Method and system for realizing new energy power generation prediction
By employing multi-level data preprocessing and an improved long short-term memory network model, combined with attention mechanisms and optimizers, the problems of low accuracy and high computational load in new energy power generation prediction are solved, achieving efficient and accurate new energy power generation prediction.
Patent Information
- Application Number
- CN202511054513.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-30
- Publication Date
- 2025-11-21
AI Technical Summary
Existing methods for predicting new energy power generation cannot fully consider complex nonlinear factors, resulting in low prediction accuracy. Furthermore, deep learning algorithms require a large amount of computation and a long training time, making them difficult to apply widely in practical engineering.
Multi-level data preprocessing, principal component analysis, and mutual information method are used to screen features. An improved long short-term memory network model is combined with an attention mechanism. An adaptive moment estimation optimizer and an early stopping mechanism are used for model training. Uncertainty analysis is performed using Monte Carlo simulation.
It improves the accuracy and efficiency of new energy power generation forecasting, reduces the computational load, and has the ability to migrate across regions and scenarios, adapting to new energy power generation systems of different regions and types.
Smart Images
Figure CN120996253A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of new energy technology, and in particular to a method and system for predicting new energy power generation. Background Technology
[0002] With the continuous growth of global demand for clean energy, the proportion of new energy sources such as solar and wind power in the energy supply is increasing. However, the intermittent and fluctuating nature of new energy power generation poses a significant challenge to the stable operation of the power system.
[0003] Currently, existing methods for predicting renewable energy generation have many shortcomings. For example, some prediction methods based on traditional statistical models cannot fully consider the complex nonlinear factors in the renewable energy generation process, resulting in low prediction accuracy. While some prediction models using deep learning algorithms have improved prediction accuracy to some extent, they suffer from problems such as high computational cost, long training time, and high hardware requirements, making them difficult to widely apply in practical engineering. Therefore, developing an efficient, accurate, and easy-to-implement renewable energy generation prediction algorithm is of significant practical importance.
[0004] Based on the above problems, this invention proposes a method and system for predicting new energy power generation. Summary of the Invention
[0005] To overcome the shortcomings of existing technologies, this invention provides a simple and efficient method for predicting new energy power generation.
[0006] This invention is achieved through the following technical solution:
[0007] A method and system for predicting new energy power generation, characterized by the following steps:
[0008] Step S1: Data Acquisition and Preprocessing
[0009] Collect new energy power generation data, preprocess the collected data, apply interpolation algorithms to fill missing values, use moving average filtering to eliminate random noise interference, and mark and correct data that exceeds the reasonable range by setting a custom data range threshold.
[0010] In step S1, the power generation of the solar panel, the wind speed measured by the anemometer, the light intensity collected by the light sensor, and the operating status data of the power generation equipment are collected in real time.
[0011] The collected data is stored and organized according to a custom time interval (e.g., 15 minutes) to form a historical dataset;
[0012] Meanwhile, a sliding window method was used to retain a portion of recent data as independent validation samples to construct a data test set for evaluating the model's predictive performance.
[0013] Step S2: Feature Extraction and Selection
[0014] Principal component analysis algorithm is used to extract features from the preprocessed data. Mutual information method is used to evaluate and select the importance of the extracted features, and the features with the strongest correlation with new energy power generation are selected.
[0015] In step S2, the principal component analysis (PCA) algorithm is used to extract features from the preprocessed data, converting high-dimensional data into low-dimensional feature vectors to reduce data dimensionality and computational complexity.
[0016] Meanwhile, the mutual information method is used to evaluate and select the importance of the extracted features, and the features with the strongest correlation with new energy power generation are selected to improve the performance and efficiency of the prediction model.
[0017] Step S3: Prediction Model Construction
[0018] An improved long short-term memory network model is constructed. Based on the traditional long short-term memory network model, an attention mechanism is introduced. An attention weight calculation layer is added to the long short-term memory network unit. By calculating the attention weight of each time step in the input sequence, the input information is weighted and fused, thereby highlighting the influence of important information on the prediction results.
[0019] In step S3, an improved long short-term memory network model is built using the TensorFlow deep learning framework in the Python environment. The hyperparameters of the improved long short-term memory network model are set, including the number of hidden layer neurons, learning rate, and number of iterations, to achieve optimization of new energy power generation prediction.
[0020] Step S4: Model Training and Optimization
[0021] The prediction model is trained using data from historical datasets, and gradient descent is performed using the adaptive moment estimation (Adam) optimizer. An early stopping mechanism is used to monitor and validate the loss curve, and the optimal number of iterations is dynamically determined to prevent overfitting.
[0022] In step S4, during the training process, an early stopping mechanism is set to monitor the evaluation metrics of the model on the test set. When the performance metrics on the validation set no longer improve within a custom number of rounds, training is stopped to prevent the model from overfitting.
[0023] An adaptive moment estimation optimization algorithm is used to automatically update the model parameters. The optimal combination of hyperparameters is determined by grid search or Bayesian optimization methods, thereby improving the model convergence efficiency and prediction accuracy.
[0024] In step S4, the trained model weights are loaded, rolling predictions are performed on the test set data, and prediction results are generated.
[0025] The prediction results are evaluated using the root mean square error (RMSE) or mean absolute error (MAE) as evaluation indicators.
[0026] Step S5: Output the prediction results
[0027] The real-time collected new energy power generation data is input into the trained prediction model, which outputs a custom prediction value of new energy power generation over a future period. Uncertainty analysis is performed on the prediction results, and the confidence interval of the prediction value is given.
[0028] In step S5, the Monte Carlo simulation method is used to perform uncertainty analysis on the prediction results and give the confidence interval of the prediction value.
[0029] A system for predicting renewable energy power generation, used to implement the above method, includes:
[0030] The data acquisition and preprocessing module is responsible for collecting new energy power generation data, preprocessing the collected data, filling missing values with interpolation algorithms, eliminating random noise interference by using moving average filtering, and marking and correcting data that exceeds the reasonable range by setting a custom data range threshold.
[0031] The feature extraction and selection module is responsible for using principal component analysis algorithm to extract features from preprocessed data, and using mutual information method to evaluate and select the importance of the extracted features, and screen out the features with the strongest correlation to new energy power generation.
[0032] The prediction model building module is responsible for building an improved long short-term memory network model. It introduces an attention mechanism on the basis of the traditional long short-term memory network model and adds an attention weight calculation layer to the long short-term memory network unit. By calculating the attention weight of each time step in the input sequence, the input information is weighted and fused, thereby highlighting the impact of important information on the prediction results.
[0033] The model training and optimization module is responsible for training the constructed prediction model using data from the historical dataset, using the adaptive moment estimation (Adam) optimizer for gradient descent, and using an early stopping mechanism to monitor and validate the loss curve, dynamically determining the optimal number of iterations to prevent overfitting.
[0034] The prediction result output module is responsible for inputting the real-time collected new energy power generation data into the trained prediction model, outputting the customized prediction value of new energy power generation for a future period of time, and performing uncertainty analysis on the prediction results to provide the confidence interval of the prediction value.
[0035] A device for predicting new energy power generation, characterized in that it includes a memory and a processor; the memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-described method steps.
[0036] A readable storage medium, characterized in that: a computer program is stored on the readable storage medium, and the computer program, when executed by a processor, implements the above-described method steps.
[0037] The beneficial effects of this invention are: the method for predicting new energy power generation can more accurately capture the complex patterns in the process of new energy power generation, effectively improve the accuracy of prediction, reduce prediction errors, and at the same time, it has a small amount of computation, high model training and prediction efficiency, can adapt to the prediction needs of different regions and different types of new energy power generation systems, and has strong generalization ability and adaptability. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0039] Appendix Figure 1 This is a schematic diagram of the method for predicting new energy power generation according to the present invention.
[0040] Appendix Figure 2 This is a schematic diagram of the improved Long Short-Term Memory network model of the present invention. Detailed Implementation
[0041] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below in conjunction with the embodiments of this invention. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.
[0042] The method for predicting new energy power generation includes the following steps:
[0043] Step S1: Data Acquisition and Preprocessing
[0044] Collect new energy power generation data, preprocess the collected data, apply interpolation algorithms to fill missing values, use moving average filtering to eliminate random noise interference, and mark and correct data that exceeds the reasonable range by setting a custom data range threshold.
[0045] In step S1, the power generation of the solar panel, the wind speed measured by the anemometer, the light intensity collected by the light sensor, and the operating status data of the power generation equipment are collected in real time.
[0046] The collected data is stored and organized according to a custom time interval (e.g., 15 minutes) to form a historical dataset;
[0047] Meanwhile, a sliding window method was used to retain a portion of recent data as independent validation samples to construct a data test set for evaluating the model's predictive performance.
[0048] Step S2: Feature Extraction and Selection
[0049] Principal component analysis algorithm is used to extract features from the preprocessed data. Mutual information method is used to evaluate and select the importance of the extracted features, and the features with the strongest correlation with new energy power generation are selected.
[0050] In step S2, the principal component analysis (PCA) algorithm is used to extract features from the preprocessed data, converting high-dimensional data into low-dimensional feature vectors to reduce data dimensionality and computational complexity.
[0051] Meanwhile, the mutual information method is used to evaluate and select the importance of the extracted features, and the features with the strongest correlation with new energy power generation are selected to improve the performance and efficiency of the prediction model.
[0052] Step S3: Prediction Model Construction
[0053] An improved Long Short-Term Memory (LSTM) network model is constructed by introducing an attention mechanism into the traditional LTM network model. This mechanism enables the model to pay more attention to historical information relevant to the current prediction time, improving its ability to capture long-term and short-term dependencies in time series data. An attention weight calculation layer is added to the LTM network unit. By calculating the attention weights at each time step in the input sequence, the input information is weighted and fused, thereby highlighting the impact of important information on the prediction results.
[0054] In step S3, an improved long short-term memory network model is built using the TensorFlow deep learning framework in the Python environment. The hyperparameters of the improved long short-term memory network model are set, including the number of hidden layer neurons, learning rate, and number of iterations, to achieve optimization of new energy power generation prediction.
[0055] Step S4: Model Training and Optimization
[0056] The prediction model is trained using data from historical datasets, and gradient descent is performed using the adaptive moment estimation (Adam) optimizer. An early stopping mechanism is used to monitor and validate the loss curve, and the optimal number of iterations is dynamically determined to prevent overfitting.
[0057] In step S4, during the training process, an early stopping mechanism is set to monitor the evaluation metrics of the model on the test set. When the performance metrics on the validation set no longer improve within a custom number of rounds, training is stopped to prevent the model from overfitting.
[0058] An adaptive moment estimation optimization algorithm is used to automatically update the model parameters. The optimal combination of hyperparameters is determined by grid search or Bayesian optimization methods, thereby improving the model convergence efficiency and prediction accuracy.
[0059] In step S4, the trained model weights are loaded, rolling predictions are performed on the test set data, and prediction results are generated.
[0060] The prediction results are evaluated using the root mean square error (RMSE) or mean absolute error (MAE) as evaluation indicators.
[0061] Step S5: Output the prediction results
[0062] The real-time collected new energy power generation data is input into the trained prediction model, which outputs a custom prediction value of new energy power generation over a future period. Uncertainty analysis is performed on the prediction results, and the confidence interval of the prediction value is given.
[0063] In step S5, the Monte Carlo simulation method is used to perform uncertainty analysis on the prediction results and give the confidence interval of the prediction value.
[0064] The system for predicting new energy power generation is used to implement the above method, including:
[0065] The data acquisition and preprocessing module is responsible for collecting new energy power generation data, preprocessing the collected data, filling missing values with interpolation algorithms, eliminating random noise interference by using moving average filtering, and marking and correcting data that exceeds the reasonable range by setting a custom data range threshold.
[0066] The feature extraction and selection module is responsible for using principal component analysis algorithm to extract features from preprocessed data, and using mutual information method to evaluate and select the importance of the extracted features, and screen out the features with the strongest correlation to new energy power generation.
[0067] The prediction model building module is responsible for building an improved long short-term memory network model. It introduces an attention mechanism on the basis of the traditional long short-term memory network model and adds an attention weight calculation layer to the long short-term memory network unit. By calculating the attention weight of each time step in the input sequence, the input information is weighted and fused, thereby highlighting the impact of important information on the prediction results.
[0068] The model training and optimization module is responsible for training the constructed prediction model using data from the historical dataset, using the adaptive moment estimation (Adam) optimizer for gradient descent, and using an early stopping mechanism to monitor and validate the loss curve, dynamically determining the optimal number of iterations to prevent overfitting.
[0069] The prediction result output module is responsible for inputting the real-time collected new energy power generation data into the trained prediction model, outputting the customized prediction value of new energy power generation for a future period of time, and performing uncertainty analysis on the prediction results to provide the confidence interval of the prediction value.
[0070] The device for predicting new energy power generation includes a memory and a processor; the memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-described method steps.
[0071] The readable storage medium stores a computer program that, when executed by a processor, implements the above-described method steps.
[0072] Compared with existing technologies, this method for predicting new energy power generation has the following characteristics:
[0073] First, it has high-precision prediction performance.
[0074] By employing multi-level data preprocessing, intelligent feature screening, and an improved long short-term memory network model, the dynamic patterns of new energy power generation are accurately captured, significantly improving prediction accuracy and reducing error indicators, thus providing a reliable basis for grid dispatch.
[0075] Secondly, it can achieve efficient computing and fast response.
[0076] Principal Component Analysis (PCA) and mutual information methods are used to achieve dimensionality compression, and the network structure is optimized by combining attention mechanism. While maintaining accuracy, the computational load is greatly reduced, the training cycle is shortened, and the real-time requirements of engineering are met.
[0077] Third, it has strong generalization and adaptability.
[0078] Deep learning training based on large-scale historical data, combined with an early stop regularization strategy to effectively suppress overfitting, enables the model to have cross-regional and cross-scenario transfer capabilities, adapting to diverse new energy power generation system prediction tasks.
[0079] The embodiments described above are merely one specific implementation of the present invention. Ordinary changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included within the protection scope of the present invention.
Claims
1. A method and system for predicting new energy power generation, characterized in that: Includes the following steps: Step S1: Data Acquisition and Preprocessing Collect new energy power generation data, preprocess the collected data, apply interpolation algorithms to fill missing values, use moving average filtering to eliminate random noise interference, and mark and correct data that exceeds the reasonable range by setting a custom data range threshold. Step S2: Feature Extraction and Selection Principal component analysis algorithm is used to extract features from the preprocessed data. Mutual information method is used to evaluate and select the importance of the extracted features, and the features with the strongest correlation with new energy power generation are selected. Step S3: Prediction Model Construction An improved long short-term memory network model is constructed. Based on the traditional long short-term memory network model, an attention mechanism is introduced. An attention weight calculation layer is added to the long short-term memory network unit. By calculating the attention weight of each time step in the input sequence, the input information is weighted and fused, thereby highlighting the influence of important information on the prediction results. Step S4: Model Training and Optimization The prediction model is trained using data from historical datasets, and gradient descent is performed using an adaptive moment estimation optimizer. An early stopping mechanism is used to monitor and validate the loss curve, and the optimal number of iterations is dynamically determined to prevent overfitting. Step S5: Output the prediction results The real-time collected new energy power generation data is input into the trained prediction model, which outputs a custom prediction value of new energy power generation over a future period. Uncertainty analysis is performed on the prediction results, and the confidence interval of the prediction value is given.
2. The method for predicting new energy power generation according to claim 1, characterized in that: In step S1, the power generation of the solar panel, the wind speed measured by the anemometer, the light intensity collected by the light sensor, and the operating status data of the power generation equipment are collected in real time. The collected data is stored and organized according to a custom time interval to form a historical dataset; Meanwhile, a sliding window method was used to retain a portion of recent data as independent validation samples to construct a data test set for evaluating the model's predictive performance.
3. The method for predicting new energy power generation according to claim 1, characterized in that: In step S2, the principal component analysis algorithm is used to extract features from the preprocessed data, converting high-dimensional data into low-dimensional feature vectors to reduce data dimensionality and computational complexity. Meanwhile, the mutual information method is used to evaluate and select the importance of the extracted features, and the features with the strongest correlation with new energy power generation are selected to improve the performance and efficiency of the prediction model.
4. The method for predicting new energy power generation according to claim 3, characterized in that: In step S3, an improved long short-term memory network model is built using the TensorFlow deep learning framework in the Python environment. The hyperparameters of the improved long short-term memory network model are set, including the number of hidden layer neurons, learning rate, and number of iterations, to achieve optimization of new energy power generation prediction.
5. The method for predicting new energy power generation according to claim 1, characterized in that: In step S4, during the training process, an early stopping mechanism is set to monitor the evaluation metrics of the model on the test set. When the performance metrics on the validation set no longer improve within a custom number of rounds, training is stopped to prevent the model from overfitting. An adaptive moment estimation optimization algorithm is used to automatically update the model parameters. The optimal combination of hyperparameters is determined by grid search or Bayesian optimization methods, thereby improving the model convergence efficiency and prediction accuracy.
6. The method for predicting new energy power generation according to claim 5, characterized in that: In step S4, the trained model weights are loaded, rolling predictions are performed on the test set data, and prediction results are generated. The prediction results are evaluated using the root mean square error (RMSE) or mean absolute error (MAE) as evaluation indicators.
7. The method for predicting new energy power generation according to claim 1, characterized in that: In step S5, the Monte Carlo simulation method is used to perform uncertainty analysis on the prediction results and give the confidence interval of the prediction value.
8. A system for predicting new energy power generation, characterized in that: To implement the method according to any one of claims 1 to 7, comprising: The data acquisition and preprocessing module is responsible for collecting new energy power generation data, preprocessing the collected data, filling missing values with interpolation algorithms, eliminating random noise interference by using moving average filtering, and marking and correcting data that exceeds the reasonable range by setting a custom data range threshold. The feature extraction and selection module is responsible for using principal component analysis algorithm to extract features from preprocessed data, and using mutual information method to evaluate and select the importance of the extracted features, and screen out the features with the strongest correlation to new energy power generation. The prediction model building module is responsible for building an improved long short-term memory network model. It introduces an attention mechanism on the basis of the traditional long short-term memory network model and adds an attention weight calculation layer to the long short-term memory network unit. By calculating the attention weight of each time step in the input sequence, the input information is weighted and fused, thereby highlighting the impact of important information on the prediction results. The model training and optimization module is responsible for training the constructed prediction model using data from the historical dataset, employing an adaptive moment estimation optimizer for gradient descent, and using an early stopping mechanism to monitor and validate the loss curve, dynamically determining the optimal number of iterations to prevent overfitting. The prediction result output module is responsible for inputting the real-time collected new energy power generation data into the trained prediction model, outputting the customized prediction value of new energy power generation for a future period of time, and performing uncertainty analysis on the prediction results to provide the confidence interval of the prediction value.
9. A device for predicting new energy power generation, characterized in that: It includes a memory and a processor; the memory is used to store a computer program, and the processor is used to execute the computer program to implement the steps of the method as described in any one of claims 1 to 7.
10. A readable storage medium, characterized in that: The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 7.