An integrated optimization method for new energy power generation prediction

Abstract

The present application relates to the technical field of new energy power generation prediction, and specifically provides a kind of integrated optimization new energy power generation prediction method, comprising: obtaining the time series data of the power generation of new energy station in historical time period;Determine the category to which the time series data belongs;Time series data is input into the integrated prediction model of corresponding category, and the prediction result corresponding to each prediction model in the integrated prediction model is obtained in each prediction model in the integrated prediction model in the to-be-measured time period;Wherein, the prediction model includes global prediction model and local prediction model;The average of each prediction result is calculated to obtain the prediction result of the power generation of new energy station in the to-be-measured time period.The present application provides a kind of integrated optimization new energy power generation prediction method, which can improve the prediction accuracy of new energy power generation and improve the synergistic advantage of multiple models.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation prediction technology, and in particular to an integrated and optimized method for predicting new energy power generation. Background Technology

[0002] With the increasing penetration of renewable energy sources such as wind and solar power into the power system, high-precision forecasting of new energy power generation has become a key link in realizing the optimized dispatch of smart grids. However, current new energy forecasting technologies still face the following two prominent technical bottlenecks.

[0003] First, due to the significant spatiotemporal heterogeneity of new energy power generation, different prediction models exhibit significant differences in performance across different time series. Traditional fixed-weight integration or simple arithmetic averaging strategies are difficult to achieve dynamic adaptation between models, resulting in limited prediction accuracy.

[0004] Secondly, existing prediction frameworks lack an intelligent optimal model selection mechanism, and cannot adaptively call the most suitable prediction model based on real-time data characteristics, which restricts the full realization of the advantages of multi-model collaboration.

[0005] There is currently no effective solution to the problem that the prediction accuracy of new energy power generation in related technologies is limited and that the most suitable prediction model cannot be adaptively called. Summary of the Invention

[0006] The present invention provides an integrated and optimized method for predicting new energy power generation, which at least solves the problems of limited prediction accuracy and insufficient synergistic advantages of multiple models.

[0007] To achieve the above objectives, the present invention provides the following technical solution.

[0008] This invention provides an integrated and optimized method for predicting renewable energy power generation, comprising: acquiring time-series data of power generation from renewable energy power plants over a historical period; determining the category to which the time-series data belongs; inputting the time-series data into an integrated prediction model corresponding to the category, and obtaining a prediction result for each prediction model in the time period to be measured from each prediction model in the integrated prediction model; wherein the prediction model includes a global prediction model and a local prediction model; the global prediction model is trained using training data from multiple categories; the local prediction model is trained using training data corresponding to a single category; the prediction models included in the integrated prediction models of different categories are trained and tested based on training data from different categories; and calculating the average value of each prediction result to obtain the prediction result of the power generation from renewable energy power plants in the time period to be measured.

[0009] Preferably, determining the category to which the time series data belongs includes: extracting statistical features of the time series data; wherein the statistical features include: mean, variance, maximum value, periodicity and / or autocorrelation coefficient; determining the data subset to which the time series data belongs based on the statistical features to obtain the category; wherein the data subset is obtained by classifying based on the statistical features of training data; the training data includes: historical time series data of power generation of various new energy power plants.

[0010] Preferably, determining the data subset to which the time series data belongs based on the statistical features includes: collecting historical time series data of power generation from various new energy power plants as training data; extracting statistical features from the training data, classifying the training data based on the similarity between the statistical features to obtain multiple data subsets; and determining the data subset to which the time series data belongs based on the statistical features of the time series data.

[0011] Preferably, before inputting the time series data into the ensemble prediction model for the corresponding category, the prediction method further includes: constructing a global prediction model and multiple local prediction models to obtain various prediction models; inputting the training data for each category into each prediction model, and outputting the corresponding prediction results through each prediction model; ranking the performance of each prediction model based on the prediction error of each prediction result; gradually integrating the second-best-ranked prediction model with the best-ranked prediction model in a top-down order to obtain the prediction model for each integration and the prediction error of the corresponding output prediction result; until the prediction error no longer decreases, the integration terminates, and the ensemble prediction model corresponding to each category is obtained.

[0012] Preferably, constructing a global prediction model includes: training a neural network model using training data from each category; stopping training when the training conditions are met, thus obtaining a global prediction model.

[0013] Preferably, constructing a local prediction model includes: training the neural network model separately using training data of different categories; stopping the training when the training conditions are met, and obtaining each local prediction model.

[0014] Preferably, the merits of each prediction model are ranked based on the prediction error of each prediction result, including: calculating the prediction error of each prediction result on the validation set using a loss function; wherein the validation set includes: historical time-series data of power generation of each new energy power station; and ranking the prediction models in order of prediction error from smallest to largest according to the magnitude of the prediction error of each prediction result.

[0015] Preferably, the second-best prediction model is integrated with the best prediction model in a top-down order to obtain the prediction model for each integration and the prediction error of the corresponding output prediction result. This includes: using the best prediction model as the initial model; integrating the second-best prediction model with the initial model in a top-down order to obtain the prediction model for each integration; calculating the prediction result of the prediction model for each integration; and calculating the corresponding prediction error based on the prediction result.

[0016] Preferably, calculating the prediction result of each integrated prediction model and calculating the corresponding prediction error based on the prediction result includes: calculating the average value of the prediction results output by each prediction model in each integrated prediction model as the prediction result of each integrated prediction model; and calculating the prediction error of each prediction result on the validation set as the prediction error of each integrated prediction model.

[0017] Preferably, the integration process terminates when the prediction error no longer decreases, resulting in an integrated prediction model for each category. This includes: if the prediction error of the currently integrated prediction model is less than the prediction error of the previously integrated prediction model, then the next best-ranked prediction model is integrated; if the prediction error of the currently integrated prediction model is greater than or equal to the prediction error of the previously integrated prediction model, then the integration terminates, resulting in an integrated prediction model for each category.

[0018] This invention provides an integrated and optimized method for predicting renewable energy power generation. It categorizes time-series data of power generation from renewable energy plants and constructs corresponding integrated prediction models for each category. The integrated prediction models then output prediction results, adapting to the differences in time-series data. Furthermore, the integrated prediction models constructed in this invention integrate the prediction errors of various prediction models for different categories of time-series data. The optimal model combination is selected during model integration. When using this method to predict the power generation of renewable energy plants, users only need to select the corresponding integrated prediction model based on the category of the current time-series data. This adaptively selects the most suitable prediction model, improving the prediction accuracy of the model under different time-series data. It avoids the problem of difficulty in adapting to dynamic models caused by fixed-weight integration or simple arithmetic averaging, while leveraging the advantages of different prediction models to further improve the prediction accuracy of the integrated prediction model. In addition, the prediction model of this invention includes a global prediction model and multiple local prediction models, preserving the stability of model prediction while improving local adaptability and prediction accuracy. This solves the problem of limited prediction accuracy for renewable energy power generation and the inability to adaptively select the most suitable prediction model. Attached Figure Description

[0019] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other embodiments based on these drawings without creative effort.

[0020] Figure 1 This is a schematic flowchart of an integrated and optimized method for predicting new energy power generation, which is an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0022] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0023] In related technologies, ensemble learning-based methods for predicting new energy power generation typically construct multiple prediction models based on the characteristics of different time series data of new energy power generation. Then, the prediction results of multiple prediction models are weighted and averaged or simply arithmetic averaged to obtain the final prediction result.

[0024] This approach often uses fixed weights for different prediction models, making it difficult to adapt to the differences in power generation data across different time series. This results in the integrated prediction model's prediction results lacking dynamic adaptability, thereby reducing the model's prediction accuracy.

[0025] Furthermore, the prediction methods in related technologies typically integrate fixed prediction models for different time series data, lacking an intelligent optimal model selection mechanism, which fails to leverage the advantages of different prediction models, resulting in insufficient prediction accuracy.

[0026] like Figure 1 As shown, in order to improve the prediction accuracy of new energy power generation and enhance the synergistic advantages of multiple models, an embodiment of the present invention provides an integrated and optimized method for predicting new energy power generation, including the following steps.

[0027] Step S1: Obtain time series data of power generation of new energy power plants for historical time periods.

[0028] Step S2: Determine the category to which the time series data belongs.

[0029] Step S3: Input the time series data into the ensemble prediction model of the corresponding category. Each prediction model in the ensemble prediction model will generate a prediction result for the time period to be tested. The prediction model includes a global prediction model and a local prediction model. The global prediction model is trained using training data from multiple categories. The local prediction model is trained using training data from a single category. The prediction models included in the ensemble prediction models of different categories are trained and detected based on training data from different categories.

[0030] Step S4: Calculate the average value of each prediction result to obtain the prediction result of the power generation of the new energy power station during the test period.

[0031] Specifically, in step S1, the historical time period refers to the most recent period of time data in the past when it is necessary to predict the power generation of new energy sources. The time period to be measured in step S4 refers to the next time period after the historical time period.

[0032] The length of the historical time period is the same as the length of the time period to be measured.

[0033] The present invention provides an integrated and optimized method for predicting the power generation of new energy sources. This method classifies the time series data of power generation from new energy power plants and constructs corresponding integrated prediction models for different categories of time series data. The method then outputs the corresponding prediction results through the integrated prediction models, which can adapt to the differences in different time series data.

[0034] Meanwhile, the integrated prediction model constructed in this invention integrates the prediction errors of various prediction models based on different categories of time series data. The optimal model combination method has been selected during model integration. When using this prediction method to predict the power generation of new energy power plants, it is only necessary to select the corresponding integrated prediction model according to the category to which the current time series data belongs. The most suitable prediction model can be adaptively selected, thereby improving the prediction accuracy of the prediction model under different time series data. This avoids the problem of difficulty in adapting to dynamic adaptation between models caused by fixed weight integration or simple arithmetic average. At the same time, it leverages the advantages of different prediction models and can further improve the prediction accuracy of the integrated prediction model.

[0035] In addition, the prediction model of the present invention includes a global prediction model and multiple local prediction models, which not only retains the stability of the model prediction, but also improves local adaptability and improves prediction accuracy.

[0036] In step S1, the power generation data of new energy power plants for historical time periods are first collected; then, the time series data of power generation data is obtained by using a sliding window method.

[0037] The power generation data can come from data acquisition and monitoring systems (such as SCADA systems), electricity metering devices (such as smart meters), and energy management systems (such as EMS / BMS).

[0038] The length of the historical period can be selected according to actual needs. Preferably, the length of the historical period can be one day or one week, etc.

[0039] New energy power plants refer to facilities that generate electricity using renewable energy, mainly including: wind power plants, solar power plants, hydropower stations, biomass power stations, geothermal power stations, and ocean power stations.

[0040] In a preferred but non-limiting embodiment of the present invention, step S2 includes the following steps.

[0041] Step S21: Extract the statistical features of the time series data; wherein, the statistical features include: mean, variance, maximum value, periodicity and / or autocorrelation coefficient.

[0042] Step S22: Determine the data subset to which the time series data belongs based on statistical characteristics, and obtain the category to which it belongs; wherein, the data subset is obtained by classifying based on the statistical characteristics of the training data; the training data includes: historical time series data of power generation of each new energy power station.

[0043] The historical time-series data was extracted from the historical power generation data of various new energy power plants using a sliding window method.

[0044] In step S22, the training data can be classified using hierarchical clustering. Based on the similarity between the statistical features of each training data set, the training data is divided into several data subsets, and each data subset is a category.

[0045] Furthermore, the hierarchical clustering method adopted in this embodiment of the invention is preferably single-layer clustering, that is, the training data is directly divided into different cluster subsets of data after one clustering process.

[0046] Furthermore, in step S22, the data subset to which the time series data belongs is determined based on statistical characteristics, including the following steps.

[0047] Step S221: Collect historical time-series data of power generation from each new energy power station as training data.

[0048] Step S222: Extract statistical features from the training data, classify the training data based on the similarity between the statistical features, and obtain multiple data subsets.

[0049] Step S223: Based on the statistical characteristics of the time series data, determine the data subset to which the time series data belongs.

[0050] In the model training phase of this invention, statistical features are extracted from the collected historical time-series data. These features are then used to classify the historical time-series data, resulting in different data subsets. Based on the statistical features of the collected historical time-series data, the subsets to which the time-series data belongs are matched to determine its category. According to the category of the time-series data, the corresponding ensemble prediction model can be adaptively invoked, thus adapting to the characteristics of the current time-series data and meeting the prediction accuracy requirements.

[0051] In a preferred but non-limiting embodiment of the present invention, the prediction method further includes the following steps before step S3.

[0052] Step S01: Construct a global prediction model and multiple local prediction models to obtain each prediction model.

[0053] Step S02: Input the training data for each category into each prediction model, and output the corresponding prediction results through each prediction model.

[0054] Step S03: Based on the prediction error of each prediction result, rank the superiority and inferiority of each prediction model.

[0055] Step S04: Following a top-to-bottom order, the second-best prediction model is integrated with the best prediction model step by step to obtain the prediction model for each integration and the prediction error of the corresponding output prediction result.

[0056] Step S05 continues until the prediction error no longer decreases, at which point the integration terminates, yielding the integrated prediction model for each category.

[0057] In this embodiment of the invention, during model training, the optimal combination of prediction models is selected for each category of training data to obtain ensemble prediction models corresponding to different categories. Compared to selecting the optimal model ensemble method based on the current time series data during application, this embodiment of the invention requires less computation, has lower model complexity, and can obtain prediction results faster during prediction, thus improving prediction efficiency while meeting prediction accuracy requirements.

[0058] Step S01, which involves constructing a global prediction model, includes training the neural network model using training data from each category. Training stops when the training conditions are met, resulting in a global prediction model.

[0059] Furthermore, the training objective of the global prediction model can be expressed as the following formula.

[0060] (1).

[0061] in, Represents the training data sample matrix. Represents the target matrix for prediction. Represents the parameters of the global prediction model. This represents the global prediction model function.

[0062] Step S01 involves constructing local prediction models, including training the neural network model using training data of different categories. Training stops when the training conditions are met, resulting in the individual local prediction models.

[0063] Furthermore, the training objective of the local prediction model can be expressed as the following formula.

[0064] (2).

[0065] in, Indicates the first The training data sample matrix of the i-th cluster in layer 1. Indicates the first The predicted target matrix of the i-th cluster in layer 1. Indicates the first The local prediction model parameters for the i-th cluster in layer 1; Indicates the first The local prediction function corresponding to the i-th cluster in the layer.

[0066] The training termination condition for the global and local prediction models can be a preset upper limit on the number of training iterations. This upper limit can be adjusted based on the data size and complexity provided in the actual project. Training stops when the preset upper limit is reached, thus obtaining the corresponding global and local prediction models.

[0067] The global prediction model constructed in this embodiment of the invention has relatively stable prediction capabilities because it is trained using training data from various categories. However, it may have insufficient local adaptability when dealing with specific fluctuation patterns of individual new energy power plants.

[0068] Therefore, the embodiments of the present invention also construct multiple local prediction models. By using different categories of training data to train the neural network model respectively, multiple local prediction models are obtained, which can optimize the training for different categories of training data, thereby improving the local adaptability and prediction accuracy of the model.

[0069] The global prediction model and each local prediction model constructed by the embodiments of the present invention are both a prediction model. The candidate model set including all prediction models is constructed as follows.

[0070] (3).

[0071] Among them, represents the candidate model set; represents the i-th prediction model trained for the time series , C represents the number of prediction models, 0 < C < N, and N is a natural number.

[0072] In a preferred but non-limiting embodiment of the present invention, in step S03, based on the prediction errors of each prediction result, the advantages and disadvantages of each prediction model are ranked, including the following steps.

[0073] Step S031, use the loss function to calculate the prediction error of each prediction result on the validation set; where the validation set includes: historical time series data of the power generation of each new energy power station.

[0074] Step S032, according to the magnitude of the prediction error of each prediction result, rank the advantages and disadvantages of the prediction models in ascending order of error.

[0075] Specifically, the validation set is part of the historical time series data extracted from the historical time series data of the power generation of each new energy power station collected.

[0076] Using the loss function to calculate the prediction error of each prediction result on the validation set can be expressed by the following formula. (4).

[0077] Among them, represents the prediction error of the i-th prediction model trained for the time series ; represents the loss function of the i-th prediction model trained for the time series and .

[0078] Specifically, the loss function can be the mean square error function (MSE) or the mean absolute error function (MAE).

[0079] The prediction model with a smaller prediction error and a higher ranking is better.

[0080] The ranking of prediction models based on their error from smallest to largest can be expressed by the following formula.

[0081] (5).

[0082] in, Indicates time series A set of prediction models sorted by prediction error from smallest to largest; This indicates that the top-ranked prediction model has the smallest prediction error.

[0083] This invention utilizes a global prediction model and various local prediction models to obtain corresponding prediction results for each category's training time series. The prediction models are then ranked based on their prediction errors, enabling the selection of the optimal model combination for different time series data. Furthermore, by using both the global and local prediction models as candidate models, both prediction stability and local prediction accuracy are guaranteed.

[0084] In a preferred but non-limiting embodiment of the present invention, in step S04, the second-best ranked prediction model is integrated with the best ranked prediction model in a top-to-bottom order to obtain the prediction model and the prediction error of the corresponding output prediction result for each integration, including the following steps.

[0085] Step S041: Use the best-ranked prediction model as the initial model.

[0086] Step S042: Following a top-down order, the second-best prediction model is integrated with the initial model step by step to obtain the prediction model for each integration.

[0087] Step S043: Calculate the prediction result of each integrated prediction model and calculate the corresponding prediction error based on the prediction result.

[0088] Specifically, in step S041, the optimal initial model can be represented as: ,in, Indicates the initial model. This indicates the top-ranked prediction model.

[0089] In step S042, the combination of the second-best ranked prediction model and the initial model can be expressed as the following formula.

[0090] (6).

[0091] in, This represents the prediction model obtained from the first integration. Indicates the initial model. This indicates the prediction model that ranks second.

[0092] Furthermore, in step S043, the prediction result of each integrated prediction model is calculated, and the corresponding prediction error is calculated based on the prediction result, including the following steps.

[0093] Step S0431: Calculate the average value of the prediction results output by each prediction model in each integrated prediction model, and use it as the prediction result of each integrated prediction model.

[0094] Step S0432: Calculate the prediction error of each prediction result on the validation set, which is used as the prediction error of the prediction model for each integration.

[0095] Step S0431 can be expressed as the following formula.

[0096] (7).

[0097] in, Indicates time series The prediction result of the prediction model in the nth ensemble, where n represents the number of ensembles; Indicates time series The prediction model obtained from the nth ensemble; Indicates the first element in the integrated prediction model. The prediction results of the prediction model This represents the sum of the prediction results output by each prediction model.

[0098] It is understood that the way this invention integrates various prediction models is by summing the prediction results of each prediction model and taking the average value. By integrating the models, the errors of multiple prediction models can be canceled out and a more stable prediction result can be obtained.

[0099] Furthermore, in a preferred but non-limiting embodiment of the present invention, step S05 includes the following steps.

[0100] Step S051: If the prediction error of the currently integrated prediction model is less than the prediction error of the previously integrated prediction model, then continue to integrate the second-best ranking prediction model.

[0101] Step S052: If the prediction error of the current integrated prediction model is greater than or equal to the prediction error of the previous integrated prediction model, then the integration terminates, and the integrated prediction model corresponding to each category is obtained.

[0102] This invention, through the gradual integration of the second-best prediction model and the cessation of integration when the prediction error of the integrated prediction model no longer decreases, ensures that the prediction performance of the integrated prediction model is optimal for the current time series data. This not only fully leverages the advantages of each prediction model but also guarantees the accuracy of the prediction model.

[0103] Furthermore, in step S3, the corresponding prediction results are obtained from each prediction model in the integrated prediction model, which are the prediction results of the power generation of the new energy power station for a future period of time output by each prediction model.

[0104] In step S4, the average of each prediction result is calculated to obtain the final prediction result of the integrated prediction model. This prediction result represents the final prediction result of the integrated prediction model for the future time period.

[0105] Another embodiment of the present invention provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform the integrated and optimized new energy power generation prediction method of the present invention.

[0106] refer to Figure 2 The block diagram of an electronic device, representing an embodiment of the present invention, is an example of a hardware device applicable to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0107] like Figure 2 As shown, the electronic device includes a computing unit 101, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 102 or a computer program loaded from a storage unit 108 into a random access memory (RAM) 103. The RAM 103 may also store various programs and data required for the operation of the electronic device. The computing unit 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0108] Multiple components in the electronic device are connected to I / O interface 105, including: input unit 106, output unit 107, storage unit 108, and communication unit 109. Input unit 106 can be any type of device capable of inputting information into the electronic device. Input unit 106 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of the electronic device. Output unit 107 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 108 may include, but is not limited to, disks and optical discs. Communication unit 109 allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, and / or wireless communication transceivers, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0109] The computing unit 101 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 101 include, but are not limited to, CPUs, graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing units, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing unit 101 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention can be implemented as computer programs tangibly contained in a machine-readable medium, such as storage unit 108. In some embodiments, part or all of the computer program can be loaded and / or installed on an electronic device via ROM 102 and / or communication unit 109. In some embodiments, the computing unit 101 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0110] Computer programs for implementing the methods of embodiments of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0111] In the context of embodiments of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, or infrared systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0112] It should be noted that the term "comprising" and its variations used in the embodiments of this invention are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "a plurality" mentioned in the embodiments of this invention are illustrative and not restrictive, and those skilled in the art should understand that unless explicitly indicated otherwise in the context, they should be understood as "one or more".

[0113] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this invention are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0114] The steps described in the method embodiments provided by the present invention can be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0115] The term "embodiment" in this specification refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily imply the same embodiment, nor does it imply independence or alternativeity from other embodiments. The various embodiments in this specification are described in a related manner, with reference to each other for similar or identical parts. In particular, for apparatus, device, and system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, and relevant details are referred to in the description of the method embodiments.

[0116] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. An integrated optimization method for new energy power generation prediction, characterized in that, include: Obtain time-series data on the power generation of renewable energy power plants over historical periods; Determine the category to which the time series data belongs; Construct a global prediction model and multiple local prediction models to obtain various prediction models; The training data for each category is input into each prediction model, and each prediction model outputs the corresponding prediction result. Based on the prediction errors of each prediction result, the advantages and disadvantages of each prediction model are ranked. Use the best-ranked prediction model as the initial model; By adopting a top-down approach, the second-best prediction model is gradually integrated with the initial model to obtain the prediction model for each integration. Calculate the prediction results of the prediction model for each integration, and calculate the corresponding prediction error based on the prediction results; The ensemble continues until the prediction error no longer decreases, at which point the ensemble terminates, yielding the ensemble prediction model for each category. The time series data is input into the ensemble prediction model of the corresponding category. Each prediction model in the ensemble prediction model obtains the prediction result corresponding to each prediction model in the time period to be tested. The prediction model includes a global prediction model and a local prediction model. The global prediction model is trained using training data from multiple categories. The neural network model is trained using training data from each category to obtain the global prediction model. The local prediction model is trained using training data corresponding to a single category; different categories of training data are used to train the neural network model separately to obtain each of the local prediction models; the prediction models included in the integrated prediction model of different categories are trained and detected based on training data of different categories. The average value of each prediction result is calculated to obtain the predicted power generation of the new energy power station during the time period to be measured. 2.The method of claim 1, wherein, Determining the category to which the time series data belongs includes: Extract the statistical features of the time series data; wherein the statistical features include: mean, variance, maximum value, periodicity and / or autocorrelation coefficient; Based on the statistical characteristics, the data subset to which the time series data belongs is determined, and the category to which it belongs is obtained; wherein, the data subset is obtained by classifying based on the statistical characteristics of the training data; the training data includes: historical time series data of power generation of each new energy power station. 3.The method of claim 2, wherein, Determining the data subset to which the time series data belongs based on the statistical characteristics includes: Historical time-series data of power generation from various new energy power plants were collected as training data. Statistical features are extracted from the training data, and the training data is classified based on the similarity between the statistical features to obtain multiple data subsets; Based on the statistical characteristics of the time series data, the data subset to which the time series data belongs is determined. 4.The method of claim 1, wherein, Construct a global prediction model, including: The neural network model is trained using training data from various categories. Training stops when the training conditions are met, resulting in a global prediction model. 5.The method of claim 1, wherein, Constructing a local prediction model includes: The neural network model was trained using different categories of training data. Training stops when the training conditions are met, and the local prediction models are obtained. 6.The method of claim 1, wherein, Ranking the advantages and disadvantages of each prediction model based on the prediction errors of each prediction result, comprising: calculating the prediction error of each prediction result on the validation set using a loss function; wherein the validation set comprises: historical time series data of the power generation of each new energy station; According to the prediction error of each prediction result, the advantages and disadvantages of the prediction model are ranked in order of prediction error from small to large.

7. The method of claim 1, wherein, Calculate the prediction result of each integrated prediction model, and calculate the corresponding prediction error according to the prediction result, comprising: Calculate the average value of the prediction results output by each prediction model in each integrated prediction model as the prediction result of each integrated prediction model; Calculate the prediction error of each prediction result on the validation set as the prediction error of each integrated prediction model.

8. The method of claim 1, wherein, Until the prediction error no longer decreases, the integration is terminated, and the integrated prediction model corresponding to each category is obtained, comprising: If the prediction error of the current integrated prediction model is less than the prediction error of the last integrated prediction model, then continue to integrate the next optimal prediction model; If the prediction error of the current integrated prediction model is greater than or equal to the prediction error of the last integrated prediction model, then the integration is terminated, and the integrated prediction model corresponding to each category is obtained.

Images