Quantile Probabilistic Power Load Ensemble Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current power load forecasting methods lack effective ensemble forecasting techniques for quantile probabilistic short-term predictions, especially with increasing load uncertainty due to distributed renewable energy and energy storage, which limits their accuracy and operational efficiency.
Innovation Solution
A method for quantile probabilistic short-term power load ensemble forecasting that involves dividing historical data into subsets, performing bootstrap sampling, training neural network, random forest, and gradient boosting regression tree models, and determining weights for these models to minimize pinball loss, thereby establishing an ensemble forecasting model across different quantiles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple forecasting models are integrated using ensemble forecasting, then forecasting accuracy and generalization performance are improved, but device complexity and computational burden increase
Solution Approach 1:
The patent combines multiple diverse forecasting models (linear regression, neural network, support vector machine, random forest, gradient boosting) into an ensemble forecasting system. Each model processes the same input features but produces different predictions, which are then aggregated through weighted averaging to generate the final forecast, thereby improving accuracy while managing complexity through systematic integration
Solution Approach 2:
The patent optimizes the weight parameters of individual models within the ensemble by minimizing pinball loss function. This parameter adjustment allows the system to dynamically determine the contribution of each model based on its performance characteristics, improving forecasting accuracy while adapting to different quantile requirements without increasing structural complexity
2Adaptability or versatility
If quantile regression models are trained on multiple training sets with different parameters, then model diversity and ensemble performance are improved, but training time and computational resources increase
Solution Approach 1:
The patent performs bootstrap sampling on the training data to generate multiple training sets before model training. This preliminary action creates diverse training datasets that enable the subsequent training of multiple models with different characteristics, ensuring model diversity is achieved through pre-prepared varied training data rather than through iterative trial-and-error processes
Solution Approach 2:
The patent creates multiple copies of the training data through bootstrap sampling, where each training set is a resampled version of the original data with different random selections. This copying approach efficiently generates diverse training sets without requiring additional data collection or complex data transformation processes, reducing the time cost of creating diverse training data
3Device complexity
If point forecasting is used to provide a single estimated value, then computational simplicity is maintained, but uncertainty characterization and information content are insufficient
Solution Approach 1:
The patent transitions from point forecasting (single value prediction) to quantile forecasting by introducing a new dimension - the quantile level. Instead of predicting only the mean or median load, the system predicts multiple quantiles (e.g., 10th, 50th, 90th percentiles), thereby characterizing the uncertainty distribution of future load without significantly increasing computational complexity. This dimensional expansion provides rich uncertainty information while maintaining the simplicity of regression-based forecasting
Data Source
AI summary
The disclosure relates to a quantile probabilistic short-term power load ensemble forecasting method. The method includes: dividing historical power load data of a power system into a first data set and a second data set; performing bootstrap sampling on the first data set to generate multiple training data sets; training a neural network quantile regression model, a random forest quantile regression model and a gradient boosting regression tree regression model for the each training data set to obtain quantile forecasting models; establishing an optimization model with an objective function for minimizing the quantile loss for the second data set, and determining a weight for each of the quantile regression models, to calculate a power load ensemble forecasting model for predicting the power load in the power system.