This work proposes a novel and sustainable energy development strategy for addressing the energy shortages in rural areas and the low energy efficiency of off-grid solar power systems. This study combines the analysis of power consumption type with consumption anomaly detection to characterize households’ power consumption habits and ensure the safety of a system. Specifically, the proposed anomaly detection method is a hybrid nonintrusive model.
The home power usage data are collected and processed by auto-data-binning without manual labeling, and thus, the training cost is reduced to enable the application of machine learning technologies in underdeveloped areas with limited computational resources. With the premise of limited energy sources in off-grid areas, the proposed power consumption analysis method divides home power usage habits into four different types. Different feedback mechanisms are adopted to extend the microgrid’s supply time according to the analysis results.
The proposed method significantly increases the utilization of local renewable energy and improves residents’ experience. The proposed method is implemented in a rural village in Tanzania; after long-term monitoring, the validity of the proposed method is demonstrated.
Read more: