Multi-task deep learning for large-scale buildings energy management
Bibliographic record
Abstract
Building energy management acts as the brain of the building, which controls the energy supply based on sensor data and algorithms. However, existing methods only focus on single-task prediction like load forecasting. As more multi-variable data is collected from ubiquitous sensors, building energy management needs to extend functionality from single-task to multi-purpose predictions. This study designs a multi-task learning system to tackle four different tasks: 1. Electricity load forecasting; 2. Air temperature forecasting; 3. Energy anomaly detection; 4. Energy anomaly prediction. A mixture-of-experts framework with the self-attention mechanism is proposed for learning heterogeneous tasks. A new comprehensive dataset has been created with real data to demonstrate the heterogeneous tasks' efficacy of the suggested framework. Extensive experiments are conducted with various deep learning models, which shows our proposed model achieves superior prediction performance overall tasks. Comparative studies are performed to explore the correlations between forecasting and anomaly learning, which reveal the benefits of multi-task learning for heterogeneous tasks. Anomaly detection and prediction both achieve 98% accuracy and 95% F1-score, while the electricity load forecasting single-task error is reduced by almost 60% through the multi-task model. Nonetheless, the tasks' training difficulties and resource consumption are also investigated and the deeper network doesn't ensure better performances. The dataset is open-sourced at: https://github.com/rekingbc/Multi-task-building.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".