Data Fusion of IoT Appliance Traces and Satellite Solar Irradiance for Energy-Efficient and Robust PV–Battery Sizing Under Uncertainty

Abstract—Accurate integration of distributed renewable generation in residential energy systems requires realistic joint
modelling of electricity demand and solar availability. Most of
the existing approaches are based on synthetic or averaged load
profiles, which reduce design robustness. This study introduces a
quantitative data fusion framework that combines real appliancelevel IoT power data with satellite solar information to create more realistic and uncertainty-aware design scenarios. The
framework incorporates three stages: it first cleans and combines
household energy readings, then aligns them with daily solar
data, and finally generates representative design cases based on
statistical ranges (for example, typical, low and high energy use
or sunlight conditions). Validation tests to compare this method
with conventional averages using the Wasserstein distance and
Spearman correlation confirm improved distribution fidelity and
uncertainty preservation. The outcome is a fully reproducible
open-source pipeline that allows engineers and researchers to size
PV–battery systems using real-world data rather than simplified
assumptions. The approach improves the accuracy, reliability,
and reproducibility of the design of data-driven energy systems.