New energy power generation prediction method combining global modeling and local optimization

Abstract

The invention relates to the technical field of new energy power generation prediction, and provides a new energy power generation prediction method combining global modeling and local optimization, and the method comprises the steps: obtaining the current time series data of a target new energy station; the current time sequence data is input into a prediction model to obtain corresponding prediction data, the prediction model is obtained by locally optimizing a global model based on a time sequence subset, and the time sequence subset is obtained by clustering analysis of historical time sequence data of a plurality of new energy stations including the target new energy station; the global model is obtained by training according to lagging features and category features of the historical time sequence data; and determining a new energy power generation prediction result based on the prediction data corresponding to the current time sequence data of each new energy station. The problems that the power generation time sequence of a plurality of new energy stations cannot be accurately predicted when new energy power generation prediction is carried out in related technologies, and the calculation efficiency cannot be improved while the prediction precision is ensured are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of renewable energy power generation prediction, and in particular to a renewable energy power generation prediction method combining global modeling and local optimization. Background Art

[0002] With the rapid development of renewable energy such as wind energy and solar energy, the number of new energy stations has greatly increased, and it is necessary to conduct new energy power generation forecasts.

[0003] Renewable energy generation forecasting methods based on large pre-trained models have promising application prospects, but these methods typically require separate modeling for each renewable energy site, resulting in high computational complexity, difficulty in scalability, and insufficient adaptability to renewable energy volatility and regional differences. Specifically, renewable energy generation forecasting methods based on deep learning suffer from insufficient prediction accuracy and generalization capabilities when simultaneously forecasting the generation time series of multiple renewable energy sites. Renewable energy generation forecasting methods based on transfer learning also suffer from low model adaptability and computational efficiency when dealing with diverse renewable energy sites and regional differences.

[0004] There is currently no effective solution to the problem of being unable to accurately predict the power generation time series of multiple new energy sites when using relevant technologies to predict new energy power generation, and to improve computing efficiency while ensuring prediction accuracy. Summary of the Invention

[0005] The embodiment of the present invention provides a new energy power generation prediction method that combines global modeling and local optimization, which at least solves the problem that related technologies cannot accurately predict the power generation time series of multiple new energy stations when performing new energy power generation prediction, and improves computing efficiency while ensuring prediction accuracy.

[0006] The present invention provides an embodiment of a new energy power generation prediction method that combines global modeling and local optimization, including: obtaining current time series data of a target new energy station; inputting the current time series data into a prediction model to obtain corresponding prediction data, wherein the prediction model is obtained by locally optimizing a global model based on a time series subset, the time series subset is obtained by clustering analysis of historical time series data of multiple new energy stations, the global model is trained based on the lag characteristics and category characteristics of the historical time series data of multiple new energy stations, and the target new energy station is a new energy station among the multiple new energy stations; and the new energy power generation prediction result is determined based on the prediction data corresponding to the current time series data of each new energy station.

[0007] The embodiment of the present invention provides a new energy power generation prediction method that combines global modeling and local optimization. Before obtaining the current time series data of the target new energy station, the method also includes: obtaining the unprocessed historical time series data of multiple new energy stations; preprocessing the unprocessed historical time series data to obtain historical time series data; extracting features from the historical time series data to obtain lag features and category features; training the deep learning big model based on the lag features and category features to obtain a global model, wherein the deep learning big model is a deep learning big model that adopts a neural-based analysis architecture, and the global model is the above-mentioned deep learning big model that achieves the preset learning goals.

[0008] The invention provides a new energy power generation prediction method combining global modeling and local optimization, which preprocesses the historical time series data to be processed to obtain the historical time series data, including: uniformly formatting the historical time series data to be processed to obtain formatted data; cleaning, denoising, outlier detection, and missing value filling processing of the formatted data in sequence to obtain preliminary processed data; and standardizing the preliminary processed data based on a standard score normalization algorithm or a minimum-maximum normalization algorithm to obtain the historical time series data.

[0009] The embodiment of the present invention provides a new energy power generation prediction method that combines global modeling and local optimization, which extracts features from historical time series data to obtain lag features and category features, including: extracting features from historical load data in historical time series data based on a rolling window method to obtain lag features; performing one-hot encoding conversion on time information in historical time series data to obtain time features; and determining meteorological features based on meteorological information in historical time series data; wherein category features include time features and meteorological features.

[0010] The invention provides a new energy power generation prediction method combining global modeling and local optimization, which trains a deep learning large model based on hysteresis features and category features to obtain a global model, including: and categorical features Construct the input matrix ; Create input matrix The mapping relationship between the target matrix and the target matrix, where the target matrix is ​​used to represent the renewable energy power generation within the preset prediction range; the average absolute error is lower than the preset threshold as the preset learning target, combined with L 1. Regularization control: optimize and train the model parameters of the deep learning large model based on the mapping relationship to obtain the global model.

[0011] The embodiment of the present invention provides a new energy power generation prediction method that combines global modeling and local optimization. Before inputting current time series data into the prediction model to obtain corresponding prediction data, the above method also includes: extracting statistical features and standardizing the historical time series data of multiple new energy stations to obtain statistical feature vectors with the same weight; clustering analysis of the statistical feature vectors based on the K-means clustering algorithm and the hierarchical clustering algorithm to obtain a time series subset; and locally optimizing the global model based on the time series subset to obtain a prediction model.

[0012] The invention provides a new energy power generation prediction method combining global modeling and local optimization, which performs cluster analysis on statistical feature vectors based on the K-means clustering algorithm and the hierarchical clustering algorithm to obtain a time series subset, including: constructing the following clustering hierarchy based on the K-means clustering algorithm and the hierarchical clustering algorithm: ; in, Y represents the target dataset for new energy power generation prediction, represents a union, C Indicates the maximum number of clustering levels, l Represents the hierarchical index of the cluster, from l =1 to l=C -1 represents different levels of clustering, Indicates the l In hierarchical clustering i A set of data points of a time series subset is obtained; cluster analysis is performed on the statistical feature vector based on the clustering hierarchy to obtain the time series subset.

[0013] The embodiment of the present invention provides a new energy power generation prediction method that combines global modeling and local optimization. The global model is locally optimized based on a time series subset to obtain a prediction model, including: initializing the local model based on the model parameters of the global model; training and optimizing the local model based on the time series subset to obtain a prediction model, wherein the prediction model is a local model that meets the preset prediction accuracy requirements.

[0014] An embodiment of the present invention provides an electronic device, comprising: a processor, and a memory for storing a program, wherein the program comprises instructions, and when the instructions are executed by the processor, the processor executes any of the above methods.

[0015] The present invention provides a non-transitory machine-readable medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute any of the above methods.

[0016] The present invention provides a new energy power generation prediction method that combines global modeling and local optimization. The current time series data of the target new energy station is input into the prediction model to obtain the corresponding prediction data, wherein the prediction model is obtained by locally optimizing the global model based on the time series subset, the time series subset is obtained by clustering the historical time series data of multiple new energy stations including the target new energy station, and the global model is trained based on the lag characteristics and category characteristics of the historical time series data of multiple new energy stations. Through the above-mentioned global modeling, time series clustering and local fine-tuning methods, it is possible to optimize the data patterns of different new energy stations while making full use of the global data features, improve the adaptive ability and stability of the prediction model in the face of different new energy stations and regional differences, and reduce the computational complexity, thereby achieving high-precision and high-efficiency calculation of new energy power generation prediction, especially in the case of data scarcity and strong volatility of new energy. It has obvious advantages. It solves the problem that the related technology cannot accurately predict the power generation time series of multiple new energy stations when performing new energy power generation prediction, and improves the computational efficiency while ensuring the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without inventive effort.

[0018] Figure 1 It is a flowchart of the steps of a new energy power generation prediction method combining global modeling and local optimization in an embodiment of the present invention.

[0019] Figure 2 It is a structural diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following describes embodiments of the present invention in more detail with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying 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 described 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.

[0021] Renewable energy power generation prediction methods based on pre-trained large models have good application prospects, but related methods usually require separate modeling for each renewable energy site, have high computational complexity, are difficult to expand, and lack the ability to adapt to the volatility and regional differences of renewable energy.

[0022] For example, deep learning-based renewable energy generation forecasting methods suffer from insufficient accuracy and generalization when simultaneously predicting the generation time series of multiple renewable energy sites. Transfer learning-based renewable energy generation forecasting methods also suffer from low model adaptability and computational efficiency when dealing with differences between renewable energy sites and regions.

[0023] To do this, please refer to Figure 1 As shown, the embodiment of the present invention provides a new energy power generation prediction method combining global modeling and local optimization, including steps S101 to S103.

[0024] Step S101: Acquire the current time series data of the target new energy station.

[0025] Step S102: Input the current time series data into the prediction model to obtain corresponding prediction data, wherein the prediction model is obtained by locally optimizing the global model based on the time series subset, the time series subset is obtained by clustering analysis of the historical time series data of multiple new energy stations, the global model is trained based on the lag characteristics and category characteristics of the historical time series data of multiple new energy stations, and the target new energy station is a new energy station among the multiple new energy stations.

[0026] Step S103 : determining a new energy power generation prediction result based on the prediction data corresponding to the current time series data of each new energy station.

[0027] Specifically, a new energy station refers to a location that uses new energy to generate electricity and centrally controls and manages the power generation process. New energy refers to renewable energy, including but not limited to wind, solar, and biomass energy.

[0028] Time series data refers to a series of observations recorded in chronological order, with each data point associated with a corresponding timestamp. Time series data is characterized by temporal dependence, whereby data from previous and subsequent times exhibit relationships or patterns.

[0029] Current time series data is not limited to the current power generation data and corresponding meteorological data of the target new energy station. Time series data for new energy stations is collected continuously at fixed intervals. Current time series data can refer to time series data collected within a period of time close to the current moment. Those skilled in the art can determine the specific length of this period based on prior values ​​and actual circumstances.

[0030] It is understandable that preprocessing the acquired current time series data helps improve prediction efficiency and accuracy. The method for preprocessing the current time series data can refer to the method used for preprocessing historical time series data.

[0031] It can be understood that hysteresis characteristics and categorical features are common features of historical time series data from various renewable energy stations. Hysteresis characteristics are used to capture temporal dependencies. Categorical features consist of temporal and meteorological features. Temporal features are used to capture seasonal and cyclical patterns in renewable energy generation. Meteorological features reflect the impact of external meteorological conditions on renewable energy generation. The preferred method for extracting hysteresis and categorical features from historical time series data from various renewable energy stations will be described in detail later.

[0032] Exemplarily, the new energy power generation prediction method based on pre-trained large models includes but is not limited to: new energy power generation prediction based on deep learning, new energy power generation prediction based on convolutional neural network (CNN), new energy power generation prediction based on long short-term memory network (LSTM), new energy power generation prediction based on generative adversarial network (GAN), and new energy power generation prediction based on transfer learning.

[0033] This embodiment is based on deep learning-based renewable energy generation forecasting and uses a deep learning large model to train a global model, which has better generalization capabilities. Those skilled in the art should be aware of the difference between deep learning models and deep learning large models, and this embodiment will not be repeated here.

[0034] The global model is trained using the common characteristics of historical time series data from various new energy stations. Compared with the local model optimized using a subset of the time series, the global model has more parameters and can achieve better performance over a wider range of data, and effectively prevent overfitting.

[0035] Based on the global model obtained by training based on common features, local optimization of the global model based on time series subsets can enhance the adaptability of the obtained prediction model, improve the prediction accuracy of the prediction model, reduce the computational cost of the prediction model, and facilitate the optimal scheduling of smart grids.

[0036] In other words, cluster analysis is performed on the historical time series data of multiple new energy stations to obtain time series subsets, and local optimization of the global model based on the time series subsets is performed to obtain a prediction model. The technical effects achieved include at least: analyzing the historical time series data by applying a clustering algorithm to discover time series subsets with similar characteristics, and then locally adjusting the global model according to the time series subsets, which can improve the prediction performance of the prediction model when processing time series with corresponding characteristics.

[0037] Based on the prediction data corresponding to the current time series data of each new energy site, the new energy power generation prediction result can be determined by referring to the existing technology, and this embodiment will not be repeated here.

[0038] The renewable energy power generation forecast results include, but are not limited to, the generation time series, predicted power generation, predicted power, and corresponding confidence intervals and uncertainty assessments for each renewable energy site in the future time period. The length of the future time period is determined by the model parameters of the forecast model.

[0039] In summary, the above-mentioned new energy power generation prediction method combining global modeling and local optimization provided in this embodiment inputs the current time series data of the target new energy station into the prediction model to obtain the corresponding prediction data, wherein the prediction model is obtained by local optimization of the global model based on the time series subset, the time series subset is obtained by clustering the historical time series data of multiple new energy stations including the target new energy station, and the global model is trained based on the lag characteristics and category characteristics of the historical time series data of multiple new energy stations. Through the above-mentioned method of global modeling, time series clustering and localized fine-tuning, it is possible to optimize the data patterns of different new energy stations while making full use of the global data features, improve the adaptive ability and stability of the prediction model in the face of different new energy stations and regional differences, and reduce the computational complexity, thereby achieving high-precision and high-efficiency calculation of new energy power generation prediction, which has obvious advantages especially in the case of data scarcity and strong volatility of new energy.

[0040] The above method provided in this embodiment can solve the problem that related technologies cannot accurately predict the power generation time series of multiple new energy sites when performing new energy power generation prediction, and improve computing efficiency while ensuring prediction accuracy.

[0041] Preferably, before step S101, obtaining the current time series data of the target new energy station, the above method further includes steps S001 to S004.

[0042] Step S001: Obtain historical time series data to be processed of multiple new energy stations.

[0043] Step S002: pre-process the historical time series data to be processed to obtain historical time series data.

[0044] Step S003: extract features from the historical time series data to obtain lag features and category features.

[0045] Step S004: train the deep learning big model based on the lag features and category features to obtain a global model, wherein the deep learning big model is a deep learning big model that adopts a neural-based expansion analysis architecture, and the global model is the above-mentioned deep learning big model that achieves the preset learning goals.

[0046] Methods for obtaining the historical time series data to be processed include but are not limited to sensor acquisition or monitoring system acquisition. Those skilled in the art may refer to the existing technology for implementation, and this embodiment will not be described in detail here.

[0047] The Neural-Based Unfolding Analysis Architecture, also known as the N-BEATS architecture, with its residual connection structure, enhances the expressive power of the global model and helps improve the accuracy of renewable energy generation forecasts through end-to-end learning of seasonal and trend information. Furthermore, the global model trained based on lagged and categorical features exhibits good generalization and stability.

[0048] Preferably, step S002, preprocessing the historical time series data to be processed to obtain the historical time series data, includes steps S0021 to S0023.

[0049] Step S0021: Perform unified formatting processing on the historical time series data to be processed to obtain formatted data.

[0050] Formatting includes but is not limited to: converting all numeric data to the same numeric type, such as floating point numbers or integers; and unifying time data from different sources into the format of year-month-day-hour:minute:second.

[0051] It is understandable that uniform formatting of historical time series data to be processed can improve the quality and usability of the data.

[0052] In step S0022, the formatted data is cleaned, denoised, detected for outliers, and filled with missing values ​​to obtain preliminary processed data.

[0053] Cleaning refers to removing duplicate and irregular data. For example, it removes duplicate records reported by sensors within the station and removes negative PV output values.

[0054] Denoising refers to eliminating data noise introduced by external interference or equipment errors through filtering or statistical methods. For example, it can smooth abnormally fluctuating voltage data.

[0055] Outlier detection uses rules or algorithms, such as box plot analysis, to identify data points that do not conform to physical or operational rules. For example, it can detect sudden surges in load values. The criteria for determining sudden surges can be determined by those skilled in the art based on prior knowledge and actual conditions.

[0056] Missing value filling refers to filling in missing data through interpolation, mean filling, or other methods. For example, when load data for a certain period is missing, linear interpolation can be performed based on the load data before and after this period.

[0057] It is understandable that the preliminary processed data obtained through the above-mentioned cleaning, denoising, outlier detection, and missing value filling processes are more accurate and complete than the formatted data, and can more accurately reflect the power generation status of the new energy station, avoiding the impact of low-quality data on subsequent modeling and learning.

[0058] Step S0023 , based on the standard score normalization algorithm or the minimum-maximum normalization algorithm, the preliminary processed data is normalized to obtain historical time series data.

[0059] The standard score normalization algorithm is also called the Z-score normalization algorithm.

[0060] It is understandable that standardizing the initially processed data ensures data consistency and usability. By converting different feature data to the same order of magnitude, we can prevent certain feature data from having a significant impact on global model training due to their large dimensions, thereby improving the convergence speed and stability of the learning algorithm.

[0061] Preferably, step S003, extracting features from the historical time series data to obtain lag features and categorical features, includes: extracting features from the historical load data in the historical time series data using a rolling window method to obtain lag features; performing one-hot encoding conversion on the time information in the historical time series data to obtain time features; and determining meteorological features based on the meteorological information in the historical time series data. The categorical features include time features and meteorological features.

[0062] It can be understood that the lag features extracted from historical time series data based on the rolling window method can capture time dependence.

[0063] Temporal information, including but not limited to the month, day of the week, and hour extracted from the timestamp, is converted using one-hot encoding to generate temporal features that can capture seasonal and cyclical patterns in renewable energy generation.

[0064] Meteorological information includes but is not limited to temperature, wind speed, and light irradiance. Determining meteorological characteristics based on this meteorological information can reflect the impact of external meteorological conditions on renewable energy power generation.

[0065] Preferably, step S004, training the deep learning large model based on the lag features and category features to obtain a global model, includes steps S0041 to S0043.

[0066] Step S0041, based on hysteresis characteristics and categorical features Construct the input matrix .

[0067] Step S0042: Create an input matrix The mapping relationship between the target matrix and the target matrix, where the target matrix is ​​used to represent the renewable energy power generation within the preset prediction range.

[0068] Input Matrix With the target matrix The mapping relationship between them can be expressed as: ; In addition, the preset prediction range can be adjusted by setting the model structure parameters of the deep learning model.

[0069] Step S0043: Taking the mean absolute error below the preset threshold as the preset learning target, combined with L 1. Regularization control: Based on the above mapping relationship, the model parameters of the deep learning large model are optimized and trained to obtain the global model.

[0070] The formula for optimizing training can be expressed as: ; ; in, Represents a deep learning model using a neural-based analysis architecture. represents the model parameters, L represents the loss function, represents the regularization factor, Represents the model prediction results.

[0071] The mean absolute error (MAE) measures the average absolute difference between the model's predicted values ​​and the true values, and is used to optimize the model's prediction accuracy. The above-mentioned preset threshold can be determined by those skilled in the art based on prior values ​​and actual conditions.

[0072] L 1Regularization control refers to the L 1 Regularization is used to control the complexity of the model and to optimize the sparsity of the model parameters. Regularization factor For control L 1 The strength of regularization can be set to 0.0001.

[0073] By minimizing the loss function L, which can simultaneously optimize the model's prediction accuracy and the sparsity of the model parameters, while avoiding overfitting while achieving the preset learning goal of keeping the mean absolute error below a preset threshold.

[0074] Preferably, before step S102, inputting the current time series data into the prediction model to obtain the corresponding prediction data, the above method further includes steps S10121 to S10123.

[0075] Step S10121: extract statistical features and perform standardization processing on the historical time series data of multiple new energy stations to obtain statistical feature vectors with the same weight.

[0076] Statistical characteristics include but are not limited to mean, variance, first-order autocorrelation, trend, and linearity, which can reflect the main change patterns of time series.

[0077] Standardization can ensure that different statistical features have the same weight in the subsequent clustering process, thereby avoiding excessive influence of certain statistical features on the clustering results.

[0078] Step S10122: Perform cluster analysis on the statistical feature vector based on the K-means clustering algorithm and the hierarchical clustering algorithm to obtain a time series subset.

[0079] It is understandable that by performing cluster analysis on statistical feature vectors, time series data with similar characteristics can be grouped into the same group. Among them, the hierarchical clustering algorithm can provide a stable and repeatable way to generate time series subsets, avoiding the uncertainty brought by the initialization process.

[0080] Step S10123: perform local optimization on the global model based on the time series subset to obtain a prediction model.

[0081] It can be understood that by executing step S10123, while retaining the global model knowledge, it is possible to optimize the data features of different scenarios, improve the prediction accuracy, and avoid the high computational cost brought by training from scratch.

[0082] Furthermore, step S10122 performs cluster analysis on the statistical feature vector based on the K-means clustering algorithm and the hierarchical clustering algorithm to obtain a time series subset, including: constructing the following clustering hierarchy based on the K-means clustering algorithm and the hierarchical clustering algorithm: ; in, Y The target dataset for new energy power generation forecasting includes the historical time series data of all new energy stations. represents a union, C Indicates the maximum number of clustering levels, lRepresents the hierarchical index of the cluster, from l =1 to l=C -1 represents different levels of clustering, Indicates the l In hierarchical clustering i A collection of data points from a subset of time series is used to define clustering structures at different levels.

[0083] Cluster analysis is performed on statistical feature vectors based on the clustering hierarchy to obtain time series subsets.

[0084] It is understandable that the above clustering hierarchy can help the model identify subsets of time series with potential similarities, which will be used as the basis for subsequent model optimization training.

[0085] In other words, through cluster analysis, the above method provided in this embodiment can enable the prediction model obtained by subsequent optimization to better adapt to different types of time series, thereby improving the generalization ability of the prediction model when processing diversified new energy power generation data.

[0086] Furthermore, step S10123 locally optimizes the global model based on the time series subset to obtain a prediction model, including initializing the local model based on the model parameters of the global model. Training and optimizing the local model based on the time series subset to obtain the prediction model, wherein the prediction model is the local model that meets the preset prediction accuracy requirements.

[0087] The mapping relationship for training and optimizing the local model based on a subset of time series can be expressed as: ; in, Represents the prediction model, which can be understood as a localized version of the global model. can provide better prediction performance. express The corresponding feature matrix. Represents a prediction model The model parameters. express The corresponding target matrix.

[0088] The indicators used to evaluate prediction accuracy include, but are not limited to, mean absolute error, root mean square error, and mean absolute percentage error. Setting the preset prediction accuracy requirement and determining whether the prediction model meets the preset prediction accuracy requirement can be implemented by referring to existing model evaluation methods and will not be further described in this embodiment.

[0089] It is understood that step S10123 can be understood as a model adaptation process, providing more appropriate new model parameters for the time series subset based on the model parameters of the global model. The prediction model constructed based on these new model parameters can optimize the data characteristics of different scenarios while retaining the knowledge of the global model, thereby improving prediction accuracy and avoiding the high computational cost of training from scratch, achieving high-precision and high-efficiency calculations for renewable energy generation forecasts.

[0090] The present invention also 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, wherein the computer program, when executed by the at least one processor, causes the electronic device to perform the method of the present invention.

[0091] refer to Figure 2 , a structural block diagram of an electronic device that can be used as a server or client of an embodiment of the present invention will now be described, which is an example of a hardware device that can be applied 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 processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0092] like Figure 2 As shown, the electronic device includes a computing unit 201, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 202 or a computer program loaded from a storage unit 208 into a random access memory (RAM) 203. RAM 203 may also store various programs and data required for the operation of the electronic device. The computing unit 201, ROM 202, and RAM 203 are interconnected via a bus 204. An input / output (I / O) interface 205 is also connected to the bus 204.

[0093] Multiple components within the electronic device are connected to the I / O interface 205, including an input unit 206, an output unit 207, a storage unit 208, and a communication unit 209. The input unit 206 can be any type of device capable of inputting information into the electronic device. The input unit 206 can receive input numeric or character information and generate key signal inputs related to user settings and / or function control of the electronic device. The output unit 207 can be any type of device capable of presenting information and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 208 can include, but is not limited to, a magnetic disk or an optical disk. The communication unit 209 allows the electronic device to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks, and can include, but is not limited to, a modem, a network card, an infrared communication device, and / or a wireless communication transceiver, such as a Bluetooth device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0094] The computing unit 201 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 201 include, but are not limited to, a CPU, a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing units, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 201 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention may be implemented as a computer program tangibly embodied in a machine-readable medium, such as the storage unit 208. In some embodiments, part or all of the computer program may be loaded and / or installed onto the electronic device via the ROM 202 and / or the communication unit 209. In some embodiments, the computing unit 201 may be configured to perform the above-described methods by any other suitable means (e.g., via firmware).

[0095] The present invention also provides a non-transitory machine-readable medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to perform the method of the present invention.

[0096] The present invention also provides a computer program product including a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to execute the method of the present invention.

[0097] The computer programs for implementing the methods of the 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, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0098] In the context of the present invention, a machine-readable medium may 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. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable signal medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, or infrared system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0099] It should be noted that the term "including" and its variations used in the embodiments of the present invention are open inclusions, that is, "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 other embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "multiple" mentioned in the embodiments of the present invention are illustrative and not restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more". The descriptions of the terms "first", "second", etc. are for descriptive purposes only and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features.

[0100] 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 the present invention are all information and data authorized by the user or fully authorized by all parties.

[0101] The various steps described in the method implementation methods provided by the embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method implementation methods may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0102] The term "embodiment" in this specification refers to specific features, structures or characteristics described in conjunction with the embodiment that can be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. The various embodiments in this specification are described in a related manner, and the same or similar parts between the various embodiments are referenced to each other. In particular, for the device, equipment, and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts refer to the partial description of the method embodiment.

[0103] The above-described embodiments merely represent several implementation methods of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of protection. It should be noted that a person of ordinary skill in the art would be able to make various modifications and improvements without departing from the scope of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A new energy power generation prediction method combining global modeling and local optimization, characterized in that: include: Obtain the current time series data of the target new energy station; Inputting the current time series data into a prediction model to obtain corresponding prediction data, wherein the prediction model is obtained by locally optimizing a global model based on a time series subset, the time series subset is obtained by clustering analysis of historical time series data of multiple new energy stations, the global model is trained based on the lag characteristics and category characteristics of the historical time series data of the multiple new energy stations, and the target new energy station is a new energy station among the multiple new energy stations; The new energy power generation prediction results are determined based on the prediction data corresponding to the current time series data of each new energy station.

2. The method according to claim 1, characterized in that Before obtaining the current time series data of the target new energy station, the method further includes: Obtaining the historical time series data to be processed of the plurality of new energy stations; Preprocessing the to-be-processed historical time series data to obtain the historical time series data; Performing feature extraction on the historical time series data to obtain the lag feature and the category feature; The deep learning big model is trained based on the lag feature and the category feature to obtain the global model, wherein the deep learning big model is a deep learning big model that adopts a neural-based expansion analysis architecture, and the global model is the deep learning big model that achieves the preset learning goal.

3. The method according to claim 2, characterized in that Preprocessing the historical time series data to be processed to obtain the historical time series data includes: Performing unified formatting processing on the historical time series data to be processed to obtain formatted data; The formatted data is sequentially cleaned, denoised, detected for outliers, and filled with missing values ​​to obtain preliminary processed data; The preliminary processed data is standardized based on a standard score standardization algorithm or a minimum-maximum normalization algorithm to obtain the historical time series data.

4. The method according to claim 2, characterized in that Performing feature extraction on the historical time series data to obtain the lag feature and the category feature includes: Extracting features from the historical load data in the historical time series data based on a rolling window method to obtain the hysteresis features; Performing one-hot encoding conversion on the time information in the historical time series data to obtain time features; determining meteorological characteristics based on meteorological information in the historical time series data; The category features include the time features and the meteorological features.

5. The method according to claim 2, characterized in that The deep learning large model is trained based on the hysteresis feature and the category feature to obtain the global model, including: Based on hysteresis characteristics and categorical features Construct the input matrix ; Create the input matrix A mapping relationship between the target matrix and the target matrix, wherein the target matrix is ​​used to represent the renewable energy power generation within a preset prediction range; The mean absolute error is lower than the preset threshold as the preset learning goal, combined with L 1. Regularization control, optimizing and training the model parameters of the deep learning large model based on the mapping relationship to obtain the global model.

6. The method according to claim 1, characterized in that Before inputting the current time series data into the prediction model to obtain corresponding prediction data, the method further includes: Performing statistical feature extraction and standardization on the historical time series data of the plurality of new energy stations to obtain statistical feature vectors with the same weight; Performing cluster analysis on the statistical feature vector based on a K-means clustering algorithm and a hierarchical clustering algorithm to obtain the time series subset; The global model is locally optimized based on the time series subset to obtain the prediction model.

7. The method according to claim 6, characterized in that Cluster analysis is performed on the statistical feature vector based on the K-means clustering algorithm and the hierarchical clustering algorithm to obtain the time series subset, including: The following clustering hierarchy is constructed based on the K-means clustering algorithm and the hierarchical clustering algorithm: ； in, Y represents the target dataset for new energy power generation prediction, represents a union, C Indicates the maximum number of clustering levels, l Represents the hierarchical index of the cluster, from l =1 to l=C -1 represents different levels of clustering, Indicates the l In hierarchical clustering i A set of data points for a subset of the time series; Cluster analysis is performed on the statistical feature vector based on the clustering hierarchy to obtain the time series subset.

8. The method according to claim 6, characterized in that Locally optimizing the global model based on the time series subset to obtain the prediction model includes: Initializing a local model based on model parameters of the global model; The local model is trained and optimized based on the time series subset to obtain the prediction model, wherein the prediction model is the local model that meets the preset prediction accuracy requirement.

9. An electronic device comprising: A processor and a memory storing a program, wherein the program comprises instructions which, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 8.

10. A non-transitory machine-readable medium storing computer instructions, characterized in that The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 8.

Images