Power Prediction Model Training with Multi-Scale Data Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current power prediction methods for solar energy integration into power grids face challenges due to data loss from random sampling and the need for manual hyperparameter tuning, requiring large datasets and being sensitive to precision of domain knowledge constraints.
Innovation Solution
A model training method that utilizes historical power and meteorological data from multiple sources and time scales, employing multiple models with adaptive weights to optimize learning and enhance prediction precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If random sampling is performed in frequency domain to reduce calculation complexity, then computation time is reduced, but information loss occurs in the input data
Solution Approach 1:
The input data is segmented into multiple frequency components through Fourier transformation, allowing selective processing of different frequency bands. This segmentation enables the model to retain important information while reducing computational burden by focusing on relevant frequency ranges rather than processing all data uniformly.
Solution Approach 2:
The patent transforms the data from time domain to frequency domain, changing the representation parameters. This parameter transformation allows the model to capture periodic patterns and trends more effectively while reducing the dimensionality of the input space, thereby maintaining information integrity with reduced computational complexity.
2Measurement precision
If domain knowledge is added as prior knowledge to restrict and optimize the model, then model accuracy is improved, but the method becomes sensitive to precision of constraint functions
Solution Approach 1:
The patent employs adaptive weighting mechanisms that dynamically adjust the influence of different loss functions during training. Instead of using fixed domain knowledge constraints, the model learns optimal weights for different constraint terms, making the incorporation of domain knowledge flexible and less sensitive to initial precision assumptions while maintaining accuracy improvements.
Solution Approach 2:
The model incorporates feedback loops where prediction errors are used to adjust the weighting of domain knowledge constraints. This feedback mechanism allows the system to automatically tune the influence of prior knowledge based on actual performance, reducing sensitivity to initial constraint precision while maintaining accuracy benefits.
3Measurement precision
If multiple models with different weights are used to represent input data, then prediction precision is improved, but training complexity increases
Solution Approach 1:
The patent merges multiple model outputs through a weighted combination mechanism, where different models process different aspects of the input data (e.g., different frequency bands or time scales). The results are then aggregated with learned weights, achieving improved prediction precision while managing training complexity through shared parameter initialization and progressive training strategies.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This application discloses a model training method, a power prediction method, and a device, which may be applied to the field of artificial intelligence. A model includes at least a first model and a second model. The method includes: obtaining a dataset including historical power data and historical meteorological data in preset duration; determining a weight of a loss function of each model based on the dataset; constructing a target loss function, where the first model and the second model may be used to represent input data, and data represented by the first model is not completely the same as data represented by the second model; and finally training the models based on the target loss function. More complete information is obtained from different modal data (that is, the historical power data and the historical meteorological data) and different time scales (that is, the preset duration) based on a currently available data source, to optimize the learning model and achieve higher prediction precision.