Support Vector Regression Feature Weighting for Load Forecast
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current load forecasting methods, such as artificial neural networks and support vector regression, face challenges in accurately predicting short-term power system loads due to the unequal importance of input features, which affects prediction accuracy and efficiency in practical applications.
Innovation Solution
A novel feature learning scheme is introduced that computes correlation scores for each input feature with the power system load, using these scores to weight features in a non-linear support vector regression framework, enhancing prediction accuracy by prioritizing features with higher correlation to the target variable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional load forecasting methods (neural networks, support vector regression) are used without feature weighting, then the model structure is simple and easy to implement, but prediction accuracy deteriorates due to unequal importance of input features
Solution Approach 1:
The patent applies parameter changes by introducing correlation scores as weighting parameters for input features in the support vector regression model. Instead of treating all features equally, the method computes correlation scores between each input feature and the target variable (power system load), then uses these scores to weight the features during model training. This transforms the standard SVR model into a weighted SVR model, improving prediction accuracy by adjusting parameter weights based on feature relevance.
2Measurement precision
If all input features are treated with equal weight, then the computational process is simple and fast, but prediction accuracy deteriorates because features with higher correlation to load are not prioritized
Solution Approach 1:
The patent applies preliminary action by pre-computing correlation scores between input features and the target variable before performing the support vector regression. This preliminary step identifies and ranks features based on their correlation strength, allowing the model to focus computational resources on the most relevant features. By performing this feature selection and weighting beforehand, the subsequent SVR training becomes more efficient and accurate, avoiding the need to process all features equally in detail.
3Measurement precision
If feature weighting based on correlation scores is implemented, then prediction accuracy improves, but the complexity of computing correlation scores and weighting features increases
Solution Approach 1:
The patent applies mechanics substitution by replacing complex manual feature selection and weighting mechanisms with automated correlation score computation. Instead of requiring domain experts to manually identify important features or use complex feature engineering techniques, the method substitutes these mechanical processes with automated statistical correlation analysis. The correlation scores are computed automatically from the training data, and the weighting is applied automatically during SVR training, simplifying the overall process while improving accuracy.
Data Source
AI summary
In a support vector regression approach to forecasting power load in an electrical grid, a feature learning scheme weights each feature in the input data with its correlation with the predicted load, increasing the prediction accuracy. The kernel matrix for the input training data is computed such that features that align better with the target variable are given greater weight. The resulting load forecast may be used to compute commands sent to demand response modules.