Adaptive and dynamic self-correction method and system for new energy power prediction model

Abstract

The invention discloses an adaptive and dynamic self-correction method and system for a new energy power prediction model, and belongs to the technical field of new energy power prediction and artificial intelligence, and the method comprises the steps: constructing a feature portrait vector of a new energy station; establishing a power prediction evaluation index system, and performing quantitative evaluation on the power prediction model; adaptively matching an optimal power prediction model from a pre-established power prediction model library; performing dynamic self-correction on the model; the data scheduling module interacts with a station SCADA system or a weather forecast system to form a self-adaptive system with feature portrait construction, model matching, power prediction, evaluation, correction and updating as cycles, and self-adaption and dynamic self-correction of the model are achieved. According to the method, intelligent model selection, self-learning and dynamic parameter correction among different stations can be realized, the robustness of the model under a multi-scene condition is improved, and the method is suitable for power prediction optimization of multiple types of new energy stations and intelligent operation and maintenance of the model.

Description

Technical Field

[0001] This invention belongs to the field of new energy power prediction and artificial intelligence technology, specifically relating to an adaptive and dynamic self-correction method and system for a new energy power prediction model. Background Technology

[0002] With the rapid growth of installed capacity of new energy sources, the high proportion of renewable energy such as wind power and photovoltaics has brought significant challenges to the safety and economic dispatch of the power system. Power forecasting, as a key link supporting the consumption of new energy and electricity market transactions, directly affects the operational safety and economy of the system due to its accuracy and stability. Existing new energy power forecasting models typically employ fixed structures and parameters, undergoing offline training based on historical data during the initial modeling phase. However, due to significant differences in meteorological conditions, geographical environment, equipment type, and operation and maintenance strategies among different power plants, the adaptability of new energy power forecasting models is poor when migrating across power plants, leading to a significant increase in prediction errors. Furthermore, with the passage of time, meteorological patterns and operational characteristics change frequently, making it difficult for static models to respond to environmental changes in real time, lacking online self-correction and dynamic optimization mechanisms. In addition, current power forecasting evaluation systems mostly focus on single accuracy indicators, failing to comprehensively reflect forecasting performance under different time scales, trading demands, and risk constraints, and lacking a unified indicator library and dynamic evaluation framework. In summary, current power forecasting for new energy sources struggles to respond to environmental changes in real time, lacks online adaptive and correction capabilities and dynamic optimization mechanisms, and fails to comprehensively reflect forecasting performance under different time scales, trading demands, and risk constraints. The indicator library and dynamic evaluation framework need further optimization. Summary of the Invention

[0003] This invention provides an adaptive and dynamic self-correction method and system for new energy power prediction models. The aim is to solve the problems in current new energy power prediction, such as difficulty in responding to environmental changes in real time, lack of online adaptive and correction capabilities and dynamic optimization mechanisms, difficulty in comprehensively reflecting prediction performance under different time scales, trading demands and risk constraints, and the need for further optimization of the indicator library and dynamic evaluation framework.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides an adaptive and dynamic self-correction method for a new energy power prediction model, comprising the following steps: S1. Construct a feature profile vector for new energy power stations. This involves extracting multi-dimensional features from the meteorological, geographical, equipment, and operational data of the target power station, and then performing feature normalization, dimensionality reduction, and cluster analysis to obtain the feature profile vector. S2. Establish a power prediction and evaluation index system; The power prediction evaluation index system includes multiple evaluation indicators, which are used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability, reliability and economy. S3. Based on the feature profile vector and power prediction evaluation index system, adaptively match the optimal power prediction model from the pre-established power prediction model library. The power prediction model library contains models of various algorithm types, and each algorithm type comes with its optimal combination of applicable tags. S4. Based on real-time monitored meteorological and operational status data, the optimal power prediction model is dynamically self-corrected. S5. Through the data scheduling module, it interacts with the station's SCADA system or meteorological forecasting system to form an adaptive system with a cycle of feature profile construction, model matching, power prediction, evaluation, correction and updating, so as to realize the self-adaptation and dynamic self-correction of the power prediction model.

[0005] In some implementations, in S1, multidimensional feature extraction includes meteorological features, geographical features, equipment features, and operational features; wherein: Meteorological features include irradiance, temperature, humidity, wind speed, air pressure, and cloud cover; geographical features include latitude, altitude, terrain slope, and orientation; equipment features include component type, capacity, inverter configuration, and grid topology; and operational features include output rate, power fluctuation rate, equipment health, and historical deviation distribution. The clustering analysis uses K-means++ or graph embedding methods.

[0006] In some implementations, in S2, the evaluation index system includes root mean square error (RMSE), mean absolute error (MAE), continuous sort probability score (CRPS), quantile error, event prediction accuracy, and reliability index.

[0007] In some implementations, in S3, the power prediction model library includes LSTM, Transformer, TimeXer, and LightGBM algorithm models; the optimal applicable label combination includes meteorological intervals, geographical zones, station capacity, and operating modes.

[0008] In some implementations, in S3, adaptive matching employs meta-learning or transfer learning strategies; meta-learning includes cross-task learning to extract general knowledge; transfer learning includes transferring model knowledge trained in one field or region to a new target field.

[0009] In some implementations, in S4, the dynamic self-correction includes an online optimization module, a self-learning module, and a label update module; wherein: The online optimization module is used to evaluate the latest prediction error using a sliding window and trigger fine-tuning of local parameters; the self-learning module is used to introduce a small number of new samples for incremental training or parameter transfer; and the label update module is used to dynamically update the station profile labels based on meteorological change patterns.

[0010] Furthermore, in S4, the online optimization module uses a sliding window to evaluate the latest prediction error. Specifically, within a sliding window of a preset time length, it calculates the error index between the predicted value and the actual value of the optimal power prediction model. When the error index exceeds a preset threshold, it triggers a local fine-tuning of the parameters of the optimal power prediction model.

[0011] Furthermore, in S4, the self-learning module performs incremental training or parameter transfer specifically by: collecting the latest meteorological and power data from the station as new samples, using the new samples to incrementally update the matched optimal power prediction model, or transferring the parameters learned from the source station model to the target station model.

[0012] Furthermore, in S4, the label update module dynamically updates the site profile labels by adjusting the meteorological-related feature labels in the feature profile vector based on real-time monitored changes in meteorological patterns, so that the optimal power prediction model match is more consistent with the current operating scenario.

[0013] This invention also provides an adaptive and dynamic self-correcting system for a new energy power prediction model, used to implement the aforementioned adaptive and dynamic self-correcting method for the new energy power prediction model. The system includes a feature profiling vector module, a quantitative evaluation module, an adaptive matching module, a dynamic self-correcting module, and a continuous adjustment module, wherein: Feature Profile Vector Module: Used to construct feature profile vectors for new energy power stations. It extracts multi-dimensional features from meteorological, geographical, equipment, and operational data of the target power station, and then performs feature normalization, dimensionality reduction, and cluster analysis to obtain the feature profile vector. Quantitative evaluation module: used to establish a power prediction evaluation index system; The power prediction evaluation index system includes multiple evaluation indicators, which are used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability, reliability and economy. Adaptive matching module: Used to adaptively match the optimal power prediction model from a pre-established power prediction model library based on feature profile vectors and power prediction evaluation index system; The power prediction model library contains models of various algorithm types, and each algorithm type comes with its optimal combination of applicable tags. Dynamic self-calibration module: used to dynamically self-calibrate the optimal power prediction model based on real-time monitored meteorological and operational status data; Continuous Adjustment Module: This module interacts with the station's SCADA system or weather forecasting system through the data scheduling module to form an adaptive system that iterates through feature profile construction, model matching, power prediction, evaluation, correction, and updates. This enables the power prediction model to adapt and dynamically self-correct.

[0014] Compared with existing technologies, the adaptive and dynamic self-correction method and system for predicting new energy power in this invention have the following advantages: The adaptive and dynamic self-correction method of the new energy power prediction model of this invention constructs a feature profile vector of new energy power stations, extracts multi-dimensional features from the meteorological, geographical, equipment and operation data of the target power station, and performs normalization, dimensionality reduction and cluster analysis to form a unified and quantitative expression of the power station features. This overcomes the defects of poor adaptability and increased prediction error when the model is transferred across power stations due to differences in power station features. By establishing a power prediction evaluation index system covering accuracy, stability, reliability, and economy, multi-index comprehensive evaluation replaces the limitations of traditional single-accuracy indexes, better meeting the needs of power system dispatch for prediction results. Furthermore, based on feature profile vectors and the evaluation index system, the optimal power prediction model is adaptively matched from a pre-established model library containing various algorithm types and optimal applicable label combinations, achieving intelligent matching between the model and station characteristics and avoiding the insufficient generalization ability problem caused by fixed model structures. Through dynamic self-correction of the optimal model based on real-time monitored meteorological and operational status data, the model can respond in real-time to environmental changes and operational status fluctuations, maintaining prediction performance through online adjustments. Through interaction with the station's SCADA system or meteorological forecasting system via the data dispatch module, an adaptive system is formed, with feature profile construction, model matching, power prediction, evaluation, correction, and updating as its cycle. This achieves closed-loop management and continuous optimization of the entire prediction process, ensuring the stability and reliability of the model in long-term operation and improving the overall economy and dispatch efficiency of new energy power prediction. Attached Figure Description

[0015] The accompanying drawings are provided to further understand the invention and constitute a part of this invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0016] Figure 1 This is a flowchart illustrating an adaptive and dynamic self-correction method for a new energy power prediction model according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0019] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0020] It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this regard, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0021] In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0022] How can we provide a method that can integrate power plant feature profiles, adaptively select the optimal prediction strategy, and have dynamic self-correction capabilities to achieve intelligent management and continuous optimization of power prediction models for new energy power plants, realize intelligent model selection, self-learning, and dynamic parameter correction among different types of power plants, and improve the applicability and robustness of prediction models under multi-temporal and multi-scenario conditions?

[0023] like Figure 1 As shown, the present invention provides an adaptive and dynamic self-correction method for a new energy power prediction model, comprising the following steps: S1. Construct a feature profile vector for new energy power stations. This involves extracting multi-dimensional features from the meteorological, geographical, equipment, and operational data of the target power station, and then performing feature normalization, dimensionality reduction, and cluster analysis to obtain the feature profile vector. S2. Establish a power prediction and evaluation index system. The power prediction evaluation index system is a comprehensive assessment framework consisting of various evaluation indicators (RMSE, MAE, CRPS, reliability index, etc.) used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability and economy. The power prediction evaluation index system includes multiple evaluation indicators, which are used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability, reliability and economy. S3. Based on the feature profile vector and power prediction evaluation index system, adaptively match the optimal power prediction model from the pre-established power prediction model library. The power prediction model library contains models of various algorithm types, and each algorithm type comes with its optimal combination of applicable tags. S4. Based on real-time monitored meteorological and operational status data, perform dynamic self-calibration on the optimal power prediction model (a dynamic optimization mechanism in which the model automatically adjusts its parameters or structure according to the latest data or environmental changes during operation to maintain prediction performance). S5. Through the data scheduling module, it interacts with the station's SCADA (Supervisory Control and Data Acquisition System) or meteorological forecasting system to form an adaptive system with a cycle of feature profile construction, model matching, power prediction, evaluation, correction and updating, so as to realize the self-adaptation and dynamic self-correction of the power prediction model.

[0024] This invention presents an adaptive and dynamic self-correction method for a new energy power prediction model. By leveraging power plant feature profiling, the model rapidly adapts to different geographical regions and equipment conditions, exhibiting strong generalization capabilities. It enables real-time self-correction and parameter self-evolution during model operation. Combining power trading and provincial assessment requirements, it provides a multi-indicator fusion evaluation system, achieving multi-dimensional evaluation and optimization. It supports various algorithm frameworks and prediction time scales, allowing for flexible deployment on new energy prediction platforms and demonstrating high scalability. Through continuous optimization of model parameters and prediction strategies, this invention makes power prediction results more stable and reliable, thereby reducing system reserve capacity requirements and improving economic and dispatch efficiency.

[0025] This invention presents an adaptive and dynamic self-correction method for new energy power prediction models. It proposes a multi-dimensional profiling method integrating meteorological, geographical, equipment, and operational characteristics to provide a unified feature representation for the adaptive selection of power prediction models. A multi-objective index library covering RMSE, MAE, CRPS, and event prediction accuracy is established to achieve comprehensive evaluation of prediction accuracy, stability, and reliability. A model library including algorithms such as LSTM, Transformer, TimeXer, and LightGBM is constructed, and automatic matching between models and site profiles is achieved through meta-learning or transfer learning. Dynamic adjustment and self-evolution of model performance are realized through online optimization, self-learning, and label updates. An automatic loop of feature profiling, model matching, prediction, evaluation, correction, and updating is formed to achieve continuous optimization of the prediction model. Among them, RMSE (Root Mean Square Error) measures the deviation between predicted and actual values, reflecting the overall level of prediction error. A smaller value indicates higher prediction accuracy. MAE (Mean Absolute Error) represents the average absolute difference between predicted and actual values, used to evaluate the average bias of the model and has good interpretability. CRPS (Continuous Ranked Probability Score) is a performance metric for probabilistic prediction, measuring the distance between the predicted distribution and the actual observations, comprehensively reflecting the accuracy and uncertainty of the prediction results. A smaller CRPS indicates a prediction distribution closer to the actual results. Transformer: A deep learning model based on self-attention mechanisms, capable of efficiently handling multi-dimensional time series and complex feature interactions, it is an important model structure in the field of new energy prediction. LSTM (Long Short-Term Memory): A recurrent neural network structure used to capture long-term dependencies in time series, commonly used in new energy power prediction tasks. TimeXer: A time-series forecasting model enhanced by exogenous variables. Through time embedding and the interaction structure of exogenous driving features, it improves the stability and generalization ability of renewable energy power forecasting under severe meteorological fluctuations. LightGBM (Light Gradient Boosting Machine): An efficient machine learning algorithm based on a gradient boosting framework, characterized by fast training speed and strong generalization performance, suitable for processing large-scale renewable energy forecasting data. Adaptive Selection: Automatically selects the optimal model and parameter combination based on different site characteristics and evaluation requirements, achieving personalized model deployment and optimal matching.

[0026] This invention presents an adaptive and dynamic self-correcting method for new energy power prediction models. This method constructs profiles based on multi-dimensional features and achieves automatic model matching. Through a multi-objective power prediction evaluation system, it integrates indicators such as CRPS, RMSE, and MAE to guide dynamic model optimization. The model's dynamic self-correcting mechanism includes online optimization, self-learning, and label update modules. Ultimately, an adaptive loop system is formed, enabling continuous self-optimization and stable operation of the power prediction model.

[0027] This invention provides an adaptive and dynamic self-correcting system for new energy power prediction models, comprising a feature profiling vector module, a quantitative evaluation module, an adaptive matching module, a dynamic self-correcting module, and a continuous adjustment module, wherein: Feature Profile Vector Module: Used to construct feature profile vectors for new energy power stations. It extracts multi-dimensional features from meteorological, geographical, equipment, and operational data of the target power station, and then performs feature normalization, dimensionality reduction, and cluster analysis to obtain the feature profile vector. Quantitative evaluation module: used to establish a power prediction evaluation index system; The power prediction evaluation index system includes multiple evaluation indicators, which are used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability, reliability and economy. Adaptive matching module: Used to adaptively match the optimal power prediction model from a pre-established power prediction model library based on feature profile vectors and power prediction evaluation index system; The power prediction model library contains models of various algorithm types, and each algorithm type comes with its optimal combination of applicable tags. Dynamic self-calibration module: used to dynamically self-calibrate the optimal power prediction model based on real-time monitored meteorological and operational status data; Continuous Adjustment Module: This module interacts with the station's SCADA system or weather forecasting system through the data scheduling module to form an adaptive system that iterates through feature profile construction, model matching, power prediction, evaluation, correction, and updates. This enables the power prediction model to adapt and dynamically self-correct.

[0028] The adaptive and dynamic self-correction method of the new energy power prediction model of the present invention will be further described in detail below through specific embodiments.

[0029] In this embodiment, the adaptive and dynamic self-correction method of the new energy power prediction model is performed according to the following steps: Step 1: Constructing a characteristic profile of new energy power stations; Multidimensional feature extraction is performed on the meteorological, geographical, equipment, and operational data of the target site, including but not limited to: meteorological features (irradiance, temperature, humidity, wind speed, air pressure, cloud cover); geographical features (latitude, altitude, terrain slope, orientation); equipment features (component type, capacity, inverter configuration, grid topology); and operational features (output rate, power fluctuation rate, equipment health, historical deviation distribution).

[0030] Feature normalization, dimensionality reduction, and cluster analysis are used to form feature profile vectors for new energy power stations, which are then used for subsequent model selection and adaptive mapping. Cluster analysis employs K-means++ (K-means Plus Plus, an algorithm that increases the K-means increment to keep the initial cluster centers as far apart as possible, improving the quality of clustering results and the convergence speed of the algorithm, and avoiding the algorithm from getting trapped in poor local optima) or graph embedding methods; Step 2: Establishment of a power prediction and evaluation index system; A power forecasting evaluation index library is established, encompassing multiple time scales (ultra-short-term, short-term, intraday) and multiple objectives (accuracy, reliability, stability, and economy). Indicators include, but are not limited to, RMSE, MAE, CRPS, quantile error, event prediction accuracy, and reliability index. Among these, CRPS, as an indicator comprehensively reflecting the accuracy of forecast distribution and the ability to characterize uncertainty, further enhances the comprehensiveness and diversity of model forecasting evaluation. Based on the assessment requirements of different provinces and power trading scenarios, a hierarchical weighted evaluation system is constructed to guide the dynamic selection and optimization of models.

[0031] Step 3: Matching power prediction model library with labels; A power prediction model library was established, encompassing various algorithm types (such as LSTM, Transformer, Graph Xer, LightGBM, etc.). Each model was accompanied by its optimal combination of applicable labels (meteorological intervals, geographical zones, station capacity, operating mode, etc.). Based on the station profile vector and indicator priorities, meta-learning or transfer learning strategies were employed to achieve adaptive matching and initialization of the optimal model for the target station.

[0032] Meta-Learning is an AI method that learns how to learn, extracting general knowledge through cross-task learning to enable models to quickly adapt to new weather stations or meteorological scenarios. Transfer Learning involves transferring knowledge from a model trained at one weather station or region to a new target weather station to reduce retraining costs and improve prediction performance.

[0033] Step 4: Model dynamic self-calibration; A time-series-based operational label library is constructed to monitor changes in meteorological and operational status in real time. When model performance deteriorates or scene labels shift, self-correction is achieved through the following mechanisms: 1) Online optimization module: The latest prediction error is evaluated using a sliding window, triggering fine-tuning of local parameters; 2) Self-learning module: A small number of new samples are introduced for incremental training or parameter transfer; 3) Label update module: The station profile labels are dynamically updated according to meteorological change patterns, enabling the prediction model to evolve automatically.

[0034] Step 5: Model Adaptive Loop and System Implementation; This invention constructs a closed-loop system through the above modules. It achieves interaction with the power plant SCADA system or meteorological forecasting system through API interface and data scheduling module, forming an adaptive loop of feature profiling, model matching, prediction, evaluation, correction and updating. This enables intelligent management and continuous optimization of the power prediction model of new energy power plants, thereby improving the reliability of new energy power plants.

[0035] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Anyone skilled in the art can readily implement the present invention according to the description and above. Any modifications, alterations, or equivalent variations made using the technical content disclosed above are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. An adaptive and dynamic self-correcting method for new energy power prediction models, characterized in that, Includes the following steps: S1. Construct a feature profile vector for new energy power stations. This involves extracting multi-dimensional features from the meteorological, geographical, equipment, and operational data of the target power station, and then performing feature normalization, dimensionality reduction, and cluster analysis to obtain the feature profile vector. S2. Establish a power prediction and evaluation index system; The power prediction evaluation index system includes multiple evaluation indicators, which are used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability, reliability and economy. S3. Based on the feature profile vector and power prediction evaluation index system, adaptively match the optimal power prediction model from the pre-established power prediction model library. The power prediction model library contains models of various algorithm types, and each algorithm type comes with its optimal combination of applicable tags. S4. Based on real-time monitored meteorological and operational status data, the optimal power prediction model is dynamically self-corrected. S5. Through the data scheduling module, it interacts with the station's SCADA system or meteorological forecasting system to form an adaptive system with a cycle of feature profile construction, model matching, power prediction, evaluation, correction and updating, so as to realize the self-adaptation and dynamic self-correction of the power prediction model.

2. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 1, characterized in that, In S1, multidimensional feature extraction includes meteorological features, geographical features, equipment features, and operational features; wherein: Meteorological features include irradiance, temperature, humidity, wind speed, air pressure, and cloud cover; geographical features include latitude, altitude, terrain slope, and orientation; equipment features include component type, capacity, inverter configuration, and grid topology; and operational features include output rate, power fluctuation rate, equipment health, and historical deviation distribution. The clustering analysis uses K-means++ or graph embedding methods.

3. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 1, characterized in that, In S2, the evaluation index system includes root mean square error (RMSE), mean absolute error (MAE), continuous sort probability score (CRPS), quantile error, event prediction accuracy, and reliability index.

4. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 1, characterized in that, In S3, the power prediction model library includes LSTM, Transformer, TimeXer and LightGBM algorithm models; the optimal applicable label combination includes meteorological interval, geographical region, station capacity and operation mode.

5. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 1, characterized in that, In S3, adaptive matching employs a meta-learning or transfer learning strategy; Meta-learning includes cross-task learning to extract general knowledge; transfer learning includes transferring model knowledge that has been trained in one field or region to a new target field.

6. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 1, characterized in that, In step S4, the dynamic self-correction includes an online optimization module, a self-learning module, and a label update module; wherein: The online optimization module is used to evaluate the latest prediction error using a sliding window and trigger fine-tuning of local parameters; the self-learning module is used to introduce a small number of new samples for incremental training or parameter transfer; and the label update module is used to dynamically update the station profile labels based on meteorological change patterns.

7. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 6, characterized in that, The online optimization module uses a sliding window to evaluate the latest prediction error. Specifically, within a sliding window of a preset time length, it calculates the error index between the predicted value and the actual value of the optimal power prediction model. When the error index exceeds a preset threshold, it triggers a local fine-tuning of the parameters of the optimal power prediction model.

8. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 6, characterized in that, The self-learning module performs incremental training or parameter transfer in the following ways: collecting the latest meteorological and power data from the station as new samples, using the new samples to incrementally update the matched optimal power prediction model, or transferring the parameters learned from the source station model to the target station model.

9. The adaptive and dynamic self-correcting method for the new energy power prediction model according to claim 6, characterized in that, The tag update module dynamically updates the site profile tags, specifically by adjusting the meteorological-related feature tags in the feature profile vector based on real-time monitored changes in meteorological patterns, so that the optimal power prediction model matches the current operating scenario more closely.

10. An adaptive and dynamic self-correcting system for a new energy power prediction model, used to implement the adaptive and dynamic self-correcting method for the new energy power prediction model according to any one of claims 1-9, characterized in that, It includes a feature profiling vector module, a quantization evaluation module, an adaptive matching module, a dynamic self-correction module, and a continuous adjustment module, among which: Feature Profile Vector Module: Used to construct feature profile vectors for new energy power stations. It extracts multi-dimensional features from meteorological, geographical, equipment, and operational data of the target power station, and then performs feature normalization, dimensionality reduction, and cluster analysis to obtain the feature profile vector. Quantitative evaluation module: used to establish a power prediction evaluation index system; The power prediction evaluation index system includes multiple evaluation indicators, which are used to quantitatively evaluate the power prediction model from the dimensions of accuracy, stability, reliability and economy. Adaptive matching module: Used to adaptively match the optimal power prediction model from a pre-established power prediction model library based on feature profile vectors and power prediction evaluation index system; The power prediction model library contains models of various algorithm types, and each algorithm type comes with its optimal combination of applicable tags. Dynamic self-calibration module: used to dynamically self-calibrate the optimal power prediction model based on real-time monitored meteorological and operational status data; Continuous Adjustment Module: This module interacts with the station's SCADA system or weather forecasting system through the data scheduling module to form an adaptive system that iterates through feature profile construction, model matching, power prediction, evaluation, correction, and updates. This enables the power prediction model to adapt and dynamically self-correct.

Images