建筑暖通系统负荷信息的预测方法和装置
By constructing a loss function that includes prediction error information and dynamically adjustable constraint information, an initial prediction model is established and regression coefficients are updated in real time. This solves the problem of data scarcity in new building projects and achieves high-precision load prediction and adaptive capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TIANJIN UNIV
- Filing Date
- 2026-05-06
- Publication Date
- 2026-07-17
AI Technical Summary
In new building projects, due to the fact that building energy efficiency equipment has not been fully put into operation and data acquisition equipment has not been fully installed, historical data is scarce. Existing technologies are difficult to apply to building renovation and new projects, resulting in low efficiency and accuracy of energy efficiency management and equipment control, and difficulty in quickly adjusting operating strategies.
By constructing a loss function based on prediction error information and constraint information with dynamically adjustable intensity, multiple initial regression coefficients are solved to establish an initial prediction model. The regression coefficients are then updated in real time based on the current error to form a prediction model, enabling online incremental learning, adapting to nonlinear changes in the environment and load, and continuously tracking the dynamic characteristics of the system.
Under conditions of limited data, this method improves the accuracy and time series adaptability of load forecasting results, solves the problem of adaptive load forecasting in the case of scarce data, and achieves accurate forecasting with low computational complexity.
Smart Images
Figure CN122133112B_ABST