Wind power prediction method and device, equipment and storage medium

Abstract

The invention relates to a wind power prediction method and device based on a time sequence large model, equipment and a storage medium. The wind power prediction method comprises the following steps: acquiring a time sequence large model and a first training data set of a plurality of source wind power plants; training the time sequence large model by using the first training data set to obtain a basic large model for wind power prediction; acquiring a historical wind power array and a future weather forecast array of the target wind power plant, wherein the plurality of source wind power plants do not include the target wind power plant; and taking the historical wind power array and the future weather forecast array as input of the target large model, and predicting a future wind power array of the target wind power plant. According to the method, the dependency of model training on the training data volume is reduced to a certain extent, meanwhile, the problem that the number of model parameters is different is effectively solved, and the application flexibility and effectiveness of the model in variable scenes are further improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a wind power prediction method, device, equipment and storage medium based on a time series large model. Background Art

[0002] Current wind power prediction methods primarily rely on data-driven approaches. High-precision wind power prediction models rely on large amounts of training data. However, newly built, expanded, or renovated wind farms often have short operating times and may not provide sufficient historical data, or even no historical data at all, for wind power prediction model training.

[0003] Currently, the most commonly used method for small-sample wind power forecasting is the "pre-training-fine-tuning" approach of transfer learning. This involves first training a wind power forecasting model with sufficient training samples from the source wind farm, and then transferring the power forecasting knowledge learned by the model to the target wind farm. However, using transfer learning alone to solve this problem has certain limitations: it requires strong "consistency" between wind farms and requires "one target farm, one pre-training" approach. On the one hand, transfer learning relies heavily on the sample quality and relevance between the source and target wind farms, and due to issues such as data privacy, it is sometimes difficult to obtain source wind farm data with high similarity. On the other hand, ordinary transfer learning alone cannot solve the inconsistencies in model input length, prediction scale, temporal resolution, and number of prediction variables between wind farms, limiting the flexibility and effectiveness of applications in changing scenarios. Summary of the Invention

[0004] In order to solve the above technical problems, the embodiments of the present disclosure provide a wind power prediction method, apparatus, device and storage medium based on a time series large model.

[0005] In a first aspect, an embodiment of the present disclosure provides a wind power prediction method based on a time series large model, comprising: Obtaining a first training data set of a time series large model and multiple source wind farms; Pre-training the time series large model using the first training data set to obtain a basic large model for wind power prediction; Obtaining a historical wind power array and a future weather forecast array of a target wind farm, wherein the multiple source wind farms do not include the target wind farm; The historical wind power array and the future weather forecast array are used as inputs of a target large model to predict the future wind power array of the target wind farm, wherein the target large model is a basic large model or is obtained by fine-tuning the basic large model using a second training data set of the target wind farm.

[0006] In a second aspect, an embodiment of the present disclosure provides a wind power prediction device based on a time series large model, comprising: A first acquisition unit is used to acquire a first training data set of a large time series model and multiple source wind farms; A training unit, configured to pre-train the time series large model using the first training data set to obtain a basic large model for wind power prediction; a second acquisition unit, configured to acquire a historical wind power array and a future weather forecast array of a target wind farm, wherein the plurality of source wind farms does not include the target wind farm; The prediction unit is used to use the historical wind power array and the future weather forecast array as inputs of a target large model to predict the future wind power array of the target wind farm, wherein the target large model is a basic large model or is obtained by fine-tuning the basic large model using a second training data set of the target wind farm.

[0007] In a third aspect, an embodiment of the present disclosure provides an electronic device, including: Memory; processor; and computer programs; The computer program is stored in the memory and is configured to be executed by the processor to implement the method of the first aspect as described above.

[0008] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method of the first aspect described above when the computer program is executed by a processor.

[0009] The wind power prediction method provided by the present disclosure includes: obtaining a time series large model and a first training data set of multiple source wind farms; using the first training data set to train the time series large model to obtain a basic large model for wind power prediction; obtaining a historical wind power array and a future weather forecast array of a target wind farm, wherein the multiple source wind farms do not include the target wind farm; using the historical wind power array and the future weather forecast array as inputs to the target large model to predict the future wind power array of the target wind farm, wherein the target large model is a basic large model or is obtained by fine-tuning the basic large model using a second training data set of the target wind farm. The present application reduces the dependence of model training on the amount of training data to a certain extent, while effectively solving the problem of quantitative differences in model parameters, and further improves the application flexibility and effectiveness of the model in variable scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0011] In order to more clearly illustrate the embodiments of the present disclosure 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, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0012] Figure 1 A schematic flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure; Figure 2 A schematic diagram of the structure of a time series large model provided by an embodiment of the present disclosure; Figure 3 A schematic diagram of the structure of a basic large model provided in an embodiment of the present disclosure; Figure 4 A schematic flow chart of a wind power prediction method based on a target large model provided in an embodiment of the present disclosure; Figure 5 A schematic flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure; Figure 6 A two-dimensional structural diagram of a historical wind power training sample provided in an embodiment of the present disclosure; Figure 7 A two-dimensional structural diagram of a weather forecast training sample provided by an embodiment of the present disclosure; Figure 8 A schematic flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure; Figure 9 A schematic flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure; Figure 10 A schematic flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure; Figure 11 A schematic structural diagram of a wind power prediction device based on a time series large model provided by an embodiment of the present disclosure; Figure 12 A schematic structural diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0013] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features therein can be combined with each other in the absence of conflict.

[0014] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0015] In response to the above technical problems, the embodiments of the present disclosure provide a wind power prediction method based on a time series large model, construct a relatively general wind power prediction model (the above-mentioned basic large model), and realize zero-sample / small-sample wind power prediction. When there is a forecast demand for the construction, expansion, or reconstruction of a target wind farm, the wind power prediction model no longer relies on sufficient historical data and a large amount of training time for the target wind farm. By relying on the experience of the wind power prediction model's previous training data, it can directly or after fine-tuning achieve low-time, low-cost, highly adaptable, and high-precision wind power prediction, while effectively solving the problem of quantitative differences in model input length, prediction scale, time resolution, and variables between wind farms. This will be explained in detail through one or more of the following embodiments.

[0016] Before describing the methods involved in this application in detail, the following terms are explained: A wind farm (wind power plant) is a power plant that converts wind energy into mechanical energy, which is then converted back into electrical energy. The wind farm provided by the present invention comprises, but is not limited to, a wind tower, wind turbines, collection lines, and a booster station. The wind turbines receive wind energy and convert it into electrical energy, which is then output through the output of the booster station. The wind power generated by the wind farm is the total power output by the booster station.

[0017] Wind power forecasting refers to the technology of predicting the power generation of a wind farm in the future, providing a reference for the operation and control of the wind farm, the dispatching and operation of the power system, and the power market transactions.

[0018] Moirai refers to a large deep learning model for time series forecasting. It is trained with the LOSTA dataset (more than 27.6 billion data points covering transportation, energy, electricity, finance, healthcare and other fields). It has 14 million parameters and powerful modeling capabilities, capable of processing data of different resolutions, different numbers of variables and different distribution characteristics.

[0019] Patch refers to a technique that divides raw data into multiple smaller fixed-size fragments when processing data (such as images, time series, or text).

[0020] Transfer learning refers to a machine learning method that aims to use model knowledge trained in one field and apply it to related fields to improve training efficiency and prediction performance.

[0021] Source wind farms refer to wind farms that are distributed across the country and have been operating for a long time and have sufficient training data.

[0022] The target wind farm refers to a wind farm that has been newly built, expanded, or rebuilt and has no or only a small amount of training data.

[0023] Pre-training refers to the process of training a wind power prediction model using training data from the source wind farm.

[0024] The wind power prediction large model base refers to a large model suitable for wind power prediction obtained by pre-training the time series large model Moirai.

[0025] Fine-tuning refers to the process of updating the model parameters in the wind power prediction large model base using the training data of the target wind farm.

[0026] The exclusive wind power prediction model refers to a high-precision wind power prediction model applied to the target wind farm.

[0027] Numerical weather forecast data refers to meteorological forecast data that reflects the state of atmospheric movement in a certain period of time in the future, obtained by solving the set of fluid mechanics and thermodynamics equations that describe the weather evolution process based on the actual atmospheric conditions under certain initial and boundary conditions through numerical calculations.

[0028] The wind power prediction method based on a large time series model provided in the embodiments of the present disclosure is applicable to wind power prediction scenarios based on large time series models. The method can be performed by a wind power prediction device based on a large time series model. The device can be implemented in software and / or hardware, and the device can be integrated into an electronic device. The electronic device can include, but is not limited to, mobile terminals such as smartphones, laptops, digital broadcast receivers, personal digital assistants (PDAs), tablet personal computers (Tablet PCs), PMPs (portable multimedia players), in-vehicle terminals (such as in-vehicle navigation terminals), wearable devices, and fixed terminals such as digital televisions, desktop computers, smart home devices, and the like.

[0029] Figure 1 A flow chart of a wind power prediction method based on a time series large model provided by an embodiment of the present disclosure, specifically including the following steps: Figure 1 The following steps are shown: S110 : Acquire a large time series model and a first training data set of multiple source wind farms.

[0030] It is understandable that the time series pre-training large model is a large-scale neural network model that has been pre-trained with massive historical data. The model can automatically learn complex time series patterns, and can then quickly adapt to the sample characteristics of different fields through simple adjustments. Due to the structures such as the projection layer with multiple patch sizes and the arbitrary variable attention mechanism within the model, as well as the random sampling strategy of the model input length and prediction scale set during pre-training, the time series pre-training large model can adapt to different prediction spans, model input and output structures and characteristics. The multiple source wind farms can specifically be multiple source wind farms across the country. The first training data set includes sufficient and massive training data from multiple source wind farms across the country under all weather conditions, so that the trained basic large model can be applicable to wind power prediction under all weather conditions and has universal applicability.

[0031] Among them, the time series large model includes the input feature extraction layer, the time series modeling layer and the prediction output mapping layer.

[0032] For example, see Figure 2 , Figure 2 A schematic diagram of the structure of a large time series model provided in an embodiment of the present disclosure is provided. The large time series model includes an input feature extraction layer, a time series modeling layer, and a prediction output mapping layer. The large time series model first preprocesses historical data through the input feature extraction layer, and uses a variety of patch sizes and variable encoding techniques to extract key features across frequencies and variables. Subsequently, the time series modeling layer uses an improved Transformer (attention mechanism model) encoder and an arbitrary variable attention mechanism to perform deep temporal dependency modeling on the extracted key features. Finally, the internal representation is decoded into mixed distribution parameters through the prediction output mapping layer to generate prediction results for future time steps. Due to the general temporal evolution laws learned during the large-scale time series data training process, it can perform targeted adjustments to the feature laws for time series prediction tasks in specific fields, such as wind power prediction tasks, to achieve domain-adaptive prediction accuracy and robustness.

[0033] S120: Pre-train the time series large model using the first training data set to obtain a basic large model for wind power prediction.

[0034] It is understandable that, based on the above-mentioned S110, the time series large model is pre-trained sequentially using sufficient training samples from multiple source wind farms across the country to obtain a basic large model. That is, the time series large model is pre-trained into a wind power prediction model suitable for wind power prediction, wherein the basic large model can be understood as the base of the above-mentioned wind power prediction large model. Sequentially training the time series large model means that the current source wind farm is trained based on the model parameters obtained by the previous source wind farm training. Specifically, one scenario is to train with the training data of one source wind farm at a time. That is, after pre-training the time series large model with the training data of source wind farm a, pre-training is continued with the training data of source wind farm b based on this pre-training (including model parameters and network structure, etc.). It is also possible to select a certain number of source wind farms from multiple source wind farms as the current source wind farms for batch training. Other possible training processes are not limited. This basic large model can adapt to different prediction spans, model input and output structures and characteristics, etc., which improves the application flexibility and effectiveness in variable scenarios, and can thus achieve zero-sample or small-sample wind power prediction for the target wind farm.

[0035] For example, see Figure 3 , Figure 3 A structural schematic diagram of a basic large model provided in an embodiment of the present disclosure is provided. The basic large model is obtained based on pre-training of a time series large model. Accordingly, the basic large model includes a feature extraction layer, a wind power time series modeling layer, and a wind power mapping layer. The feature extraction layer is used to extract features of historical wind power and numerical weather forecast data, the wind power time series modeling layer is used to deeply fuse the extracted features, and the wind power mapping layer is used to map the fused features to the prediction distribution parameter space to generate the final wind power prediction sequence.

[0036] S130: Obtain a historical wind power array and a future weather forecast array of the target wind farm.

[0037] The multiple source wind farms do not include the target wind farm.

[0038] It is understood that, based on the above S120, the target wind farm refers to a newly built, expanded, or renovated wind farm for which there is no training data or only a small amount of training data. The source wind farm refers to a long-established wind farm for which there is a large amount of training data under all weather conditions. The multiple source wind farms mentioned above do not include the target wind farm. A two-dimensional array of historical wind power and a two-dimensional array of future weather forecasts are obtained for the target wind farm.

[0039] Optionally, before obtaining the historical wind power array and future weather forecast array of the target wind farm, the method further includes: A second training data set for the target wind farm is obtained, wherein the sample data volume and sampling period of the second training data set are both smaller than those of the first training data set; when the second training data set includes at least one historical wind power training array, the parameters of the basic large model are fine-tuned using the second training data set to obtain a target large model, wherein the target large model is specifically used for wind power prediction of the target wind farm; or, when the second training data set does not include the historical wind power training array, the basic large model is directly used as the target large model.

[0040] It is understandable that a second training data set is obtained for the target wind farm. The sample size and sampling period of the second training data set are much smaller than the sample size and sampling period of the first training data set. The second training data set includes limited training data, which can be understood as a zero-sample set or a small sample set, while the first training sample set includes training data with sufficient weather conditions, which can be understood as a large sample set or a massive sample set. If the second training sample set does not include the historical wind power training array, the general time series features of the massive wind power data learned by the basic large model in the pre-training of training data of multiple source wind farms are directly used to perform zero-sample prediction of the future wind power of the target wind farm. For scenarios where the second training sample set does not include the historical wind power training array but includes the weather forecast training array, the basic large model is directly used for zero-sample prediction.

[0041] It is understandable that if the second training data set includes a historical wind power training array, that is, the target wind farm has limited historical power data, then the small sample data of the target wind farm is used to fine-tune the basic large model. In this way, it can not only reflect the general time series characteristics of the massive time series data in the pre-training, but also take into account the differentiated characteristics of the specific geographical environment, meteorological conditions, units, etc., that is, the basic large model after fine-tuning to a certain extent has both general time series characteristics and meets the differentiated characteristics of the target wind farm. The second training data set includes input data and corresponding output data, wherein the input data is the historical wind power training array and the weather forecast training array, and the output data is the actual wind power training array. In one possible application scenario, the second training data set includes both the historical wind power training array and the weather forecast training array. In another possible application scenario, the second training data set only includes the historical wind power training array. In both cases, the basic large model can be fine-tuned according to the second training data set.

[0042] Optionally, obtain the historical wind power array and future weather forecast array of the target wind farm. This can be achieved through the following steps: Obtain the historical wind power array of the target wind farm within a historical set period and the future weather forecast array within a future set period.

[0043] It is understandable that the historical set period refers to the collection period of the historical wind power array. For example, the current moment is moment B, and the historical set period refers to the T period before moment B, which can be specifically T hours. The future set period refers to the forecast period of wind power. For example, the future set period refers to the N period after moment B, which can be specifically N hours. The above-mentioned first training data set also includes a historical wind power training array of T hours and a weather forecast training array of N hours. Among them, when pre-training the time series large model using sufficient training samples from multiple source wind farms (i.e., the first training data set), the above-mentioned T and N do not need to be set manually, and the lengths of T and N can be randomly sampled within the large model. In the process of fine-tuning the basic large model based on the small sample training samples of the target wind farm (i.e., the second training data set), the lengths of T and N can be set according to actual needs.

[0044] S140: Use the historical wind power array and the future weather forecast array as inputs of the target large model to predict the future wind power array of the target wind farm.

[0045] The target large model is a basic large model or is obtained by fine-tuning the basic large model using the second training data set of the target wind farm.

[0046] Optionally, the historical wind power array and the future weather forecast array are used as inputs of the target large model to predict the future wind power array of the target wind farm. This can be achieved through the following steps: The historical wind power array and the future weather forecast array are used as inputs of the target large model to predict the future wind power array of the target wind farm within a set period of time in the future.

[0047] It is understood that, based on the above S130, the historical wind power array includes the wind power values recorded for the target wind farm over a period of time (e.g., the past three hours, eight hours, or a week), i.e., the historical power for the past short period of time. The future weather forecast array includes meteorological forecast data for a specific time period in the near future, such as wind speed, wind direction, temperature, humidity, and other key factors affecting wind power output, i.e., the weather conditions for the near future. The historical wind power array and the future weather forecast array are fed as input into a pre-trained target large model. The target large model then infers the future wind power array based on the input data, thereby predicting the wind power output of the target wind farm for the specific time period in the future, i.e., the wind power for the near future. The forecast period for the weather forecast and the wind power prediction period are the same. For example, the historical power for the past eight hours and the weather forecast for the next six hours are collected, and the wind power for the next six hours is predicted using the target large model. This approach effectively combines historical wind power data with future weather forecast data, improving the accuracy and robustness of future wind power predictions. It is particularly suitable for small-sample wind farms with data scarcity problems. By introducing weather forecast data as an external variable, it not only broadens the information source of the large model, but also improves its generalization ability.

[0048] For example, see Figure 4 , Figure 4 A flow chart of a wind power prediction method based on a target large model provided in an embodiment of the present disclosure is provided, in which historical wind power data for the past T hours is used as the main input, and numerical weather forecast data for the next N hours is used as the dynamic covariate input, which serves as a secondary input and plays a secondary auxiliary role. The wind power prediction data for the next N hours is calculated through the target large model (the wind power prediction model in the figure).

[0049] The embodiment of the present disclosure provides a wind power prediction method based on a time series large model. Based on the open source time series large model, it fully utilizes its general time series pattern representation ability learned in large-scale time series data pre-training; at the same time, it introduces a "pre-training-fine-tuning" strategy of transfer learning, and uses the massive and diverse historical wind power data collected from multiple source wind farms across the country to pre-train the time series large model, so that it can fully learn and refine the universal characteristics and differentiated laws presented by different geographical environments, meteorological conditions, and wind turbine configurations, significantly improving the generalization ability and robustness of the model; then, for zero-sample wind farms, zero-sample prediction can be directly performed. For small-sample wind farms, only a small amount of training data from the target wind farm is required to enable the pre-trained basic large model to find the parameter update direction that best reflects the characteristics of the target wind farm, thereby achieving low-time, low-cost, high-precision, and highly adaptable small-sample wind power prediction. Therefore, while fully utilizing the existing massive training data from multiple source wind farms, the present disclosure significantly reduces the dependence on long-term, large-scale training data from the target wind farm, providing a more flexible and feasible solution for actual deployment and application.

[0050] Based on the above embodiments, Figure 5 A flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure is provided. Optionally, the time series large model is pre-trained using a first training data set to obtain a basic large model for wind power prediction, specifically including the following steps: Figure 5 The following steps are shown: S501: Inputting the training data sets corresponding to the multiple wind farms in the first training data set into the time series large model in sequence, targeting the current wind farm among the multiple wind farms.

[0051] Among them, there are historical wind power training samples, weather forecast training samples and actual wind power training samples; the current wind farm contains sufficient training samples under all weather conditions.

[0052] It can be understood that the first training data set includes a large number of historical wind power training samples, weather forecast training samples and corresponding actual wind power training samples for each source wind farm. The historical wind power training samples can be understood as a two-dimensional array of historical power, and the weather forecast training samples can be understood as a two-dimensional array of numerical weather forecasts. Among them, the historical wind power training samples and weather forecast training samples are input data, and the actual wind power training samples are output data.

[0053] Among them, the historical wind power training samples and weather forecast training samples refer to two-dimensional arrays, wherein the historical wind power training samples include a first time dimension and a power category dimension, the first time dimension includes multiple first moments for power sampling, and the multiple first moments differ by a first set period of time, the power category dimension includes a first time feature code, historical power data and a wind farm number, and the first time feature code corresponds to the first time dimension.

[0054] It is understandable that the historical wind power training sample can be understood as a historical power two-dimensional array, which includes a first time dimension and a power category dimension, such as Figure 6 As shown, Figure 6 A two-dimensional structural diagram of a historical wind power training sample provided in an embodiment of the present disclosure. Figure 6 The horizontal dimension represents the first time dimension, where 1, 2, 3, ..., T (the past T hours) represent the first, second, third, ..., and Mth moments, respectively. The difference between two adjacent moments is 15 minutes, and 15 minutes refers to the first set period mentioned above. The vertical dimension represents the power category, including the first time feature code, historical power data, and wind farm number. Historical power data can be specific power values. To distinguish between multiple source wind farms across the country in the source domain, different numbers can be assigned to each source wind farm. The specific setting method is not limited. For the time feature code, the time feature code of the first sampling time point among the M time points can be defined as 1, the time code of the second sampling time point as 2, and so on. This is not detailed here.

[0055] Optionally, the weather forecast training sample includes a second time dimension and a meteorological category dimension, wherein the second time dimension includes multiple second moments for weather sampling, and the multiple second moments are separated by a second set period; the meteorological category dimension includes a second time feature code and multiple meteorological forecast data types, the multiple meteorological forecast data types include forecast wind condition data and other meteorological forecast data, the forecast wind condition data includes forecast wind speed data and forecast wind direction data, and the other meteorological forecast data includes at least one of forecast air pressure data, forecast temperature data and forecast humidity data.

[0056] It is understandable that the weather forecast training sample can be understood as a numerical weather forecast two-dimensional array (weather feature two-dimensional array), and the weather forecast two-dimensional array includes a second time dimension and a meteorological category dimension, such as Figure 7 As shown, Figure 7 A two-dimensional structural diagram of a weather forecast training sample provided in an embodiment of the present disclosure. Figure 7The horizontal axis in the middle represents the second time dimension, where 1, 2, 3, ..., N represent the first moment, the second moment, the third moment, ..., the Nth moment, respectively. The difference between two adjacent moments is 15 minutes, that is, the second set period can also be set to 15 minutes. The first set period and the second set period can be the same or different. The meteorological category dimension includes a second time feature code and multiple meteorological forecast data types. The multiple categories of meteorological forecast data that affect wind power include, but are not limited to: forecast wind condition data and other meteorological forecast data. The forecast wind condition data includes forecast wind speed data and forecast wind direction data. Other meteorological forecast data includes, but is not limited to, one or more of the following: forecast air pressure data, forecast temperature data, and forecast humidity data. Figure 7 Each line in the is a type of weather forecast data. For each weather type, the forecast data of N time points are arranged in chronological order to obtain a one-dimensional array, which can also be called a sequence. The sequence of forecast data of each weather type is arranged according to the rule that the same sampling time point corresponds to the same time dimension to obtain a two-dimensional array of numerical weather forecasts.

[0057] It can be understood that the training process of the basic large model is as follows: copy the parameters of the time series large model, use transfer learning for pre-training, and use the training data of multiple source wind farms to update the parameters of all layers of the time series large model E times at a time. Here, 1≤E≤10000 is set, and the specific value of E is determined by the "early stopping method". The "early stopping method" is an effective strategy used to prevent overfitting during the training of machine learning models. Specifically, the current training data can be divided into a training set and a validation set. The model parameters of the time series large model are updated using the training set, and the accuracy is verified using the validation set. When the accuracy of the validation set decreases p times continuously, the iteration process is terminated and the time series large model is no longer updated. Specifically, multiple source wind farms are selected to perform transfer learning pre-training on the time series prediction large model in sequence until all source wind farms are trained. The training process is as follows: First, the current wind farm is determined from multiple source wind farms. The training data of each wind farm in the source domain is sufficient to be data-driven by the time series large model. The current wind farm refers to the wind farm whose corresponding training data is to be used to pre-train the time series model in the current round. All training data for the current wind farm is recorded as the current training data. Subsequently, the optimal parameters of the time series model for the current wind farm are calculated, specifically including steps S502 to S505.

[0058] S502: Select a group of training samples from sufficient training samples.

[0059] A set of training samples includes historical wind power training samples, weather forecast training samples and actual wind power training samples.

[0060] As can be understood, based on the above S501, a group of training samples is randomly selected from the sufficient training samples under all weather conditions of the current wind farm, denoted as group A training samples. Each group of training samples includes a historical wind power training sample and a weather forecast training sample. Where A is an integer greater than or equal to 1 and can be set as needed. In addition, the set A should not be greater than the total amount of training data contained in the current wind farm.

[0061] S503: Input the characteristic parameters of the historical wind power training samples and the weather forecast training samples into the time series large model to obtain a set of corresponding wind power prediction training samples.

[0062] The time series large model has model parameters trained based on a training sample set corresponding to a previous wind farm.

[0063] It is understandable that, based on the above S502, the characteristic parameters of A historical wind power training samples and A weather forecast training samples in group A of training samples are input into the time series large model to obtain corresponding A wind power prediction training samples.

[0064] S504: Calculate a first average prediction error between wind power prediction training samples and wind power actual training samples.

[0065] As can be understood, based on the above S503, the wind power prediction training samples refer to the predicted values output by the large model, and the wind power actual training samples refer to the actual values. The average prediction error between the predicted values and the actual values of Group A is calculated and recorded as the first average prediction error. The error calculation method can be mean square error, mean absolute error, etc., and is not limited here.

[0066] S505. Update all layer parameters of the time series large model according to the first average prediction error to obtain the optimal parameters under all weather conditions of the current wind farm, until the training of the time series large model with the training data set corresponding to the last wind farm is completed to obtain the basic large model for wind power prediction.

[0067] It can be understood that, based on the above S504, the parameters of all layers of the large model are updated according to the first average prediction error to obtain the optimal parameters under all weather conditions of the current wind farm. Among them, the parameter update algorithm can adopt the parameter update method in the traditional neural network training process, such as Adam. Other possible parameter update methods are not described in detail.

[0068] It is understandable that after the optimal parameter calculation is completed, the optimal parameters are used as the final parameters of the time series large model. After traversing the training data of multiple source wind farms in the source domain in turn, the optimal parameters at this time are used as the final parameters of the time series large model to obtain the wind power prediction large model base (basic large model). At this time, the parameters of the basic large model are the parameters after the last pre-training update of the time series prediction large model.

[0069] The embodiments of the present disclosure provide a wind power prediction method based on a time series large model. The time series large model is pre-trained using massive training data from multiple source wind farms, making it suitable for wind power prediction. A basic large model that conforms to different geographical environments, meteorological conditions, and wind turbine configurations is obtained, significantly improving the generalization and robustness of the model.

[0070] Based on the above embodiments, Figure 8 A flow chart of a wind power prediction method based on a time series large model provided in an embodiment of the present disclosure is provided. Optionally, characteristic parameters of historical wind power training samples and weather forecast training samples are input into the time series large model to obtain a set of corresponding wind power prediction training samples, specifically including the following: Figure 8 The following steps are shown: S801: Use the input feature extraction layer to extract features from historical wind power training samples and weather forecast training samples to obtain power feature training samples and weather feature training samples with time series and variable distinguishability.

[0071] As you can understand, both the power feature training samples (power feature 2D array) and the weather feature training samples (weather feature 2D array) include a time dimension and a feature dimension. The time dimension corresponds to the time dimension of the historical power feature 2D array and the weather feature 2D array. Each dimension represents an extracted feature. The number of feature dimensions can be customized, typically 8, 16, 32, 64, and so on, which are powers of 2.

[0072] As can be understood, during pre-training of large time series models, the feature extraction layer uses a multi-patch input projection module to pre-process historical wind power training samples (a two-dimensional array of historical power) and weather forecast training samples (a two-dimensional array of numerical weather forecasts). Specifically, the input two-dimensional array is divided into non-overlapping patches, patch sizes are selected based on the data sampling frequency, and each patch is mapped to a unified high-dimensional embedding representation using linear projection. This generates a two-dimensional array of historical power features and a two-dimensional array of weather features that distinguish time series and variables. This feature extraction layer not only effectively captures trends, cycles, and local features in the data, but also provides a unified high-dimensional embedding representation for subsequent layers.

[0073] S802. Input the power feature training samples and the weather feature training samples into the time series modeling layer, perform deep feature fusion through the multi-head attention module and the time series modeling module included in the time series modeling layer, and obtain the fused time series context feature training samples.

[0074] As can be understood, building on the aforementioned S801, the time series modeling layer employs an attention mechanism (Transformer) architecture specifically designed for modeling long- and short-term dependencies in time series. First, this layer uses an arbitrary variable attention mechanism to convert the input two-dimensional array of historical power features and weather features into a unified sequence representation, assigning each variable a unique variable ID to encode the relationship between them. Subsequently, a fully automatic attention mechanism, combined with Rotational Position Encoding (RoPE), enhances the modeling of temporal dependencies between different time steps, resulting in a fused two-dimensional array of temporal context features (temporal context feature training samples). Furthermore, to better integrate multivariate information, the time series modeling layer can also introduce a binary attention bias to ensure variable index invariance and enable efficient modeling of any number of variables. Furthermore, during computation, the time series modeling layer employs a feedforward network architecture, with the nonlinear activation function Swiglu replacing the traditional ReLU to improve expressiveness. Furthermore, RMS Norm and query-key normalization are employed to enhance the stability and training efficiency of the deep Transformer. After stacking multiple layers of Transformers, the time series modeling layer outputs a two-dimensional array of fused temporal context features, providing sufficient semantic information for the final power map.

[0075] S803: Utilize the prediction output mapping layer to map the high-dimensional temporal context feature training samples to the prediction distribution parameter space to obtain wind power prediction training samples.

[0076] As can be understood, based on the above-mentioned S802, the prediction output mapping layer utilizes a multi-patch-size output projection module to map the high-dimensional contextual features output by the time series modeling layer into the parameter space of the prediction distribution, thereby generating the final wind power prediction sequence (wind power prediction training samples). The specific implementation includes: first, a multi-patch-size output projection mechanism is adopted to adapt to the characteristics of data with different temporal resolutions, ensuring that both high-frequency and low-frequency time series information are fully utilized. Subsequently, a linear projection is used to map the hidden states of the Transformer encoder into the parameter space of the prediction distribution. In the output module, Moirai uses a mixed probability distribution scheme to accommodate the potential asymmetry, multimodality, and uncertainty of wind power data. Specifically, the prediction distribution can be composed of multiple parameterized distributions (such as Student's distribution, negative binomial distribution, lognormal distribution, or low-variance normal distribution), and optimal fit is achieved through learning weights and parameters. Finally, the prediction output mapping layer decodes the parameters of the prediction distribution and outputs the wind power prediction training samples.

[0077] The embodiments of the present disclosure provide a wind power prediction method based on a large time series model, which can accurately predict wind power.

[0078] Based on the above embodiments, Figure 9 A flow chart of a wind power prediction method based on a time series large model provided by an embodiment of the present disclosure, optionally, using a second training data set to fine-tune the parameters of the basic large model to obtain a target large model, specifically including the following steps: Figure 9 The following steps are shown: As can be understood, the parameters of the basic large model are updated in U steps using the training dataset of the target wind farm, where U is an integer greater than or equal to 1 and can be set as required. The parameter update instructions for the basic large model are as follows.

[0079] S901: Select a set number of training samples from a second training data set.

[0080] It can be understood that M training samples are randomly selected from all training samples of the target wind farm, where M is the set number, wherein M is an integer greater than or equal to 1, and the set M should not be greater than the total number of training samples contained in the target wind farm, and can be set according to needs.

[0081] S902 : Inputting the historical power training arrays and weather forecast training arrays in the set number of training samples into the basic large model respectively to obtain the corresponding set number of wind power prediction training arrays.

[0082] It is understandable that, based on the above S901, the historical power training array and weather forecast training array of the selected M training samples are respectively input into the basic large model to obtain corresponding M wind power prediction sequences (wind power prediction training arrays).

[0083] S903: Calculate a second average prediction error between a set number of wind power prediction training arrays and a set number of wind power actual training arrays.

[0084] It is understandable that based on the above S902, the average prediction error between the M wind power prediction training arrays (predicted values) and the corresponding M wind power actual training arrays (actual values) is calculated, wherein the error calculation method can adopt mean square error, mean absolute error, etc.

[0085] S904: Update the parameters of the basic large model according to the second average prediction error to obtain a target large model.

[0086] It can be understood that, based on the above S903, the parameters of the basic large model are updated according to the second average prediction error to obtain the target large model.

[0087] It is understandable that the training process of pre-training the time series large model to obtain the basic large model is similar to the training process of fine-tuning the basic large model to obtain the target large model, and will not be described in detail here.

[0088] The embodiments of the present disclosure provide a wind power prediction method based on a time series large model, which trains a target large model specific to a target wind farm.

[0089] Figure 10 A flow chart of a wind power prediction method based on a time series large model provided by an embodiment of the present disclosure, specifically including the following steps: Figure 10 The following steps are shown: S101. Obtain the time series pre-trained large model Moirai.

[0090] S102: Obtain source wind farm training data.

[0091] S103. Pre-train the time series prediction large model Moirai in sequence using sufficient training samples from multiple source wind farms across the country to obtain a large model base for wind power prediction.

[0092] S104: Obtain a two-dimensional array of historical power and a two-dimensional array of real-time weather forecasts of the target wind farm.

[0093] S105 , inputting the historical power two-dimensional array and the real-time weather forecast two-dimensional array of the target wind farm into the wind power prediction large model base to obtain the real-time wind power prediction sequence of the target wind farm.

[0094] S106: Obtain small sample training data of the target wind farm.

[0095] S107: Fine-tune the base parameters of the large-scale wind power prediction model using the limited training samples of the target wind farm to obtain a large-scale wind power prediction model exclusive to the target wind farm.

[0096] S108: Obtain a two-dimensional array of historical power and a two-dimensional array of real-time weather forecasts of the target wind farm.

[0097] S109 , inputting the historical power two-dimensional array and the real-time weather forecast two-dimensional array of the target wind farm into the exclusive wind power prediction large model of the target wind farm to obtain the real-time wind power prediction sequence of the target wind farm.

[0098] It is understandable that the specific implementation description of the above S101 to S109 can be found in the above embodiment and will not be repeated here.

[0099] Figure 11 The schematic diagram of the structure of a wind power prediction device based on a time series large model provided by an embodiment of the present disclosure. The wind power prediction device based on a time series large model provided by an embodiment of the present disclosure can execute the processing flow provided by the embodiment of the wind power prediction method based on a time series large model, such as Figure 11 As shown, the apparatus 1100 includes: A first acquisition unit 1101 is configured to acquire a large time series model and a first training data set of multiple source wind farms; A training unit 1102 is configured to pre-train the time series large model using the first training data set to obtain a basic large model for wind power prediction; A second acquiring unit 1103 is configured to acquire a historical wind power array and a future weather forecast array of a target wind farm, wherein the multiple source wind farms do not include the target wind farm; The prediction unit 1104 is used to use the historical wind power array and the future weather forecast array as inputs of a target large model to predict the future wind power array of the target wind farm, wherein the target large model is a basic large model or is obtained by fine-tuning the basic large model using a second training data set of the target wind farm.

[0100] Optionally, the apparatus 1100 is further configured to: Acquire a second training data set of the target wind farm, wherein the sample data volume and sampling period of the second training data set are smaller than those of the first training data set; In a case where the second training data set includes at least one historical wind power training array, the second training data set is used to fine-tune the parameters of the basic large model to obtain a target large model, wherein the target large model is specifically used for wind power prediction of the target wind farm; or When the second training data set does not include the historical wind power training array, the basic large model is directly used as the target large model.

[0101] Optionally, the training unit 1102 is configured to: Inputting the training data sets corresponding to the multiple wind farms in the first training data set into the time series large model in sequence, targeting a current wind farm among the multiple wind farms, wherein the current wind farm contains sufficient training samples under all weather conditions; Selecting a group of training samples from sufficient training samples, wherein the group of training samples includes historical wind power training samples, weather forecast training samples, and actual wind power training samples; Inputting characteristic parameters of historical wind power training samples and weather forecast training samples into a time series large model to obtain a set of corresponding wind power prediction training samples, wherein the time series large model has model parameters trained based on the corresponding training sample set of the previous wind farm; Calculating a first average prediction error between wind power prediction training samples and wind power actual training samples; Based on the first average prediction error, all layer parameters of the time series large model are updated to obtain the optimal parameters under all weather conditions of the current wind farm, until the training of the time series large model with the training data set corresponding to the last wind farm is completed, and the basic large model for wind power prediction is obtained.

[0102] Among them, the time series large model includes the input feature extraction layer, the time series modeling layer and the prediction output mapping layer.

[0103] Optionally, the training unit 1102 is configured to: The input feature extraction layer is used to extract features from historical wind power training samples and weather forecast training samples to obtain power feature training samples and weather feature training samples with time series and variable distinguishability. The power feature training samples and weather feature training samples are input into the time series modeling layer. The multi-head attention module and the time series modeling module included in the time series modeling layer perform deep feature fusion to obtain the fused time series context feature training samples. The prediction output mapping layer is used to map high-dimensional temporal context feature training samples to the prediction distribution parameter space to obtain wind power prediction training samples.

[0104] Optionally, the apparatus 1100 is further configured to: Selecting a set number of training samples from the second training data set; Inputting the historical power training array and the weather forecast training array in the set number of training samples into the basic large model respectively to obtain the corresponding set number of wind power prediction training arrays; Calculating a second average prediction error between a set number of wind power prediction training arrays and a set number of wind power actual training arrays; The parameters of the basic large model are updated according to the second average prediction error to obtain the target large model.

[0105] Among them, the historical wind power training samples and weather forecast training samples refer to two-dimensional arrays; The historical wind power training sample includes a first time dimension and a power category dimension. The first time dimension includes multiple first moments for power sampling, and the multiple first moments differ by a first set period. The power category dimension includes a first time feature code, historical power data, and a wind farm number. The first time feature code corresponds to the first time dimension. The weather forecast training samples include a second time dimension and a meteorological category dimension, wherein the second time dimension includes multiple second moments for weather sampling, and the multiple second moments are separated by a second set period. The meteorological category dimension includes a second time feature code and multiple meteorological forecast data types. The multiple meteorological forecast data types include forecast wind condition data and other meteorological forecast data. The forecast wind condition data includes forecast wind speed data and forecast wind direction data. The other meteorological forecast data includes at least one of forecast air pressure data, forecast temperature data and forecast humidity data.

[0106] Optionally, the second acquiring unit 1103 is configured to: Obtaining a historical wind power array of a target wind farm within a set historical period and a future weather forecast array within a set future period; The historical wind power array and the future weather forecast array are used as inputs of the target large model to predict the future wind power array of the target wind farm within a set period of time in the future.

[0107] Figure 11 The wind power prediction device based on the time series large model of the illustrated embodiment can be used to implement the technical solution of the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0108] Figure 12 This is a schematic diagram of the structure of the electronic device provided by the embodiment of the present disclosure. Figure 12, which shows a schematic structural diagram of an electronic device 1200 suitable for implementing the embodiments of the present disclosure. The electronic device 1200 in the embodiments of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (such as in-vehicle navigation terminals), wearable electronic devices, and fixed terminals such as digital TVs, desktop computers, smart home devices, and the like. Figure 12 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0109] like Figure 12 As shown, electronic device 1200 may include a processing device 1201 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage device 1208 into a random access memory (RAM) 1203 to implement the wind power forecasting method based on a large time series model according to the embodiments of the present disclosure. RAM 1203 also stores various programs and data required for the operation of electronic device 1200. Processing device 1201, ROM 1202, and RAM 1203 are interconnected via a bus 1204. An input / output (I / O) interface 1205 is also connected to bus 1204.

[0110] Typically, the following devices may be connected to the I / O interface 1205: an input device 1206 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 1207 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1208 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1209. The communication device 1209 may allow the electronic device 1200 to communicate with other devices wirelessly or by wire to exchange data. Figure 12 The electronic device 1200 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0111] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart, thereby implementing the wind power prediction method based on the time series large model as described above. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 1209, or installed from the storage device 1208, or installed from the ROM 1202. When the computer program is executed by the processing device 1201, the above-mentioned functions defined in the method of the embodiment of the present disclosure are performed.

[0112] It should be noted that the computer-readable medium described above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable storage media may include, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wire, optical cable, RF (radio frequency), or any suitable combination thereof.

[0113] In some embodiments, the client and server can communicate using any currently known or later developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or later developed network.

[0114] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0115] Optionally, when the above one or more programs are executed by the electronic device, the electronic device may also execute other steps described in the above embodiments.

[0116] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including, but not limited to, object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0117] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0118] The units involved in the embodiments described in this disclosure may be implemented in software or hardware, wherein the name of a unit does not necessarily limit the unit itself.

[0119] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0120] In the context of the present disclosure, 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. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor 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.

[0121] It should be noted that, in this document, relational terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or gateway that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or gateway. Without further restriction, the elements defined by the sentence "including a wind power forecast based on a large time series model" do not exclude the existence of other identical elements in the process, method, article or gateway that includes the elements.

[0122] The foregoing description is intended only to provide specific embodiments of the present disclosure, intended to enable those skilled in the art to understand and implement the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the embodiments described herein, but rather to be construed in the broadest manner consistent with the principles and novel features disclosed herein.

Claims

1. A wind power prediction method based on a time series large model, characterized in that: include: Obtaining a first training data set of a time series large model and multiple source wind farms; Pre-training the time series large model using the first training data set to obtain a basic large model for wind power prediction; wherein the basic large model includes a feature extraction layer, a wind power time series modeling layer, and a wind power mapping layer, the feature extraction layer is used to extract features of wind power and weather forecast data, the wind power time series modeling layer is used to fuse the extracted features, and the wind power mapping layer is used to map the fused features to a prediction distribution parameter space to generate a final wind power prediction sequence; Acquire a historical wind power array and a future weather forecast array of a target wind farm, wherein the multiple source wind farms do not include the target wind farm; The historical wind power array and the future weather forecast array are used as inputs of a target large model to predict the future wind power array of the target wind farm, wherein the target large model is the basic large model or is obtained by fine-tuning the basic large model using a second training data set of the target wind farm.

2. The method according to claim 1, characterized in that Before obtaining the historical wind power array and the future weather forecast array of the target wind farm, the method further includes: Acquire a second training data set of the target wind farm, wherein the sample data volume and sampling period of the second training data set are smaller than those of the first training data set; In a case where the second training data set includes at least one historical wind power training array, the parameters of the basic large model are fine-tuned using the second training data set to obtain a target large model, wherein the target large model is specifically used for wind power prediction of the target wind farm; or In the case that the second training data set does not include a historical wind power training array, the basic large model is directly used as the target large model.

3. The method according to claim 1, characterized in that The method of pre-training the time series large model using the first training data set to obtain a basic large model for wind power prediction includes: Inputting the training data sets corresponding to the plurality of wind farms in the first training data set into the time series large model in sequence, for a current wind farm among the plurality of wind farms, wherein the current wind farm contains sufficient training samples under all weather conditions; Selecting a group of training samples from the sufficient training samples, wherein the group of training samples includes historical wind power training samples, weather forecast training samples, and actual wind power training samples; Inputting characteristic parameters of the historical wind power training samples and the weather forecast training samples into the time series large model to obtain a set of corresponding wind power prediction training samples, wherein the time series large model has model parameters trained based on a corresponding training sample set of a previous wind farm; Calculating a first average prediction error between the wind power prediction training samples and the wind power actual training samples; All layer parameters of the time series large model are updated according to the first average prediction error to obtain the optimal parameters under all weather conditions of the current wind farm, until the training of the time series large model with the training data set corresponding to the last wind farm is completed to obtain the basic large model for wind power prediction.

4. The method according to claim 3, characterized in that The time series large model includes an input feature extraction layer, a time series modeling layer, and a prediction output mapping layer. The feature parameters of the historical wind power training samples and the weather forecast training samples are input into the time series large model to obtain a set of corresponding wind power prediction training samples, including: Using the input feature extraction layer to perform feature extraction on the historical wind power training samples and the weather forecast training samples to obtain power feature training samples and weather feature training samples with time series and variable distinguishability; Inputting the power feature training samples and the weather feature training samples into the time series modeling layer, performing deep feature fusion through the multi-head attention module and the time series modeling module included in the time series modeling layer to obtain fused time series context feature training samples; The prediction output mapping layer is used to map the high-dimensional temporal context feature training samples to the prediction distribution parameter space to obtain wind power prediction training samples.

5. The method according to claim 2, characterized in that The method of fine-tuning the parameters of the basic large model using the second training data set to obtain a target large model includes: Selecting a set number of training samples from the second training data set; Inputting the historical power training array and the weather forecast training array in the set number of training samples into the basic large model respectively to obtain the corresponding set number of wind power prediction training arrays; Calculating a second average prediction error between the set number of wind power prediction training arrays and the set number of wind power actual training arrays; The parameters of the basic large model are updated according to the second average prediction error to obtain a target large model.

6. The method according to claim 4, characterized in that The historical wind power training samples and the weather forecast training samples refer to two-dimensional arrays, wherein, The historical wind power training sample includes a first time dimension and a power category dimension, wherein the first time dimension includes a plurality of first moments for power sampling, and the plurality of first moments are separated by a first set period, and the power category dimension includes a first time feature code, historical power data, and a wind farm number, and the first time feature code corresponds to the first time dimension; The weather forecast training sample includes a second time dimension and a meteorological category dimension, wherein the second time dimension includes multiple second moments for weather sampling, and the multiple second moments are separated by a second set period. The meteorological category dimension includes a second time feature code and multiple meteorological forecast data types, and the multiple meteorological forecast data types include forecast wind condition data and other meteorological forecast data. The forecast wind condition data includes forecast wind speed data and forecast wind direction data, and the other meteorological forecast data includes at least one of forecast air pressure data, forecast temperature data and forecast humidity data.

7. The method according to claim 1, characterized in that The step of obtaining the historical wind power array and future weather forecast array of the target wind farm includes: Obtaining a historical wind power array of a target wind farm within a set historical period and a future weather forecast array within a set future period; The method of using the historical wind power array and the future weather forecast array as inputs of a target large model to predict the future wind power array of the target wind farm includes: The historical wind power array and the future weather forecast array are used as inputs of the target large model to predict the future wind power array of the target wind farm within the future set time period.

8. A wind power prediction device based on a time series large model, characterized in that: include: A first acquisition unit is used to acquire a first training data set of a large time series model and multiple source wind farms; a training unit, configured to pre-train the time series large model using the first training data set to obtain a basic large model for wind power prediction; wherein the basic large model includes a feature extraction layer, a wind power time series modeling layer, and a wind power mapping layer; the feature extraction layer is configured to extract features of wind power and weather forecast data; the wind power time series modeling layer is configured to fuse the extracted features; and the wind power mapping layer is configured to map the fused features to a prediction distribution parameter space to generate a final wind power prediction sequence; a second acquiring unit, configured to acquire a historical wind power array and a future weather forecast array of a target wind farm, wherein the plurality of source wind farms does not include the target wind farm; a prediction unit, configured to use the historical wind power array and the future weather forecast array as inputs of a target large model to predict the future wind power array of the target wind farm, wherein the target large model is the basic large model or is obtained by fine-tuning the basic large model using a second training data set of the target wind farm.

9. An electronic device, characterized in that: include: Memory; processor; as well as computer programs; The computer program is stored in the memory and is configured to be executed by the processor to implement the wind power prediction method based on a time series large model as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind power prediction method based on a time series large model as claimed in any one of claims 1 to 7 are implemented.

Images