Power load prediction method, system, medium and device based on quantile regression
By using a power load forecasting method based on quantile regression and employing models such as bidirectional temporal convolutional networks and attention mechanism layers to dynamically allocate feature weights, the method addresses the problem of insufficient adaptability of load forecasting in coal mine scenarios. It achieves accurate identification of load abrupt changes and high load clustering, thereby improving the operational stability of the power grid.
CN122393910APending Publication Date: 2026-07-14INFORMATION & COMM CO OF STATE GRID XINJIANG ELECTRIC POWER CO LTD +1
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INFORMATION & COMM CO OF STATE GRID XINJIANG ELECTRIC POWER CO LTD
- Filing Date
- 2026-04-17
- Publication Date
- 2026-07-14
Smart Images
Figure CN122393910A_ABST
Abstract
The application discloses a power load prediction method and system based on quantile regression, a medium and equipment, and relates to the technical field of power system load prediction.The method comprises the following steps: acquiring historical power load data and corresponding time characteristic data of coal mine power load; preprocessing the historical power load data to construct a power load feature set; constructing a learning prediction model based on quantile regression; iteratively training the learning prediction model according to a preset quantile and a loss function in a quantile regression output layer based on a training set; inputting corresponding historical power load values and time characteristic data at a to-be-predicted moment into the trained learning prediction model, outputting a load prediction value in the quantile regression output layer, and constructing a probability prediction interval according to multiple quantiles.The application focuses on the key time point that has the greatest influence on the prediction result in the historical power load data, dynamically allocates feature weights, and significantly improves the accuracy of point prediction.
Need to check novelty before this filing date? Find Prior Art