor; In step S42, the local response tensor is calculated by using a time window sliding method, based on the original disturbance response tensor, calculating the average time response of each error factor to obtain a local sensitive response tensor; In step S43, the error influencing factor heterogeneity calculation is specifically performed based on the error type, by combining the improved response dispersion calculation method with the error type, to calculate the heterogeneity of the error effect factor, and obtain the error response heterogeneity characteristics; In step S44, the error-influencing dominant factors are screened, specifically by calculating the average response intensity of each error factor under the error type based on the error response heterogeneity characteristics, and screening the error-influencing dominant factors by constructing a dominant scoring index function to obtain dominant screening index characteristics, and based on the dominant screening index characteristics, screening the error-influencing dominant factors from the disturbance response original tensor to obtain error factor identification data.

4. The intelligent new energy power generation prediction curve decomposition and analysis method according to claim 3 is characterized by: In step S3, the error distribution analysis is specifically performed using an error statistical calculation method combined with cluster modeling to obtain error pattern classification reference data, including the following steps: Step S31: basic error calculation, specifically, based on the reconstructed power prediction curve, by comparing the difference between the reconstructed power prediction value and the actual power at each time point t to obtain the basic error; Step S32: Error cluster identification, specifically using a density peak clustering algorithm to cluster the basic errors, extract the latent class structure of the errors, and generate error types; The error types include excessively high errors, excessively low errors, violently fluctuating errors, systematic offset errors, oscillatory errors, and sudden errors; Step S33: Error probability distribution calculation, specifically, using the probability density function to model the probability density distribution of the error type to obtain an error probability density function model, and by using the error probability density function model, according to the basic error and error type, perform error probability distribution calculation to obtain error pattern classification reference data.

5. The intelligent new energy power generation prediction curve decomposition and analysis method according to claim 4 is characterized by: In step S5, the correction suggestion is generated to provide prediction correction optimization measures, specifically, prediction correction suggestions are generated based on the error effect factor identification data and the reconstructed power prediction curve to obtain structured prediction correction reference data.

6. The intelligent new energy power generation prediction curve decomposition and analysis method according to claim 5 is characterized by: In step S1, the data collection is specifically to obtain a standardized time-series power generation prediction data set through multi-source power generation data collection and data optimization processing; The multi-source power generation data collection specifically includes collecting new energy power generation power curves, meteorological data, operating condition data and geographic time and space information; The new energy power generation curve includes a power generation prediction curve and an actual power generation curve; The data optimization processing includes data time series alignment, outlier removal, data standardization and error pattern annotation; The standardized time-series power generation prediction data set includes power generation prediction input samples, new energy power generation curves, error annotation data, space annotation data, and time annotation data.

7. An intelligent new energy power generation prediction curve decomposition and analysis system, used to implement an intelligent new energy power generation prediction curve decomposition and analysis method according to any one of claims 1 to 6, characterized in that: It includes a data collection module, a power curve decomposition module, an error distribution analysis module, an intelligent factor identification module and a correction suggestion generation module.

8. The intelligent new energy power generation prediction curve decomposition and analysis system according to claim 7 is characterized by: The data collection module is used to collect data, obtain a standardized time-series power generation prediction data set through data collection, and send the standardized time-series power generation prediction data set to the power curve decomposition module, the error distribution analysis module and the intelligent factor identification module; The power curve decomposition module is used to decompose the power curve, obtain a reconstructed power prediction curve through the power curve decomposition, and send the reconstructed power prediction curve to the error distribution analysis module, the intelligent factor identification module and the correction suggestion generation module; The error distribution analysis module is used for error distribution analysis, obtains error pattern classification reference data through error distribution analysis, and sends the error pattern classification reference data to the intelligent factor identification module; The intelligent factor identification module is used for intelligent factor identification, obtains error effect factor identification data through intelligent factor identification, and sends the error effect factor identification data to the correction suggestion generation module; The correction suggestion generation module is used to generate correction suggestions, and obtains structured prediction correction reference data through correction suggestion generation.

Images