本研究主要探討基於慢智慧架構之空調節能,利用異質感測器收集空調系統運轉參數、環境參數及居住者行為,藉由決策樹演算法建立預測模型,提出應用於空調節能之慢智慧系統架構,使小型空調機達到智慧節能及舒適度控制。研究所提之慢智慧系統架構包含列舉、適應、排除、集中及傳播等功能模組,並由這些功能模組建構快速決策週期及緩慢的決策週期。在空調的節能控制上,慢智慧系統之快速決策反應可以及時控制風扇及壓縮機運轉模式,以反應空調系統能量的需求,達到即時的節能控制。而緩慢決策則利用長時間學習使用者使用空調的習性,藉以反饋到空調運轉的控制策略,以達到舒適及節能之智能控制。同時,緩慢的決策也會長時間記錄及解析空調系統運作的參數,以降低空調系統因老化及故障所造成的運轉效率低落,使空調運轉達到節能及舒適的目標。
This research studies air-conditioning energy conservation based on Slow Intelligence Framework (SIF). Heterogeneous sensors are employed to collect operating parameters of air-conditioning systems, environment parameters and resident behavior. Then a decision tree algorithm is applied to construct a prediction model that can be integrated to the energy saving SIF of room air conditioners in order to accomplish the energy and comfort goal of intelligent control. The SIF system developed in this research includes such function modules as enumeration, adaptation, elimination, concentration and propagation, which can help to construct a fast decision cycle and a slow decision cycle. To facilitate real-time energy-saving, SIF's fast decision could provide dynamic control over the operation mode of the fan and the compressor. SIF's slow decision, on the other hand, spends a long time learning users' air-conditioning habits. The learning results fed back to the operation control strategy could help achieve high efficiency in intelligent comfort and energy control. Meanwhile, slow decision could also keep longitudinal records and analyze operating parameters of air-conditioning system to reduce low efficiency operation caused by aging problems and failures, thus enhancing operational performance in comfort and energy management.